use AssessRcyT icnioooucr rat
imuiuiMis vmm mmemrnt
By
S4wmr4 W. Uvllll
FINAL ftSFOftT
Ccmcr.tct Bo»
Ink Orfer Ito. fS-OlHUff
mi Ni«t i«. tiro-uo
S*pCMltor Uf Mil
fw
Offiett of tolicy, rt4nittt «ft4 tv»l«fttUa
l.». hvinaaul fnuctiw Acancy
*fli m strMt, t.v.
i.e. IOMO
oa CkivirmMAMi Qw*Ucy
tocurfwn Offita ®f cbt 9r«si4Mt
122 J«k|M tl«M, M.tf.
WmM»i§e««if §~€• 200M
-------
-------
disclaimer
Thi» r«port ha« b€«o r«vi«w«4 foe ctctwical awftt by chc Councit on
Enwironptnttl Quality nod the U.S.. Cnvi ronn«nC«l Protection A|encj, and
ipprav«4 for publication. Approval do«* not ii|fii£y chic cht contents neces-
«»nly r«fl«ct cht v;«vi an4 peltci«s of th« Council or ch« Agency, nor does
••fttiort of trad* najtm or comntrciiil produces contticwci «ndor»«ftttnt or r«co»"
fMftdtcioo for a»».
-------
-------
PREFACE
Results are prMMtad in this report of a acody by Midwest fce«e*cch
Institute (HEX) entitled "Ink Asiefjutat Kethodologiee for I€M«W uo Turt,
Order Ma. 85-EDl-AMF, afider Contract No. BQ4CIS "tunic Ordering Agreeswnt for
Epvirqazaentsl Studies, Reaearch, and Analy»i«," KBI Project Ho. 817Q-L(4).
the work was perfortfad for the Office nt Policy Analysis, Office o£ Policy,
Planning And Evaluation, U.S. 6nvirc»M&nni:al Protection Agency. Dr. Ann M.
Plahar was EPA Task Officer, end 9r. illliam Mill* was Technical ftepreeanca-
tive for the Council on Environmental Quality on thia task.
The results herein draw snbecantially on related reaearch ic MRI for
che Office of Poli-cy Analysis, under £PA Contract Not. #§«•<] 1 "-Mi 21 (Kfll Project
Kot. I3*f-L(2) and 75A9-U1Z}) *ad on *a-al-*55* (mi Project Ho. IX51-LK 4
report of that work, "Ceatparlcon of tiikl sad Coata of Kaaardous Wait# Alter-
aacivea: Kattioda Devalopawac and Pilot Studies,1* is available through the
National Technical Information Service CUTIS tfo. P&B6-19B,912). Dr. Fisher
was Project Officer or co^Project Officer with Ma. Jeanne Briakin throughout
the earlier work.
Many people contributed co chis study. Kr. Thomaa L. Perguaon, thee
8e«4» MRl*s Process Assessment lection, wet Project Leader Cor all Che
research under the earlier related coBtracca noted abova, Dr. Thomaa W* Lapp,
Senior CheaJLat, was a major contributor to thac research. Other HX1 con*
tributor* co the methodology development included: Nr. Chriacopher J. Cole,
Aaiiaeaac £2wirotbM6t4L feiiteiit N*. Cxry L. XaUo, Senior Chenieai
Cosinear in the area* of aource aaiass
-------
Although this itaif «ii Mdi in aupport «f EPA'i oatdi to amti ebt
rltfc* Associated witit •ItacnACiv* hauriBtta uuti aaciac^Htit «ppro*ch**t ffllX
bip»# time ch« research aad Mtbodoloiy 4«i®crih«ii will to mmtml Aim in othir
•nvirooeeetal r*sul*torr p*oc«a««B nod in ocbtr dwiiioa iktm npfliaciost.
KIDUCST lESlMCfi CV8T1T0TE
ZdvmU V. U«l<»> -f* *** l**/€rr
l#*it hreciti Ait4smnt Section
A#pc*wtif
tt4CC«o C*wfear4, BirccCor
Kavirow—nt»l Jriten DapsrCMat
ii^Mbtr 23, 1181
#
-------
TABLE Of COtTTEYTS
AblCrdCCvaaaaataiatvtttfifi^iiiitiiiiiiiat*!**# tf L £ I
AC tiAOtfL ICIfttAlttivvataiaiat •¦•¦«•••¦••»• | V
SuflM r]f >lk<*»a«ai»»a(«iV|«|«««»|9|||t l«l»4»l«»l»l«»aaaiai«aia(aa a • a • • • S™"1
I*1 rcit rOdlJCC XOQ • • a • a • « I"1
h i ftiik A*s*49tn£n£ ftol* in Decision Making. • I-i
B • Rink Aiitfi^Mnc Needs Under HGRA itij Su^erf un4 . -.« 1-2
Ct Study Objiciivtitt Origins, nod Scope 1-^
I. Objcccivts* . »• •« 1-4
2 • Of l |1 HJ « I « I • l«»«»<>«»l»l«l«»a»t*aa»«»«aal*«aa X "^
3 • Sdpi 1-5
Rcfj *o Cbjipeer r — ^
l!i Bisk Assessment tnd Risk Hinagi* tACIl4tl»l*l4f«»4#l*l4>l*l«l I!"I
A • T € nol ojy«¦ ¦ * • ¦ »«.*»«$»<.*»*«.».««..«* 11*1
1 • tot U^di£vi JL L M L
2 j Da£ i&l c loos .•»..«••*..*•.•«**»•»»«.»* £ I. m 2
B. Feant^orn for Assessing and rlinaging Risks !l-ll
1. Comparative riste-cosc-btnef it assessment 11 — 1 i
2. Decision waking «... — ... 11-14
3i !iik nanigcawfiL strategies — ... 11-14
4. Decision i op Indentation . Il-Lfc
References co Chapetr II 11-18
III. Approach to Comparative Risk Assessncnt of Hazardous
Haste Management Alternatives III-1
kt He chiodo Logical Objtotivtl................... III-l
1. Rational* and guidelines Ill-L
2. Bounding the assessment .. II1-3
B» MechodologxcaL FcaoevoThc»«»••••••*«••••»•..~•••••• £ IX
w
-------
TABLE OF CONTENTS (Continued)
IV. An Integrated Comparative Risk Assessment Methodology IV-1
A. Source Assessment (Hazardous ChJfletecinlioT*).... IV-I
1. C&neraL approach* fV-1
2. Special scenarios IV-3
B. Envirofuocnt#l Transport and Fate Anal/li) IV-8
|. Sice eharacteriaat ion IV-B
2 , Model selection. IV-10
C. Exposure Prediction. fV-14
1. Exposed workers ......... IV-L5
2. Exposed publi ..... .... IV-16
D. Health and Environment*! Effects Analyses IV—1.9
1. Search scope. IV-19
Z. Identify likely responses IV-2J
3® Develop dose-response rtlac ;Oftthipl IV-2J
£. Health end Envi rsoiiMial Iapacta Eat imac ion
and Integration IV-33
1. Effects per individual . . . . CV-31
2. Effects on populations . — — IV-3A
1. Special con*(derations IV-36
4. EavirooflMtncal ingpects estimation — IV-37
P. Uncertainty AnaLyti s • • •. • IV-38
1. Sources of uncertainty IV-J8
2. Sensitivity analysis of key variables IV-39
3. Aggregation of uncertainties IV-41
V. Source Asteisaest (U«Mrd Cfcera'Cterixation) , V-l
A. Hazard Identification aad Description V-l
1. Technoecoooaic characterization V-l
2. Chemicals af concern and their properties.... V-2
3. Release mechanisi&i and points. V-2
B. Quantification of EleLeasei V-l
References for Chapter V. V-?
vi
-------
TABLE OF COMTEWTS (Concinuitd)
Prediction of Environment#! Transport and Fact.
A.
a.
Central Data RiquiroseDti Sources
Groundwater Mcdela. ..
1. HaChematictl aod«ll<
2. Ranking oodeb.
Surface Vdf«r Models.
1. FoLlucanc diapers ion wodei ».
2. Uat«rfhed runoff model).....
l « KO(l
1. Screening aodel*.. ...»
2. Refined wodel.s ....
3. Specialized nod*La
4. Criteria for model selection
5. Accuracy and linicaciona of tie models....
$ 6
$ 9
Ecological Models ..
Zncaroadia ami ttultijKdU Model
9 ® • ® %
Rif«rmnciJ to Chapter VI
Exposure Prediction. ~
A. Populations Potancially Exposed,
1. Occupational aurpoturea.......
2 4 C^ne ra I po puLat l oft expo Aufa4 <
Estimates Exposure!!,
1. Czpoiurc proEiLe......
2. Expoaura i.acegrat ion..
• • a • •
l«f a t^eci ca a tnj ^#l^ai ^^t a (? r T
Health Cffecta Prediction H«chod*
A. Character!atica of Health Effects Data
**ia»*»e
1. Types of aixpaauraa
2. Types of effaicta <
3. Type* of dosat-response relationships.
4. Data sources and quality...
I 4 • I
see*
» e » «
> < a a
P«ae
VI-L
Vl-1
VI-J
VI~*
V1 -19
Vl-21
VI -21
VI-22
VI-28
VI-29
VI-3L
VI-33
VI-33
VI-3S
VI-36
VI'57
VI-^0
vri-i
VIt-1
vri-i
VII-2
vii-5
vf 1-3
VII-5
VII-7
viri-i
VIII-l
VIII-2
VI1I-3
vin-5
VIII-U
vii
-------
TABLE W COKOTTI (Ce»cin«iadJ
£sss
I. Prtdicctvt H9deU..»*> VTII-19
1. Ifp«i of excr«p6l*ri-oa* oaadcd* VIII-19
2. Mathematical extrapolation newels.... VIIl-27
lafaraitcts to Chapter VIII ...... ¥111^61
IX. Eacintacion and Suaoucicm of Advene Zopacti on Expocad
Populations* <« ~;..~.»•».»•«• I3C-1
A. tfua»o Health Ivpacts **-i
1. Public health eff«ci IX-1
2m Occupation*! baalch. efface*-......... EX"*
B. Ecological and Otitcr CftvirooMBtal [ap4cc>........ IX-6
X. AnaIyeis of Oncert»il\C i*» Lfl tiftk Auatsnaat of Hawrdou*
Wait* X~1
A. Overview qC UneerraiBcy Analytis X-l
I. Taraifialocy X-l
2# Approaches to uncertainty enalyata. 1-2
3. Consideration* «fle«tla» cJholc* of approach.. X-l
9» Sources of Uncertainty *. • *~7
U Scenario* definition and applicabili ty....... *-9
2» Pollutant 2-9
1. CcrviroodHuical tranaporc and Uc« 1-10
4. Exposure prediction X-10
3# Health effects d«ca And oodel Availability... X-10
Health iop4«t eacLaacioa And tumation .. X^IO
C«. Aggregation of Uncare«lfl£i*B X-U
D. Relatival CJaeertaiaty end Bilk.., 1-15
E. Um of SurrofatM in Comparative ftUk Aseesaeeat.. X-l7
lifirioctf Co Chapter X 1-19
Appendix A - Risk Aiseimnc Terminology. A—I
Appendix B - Development of fceoatrioa far Comparative Bti»k.
Aitumsl 9t Kaaardeue Waste K«U|Mtitc B-I
Appendix C - Ecological attd Soeioeeoasadc tapaet Aiie
-------
TABLE OF CONTESTS (Continued]
y.»c of FLgurm*
pi8life IjtlS P«e
S-l Frracvork for Atitniog Techaalogical
Risk#.... ..... i-2
S-2 ElenenO in Health liak AiuiniDt of Hazardous tfuCc
Disposal Methods......... 8*5
S-3 Schema for Assessing Haterdoua Wastes a< Sources of
gQviroiuienc«l Pollution ........ S-€
S~4 Schema for Selectioo of Models for AMljrtin of
Tr«nftj»frt and fata of Pollutant* Bat arias the
Eevironaeoc S-8
5-5 ScheMt far Selective of Health Effects C»ciB*cton
Model• Depending ©a Availability of Diet............. 8-9
Il'i Prauiawirk, for Assessing and Kanagiog Technological
litkt ~ 11-12
IV-l Schema for Assessing U^xarrdous Waste* as Sources of
Environmental Pollution.. tV-3
IV-2 Schema Cor Selection of Hodala for Analysis of
Trvasporc and Fata of Vollutaiits Entering the
Environment»• IV-U
IV-3 Schema for Selection of Health Effects- Eatlnation
Models Depdadtaf m of Diet ; IV-20
VIIl-l Jtttprt*«ncac£v« Po«rIKf<«*po(M« for Health
Effects of Chcaueals................. VlII-6
VI11-1 0Qse-fcesponse (lata Eor Careinogeaic Effects of
Z-Aceryiaminofltjorcne in Kice. vm-12
VIII-1 Factors that M©4»fj tfcue Death ®aCe from tspositre to
Cttaadeal VXII-29
YIII-4 Illustrative Increase la tfaedprtaiflty Eor Law-fiat*
Extrapolation....... VIII-50
mi-S Sbifl-Jtupnta Data *»
-------
t T
MM <
1 I H
<
a
»-*
i ~-
^55
M I i
I «» W»
5
1
S S3
i i
<
i
o»
3
i
f ? ?
*» fes M
I I
D
- !?
O 6
ms ,
p n f « c a
nt 2 n Ml
g r*
4 « « 2) ¥*
9 **~ *~ *-« * ®
4 « < W O
9 # r* #
r» c
a X •** »*
2 *•
* m
rt
*
r»»
a
• » * H
1 n n n
flP|
. . ^ •»!
•*» »»• • m a •
0 o «r f* o a
A P t» *- # #?
««8l3
H & m» m
V ft B IS* n J?
k- O » H
fir* I
m • |r **•#
9 • 5 9 A
X »* »
11
-I
1 1
I 3
II
K P.
5*8
r-
r» c
§e 6 5 o
ft •» «® O • »•»
¦ fnfl O
S •«» w ~
o
H
«*
8
m
t
«t
rt
§
ft
#n
A
P
as
i
I
*:
m
?.
5
»
s
? 7 t
«j #» -i
X
I
X
I
9< «c < < < <
I M H M H H
77 I
S
<<<
K M «
I I i
Ui M H-
O a» U
5
I
< y
*¦* <
t * 4
! 2
W o
9~f
9-f
#^4
3
<
<
i
1
i
rai
-------
CONTEXTS (Continued)
List of Tables
Table T: 119 Patji
5-3 liluitfitive Physical Oilraccucitaci©a of Land
Disposal Sitis for CCI% Product Lor Vl«*cea......... B-IG
8-4 lUmtrjc w« Desiyi o£ Landfill wich a Siftfte
3yncc 1 c tit?••••¦• • B"*l3
-------
-------
M8T1ACT
A *achomology Is described for ¦yotsoatIcslly nuiiint and cow-
pa rim ebt tlm§ to tana* health and the auvironmmt of hazardous waste «an-
agoMnc «t6«fH*ciwtf• the ncthpdology solacts and links «ppropri»te nodeIs
#®4 coctxaiqua* foe performing tbo nvml M^ciMi o£ eba comparative tucii*
¦nit proem. Ttra •election of cMpotMt Mtbodi «a» bo sod on incensi**
review and oveluaciea of Owe ccdnteil and iciMct policy literature, ptrtveu-
larlp works oa eaviroemeat*! transport Modeling, boeltb effects •odd Log, *od
risk luaitOMl IN risk a«nj|gettnc concepts* Tfce goal of Cbe Methodology is
eo dovalop both bast (itLMtu of tha risks co potentially upsMdi individuals
Odd populations for «®*fc idaacifind wasta aonagO*eoc alcarnaciwo. and also
ostiMtoo of tho up par and iontt luri.es of cb* rifb rii|o «c ooo or sort con-
fidence levels, eooii4trio| botb rudoo and systoMtlc sources of encertainty.
Tho Mtbodology as prttntlf 4mlops4 is oriented toward site-specific
«i«OMisoti of sltoriMtivc creatncot, storage* ami disposal faeilicias, and
eosuifti i«v«t na jor atapix (I) Sowru Aimtsini (baserd characterisation);
(2) Invlrowneotsl transport sad fwm Analfilit CD Capo sura Prediction;
U) Health and environmental Effaces Analysis! IS) Adverts Iapace Escinacion
sod f unmet ion for Capoied Individual* and fopwltclaftil C6) Uncertainty Analy-
siot and (7) fteport and Conpare tesulti «« Appropriate. Uncertainties for
eaoh step aro A||ritit«d to 7iold the uncertainty range about tba best risk
utiiat(« An eatansiva bibliography is included* Careful appraisal of the
concepts of cbe Mtbodology doneastretion of les utility in decision Bak-
ing sre now needed, particularly in casai where alcertiect*a ratsedial actions
aro being considered for existing hazardous weece lice*.
aiii
-------
-------
ACXHOVLTDCKElfTS
Midwest Research Inscieucc It pleated to acknowledge che Helpful
r«vi«w« of pr«lien nary drafts of portion* of chit report pntvid«4 by m»my
Iadi viduaLa fro® «evar«l offices of the IJ.S* Eavlronaental Protection Agency
<¦ well ii other organic does *c EPA's request, Tb« ZTK Project Officers,
Br. Ann Fisher end H». Je*noe RriaVin, arranged Co have th*ie revleva Mde,
afld «JL«o provided technical iapot throughout this work. The reviewers oote4
Latuea chit shqu.ld t>e addrtned, ib|^1 i*d cev references to related work,
offered coA»cmcciv« criticiama, and provided valuable inaights. The report
ha» beneficed subetantially from theie *a*iaeaJica»
Individual* who iistttid in one or sore icag«« ia cJiLs work Are
liated belowT grouped by affiliation.
Environ—ncel Protection iliwey
Office of iolid tfeace
Jmi Antiszo
Jma To Bachmaier
9msm C, 8rom
Jeaea Buloum
Michael t, lurai
Eileen D» Claussen
David Friedman
Francine Jetoff
5r»c Rmlli
Harfaret A. Podolak
Cliff ftothenstein
Dale &uhter
Matthew Straus
Office of E—rgeocy end teaedi*l
—^ __
K. Jack IsefmttijliiB
littf Srucmeo
Suaan A. tbvrnalo#
William Vatavuc
Office of Pesticides
-------
Environmental Protection Agency (continued)
Office of Policy* ?l«utin« #ni Evtl»«tt<>o
D»vi4 Sais«r4
Joha CfcLsabarlcia
V, Jafl*9 COgLlAQO
Lor #31 Du&a
Alas Ho Ehrlich
SAsmml Buf-slltikfle
Hop*
Seacc Snydar
Other Organisation!
I^srry Bute
Sabocka tod Ccetp«cy,
UaJhiflgcoa, DC
lac »
Edwiti A. Comet, K.D.
C«u«r lor Occupational and
Environmental Kaalth
Jota» Hopkins tjniv«r»ifcy
Siicioori. MO
tabnec I. Rubin
School of Hygiene and Public
Health
John* Hopkins Quiver** ty
Baltimore, KB
Bill Rohrar
Pope-Raid 4»soci#Ces
SC. Paul. KM
Robert J* Coldan oad
Nathan «L Karcb
f»rch i Axeciate*
Washington, DC
Pane? (Ubumcc
Industrial Economic*» lac.,
C«MbriiB«e HA
Carolyn D. Seising
Coeriy Laboratory
Massachusetts Institute o£
TsciUMlogy
Cesfrridge, MA
Wlllia* Umm
Mwm l«*earcfe and Engineering
AasoeittiM, lac*
Hashiagcoa, DC
Itacvin $clm«Worm®n
Environmental lav tescicaca
Washington, DC
Robtrc Tardifl
Lift Sytcena* Inc.
Washington, DC
Daniel Violette
Energy And Resource Consultants,
Inc.
SouIdar, CO
Center lor IcoMnici Rajearth
Research Triangle Institute
Research Triatigla Park, IG
JIB Associates
McLean, VA
-------
CM ScitOC* Advisory lotN, tllvtrQIMMOCal fta« $m»CO—flCCce. W«»c«
la«nuo« Subco—nccoa Tho »AJ Svbcoaauctaa tovtawaii rha raoorc "Co—nn-
*o« of iitk* uU Coitt «f brurimii tiaica iteehota Dtv*l~
opmat *n4L Pilot Scwlia*" (4ac«4 SowtMrr It, If 141» mi «4cJ» the project
ceee in January afti f«bmrr I WW, and iuMd a report on it* mitv July
11, itIS. ComittM
-------
Finally, «t*im of « 4rafc of eha ptmmt raport.
-------
SUMMARY
This report prittaci the result.* of a study performed by Midwest
Research Institute (MM) far the U.8« Environmental Protection Agency (CPA) in
support of CPA's n«d» to isnai And co«pdeis that, might be used for key portions of this methodology
c. To Identify the nose g«oer«llj appliotbi#, efficientT «nd
scientifically defensible technique* mA models for each «c«#>
d. To link selected coapooettc methods into 4 flexible general
methodology that should provide support for « range of regtf*
latory decisions involving haeardous wastes
Hi overall goal mm *0 develop a methodology that could enable CPA
Co conpare both the bete of the risk* of specified mici management
alternative! and the espilaU a&eercaiocies associated wtch thaae buc esti-
mates.
Approach
A foal of our risk assessment methodology was to incorporate tech-
niques and •odela chat will yield scientifically defensible cooperative
assessmentt, and to ovoid mixing in coeiponents store properly reserved to the
riilt management portion* at the overall regulatory procees. Careful consider-
ation was gives co bow the components were defined and hov they were related.
The tarainalogy adopted in this study was systematically selected following an
extensive review and evaluation of cite evolving and still widely varying
nomenclature La use in the literature; on risk assessment and risk, suuurgement.
A coepacible, self-tonalscant sec of definitions were developed that reflect
the beat of recant usage in die severe! disciplines associated with risk
asaasseest/nanageflMnt concepts, both la the Doited States end internationally.
A framework. for the assessing and of technological risks in general
wea developed as uhmm In figure 5-1, (The health and environmental rlak
assessment and uncertainty analysis portions chat are the subject of this
report are vichin the heavy Iioes.)
Several important objectives and guidelines were identified for fit-
tint e methodology for comparative risk assessment of hazardous waste manager*
¦ant «lee*oaciv«g vichia dUs frac*eorfc!
1-1
-------
«1|§!U |«3|fo|8W|>»x tujlifUin put Vutata««y jo; m
-------
* The approach ihftuld be applicable to $ range of potential regu-
latory decision*, but ahould roc focua on making vnti etanage-
oenc 4«ciiiont on a natioaei level oo ipAcific vestes« apacific
wast* sources, or specific hind* ol waste craadaeoc, tcorigt
•cud 4ispo»»| facilities (TSDfs). The ©ethcdology should be
tticablc, however, oo apcxif ic wastes iod TSDF eeciuwelogie* at
specific site®.
* The oethadology should be sufficiently flexible to address dif-
fitwt, baMxdous wast««, TSDF technologies» modes of toxicant
release aod transport, exposure e«*4icio«s» and health or envi-
renaeatal effects.
* The methodology ahoulcft be applicable t« a raage of TSDF tech-
oologies, Including: laodfx list tor face iftpfrucdeencs; surface
SpreadingI storage in piles or containers; deep well La lectionI
cheaical and biological treaca*ftc«{ incineration on land or at.
itat and coabustion In boilers.
* The as»«ssweat of each option should coeaidcr all kind! of
rilittn, awJUiaedla environmental traasportr and health and
ettvifotumBcel effects (not just carcioogenieicy, for eaaaple)
for the life cycle of the technology or beyond if che re are
long cem effect;** Comparable contributing factors need to be
ereaced aseparably Corr eachi optiob.
* Hie Mthodblofy should yitld quantitative ratal Era when necea-
sary deta are available and yield che most ucefuL reaults pos-
sible aim data (or che ties and resources needed Co compile
dace) are iiauted.
* The eiechodology should yield information on che «oat likely
outcomes for each option, and also on che uncertaioties ia such
mtimmt for ell options considered.
Failure to be cae^refcenjlve end consistent throughout the aesesanenc
could lead to decisions chat are less protective of health and che environ-
nent« and poaaibly also acre coatly to aociety than another decision would
have been. AaseeaaMMt* aixed chenical uaitet or diapoaal options
that could Lead to axpesuree to different chemicals with different health
effects anise be particularly careful co avoid ucMiiitMC analysis that could
lead to perverse decisions.
The Methodology
The auithodology for eeseasiAg the health and environmental risks of
haaardoua waste auanagwusnt alternatives depends on four fuodaaental assueip-
tiona: (a) aequeatiel analysis step* can be defined and linked to provide an
overall aasesaa»uit of the health and eirviroweeoCal rialut of basardbue waste
BMUiagaflMtat alternativest (b) eech seep caa be performed by utilizing a com-
bination of acaoerioe, available or estiaeted date, and predictive andels;
-------
(c) best "poLac" #1 cha ritfca can be obtained by using the most
li-ktif cat* assumption*, lofuc daca* tad predictive c»o da Li throughout» rather
th*a inceotionally "cewwrvnciW or "opt iaistic** choices; and upper and
lewtr confidence Ii*its of ttat overall ri#k can be obtained by licit
Lng eh* uncertainties fro® boeh randoe «a4 systematic source* of error in th#
assumption a, input data and aodelt, and then *|gr
-------
T
©•unify and
Sootltlva
Data llIM
GkOmctOfita
Hotafrftow Wad*
md Tfoatmont/
Tachnoloflttft
Htalffc
Effect*
Onto
Idanftlfy and
' Analyse
Character! za
Environmental
Prod let
«p*
PalManti i
'«M
Ttonipart and !
¦
tirpMtifM
Ratauad
For*
ImimHIItKill
Madia Me
.1
l> #inn iiTft flfr-rE'—**¦
jiiyj iiii ii qrj W^J
Oala Im
%fly Oom/
Funcilom I®
IllSSI-I^Ssrf
Individual*
Cttimaf*
Htafrit Effactf
fm
PapuioNoftt
Hua»n Dds«/
Functions
/ Popart Bat r * v
H' (|tf imt«« and ^
< Uncertainty I
Hor9«(
• u
rfwS»S'«i: i::5W::::SfS5« ^.SJVSI
linc«rtointy Analyils at 6or« l-J - II«e»H ia Health Ilfl Atitiimil #f fUnrdou* Matte DtapaMl Haihadt
-------
m
i
p.
ffi
in
m
I!"
ill
III
lift
c
o
u
ft
u
0
in
2
¦n
<
i#
0
a
1
jb
u
m
&>
!«
H
-------
Identify i»ajor •avireotttar®' ~ransport routes and transforma-
tion pathways for chejaical.9 af concern
Esc intern concentration* of chemical* (or their haurdMii trAfla-
fornatloa products) along these route* over time
I4«neify Iqmcioo) eLong these route* where Che cfes«ic®l5 may
reach susceptible populations
A »chana for the selectioa of suitable models for - anal?ait of pol-
lutant traMfttrc in diffcrwc stadia in given in Figure 8-4.
1® Exposure prediction involves estimation of tbe d«grt« co which
pollutants reach humans or other organLsas, including?
Estimate numbers of people or org«n£«n» thac may be contacted
Predict frequency, intansicyt and duration of tbe exposures
that any occur for population* or subpopulacion*
^ ta^irmiKBiil dfictt a—I wit invelvaat
B*vie%» iactetiwtiy ch* literature on chc health and environ-
mental effects for the ahMaical* of concern
Identify ranges of reepo&Jat known, md especially those that
could occur et pradictad environmental. exposure}
Select or develop chamioai-ap#«tfic dose-reaponse relationships
for effects of concero
Develop (throttfh selee-ced evbrapolacioft nethoda) -rick factors
for specific predicted environmental doses
k scheme feT selection of health effects estimation models, depend-
ing on the availability of data i* shown in figure %-$.
The preferred dlot
-------
fete
¦tb
If If AM*
IfJOli
>n.
GWriter
riUMt,
AlillOw
•TMlllM
co*»
U?M, m
UJM-fOW
Figwra S-4 - Sckm lor Selecting of Ntdali far Analysis of Triaifntt «al fate
•f foilut«at« Catering Um SavlrMMcat
-------
¦a#
-------
'• impact aseiatcioB and summation is an Integrative step
that Involves estumciofi of EEe-prebebllitLea and extent a of adv«r»« health
•ltd environmental effects aec«*lly axpactad from the pradacted txpoiurti,
including:
Apply ikK(rrSifonM functional to exposed iftdividttala by
tspo«ur« gfOvp
Id«ati£y any »p»ci#ily aemslciv* 4ubpopalations o£ concern
Hake best estimate* of the risks for each efface to catch
subpopulatioa
Make best estimate* of zhm total number of c«s«« of each effect
Estimate the uaeertgicty rang* lor each best estisaw
6. Uncertainty analysis ccpaidara both ayscamatle and random
sources of error? is *11 «i
-------
~ Provide aittil®.? best •itiiwcta and uoctrtiincin. If appro-
priate for ocolofical or other envi-roaiusBCal riika, coat* of
control cacthnolog tea, amd ochtt «iiaeiic and ica raeults «ad conclusions
Make oral presentation* «i appropriate
M«ke special effort co place the rialts, costs, and, benefit* of
the alternatives Lace piiripBctivei helpful to the reader,
eudiettee, and decision mkir
-------
factor** (Thm ruutt of auch a coebinecioe mN to m «ici«ic« of ov*r«l I
•iftcc ac >om ill-4«fia«d point battwawa aa MLC aim! « writ cmi.)
The JojJLytis 4mb addrta* coolidmco llanta in tbt u&cartAincy
Mslftis* a ad both the tost titiMU and ttui mc«rt«(aCf «•*»§• cm, b« provided
co cto tfftciiion awfcer. te«lf«ia of cmcartAintiaa conaidara boch ijicmicIc
*fti rtndM sources of error. tfneertaiotiaa are aatirtAtad lee eech factor
CMirlbuciat Co the rial* for aacb option, ufiiiad. and cKan scatad with th«
(ABI confidence bounda for all factors** The tukcereaiatiaa mr« «itr«|i(id
icron the entire (iigimot for eecb «ptioo «od risk rafl|«i for opciona are
ceapo*«4«
tbo attctodology focuaea m, pmMic toAitk i*M«cs of h**aw4^*»
CcolOficeL Ml sociearcocMaie iapacti
-------
x« imooucrzoii
Decision staking to cM public interest requires consideration of che
rant* *f consequences — biotfici, conrs® and risks to individuals* groups,
business, and the environment — posed by alternative taohiiQlof ieal choicee*
legielacioa of cm requires that govtnunat «|eAcici and industry reduce riaks
to hiwu health and the environment to "reasonable" UwU. Executive Order
11291 directs federal agencies to prtpitt tegMLlacory lepact Analyses (RIAs)
that deecribe the benefit* and costs to society for prop#*** major regula-
tions* A eejor policy issue c«nfrottcxi>| regulatory agencies it how base to
choose snong multiple regulatory end nonreguletory 4pf«iaata« ia faanagiag oar
diverse technologies when the
A. Bisk.tic 8oU 4» Decision Hakine
Keceat studies by coemrictees ol the National Research Council. a
study group of the loyal Sodsty, and by others haw concluded that the
3cicnce~besad parts of thai risk. assesasMoic process should be as independent as
possible of ttur ultimate decision-eiakxag scepe, although tbe latter obviously
should draw extensively on the former foee MAC/HAS, 19821 tfRC/NAS, 1983; and
Royal Society* 1983)» litis introduced in recent U-S- Congresses (e.g.r the
"Risk Assessment Research sod Demonstration Act of 196J") also reflect « con-
clusion that information on the rifle* of m given choice should be cotrpared
with the riaka of other pventical choices «nd with acker everyday risks* so
that the decision outers and che public can gain a useful perspective of che
risksr costs* and benefits of tha available options (l/.S. Conireat* I98SK
teacher lesson Croat recent study is that the uncertain tiara in cstiaaeiaf tlxe
riaka should ba aede explicit• A stateoent of estimeted adverse health
efface a should identify not juae a sioflc-velumd prediction* but a range of
reesonably probable forecast** The decision maker can then compare the ranges
well as the palet estimates,
Recent research has else stove e need to reevaluate the role of
in mttrnming end *l»k« Hekiof a wco»»«rv«ci*« 4eci-
«ioow one that is likely to be eore protective of health «ad the envi-
ronaeQc then an alternative decision) is videly accepted as a prxadenc practice
Lb riak. fluuugeaMst. It) keepiftf «ith che reooeaieiided aeparacioo of risk
assessaenc and risk nenageeusnt eetivitiee* ho«revert (Kmeervetive assuatptions*
coaaewative andela# cooumtivi aseiaecea, etc.* should noc be key elessents
in Cte science^beaed risk estimation steps. " A catenation of conservative
aastaepciena * models and estinatea thronf^hout a risk aasesssiant can lead to a
-------
"woi'st-csse" (or even worsc-ef-cf»«-«e«,at-*eaisi§i3 prediction that may be of
lied* value Cor possibly Misleading) co the decision naker. ?or chis reason,
the concept of "verst-credibLa ease" bas received sooe emphasis recently.*
Moat decisions on hscardous «iiLt sitea actually involve "either-or"
ahxrices btc»««n technologic*! alternative# with different risk ltvtit rethec
Chan « "yea^no™ cbotee on « ttngla elite. When disaieiler alternatives require
different analysis proca4ur«tt comitvatiin ambiguously o* inconsistently
applied could lead to biased reeults and poor decisions — even to the chains
of « technology chat is Ian protective of human tieelth and the environment
and possibly wore coccly co society tit an an available alternative* Best
esciamce* of tha risks, costs, and benefit* for the alternatives, coupled with
consideration of their uncertainties (including wr»c-credibLe uia consider-
ations >, should produce the option! basis ipr decision staking* The Council on
Efivirofliaafltal Quality has recently noted thac "rules of reason" shoold refines
worst case analysis as Che besie of r«t*L*ca«7 decision making (CEQ,. 1985).
The Council has also noted dm need to and tha difficulty of staking environ-
mental regulations in the i«ea of ieacnpLece or unavailable information, and
haa made suggestions (CEQ, 1984J.
a. gjufc Aase^aaeng Meeds Under RCRA end
The Resource Conservafc ion and Recovery Act (Uli) of 1974 and Haz-
ardous and Sol id Haiti Anandnents (USVA) of 1984 are ifctjor legislation con-
trol I in§ hazardous waste disposal* Under RCIA, 6fA designate* specific cliir
ice Is and specific industrial waste streams as bstrMM end regulates their
traeceanc, storage, and disposal* RCRA end H5WA ar« less specific about the
need co determine the ress cause or significantly coa-
cribute to aa Increase in a»rcality or to an Lacrosse in serious irreversible
or incepacicetiog reversible illness) or (b) pose a substantial present «r
potential baierd** co buaun health or the environsiBnt when improperly treated,
stored, transported, disposed of* or otherwise niisunagcd**1
Sec* 3013 of 1C1A, "Monitoring, Analysis, and Testing," addresses
the issue of risk indirectly* It states that upon receiving information that
d»® preaeace or release of ^h»«c4om0 te at e treatment, eeorage* or
*• Efforts to define «eti(-ercdibU caaei cm, elso be controveraiali tbe
eourca hed to ftre^ple «ich cMs iejae*
** tGLA does net define heeerd,1 but m u«ed here to define 'haaardous waste*
it epfteers to be esaentielly aynaaoaMua eith cotiventioaal definitions of
'riek.'
-------
disposal facility (TSDP) may present a substantial laiacd to human health or
the aavircHuaetic, due EPA Administrator can repair* chat the owner/operator of
the site conduat (fcHiitorlwf# testing mi *naiy»ii to determine eh« nature and
uctnc of such haaard.
Sec. 3019, n£jcpo*ure Xafonaetioo and -temitfe Asuituncs" (mMni to
the 1984 jKMmdnents), speaks loose directly Co risrk needs. tt
identifies the kinda of information valeted Co potential huawn eapasure to
aavtroawaCAl contaminant* clMt m mmm of a hasardous wiau TSOP
n»ac »ubmit to EPA before a permit to operate is granted. It provides proce-
te« for determining If health isxisusti tri needed, for establishing pri-
orities aaMg aoeh e««4s» and for initiating itKh itsttiiNACi. It also
defines the general contents o£ health assessments to include evaluation o£
riakj to potentialIy effected populations*
The MWA directed a phase-ovt; bf 1990 of pcft»i
-------
C. Study Objaccive*. OrigLoa. and Scope
Ttif cooaicmanc of EPA to mini risk asses-ssiant effectively in Making
coa«iitmc regulatory decisions in all of its prOgrj« arui has been expressed
hy present And previous EPA Administrators And La described in a Wi4 report
to cha public
-------
3. Sgojgjj: The uop< of this studf had three major casks* (I) an
overview ol the risk ittittMiic proofs* (including m»ce»s*ty definitions) and
of the components gf riik siiitiatiti; far. hazardous wastes disposal; it) re*
views at (methods far specific steps in M*«rdoa* wa«ce rink, nwiiaeBt includ"
int sources assessment, analysis of environmental transport and f«tq, exposure
prediction, analysis of healed ifftcta and prediction, asaestsient of environ-
mental and other iap*ce«., ietpact integration, and uncertainty analysis; and
(3) developneot of a ganaral #pprofck foe assessing rifki in haaardous uasta
management*
Our study wai c«Mtr»iae4 ia scope by available resource®. Several
assumption* and limitations h#« made, tha oast significant of ufeich are noted
here, because fi«y nay emquira further comid^mcioo La other specific risk
«i«es»uc applications.
" A solid wiit# w&i defined ** hazardous If it is IUt«4 in the
appropriate taction of the Code of Federal Regulations or
exhibita j significant of igai6«i»tLiey, carraairity,
re*et£vic,f or to*icitj»
The risk assessment approach should b« orianted toward RCRA-
r«lac*d decisions concerning hazardous waste disposal.
• Tha approach should be applicable to specific ha^ardou* waste
treatment, storage md disposal facilities (TlOFi) or to speci-
fied hazardous wastes.
• Tha approach should be applicable to multiple technologies,
including* iiliti surface jatpoumimanti; surface spreading!
icora|i in piles or container*} detp «mLI injection; chemical
and biological treatments; incineration oa land or at sea; and
coobusticn la boilers.
• tha approach shoald be applicable to existing sites containing
to«art»tja wastes, e&istiag TSDFs« and pr«^i«4 new central or
dispersed facilities.
• Tha approach should be applicable Co a range of potential regu-
latory decisions, suck est exclusion or acceptance of desig-
nated >ut«i (e.g., fr«« a landfill or incloeracor); proeolga-
cioo of specification* or standards for given diapoaal tech-
nologies uars# hjrdvogaoLogical and nacaorological caoditioMf human
papulation diatYibutioa and othuatr aovironaMoeal factors).
• That approach naad not be oriantad now to (taking uaita manage-
awnt daciaiona on a national Iaval on specific testes or waste
sources. Vacional consideracions vill be cakaa into account as
needed in fucure decisions*
1-5
-------
the ipprotch should adicewi all release* of pollatuoti from
TSDFs to air and water tlwc ««y effect ch* public. ton**#**,
rtleiMJ Icon transportation of haxarctoue wiitti are noted only
briefly. they auk? iw*d to ' be aoesid*r»d ic ItnfCt* |« n*ny
rwl-tife 4ecisl«n«#
Ttui AfpfOACto should take & I oar« of huipaii or nonkuman r*c«ptars.
The approach, should focus on the potential Huron health effects
of roxic pollutants. It should '«44r*«a pri^arxly the health
ritki of the icnaral population in cite vielnicy of TSt>P.
The pttblic is assumed bqC Co have access co* tit# sic*. Occupa-
tional exposures ac Che TS-W a*e noted only briefly* ehmy may
rrquire further co«tldwi« in «or*s decisions.
Methods for assessing iapqccs on the physical and ecological
enviroMMDt are examined only briefly. Inpacts on such factors
as cht atmosphere and eha flora and fauna otay require detailed
evaluation in sobs decisions.
Tha study notes only vary briefly sociaecoooaic isipaccs. Sucb
factors as b«n*flt~ca*C ratios, equity considerations, land-use
planning and social acceptability of alternatives way require
io~depth consideration in sone decisions.
-------
ftcfBrancaa to Chapter I
CEQ. "Nation*I EnvirDDBcncil Policy Act UjuUcioo): PropoiKt Amodmat to
AO CF1 1502.22." Council on Environmental Quality, Enecutiv# Of Pica of
th« Praaidact. federal Ruintf. 50(154) 32234-B, Aofujit 9, ltl5.
CEQ» "National Eavironaumcal Policy Act Elagolacieoa: Incooplata or Ua«v«il-
itbli InformationCouncil on Environmental Policy. Padaral itfiftir,
SIC80> l»l9»13*2t, April 25, 1916. —
EW. 11 tk Assos—nt and HtmjtMBtj ffmaworit for D^ciiiwi Making. Envi-
ronmental Protection Agency. EPA 660/9-85-002. ' Washington, D.C.
DfOMbtr 1994* 31 pp.
Mil, "Health Sisk AiMftntnc Mathcdologian for ttCRA fctfulmtorj Analytis"
ttrafc Import. Hidvest leaaarch Institute. EPA Contract Vo» 68-01-6621
(Subcontract Ho. 30*6, Vteck AtiitooMci 1J and 27 fm rcf, Inc. I, Office
of Policy Analysis, U.5. EOvironaantal Protection Agency. Washington,
D.C, VovBrtMr 7, 1983.
MUX* "Pilot tiik Aralyaia for Land Disposal Prohibition of Certain Hazardous
Wastes." Draft lepcrc. Kid«ra$c Research. Institute. CPA Cant:race Ho,
(Subcontract tie. 3G-f, Work Assignments Ma. 9S fron tCT, tnc#).
Office of Policy Analysia, U.S. Esvrironaeatal Protection Agexicy.
UmUubCoQi D.C. April 19, 1984*
Mil. "Caspar!son of Risks and Coaca o£ Haaardeua Haste Alternacivea: Kathode
Development and Pilot Studies.11 Draft Report. Nidwaat Raaaarcti Insti-
tute. CPA Contract No. 66-01-6558*(Subcontract Ho. 133.1S5, Work Asaign-
isent So- 24, fro* Sabo(ki and Cesrpacy), Office of Policy Analysis, U.S.
Environmental Protection Agency. Washington. D.C. November 19, 1984.
NRC/nas. Risk Peciaioa Kafcig*1 Perception* aad Research. Conaittee on
liak nod Decisionmaking tffoward Raiffa, Chainaan). AaaciabLy of behavioral
and Social Science#, Rational Research Council• Itational Academy
Praaa. Washington, D.C. 1942* 68 pp*
dlC/NAS. iitk A*aasaBent in the FadaraL Covarrmaoc: Hanaging tha Procesa.
Ceanittoa on tha Institutional Haana for Aasassoant ot lisfca to Public
SealEh (RaueL A. Stallones, Chairman), Comix* ioo an Lifa Sciences,
National Research CounciL. National Acadasty Praaa. Washington, D.C.
1963. 191 pp.
Royal Sociaey. tiak As sea feat. Study Group on iiik (Frederick Varncr*
Chairman). Tha Koyal Society* London. January 1983. 198 pp.
i.S» CMgrtM, "liak Asaessaaat Research and DcaoBSCracion Act of 1985*" Kill
M.1.2749. Introduced in 99th Coogress by Congressaaa Donald litter and
ochara* Jnae 12, 1985. 13 pp.
lm7
-------
-------
It. If*
jMPtT Mm mi ft mmmomtr
*~ Towiaolour
list tiiiianaflc/uoiguific (Mcifci 4«*alop«4 ia4*peo4*4cly ia
4iv*ru fialda with atsbstaoeially distiact llvoracuroa aa4 diffariog {urtiMl-
ogiaa* Tba riatcralatad literature has (row ripiily 1b tha pair decade or
•o, mm4 lateractiao* tat«f«ea remttbiri is fiol4a have iacmit^i
but a | A and taHariMi to the t*»o rabsectioea follotriag.
I. ¦•aaroea utlnri feavt defined riak-rel«C04
mfm, tat cb«iTTi««~MiIdwa will aach ochor aa* b*v« ac ciaai baaa
iocooiitKac m ctofvaiif «4it>4i» (Mr *m ••«§•» A •** WMbur^kjf
of usage Ia tbe Ucerature is ia Apf dia A. Overall, the
licoratur* reveala tbac risk himsmbc ia m loaatur* 4ie*iptiac without Ota
klml of Itafv4er4tse4 n iwasncUtunre aoa flats Is an tUir *iaaiplioa *«ch as one
of lb* physical* biological, or aocial tdmn.
fa vara I recant publications have attempted to define cams, tat none
can b« r«ftri«4 41 the final word. In above* thai leroift&logy is still evolv-
ing. It lacks aaasiatancy, a van for auch basic eem*i aa "risk," **ri»k <(¦•«!-
nent," and "risk ¦aaageamnt."
Sooni authors faal that "riotf* oust ba a quantitative cifrtsilw of
probability* while others do noti a far* equate rial* with "uSK^rtaioCy," aiMn
vitb "kmHi" Saaa tafiaa "bastard"1 aa laJtrisaie Ctticicy( ethers a* a iU«4f
ei«i Ult
-------
nacageeiant. for example* M4UtL«wcio^" in call*4 "chjufactarisac ion^j "evalua-
tion" is used la wsv^rgl ways; «-"* "MBageftcne" is of tan Urn tad to "reaching
a decision." One recent «uthor classified "risk anAlysia" as a subeet of coat-
baoefit analysis, vhile aaatkar curiously subsumed risk assessment ®n4 miiagi-
unc undar "analysis•"
fltff*r«nt i|«oci«i of |«rvtnrnawnc *od different cnaesieteae of the
Nation*! bmrch Council, latismal A««4**y of Science*, have u**d substacr*
cially diffttiat ctnu and definitions, to illustration «f tk« diversity that
oitci among Authors oo basic earns i« seen In Tabla II-l. In addition, the
ever*LI process of scudyiag and auuiaging technological risk* tends naturally
to hava several sequential seepst fWBY analysts c
-------
TABLE II-1
EXAMPLES Of PITVOISE TEHMMMjOCy li RISK ASSESSMENT UTCTMUttE
lefcrcace
OtMmj § fafcaer, 1976
Roue, If7?J 1583
Ca^nrehettmivc
MA Tcfiut
Aaa«*soenl
Asaetaafeftl
K»tea, lf?lj
Mbyte 5 Bortoct, J 910
Sancrs, 1579
CoiiMi, 1980;
Greer-Wonttea, 1980 i
et *1., I|12
Gmmd et, al. , 1980
Aatesattedt
HiM|figuit
Aitptineol
A»se«Meot
NK/1IAS, lit®
Aiiemeot
frittCjpal Coappogala
Kstiawti«ut
Eva}u»Ltoo
Deteratoalion
ldentificalion
E.*lUMtioR
Evaluation
Ha card identification
ftiii es Lioution
Kiak eveluMion
Aoalyel* (or
Evaluation
tUiMfescat
(ilwftim)
Exposure MiciSKat
Dotage (bayard)
aaaesataenl
Expgatsrv analysia
Pathological activity or
besard appraiMl
Combination of cxpoaure
and hazard eatlnetee
Subtena*
i
Identi ficatloo
SstlMtiMI
I
Averaion
Acceptance
i
I
Magnitude
Probability
Adverse effect
Probabtlily
Partar et al., 1980
Aaaeaaaeat tequaiei with rlsfc/bcncfIt analysis)
-------
liiiaggaae®
Cam—f, I til
MKfllAS, 1112
fVRt/ftMt 1963
cwt nai
loytl S*d*ty» Wl
nsMCp imi
D«Tiei» IIM
CMfralcttive
lUfc Tmwmm
tealftfts
*§•••**01
AM**M«nL/
tulyiii
ANIMMat
Aaatyste
TAJLE I 1-1 (Cootinuvd)
AtMlyfti* («V*|lMtiM of
•CiMtlflC 4tl«)
Kv«twitlM of
tiiicai
f«ct»r«
E«ltult(OB
Utard licit loo
Ooa*~r««po«tao iiitiuuni
Expoxtia iimiMnt
HIml cb« radix Jut ion
8ourC« ilMMHRl
PfttfcMy- 4k Wt H C 1 ft t
1 apart mmkmaI
biianii*
iittuaaat
RiducUcu
Man«|M>ant
htiutioR ef pro!
of baatird occuri
Dcuniution of ljfpt» of
hiiirdi piMtd
ildralifkillM tf nUbwi
Eftl.iBal.loA of Mtpiluda o( tMHfMK4%
ItiwUw of probibiiuiai of Miicofer
Ovieniae »i|li(iCMc«
pts4y tr«
-------
Meferttflca
Cocprefaeaiivc
IU»k T*mm
Ikvli, 1984
(cottiianad)
Hoiiilui, IBM,
Park & Soee, 19(4
Asse«8ieot
(continued)
tfa»®je*eist
Control
AaMlysia
Aiteineot
tl*n*gc*irat (or
regulatory rt-
aponte)
Snitt, 1914
AlKDSMUlt
TABLE It-1 (CmUiiuimI)
Principal Conwnewt*
EsLiiatioa of auaher of
people cxpoteU and
oouter incurring
adverse
Ac^eplaliiIlly judj*«ni
Action aelerttoo
lopleaentation
Evaluation of reaulls
Priority setting
Aaseaaoent
NuifCMnt
AJtetfnent
Itaiard tdtntificat ion
Hazard evaluation
Rick evaluation
kMtuitioo of benefit#
wrw» riaks* casta,
•nil alternatives
Hiurtf Identification
Hazard evaluation
Exposure 1dentilirati mi
Exposure evaluation
SubtcrvK
!Engineering failure ww^went
Expoaure aaaeaaroeni
Effects nteBtaeat
Risk characterization
AiMtment policy
/ RUsk value
I Unce rtJftinty-of-risk v*luc
f Coat iflipoct
ICatt/fctfneFil analyaie
I Perceptions, constraint*, intangibles
-------
Comprehensive
Itefimocg Riah Touiki
USMRC, IfM m4 1986 Probabilistic
riftk aseessncnt
USMRC, 1915 Bisk **iu|«Reat
. OSTTj If85 Ajiteaaaieat
lied mt ml., IMS
lUilar & Thanaa, J9US AoalysJ*
TABLE II-1 {Cant intMd)
Principal CogpQfteftf
Sy&tcuui analytic
Fiult Lrea/evcflt Ire#
aaalyais
llumn factor analysis
Accident precursor
analysis
Accident sequence
analysis
CofltataoKMil analysis
Accident prevention
Accident Bsnafenent
Consequence Mitigation
Hazard (toxicity) ass«ss-
neat
Exposure Mattfiwm
Ksaird
Context
Couaqueftcei
Untert*inty
Scf*rit.-y
tti]ptiLude
Study
Interpretation
8ubter»a
Hazard identification
Risk sstessncnt t«ife.vAlofp£fti of condi
lionol probabilities)
Risk-benefit analysis
/Risk appraisal
^Translation (into relevant tenss of
I political theory, Uw, ecoaomct,
' and huMM bditvior)
-------
I
I
II
*i s
u
18
•
,f
iz
mt m
Is
•at
8" 52
11
;?
21
n
mm
m-
"2 *
a sf
** «•
<*¦ a
> «* -«
e
4
%
«
i 4
Jci s:
•mm ¦ ,
c «*
¦ 8 i» «* I • i
m -* m m m I
!i« ii ,» :!,? <*4 ,p
«t •¦ u «- « *» i
S 3 to 5 1
*g£
1
s i*s"
O to
%, m
mt i>K
m
8 s
fcM
# ia#
•m
IN ^
5 1
»» «
^ to
v m u
m m m ti
#» u u m
•eir
"P * " «,
*2
3«i
U « 9
d <¦ *
£-«•«*
b b u
* h»
#ll
1
J
Ml
tf!
1
j
i
W
« <• *t w
• 8 8 2-
8 ZS S2 1
• «t v v ¦* - ¦
A 2 B j
«, o « 5
liflff* -
" • 2 5 M H
W m m > «» u «•
8! a 111 B Si: (I M
I
a
-F
«
1
s
<1
«
2
at
0
¦
¦
2
«t
m
rt
2
tf
a <•# «* «<
111
11 5 f
: S
«* —i
1
1
t
2
g
«*
i 1
¦*
i
i
s
Mi
3
u
to
a
:»(!
[
2
1
>»
s
•n
Bl
m
u
s
«l
(M
X
3
tf
a
to
5
to
to •
S i
8
i
i
ii-i
-------
Reference!
USOHHS, 19M
Co^rdbeoaiv®
II*It. Tew
Aaae«sneiit
Kalleobeck & CuofilBitea, Asicstaeat
1986
Saith ct al., 1981
Aiseaiwnt
TABLE 11-1 (Concluded)
Principal Cowpoaentg
ji.iiujiu „L.uJuiiu niiiaiAuiwiiWaM jmHMit.iniii .aui¦ uiu, jiiil j
Subtenns
Hazard iifantificatton
HiMrd characterisation
Kvpoitire charactEriutloc
Riah deteraination
Qualitative or quantitative detennina-
tion of potential haxard froc •
condition or aob*tancc
Characterize action of hazard
Dose-rcaponae asaeasaienC
Detrraioe difference* to riak -across
subpopulatiaas
Qualitative and quantitative evaluation
1 of likely exposure
Integration
Binary (riafc/no tiak) conctueiana
Quantitative aajltitiloenaional conctu-
aiona (including sensitivity testing
a»d uncertainty cbaimct»r ixat I on)
Exposure characterization
Heal lb off ecu charocter-
ixatiOA
Ruk analysi m
AuapUbis concent ml ion
calculation
Qu0lit»livc evaluations
Quant1Cattve evaiuationa
Individual excesa ruk
Number excesa caaea
Scnaitivity aiulyaia
Uiliril identifi cation
Ke?a*rf iccoaatioi
(defining
ayeten)
Kaviroiueeotal pathway
evaluation
Riak cliaracterizatioa
(Iiumb health mJ ew
ayeten)
Riak iun«g«*ei»t
-------
An ><11011 or combination of actions (events, condlcLootft challenges.
decision!, dc ceuoes) produces one or, nre cottamouancea (outcomes, responses,
impacts, or aff«tf?s)« These consequence* ere judged to be beneficial (favor-
advantageous, propitious or tMk), tdnrrt (detrimental, 'harmful ,
injurious, or bad), or neutral, on che *elue jysfrwt-of che parcy or
pi'tiu dotftf the judfing** A potential consequence deemed to be adverse is
vieved aa a the—c.** An action or condition posing a chtuc la a hazard,
!«••» a luuurdl ii « poctaciil lourca of sdvtrsn respects, or in popular t«r-
minology, a lourcc of ri»k.m Thia atudy foeuses on tdvrt«)
• A quaiicetlvg jtatewant of che livelihood or poesibilitT of
occurrence of one or aore identified idvtrse effects» based. on
partial or minimal iiifjocmeCion or historical perspective-
In many cases of ratulttoey toc#r#« 4 sequence of several atep*
(actions and their consequences) may be involved beewften the initial source
and ultimate consequence* of concern, tueh ae actual effects on human
health. For environmentally releted wlska, a designated location, activity or
• A consequence deemed beneficial by one person or group mey be deemed o£ no
velum or adverse by enother (or even by the seme person at another tine),
e.g., a "small darter>n a- aero sum gam# an election moult, or a eui-*
cidm* Values thus enter into the stodiy of risk* at the cerlicst stage,
and ameer subsequently in many ways, either explicitly or subtly.
Adverse health effects should generally be objectively definable.
** "Poceaciel1" meane "couW occur in the future*"
**•4 ceaadctee of the Ketionel Research Council stated thee the heejard of a
chmoiieel U a function of both ics incri«*£« ce^ieicy endl use pattern
-------
uee Croat «hich a toxic ubitaflaa encers thl» sequence mj be c«osld«r*i mm a
hazard. 6«po»urc is »aid to occur at the p#i«e in the sequence at which the
coxic substance it pceee&t »t the interface tie environment aoJ the
biological orgeaisxu The ettviroo—gsal «toat La the 4*oubc of che substance
chat Actually contacts or enters the organie* through bodily fMubrenee end
portalfti intact or broken eyex* na»e, and ¦autlu* Subsequent steps
eithiu the arf»is« mj be involved before the »p«eified- *ffecc occurs, The
risk for itui effect i« then estiaaeed by -appropriate oeLoulation hased on the
cheaicala, Kipi and coodiclone involved and assumption* required (as will be
diic*uf«4 later).
"Expoaure moiBcnc" is often used to deacribe different uc* of
acciviciea by researchers with different viewpoints* Some r«»e«rchera con-
aider exposure aneiuaitc to beg la tricki cheeical analysis of eta contasinants
la the air. vatirr food, $orf«oM, etc. , vhich people actually inhale, ingest,
touchy arc, if teowiaf concentrations and aaauaiing the daily volume of air
breathed, capwater consvaeed, etc. r these researchers e*ciiaeta the exposure
rate and exposure dose over e given duration Cor each exposed individual. The
exposure assessment is usually completed by quantifyiag the known or assuxied
population at risk. Other reseerchars, hoiwever, may piece greater emphasis in
exposure assessment oo eoalyxinf or o&deHog che transport #1*4 interact ions of
chenical subataacea fro*» this point et which 6hey enter (he environment through
a^r, weter, land* «nd ecoayeteae until tfuty reach receptor papulations* i,
they quantify the pointa in apace and tiee thet toxiceocs and populations
intercept each at her, aa well as predict the. exteftta of exposures chac could
result. Still other researchers amy begin the expoeure assessment %rith ^ueli-
teclve tod quantitative wuipe* of natty of cmtuiMiiy ineg the
environment, i.e., with an es*e*«Mtnt of the sources. Thus, Hushon end
Clsraan (1981) divided exposure uieiseaot inco five stepat chenieal descrip-
tion; naterials baieneef petheeys Of emri renown tai releasel population pro-
f ileal and "a4Mtesextent" 1a which the environmental concentrations and popula-
tion profiles are ceieb ined m live exposure profiles. In one recent report
(OIAt 1964), the exposure easesssamc vat extended to include absorption m4
transport within the organise to thje $it« of coxic activity. Hany additia&el
factors can effect biologituU. upceke af a given chaaical, including the
preaence of other eJt«ei«#l*»
to the preaent research, the tama ^source esaesusent," wenrvdron-
eMastel trsnapart and faca analyais," and Mexposure predLccion'* ere ujied far
clarity inacead of the less dafinitive Maotpeaure ae*essaeacM terstinology.
He edopt *rirt eaeessveat" to describe a hro«d efcudy procese** con-
aisting of aevsHral eAeiyticel conpeoents. The uiwi of tpeeific aselytical
coo^oneata varies with a given rltk aiienwc probl«nt hue generally
# E&virooaHMital doae as defined here ie eaelojgous to the ededniatered doee in
a controlled toxicological study.
•* The eanaa "risk enelyaia" and "riaVi evaluation,*1 uhieh have bean defined
¦eny ways in the litmnuure, are nQt uaed* Neither of theae ceraa ara
included in IF*'# xtm "Cloeaery «f EaviroemmteL terma,'* but "riak
asaeaaoenc" Le include! (Ft4, ifM).
II-XO
-------
includes: identification arid charact*nzat i on of sonrcis of pntantial ehr«4cs
(i.e., Kicirdi) 1 idencIficacion cifli«« expressed as estimated probabilities of iapaces par average or (test
aapoaed UHtifridual, or n integrated estioates of the total lapacca on an
exposed papulation. Io either case. the -results should provide act only chat
peine estimates but also the ran|c« of uncertainty for each estimate. k risk
•tody to chit point consiscs of an analysis—la as scienci firstly defensible,
veiueHfree* a Banner as poasihla—of several elements associated "with poten-
tial adwrsa affacts of a situation. Variation* Of the risk, assessment pro-
cess cm be vade, for exanpla, to yiald escinacea of suucifLun exposure levels
or source strengths that can be permictad while achieving a seated acceptable
risk level.
B. Framework for Assessing and Managing
The making and LapleoancinB. of decisidns to control risks requires
—ich more information then tha estimates of their magnitude. It uaually
requires abjective estiMtes and comparison of the costs and benefits of
available alternatives, i.e., conpsrtdvc risk-cast-bene*it assesamenc**,
inforaacioe on public perceptions of cbe risk and on feasibility of iaplesMuit-
i0J| V&fiQua HUfia^iBArtC LS «Ud uSuaLly COfi*IddCAd. Nil1 a view of
an overall framework for assessing and managing technological risk ia outlined
is Figure 11-1. The healch and tnvirenaeotal risk assaasaenc, an important
pare; ia the overall procui, is ihova with seven coaponencs aa developed in
thic study for hazardous waates: (11 source (haxard) assessment; (2) trans-
port and face analysis; (3) exposure prediction; (4) healcb and environmental
effects analysia; (5) adverse l®p*ce estimation integration; (6} uncertainty
analysisi and <7} report and compere the results as necessary. Some of these
consonance would differ in assessing tha riska of other technologic*.
Important aspects of comparrartlve risk-cosc-benefit assessmenc end
risk management are discussed briefly below.
I. Comparative gisfc-cost-benefit _assessmeat* Cooperative risk-
co*t~benefit assessment is a sciemceHsmsmd ^ascription and comparison of tbe
estimated or forecasted risk, costst sod benefice of « policy, course of
action, or technological or siting alternative. It can include!
* A reviewer has noted that w toon m one specifies the use of conservative
assumptions, a confidence lladt m the range of uncertainty, or the use
of an upper bound estimate, one is introducing a valae judgment into the
analysis.
*+ tha "cooperative risk assessment" described io tha proposed Mlisk
Assessment Research m4 Oenkaastration Act" (U.S.. Congress, 1983) noted
Sumy of chase elements* A frill.- for A "Coaqwehensiv* environmental Risk
MAajftMMaB Act" «as flammed Cw I® Uu 1987-89 Coofreaa,
11-11
-------
PItufti tl-1 - frMBWit for AiMitlni and Hanjigttig Technological Riakfl
-------
Co«p«riaon of tkm reduction in risks to health and the
environment that auty b# achieved by one or more technology
control options
CeKjMrison of C.t»« risks. and coats po»ed by two or more
technologies thee $mt*m the aasie huuean. meetll, i.e., pro-
vidm aimilar identified benefit a*
Conperlaon of the risk*, cost?, and benefita of alterna-
tive f«chaalati«* or caohnology controls, noting both
equity of diatribution UMg stakeholders and net #oci«tal
effect* (reduction of rlik is also a benefit)
Comparison of n giwen. rf#k with. naturally occurring or
risks, enabling the decision Mkers and the
public to place tlui rtska, coici, end benefice of miter-
active techno Logical flMeai iota tetter perspective
Ideally, a comparative risk-cost-benefit assessment methodology
should be characterized byi
A tyataaiAtie *ppro»th that it sufficiently flexible and
nonmechaniacic chat it can be applied to a variety of
decision -queacione
A realIatic techno social «ad institutional yi«« of deci-
sion objiccivti, options, and inpLaatataJbiLicp
A comprehensive, long range vlev of risks, coats, and
benefits (
-------
Z. Ptcitlop awikict: iisfet decision staking involves deciding
wftather er floe to actually do swathing jJbout the »ourc« of risk that haa been
assessed. It raqoires a combination of substantially science-based esxiwates*
of healcJti and environmentsL risk, coses, and benefits (developed in preceding
seeps) with gil»r consideration* that affect ruh
-------
Ii»k av«r»iaai Cm milae-tp. tUernativ«s that avoid or
oisintt* tha risk* Achievement «f sero risk is often not
possible, but the concept of 4m ffljttimit or negligible risk
Is well-recognized by Ch« courtb and ia daily lite.
Social control¦! Governmental igtciclit, trade associ*-
ci.
Administrative control* I An organisation limit* che
secure or degree of eipOturt Co eisk tor tocM or ail
employee*, cuaconera, or visitors, e.g.» by employment
policies tlmt requite deaignated tevala of physical or
mental abilities; by reatricting access eo designated
areas
* practices: Employee* working at dangerous t®sk«
are required to follow * pacific procedures and
sequence* dtiipad to Minn tfae chances oi eeoi-
deot.
* •Forgicmal .protection: An individual or group u*«
dMisea utile being enpo*«4 to
hazards, a.gM dust mask; sn organization decree*
cartau) protective device* for employees or person*
is it* charge in heaerdoue situations, e.g., rubber
clothing, safety goggle*.
* Mwc4tio# mi> training: Person* likely to be placed
ac *i«ti or to be present when otbera say be at risk
are provide* epeciei training and education to aa tr-
iad «e ri*k m4 adverse consequence ««
n-o
-------
• Hedjcal surveillance. tr fcention* Individuals and corpore-
tions bawe long ibc«4 decisions on wbat levels of risk
should b€ avoided and what levels can be tceapcri, giv«n
that redaction of a risk Loipoaes a cost. Vhen a. risk
casnftt be reduced. eeer-effectively, a decision my be made
CO distribute It a/aoctg several partiee so that no one
bears an Intolerable coat. Such decisions art frequently
made In Che Booooiie context of establishing an insurance
program. Personal be-alch, «ccidactt fire, and liability
insurance are coomb nftthods of >p(«ading risk* it we 11 as
limicui§ ttoe hours or j*riM of uoeker exposure.
Corporetioo* ia particuler have faced increasingly dif-
ficult deeiaioos to the last decade as their liability for
ocoupacional and conauaiar health and ufeCy claims and
•ftvirouaencal effects has eecalated enormously, and th*
costs of insurance have reflected the dramatic rise in
health cere costs and; envirotsaencal m, mm other regulatory cMop
Establishing reportinjl requirements and schedules
Determining loopaction and monitoring needs
12*16
-------
Establishing effective focusti for receiving public caimcnt
«nd other rrochaoi sbs far negoc iac ion and dispute resolu-
tion
Ssti&lithiog sntorceaenc and penalty procedures
Twplsnenc ing periodic review of d«ci*iont and itntcgics
in light of new irtfarraat ion «nd progress CO date
The ri »k asBBssmenc/n*n*genaeni process has been described here in A
t«quenc.i*l order, buc in retlity. It u usually an iterative process with
nulcipU feedback loops. * Uncertainty «jlyts.i should b® a pmri ot each step
in ch« Jise$»n«nt: unccrtai&cies should be identified and qu«r>ti.t~ied through-
out. and aggregactd scroa• th« process -
A isn proscripciv* decision cric
-------
References to Chapter ft
Hail4tt J# C» III, and S. 1® Thostas. What Are We Doing When W* Thiok We Are
Doing Risk Analysis? pp. 65-74 in Assessment qf Risk fro« Low-Laval
Exposure to Radiation and .Chewicajst A Critical Ovarviev, A, D.
Woodhead, C. J. SheLlabarger, V. Pond, aod A. EoUaender, eds. Plenua
?r«u. thru York, KY. 1985.
BNL. tfofluhop on Probiets Areas Associated with Developing Carelnogan Culde-
lia««T Syifcal'
(Canards* Irookhavan Rational Laboratory. Upson, UY, iun« ifi4.
loyfcin, I# f# tisk Analysis foe Pfcysioei fata cards* 4 Toxic Cheaical Case
Study® pp. 215-240 In ai»k Analysta in tbe Cfiaaicai Industry* Ch«®is»i
Hanufaccurers Association. Washington, DC. September 1985®
CDH5. Tha California Sice Mitigation Decision Tree: A Basis- for Decision
Hawing* Toaic Substances Control Division, California Departaant of
Moaltiv Services* Sacrenento, CA. 1986.
C6Q. National Eavirooaeneal Policy Act Regulations: IncQBplete or UnavaiI-
able lafarmacion- Council on Cavirocosental Qoallty. Federal istuetr
51(80) 15118-15111. April 25, 1986.
Conrad, Jobet* "Society and tisk Assessment: An Atteapt eC Interpretation."
pp. 241-2Tl in Society, TechnoloeT ond Risk Aaseyineot. J. Conrad, ad.
Acadeaic Press, Inc. London. 1980,
Convey, R. A* p. 4 in Eg»iran—meal giak Analysis for Chemicals. 1. A.
Conway, ed. Van Noserand Reinhold Cosipeny. Weu York. r587T~~
ClS. A Review eg tisk AjiiKiasngisg Methodologies. Congrstssional Research
Service, Libxajry of Congress. Prepared by C. H. Marcus, mt alfor the
Subaosnittae on Science, Research, and Technology, U.S. House of Repre-
sentatives, 9ith Congress. U.S. CovtftMi&c Printing Office. Washington,
D.C. Kerch Ifll. 71 pp.
Cullingford, Michael# Frederick Miahas, and Sappo Vuori. "Use of &iak Analy-
¦ ii in Salaty Decisions." Presented for International AtoauLc Energy
Agency at the Annual Congrass of tha Federal Society of Radioprotection*
Avignon, France. October 19~22« 19*82.
Devias, J. C* "Science and Policy i«* Risk coecrol*" pp. I3~2i in ti»k Maaege-
wgat of Baietina Chcsucais- Chesdcal Manufacturers Association,
Washington, 0«C. June 1984*
DSKC, tisk AeeeMiwmc T5iriorrVAr' Jt*Iir
1963.
EPA. lt»fe Aeacssswac aatd.Kanat«*eqgi graswnwacfc for Decision Makici. Snvi-
roonental Protection-Agaacy. SPA MO/>-8*-002. Oeceabec 1984 ~ 33 pp.
rwe
-------
EPA, ClosMry of Extvironaantal Tar**. Office of Public Affiin. 0PA-87-017,
ExmroraantJl Proeaccioc Ag#ttey; ttartujjftem, DC. March 1943. 20 pp.
Cr®«r-Woocten, Bryn. "Cont«xt» Croctpt and Coasequence in Risk Aikiikmc
2tiwch.* pp. 61-101 In Soaiatr. Technology and Rjgk Ajttiwwt.
J, Coor*ip e4« Ae«4«aic Prtiii Inc.
Gasmant S., K. W, Molekg* ?~ Irwin. and C. Whitehead. Public Policy toy
Cb—ittali - Itotiawal hb< tccgrnatiotial I»ua». Th« CorrMarvaeion Found*-
cian. Maabicgcon, D.C. I98tt. 144 pp.
Hadlay, P. V.» SCOPE Seporc Bo. 8.
Jofa» Ml«f I mm, Tarfc. 1971.
MoghJLa»i, 4. Alan* "lijk Karti«f«*»eftc - Practice end Preap« 11-23. November lff4»
Morgan, M« C. "ftisk A»ae*««titE vn4 ftiak Haa«g«*tLOC 0«ci«ioo-*4kioi lor Cham-
ieal Expoture." pp. 107-143 JLo Eftvironanracal bpuwrt trow Chanlcalt,
Vol. II, «» 1* tf«el7( md C. E. lUu, ada. CMC PrtoA, Boca (Ucon, PL.
1943®
Muopover, J. Am Aoalyfti* of thai cl»_Bittiadi Strategy for ftiafc Kan^genent.
Biaj^ioAlzaia t(4) 437~4A6.' [feceaber !?&.
Nul t K. Am Baeard Aiiciibmc; Ctroeideracioni id Etc faslga and
Interpretation of Studiu. pp. ll»*J in Ri»k AaMKlya 1» lo elm Qm«£cal
loduacry. Chewical Kainifaocurcri At•ie&ciott» ^aSIStcottT" DC?
Sapeaibar If&S*
KRC/BA8. Peciaion Halung _for Outntoili la tb« Bnvironniaoe. Co*-
aietaa oo Prkflcipl«a of Q*et*ioa Kakinf for ft«fulacinf Chauaicals ia cba
Environaenc (J. C« til* ChaxnNui)» -lfacioa«l S*aAareb Cmmell,
national Acadaay of icitseei. 19'15 • 2J2 pp.
HKC/VA9. **tiak Asaeasttanc." pp. 45-98 in iagulating Paaticidaa. 'Comictec
on Protocypa Svplicit Aaalyiaui for Paacicidai (kobart Dor^man, Chaitrauun),
Hatioiul &M«areh Council. Baciosal Acadaay of flciancaa. Vaahingtoo,
D.C. WW®
El-ll
-------
KRC/MA3« litfe m4 Pfcitlgnwiltiwit Perception* m4 l>ii|A«
tove, M. ft* to 4ii.atfe5iT of &iik« John Kiley fc 8004• Hew York. 197?. 468
PP«
love, U, D., wich QtiWr*. gveluaciao Ifohode. fay Smriraoaectal Standard*.
CKC Prtii. Boce ftacoe, FLl 1^1. £|2 pp.
loyal Society. ftiek Amcwtoc* Study Croup 00 il«k (Frederick Werner,
Chairman), The Royei So
-------
Saicti, J» H. wRol« of Risk Aaaaastaaiic ac Ch«pical Kanufaccurara Aacoci&tioa."
pp. f-10 in 81 »lt Mmm-memtix of gxiatlng C^«mical«. Cham c* I Manufac-
corefs Association. VaafciBttoQ* O.C. J una 1984.
S*ith, K. I., I. A* Carp«ocar» *ad M, S. FiulsCieJt. ttjgk AjiowenC nf
Ha»*rdou« Chataical tystwn ia Developing covntrio. S*ac West Envircto-
Staich, V. K. Butfu AnaLyai a for N«tur*l Hazarda. Rnk An»ly
-------
-------
III. AmflACH TO COMFIMTX¥E MS* ASSEtSMQfT Of BMMP008
An Approach Cor and coopering the risks to human health
and tha anvironsMnc of fc*«**4oiM vast ¦ MiM|amnc alcermriv^i Is out lipid in
this chapter. Kecluxlologicai objectives art briefly oot«d, fallowed if a sum-
mary of the ntntkotljjLagi&ti La due ncuntju
4» Methodological Objectivwa
The isethodology should be be^etl 60 the rtjectivti and scope ®f the
scurfy stated in Chapter 1 and should. fit wichin the f raneuorlc lor assessing
4ixi iwtugioi risk,* out lined In Chapter II * A rtcioMle Un<>: Itattctout ctaaageseac deoisioa«
will typically involve s cniice aaoog alternatives, Mttuir chac a *yes" or
"no" choice. Therefore, the goei of Che vethodology it to provide
»fl that ch* decision naker can coopers the likely (and possible) outcomes of
alternative decisions* to objective it to incorporate technique* thac vill
yield scientifically defensible cowperttiyf risk assessments, and to avoid
mixing ia components aore properly reserved to the risk wanagegenc portion of
the overall regulatory process* Key points ia Che raclonale are J
* The aathodology should be eospreheaslve end aoasiscenc, yec
fleaible.
* The methodology should niainx# introduction of personal values
and identify unavoidable value-laden decisions and assuiaptions.
* The isethodology should field information on the moat likely
outcome Car each dpfcion, and also on the uncertainties in such
estimates for all options considered. Ml are needed in risk-
based decision leaking.$ (Note that the choice of options will
t'llflltct sQBeone' S valttaa*}
* Ccmprebensivenesa require* Chat the assessment of each option
consider ail kinds of releases* transport routes* ami health
and e&viroaeeotai effects (not jeit carciaogeiiidity, for axast-
fU> for the life cycle of the technology or beyond if there
are long care effects. The assessment of effects, particularly
secondary or higher order affeats, will be lieited by ciiae,
resources md data. (Vote thac the allocation of tine and
resources can reflect values.)
t The cost-benefit analyses in support of 4f«iaioo salting, should be siailerly
caaprehonsive, conai stone and eeylicit a boat uncertainties*
Itt-1
-------
Consistency retires chac in aaaiyeiag the rijlcj, co«p#r«bl#
contributing be tr««ce4 co«p«r#Mly for each option.
For exaaipLe, assumptions wst be consistent regarding quanti-
citf m4 compositions of wastes* jc«le of disposal cachoolo-
gies* renedial or corrective actions c«k«a» te*por*l up«ec«,
and do#« response relationships* ItocerteinciM «u«t alto be
consistently treatedt
The Methodology wwt be flexible because of the variability of
the haurdoiu waste probLeftf for some factors uxth upper confidence limit
estimates (e.g., WZ) for other factors*t The resulting
estiauitc of overall effect would be at seM ill-defined point
between an KLE and a worse case estimate. The analysis should,
however, address confidence limits in the uncertainty anslysie,
and both the best estimate and the umeerceinty range should he
provided to cha decision maker*
t The uncertainty for each factor can be stated to within ooe standard devia-
tion <± I e) of the base estimate or some higher coafldence interval suah
as 2 e or 3 o. The larger intervals efcar aggregation will tend to
increese the overlap of risk estimates for cha options being comparedT
and the choice of interval probably won't atake a big difference in most
dacisione between technologies.
?IM
-------
• of uncertainties fltuic consider both systematic
-------
develop rspresentativc sices? the greater the nunber and
dive*»itf of lien, the greater the effort required.
Taaporei consideration* - (Java the ciae of waste
disposalf exposure isiei^vac and health iupsets been defined,
o* don the project tea* oeerf cp uioM reasonsble periods!
The tfxtcr the cine periods la general* the greater the
uncertainties.
A siarpler nethodology, whil« • highly desirable gs>sl, #pp@*rs aecas-
co require more restrictive bound las ^od 4 Urt«t number of assumptions
to be built in.
B. Methodological frn>em>rti
The methodology outlined below is described in |<«tc«r detail la
Chapter If wad is based on the literature revl«*ta «rtd diacussions io Chatr-
ten V through X. Ic is intended Co be as i^neric as practicable, but co
pcrwit modification* in accord with Uw nature of particular i»us}iwo:
probLesui and with the quality of the 4*te aval labia. Important feature*
aret
~ It provides a module* {rtfemrk for a saris a of steps, most of
which must be performed in. given a«suiMnii*
• It can assess the consequences of disposing of specific wastes
at specific sites with specific technologies, through use of
Actu»iit case data ct carefully constructed scenarios.
• Ic contains procedures (Tor selecting sooni sveitebLe analytical
technique* for each step* so that reaxinan use can be made ot the
data available.
• It Mulywts OQctruittCies in all steps and iitrtgtMs then
across ail.steps in. each assessment.
The general framework for the health and enviromnantsl Ttsk as-
ses soent consists bvQ*dlf of the n«en steps ss fallow**
« Source Assessment (Resard Characterisation)
• Kavirrmmentel Transport and Fete Analysis
• Kzpdnrt Predict ion
Health mod Bnviroewasitai BLffaots Analysis
• Adverse larpact Estimation and Suaaeation
• Hacefcainej Analysis
• Upotc and Caspar* ths Results as Appropriate
Eaeh of these steps contains oore than one element, however, so that the over-
all scape of the risk assessment can con tain suy eloaaeats, dapendini on the
decision pnblas. Typical activities under each step are as follows!
OI-4
-------
Source thaaard
-------
i
te I ifc m
sli{
•«* a#- • II
si 11 &
Sfil f
18 alls
-&* 8 £ t
Ui'.ii
'"III
- " r • m **
*s?itf
• - ¦<
|?ffs|s
2 A i»» _
iil:!
" 5.» J
Iss|3f
Jiff! S
I
m*
I
5s-
s If
?|E
rr **
0 n o
: 8 i
iff
< ¦ ¦
1 ¦ •
r*
a
s
r«
5
o
f
? ?
<
s t
li! I
i
m
»
2. S
5
6
!•
»?
Is
£ 3.
s •
i:
*
»«•
:i»«
St
I f
ii
i I
m m
m n
is
ft
H
o
s
n
B
on
u
0
1
V
s
H
m
!
ft
»
r*
a
2
m
£
<
i
**
• •
&
§£*
3
8
s
•«
r®
i
f r
ft
H* **
* S
I 4
I
I 1
s. &
% $
m
1
p-»
3
i
S
r» #«i
St
2~
e*
f.
o
f*
o
m
»
n
s
s*
fi
I
*! i
si a
fi "
3.o
• ¦
• %
m *¦
-I
!
1
ft
•
I
r
m
rt
®-s
1
»
B
r
•
M
n
m*
«
r
?
rf
i
m*
8
fr
2
2
5
o
~41
ft
ff
n
1
•
•
4^
»
tk
w
m
rt
I
«%
o
•»
m*
1
o
•
s
tr
•
§
1
r>
£
•
t
m
A
O
•
a
1
t
5*
9
m
fr
s
rt
a
•
tr
a
n
s
rt
t
m
a
e-f
il
{s
»-> *»
i
&*8
fr
I«i
yrl
tgi
lis
p*
I
i
a
t
i
f
£
2
if
CS
II
3 "
¦
l
I l«
!
8
El'.
jg ft •* a r-
hfi it
** 9 SL ^
• 3 2, #» -
iS% n* ^ ma m
jtpr
ft M ^
«8S
«- s*
IS
IgS
MS"
m* •
m r>
§9 •
ll
It
ft
t- 5
m if
^ %«#
-------
GbaMtttftrfea
MasordetM Watia
wid Treatment/
Dapeitl
fodhnologiet
Phytic*-
Qmieal
P«op«rtiM
Doto 8m
Identify and
Chorocterize
PoUvtonh
Xeleated
Health
£ffeo*i
Data tew
Ptopwleflon localiad Daift Bm#
Analyze
Erwlro(«ntftiol
Transport and
Fate
Predict
Exposure*
(nviforiMiental
Medio Doto
Sam
Expenm
Monitoring
Data Ism
Uncertainty Analysis ol Each Element
Population
Derutfy and
S«mlllv«
SubpopuloHon
Oota kn«i
Apply Date/
ReipoOM
FuncMoni lo
ExpOSCd
Individual!
Eitimale
Heolth EffecH
lor Expoted
Fopulaiiortt
Oevetop
Predictive
Ifcnon Do««/
Function*
/ Report l«»i ^
Efttimahlt atrtd
Uncertainty j
; femget /
«••£! s#
Imtui Mil,
Figure ni-1 - Eltvenca io Health Rick Asmimim of tUsexdou* Waace Disposal Methode
-------
carefully bounded, impropriate d«ca must be cdmpL L«d, AMiyiical techniques
nuit b« >sL#cc«d »nd ciicul-ttigns Biader, And conclusions drawn in a scien-
tifically manner. Chapter IV develops ch«s« seeps «nd elements
further.
111-3
-------
if. am .imciATED mam/mm n« awiwt wgnioootfleif
A eecbodology for assessing mi eoap«riB| ch« risks to hiaiti health
a*d (hi Mvirooaent of tesartets wtiu «ni|n provided
la Kb«n thiftm. The aathodology should be •pplied coeaiatently to each of
the so that their ritbs can be compared realistically.
Mil' a tuggdsted fr*aw«rii for health and environmental risk*
Mitca of h#»«e<«iu u»u« *»o*f«*«oc •it#r»icl%«i ceoilid of seven *«Jor
atepat (I) iai«n pollutant sources and ril««a«it (2) aiciawce transport end
fata of pollueanta in the environment 3 (J) aasaas espoauraa -of humane, other
special, Or other things to the pollutants! (4) review tod evaluate the
literature base for the pollutants and select or develop predictive nodels for
adverse Uepaatsl IS) eatiswta actual ijapacts on individuals and population* of
projected espesercs to these substances! (t) analyse uncertainties for each
step and aggregate tbe uncertainty across all stepsI and (II jania and
report tbe results.
A. Itefca Isififlgegit lls;a.ag#ea»s Cbarecteriaatidg)
H» pollutants oatteriflg dm mrttummt ff«» th* source mmt first
be determined. tci »oe* a»j« naawn, the secure •! tbe pollutants, tboir
-------
TABU IV-1
tififfjafT tapsiaetiificitts m mmmm wste pistocal sombce Assessigirr
- * ' IKS!5«
•I Coacera
OuriCUtllr Be leases
Ceacr* tl— fcctlvltUi
Pr«4B€tiM and dis-
tribution processes
Product use patUrsi
Hazardous vai(> TSDfs
Locations
QuistiUci
Cnfosltioas
frsftrtin
lglease WatffeafiJ
Eainioai
Vapors
Particulates
Effluents
Solution*
Suspensions
Land liip#i»l
Acc identic
Lrals
Spilla
Ej^lovleaa
Firr*
Floods
Viwfa
Been!v ami oceans
OrouaAntor mmmm
Unnaturatcd
Saturated
land
Surfict
Subsurface
Deep strata
lUaatJ Dtiiw*Uc«
Oinutili
Quantities aval I*
able
Propertied
Physical
CbenlCe!
Ta^wral C«n»IJur-
ation*
Hf)c*«c
Quant I Ural ion
CheMicaU
Kale*
Quant 111
Physico-ctakUal
fona
-------
ft.
f
i
ffgwrt IV-1 - Schtaa for A#««»®ing tUurdaua W«st«» ti S#urc«« of Rnvir©qn»nt«l Pollution
-------
• Characterise hazardous substance aoure* of concern, lac lode <>
necessary waste generation activities (e.g., production and
distribution proensttt product use patterns) and hazardous
waste treatment storage and disposal facilities (TSDFs).
Characterise location, generation rate, quantity, composition,
and properties of waste to be assessed. Composition includes
qualitative and quantitative information on the specific chem-
icals present.
• Characterise release of contaaiunti to ch« environment Eroi
the source* Identify release mechanists*, quantities, and the
receiving media.
• Identify the chemicals likely to be released based on prelim-
inary knowledge of their physical and chemical properties and
their concentrations In thet uaste.
Hake a preliminary review of the toxicological and ecological
properties of those chemicals likely to be released la
significant amounts.
" Designate for detailed assessment those chemical* vhith ere
present to such quantities end have such properties that their
release Day cauae li^itificiAC adverse isfxcci. Pr«lLainery
coneiderations of rMeiviag Mdi* «od racapcor populations vill
probably be raquirrt htvt. Thai# designations should be re-
evaluated during the inpact analysis.
Quantify the source; i.e., estimate the rate, concentration*
quantity, and physical cJsamical form of eacn hazardous con-
stituent released to each receiving medium.
In practice, the source assessment should use the best available
information and data in constructlog scenarios and making calculations.
Search and retrieval of Inform* tide* for source and waste characterization may
require extensive literature search and personal contacts. Sources of data,
methods of estimation and reasons for aasuoptions should be carefully docu-
mented. Information should also be referenced pertaining to the uncertainty
range about best estimates for each value.
teXeeae quantification may well be the moMt difficult part of sourte
asseesaaac anlee* atoplifyimj assumptions are eiada. Balaasa may be site as
•all as facility specific. Particularly difficult nay be releases during
transport or at storage and transfer points.
Considerable data arc now available for discharges and emission?
from maoy typical iaduacriai processes and TSDfs, bat the dace may not be
definitive la a given case, and very little data may ba available in others.
For example, amissions to air have been osaasurad for many incineration facil-
ities, but much uncertainty exists over the level of chlaradioxin release
under many operating condition*. Similarly, fugitive particulate aoisgioa* to
XV-4
-------
air fr* roadway* tu«d by vehicla* hn; The petsiblllties that corrective actions
will be taken when reLease* to che environment era discovered poi« difficult
methodological probIdea in assess lot risks, coici, and cht benefit impact of
TfOTt. Problees of a different nature are posed by the poadibllity of opera-
cloa odder upset eooditiona or of catastrophic failures. All of chose
preblees require scenario approaches aa discussed below*
a. Corrective actioa fefeaejoil Under ECtA regulations AO
Cfl Part 244*100, corrective action is required when ami coring wells are
found to caetata hatardoaa constituents thac exceed tbeir respective co«ce«t-
tret lea Liddts. Corrective actions can rente from groundwater pa»pi«ig to
retrieval e®d redisposel of the waste. Including future corrective action* im
the ft A assessment and cost analysis peaes methodological difficvltiea ia
«n§i»4«t a eMHCAci«« eaeeossmat of alternative diapteei techmelogiee and
a^metei. These dlfficalclea eri>e fro* two areas> temporel factors and
tlue efficiency of the given
-------
Tttc netbodoiogical problem La comparative risk
is that the risks to beelch ud cbe tost to the ofdirncoc may vary greatly
depending on whether or not a leak is detected and on the type of corrective
action cekea. Furthermore, leaks (or accidents) have a probability distribu-
tion over time® and the few leek* occurring airly oty hav« uny different
(probably greater) impact* than chase occurring later. The detection of «
liner leak during tha w«*te deposition period, for example, my dictate chac
the landfill be excavated* and repaira made Co the liner and its base.** The
possible health effects and cost of sucb remedial actiona could v«ry widely
vich che scenario choaan and could conceivably b« substantial. la general,
hovevmr, change* in health impacts Cor corr*ociv« actioo*. for leaking land-
fills or inrpoundmcats during a 20-year operating period can be reasonably
aasumed to be negligible* tfee probability of leaking during the operating
period is smell* and is accounted For in the Pop*-Reid model used to estimate
release of pollutant* from the landfill. Tha medal assumes a degree of pre-
closure repairs of detected leeks which reduces sonevhat the volume of
leachete from the landfill that would ocbervise occur from early liner
failures* If eoata are being aoalyxed, the assesaments should state
explicitly if preclosura repair ooata are included or omitted.
The detection of a liner leak during the post-closure
monitoring period alto requires tone type of corrective action. An assusrptioo
can be reaaonably be mada that only tha Riiniaun corrective action will be
taken to aatiafy the hazardous uaate dispoaal regulations (AO Cft 2M). In
a one caaea, groundwater pumping can be assumed to be the only corrective
action taken, In others, excavation and. repair nay need to be aatuned. In
ftlli otters» consequence sucigACioa netsurca* lucJh is treatment of vattr
supplies drawn from cOftceiimct4 aquiJfera, may need to be considered.*** In
all cases, the health effects aad cost enelytes should state explicitly what
has been included and what haa been omitted.
Liner leaks chat have not been detected by the end of the
pose-closure care period or that develop during the poet-closure care period
poaa a further methodological problem. In aoma atudies, one slight want to
asauase that no corrective action was taketn became there was little probabil-
ity of discovering a leak after SO years (monitoring by the operator ceases
10 years after closure). In reality, of course, the discovery of a major leak
in cha post~cloaure care years that poaed aeriou* huyman health ha sards to the
public veuid almost cmrtaioly receive son* kind of corrective action* probably
under the Superfvod legislation, i.e., CtKCLA of 1980 and SABA of 1986.$
Aiaia* ell assumption* should be explicitly stated.
* The appearance of contaminated laachace ia tha leachate collection ayateei
may be aufficient reason co dictate action beyond stuple pumping.
** Id practice, the facility my be clos.ed by the operator or govertmamt and
tha waates relocated.
***In reality additional treatment would probably be perforated If deeaed
advisable.
X If one did not assume audi subsequent corrective action* m paradox would
exist: a poorly designed landfill that failed quickly but was detected
and repaired would be safer than a better designed one chat failed later.
IV-6
-------
(2) tfflgf—Cf a( correctiyeactiDat Tbe tEfieicaey of
earroctlva action ^w
-------
In die case of luMrdQW waste disposal, nlcixi, exposures,
and unpacts will be dependent an the nature of the waste in a given assess-
ment, on Che site of catastrophic rele«a«, and on «¦ sunup c iotas, concerning the
tiaing and efficiency of protective and corrective actions. In some cases,
tie nature of the waste and tie disposal technologies aaghc make aegligibLe
Che probability of catastrophic releases Cro® many ar even most csuisi~ For
example, a nenflansable sludge of low volatility, low solubility in water, and
high viscosity could be cleaned up with ainiouni risk in event of a large spill
frws an overturned truck, In view of the usual tim and resource Iiaication*,
efforts Co assess such risks could be reuoi)«bly ainiaized in the health and
cost uiatanenc.
On Che other hand, attention nay he required for the risks of
flooding during cleanup of an old haeardeue waste disposal site or for the
risks of loading/unloading activities in transporting wastes for at-se« incin-
eration. Ac a nininua, transportation risks should be discussed qualita-
tively! subsequent quantitative analysis aey be desirable before ultimate
decisions arc reached* The qualitative analysis should note the worst reason-
able case scenario, e.g., a naxinua release at the sioet populous or most
difficult CO ciea*~up point in the route. In general, the Bore decentralised
the hazardous wastes, the lower the risks of catastrophic release, the taore
centralized the waste disposal technology, the greater the potential for
catastrophe.
B. Environmental Transport and rate Analysis
The aovcoeac and interactions of pollutants entering the environment
axe effected by phy sveoehesaical and biological processes which can vary with
the nacarae of the pollutants, how duty are released, and Che stadia they
enter. These conditions Mil already have been defined in the source assess-
sient. The eavironaentai processes will also depend on characteristics of
spacific sites, starting with the area iavediately around die source and
extending to areas wfeare population* tsay be exposed. Hence the next element
in assessing tha risks of hazardous waste oanegenent alternatives is dwrsc
terization of the sites under study. Depending on the waste and site charac-
terization*, appropriate transport model i are then selected to estinate Che
aovwnant ef pollutanca to population** Sice characterization* and node I
•election are discuaaad below.
I. Site characterisation: Tha general data requirtaanes for
aodeling transport a«>d face of clb—icala released inco tha environment ab4
sources for finding such information are discussed in Sactioo VI.4. taquire-
OMnts could include pfeyeieal characteristics of tha T3DF and surrounding
terrain, physico^heaieal properties of soils, vatmr and air, and biological
characteristics of Che area. The checklists In Table If-2 of processes
affecting transport and fate of environaeotal contaaioants suggest the range
of possible requirement*.
tf-«
-------
TitU IV-2
mocissq AFFCCTme oviRomnttAL
Mttritcacftaaical jpfgc««»#<
Tr»it«r
Oi»per»iee
Dilution
Adaorpeioa
Acciwtioa
SeabllLMtion
LaaobiLiiAC lea
ieaCtlcMI
iytrolyaia
OtUiCUB
ftaducciaa
ffcetoLjraia
Ot§eiiAitI«m
Ka4i*aCtl*e iMf
lAteratfia
VoUcilis«CiOQ
Dvttlaf
Crttioft
Sedimentation
Frteipiteclcra
feUclon
Aep*ci«i
Ni«roor|*iu>mi
• Other blote
KtctbellfA
Activation
Ceevartioa
Dtfr«Utlo«
Dtceap»»Ui*«
Eacretlaa
Tnic Sffecta
liaaecuMlaeiM
tie««»l fleet lea I*
IwMt eteiss
bttrli otic t lattiMdia
trwihri
for bom tfec i a los-mfcinc fur^Kf, M«»r«l U. Bate that bioavailability of « ebaelcal cm change with
rlM Mil teftditloe*. If due palletaBC It paraiateftt, eoMidara*
tiea of aatabelleifti artMiM* aay ba uaaecaaeary (alcheaih biotraajforeacioea
aay require MalytU).
XV-f
-------
TeaporeI •sfuwptiois should ba implicitly seated. That is, the
analyst should sey if the characteristics of the ilci ir« assumed to stay cha
save or co change over Che time honson of the assessment.
2m Model selection: The selection of mo4el» for predicting
environmental transport end («ce will 4«p«o4 not only on the ha Cure of the
pollutABC and the receiving »edie, as oote4 above, but also on Che quality end
quantity of the data base available. To be as generieally useful as possible
a ultctid aodeL ibould be able co ucvliie data likely to be available la nait
isiiimwit probLews and yield quaneitative calculations of Che concentrations
of a contaminant as it nvci through a particular nediuxa of Lnuriit. Itt
addition a selected modal should be «ell-validated, if possible, in the con-
teat of a number of representative east studies.
Available oodels for the nain environmental conpartisents ire dis-
cussed and feonpe'ftd ia Chapter ¥1# Many of these models were developed for
special problems in s single suedium while others have a broad* nmltiaedia
scopa. Their eapabilitres and limitations vary substantially. The choice oC
sndals in j given analysis depends not only on the scope of the problems and
che level of detail and certainty desired in the results, hue also on the
data, time, and resources available fat" the job. A schema for helping select
appropriate Bedels for envirotuaental transport analysis in exposure issestoeot
of hazardous uasta TSOfs is ftiven in figure lf-2. Preferred aodels far
groundwater, surface caters, and air dispersion are sudssarised belov.
(a) Cj^juidvacer models? Prediction of transport: of pollutants
m subsurface waters ihouLd addrasa iaquMtiilly aoveaeeci in ch« upper,
unsacarated or partially saturated vadosa cone and la cha deeper, saturated
sones or aquifer*. fath tones can be addressed by either analytical or
numerical swscheciacical models (see Chapter VI)| the choice in a given case de-
pending on the daca avalLable and outputs desired. Banking models can also be
of use in special cases as discussed below.
CD Preferred unsaturated gone models: The rccocncnded
models for estimating contaainant residence time and movement in the un-
saturated sone are analytical matbeastisal tmdels such as PKS1A* or SESOIL
Csee Section VT.B.l.c). A nachsoBaclcaL hydraulic expression such as that
developed by Kctfhorter and Nelson can b* useful in situations in vhich the
¦oil *• sandy-gravelly, and groundwater levels vidua a few meter* of cha
ittffics. Mathematical transport nodeIs asy be inapplicable te some cos*»ina-
eiona of pollitusti and unsaturated a one soils, e.g*, lar|i releases of sane
highly chlorinated solvents into clay aetata can cause fraeturing and rapid
transport by ahannallag, ttt cases shsis nodals or aquae ions cannoc be med,
estimates of race of transport and residence tine have co ba made. Por exam-
ple# if extensive channeling Is predicted« transit tine in the unsacarated
aone night be assumed to be negligible*
(Z) Pfegarrtd saturated lone swdalst The eonplexity re-
f«4r®4 Is tha exposure and risk tssassaeat scenarios influences the selection
of th* frwmdnatat eransport nodal* In p»rtlt an analytical nodal trill be
*Me to nodal a onntaatinaBt pi una in a homogeneous aquifer, but sill not be
IV-10
¦$
-------
(hwlwWB
t »¦!««*•
McMlfjl
Mki
ms»k
m
•. i *»—
Air
CotnpUx ItrralA
Flgur* tV-2 - SctvMi lor S«l®ction o4 Haiti* f6r An«)y»U of Traiuftpoet Fat«
of Pollutant* Entering th« Invlromnenc
-------
«ble to show the effeces of *d4iftg corrective actio*! recovery wells to the
$ten«rio. A numerical model will be able CO predict the effect of corrective
action wells on the contaminant plume boc requires significant increases in
input dttt, allocation of time and fi»ds» end, possibly, computer r*«otjrc«i.
Many munerical aodels Are able to handle heterogeneous aquifers under water
table, irciaiu, or leaky urtejiiw conditions.
the decisioq of whecher to use *o analyc Leal «o4«l or a
aunericai node I can depend on tha ^wj»ttcy »f available data and oo the
availability of personnel kaowl«4gMblt La cha u<« sf » particular numerical
model, It may also depend oo eh* need to use tha least time-consuming tech-
nique to provide good initial estimates of che dispersion, adveccion, mn4
adsorption characteristics of a specific disposal alee, to, analytical model
is highly desirable and useful for screening alternative waste disposal sites,
for esanple, and far detailed planning and design of field measurements and
iwbIcoring programs. Bowavar» it may act be as useful as a numerical nodal in
predicting specific conditions at points of e*po»ttre.
Models chat can be recoeeaended for use In a generic
methodology are:
- A cottbiaatiflfl of two analytical «o4«l»» PESTM lor
unsaturated flow and KUMf for saturated flow, can ba saperior for some expo*
sure assessments because of thair user friendliness and minimal data require-
b*oc$. The FDfWATCH/FEMtJASTE coupla appaars to ba another useful system.
- Tfci adalycieai aedal AT12JD mcci well because it can
provide oner-, two-, or ehrae-disMHisional modeling of a contaminant plume? data
needs are greater with the AT123D, but not unmanageable.
• The numerical Random Walk Solute Transport Kodel or
TBAMS aay be preferred for certain exposure i«3«s»mnts because it permit• a
determination of the affects on the QftJttamnant pluxwe of corrective action
veils, or a large-capacity «ete«-supply *ell-field near cha source sice. The
&*odoa Ualk Model requires « greater mount of groundwater data (e.g., well
capacities, well spacing*), hut most aquifer parasaeters are sisullar to those
required by- ATI23D. tloce chat la soon cases characterization of present or
llkaly future groundwater withdrawals frtn cha aquifer, recharge rates, and
water table levels or dapletions «ay be important to the Impact assessment.
(3) lank|ng-ty»a models* Although noc strictly ground-
water transport models, ranking-cype medals oan often provide tha bast avail-
able measure of potential groundwater problau in emergency situations.
Combination of contamination ranking nodala and tha groundvatar-related por-
tions of hasardoua waste sice ranking models can provide a method for anaLyi-
iag an immediate problaa. Potentially useful ranking models include! Che
current forme of tha LcGrand Motel) the MXTKC system; and the JIB model (see
Section Vt.8-2). Th«se models require ednioMil data Input, equipment, time,
and expense, and yield quas("quantitative sica evaluations.
iv-«
-------
Data aeeda far reflfcing-cype m4«Ii 4rt Itee ri|«rou> then
the corresponding inputs required far «kJmmcIdels* Tb« mduc itg» or MACa for toxicants.
c. Airditearsi 00 aftd deposition pn4»lit A mty hrj« rumber
of air todol* arc available* Models considered appropriate for a generic
aethodelegy *>aret (a) those approved by EPA oad included ti tbe Ue*rr* Het-
mtk for Applied Modaling of Mr folletcioai (UUUttf-d) itriut aad (b) tMm
AcaMpfcarlc TraMfftrt Model «§«w»iepiMl by Oak Kidge HatlOMi taA
eo«aiddro4 a ss6«»il ia tnii OiOPi GilteliM* Sortoa* on e»a
rmim ia VI.B, nKMMiM a»dala mrci
- IflMhaatriol Umrt* Cooplex (IBC) U oitlhtr Ita abort- tot* Or
lo«t*cav« ««r)iasa (ZSCR or I1CLT) waa rated oa« of (bo aoot oaafui oodeta in
cbo UVAKA? oorioa far area aOorco appticaticm inv9l«iflg htiirdaue woate TSDfa.
The 1ICIT bao cha Mpability to aianlatt tbe pi use depletion and
parelcuiaco depoaitioo proeaaaea likely to be esaentlal for modeling aadaaiona
frcMB abiny loaLg-tem hesardout watu eperacloMi. It if currently beiog uaed aa
cbe coeceacracioB aodeliaig coe^oftemt ii the Inhdldtioe Xnpeaure Mod>eling (Iftl
syetOB. The IBC8T ia pdreiculdrly woeful for abort teata (e.g., trial bumo
U wuit iatiMrtcU*) fe* fflMftli of thm effects of awlfunc*
uom and fyitaa upoeco*
W-IJ
-------
- Atmospheric Transport Model CAW) tea s«mil features that
nuke It potentially useful for dispefjiou node Ling of coxic pollutants. I c
bts analytical capabilities similar Co the fSC^T to simulate the plume d«pl#-
Cion uui pirclcaUce dspaiiCioD proe«asesf but wti judged mot# ultful ifl point
source applications* A nodi-fied of the ATM allied ATH80 it preferred.
Die AIM is *. aasffttteac of CEhS (sea eultiauKLia models, btlmi).
- CauAilao Plume OiaperaLoo Algorithm (VALLEY) of the l/NAMAP
iciriM it recoeDended (despite its limitations) far applications involving
costplax terrain at tha aourca. Improved aodels should be considered as they
baccuDs validated.
d. Xaltiaarfia ipodelst Multimedia oodels ar« in an early
stage development ¦ Tai«i anin Section VI.F, reconw ended mode Us
aret
- Graphic bpoiure Modality Syitaa (CEMS) is a cooprchtnti*«,
multimedia nodal §ystea developed by EPA'a Off ice of Toxic Subatineas. It
integrate* aeleetad tingle madia model*. Ze it particularly uiafal lor air-
related transport (it «i«i tha ATM lor the air dispart Ion component).
- Unified Transport Nodal (UTN) and it* organic chemical
version, I7TH-TOX, combines selected models in aeries to give multimedia
capability, particularly la the water—related cransporc. Data needs art
extensive.
e. Uftceygaiwtieet Prediction of «nvirct»*«otal transport and
fate of contaminants are subject Co uncertainties fro* savaral causes includ-
ing^ deficiencies In input detal selecCion of optimum model lor .§ §£v«m
aspect 1 inherent deficiencies In the oodels selected; and difficulties In
liokiog ta lac tad tiodali. Data 4«flci«eaie* ir« expected routinely. Of the
available models* vat idatlon of a one of tha air add water 44.spersi.an models
appears adequate* but validation efforts for watershed runoff ul groundwater
nodeI1 is just beginning for oonreactive pollutants; reactive pollutants are
nodeled less confidently. Data limitations can determine model -selection to a
significant extent. Uncertainties arising fron lickibg transport amdels h«w
not been studied extensively, but are ooc believed to be significantly larger
than uncertainties in the flodela thaueelvea.
C. Empoaitre ?rwiicCxoa
Exposure* suae be predicted Ear two toy human population categories!
(a) workers vbo are directly involved in activities involving hazardous
wastes; and (b) members of the public who nay esse into contact with contam-
inants niaiifedl ftw hacardoua waste TSOfs #p4 CT
-------
Qualitative idtntificat ion if population groups litely co be
significantly exposed* *
* Quantitative e»ctn*cian of the nuaber of individuals id each
group.
* Quantitative estimation. of eta mmptitude of exposure of each
group.
Available mechoda and resources for predicting populations exposed
»n4 «efnicc«le of exposure* are reviewed is Qui peer Til - ()uaz»ticative exposure
prediction requirea characterisat iocs of aeny modifying factor* &a suggeacad by
Che checklist in Tablt IV-3 for human exposures. The discussion bale*
addresses methods only for workers and the public, ami focuses oa exposures
that amy cause adverse heelth effects* Mary of the concepts art ilso
applicable, however, to prediction of exposures for ftOohiUBan receptors*
fMLE ft-3
CWStBgtATtOWa fl CTAUCTttlZIIC WHAM BXPOWKK TO CHO
-------
CipoMNf ere difficult to eaciaace 4«aacltativoly if ¦onitoriag
data «fl uOOveliable. The primary route Of OSpOture if M«t Of COO by Lfthala-
tito, but dbtorpBloo through tho »ma. as* (uM-cft-mich erutfirt »rt alt©
^o«iUl«. Cayoduro* will nry oicfc cIm ohaaiddl «ad phyaical ueurt of tlw
••eerieIs talB( fcoadlod, the CKtnotoiy used, and control Measures and
trdialag ^ro|r«M Lot pLut. Tk*y will elao wary wid the iadividust *» use #1
milabli control MMoftt ami prttcriM tract less !•.§.» uae of d«*c aaski),
la (imtiI, estLootion teebniquee ere nlati*«Lf erode for predicting worker
exposures iioumI tlDf>. Well-validated aathenatlcei M4ili are aot av«ilablt.
Secondary oapoauroe of worker's faoillei can be cooiidtred; tbay are
extreaely difficult to quantify c«ofidntttlj» hOMvir* «•!•!» very good aooi-
eoring dau are available.
Occupational exposures are a||ri|ii(d In subgroup* by type or source
of eapesurs (e.g« i chooicais, work station, route* ece*>» m4 itunii! by oatni~
cude Of exposure (iocenaity. (rmqwrnty, duration), aa appropriate for sub-
sequent uae io the health effects prediction. la isse comparative risk,
umisisti, qualitative eitimcti of wrkar exposures aight be quite »uffi«
cient as aa input to decisioo oaking, because public health eoacerni m»f
predflMMfe. la other cases, quantitative estiante* any be needed for both
ebo average <»**nr mtfmatmm aad (or the umIjbiIIj esposod worker. Special
note should bo aadto of any pelleted levels ibat are above regolorory acon-
dard* for dborB-toin or lonc-tera oxpoeoras.
I. taaoaod public; fcc
-------
A totalled population ptofilt is then developed, bai«d on the
feograpkical boundaries determined far che pollutants' dispart ion through air
and eater away from tha lifts of hazardous wait* disposal iccivicitt. Spe-
cific coordinates trie desirable in defining cha study area sectors, tinea
iivtnl daca bu«a, Including that of the U.S. Bureau of Census, can correlate
Chair daca by latitude and longitude. la the J.S. Census data* population
enumerations arc broken dove by urban and rural areas, i|e, race, 4bz, coemut-
in| patterns, and houeehold daca. Ml data c«H be related ca geographical
coordinates. Only
-------
UocerE4iories la identifying the general exposed population
will dipnd oo che quality and completeness of eh* data available, tec should
be saall co»f«rcd Co chose In other of the tltk assesenent. Uncertain"
ties will be greater in crying cq quaacify »upopulations lor deteiled exposure
cither because of difficulty La specifying the subpopulacioos or Lb
obtaining geographically coded information.
b. Exposure profile* An exposure profile omc b# developed
for each of the identified exposed pcpulxcioft groups. This prof lit should
identify three chaacteristici of the exposure*
• Route of exposureI ®|r» water, foods, toil( or other
routes
• linns over which exposure will occur! ci*e Co masec and
duration
• CoqeeQtretion cf coneaainant received: concentration by
route Mini variation ovar tune
Appropriate models or other cstieestion techniques are used. la many cai«t#
estimates <#ili be needed lor the average exposure* for residents La an area,
Cor the maximally exposed resident, a ad for eapeeially sensitive subpopuis-
tione as discussed in the next subsection.
c. Exposure integration! The exposure integration process
involve* «f|Te|«tim| exposure ce the «se-mt possible avtr fcueet, eoneantra-
tione, time, and population groups* The degree of aggregation chat is rea-
sonable ceo depend on several aspects of the contaminants. The toxicological
pro parries of the chereicals of concern will be particularly important, tec
physicel-chenicel properties and route of exposure my also require considera-
tion.
Aggregation across exposure routes and concentrations should
usually be straightforward unless aeny eubpepula.cions have been identified,
the simplest case would involve one popular ion group uniformly exposed to *
constant concentration of contaeineot, e.g. , in Che «ir breeched or dmuokieg
water ctmsuaed* In csiu where exposure Co the chemoal by different routes
is likely to h«ve Hilfer«®c effects, aggregation across routes should be
avoided* la cases where « vide range of exposure levels occurs that are
likely to have different effects, aggregation should be avoided*
Aggregation acroai populetloo groups eaa be more difficult* If
exposures rmry videiy across groups, the "(net exposed" groups pr individuals
(HEX*) should be identified. They generally should not be eggregated with all
other groups to £*loul*ca average or aexn exposure level, because eos-t of
the adverse health impact* could ha incurred by a smell masher of highly or
frequently exposed persons. Bounding beeween Mill# less-exposed populations(
and populations with de Minima exposures ie desirable.
Aggregation across tiew also requires care in that aone health
effeces are associated primarily srith short-term exposures at relatively high
rv-ia
-------
concentrations, while others are msioci«t«il with long-tens (ivto lifetime)
azpo&ures at lower levels. Population group* with substantially different
temporal aspect* of exposure (e.g., 9hart vs. lottf ternf intensitt«nt vs.
continuous) should be treated io the health, effect* estimation end
should not be aggregated La toe exposure prediction.
In Humify, the integrated exposure tabulae#! all
sifoificmt population group* according to the relevane environmental dot*
chat each is estimated co receive.
D- Wealth an4 Environieentai Effects,taalyses
The toxic and environasental pcop»rti«« of the pollutants of concern
will have been partially determined la 5t«p I, Sourca Assessment. Information
art these chemicals roust now be asseabLed, evaluated* end converted into forms
that Mill bm useful l«r predicting the health and efivirnaaencel effects of
estitteted exposures of future populations. Specific activities will include*
review the health effects literature for the specific chemical*l identify the
klfisda of responses Likely and of concent La the expected exposure rangef
develop dose-response ralationshi pa covering the range of exposures; end
develop risk factors for specific ercrirBmeccal doxea lor exposed individuals•
Figure IV-3 shows schematically tha relationship of these activities and out-
li.net the procedure (to be discussed below) Cor selecting the generally best
available data for developing the dose~response isodils. The several elements
of this schesva are discussed below La temi of assessing huaao health nMti«
but auLoy of the eleaencs would be auilar for assessing healch ietpeets on
other species. Chapter VIII provides an extended discussion (with references)
of the theoretical aspects of health effects prediction epproechei-
I. Ewaluaea available iiegcatures A thorough search for end
review of tha literature describing the biological effects of the chemical* of
concern ii made.
a. Search scope! Tha scope of the search should include all
chemicals likely to W present in relatively high concentration ac point of
exposure and other cJMMticals that have special toxic or environmental prop-
erties. The fulL reoge of health effects will be of interest in the assess-
ment. 4 given chemical of concern auiy produce a variety of effects depending
on exposure conditions. let aany cases, the researcher will be able to focus
on one or two of the ao»c significant effects, particularly any irreversible
ones Likely to occur from losr doaa, long-tern exposures through environmental
routes* If exposure* are apt to be widely variable ov«r elm, however,
effect! of acute exposures nay also require consideration and say even pre-
doaioete in the Mseessieot (e.g., if a spill of e concencraced toxicant slight
occur). In general, the most significant concerns will be carcinogenesis,
reproductive effecta (including sotagenic, teratogenic* and fertility
effects), and any effects associated vlth bioacouBulacieo. If Che literature
search does not develop adequate data 00 the chenicals of concern co apply
preferred estivation anithods, it mf be necessary to expand the search to
include selected sisdiar cheadcals s© that less preferred methoda my be
attempted.
W-lf
-------
Is IKfc>«
4JSST
ifca
h=.
10*1
€fe#8®®4
1
I
fgns;!#ft
IhI 4r~J
MtWhH
tMW! Ktefc#
I
T €¦*•««> imWM
¦miim4 ««niK«M umv
Flgur* IV-3 - Schema Cor Selection of Health Effects- Estimation Models Depending on Availability of Data
-------
b ~ Hfrattree uiftlt; literature search teehexques will
depend rn httw weLL-VuDCwn (err how obscure) the cheat! cals of concern *re. Many
coflvtercial compounds aad cam » thac are aa ob-
jective as possible. Data that are not fully satisfactory aay have to ba used
sometimes in the absence of adequate data, but deficiencies m cha data will
affect the uncertainty in the Hsk sicinacn and oust be noced. laportant
checkpoints of the acceptability guidelines for bwmm and aai«al data are
listed below.
A. Wttmm studies? The conclusion* from even good quality
hunan studies c«n be lasa certain that those fro* animal studies because of
the nqnber of potential confounding factors. Poiots to vaccb includes
• Tha atody (exposed) population should be carefully docu-
nantad as to source and characteristics.
• The control (ua«rposed) population in analytical epidemi-
ological studies or reference population in descriptive
apideadologscal and clinical case report studies should be
documented and dea^ribed.
• Tha; saasple »lti of tha scody and control populacloe*
should have been sufficient to meet statistical criteria.
fmmtL~Hf confounding factors such as age distribution,
sea, «oA lifestyle aootURg and use of con-
trailed aubataaces, If known) should be controlled in the
study design.
The a tody population should haw expoaed so as few
taaicanta as possible — preferably just the chamicaL of
concern —» and to a odoiaaMi of other variable itresaas.
If-JI
-------
• Exposures shoold have been quantified* Tha Exposure
iiKiiMoc should be documented* Preferably, data will be
available froa appropriate asapling and analytical (Moni-
tor log At tha point of exposure or in bodily fluid* or
tissues. The duration and tine sequences of e*po*ur*s
chould be noted.
• Source of rciponfte d^ri should he docuatncid. All cliff*
leal ii|pi should ba reported. The length a£ the follow-
up period should, be reported m that latent sffecti, If
any, can be properly evaluated.
• The response of tha control population should have been
iimiiMtlf Jtodiedf purticuliirij for efface* of chronic
exposures.
• Scaciitical analysis should kw been «adeT consistent
•Ith the study design. Causal Associations between expo-
sure and response should have be eo inferred only if
chance, biaa« and confounding factors have bean ruled out
as explanations.
One or no re dos«-re»p«oti*e relationships should b« evident
La che d«t«, and any effects of other variables should
taav« h«en reported.
i. Inlatl nniini Poiati on which to judge ch* qotlitr of
mimal studies Include:
• The number of animals of a given study shouLd be ade-
quate. In a chronic study* there should be 50 or more
aninaia per dose group for each sex. In a teratology
study* there should be 10 to 12 female rabbits or 20 or
taore fesale rodents per dose group. If one is particu-
larly interested in effects at low dose, studies using
larger numbers of animals are highly desirable, although
generally not available*
• There should be ic l*aac three dose Levels (depending on
the type of study) La addition to the controls. The
highaat dose level should produce soaia toxic syeptoms but
not cause sure than 10Z unscheduled dee tha. The lowest
dose, ideally, should produce essentially no toxic synrp-
c«s,
• There should be m untreated control group and its nuaiber
ahould be the same a* that of tha test group. This group
should have recaLived vehicle doses If one is ua«4 for coar-
pound a4aiaiaeration.
• the route of adnLniet ration ihouLd be the sane as the
expected exposure route for hwaans.
IV-22
-------
• there should not be an ww»iv« number (> 101) of "dii-
••¦ed" «Qin«l5 or eerly deaths in the control grou^,
e*p*ciallj if the lesioa is similar to or the saae 19 Che
laaioti observed in cite treated (roup, the response oC the
cr««c»ent group* should be dose dependent.
• All Asjar orga&s fhcuU be Listed ia the pitholajj cables «
This is 4 tieature of tiu» quality of the study. Failure tc
noci numerous srgui can be ao indication of poor necropsy
or tissue ptfaeaM&Djt techniques.
• My clinical aigne, such t< anorexia, Alopecia, or con-
junctivitis, 9 ho old be listed in tabular form. The seme
should be Cra« fdr clinical chemistry and hematology data
when tbe«e are pare of the needy*
• All studies ahould. have ivf^rtLB| date etutlysla data and
cheeieal Mthodalogy to ensure quality control. {These
are least critical for acute exposure ttudiea.)
• A doae-reepocae relationship should ba evident in the
dac*.
• Peer review eoitOManta, if Available, should be closely
examined for possible short etmingi in the study.
• A study1a swteriels and! methods section, th* jeldoa
explicit "untotflsrdl results" section*, end the Gonelueicns
section, should be eere£ully read to evaluate ic*
In practice, the-risk assessment ceo® will find coaicicy studies that satisfy
all of these criteria are available lor few charticala. Ic should take any
deficiencies into account in the uncertainty analysis.
2. Identify likely r« a port—if The literature review generally
will have revealed that the chearfcel* of concern cause one or note adverse
responses ia test otpaisu under one -Or
-------
exposure conditions# In general, data from on* or more chronic exposure
stadias arc highly desirable, and data fron human »tu4i«> would bm preferable
to data of comparable quaI icy fraa animal studies, aa was indicated io
Figure IV'rJ« If fata (ram appropriate human or enioel studies art mz
available, then oat ttiorti progreaaively t® less preferable suit hods that
ictvolve greater extrapolations or agsuatpcions. Crosschecks between on# or
mare of these alternative taethods should be made, where possible, co improve
cba prediction m£ the Vtinds of «f(«cc«.
Information m the dose~respoate patterns for potential affect* af a
chemical is comparad trieh predicted exposure patterns to identify those
effects likely co occur at estimated environmental exposures. Crsphlcal ploct
of che arifiaal data, if not already provided, will be helpful, particularly
Uinta several studies are available* One aay be able to wait effect* litely to
be found only jc very high acute tzpoaurtt if only tow level exposures irt
possible, but caution ntusc be exerted at chit point not to discard information
chat aould ba useful in a subset of the imauKQi (e.g., a catastrophic
release)' Studies shoving p^iiciw response Lev humani waul4 be giveo greater
than stn4j.es La lamer aanali; studies showing positive
response in short-teno microbioessay or rodant studies but nor in higher
frumasals are difficulc co generalixe since the tyualicy of eta specific data and
the toxicological principle* involved ara critical. If the chemical ha a
apaciai toxicicy (such aa carcinogenicity or reproductive effect*), low dose
exposures always will ba of concern.
Tha effaces can usually be identified with reasonable confidence if
toxicologic*! test daca in animal* or ocher organism are available for the
taet chemical or for cogsace chemicals (i.e., other efaeaicals with similar
chemical structure and physicochaodcal properties) f or i! toxicokaoecie
(pharaacoki.net ic} information 00 the checaicals' abaorbabiiLty, ¦ability, and
biotransformations within the human body or other appropriate species are
available.
3. Develop relationships'. Dose~response relation-
ships must be developed for tha Best significantkinds of effects identified
in the preceding section. All the available data should be considered in
developing the most reliable doae-reapoaae model in humana for each signif-
icant effect of each chemical being assessed under the exposure conditions
predicted* Ideally, a quantal dose*respoa*e relationship will be available
that can ba expressed graphically and mathematically.
The procedures used lo developing the doee-respoose relationship
eilL depend on Che nature of the available data, toxic effect, and the assump-
tions and extrapolations that auac be made ior predictive purposes. Tha
procedures io each case are outlined below*
a. P«tm preference» Ef adequate doee-response data ara
available frea «tuiiwef*',^^w*™^^e«urs to th* ektaiejii of eoaeera, thmy can
be used directly lo calculate the dose-Tesponee function using an appropriate
mathematical model. The dose-xespoese fraction based 00 the Literature data
is then applied (after appropriate conversion of dosage units, if necessary)
rf-M
-------
to the estixiatcd mccinl exposure f ro« the hasardoue watte disposal activity to
give «b estimated risk factor lor m upond individual or population.
If data on tmMDi are not ava Liable, the inimt data arm Che
next choice. Any available toxicokinetic data chat jri available should be
considered* If chronic aniaal test data arc available and adequate* the jim
appropriate mathematical model is used as with human data far dase-coMose
interpolation aod extrapolation as above. II chronic animal data are limited,
but am indicative of an effect, a oimpter model still might be useful in
aacioiaclng che risk approximately. If chronic studies ware made but no
adverse efface* were observed, these data can be used to caLculace an upper
liait risk level.
If no chronic studies have been made, leas accurate methods are
used. In soae cases oonchronic exposure data on animals are available that
may be useful In eatiaaeing effects of chronic exposures. Extrapolation of
the results frsa short-term studies to long-term exposures is a difficult step
because, among ochar problems, sove effects (c.|., carcinogenesis) are seen
only in longer term studies. Therefore, this type of extrepalatioo should be
done only if absolutely necessary (see Section Vlll.a.l).
If chronic test data an several cogmatea are available, quanti-
tative structure-activity relationship* (QSAfiL) may be u««d co estimate the
chronic toxicity of the chemical of coocira relative co che cognacs. If non-
chronic exposure date are available for the chemical and similar data plus
chronic data are avellable for one or a few prototypes, their relacive do»e-
reiponae function* eat be uaad to eitiuea chronic toxicity of the chemical.
If data on cognates are unavailable but nonchronic exposure data are available
for the chemical, chase are extrapolated approximately on cite basia of general
coxicologioai principles, if possible, to estimate chrtNiic toxicity. Finally,
ranking method* are uaad aa a last choice to coupe-re the health risks of
alternative decisions if no methods are available for generating numerical
estimates. Banking* are baaed on the use of in vitro case data on che
chemical or Lbs uaa of oonparametric methods (see SecCion VIII ,B.I.e.(6) ].
The uncertainty associated with estimates from these alterna-
tive methodologies will be greater than with the primary method, and perhaps
by two extra orders of magnitude or «mre in addition co other uncertainties*
b. Thrcahttld and nantMrmaimM effecta: A decision oust be
iu4i on mhethar function should be dev^lopnti with a
threshold dose or as a nomcbra ahold model. This decision generally cannot be
made strictly frem the obtarved doaaratponae data, but aiuat be made on
theoretical ground*.* While threshoLda were once believed inherent in bio-
logical response to disturblag stimuli, nonthreshold models of cancer gained
substantial acceptance over the past 40 years* Wore recently, nontbreshold
vodels have been proposed for levtrtl other effects involving aoleculsr
* See discussion ia Section Viri.A.3. In some instances, the analyst may
oish co asciamce risk* by both threshold and ooathraahold models for
eampafieoa.
IV-25
-------
biology and genetics, and even for some effect* previously presumed widely to-
have thresholds (e.g., brain dwugc from lead* teratogtaic effects). The
preaant methodology uiuiiei chat «any chemicals will enfaibit systemic effects
for which 4 thre-shald model is »pfcopti«C« end Chat ocher chemicals vill
exhibit carcinogenic or other effects foe which a ooothresboid model n»y be
daoanded.
(I) KonthrMhold model: for health effects thac «rc as-
suatad noc to have dot* thr• ai«l4a„ aeacSiskaId versions of the multistage «n4
Weibull nodal a are used. TIM EPA has adopted the "limsriMil" multistage to
eatimate upper confidence Unit* on ri«k And lower confidence Hair® oe vir-
tually lift doaea" for numerous carcinogen*; CPATi Carcinogen Aa*ei*ment Croup
has developed substantial documentation with thia method in support of figtt-
latory decision making. Hie CM method is therefore adopted her* to utinici
the rsodel-based upper confidence limit on risk m pert of the uncertainty
analysis, lest estimates of the risk «rc Chen also made using either or both
cha UeibuLl nad«l or the conventional multistage model. Item of these models
ia ' diacussed belov.
The Wmibull node I is usually an excellent choice for
making risk extrapolations to very low exposures, provided the available data
sec* are not deficient in daae groups La the juld and low dose range- far data
seta vi tb conventional spread and shape, the Weibuil general I y yieldsr a goad
lit to data pointi (avan if cba data h4va a threahold-like appearance)t a law
dose extrapolation that ii nearly Ikur; and risk estimates that are near the
average Cor those of several othar nodal a used (i.e., intermediate between
diogc of the wir-tut in4 miltiitts* ool»i on cte high rlik side and thorn of
the logit and prtbit to the low risk side)*
The biological rationale for the Veibull (once considered
a weakreaa) haa been akpanded co a point [see Section VIII.1,2.a.C3)] thai It
appears to be of comparable stature to chat of the multistage (it has fewer
introduced constraints than tha multistage and linearised wUistap*}. If
the data set lacks a dosa group in the low response range, the multistage
model may provide a better representation of the probable dose-response rela-
tionship and can be used. (Alternatively, both models can be saeployed and a
geometrical average taken for the risks eecUated at predicted exposures.)
The cimm-independent form** of the Weibull model ia
exfrteemd by the equa t ion:
* 1. C. Albert, Chairmen of CFA's Carcinogen Assessment Croup recently noted
that an overall faeliog that the biological foundation is flimsy
for CPA's current method of Low level risk estimation [see Section Vtlt.-
S.2.a.(S)).
** A general for* of the Wtibudt distribution is:
r(t,d) - l - «*<• ~ MmUt - »!k 1
whan t i« the tlmm after dosing starts, v is tumor growth cioe, aod k is
a minher of discrete changes leading to tuaon.
IV-26
-------
Mil
I - e
-Cm ~ •<"!
uhara P(d) it tfea probability ot ratfWM at 4lMt 4 aOd a( !¦ ¦ ace par««c«ri
to ba aatiaatad (|,i > 0). Mpiyi (a) Is datamiaad by tbt twdmrMzui iacl-
1, and
ceoeivi with « < 1» A ralaa ¦ < i mmj riflioc an abaanca tf iu!fici«ic low
da*a data polftti ir tba «isteoee of anuaual toaieokinatLc features of chi
aubacancat ceaparlaon with reauici frc« cfct «ult£*ta§a would h* d<5Lr«bl« la
CKll CIM.
II tit® backfrouad IncUanaa t« naftlibia (® z 0>» if <»«•
vistitt t« itprttt cha utrt ri»k h« - I -
tf-11
-------
•mkmm td ¦ expected uwber if hits st dosa, d. Hut aoOl e«ia tee used wick any
p*lltlv« rtiponit data tva if oedLy oa« useful data paiac is available,
because, lor pitrpoMi of rmfk ticiMtioa, cbe slope af ch* com caa be
iiiuiMl. Liaaar interpolation between the iiti point and the origin or
background gi«es essentially (Im iim slope; both ftmbAly svmicunit the
rials at »iry low dose. 4 ceabiaetion convex-linear interpolation (m Sue-
cioa can be used for a teat estiamte.
The Modeling procedure is generally iaca aod
«s«ws «r iiisrst should fc» available froaa co esciaaca ste;
yrababllity af ©eels tiMmrim affact of AaalcAl for popalacioas*
If aaaltipla 4au sacs ara a**ilabia aad are of coayarabla quality» tha proba*
bilitias «c a giv«A »l«»« eaft l« m*mt«i aeueaitrieally.
(2) Thrtsiiold siwtals If the desar-rasponse relacionship
4i baliavad co hava a cniresKoTd doea« a tbraahold vara ion of eiciiar tha
• Dm «hl-«(|uara statiacie i« 4iem«a«d altib UM eultistaga awnAai
I® taction
pmb
-------
Mt I toil I ®r aBltietage MtihMacical Model is iih4« 7be tleiball andel |»
|e*erelly mi eacelleet choice Nu«m of the United BuaNr of variable* and
generally good flexibility In lict mi. The equation i*s
• t - .-<• * »"" V"
wter« d^ |» the threshold doae to be provided eat e« I, And m arc variables
dttenained by r*frcoaio« analysis. The Multistage «od«l nay bo freferable If
tha date set !~ deficient ia the Mid end imm response range. b*»t *ey be uadd*
tirebls If the rarm rite* rapidly and than pleceae* (tee Section Vtlt.ft.2.*
¦.(1)1 for further dlacwssien). The threshold version of the Multistage Model
In
MO . I - .*'•» * ••'•"V * •»«-«,>' * •" * V^V"'
•here it it cha threshold dote. for cither node I, cha risk is zero for
txposures be lev the threshold dose* Above the threshold. eh* risk can be
decemtaed directly fro* « eurve Htced eo the eaperiateecel i*c* or fro® tbm
•odel.
He identifieatioo of eke threshold dose co be used is
Wfthilly • iffvbla. lh« mil4|U data ere oitm lor continooas or graded
respoeee*, end queatal dete caai be rfifficalt to interpolate becauae of Che
Mnil aayeptocic spievwili of tha daaot-eaepoMea euin to the has* or hack-
gwMatf line.
Acceptable deity intakes (ADXs) will have been established
for i«eny chenicals beaed on the application of en appropriate safety fatter to
a no-observed adverse effect level (HOAil5 or lowest observed adverse affeet
level (10ACL) in experimental data* The risk li essuaed to be *«re at the
ADI, ac lease for aUsoet everyone* The threshold doaa Itself it *ss«naed to be
well above the ADI, but below mf Wfl. The precise nuafeer ia usually rather
uneereaia, if a Mtfl it raperead. For predicted aepostrras chat happen
to be ie thm th**ebaid doee ftfiaa, tha estimated fit* ia highly aeaaitlve to
assail cMMfaa in daia aM to tint Maigaart thraahol* nunber. M the internet
of coMSiatancy 4m aasNMaiaf riaha ehan definitive data ere lachiog( tha
following tfe*a*-tia* peoeaae ia atatfaftted*
• If en A0X has beeo established, the riek is asauMed
to be eero st it and op to the t&resfteld defined at )
cisiiii ehe ADI. Ills valise of 5 ia net tes«4 cm mf
r«fMl«tlonf but eppeera to be a reeaoneble ohoiee*
heaed on ehe pveetice of dividioi a N0A£L by « aafety
faseor of 10 (or ax>re) ce obtain en API, • practice
lissaii m ei»?icai>l«igic«l alii puhlie health immem,
Xa eha lailihdly evaat that tlw MftviMtetal data show
nil affacc et 5 tiMM A01 leveL, a eMallar
•altl^la ahOMtd fc« Bils ia eiw
ini Ia cltii tisiissll. tm isiilii,I m4«I.
W-M
-------
• I! IB ADl ha* not been established, the threshold
doit is detenined !rdi the DOA£L(a) if one or sore
it reported* The chntilsoU can g«n*r*llj be taken
-------
¦quival«ftc. If total dosage 4Lff«r» because of suoh variations, the e€C«ct is
assumed to be coBparaole on 4 prorated (cijttft-weLghted average) basis* For
reversible acute effects, the peak exposure dot* Li of prinary slgdific&ftc*,
but tima-veighted averages sr« required for other doses. If warranted by
informtion on the chesicai of concern or aspoauri condition*, adjustments
should ta made. For exaople, Intermittent exposures my be nearly equivalent
Co continuous doses if cbe chemical is strongly retained or iif be nearly
without effect if it is rapidly excreted.
If the conditions of exposure differ significantly between
ceic end subject population* all available data should be considered io devel-
oping adjustment factors and in estimating the uncertainty of the result*.
Prorating such factors require special care when extrapolation chronic ten
data to short-tern (even single) exposures of the subject papulaeioo to
carcinogens. Special care is siwilerlj required when cztrtfolatins minimal
chronic Celt data to lifetime hiususn exposures, especially if the chemical
bioaccuff&laces, if the chesrleal Is a suspect carcinogen on other grounds, or
if the body develops a tolerance to it. The analysis tends to become chenical
specific.
(3) Animal to human extrapolation: the detault assump-
tion is that the effects of a given dose are the seme in huawn* as in test
anisials at equivalent doses (sec * units" below). If there it clear evidence
in a given case that this assumption is inaccurate, then an appropriate
adjustment factor is used in estixseEioB rish.
(4) fapoiuTB units: fh< hwiafl exposure levels reported
in epidemiological studies or test doseges in animal studies and the antici-
paced environmental exposures are converted to ftonaistent dose units for com-
parison. Appropriate units are: milligram per square meter (wg/«2) of body
surface area of the tpeoies per day exposed* for most toxic effects^ mgfm per
pregnancy** for teratogenic effaces; f»g/is* for effects of acuta inhalation
exposures; and concentration in carrier (e.g.« ng/«t) lor acute irritation
effftccs. If tbat teat data are reported in terns of dose per unit weight (leg),
the conversion factor uaed for interepecies is: a ¦ 0.106(kg)' '«
* These are tha units currently preferred by EPA, The i»ore widely—used
unite milligram per kilotraoi (m§/fcg) of body weight gi*e lower estuaetes
of absolute risk* In many assessments the costparacive risks will be
inaenaitive to the units used sUvee exposure to the same chenicalCa} will
occur under all waste n*anegef»enc alternatives*
** FetseLe test animals (mice and rats) say be dosed over eichar the entire
gestation period or oa ju$t 1 to 7 days of their sensitive perioda which
is about oott*third of che total gestation. Total dote per pregnancy is s
convenient test statistic that would, at worst, overestimate, by a factor
of ehrmt, the effect of human ecpoture to the lam dose (i.e., in a case
where the dose «u received entirely during the sensitive period}. It
would generally be aore aocuraee for low level environmental doses.
CT-31
-------
Standard weight aod factors Arc tuuMii lor All
iptcin. I*lanac« «n JO kg, cootunes 2 L of «aetc per day, «nd
breaches 20 ¦' of air per day. Factor* far other test apecias art available
(sea Teble V111-2).
(5) S»poi\ire to »i»tiirn: Die default *jsimtption is chat
for chemicals causing the saae effect^ Che doses ire sisply additive* For
chemicals having different effectsr the effest of each *>c its estimated
exposure* Ifr how®*#*", che oheaiaels are knovo to Interact (e.g., tynargia-
cically), than che aaciaaead effect «houl4 be adjusted cq reflect chia Lnfor-
eaticm. Alternatively, if one (or a fev) of the cheaical* La *ufficiencly
conic and ia such high concentration that it doniucii the health iffKti,
then th« ifficci of the less patenc chemical my be 4m miniwus and can be
ignored.
(6) Sensitivity differences; The default position is
thac explicit allowance ia oat aede for sensitivity differences between indi-
vidual a Id tha raEtftnci group and eh* exposed population. If tha available
data art for a general pepulstioo or a population e£ low susceptibility for
tha given affect healthy workers for respiratory dysfunction), then, the
risk estimate oust be increased by so*® factor appropriate to the data. IE
Che subject populetion il known to contain a significant percentage of sub-
populatiooa of especially su*p«ctible individuals> then this face should be
used in the risk assessment. Theae faneor* should be derived after detailed
considaracion of any special sttbrpopulecioa*. For exanple, in 1983 children
constituted 28.12 of the population aod the feirchrace mi 15.9 live birch* per
1*000 total population par year*
(7) Combining response data? la §o§» cases the response
data raported in original publications oan be conbined to improve the reli-
ability of tha huun rish estimations. The data« for exampler my show cutters
at sore than one site in the aniitals teaced, multiple teratogenic effaces, or
different neurotoxic synptoes in differenc persona exposed. Since ehe ulti-
mate objective is to estimate tha number of persons affected by environmental
exposures, rather then the subcategories of elface, the varieties of an effect
ara combined. The data are summed aasumini independence, but using the fol-
lowing equation co elLaiMte errors of multiple counting of the saate indi-
viduals:
a a
PfBAjJ - X PU.) - H PU.)» *
- czuc P(Ai)P(Aj)P
-------
If differing estLcuaces are available Iron ituLcipL* studies
and noee can be rejected on th« basis «f quality, prob«biIici«i art calculated
from each data set and then averaged.* If the various effects arc frow dif-
fiftflt poo I a of toiiali «i in teratology studies, the sanple sixe will yrj
*nd the results are not readily combined. These results ire therefore cal-
culated separately and the probebilitLej of an adverse effect are chaa smini
to give a total rial.® This practice of combining data Iw* been discussed as
an approach in obtaining total tftiiaccs of carcinogenic risk, in iciitli with
one or isor* twaor sites {C?Ar 1984). flhen both sexes of « species reeeive the
lam dote and express Che saae effects» data frd« Che two ae*e« are combined
for further calculations.
£. Healch and Environmental iap^ccs Eseieetioa end Iotagratioo
The ftimki of the exposure! prediction acap are no« combined with
the health effects nodal to eaci.ata.tai the adverse effaces on exposed indi-
viduals, and with information on exposed papulation groups tt> escLxiate the
total crass be r of cases of each hind of effect associated erith each vasts
¦uuvagenent alternative. The procedures are discussed In this section for
exposed individuals and populations. Analogous procedures eoald, in theory,
ba applied to estiaata ecological or other environmental effects, but are, in
practice, ouch no re difficult to ijeplessent quantitatively, and are discussed
only briefly Is this section.
1. gff
-------
appropriate, the sub (in appropriately converted units) is used For C. If
exposures Are or Iftfti tljaa lifetime end lifetime risk* art
desired, appropriate adjustments are locLudad itt the limt function.
Calcalacions art udt for representative individuals in etch signif-
icant exposure group identified. Typical representeeiwe* would be an avarage
exposed individual, the nusc exposed individual, and average «>»t azposad
child. Additional calculation® are fwwMe nut necessary for Individuals in
especially aansitive jub|roap( (see below). The reaelts of tha risk celcu.-
LtCiopi are tabulated and aoefciaed aa appropriate and reported in a iorm
useful for decision staking. For torn® decisions, the risks per Individual siay
be adequate input, while (or ochars, the risks to populations described belov
«iU be required-
It. Effects op populations! Cotspariaon of the total adverse health
«£fe«c» of hazardous uasta manageeanc alternatives requires applying tka
health risk factors aa determined in the preceding section* rq inf creation an
exposed populations. The praaenc methodology focuses on exposures to the
cheoicala of fioncm incurred prlearily by the public and secondarily by
workers. ta either case, the analyais ie siarplest for e homogeneous popula-
tion uniformly expoaad to a fixed level of a single chemical that produces one
health affect. The case becoeies more coatplex as eultiple subpopolaeions,
eultiple exposure conditions, tiulcipU eheiicalj« or imilclpLe health effects
muat be considered.
The analysis La further conplicaeed by the likelihood chat the
population will not raaain of constant sis*, composition Or location for tha
duration of the potential eapoaure period. Temporal assumptions are required
before proceeding vith the analysis. The siapie asauaption is often «ede that
the present population pactem will continue throughout the period being
asaessed. In cooperative assesaaent of haaardoua waste management alterna-
tives, however, special care ahould be takes to dec amine if this- is a reason-
able aasuspcion. For exapple* changing residential, industriel* recreational
or transportation patterns in an area oyer several decades could substantially
affect tha rialta of having a landfill, incinerator or storage facility nearby.
bEence, explicit atateaicat of future population essuapcions mist be made.
A statement af tha population risk, is totaetiaae derived from an
astiiietad individual risk, by staple scaling. For exaapla, the probability is
aultipliad by 10* to inrfiaete thai risk, jie* Billion exposed persona, or the
probability is mnitlpIltA by the total eaiposed population to estimate the
nuaber of cases. A sure systematic procedure takes into account tha varia-
tion* la exposure conditions and of spacer? ti bill tiea for exposed individuals,
aa described below. '
a €mmm3, populations The naaber of caaaa of a given
adverse health effedt incurred by tha •*po or its doaa equiva-
lent f ron the exposure analysis; a risk. factor k.(d) from the doee-reapoase
relationahip; and cha nuabar of peraona
-------
Mo. of «icr* coei - C.*&(d) . »N. ~ C««R(d}Y»lt- ~ ... C -4(d) ¦*
• 4 * * * 1 a n n
H
or e I *dj.
Pee • nonthreshold effect such ia cancar where low dose line*r-
ity I* Jtiu®i4t the calculation foe ¦ rciidsneial iraa can b# simplified by
using the alope, ks of the dose~rcspoase lia« at low doae and the population-
weighted avingt concentration, C« The risk equation for Eh* average exposed
resident be&omes:
Awrap probability of efface ¦ C*k
and that for the population, N, btcoin.'
« _
Nuuaber of extra cases * k t C-i
i
The potentially exposed populations can be dtcinuatd by
Identifying relevant geographical boundaries aad by using current detailed
population profiles, supplemented if aecexsary by extrapolated data- Several
sourcaa my te ua«4 fa* 4*»«?Lopa«ot; mi fjt» 4twic|«i §m4 th®
identification of subpopulatioiLs exposed under different «M4iciea< {a.g.,
different routea, levels, peaks, Mid continuity).
Result! of eh* riak calculations art tabulated and combined as
appropriate Co reflect acuity issues («.|., geographical distribution).
b. Special subnocolations: It the axposed population con-
taina subgroups that differ greatly {^characteristics, then theae also we
considered, because they Bay be no re susceptible to adverse effects, S«w
subgroups May to more ausceptible in general (e.g., because of age or special
health conditiona). Others aay be sore susceptible to specific kinds of toxi-
cants (e.g., teratogens for fcatales in early pregnancy) allergens for anfty
people), oiore susceptible because of exposures Co chemicals frota octet
sources, or likely Co have a wuch higher uptake at a given exposure (e.g.,
children exposed to contaminated full)# Gxtrexjely difficult to quantify are
those trfko amy be an re suserpclble because of dietary deficiencies or prefer-
ence a, or who are deficient ia production of entymes or horsiones that are
iaportant to biological rsfmif iHiciusalaims. A satspla worksheet for suevcyia^
the potential health effects on a ,
hospitals, schoola, ouning hocses, recreational areas, and private wella)
within the area is sufficient to identify aany specific Locations where the
general population densities can be modified for the aubpopulation» Health
data, aoise of which is regionally and locally specific, are available KJtrough
IV-JJ
-------
eke K&tional Cantar for Health S.taciicics, Center for Disease Control> and
state «jr«I county health iftnciit. Statistical data are aveilable for ouobera
of individuals in different age groups, for birth races* etc.
Determinetion of the exposure of the tpeelmi subpopulatiooa may
be considerably aiore iaprecite chain that of the generalLy e*fitta«r
-------
effect may be of gr««ceac concern because of Icj nature or because of the
level of exposure to the chemical causing it, and the analysis is simplified.
If the effects are similar io severity, calculate the «{f>cti at
each exposure and *u/a them, but II thmj are greatly dissimilar. aggregate but
report separately foe aubseqaeot comparison of decision option*, i.e»r one may
need to consider the treatment and disposal option# ca sain* extent; on an
efface-by-effect basis.
If nre Chan on« cheatiaei is involved but they do not interact,
several assesM&ent alternatives niat. If the chemicals,, their tfftcct ami
exposure conditions art sufficiently similar (c.gM chlorinated solvents)t one
can aisply sum chair ulcuUtad tipoiufti to estimate an overall health
effect. As above, orb chemical or one effect atay predoatlaate and the calcula-
tion can be simplified accordingly. If these factor* are not sufficiently
similar, thee the different major combinations muat be considered separately.
Accounting procedure* t*re require careful attention. Id addition, tins
analysis would be further complicated if any of the chaauLeals present do
interact, i.e., produce synergistic or antagonistic effects. these nust also
be taken In account. Adjuatmeat factors of several kinds could be juseIfied.
*. Environmental impacts cat;auicloot The dispoaal of hazardous
waste can affect the environment in many ways in addition to huaian health
risks. thua, environmental effects should be considered in a coaprehenaive
risk assessment. On the simplest level, the wrviroiteentel consequencea can be
addressed by identifying their prtaiDc* tad i&g a mini based on ttwir
potential to damage the environment. Quantification of impacts, however, c«a
b« awdh so re difficult to perform. A rudimentary exaiii nation of aavironnental
effects lt» conjunction with a health effecta aasaaament can be accomplished by
following an approach similar to thait discussed for human health effects:
(*> data gathering, (b) identification of routes and Levels of exposure, and
(c) assessment of effects en exposed populations.
Data gathered for the purpoaa of racming environmental tranapore
models generally will be adequate to examine environmental effects. Sita
characteristics such as topography, depth to groundwater, aoil type, vegeta~
cive cover, and distance to Che nearest body of water are exaaiplea of infor**
eation thee will have been coeipiled* Xdentification of roucea of expoeura
will have been one of the most important steps in the environmental
aaaasamene.
Using the information at hand, rating* that reflect the degree of
perceived haaard should be aaaigned tc exposure routea. Ratings definitions
can be fairly broadt but still provide an indication, of the level of
environmental risk poeed hy a diepae»l alternative for a particular waste.
Batings of low, axxlerately Low, moderately hight and hifh are convenient. A
low racing indJLcatae that either no or de adninit effects to the anviroraaantal
media are predicted* A anderataly lew rating Indictees that adverae effects
will probably be noticeable but not of autjor aignifieaaoe to either the iaa»e~
diate locale or the general environment. A moderately high rating indicates
chat readily apparent environmental effecta ere predicted that will have a
significant impact on the ioaaediate aita or transcend other enviroaaMntai
W-J7
-------
ie]Mnti. A high r«cin® reflects significant anvirotusental effects chat would
lead to permanent co the surrounding taviroammc I a high rating Lapiiat
that a ma jor environmental raason iiifti not to dispose of 4 wmfit® via this
management jeemirio» These tracings alio* comparison among *lc«rAativ« TSD
approaches for « given waste, but should not be used to compare different
waste jcrews.
f. Uncertainty Analyiii
The several Available approaches fa unceminty analysis are
revieved in Chapter *# They range fr«n largely qualitative discussion and
survey of expert judgment tJsrtMjjh varUtiA l«vtli of sens i tivity (parametric)
and statistic*! uiljies, and aggregation method*. The choice of appreath
depends on several coasiderec ion*t iaclodini the quality and quantity of"
inforwecicn and d«t* available co be Analyzed, the type and of risk*
being considered, the tiae and resources available to perform the analysis,
and the purpose the analysis oust nrve, to note but a few.
Th« number and rang* of variables involved in estimating the health
and envlronjaantal risks and the costs associated with haeardoua v««c« manage-
near alternatives, coupled «rith the limitations of data and. cine that usually
exist io reel world decisions, will generally preclude a truly rigorous
systematic approach. The present method, therefore, focuses on uncertainties
ifl human health impacts (which often parallel but are wore highly valued than
environmental lapects, end of greater magnitude than uncertainties in monetary
costs), and strives to give useful re*uit» within the existing liieitattons.
The approach is baaed on a aonbination of nethods end involves three
steps*
• Step 1 consists of a identification and qualitative discussion
of sources of uncertainty in the analysis at each scenario.
Best judgment estimates are then made of reasonable ranges of
possible valuas of individual factora and pcrameters#
« Step 2 consists of a sensitivity analysis of selected factor#
and variables to define further their potential impects-
• Step 1 consists of an aggregation of uncertainties; across a
given scenario by using a For* of the propagation (cascading)
of errors method*
These steps are deacribed further below*
I. Sources of uncertainty: TM# step of the uncertainty analysis
involves a systematic turvfy of each aceaaric and of the savaral parts of the
aaaociatad ri»n analysis. fadh of che bos«* in the overall health assessment
proeeas shown schematically in £tfura Xlt-I requires cpnaideracion. Ill input
d«ca* calculation* or models, and aasuaqpcioa* aade that could significantly
affect the flat! result should be identified.
XWi
-------
The pot€«ciiil fourcei of ttuetrtmioey can be tabulated under firt
major topic »cm) or factor®. These factors ptrallil tto ««««BtiaL tcipi in
the sequential process of estimating the overall risk (i.«., a unbar of ciici
of advir«« health affects) of the alternative hasardous wait* Mfugtuni
methods. These major uncertainty factors are*
1. Pollutant Release - An«it|i^iont «nd viriibUs lit tlut source
4S5(}iBcot relating to the quantity and rate e£ release of ha&erdoaa sub-
stance* to the environment from a hazardous «att# TSDF tic*.
2. Environmental Transport «»4 fiCt Analysis Hethoda - AsauKptiona
and variables relaxing to the environmental NdU4 catea, and directions of
transport away Crow ckc site'end nrodel validity*
3. Exposure Prediction - AasuaipCions and variables relating to the
point* in space and time that the chemicals of concern nay interact wick
receptor papulations and the resultant exposure levels*
fc. Health Effects Analyses Information and Methods - loEarrtacioa
and data from Che literature relating ta the health effects of c.be chemical*
of concern, and. to Methods and codeLs for converting this Information into
predictive dose—reaponae relationships.
S. Health Iepact Estimation and Integration - Assumption* and
variable! In estimating iepacts of exposure on the individual population* and
especially sanaieivtt subpopulations exposed, and in integrating the i*paeta
acroas all exposure levels, populations and subpepolations to yield an
estimate of the number of potential cases of each type of adverse health
effect.
Within each of the factors, numerous *ubfactors or variables exist
that vary aocng wastes. TSD technologies, and management scenarios. A check."
li»t of such subfactors is sbovn in Table tV-4* Relevant sources of uncer-
tainty are tabulated and evaluated.
The number of variables will usually be la.rge and atatistical data
Mill be marginal« Although ezcepciona may be possible, this usually precludes
a detailed estimation of the probability distribution curve for eadi variable.
Estimates baaed m overall Lnforaation and judgMrac are made, however, of cha
mast likely range in which the velum would fall, e.g., within an estimated
standard deviation (a) of the beat pvine estimate for a vtlue. (About 6t.3t
of the valuaa would fall withio 11 a of Che Bean -) The axtresie vaLuea would
likaly occur within two to three standard deviations <95 to 99X of the values)
from the best point eatiomte. These ranges are based on the aaameption of a
normal (Cauaaias or ball-shaped) distribution of values, a condition that will
probabLy not preaiseLy exist for many variables in Che present kind of analy~
sit* For the general dietributian case (with «f less than infinity)* the
extreme value occurrences era given by Ghebychev1a inequality as follovs:
±3 a contains BS.9S of the values and ±4 or contain* 942.
2. Sensitivity analysis »f km w«rl*ble<« Sensitivity or para-
metric analyses am a*st useful ufeoft a ejaculation, kalvl, or laboratory
IV-39
-------
TASLB 11-4
yiCTOtt AMD fricms TO COWBIPtt II AMtttHS Of tfjCPtTAliTlf li HEALTH
nrs>t assessment now hazakpous ma3ti pisposai
ygmmjmi
9wt§mm,u»
Mlfiul hliMilCI
Tfdwlegy
lit*
TcM|p»rt»liM
tecMatt fwim&lfil.
ht4Mli|
k
Mr MIMIm
hut
Ss-a*
WmAa.
QlMtiMH Itlnui
U Air
ft 0nti4
Ti Sarf«M Msifsf
i»<
hllMMOT
* ftm
*» SayUM**
(!•
SettnailftLttiiiesL^Jtts
MMiutiae Nti
dfM4
0«la
NaJala
bull
Mmun^
Mill
tmmsisftteisim
¦attafeitaa
mmmmum
4ccw>l«fc|«a ftmnrnutwrn,
IktMCMllt lm
•AtMfslfltaftlM
VtyalcaatfftUal V(«oe««
hda UIkU
yfiissis
iMMMlt l^alillM hlMdlll
imM/tm, m it»ti itwtiMti
la tywa w»4 Tmw
MmIMtIm hti
bMWtl funrtm
muli
frvvaacy
Oa««l!M
•¦•klft t«pM* KM-MtatMia
—gal. IflaKalte--
ItlMttaM l«M
typm »| iffaclt
B«ur-R«apMM Bil«
(yt«in
fv^Mwy
CaMaiMltM
Vivemmd S4Imc«
IhIIiM*
IIm Na4«ta
fca«tt««
mm •It'll*
ll klKtiM/
Or»al*faaM
ktc* Ml ltl»itylu
MftfiUtiw l4mili(Mu
tffrtu
toll Ib4<«|JimI
MfvuUlUna
09CVf4l|aM)
Na|l*»tr litem
ittlfil
IftMfratiwi «• Tot#! Cflccta
r«l N#»Ut«««
-------
experimental system contain* a r«l»ci*®ly scull nu*b«r of wtri«bi« components
vich clurLj defined (or obitrvtbli) IncsrrBltciaaihipi. In ixwctpiriiwittl
• JfitiMt tennitivxty analysis can address only chose variables wichin « nodel
or nmch«MCic«i aquaeion. In r««l decision (taking cituaciont, however, the
«*joc uncertainties otey lie in factors that an analj«i» finds iapoitlbl* cc
fit into & quantitative model. As aate4 previously, the nuaber of variables
itt the present co«p*fAtiv® risk usessnenc of waste treatment, storage, and
disposal scenarios is so large and the iaterrelacionships 14 ietprecisely luxown
in Boise cut) that a rigorous sensitivity «n«lyjis of &11 variables is
precluded®
to analysis can be nade, however, ol the sensitivity of the final
result Co changes In key variables identified In the pcectdittg seep, fa# mmj
ba quite valuable In the determination of ov«r«Ll uncertainty.* This analysi.4
cm determine the critical ranges of these key Ucccri. It amy be perticu-
larly helpful in estimating the alteram value of a given variable that w#ul4
have co occur before tha overall risk of a disposal technology rose to soom
predetermined level of concern. An analysis of the Lupacc of tbe simultaneous
variation of Multiple variables is precLaded on a routine basis whan the
number of variabilis Ii acre than a do sen as in tha analysis of isost examples
of hazardous waste (bAa|eMne.
Sources of uncertainty co be selected for sensitivity *n*Lysi« vary
by wasta stream, scenario* aapoaure route, type of effect, and population
characteristics. Tha (tontl checklist of possible source* (Table IV-11 and
the analysis of tha pr settling icip will be Mtlul Lit identifying Wiy »atiablia
and assumptions for analysis.
3® Aggregation of uncertainties: Tha affirmation of uncertainty
across a complete risk asaesscent for a given scenario i< based on tba prop*-*
gatioo (or cascading) of errors methods. Quantitative measures of the uncer-
tainty of aany variables are not indisputably apparent, of course, so the
method uses qualitative discvasions , expert judgments by project stuff
members. and tha results of Cha sensicirity analyse# to estimate such values.
The overall risk calculation is structured, as shown in Chapcer lt
as • product of a series of factors, each representing art essential step in a
series of steps* The equation t»z
tt ¦ Fj a P ^ x > >• Fjj
The upper «M Lower I laics of the range of the risk (i.e., the
Boeartaiocy) may, vich appropriate assumptions, ba expirees ed exponentially a«
followss
* This statement uiumi the key variables have been accurately Identified.
If they have not, the sensitivity analysis nay given an ioaccvrate
indication of the true atate o£ uncertainty regarding the decision.
rw-4i
-------
2isk rant* »
m.10l=l !«f * ui * •••
j»l0M tUi * U* * "• V
1/2
1/2
for c > Q
for c < 0
where I I« the risk mt calculated by ,fbe®t esiinjuisc#" values for all variables,
the tl| tens arc the virimcei* of the Logirithiii of the uidlvldutL £*ccorif
I el it the absolute value of a. decision ptrmccr t that reflects tin
dt|r«e of caafidence dtlir<4 la. tht data."**
Although |t| varlM sooewhat with the distribution (i.e., normal or
various no-anomal distributions), decisions Involving the oOBparieotis of
alternatives art relatively insensitive Co the vaLye of | c | a* si. toed. That
is, if one ii coatpjariuag the risks of four alternative wim 4i»paiat
technologies for a given met, one would wane to use the t
n
K}
i/t
i/t
or cha aggregation of uncertainty it shown sinply by the expression
10
cuf ~ u| ~ ... o*>
1/2
Jf on« considers the five ma jar factory listed io Table ¥1-4 that
cause uncertainty in cha estimated number of ca»« of adverse health ef(ects»
than:
taoge of caaaa
J..C .f «... . 10 /"{ ~ ~ U§ » U{ ~ U|
k« «ci«» of «... . 10 /"! * "1 * »1 * Ui * "l
* Square of the standard deviations (or man square error ai appropriate).
** Note that if the best eatioate of F* for one variable it zero, the best
estinate of the risk, I, is alio *®ro and this method cannot be used la
let ei*ple for* co estiaate the range. While the best estimate of is
aero, the upper confidence limt
-------
A convaaiMt fotaac for tbe iKtaptiaa of Mctruiacics at tMi
factor immi is shova la Table tt-l»
TMtJ nr-j
SUWU&Y Of " TOM Of .WCttTJIItfriEf
Total
UlKfruiDtT factors
1
I (Bach it«c«4 ii eatlaecod ordar m fM|fiicvtf« iOu> -
4
s
The a*C«rrelati«S t( 1*0 v*ei*bUa vitbla 4 factor (Table rv-4) My
tiaiilarly cri::s=|»il<:rf II thuf ara Mdciplicaciv* MrUblM «r amy aay ba
appropriately »aaMd far thote Umc are o44Ulv«, or • coabiaacioe af opera~
clout cm bo uoo4 •« appropriate* A oaapLo M*kabaat for tabulating these
factor* md paraMccrt unit «iit« MnatoaaAc eaeoario* is illustrated i«
Tablo IV-*» ia «bich tpaca li provided for entry of nuii of up to five
v«rl«bln> Tho lite of variables In Tabto l¥*4 it a useful checklist, but
could bo expanded In diffirmt ways for different applicotioao* and eoold
probably be expanded significantly for • specific application*
four difficulties la pa*faf*4«| such an analytic aro readily idooti-
fiod. firstt the masher of Individual sources of oACortsiaty (the sabfactors
or variables noted in table lf-4) Umic caa bo idaotlflad Is a cee^reheaoive
aMlyais aey be to large (^ttbpi a 4*«a» Or aer*) fbac considerable effort
will ba nqairad to acqalra available daca ami tha probability
distribucio* of each. fMODdly, cM iaad*4«o*y of tba available daca* My
farad eaoaidarably Mra dapMdtaea m mcfiterz 6hM fwfifwi ia esti-
Mtiif oacertaiatias* thirdly, eM tec«w®l4®i«(»«hif§ Ntvwi tha sabfactors
My be diverse—jo
-------
wloi
NOflWOllNI
(nun timnii
•1<.
wornv#
?INO*SW-«Od
I.
. ,,
POJ3VJ IVKOIK3
f
r»
{•0I7VJ tfOtWYfl
.
'• *>
i«oi3*j icvviw
oiwwos
iwiwwwf mm
SMixvMimw h»33vmvm msmgftotoivzwjio JUMHss^vlisIFiinym..
« saiiSiviHiDHn m sisaivny 11 wwi jo nouvovmm « miMM
9-Al aim
-------
the probability distribution may be quite skewed. An iv«ri|c uncertainty
range I# tlsMtfow lass preferred for every variable than an explicit state*
-------
-------
V. SCHJItCE ASSESSMENT (HAZARD CBMACTCT1ZATI0I)
The source aesasstmnt portion of e risk essesament conciins t»o
«sjor components* (a) lizard Identification end de»cription I and (b) quail-"-
tification of releases of environment*! cotcaiiflUO. Source iistiiBenci can
be perforated at varioos leveLs of d*c*il, ranging froai siwpi« to costplex.
They can be global or tits specific, consider esisciag or proposed facilities,
V)4 take « single nediua or imlcinsdiji approach. Definitive source «sieu-
nencs require substantial inpaes of information and rfaea, bat isodel 9aurce*,
scenario «ppru*ch«*» end engineering neiiMUa can help yield useful results
when ch* dec a bue Is limtod.
A. Basard Identification md Description
In Chapter IIf lutirds were described as potentiil sources of ed-
tificci. Where aheaieel ccchaalogiei are involved, specific materials,
conditions, and activities »li wight be deecsed hazardous. M**ardau« materials
are those having intrinsic physical or biochemical -^*|., manufacturing,
formulation, transportation accidents
• Product Use latcami—«.g.| unused agricultural chemi cals,
•pent solvents or treatment liquors, recalled productsr acci-
dentally contaminated products*
Kaaardous Waste Treatment Scorege and Disposal Facilities
(TJDFs), e.g.# storage tanks, vests piles, incinerators,
landfills
-------
Scvrrcta My be dajcribftd if beLrrf * specific facility Or >ita» or they nay be
Aggregated on a one ba«Ls, e.g., by industry or region* Kushon audi CleoBaa
< 1981) hs*e enpfvMtzed the need to take a life cycle view in auatiing ch«
ha**rds of e assess*
narica« the chemicals may be precfcecerrnlaed by Che CM, based an a regulatory
agenda or on petitions from parties-at-Lncereat. In other catti. a prelimi-
nary evaluation of a coerplex mixture of clwwicela my be necessary to select
those to be assessed in detail» A «etpr«iw>iive a*«e»tB«nc should identity
all chemicals that reasonably oould bm expected to pose significant risks,
Tfa,« DMijumt includes evaluation of aonaafttrations , amounts present, and
possibility of release to the environamnt, and a preliminary evaluation of
thair physio*!? cfeaanLealt beaLtb effects aod environmental properties.
Handbooks, such aa that by Verschueren (1983), and environment*! ohessistry
rafarenca works (e.g., Stumn aad Morgan, 1981) are useful sources. Vben data
are unavailable, information on related ckesicais ao®eti«» c*n bm with
extrapolations, anaiagics, or tcructure-activity rel«cion«h.lpa for r«aaoaable
approxiauitLORS. Lyaan at al * (19121 provide aoe uoefui ra^ource for cbcotLcal
property astiowtion. 7ha evaluation sUtould identify thoaa cbanicala moic
Likalf to bai rvlaasad froar the source and cause adverse effaces.
Baaith afCacts Co be considered iodudei general tomltlcyt onco-
genicity (causes tuawrs) and carcioogesxicity (causes cancer or LeukeAia);
«utag«Ricity (causes aioCaciona}l teratciganieity (causes d«formad tecu*a*)i
^tarrility or decreased rc^rwiBCtivt 9uoB«sa| behavioral effect*; and calluLar
or tulMealiular affects* Useful iofotmaclao a out data «ourcaa includa litara-
curai raiporta on eootiroLlaRl toaLcole^icai atwdias, elLnicaL obaarvationa or
apidaoiiologieal studiaa on kuouwi; acuta^ aoboiiroaic, chronic and special ton"
icological tasting with labosaeory or 4oawatic anioalsi toxicoLogical tests
fiith sucroorganissM; aad perciaanc biochaadcal tests.
3. Igtlitana SMrcbaaisaia and poiats: This atap involves enginacring
analysis of production, distribution, use 7r disposal processes as necessary
to determina specific process activities and poioca at wtiicli the cbaaicals of
concern are released or escape Co the eavironoenc. Both the route of release
and Che receiving environmental nedia sbumld be Identified for each chasacaL
of eeocero.
v-a
-------
Chemicals cab b® rtl«tted Into the environment by several routes*
manufacturing emission* to air es vapors or pertLeulates and discharges to
waiCftr in folution or suspension! dissipetive uses, such 11 pesticide apfiica-
tloo; Insecure disposal of unused ontterial* *o4 viscn, followed by runoff,
leeching or voi#ciii»»tiooi «nd accidents, such as tank truck, spills, war«r*
bout* firas, ece. Careful «ttention essst be given co potential accidents and
emergency mrr«if€*LOC» i« well as to conventional operations, Safety audits,
hizird indicest buard surveys and op
-------
atanuf securing ami use ctogoriti• Sjlqcq tb«jr *re of prLairj concern ce ctvis
project „ houtvar, chty require farther coimeac her*.
• Releases during ..treati—Ptt The potential for rtlui«i during
treatment of a hazardous waste depends on the tccbaiqus teing used (e.g.*
itakitiiatioo, neutralization, reduc C ion/osida t ion, solidification« and
extraction). Frequently, trutmol metheda arc analyzed on a waste-strcam-
bsrvistcicrtu bails I u a result* tke applicability of the method to a par-
ticular casta msc be determined before the wait* release potential can be
assessed for a method. The potential release could come in the fore of fugi-
tive emissions, process enastiite#,, a spill, or an accident in which a large
quantity is reieistd* The for® of rite release also will be w«*te specific
since the waste*s chemical and phyfical eharacteriseiea will influence the
type and of release. The availability of data eo quantify the aflouac
and probability of waste ralaaaea daring treatment will be dcptndcog on the
particular vaate and cba creacsicnt techniques being studied.
• Releases during handling: In the process of handling «trtit«
several points can "exist" at which pollutants are regularly or irregularly
released to the environment. Major potential release points arc first iden~
tifLed and than uaounci released «ust be quantified. As with treatment» han-
dlini releases are dependent on the specific properties of the waste and on
the characteristics of the system. Because data for Che release of a par-
ticular waste are not always available, it is often necessary to use release
data for similar wastes subjected to similar handling practices. In some
cases, consideration of appropriate environmental transport parameters nay be
a««tai ia iiCMCifit release rtiiSi
• Releasee during transport! Release* during che transport of
hazardous wastes say occur as a result»f improper containment ar accidents.
The analysis involves estimating the probabilities of occurrence of releases
(particularly from accidents) at different location* and the anounts of ma-
terial lively to be released. Data of thta sort have been compiled La a
r#f4rc that assesses releases and costs associated with truck transport of
hazardous wastes (Abfcovica et al., 19A4a). A sifciLar report hat been con-
placed on rail and waterborne transport of hazardous wastes (Abkowitx ec al.,
19B4b). Data to compile these reports caaie primarily from the Hasardous
Material Incident File (HA2MA?) Maintained by the U.S. Department of Trans-
portation, Materials Transportatioo Bureau.
• Releases during disposalt Haste releases during disposal op-
erationa can be a significant route of exposure to human populations. The
releases cao be the result of failure of ®b« or more components of the system
(c*|m liner failure or failure of a leechate ooilectloa syerem for a land*
fill) ot a process emission (e.g., stack meistiao frae an incinerator)*
In sone cases, mathaauatical or computer models can be used to pre-
dict releases at a TSDF. fugitive air sedations from landfills, for example,
can be predicted! using an cquatioa baaed on soil bulk density, vapor flux of
the chemical from ebe soil, soil porosity, and vapor .density of cbe chemical
(Farmer eC al*T 1980). farina et el. have evaluated models for estimating air
emissioaj free hazardous waste TSDfa. DA (UtS) has assessed emission
V-4
-------
problem during incineration of hazardous liquid organic wuce}. Ehrenfeld
and Ong (X98A) h*v« evaluated eetisaioo control# for hasardoo* w«ste T5DFs.
Predicting releim to groundwater fro* I And disposal facilities is
more cooplex because there art atany eonfoflinti of che ayseeai chat can fail,
for axaatpls, iac«racci«nj between iMchtct and clay barriers can be complex
(Anderson and Jonts, 19821 Daniel, 196*). Otic Approach ij to assume that: if
Che tootaiouMnt and leachate collection have failed, than contaminant
concent rations in the leachaea released will be «ppr«*iin»tet.y cfae same as chat
given in a siaple leaching c««c on a aaepl* of tht w«*te (using leaching con-
4icl«uj essuned to tioulate those ia cbe land disposal facility). Matheaat~
ical Modeling approaches ara of recent interest alio.
A nodel under developawnt chat atteapts to stake quantitative pre-
dictions is the Pope-Raid Associates Land Bispoaal Failure Model (PRA, 1984;
1985). the model provide* estimates of leachate relaaae* from hypothetical
La (Mi disposal facilitiea (landfills, aurfaca inpoundasanca r waste pile*, and
land treatment units) having a variety of design configurations. The facil-
ities cad reflect several different clIraacic regime. The nodal can b« run to
give annuaL output*, If desired, with landmark tines
-------
For nasardous v«ate disposal» tha releases» exposures, and tmpmccf
will depend on the mture of the waste i a a givan isitDitnt. on the site of
cates traphic release, and on assumption* COIIC«min| the doing and e£ f iciericy
of protective and corzeccivt aCCions. In ton* cases( the nature of the waste
And the disposal technologies raigbc make negligible the probability of CACat-
c ropJEkic releases from many cc eveo most causes. for axarapLe, a none lamia bit
sludge of lo« volatility* Low solubility in wster, and high vijeosicy could oe
cleaned up «ich nininuiB ri^k in event of a large spill Eton an overturned
truck. In via* of the usual time and resource I ifti tac ions, «I: fores to luejf
lucli risks could b« reasonably mmi.mixed in the health and cost ajacsstnent.
On the other band, attention «iay b« required for (hi eifkj of
flooding during cleanup of an old haiardome »aici disposal site or for the
cielu of loading/unloading activities in c rac sport ing wastes far et-sea incin-
eration. At a uninia, transportation riiitf should b« discussed qualita-
tively^ sufcjequenc quanti tative analysis nay b« desirable before ultimata
decifions are reached. The quali tac ui
-------
Reference*.cp Chapter 9
Abkowicz, M., A. Kifcr, and S. Srinive»aa. Asxiiing tbe llslt^ and Co«t*
A*$oeiit«4 wich Truck Trecaport of Raaardcraj M«»cei Preft final fieport.
EPA Contract No. 68-01-6621, Off tea of Solid Waate, U.S. Envirenitental
Protection Agaacf, Washington, DC. January 1984*.
Abkcvitz, H., I. figmr, I. E&gelsttin, and. S. Srioivaaan. A*s«*iin§ the BLitlta
AMociated *ith Ijil and Wacerborne Trans-port of Hazardous Wastaa, Drafe.
Prepared under EPA Cone race No. 68-01-6621 for* the Office of Solid Waste*
U.S. Environmental Protection Agency, Washington. DC. 1964b.
Anderson, D. C. , and S. C. Jones. Clay Earrier-Laaehac• Interaction#
pp. 1S4-160 in: MsfiUigj of Uncontrolled Hazardous Waste Sites*
October 31-Noveaber 2» 1933« Washington, DC. Haxardou* Material* Control
lesearch Institute, Silver Sprist* t®- 1983#
Cfcapcaan, G. 0. r D. W«. leubeofcer, L. A. Kartines, R- T. Kacthewa, 0. A.
Oberackar, and P. Wyfeenga. Chemical Wesce Incinerator Ships: The
Interagency Progreta co develop a Capabilicy in the United Seaoi. Marine
Technology l»(4) 325-140, 1982.
Conway, ft. A., P. C. Wbitnore, and W. J. Senaen. Entry of Chemical* into the
Eavironneflt. pp. 61-14 la Envi rcneental Eiak Aaalviii_ for Chenicals,
R. A. Convey, ed. v Van Moatrand ielahoid Cowpaoy, ftaw York. 1992.
Daniel, D. E. Predicting Hydraulic Conductivity o£ Clay Liner*. Journal of
Caotechnical t&nrtg—ring 110(2) 2B5-30C, 1984*
EPA- Identification and Lilting of tUiaardoi** ttcdte Under SCKA (Raaource Con-
lervatlon And Recovery Act), Subti cle C, Section 3001s Liating of Ha*-
ardors Vaate; finilLsmtio« of Joiy 16, 198(1, Hacardoua Waste U>c (40 CfH.
261.31 and 261.32). 0-S- Environoental Protection Agency, Washington,
DC. 1961: BTTS, PMI-190G76.
EFA. U.S. Environmental Protection Agency Standard* Cor Owners and Operetors
of Kaeardowi tf«JC€ Treatment, Storage, and Disposal Facilities. Coil of
federal Title 40, Part 244. 1963a.
EPA. AP-42s Supplement Ho. 14 for Compilation of Air Pollutant Paccori»
Third Edition. Office of Air Quality Planning end Standards, U.S.
Environmental Protection Agency, Xeaearcb Triangle Perk, NC. 1983b.
EPA* U.S. Environmental Protection Agency Ocean Cuttping Panvit Progren.
Ped^rel Itiiter 4§(205) 46986-46996, October 21, lf«3c.
EPA. Assessment of Incineration u « Treatment Ketbod for Liquid Organic
Hazardous Uaitu * Suaaaary aod C
-------
EhreA-feld, i. &., and J. M® Ong. Bvaluacio-n of Eniition Controls for Baa-
ardoaa Waste Treacaent, Storage and Oiapoeal Facilities. A. D. Litcle
report to Environmental Protection Agency. EPA/*SO/3-(WO 17. Mowemtwr
19M. 167 pp.
Parian, W, , P. Spam, W. Jaainafci. and I. Murphy. Evaluation aod Selection of
KodeLa lor E»tiutin| Mr Eniisio-n froa Ha*erdaua Waste TrMcaioe,
Storage, and Dispoial Facilities, Bevij«d Or«fc Final Report. Contract
Ho* 6i*02~3l68, Environmental Protection Agency, Office of Solid Waste,
We•hi*|con DC. iS§3*
r«nu*r, tf« J., K-S. Yang, J. Letey, and V. f. Bpenccr. L§n4-Oispoe«l of
Hexaehiorobeo**n« Uticii Controlling Vapor Hoveeent in Soil- EPA-60Q/
2-BO-ll, Nhmicipel Environmental Itturch Laboratory, U.S. Environmental
Protection Aj«ncyt Cincinnati, OH. August 1986.
Huflhoa, J. H., and K. J. Clemen. Eatiaetion oE Exposure co i«s«r4om Che*-
icelc. pp. 323-388 In Beaerd AaieipBent of Cheaiical*. Vol« 1? Carme
P
-------
V»nchu«ren. X. Handbook of Ecivifonpgnc«I Pat* on Organic Chemicals. 2nd
Edition. Van Nostraod 3*innold Co«np»ny, New York, MY. 1983.
W*1ler, 1. A., and V. T. Covetlo, eds. Low PxObaDlIicy» High Cons«qu«fi£« Rl»k
toaiyiisi If sue*, Methods and Ca«# Stud i e«. Pleoua Presi, NY. 19t4
(571 pp>.
y-f
-------
-------
VI. PREDICTION Of EWTOOMWCWTAL TtAS8TOtT AMP FATE
A critical part oi mom chettical capomr* auiiiiinc* is the prr»
diction of the aovemenr end reaction! of che cheaiicals betweea their sources
(peine.* of releaae) and th« pointa in space nod ci»e at which they otighc reach
hunan dt other receptors (Meely ami Blau, 19B5), Comprehensive assessments
¦uot consider all oa jor pcthuajrs of transport and any transformation of the
tmic (material batfeea tie point* of entering che environment and poloca of
exposure (Slau, 1985}. Pathways includa aCJAOvpheric ead tqutcic transport
(fiwluni la inhalation or ingestion in driafciog water or through the skia)
and p«»M|e through [hi terrescial and aquatic food eheiaa into human foods.
TransEoraacloaj may. includa cheaicel and biological reactions and interaedia
transfers. *ultinedia txposure asiessnert car becoate the Most resource-
demanding part of *r overall risk aassasaect #1 a haaardou* material, but
partial ixpoiurc ujasatnti can oftea provide information useful far «aay
regulatory deliberation*.
Several hundred environmental transport nodels have been described,
Their claaaificatioa is noc easy. Broad categories include weter nodeIs, air
¦odels, ecological models and interaedia models. Veter Models are generally
divided into surface water and groundwater models* and the latter into chose
that address the unsaturated or variably saturated (vadose) acme end the
saturated tone or aquifer* Increasingly, however, never aodels tan address
b*th the unsaturated ead Saturated tones, or the iaceraecions between surface
water and groundwaters, or ether nnlciaedia interactions. Media nodeIs »re
Aim coMUnly ellliif£*J deeording; t« their Batkesutical basts or type of
applleecioa. This chapter diacaaaea che hinds o£ data generally required for
analyaia of environmental transport and fate and che aechenetical models chat
are available for specific pathways.
A. Ceaerel Peta leqoirewtnts and Sources
Information on the pfcysicocheaical pro^trtiat of the specifia Ar-
terial are iaportaac in evaluating cransport and fate la the environment.
These properties includet meLting and boiling points; volatility? solubility*
viacoeicyt photolysis rates; hydrolysis rates; oxidation/redaction races;
atmospheric reaction rates with ozone or hjrdroayl radical* biotransformation
rataa; vapor particle siae and density* octanel/water partition coefficient;
soil adsorption coefficient! and other sorbtive properties. 9oss of these
propsreiu ara a pec if io to transport in aqueous environments and others to
transport in air environments. Etch property plays a significant role in
predicting the transport and fate of apeolfie chemicals. Hethods are avail-
abla to assist in estiiaatioo of ehaoue^l properties (Lyekga et *l«, 19821
Peterson, 1983). Cailahaa et al. {Lf7t> review the watar-releted acviroo-
aeBtai fate of 129 priority pollutmatia. Buahoo and CLernaa (1911) haw*
relieved infonaetion sources for thsi overall esyoauro asseesaient process.
Oooigittt <1981) discuased field validation and error e&alysia in codeling che
fate oi ciunioala in the aquatic eavircmsMnt.
Vt-1
-------
• Su« climcttfltcici? Information and data are needed foe both
the major physical parameter* of the iici and the locations of population* of
concern. Physical characteristics of the site and the aurrounding terrain are
inportaot ldcc«ir» la selecting the moat, applicable transport nodel*- The
types of inforaatioti about eh« site and aurtouading locale which are of
interest ineladai
• Specific fcagraphic Location
Topographic naps
Soil amps
• liaaraac body of water
• Drainage pattern
• Types of emission sources
• Stack height and plune rise factor (if applicable)
The expo sad population, which includes wortoera at the sice as well as the
p#opit living in the vicioity, La discussed further to Chapter VTT.
• Phyjicochewical parametersi Soils data required for a variety
of transport and fate e>odels~~incluilng aeny surface water and groundwater
widelsTan be obtained fro* a awefecr of source*• The best «auras ia county
soil mt-my reports published by the U..S* Oep«*tiMAc of Agriculture. Soil
CaQtervatian Service* General toll Mpi are alto available froei the state
Soil Conservation Service of£ioe. Soil acientlscs hunowLedgesbla of particular
soil pro pert iea can also be found in this office and In the agriculture
department of the state land (rant university* Sxanpica of information needed
for save oodcls are: aoil type, organic matter content, pi, bulk density*
ncisturc content, particle sise distribution, temperature, vegetative cover,
slope and slope length, aoil erodibiiity. and aoil oanagerent practices.
Surface water data are neceasary for (torn water runoff and stream*
nodels* One of the largest data bases with this information is STOfcJST
(Storage and Retrieval for Water quality Oeta) tnaintained by CPA. Other
usefal data, such as storm hydrograpta aiud high/low stream flows, can often be
abcaieed Iron agencies soch as the fJ»S. Geological Surrey (Uses), U.S. krwy
Corp® of Engineers, Federal Insurance Administration, and state ernHLroinaencal
and wacer reaovirce agencies. Exe«plea of data and information that night be
required in aoaa ecodela arat
• Streaa flow rates, pB, taoperature, and disaolved oxygen
• Stream sediment load
¦ Background water quality
• Stora hydrographe for individual sites
• Surrounding land uaes
Groundwater transport Bodela are neceaaary to predict eiovement of a
hazardous pollutant through the soil unsaturated and saturated zones and to
VX-2
-------
predict distribution of tht coocaadnant in the groundwater Aquifer. Data to
run ch
vater resource boards, groundwater aextetaent district*» scat* ead municipal
health departiients, and jsuaicipei. vugter supply departments. Although it
•ppear* thai there are nuutry sources of inforsecion r obtaining hydrogeologic
data for speci £ic model pirimccn can be difficult bcc«us« there ere so Mny
by^rogeologic parameters to q'uanci fy« Be«t eseLoutes oust: frequently be
made. Eitcopies gf typical groundwater parameters needed by groundwater trans-
port aodels «re:
• Mf4"r»uLic conductivity
9 hydraulic gradient
• Transaissivity
« Actual aquifer pocoaity
• Effective poroaixy
• 'Depth, to groundwater
• Saturated thickness
Transverse and longitudinal dispertivities
Seepage veLocity
• Sulk density
• techarge race
« Soil permeability
The use of air transport * «odel a for wultusedia exposure «sie>smnc
requi res certain inforaaticn, Including meteorological deta, physical and
cherrical properties data for the substances, and source emissions inventories*
Meteorological anbient air data can be obtained fro«i sources such u the
Stability Array (STAB) data base of the national Oceanic and Atrsoapberic
Administration and Storage and letri«wai of AarooMtric Data (SA80AD) of the
CPA* typical meteorological data required for most air models include*
» Wind speed
• Prevailing wind direction
• Precipitation
• Atmospheric stability
• Cloud eo*er
• Median* and mioiamai del ly air tseparators*
« Klxiaf height
- Solar radiation flu*
Other ¦elected neceorologic*! date any be required for specific air transport
and dispersion aodela.
VX-3
-------
8 Biological patmi:tf«: Aiaisiiini of ativironneneal transport
and fat# procama 11 greatly complicated IC the pQlluunci Interact with
living organisma in tha environaiant. Dita requirements increase «ub*tan-
tially, but the available data btit is often frefmuitery* The Mtut« atHf
•xtene of the biotic interactions will vtry ptif lenocly wick the physical,
cheaical. and biocheeucai properties of the specific substances of interest
and with Che populations of ar$*oism« ac the »pacific tiest* being analyzed.
These interactions occur primarily in eke terrestrial And aquatic lyttuu.
although i»t«*«ccioaa in ateospherlc md grouttdvacer tyuiaa art not
precluded.
The kinds of
-------
Intarto tot ic/intaraad La r«(«rt f« tbo rvlt of orfasiwi
in Hbilitios pollscwt «r (htir MCikolilti fron aoe «a»ir«i"
until aadiaa tm aaotbar, aotabolioa of haurdmis wasto
c, &WL
(1914), im Pensceraacher ami Qtcinctti llliJK
Tfct ftimtf aecWui that §•••«• (rMl^rl Of pollutants In
groundwaters it e«meclot of diseolvad a« water m>««» through cha
Mil wcrii. Salute eraaoparc it by i«i«r-
MtiM* betMiaa tlia lolsta and £li.« Mil oatrla* Itefa internetinoe irt cm-
aaaly quantified ittiai the aatl/eatar distribution coefficient, 1^. Cbeaical
characteristica and field conditions vkieh ItctitM tb« potential for ground*
wur coataniention include: (1) (ha cbeadcal hat water solubility greater
than 30 ppau (2) the cheaical ia negatively charged at anbianc pll; (3) the
coefficient K (defined aa K. divided by toll organic carbon emtnt) ia less
than 300 to s8l( (4) tba ehJ,e®lfi degradation half~lifa is irutir than 2 Co
J weekst and (5) total precipitation it greater than 10 ta/year (Travis,
19851•
Doaena of ondala are nov available far predicting or coeparLag cha
aovaaant of gr*uad*ater conteednaatt under wtrioua condition*. They Include
aatbaactUai aodals that are fairly daaandlat af Input data and ranking nodeIs
tbnt aty be uaad la raaponae tituacloaa* 6*th typaa of aodals ara
renrit«*4 in tfcia Mecioa<
I, fticfc—tci«t aodalt: Marti—
-------
simulate physical processes)( or stochasti c (i«e.» us Log probabilistic predic-
tions b»sed on ipmlitlod of p«raaec«ri). Dtciriiniicic sodtl a
4Sfuae th*C tbe parioeCer$ J.Q d3e aauae-alfect relationships and other uncon*
troliable variable* «r« fixed or known* an4 thin determine an optim* value
Car soma variable of interest. The review below follow* eke analytical/
numerical classification.
loch anaifiieji and nuowrical models are b«»«4 on partial dif-
ferential equations describing groundwater flow ant list# ailMtr aodels that were
considered.
DataiLad information an data sianagaosect vithin the context of spe-
cific analytical or nuaaricel models It beyond the scope of cilia review, A
few of the more valuable resources InoLude He rear and Fausc (1981), Che
groundwater model data base at Holcomb Research Institute (1983), EPA (1982a),
and BWt (198*). Another valuable source of information is the EfA Robert S*
Kmtt Enviraesaeotel Research laboratory, Ada, Oklahoma. Analytical and ouner-
ical models are aoopared following potential applications* and limitations of
each type are noted.
a. tmrnlftisssl models! In analytical models, relationships
usualLy are tlatplified by assuming "steady-state conditions relative to fluid
velocity, dispersion dynaades* and other physical parameters (JCufs et al.,
1980). This simplification results in equations wfoich can be solved in func-
tional form to c*loal«ce specific values im parameters of interest, i.e.,
dilution, dispersion* and ttccaiucios of groundwater contaminants* In cases
where the broad assumptions are valid for tj*e aotuel hydrogeologieatl system
being modeled, this approach yialda rapid, results* The usefulness
* The uae «! analytic*! models doea not require the uaa of grid*, while the
oee of numerical models necessarily involves gride (Keely, I9§1).
VX-*
-------
TABLE Vl-l
G&OliNUWATEk SOLUTE TRANSPORT HOOELS EVALUATED
Co4t Appli cable Hod«l Typo t
Hubs Zone QttaracLeri si ics Reference
ATI23D SaluraCad Analyticali ' fch, 1981
1-, 2- or 3-Uintena ional
iIOFI!if Vadoae/Aqua t i c* Microbial liegradal iont Ri ((ndn €C al ., 1980
CP£ST Si(ur
-------
TABLE VI
5
i
Code Applicable
Hjjiae Zone
PI um HiMgeaent S»iur®f ed
P1ZH Root Zone*/
Unaaturaced
SESOIL Surface/
Unaaturated
SWIFT Saturated!
SUIP2 Saturated
T&ANB (Rando* Walk) Saturated
TKUSt/HLTRAN Unsaturated/
Saturated
IMS AT ID Boat Zone/
Unaaturai e
-------
tmc it-2
ADMTtttAL aOUHPATEl SOUftg IMlHWIt WBBKL1 IDOnriFlCD
mm.ro CM, MOOtLS
Greuad ifctnnnttiif
lMct«u flu* Miimcioa Pr«4ictii
tcmnisi ProctJur*
Orcit « al., If If
Kane, lf»t
rtUo «c al.» IfSI
MUHltlCAl wooits
Pollutant Movtatflt SuaolaCftrg
rejocv/tomo
CCOCUBH
laachact fraval "timm Modal
So tut* TraMftfcASratMduaea* Fla*
latclMt* Or|Mic Nifracioa *o4
AtcvauatiM Hatful
OuLattl and MheLL, Iff?
Gupta it iil.» if79
fpdaito aad Hatcigod, 1990
t*A, itlli
fartar~ if12
SyKJM* at *1., 1112
VX-f
-------
of such »#iueionf# howaver, is directly dapandtnc on adequate verification of
assumption* by field observacioom. Mal/tioal modeLa Art generally inappro-
priate fa* aquifers with etntfLn boosduy conditions or ochar characteristics
which do not permit tb* n«c#«s»ry as&unptioos«
Mod#lt which art deslgfltd for haaardoua waste sit* e"al«iations
Cor poceaciaL groundwater problem* in fr«qu«ntly applicable to several
scenarios. i«c»4s»« cheat ecdels art designed for speed u4 eiae of applica-
tion, they havm i imitations. Of the wodef« ii»ce4r the AT113& (the Analytical
frmiittc our*! two-, or chrf«*QiaemioAal model) 4p{Muf btic able to model
u«jte crtBvport in bocK s«tur«Ce4 slid LiDMturac«d tone*. It usee a simple
analytical approach eo escimca concectraciona with minimi input data.
SES01L (the Seaacnal Soil madel) iimlaiei uaier fl«u and chemical
coqcmcration* 10 the unsaturated zone. The HcVhorter-Nalaon model is a
hydraulic expreaaion uaefuL for selected applietationa in the unsaturated tone
that do net require consideration of adtorpciao inceractions. None of the
models addrasse* secondary porostcy, laniacible liquids* or multiple
concaoinancs.
Data requi*mm*nct for that use of analytical models fall into
three categories: aquifer boundary conditions, hydraulic variables« and
caoCaninant concentration variables. Boundary condition* include diatri-
butioni, types of boundaries, flux point s, and media LhickjieSse*. Hydraulic
variable* 1 oo Iude poroaicyr hydraulic conductivity, dispersion co&ifLclents,
and attenuation coefficients* CoBtaerioent conceotration variables include
initial concentration*, releaae rates, and flushing races (Kufs et al. ,
1980).
Input-output parameters for individual analytical models wary
somewhat; Table VI-3 present* an illustrative input-output breakdown for the
tandea PESf.AH/PLUME modal. Mo attempt ie made to define fully the parameters
in this example. The listing ia praaented as m illustration of the acopa and
complexity of tha Input and output paraiMttar* comaonly associated «nth analyt-
ical aiodels.
Analytical nopals are applicable to fr«uxiiw«t»r analyais where
substantial dace describing the physical system art availablt and where those
data, confirm aquifer homogeneity and chc absence of complex boundary condi-
tions. The margin of arror for enalyti-eel model output is primarily due to
the margin of arror carried loco tha mn4«l by tha uncertainty of Input param-
eters; problems presented by inherent nc4«i arc minimal in coapari"
son. Consequently, tha margin of arror accompanying mass transport outputs
for a particular modal will change m a cd4e-by-case basis* Sources of uncer-
tainty arm most coamonly related to inadequate physical characterisation of
the aquifer rathar than tha chemical analyais of the groundwater {Keely,
nmy.
b. numerical models; numerical models require more input
data Chan analytical models( an4 under some conditions poesess a potential for
taore extensive data output. MuaMrical SMdala break up function* of interest
lot© many uaaiier units* Tha csathanatioml solutions chat follow involve the
VI-10
-------
MALE TM
miwmTOT ow M oroif tmmjfxs got hp mmmicm. cwomm
lamt Pig—ttgri
t»uo4»cy conditions
SUxiaua m«A neigH* dtftbf {«»-
¦•tsriui xoo«)
^MCWSKioa paint eeordtftat#t
(Mtanled zone)
%4r—lie
folk 4M*lty »( *011
Porosity ef soil
Aquifer porosity
Solubility of contsnlaaat
Sorption constant la toil
Degradation rate coefficient 4a
toil
Rtdtirie rate
Ditynraien coefficient in mil
Stdpefg velocity in aquifer
Rt(«ci«t>fs coefficient in
•qnifer
Oitpersiea coefficients fm
»• aad y-a*ls
Decay constant la aquifer
Concentration variables
HaxuBun and ¦IbUbuiT ttass
Tim duration of watte raleaae
zone)
Frequency of vaiti release (mm
aatnrated zone)
Active lafndiail releaadd
Man race i*ii«i tin®
SCea4f-atate source race
mMt
Curve coefficient
C#ocdtMte ayateei
Units
f iflf
Selrcel Enfield et al., 1M2 m4 Vgper, 19S2
Mliiir coateat of ostCintid um
Pollataat velocity ia unsaturated
mm
leftftfe *f pallutien a lug ie mm%ti-
nted tone
Depth increMnte is uasatura ted ion*
Solatiea concentration la uaiatr
riul 1MB
Solid phase concentrations la tta~
saturated sooe
Total concentrations la unsaturated
zooa
Concentration distributloa matrix
la aquifer at tines desired tor
aceadyecace
¥1-11
-------
reduce ion of partial differefitui equations to elgebralc e^uttions which, itt
turd deiine the values of interest within each subdivision of tiw givno func-
tion. In order for this method Co approximate eenlity cLoseLy, tbe function*
under study must be subdivided into as nuif diserece uaiej «« possible* As
the nuwdcr of small units In Che mo4«J iqurcnes, the mechanics of p«rfoin»ieg
the ncceasmtj ¦*che*tic«t calculations also become nore complea.
There «ct sevsni different founts aE noiific&l irtodels» but tht
tvo nose common one* tr« reftrttd co aa fioica-diffaranca approaches and
f inita-alaiiaat jpprMeiiii. Wich both cys terns, cmiIqmw Euoctioos prii*Lou.
to the finite-differenee models, the is a differential approach; in
Che finite-element models* Che mechanics is an ioc«fr«l #pf>*dact»®
The individual (rid uaics *47 be either two or three diaen-
sioa«ll they can be square, rectangular, CriaoguUr, polygonal, or cor-
responding three dimensional shapes. All grid units witkin a linfl# model
usually ouincain the mom shape* Crid unit sutf hot»*ver, often vanes within
a model * For complex aria* of tbe aquifer under study, a greater number of
grid unita par aria or pir volute nay be designated.
Data requirements lor the use of numerical models ace very
similar ta the been dairy condition, Hydraulic variable, and concentration
variable parameters mentioned eirlier for analytical models* However, numer-
ical models «ra designed co go btf*ao4 tbe scope at the analytical «odels and
co generate sore detailed, informative outputs, e.g., to account for a greater
nuuber of attenuating factor) and complex aquifer boundary conditions, to
acbiivi chase objectives, numerical models require a commensurate Increase in
baseline data.
The input-eucpuc parameters for numerical models «re similar
from nodal to nodal, but there are verieiciontu reflecting different modeliog
approaches and unique features. For purposes of illuscracton, Table VI-4
presents an input-output breakdown for the Random Ualk Solute Transport Model
or TUJNS (ftriefcett et al., 1961). Ibis numerical model is vidaly recognized
among hydrologists and is regardid by ime authorities as among the beat
available for many solute transport modeling applicatiena (Keely, 1983).
numerical models can be ueed to model groundwater transport of
hasardoua waate la « wide variety of circumstances. From a technical stand-
point, these nodeIt have the maximum modeling capability available. The fac-
tors which I lade the use of numerical models Include availability of trained
personnel, computer facilities, substantial field data, and allocation of claie
and funda. Such models would not be appropriate for routine ««•» but would be
very valuable In the study of isolated! high priority caaea.
VX-tt
-------
I
s
H
I
§
a
o
EU
B
m
<
an
e
m
a
2
§
tti
>
S «• fc
t « >
:i|
fs *
j& > i
hi
« ®
i
«* M .n I§ a
3 V _ <- £
it*
m y
tl M ^
ill
tis
n *
• f «r
*PS i
«#! * i *
«<•«<»
s:|:
r?:I
I
t ¦
t !«•
• » •«§
. Jt
— « •
• «*
k- -1
ft
c
*
&
*
5
« s»
f ¦:
i:
« s»
&
M
rjliP:
1 te m Is mm
-¦-II " "
f-SlJ J
— %s < |i mmm
B * ¦ I ft • J
as < &
-i r-:ii«__.
lis
i 2 »»— « — • m a
""— •. it 11 i H
* t « i 5 a,? - IT "*
-ll --lis*ss
ijtjhn j;.
«f55f2l|4"l fSSil'S SiS5|tK£|
— Wfl w~»C.£6«.« i 3 w 2 • •» « • 9
* »5* is • I 4 v t «» ¦ i « * • ¦ _•
¦ w I t u «. 0 = 6 a 0 • • « o m
«s
I
1 i
¦
¦
m
• • *
-Ii
a
- 5
Is $SS
U M
€ l»
11
s .s ll'
SSsil
8"*
m
H
S ^
si-rr
«b£ «
« 9®
v «
21 -1 i
B- Sff
II
!• B ff
5 t;?t
5 - » X W
¦u e b
m 4
y k> «-
a, k •»
»
I
JC
V
i
I
k
!
•*
m
»
n
n
fe»
e
8
3?
t
\
ll
W
*
k»
a
I
n
V
! ,
« If V
*4 * *
:s::
k» a
if tee ® fc»
• ()«>•
!i litJ
• • s *j •«» *i
i - i
1
•I5S
*» &$ <^i * h • » • k
:i•t S|•S.lil
I1»!--12 • " r"
ii «# & mm # y « i i b
M m m m. J> ** g • i
m ^ Ǥ , m m ^ <0
mm 9 Mm -» «i • u
#§ H ^ g «@ S H Im «| m e S» @
51! » : 8*5 3 ! 8 "3 It
I! • 4 » tf »«*<-*a»iii»
•c t» • « «* m m* w
m
m pm -m
m & M ¥
a a ^ —« sa Jk
^i»«
WJJ1
IS® w
i
H
ii
m ^ m
i*5
2% +
xzz
Z |£
£2 Jl
Z *j
J{-
»ft.
5 55
y ^ g
4 t B
51!
£1 j
VX-U
-------
The nimtrictL oBodeit do exhibit Inherent »inor sources of
¦athenatical instability known hi rrnmr leal dispersion or oumeric«l oscil-
lation (Mercer and Paost, 1991), However, with proper aumagenenc, errors from
thesis uncertainties are negligible (£ IOX) co*p«r«^ to uncertainties intro-
duced by input partveten (£PA» i982a)« In this regard. the cwronts on
uncertainty in Che «fwiy.tic#l model discussion «pplj here « well. The reargin
of error in nodel ohjcput for verified, established aodeis listed in this sec-
tioa varies fro« case to cise* based on input error. Ch«r««t«ri*#cion of the
aquifer properties, processest and boundaries ptMtnci cht greatest diff t-
culCf. If incorr«cc parameters are applied CO a numerical aodal , the COA-
plaadty of che Add* I will serve to multiply eta errors And an erroneous output
profile will resale.
No developing methodology was ida&tified which would radically
improve or replace the solute transport models described in this section. The
greatest potential f pr increasing Che accuracy of groundwater Model outputs
lies in improving the qmlicy an4 quant icy of input data describing aquifer
properties and boundaries*
c. Sp—yl^ of aachottacicai models* Capsule descriptions of
analytical sod nuaerieil aotfels follow.
4X12301 The Analytical Transient One - Two - or Threa - Diseoiional code ia a
versatile cool for modeling tha transport of wastes in aquifers with stlnimal
input data. Developed for the Dapartaanc of Eaarfcy by Yeh (1981) at 0RKL# it
can ba applied to inatantaneoua, extended period or continuous releases froa
chemical or radioactive wastes and beat flows. It can address eight source
configurations (point! 3 lineart 3 planar; and volume), four variations of
aquifer depth and width, and transport parameters of advectien, hydrodynaaic
dispersion, adsorption, degradation or decay, and volatilisation to Che
atmosphere. Boundary conditions can include Oirichlec, tteuaaxm, adzed type,
and radiation. It is written La FQ1TM1 for IBM end DEC systems* It is well-
suited to aodeling fcaaardoua waste t ratio port f roe land disposal sices.
IIOFILK! This aodal ii applicable in crtasporc studies ohert biological fiLfss
are the controlling factor La. upcake end microbial degradation of low concen-
trations of argmie ehamicals {ros aqueoua solucioua, Developed by Bittnan
and coworkers (1980) at Stanford University for the U.S. Environmental Protec-
tion Agency, BIOPILH haa bean applied to land disposal of aqueous wastes as
well as to conventional trickling filcart in waste water treatment plants*
BlOfTLM coQtaina both steady-scat:* and non tte*4y~vc«ce models* The former is
based on Monod kinetics for subatreta utilisation, aolecular diffusion for
sub*trace transport vicMa the film* and liqutdr-iayer mass transport of sub-
strata froei bulk liquid to the film* It predicts aubatrate flux into the fila
as a fuoation of its concentration io thai balk Liquid aod tha thickness of the
fila for a given concentration of substrata. Tha nonet eady-*s tat e node! can
predict substrate flua inm an existing fila formed at a different concentra-
tion of substrate. Tha model is written in VATIV FORTRAN.
CfgfTi Tha Coupled Fluid, Energy and Solute Transport aodal was developed by
Bat telle Pacific Norchweat Laboratories for tha U.S. Departsumt of Energy*•
ltaid«f*o^4 CaergT Storage ^rogratt (Gupta at al. 1982). Davalopad particu-
VX-14
-------
larly for prediction of a confined aquifer'* response to thermal «©ecjy
stor«|«» CFEST tut also baen of Inctresc co tl» 0* S. EPA lor study of ha*-
Arsons until d£ipottd in landfills. It uts** cbe finite element method for
fluid flow, and hat two- and three-dimensional application*. It has been
verification-tested against analytical and aami analytical solution*. CFEST
cm be compiled with a one dimensional siodel developed by the iioa ar(4ni£a-
cion for the unsaturated toae UVSAT10 (Battelle, 1982) CfEST is available Id
fORTRAM and FUCS languages and !¦ operational oc cba DEC PDP 11/70»
PEMWATgH/fOWASTg: These are affiliated Finite Eleesent Method codaa lor Water
and Waste constituents» Both use a Geu»»£«n elimination solution technique
and art two diBtfliicAil modela that simulate grauMtmter dynamic* in
unsaturatad-saturated porous isadia. Developed by Yah aod Ward (1980, 1981)
for Oak Ridge National Laboratory, these acate-of-the-art codaa can be uted
either separately or La tandem. FEMWATE& can iasLoda response of the ground-
water basin to precipitation, pumping and other recharge/discharge affecta.
FEtfWAm it senaitive to tha grid discretisation and aoil characterisation ac
sharp wiaturt fronts or vertical oadia interfaces -
dynamic data base to pe dissolution, ton exchange, and generation
of gaseous 0* and C0*« It eeploys Che VevtonHtaphson numerical canhnique and
vr-15
-------
I* written in FO*T*AH for the UMIVAC 1144 and DEC ll/TO, A version is alsa
available Cor the IBM PC IT or AT and compatible -
sionel aodel far a continuous point source as well as a fesdel for a pul»e
source of coptasdnent.
-------
PR2M? The Pesticide Root Zona Model (Cartel et al., 1982) was developed by
che EPA'« Kerr Environmental Research Uborttory to predict the movement of
surface-applied or surface~Lncorporated pesticides in or frelov the plane root
eon®. PR2K should be applicable to the transport and degradation La the
unsaturated tone of other organic ch«**icals with linear adsorption behavior
and firtt order kinetic reaction*- "the soil column is divided into several
layers, oust balances are maintained for water end chemcais ia each Uyrr,
and hydrologic Factors include rainwmcer t runoff, soow aocvuoalat ion/melting,
evapocranspiration and percolation, PRIM uaes a numerical seLution to the
adjective dispersive equation for chesmicaL transport, first order reversible
sorption, and first order lumped decay kinetic* (which can iaclu4« microbial
degradation). It can take into account both pLant uptake and runoff losses of
chemicels. PtZM can be used on microcowpetere compatible with the IBK PC
XT/AT spcwit
SESOIt.: Tbe Seasonal Soil Model was designed for rapid evaluation of solute
transport in the unsaturated tone of pollutants at waste treatment and dis-
posal sites* It if a semianalytical (statistical/mathematical). coeipartmentai
model that, with minimi data requirement*, can simulate waterflow and pollut-
ant con
-------
TtJiHi taflddw Walk Soluca Tcaooport UMitlf Tmsptrt of « largo «la*« of
soiocaa io growodsater cm baaliialated is this (wmrtul aadel by
Frickatt, fcfwik and Locm^uist € IfIII far the Illinois State Water twrwrnj. It
applied a finite difference aachad m %vmm4 water flm, #nut Vartl«la-in-a-
cell" tM "rande* walk" cracking tochaiooas, riiS|>t«tivmlf» to nodal solute
ceavacciao and 4i•pur*lout affacta. TUUli can aiaailata in either oat or two
diaaocioaa bach steady or aonataady Clone lo heterogenous aquifer* under water
table, trtiaan or leaky irtiun condition*, mi can handle exchange with jur»
fact water and evapocranaplration. to ability to liflrvilate tolute transport
while grouad water puafaga ia underway ii an Unportant eentidcracLoo £or
haaardoua weata studies. The code allawa specification of concentrations of
chaaical constituents to any segment of the nodel, ami tm accoaaodata 6>jlOCl i»Ut< trmautt mut &la»Tai«> Hadal: mi* aotfal wia iawlvpW by
cha 0.$. CaaVagical Sarny ta tlaalaca tba coatcntratiaaa of diaaolvad chaai
lo«ls ia m agmism- ac ap«cifi«i paiaea ssd timrnn «M*r
flow eaMitia«f (laaiiJbow and iradafeaaft, 1978). It ia aoaKiati eallad tba
mcs-mc aodal baaaaae it «aaa tba
-------
aolmce Mwrpcion mi dttmy rente lose* ftm U.S. DA hea umM I'SOC to
ltn4i«i of bu«f Hut Model a«aua*a only tr«««ri<
(».«. vwrtioii aad borianneal} 4l»ftriiM. It dna* m% ceoeideri leafi-*
tudinal dUptriloit coaatiteent attenuation by aaturnted ••II, twpcioa, or
cbaa/bio dngradeticmt raciurge dilueioiti or restriction of verticle spreading
hy smj I* permeability oaenriala.
Tha Wf node I baa not bean validated co data* It ux for regulatory purpoaea
lit i(M xriMi (rieUini and ica future etillty ii uncertain*
W/WUt tta Variable tkiekneea Trwleatt ftmid tltm artil «ea dmlepod
for application ce mltiaiolIaBr irwalMCtr ayateaa vitb cranefere undnr
pveaaere bacnaan equifere, or m afuftr end a eurfece mtar body, vtt «m
developed by lettella Pacific RbrtJeneae laberetoriea far cbd of
Caergy, uicb particular confederation of condition* «ffactleg Muca diapoaal
ac ch* Haoford Siu In Wathintcofi Cllfp ae «i.t 19711 Kaiaanauar, 1971) and
waa rteancly updacad by cha Elaccric Tovar taaaarcb Inaeicuea (Bond, I98l>.
VTT l« • flaotibla two diaMnaional nodal uaiag ooa finiea dlffaraoc* ncthod lie
tcaady ccac* coadlcioaa and a aacond varaion for cranaiant coodicioau* TIm
.E». Tha VTT la writton in rottlA*
for tha CDC tyacaau
1. gaakina atodaiai rankiag atodala and tka
«>atar related portTo5to?"Sre ganeral waaca aita r«ki^ «odal»,
aldMMtgb not tram>d«ecer treneport andala, aigiit 1m vaL«abl« in ««aaa
potestial hmr4M« oaaea eMtrceftcies. Tk»y any p^rtfc the anlp
ianadlataly aveilabia «ey to atciaata potaeelal groundwater prol»l«ABv if tbe
uaa of awirh—iitieal aodela mc wait until wabatnacial taflpling «ad aealyaia
ara momflmtM, to pvorida occcbmx^ data* tmntt»g aadal* CMli frovid« sa«
baat nathod a^ilabia far acoping tbe problen aad davelopuog aaMrgancy
raaponao plana CKaaly, lt83)«
71-19
-------
il« or* tc«a4Ar4i<*i vchaees 4ick pcnit its
oNtrnr to jadt* «Pfrt*Uwt«ly tito btsard* par«Mat«4 If • particular problM
•Cm, fctaaplaa iaclute tb* ««rra( fninni e( tba LaCrgad Mill (Lrtroii IftO
Ml Pottyjot* «C mi. lf«l|» tto WIM • f*m. (OTA, IfII) mi tbo JIB modml
flull at al., 19*0). 4 brltf #4»e«#ti«a of tba LoCrtnd aodol mil help to
isiCvilMi dig group.
Tht LoCraad rwklsf ayitM it baaed ot» four gaologle and hfrdbrologic
character!it£ca:
I. Piscaaca Co a water fnffily;
2« Dapcb to «ator ullat
3. Hydraulic graiiMC) nd
4. PanMabilUy-Mrpcioa (iadlc*ca4 bjr cbft goolatlc totting).
Cacb of tboto charactorUiico U avaUuUorf aad rated kioHIbi to lUMtaH
•calti* for the water supply, vatir tabla, md peraaability-oorption char-
acteristic*, • ranking •«•!« of 0 to t ie uaedl for tbo hfdtmlie gradient, •
actio of 0 co 3 t§ employed. Tho ranking ocala for distance to a wacez iapply
raoioa f roa 0 for dlatancea In ixean of 6,200 ft (I 9 I#* ditcooeo* of up to
41 It. Tbo depth of eh* voter tabl# scale reage* from Q lor daptba la eacsaa
of 200 ft to t for depths of 0 ft (earfaca mtcar), Ihm paraaabllity-'oorptloa
•Ml« Is asre eaa^lai earf dopocda oo tbo type of aoil aod tho thicknui of
uacoaaelldatad aatteriala am badroclu For Wse wtar (alia |ra»ta cssplifwi for sssel^
MCiul Bodolt. Tt» awdost mm rw(oiroMuac« of rmkim§ mmis mtJui tern it mm
mtlmr Lo »« fiald than «ftalfticAl or BMsrical MCteMCieai andola. ;me«i
arov lMwa«ar, diudvanta|
-------
C. Surface tfacar Models
Studies of environmental c®ota»ln«nti in mrficc witirs are (ent-
eral ly divided into two c«teg»
The aodel most widely used to compute fate, persistence* and e*-
posure of pollutants ia freshwater eeotTictai is EXftltS (txposure Analysis
Modeling System) which waa developed by EPA'a Environmental Beaearah Labo-
ratory at Athens, Caergia. Other models which have bees used for surface
water pollutant dispersion studies Include) TOXlVASPf 81IAJ and QtJAL.
S3CIH5 is designed for the rapid screening and evaluation of the
behavior of synthetic organic chemicals in aquatic environments (Burns et al.»
Ifii, 1992)• The node I requires three types of data* chemical* environ-
mental , and loading rates. Chemical data requirements include physical con-
stants {e.g., molecular weight, solubility) an4 parameters Co c Dispute trans-
formetions such as photolysis, hydrolysis, oxidation, and biotransformation.
CXAKS is interactively linked to a data base of properties of chemicals.
Sktvireeoentei data requirements include system geoaictry, hydrology, and
¦eteorology, and definition of dispersive and advective pathveys for both
water end sediments* EXAK5 caa peetition pollutants amoog five v*lenee states
and three physical forme (adiflrbadi bioaorbed, and diaaolvetd). Pollutant
loading* can b« specified » poise eource, aottpoint source, dry fallout or
aerial drift , atmospheric 'aistetf-, ami §T9mai»sx.mw seepage. Processes such as
photolysis, hydrolysis, oxidation, volatilisation, and biotransformation are
aimuLeted as pseudo first-order kinetie reactions. Second-order effects can
be introduced also. EXAMS does not handle dynamic (transient!) flow condi-
tions, its hydraulic end sedimentation algorithms are limited, and it ia
liaited to organic chemicals. Field validation results have been reported
((James, 1982). A version of cha EXAMS model is available for use on (tiero*-
eomputers aaapetible vith IBM PC XT/AT* systems.
TOIXMiflP is a kinetic subroutine for the Water Analysis Simulation
Program
-------
are simplified by adding the paeudo fLrac-ocder races due to phocolysis,
QEidacion, biolysis, hyd-rolysi*, end voletit ieation to yield * total degra-
dation rat#. TOXIHaSP calculate* che dissolved, jorbed. «o4 biosarbed fric-
tion# of j neutral chemical only. It hes the cepability to simulate dynamic
at vail aa $tact xautlMdi. As Lo EXAMS, Laboratory «n4 licenturc
values Em cht chenical characteristics of a ootspound My be mjȤ4. WASPJ is
evailabla Cor use on IBM PC/XT and compatible «ticrrocompucer s.
SLM (or SydroQoal Model) it a simplified «od#l developed by
RydroQual, Inc. and the Chemical Kanufacctirari Associacioa lor che of
the fact of partitioning chemical* in lakes and streams (DiToto «e al. r
1992). Total degredation race of a substance is estimated by adding the
pseudo first-order rttei due to cht kinetic prQccsiei, u in TOXIVASP. Mech-
anises considered are settling, resuspeneion, and diffusion. The adsorption-
desorptien reacrioa is assumed co be tc equilibrium. The fraction of the
chaotic*! maas chat it either dissolved or adsorbed co pircicuUcei in deter-
mined by tht mass of cht adsorbing solid and che partition coefficient in the
water column and La tht sediaunc. TrtACM.Bc of colutta-aediiienc interactions
is a scrong feature of tht model. All algebraic and differential equations
are expressed Id a fora as tiitplified aa possible. Closed form solutions,
baaed on Che conservation of oiasa, are presented and operated qd. a mini-
computer linked inco graphics output devices.
QtfAL was designed to sisulate special and temporal variations of
dissolved Oxygen, biological oxygen demand, up co chree cooservaclv* minerals
and temperature in stream end caiuil syatems under steady-*cate flow condi-
tions. The model is actually a sec of ioterrelaced quality Muting spx(el?
baaed on ecspiric end kinetic considerations. Developed originally aa QUAL-I
in 1970 by che Texas U*ter Development Board chose chait
£o€Uj eloose entirely on tbm quantity of runoff Mater; and (2) these chat
Vt-22
-------
emphasize inforvat i-on on tha qaalicy of runoff wacer, i.e., phe eaviron*aflCal
eancaaiaant*, CS«arly all vacarahad isodela aimuUte erosion rates.) Runoff
models 4re sawtiMt clessifiad also According to operst.iootl chirtctafisucii
including geometrical representation of cba dr«lo«{< baiint tenrporai ripra-
sestacion of flew, «ai physical process rapreaeatation, each of which has sub-*
sets ti noted belcwti
• Cao
-------
tophi setcatioc of chaaical procaaaaa iiouUcH vary widely aocng nodela. K*ny
twcer-ahad models, htwivif, incorporate Che saaua or siailar hydrologic
algorithms.
!«¦ of the ap-pr®»«bni«t u»«4 for bydroiagical prt»a«»«ea involve
p«t dependent upon a number -of factors* particle cohesivenesii
organic matter content of the soil profile; raiofAll intensity; vegetative
cover I slop* gradient* slope length; aai aail aultural practices (Vischmeier
acui Smith,
Hwy eh«»ical processes haw an impact on contaminant behavior
because of tba coasplexity of aoil chemistry and tha broad range of contam-
inants associated vith runoff. Sana procaaaaa frequestly «ddraaaed by opcra-
tioaal acxtela
• iudaorptioo of contamiBABC to aoiL ptrttcUs hwtmlly
characterised by a pa.reicioaiog coef ficieat).
« Solubility of contaaunanta and of aoil coarponenca.
« Volatility of organic tanpouivda («ay depend on attaoro-
logicel eonditiona)*
liodtgr«4«Ciaa of eoacaauaanct by foil rt«eoo*f*al mm.
• Soil pH (Lafluaaca* eomtaadaaac aolubilieiaa).
VI- 2*
-------
• Kedo* pocntiil a£ t#|| solution (influencea awbility of
mcaIiK
tfeaee thtii aodela require l^4roli|Ut eeteorologicel,
wetersh«d data, together *ieh the dueaicel and di»trlbutio«al properties
of the cantaaioaeca. Sevarel catalogs Mid handbook* exiic chat provida noene
of evaluating w«c«rihed eadais m vol i «i other environmental «e4el« (k»ti
m* lo«n, Ifil; CPA, If MM.
b. laaof fluent I tf aioditii Five nodeIt «ra diaewaaed Mlovi
leciaoel ronnie* Ihe si#lut of all hydro logic atodelt |» cte
Ketieael fomtuTTTnt [BtroiiaeH nearly 100 yuri ago. The simplicity of
till method
-------
I
Ml
>
a
3
3
as
a
I
&
g
3
tn£
I
U
ad
-J
at)
(A
I
94
sJ
1
» «
v J m 3
5 =i 52
3= 5 I*-
iittis
3 SB |g S as »#
«
5 «
*• O
^ ^ *
* . St
iS-J
w 1 »• A
I ®3 a ®
•m m mm
SI 3-s?
is
3 5 S'Sl
111 ill!
-5
52
-ksn -
* 2* - JS »*JS
N f « M M
**> m m m <**
£1 K ^ 4
«¦ * . m ¦£
4b %*
s#
SLW ^
— ¦ ^
2s«
J
^ . e
» 1 4
• H *i
s
1.1 I
*• • §
u f® wa
«# U •»
W M •
^ 2
«
^ H m
¦ 1 y
>% ,# -N
®
S 1 2
«3 •
«
w f ^
s 3b
•» «
1# € «s
"3 Z?
JR mi U
« «®
35 3
: a ;
a — ^
K 41
U
Z 5
« «
« m
m +
^ i m
lit
m 9+ U
sm «S 0m
IU
^ ^
sis
t
^ C
zz t
?!i
i.«. 2
N •% «*
1 4 k
111
M @ 1 •
He!
# «® «
# * % •
« *•
» 9»
* 4 «
A W •
I «4««S
— m m
12 « 2 «
%8 «# « 3
3~frS ?
1#
i2i
its
m> ^ m
m a#
» •
<«# 0^
s 11
II
II
1
««£
1
*:
m m
m s
*»
?!
S3
w ^
§ »
a. ¦;
i?i|j
^ i»
i3J 31
5ISS2
I J
Jiil
*$1
all| a
l?!!s
i i
54
J*
*•! 31
111 h
mm m m m.
j-
ti
•*»
ii
3
5
I
I
:
r
s
? K
y *a «*
: 8-
# g »s
% C pi
» 4 *st
5 jl
i*iJ
V %• •»
«s W
#* i «
M Q
4 * H
te. <«£
XS|
r«.a
. ^.
^ y
3 S3
fc 53
|I
-I-
i -¦
<» i h
* cr
Hi
a is m
Is i
8 2 2
»r
3 -
« «
M <& «
<«i i
s:-
: =
l"!
• 2 *£
m -m
* j*
n-2$
-------
» w
O 'r* D
I
s.
I
I
u r4 - u u c
8 m w i °
ft c w v ft.
¦d o n § c p
K — ¦« fl •- II
- " \J -r»
: - s - -s
5«?i .
fat-?
I"!-
S • s »
n
im a
•a-**!'
? O JO fc A
SSig5i
21""lis
P * £ «#^2
Sla'3'.
** o *8. 5 m St •—•
¦ •• i |L>^ m • ¦ n
9 w -«i W ¦ JC *4 ¦ |«
w « v o -« "< • 3 »
¦ u u a lu "O V t>
O ^ U ^
B o ¦ it g
•m a — u e
8.S
Isilgls
~4 5 m •
* • »i 4i m m i ~
mM C JS Jl •
•T« C G 0 o O* i*
C *
2 * -- -
- m . ,
4.
«&,
^ y'J ?! u « « 9 « o •
1 5 S S S » W Tf X «- .4 o
|3-Jo^-lsgs§s
i«v2& u „ u *
5 3 2.5 - « - " - «•
s
8
J
| if 5 fci a
• 6 » O « a
•1
iff-
3-ifslf-
VJtiPsi.Mi
& 5 •. * .t K h& S *
"1:11 ti
j«S"»s^s"2.sS.5;
1*3*3 I;|5f!|i=
•*5 ^ ?r « 5 ® *2 1 - ® *
ti <1 ^ «»* u ® "• m —*
¦ « n *| B E ° h f> u H •
J 5 3 e SS.2S3S ss=.|
ft -M ^ u V H ^ _ 14
**$2 Z S«£ 8 8 "*$2g
8 II 8 iksn H'S-g-C IS*
p-I 41 U W I »¦* «
DBw4<)aac4
1 1* « u a h«3
5Se
« In
• M «* ®»
—' 4* 5 ®
m:
Q O «
"» $ £ 2
§ "J
lu ^ . KM L
S".£SSS"
M 5 " a U • •»
•2^ 15 • =
- s i? « ? 2
5S;S£!i»
s s |a si
^ 6 ft ^ **
e1 I « * *
MuV3£<5*l*
«p«w
1
s
u
t 2
85 i#
„slsS-u
1 § 8 1 3 • s
t*» «,2 * -S
-» ^2 -
f* §
El£,sl
- .issis
2 • X S « ^ M
g2Rlf"i
_ <# -5 fl 5
1 «
6 o
¦"* —. *» -* I
• 1*2 • E S
|J3„S~
illnl
"< - H 1''*
y *i « 8
» 2 2*2 04« 5
a • ^ 5 -j
lilllH
I
5
-------
stream* The aedel uses mathematical equations thee repneneot the physical
processes important %o oonpoioc source pollution. The MPS stodel should be
calibrated whenever it It applied to • new vitiftM. 1otiever, moat UPS model
ptraaitiri arc specific by physical wecersbed oharaaceriatics and do not
requi ra calibration (Danigian and C«wfor€» 1976).
Unified transport Model
-------
I
§ ;:n
w 6 i#
¦ o»6<*a«ut.ln-4
> -~ *» m F» -«
a.
ik
m
£ " S"m 1 « ¦ F
S3 3j?2„»?i s
« « v -j ¦ i 11
3 - " 1 2 id - il£
E?"8*»li=5 -
S2|Sj«J?.
g,
s'&flZi**
i-se js f #s^s-
3|| «S.I .*
J!
o
•t
Sill.. 153 u ?
5* • w •« o"*4*- m • 3
i|iii«! *«{;
SfcCJi bi«l»8
c s s iJ I ^ • -4
s83|iikl:^g
- - - I >4 f l» U ¦ u
n2 m8 *5
I5§fs^t as
tf
IS
I 2
u
1 6
•3 w
" i
&*$ 11
M V
¦ 3
"2 •
« g
-« u
m
El
Is
\b
lf
m 9
«
J 23
*3
ait
ajl
«
4
I
O '
h> 4
I
-------
TABLE VI-6
CAPABILITIES OF IBVIEMM MR WOOBLS*
iaa
feotr
in—mm—mm—isr-wfin fir^rgK^wroi m namttw: uro
Swt«
hinl Mircc:
S(B|U
IkiUiplt
Are* Mure*
lis* toarcv
Voluw retire*
ifciri*l«tsg
MUtinl fl»w«ttiw
SnmuiigiuT
lieli "ill#
TuituiMt kpottliti
Mtiboul
?«rli Itetrlx larUdcf aily iinaiii ni refvaMl i>4tl Eitf|«rin
ATM ¦ 4Mnlirlt Innyirt Batel. PTD16 = teiai Source Mr!
Cttl(QC) ¦ CUMtl]f|» Itifnim Mflt. PTPLlf = Mat Stent CiiifiH Plat Ifcxicl.
CUTIII ¦ SlRflc Stare* Mil. SMS - 6ttuti«o flu« thltipli E«im Ait Qwbty AJgontb*
lie ¦ litiilnil Swrt C*ap)u. icn-i - r«Mi CliMtolsfiril Multl. VcraiM 2
rtfTlR ¦ Sattipfs hiitl fioutriia Dliftftiei ?£M-I s Thii IfiMdiC Ho4»t, Vtrti^n ft.
HI * Pint, tout, line tourct V4JULEV * Ciuniui Mm* Diiftrfiit il|«titlai
I Surrai iti crtttcd ¦* M'bfii(4.
It * rartl oiJy; U* artuuri *ta)y; JC * tartl *mi wrbip
-------
Point Soarca Modal (WttM)i PTMX is a historic*! me4el that h*i
bHn »up«r^^sr^^nrTiirth5rmio»ui>.
Point Source C»ng»i» The PTPLtf code I ii ao
iaprfivtd vtfiifin of PTKAX. Applied t» 4 single polat source, Ik li utefai as
4 screening auadal for qtticklf eacicsaciag eaxirtun l-tir ground-Level ccnctne ra-
ti oas and tha iiicanc# to atasuDun concent rati one lor varying combination* of
wind speed and stability class.
Point Source Model (PTWTF}t PTKTP is a historical model thac haa
Mm superceded by PAL or by using MPTEX or 1SCS1 with user selected meteor-
ology.
Polotj Anii Line Source Atgorithn (FMJt The PAL model is aft
enhanced ^teesttMce sboct-cersi concen-
trations (1 co 2* hr) Era* point sources, area sources, two cypes of Lina
tourcci, and two types of curved path sources. The sigaricha ia not intended
far application CO entire urban treM* It I* suicable lor small-scale
analysis of the laipect of a jingle Laduatrial facility or for evaluation of
fugitive dust sources®
T*»aa Eftiaodic Model, Versions (TEM-8)? ltM-8 vi» developed as an
alternative co the PtHTP and PAL (sodel*. tt^nay b« used to escimece ground-
level short-tern (10 skin co 24 Hr) concentrations £r«n point aod area sources
to lite or gently rolling terrain. fOf-l differs fro* other nodal* • in the
UMAMAJP aerie* la its ability to account for variability in the horizontal
dispersion coefficients aa a function of averaging tine and stability. Aa an
additional capability, it can yield more realistic treatments of pluoe
restriction beneach an elevated inversion*
Taaii Cliwatolcgical Model. Version 2 (TCH-l)t TCH-2
-------
may be jppli«d in rural, a ltd urban feelings, and wilt consider uneven terrain
4i long 11 receptor eLevitions arc roc located above tbm lowest stack, height.
In addition to providing ulintUi of naxinrua hourly coo cent rat ion foe each
4»y and 24-br average a« CtSTTO produces suawary tables which include annual
¦lean concentrations* and highest and second highest l-» V-, and 24-hr con-
centrations.
Caessian Pluarsg Kultlpla S» pluaa rise due to
sootBEiifl and buoyancy as a function of downwind distance; (a) tine- dependent
exponential decay of the pollutant; and (d) the oapability to siaulate gravl-
tational settling and dry deposition. ISC is the only nodal in the UMAHAP
series that incorporates the latter feature. ISCLT is used as a cooponant in
tf*8« Inhalation Exposure Modeling (IBM) lystsa to link concentration modeling
capebilities with eanpotertsad ¦*t«o*0lotid«l and population data bases (CPA,
I9tta>.
*1-38
-------
Atmospheric Transport Model buiIding veke effects* and (c) terrain adJruscauuaes. Representation of
area and line sources is not as sophisticated ai ia sany other UKAMAP
-------
estistate concentr*Cio»B for relatively stable criteria pollutants such as
suspended particulates) are capable of onlf a fairly simplistic representation
of chef? processesT with the more r«lin«d aodslt representing plume depletion
is * function of a half-life tew and erav*l tine downwind. The AIM vodel
differs fro* the most r«£io*«d UVAKA? model* In that it has che analytical
capability to consider w«t deposition. turbulent deposition, and particle
resuepension. in aost cases tbe lack, of daca available to specify deposition
tera* Haiti tbe usefulness ol theme h«c»rfi.
Source Type(a): Source Cypa and source-receptor configuration* are
probably the scat flexibly represented aspect of the UVAKAJ? uriM. Kodela
arc available Co consider: (a) a ainfle or United number of peine sources»
(b) multiple point sources, (c) multiple point and area sources, and (d) nor*
specialised representations including line and voluste sources. In general,
when the dimensions of a stationary source arc mall coopered to the distance
«C vhich concentrations are Co be estistt«4» tba source amy be created as a
geometric point. Hazardous waste treatnent/diapotal facilities subject to
BCBla vlll generally fit this criteria, although a land surface disposal
activity or a v«ry large landfill might require treatment as area sources,
(la practice* multiple close sources that emit small amounts of criteria
pollutants are often aggregated and treated as uniformly emitting area sources
since their amber and highly variable emission rates preclude treatment as
Individual sources.)
Ctsirsglile and LeadHIee_Chjrrect:ariatics: Several geographical and
l*nd-use characteristics should ^be considered in selecting a moiEiL froas the
l/HAMAP JGriea (SPA, I9B|)« K ftimtf ceoaidericirc ia vtuitar tftt nrta of
Interest is predominantly urban or rural • So sua of the models are applicable
only to rural areasr others are appropriate ooly for urban areas* Soma of the
(•ore refined models are both urban and rural (options with the primary dif-
ferences Ln the two nodes being (a) the choice of dispersion coefficients used
to characterize the plume behavior, and (b) the treatment of meteorological
paranaters such as atmospheric stability and (sizing blights)• Another key
geographical consideration involves study area terrain. Many current disper-
sion models are based on empirical data that vera collected in flat, open
terrain, and are ehaa moat applicable to these conditions. Several of che
models do attempt to account for the effects of uneven terrain by considering
differences betveme source a&4 recejtor ground*Lev«l elevations. The VALLEf
(»odel has been designated as applicable for screening analyses in complex
terrain, although its treatment o£ plume impact ige in coeiples terrain baa been
videly oricitUed, and thus its applicability must b« considered limited, A
¦ore microtcale land-use consideration that my be Important in selecting a
model La the presence of buildings or other obstacles adjacent to the
source. The eore rt fined UHAKA? nodels have limited capability to treat tba
wake effects arented by such features»
Pace Kagnlremeiitsi All nodeis require information on the sources to
t» eodeled and the meteorological conditions to be examined (Holxvorth, 197Z).
rot- area sources, the eodels require a description of the siae of the area and
the emission rate* In many applications, source data can be obtained as
eatisaion factors frosi the "Compilation of Air Pollutant Emission factors" (Af-
42 as renrised) (EPA, 1943b)» For peine sources (e.g.* an incinerator stack).
Vl-34
-------
cte models repair# the emission race of one or ware pollutants, the height of
release, che i§i«»#E#r of the stack, nhe «*it velocity of che |«fci, and the
teeperacure si the galea as th«f axe reletsed. tn some applications,, a flow
race nay be required Intttowl of stack dierteter and velocity 1 aniltipie point
no^ilt require d location For «tcb »curce. IE dovttiMiibi is considered, tie
dinensioas of an aasociaced building will be required.
Haay roue in* application* uae data from nearby locations, sucb *«
airportsp National Weather Service atationa, and nilitary installations to
represent thai atmospheric conditions for the area of interest. The primary
source for surface and upper air meteorological data li the National Climatic
Centnr (Asheville, Itorth Carolina). tonfteru ¦o4els use a climatologiesI
juaaary of atmospheric conditions, comdMnly referred to as « "STAEl" deck. The
STAR deck luimritei tttedrological cODdicieoi ifl earns of joint frequency
di• trlbutioiis of wind speed, stability class, «od wind direction> This infor-
mation has been devel far otpy locicioai ia the United States and it also
available from the National Climatic C«oter.
Averaging Periods Averting period probably will be one of che
major criteria used id model selection. The baaio Causai*n equation calcu-
late! a short-tens concentration representative of about 10 min to 1 hr.
Hourly data are che basic e®l«®l»cioa input (or ihorc-terni model*. The con-
centration estimates are often averaged for J-> I-, aad 24-hr periods. Gen-
erally, it is alio possible to use a full Booth, season* or year of hourly
values to produce long-term estimates. As an inexpensive alternative, iaae of
the models use an integrated fona of Che basic Gaussian equation aad a
itatiitical a unary of Uie atecoraUiiul data CSTM tabaUei«n«} for rapid
calculation of annual averages• Tkete models are referred to as "leng^teno"
or "cliaatoiogicei" node la and are typically uaed to calculate 1-, 3-"month, or
annual averages.
S. Accuracy wi limitations of air Models: Three major factors
influence the accuracy of air quality fimuLatloti models according to the
Anencan Meteorological Society (ArtS, 1978, 1981). These are: (a) the capa-
bility of the algorithm to reproduce the important physical and chemical
processes; (b) Che quality of the emsrsioei data; and (c> the quality or appro-
priateness of the a*c*oralogical data. The overall accuracy of the Gaussian
dispersion model will be dependent orpoo the specific application. Tha
Gaussian model will perfowe ba
-------
M«c*oroidgicf 1 dace t^oicEBaocs are ju^»c«eti»l far diffusion
modeling feom poiac aourcaa (AM3, IJflO). With » coaplara range of roeCBoro-
logical maaaureAeoct and corceipeMiagly accurate aeiaaions dat®, trua coo-
cencratiana for the simple dispersion cut can probably be iicuuctd to within
±401 (AHS, 1978). Complicating Ititufni in the specific applic«i 10 co 20 km).
Significant improvements io dispersion Modeling will require nrore direct
observational knowledge of dispersion under the a a condition*. Hodel uaara
should be aware that cha capabilities o£ the currant UNAHAF aariaa co repre-
sent these features arc baaad on a few special caaa studies.
Perhaps Cha vest difficult ai tuatioa to model wick any accuracy
would Involve a LovHevei «cci4eot«l rftlaui of pa il u tacit i. Iirii wtctr-
taintias in the aouroe emission rates probably would be Cha dominant factor
determining accuracy of cha coanentratioa nciaatei. l« this case, the time
and space distribution of concaocratiotte would be as important as overall
concentration levela* Ordar of laagoitude estinaees of concentration within
the area covered by tba effluent
-------
Ccoiogical eedels can be cl»iifit4 in two c&eegorLaa aaoordiog Co
the ltin
-------
shich cm ioclude hydroLogic *4 oiecaaroiogic data, »it«-»p«cific rsedia
properties, and chemical-specific triaifar partitcira.
Screenim amtelt - These nodels tr> used when the data Imsb on the
chaaucal and its receiving environment «r« iudcquttt. The of theaa
an partitioning models ctue iniwe « pollutant is parcitlooed (usually at
¦qailibriuml between environmental median eaas^las .are ESPAtT, discussed
beloM> and cvo models developed at Arthur 0. Little (Lyman, 1981). A wore
cocaplax node I is T0X-5Clttlft also 4i*cw*««4 below.
Th* Environmental Partitioning model, EifMf, umi fugtcity equa~
tiona to estimate ratios of cheaical concentration in air, water* and soil
under equilibrium or dynamic (intermedia transport a ad transformation)
conditions (Wood ec al«, 1981)# The nodal providea a firsc~lcvel screening
analysis of a chemical and Its parcitiooing in the environment. Ic is an
interactive or batch model with tba ability to receive & miaimm set of input
data im the physical chemical proportion of a compound and calculate other
pmpirciai required using correlations. It also ha a th« ability Co receive
iapuc of degradation And transfer races or to supply default values. Tha
interactive tsode allows aenaitivity analyses of various parameters of a
substance to be perforated quickly and easily.
rox5Ctt£EM is a screening level mulcinedie Bedel, developed ec Oak
Bidge Matio&al Laboratory for the U.S. EPA, Of flee of Tomic Substances, to
assess tha potential face of tonic «Juecaieei« released to air, surface water,
or soil • The no del is Mid to ba simple ia mcux« and is intended to be used
m a terttnioj teviet co identify duurixali dm in imlikaly to poi«
eoviroiuneacal probleas even ua4er conservative assumptions* It integrates
8ES01L with other analytical modela to provide a profile of ohanical concan-
tracions in nultiple audi* surrounding m mlmm site* It accounts for in-
termedia transfer by volatilisation, disposition, runoff r and percolaeion
(McDowell-toyer and Hetrick, 1W4). The prograa, written ia FORTRAN Cot the
0K€ VAX 11/780 computer, is available on nainetic tape from the National
technical Information Service Conputer Products (Kaufnuun and tloarsoa,
1983b).
Sice-specific »odels - These oo4«ls include two version* of the HTM
and CBNS whieh can be used in either screening or site-specific *ppLi««-
tiona. Tha Uftlflot Trsoarort Model, HTM, is actual Ly a group of bo dels from
which a elected aodel* are used ia series, the output of one serving as the
lopite of tha next, to provide a multimedia nodal for a given scvidy problem
(Hedden, 1984). The UQl Model requires data on temporeture aod dewpoiot,
hourly precipitation, daily wind spaed* soil and vegetation characteristics,
and water shad topogray^ky. Lq additioa to daca on tha octal being node Led. The
UTM taodel raquima ilt«*$pecilie application* l/TH waa liaited in its original
version to trace taetal* (Muaro, 1976), buc has recently been eodified (aa l/TH®
TOJC) to acaosBodata argxtiies (Patterson et al«, 1964). UTM-TOX is said to be
capable of addraaaing pollucaaca froai poiac and noopoine aour«es, dispersion
in different sedia, wee and dry deposition* surface and atreaa flowa, tail
aroeiott, and percolation through tha aoiL to grooDdwatars or surfaoa wacera
CTravU, 1913 >•
VX-31
-------
The Graphical Exposure Model iiift System* GEMS? ia a network of nodals
tjjcnbltd by CPA J Office of Toxic Substances for 4$^es)ing patent i*L human
exposures CO taiic Subccancet frcrm product ion of proposed new chemicals or
iron existing production. The syicea can Integrate several data bases and
models. GEK5 cjn estimate, if necaitary. the physic*!, and envt ronmental
properties of » chc«ic»L from Its structure-properties data base. and can
uc iLixe si Ce-speci fic daca on environmental charatc«ri iticji co esci«uce
transport and trtoiformat ion races. Nodal J accessed through CEMS iaclud«:
ATM and ISC for point and area air emLtaion sources; 20XMOD for diffuse urban
emissions to air; EXAMS for surface waters; SESOIL for chat unsaturated soil
zone; AT123D and 8WIP for grcundvacer transport; and ENPART and 70XSCUEEN for
lntonijedia and multimedia phenomena. CEK5 can ucil Lze locacional data for i
ralaase site and access 1980 U.S. Census Bureau populatLon/decnographic daca to
predict exposure concentrations and numbers of people exposed Co a i r poLlut-
aact via ATX-SECPOP. GDIS also contains three subsystems far es t imat ion ~£
phyticochenical properties (including bioconcsntration factors) for chertr-
icals: CMEMEST (d«velop«taoc by A. 0. Little Company J » ALTTOCBEM, and CL0CP3.
¥1-39
-------
lefagancas eo Chapter VI
Ahlatnm, S. V*, Mid H. P. PooCe. Multicoispaneac Max Transport-Discrete
Parctl Random Walk Model Theory and Humeric«L latpltawncation. BNVL-212 7,
BactaLle Pacific Norckwest Labocatori.es, Richland, UA. TsH.
Ahlitroa, S. V., B. P. Poote, K. C. irnitci, C. I* Cole, and &. J- S«m«i
ffaaci-CotapoawiC. Has* Tr«n*port Model: Tttory and Jfawericiil Laplamenta-
tion (Discret«-?«reel-Random-Walk. Version). &NVt,-2l27, Battalia Pacific
Northwest LaboncortM, Richland, HA. H?7 (ciced III EM 1962a).
US. Accuracy of Models* A foticloo P*p*r of che AHS Cossaittee on
Atmospheric Turbulence mod Oiffusioa. Bol let-in of the Aaifitta Heceor-
otoiiMl Society MCI) 1023-16, 1978.
4MB. On-Site Kateorolotical le^uireattaca to Characterise Diffusion fro* Paint
Sourcei. Proceedings from a Workshop Held La talaigh« NC, January 15-1?t
1980. Aaerieao Meteorological Society, Boston, HA. 1980.
AMS. Air Quality Modeling and the Clean Air Act: Recommendations to EM do
Dispersion Modeling for Regulatory Applications. Mutricin Meteorological
Society, Boston, KA. 1911: NTIS, PB6j*-l06237.
Sachnat, f®, J. Bredahoeft, B. Andrew*t D. Hols* and S, Sebastian. Croupd-
water Management> Tb« Use of Mmnerical Models* Hater Resources Mono-
graph American Geophysical Qoicm, Washington, DC If SO.
Barnwell, f. 0,, Ji, An Overview of MtffP, « Simulation Modal for Chcnical
transport and Aquatic lisk Assessxs*nc> U.S. Environmental Protection
Agency, Aclwni, CA. I9S1»
Baste, D. J. and B. t. Bower, eds. tomWaioit fletural iyn«n. Resources for
the Future * Inc. Washington, DC» 1982.
Battalia Pacific Bortbwwst laboratory. CFEST/UMSAT-1D Model, Personal Contact
with C« Unsaid, Richland, WA. 1982-
Bleu, C. E. Environmental System Analysis: An Overview. pp. 1-20 in
Ba'wIrqBiiinitiil E*poattr«_ from Chamjcala. Vol. II. W. Beely and C. E.
ftauVVJs •* CtC Press, Boca Raton, 1983.
Sonasountas, M. and J. Vagner. SESOfLt A Seasonal Soil Compartment Modal.
Arthur D. Little, Inc.• Cambridge, HA. Prepared for Office of Toxic
Substances, U.S. Environmental Protection Agency. Washington, DC.
Ml.
Bond, f. y»» C. A. Mewbili, and P. J« Cut kn edit. Variable thickness Traimiat
Crowodwater flow ModelHJser's HwwL CS~£0ll, Electric Power Research
Institute, Palo Alto, CA* lWi.~
¥1-4#
-------
loud, f. tf», C<. Kw Colt, and §» J« Cutknecht. Unj»tar#tcJ Croundwattr Pljty
Model CVHSAtlD) Computer Ooda^ "Manual. CS-2434-CQt, Electric Paver
Research Institute. Paio Alto, CA. 19B2.
lowtr*, J. fM J, R. Bjorklundi and C. 8. Chancy# Industrial Source Coaplex
(ISC) Dispersion Model User1 a Guide, Voluise I. EPA-450/4-79-O3Q, U.S.
JEoviromtentil Protection Agency, Baseafoh Triau^l^ Park, tVC. 1979.
BPNL. Ceohyrfrochaoiieel Models for Solute iigraciDB® Vol. 1, Process
B««c*iftloo a*td Cooputer Code Selection. C. T. fiocai«J» J. R. Korrey,
and J® E. logtri, Volonc 1, Preliaiury Evaluation of Selected Conpute*
Codes (C. I. Kincaid, 1. 1. Marxnf, S, i. Yabtuelii, A- R. felsy, end J.
6. Rogers). Final Report, Battalle, Pacific Laboratories co
Electric Power Research Institute (CPU EA-3417, V® 1), February 1984.
(eWl EA-3417, V, 2), Koveaber 19SA.
IrQwu, H* G. , H» R. FatCersoni and T« J. Swerski. Formulae ions of the
Pfcyaicochastical Processes ia eh* ORKL Unified Transport Model Cor Toai-
cancs (UHt-TOX). Intaria Report. QRML/TM-8013* Oak Ridge National
Laboratory. Oak Ridge. TM. 198-2.
Buab, A. C., C. 1. MclTec, J. M. Sever and, J. C. (falepeska, J. I. Drover, and
&« C. Way. Contaninwia in Groundwater: Assessment of Containment and
Restoration Opticas, Conference on Management of Uncontrolled Hazardous
Hatces, Hazardous Materials Control Research Institute, Washington, DC.
1914.
Sums, L. A.| D. H. Clioe, and 1. R. Laasiter. Exposure Analysis Modeling
Systeaui { EXAMS): User Haaual and Syseas Docuaantation. U.S. Environ-
oantal Protection Agency Ehviroomecttal tesaarch Laboratory, Athens, GA.
1981.
Burns* L., 1. Lassitar, and D. Clinau Documentation for the Exposure As-
sessment Modeling Systesi (EZAH9). EPA 600/3-82-023. U.S. Environmental
Protection Agency Environmental Research Laboratory! Ax beas, GA. 1942.
Callahan, H. A., K. W. Sliaek, V. W. Cabal, I. P. Kay, C. F. Fowler, J, a.
freed, P. Jennings, R. L. Durfea, P. C. Whilaore, 6» Maestri, W* 8.
Mabay, 1. X. Holt, and C. Gould. Water-Related Environmental Pite of 129
Priority Pollutants* EPA 4*0/4-79-029aAb, 0*S« Environmental Protection
Agency, Heshington, DC* I9?9»
Cartel, ft* f»# C. N. Saith, «a4 H. N. Lorbar. Pesticide Root Zone Model
(PRZH) -» A Tranaianc Hydraulic Model for Evaluating the...Hcvewent of
PeiilETcideii ^xES ihj~to5c Md*~Xo^£r toa^curic^T^^.nes^." User * s CaldeT
Office of Pesticide Prograsts, Uaaard Evaluation Divisiop* U.S* Epv\roc-
aancal Proteetion Aganay, Uc«hlngtoa, DC. 1982.
Corps of Eagiace**. Scoraga, Trc«ta«ac, Overflow Runoff Model, Geccrelieed
Coa^atar Prograsi. User's Manual • U.S. Corps of bgioaers, Hashing con,
DC. 1971.
ft-4i
-------
Billon, ft. T.f 1. 8. LaAC*, ao4 ft® ft. Pahva. Risk Methodology for Geologic
Disposal cf Radioactive Waste: The Sandia W««ca Iiolacion flow mi
frant poVc TjyfWT ?lo«l« I. Ntflf C7 €1-04 24, SapII a Nat i cna t La bo raco ri • »8
U.St Nuclear Regulatory €«*tts§i«nf Washington, DC* October 1978.
DiToro, D. K., J. J. Fitxpacrtck, and I. V. Thooaiui. Water Quality Analysis
Simulation Prograa (WASP) and Uo4.ml Verification PrograuB (MVP ^Documenta-
tion. U.S. Environmental Protection Agency( Ouluth, MM. 1981*
DiToro, i. M., D. J* O'Connor, 1. V, Thofnann, and J. P. Sc. John. Simplified
Nodel of the Fate of Partitioning Chenicals In Lakes and Streams, p. 65
in Hodciine cl» fatt of Chtwicala in ttw Aquatic Environment. K. L.
Dick-SOo, A. Hi Haiti, and J. Ctirut Jr., ids., Axra Arbor Science
Publishers, Xac. , HI* iff!.
Douaico, P. A. i and V. V. Palciautka*. Alternative Bouadariaa in Solid Waste
fkntiMMu Groundwater 20(3) 301-JII, Kay-June 1962.
Donigian, A. 1. Field Validation and Error Aqalysis of Chemical Tate Nod-
el a. p. 303 in Modeling the Face of Chenicals jg the Aquatic Eoviron-
ownt. K. L. Dickson, A. W. Kaki, and J. Cairns, Jr., ids., Aon Arbor
Science PublLihera, Inc., HI. 1911,
Doaigian, A. S.» Jr., and K. 1. Crawford. Modeling tfonpoint Pollution ftp®
the Land Surface. O.S. CnvironsMtntal Protection Agency, Athens, GA.
19 Hi MTXS 150-566.
Donigi«o, A. S., Jr. and K. 8. Be vis > Jr. User4 * Manual for Agricultural
lunoff Management (ABM) Model* EPA-M0f3-78-080• Environmental Research
Laboratoryf U.S. Environmental Protection Agency, Athena, CA. 1978,
tonigian, A. S., Jr., and, J. D. Dean. Monpaint Source Pollution Models for
Chenicals. pp. 75-105 La Envirqnaental Entpotw from Chemicals,
Vol. II. V. 1. Heel/ and C. £• 11 aiui4i7T~c5c Pr«Ag» Boca^tat^aTHTL•
1985.
Donigian, A. 5.. Jr., J. C. Lihoff, B. t. Biokpell, and J. L. Kittle, Jr.
Guide to the Application of the KydLrologic Simulation Prograa - FORTRAN
(ISPf). EPA-600/3-64-065- B.Jk Envirauaental Protection Agency, Athens,
CA. 1983.
Enfield, C. C., 2. r, Carsel, 8* 2. Cohen, T. Phan, and 0. K. Walters,
ipprutiaicinf Pollutant Transport Co Ground Water. Ground water 20(61
711-721, KoveariMtr-Oececbec 1981.
EPA. Statu Vater Management Model* Metcalf end Eddy, Inc., UniversiCf of
Florida and Hater Resources Cttgiaoers, Inc., far U.S. Environment el
Protection Agency. 1971,
EPA. Workbook foe Cosher? eon of Air Qua!ity foddls. 8PA-4J0/2-78~028a, 0AQP8
Guideline Series Mo» 1*2-097, U,S. Environmental Protection Agency,
Research Triangle Perk, (IC« If78.
-------
EPA, Guidelines and Methodology Eor the Preparation of Bealch Effect Asiiis-
aene Chapter* of ctoi* Aabieac W«c«kt Quality Criceria Documentj» Environ-
Modeling of ttaiardou* Waace In-
cinerate ra, FinaL Report. CPA Caocracc Mo. iS-0l-4322. Office of Solid
Waste, U.S. Environmental Protection Agency, Washington, DC# 1933a.
EPA. Af-42: Supplement No. 14 for Compilation of Air Pollutant Factors,
Third- Edition. Office of Air Quality Planning and Standards, U.S. Co-
TiroflMBC«l Protection Agency, Research Trltntle P*rt» HC» 1913b,
EPA. BC&4 liek/Cast Afielypi a Model: Pha«e III Be port. ICF» lac,,
VaelUftgtoa, DC, Prepared for cbe ftfflet of Solid Vaate, U.S. Environ-
anial Protection Agency. March 1, 1914.
PeUo, J. W., L« A» Mulkey, B. B. Swank, 1. E. Llpcaei and S. W. frown* A
Screening Procedure for Aaaeasittft die Transport *od Degradation of Sol id
Waste Constituent* in Subaonrface and Surface Water*. Eavtron. Toxicol.
Cheg. I 121-124, 1982.
Farmer, W. J,, M-S» Yang, J. Letey, and W* F. Spencer. tei*d-Bi*poaal of Haxa-
chlorobenzena Vaate: Controlling Vapor Mov^nent la Soil. EPA-iOQ/ 2-80-
II, Municipal Environmental *e search laboratory, U.S. Envt roroNcntal
Protection Agency, Cincinnati, OH. Aaguet 19BQ.
Fardy, J. J., and 1. i. Sylv®. 3IAS — A Coaputcr Program tot che Generaltxed
Calculation of Speciation in Mured Metal-Ligand Aqueous Systeaa.
Faastervacher, T. and L* Ottiaatti* Eyaleation of Exposure Aeseeaaent
Mathoda. Pickard, Lee*, atel Carriek, toe.. Pt«p«rrf foe Cteeauoel Manu-
facturers Association, Washington t DC. January 1987, 15? pp.
Feljey, A. I.» D. Gicvia, mad Z* A. Janna. HXMTTO: A Coaputer ProgrAM for
Calculating Ceochemical Equilibria. Battelle, Pacific Northwest
Laboratories, U.S. Emirottaanea I Protection Agency, Vaahiof ten, DC.
1983.
VI-43
-------
G*aes, L. Field Validation of Ezpontfi Analysis Modeling Sysc«# (EXAMS) in a
fUvioi S(re»Bt p. 125 io Hodaling thaTat* of Chmu.cals ia chat Aquatic
gavirowncnt. t. L. Oickaon, A* W. Haki, arvd J. Cairaa, Jr., ad*/, Ann
Acb4t Science Publisher*, lan Arbor, Hi« 1982»
Ga*s» id» Aaoricafi
CebpefticaL C&ioo, KoahiofCOB, DC* ifS4»
vi-m
-------
Jaoae, E. A. Crou*d««cer Kodollct$t A lt«i«v. Battalia, Pacific lortiwitt
Laboratories, HH-J574, fticfeland, Ut. LM1. 41 pp.
JeiitBND, l» C. r and J. L* KUcl<« Design, ProiraaaaiBj, and Maintenance ot
tart, H€1 J. Tochnol. JmUt civil Zm. IQf 41, If 13.
JoKuiSO) ft* L« i JT . C. Iihof f i H® 8« Btvi 1 f Jr», J# L< KiCtLa, Jt< i u4 5«
Daol$Ujt, Jr. ffaar't KanuaL far HydroletleaI SUteUtUa fhragre* -
romuii {H5P.F)• KiLum Bo. 8| EJA-*G0/3-t4~0|*. U.S. Ibsvi.ratiaissttl
'ro(«eiion Aftacf, Acbni, &A. 1914,
Ktafata* 9», «»*f t. KlMriMt Mmn§bmic Yrwmpmrt Modal, for Toxic iufc*
•tanata* QaviroaaMtal Protection Agency, VmBMrntim® BC• Offici of
t«mic Subataitcoa, m$ td|M CPA/Jtf/iVt«6$/001. WflS Wo. nS€-tlt*lt,
docunaatacion. (see tbf. fcnrir. foil, ft CMt. (4> II. J«ae 5.
l»«Sii. —_____.
Kaufcaaa, D., «o4 1. Kiaarto*, TQX-SCtCt* Multimedia ScrMmitg Lawel ProgTaa*
EnrirooaMnkCAl Protection Aftner* He»Ki«$teci, OC. Otfic* of Tondc $ub~
atrnii, «•« ctpe UA/fV/KT-lftfOW. MTtl Mo. Ntft«ll2430, 4mmmacaiioa,
PB84-213750 <••• ab». Bnvlr. fell, t Cane. (3) 44-45, Jm* S, 1915b.
Koly, J* f.» 8ydrolo|ist at Bo bore S. Kerr EovlrofUMntaL tasaarcb Laboratory,
OK. PkmuI Cosaaunlcatioa tricb Fred Hopkins of Midwest Research
laacitwte, Kansas City, NO* March 2, lt!3.
tec# 0* Leodute Him Ntinilift Predictio« Nodal. PoporUMac «f CaolofT*
Scat* Stillwater,
-------
JCuf«» C., 0. Twadell, 3. Paige, X. Vet*el, P. Spoooer, R. Colonaa, and H.
KiIpatrick. Rating the Haaard Potential of Waste Disposal Facilities,
pp. 3C-41 in* Proceedings of the of Uncontrolled Hazardous
Waste Slt««, U.S. EnviroiuneDLC.al Protection Agency, Washington, DC. 1980.
LcrCraftd, K. E. A Standardized Systeei for Evaluating U«»t«->Oitpoeal 8it«s.
National Vactr Wall Association, UortUmm, OB® 1980.
Lascar, 0. , ic «1. Sysceat Analysis of Shallow Burial. Code Manual , WHWLC-CR-
1913. Volt. 1# 2, «ad 3» U.S. Huclear Regulatory Comission, Washington,
DC. i«l.
Lywan, W. J. Prediction of Cheiftical Partitioning itt the Environment* An
AsiiiiHflt of Two Screening Model I¦ U.S. Environmental Protection
Agency. 1981. (Cited in Travis, 1985.)
lyma, M» W. F. fleaekl, and 0® H. Rosenblatt. Sandbqoii of CheauLcol
fhro^nrcy Estiiatioc. McCraw-Hill look CeBpeny, New Tor*., Bt. 1932.
KcUhorter. D. B., and 1. 0. Nelaon. Unsaturated flow Beneath Tail Luge Lr-
poundJMctB. Journal of the Ceot»p.hnlcal tnaingeftng tKvision, pp. 1317-
1344, Novae*** 1979.
Nercer, tf. , and R. Faust. Groundwater Modeling. GeoTrens , Inc., Raston,
m. 19S1.
HcOowell-Boyer, L. H., and D. N. Betriek# A Nulcinedia Screening Laval Model
for Assessing the Potential Face of Cheuicals Released to eha Environ-
ment* OtHL~6041{ RPA-560/5*63-024# Prepared for the U.S. Environmental
Protection Agency by Odk Ridge National Laboratory. 1984.
Miller. C. Exposure Asseseaianc Modeling: A State-of-the-Art Review. EPA-
600/1-78-065, Environmental Research Laboratory, U.S. environmental
Protection Agency, Athene, GA« July 1978.
Monro, J. I!., f. J. Luxeoore, C. L. Segevich, K. R# Dixon, A. P. Vat son, M.
Patterson, end D. R. Jeckson. Application of tba United Transport Model
Co the Movement of Pb, Cd, An, C«i, and S through the Crooked Creek
Watershed. Oak Ridge National Laboratory. OWnJWSF/EATC-28. Oak Ridge,
T*. IPS.
Kara*tahan, I® 0., and P. A. tfi tberspoom* **Buaerical Model for Saturated-
Uoeaturaced Plow in Oef ormble Porous Media{ I. Theory." tfeter Besources
Raeaareh 13(3> 657-664? 1977.
Kareeiafean, T. V., P. A. tf I titer spoon, and A. L. Sdvards. "Suaerloal Model for
Saeurated-Ub saturated Plav Lb Deferrable Parous Media; 2. The
Algorithm." Water Resources lasearch 14(1) 135-261, 1978.
Karasitahjui, T. V., and P. A. Wither spoon. "Buaserical Modal for Saturaced-
Uneaturated Flow in Defonaable Porous Madia; 3. Applications." Water
fctaonrcea Ramaarcb 14(6) 1017-1034, 1978.
VI-4*
-------
KAS/W1C. frcinclplea for E«*lamttng Cfedtticdia 1ft clue Bo^irt>qjt»oc.. Cosmic Cce
for cive forking Conference on Principle* of Protocols for Evaluating
Choiiicii* in the Cnvlrannmc, HiticQal ludea/ of Sciences, «a. 13117
Onishi« Y. Chenicat Transport and Fate in Risk. Aimmpt. pp. U?-l)3 in
Principles of Health Bilk Assessment. P. F. Ricci, ed.. Pr«ntic<-*iell,
Inc., EagLtrvood Cliffs, HJ. 1982a-
Onishi , X, Chesucal Tirana pore and Face Models, pp. 155-23* in Principle* of
lealth liite A*m*wdc« 0. F. licci, ed.» Pranttcsr-Hall, IncTT"ljngTn»ood
Cliff*, MJ. r?ilbl
Onishi, end S. VI. Wise. Mathematical Model, SEAATM, for Sediment and
Pesticide Transport in River* and t(< ApfiiexciM to Pesticide Tranipert
id Pour Hila and Wolf Creeka in Iowa. Prepared for the 0,8. Environ
nental Protection Agency under Contract Ho. 23 Ill-ill 241 by 9attell«
Pacific Northwest Laboratory, Richland, UA. 1979,
Onishi, f»f S. H. Brown, i. 1% 0l*«n, and M» A. Perfcburst. The Chmicel
Migration end Risk Assessment Methodology Suanaery, Dtslt te^orc. 1980a.
OoliM, Y., 0. Li Schre«ber, and l» B. Codell. Hathenecical Simulation of
Sediaent end Radionuclide Transport la th« Clinch River, Tennessee.
Qu li ia Contaminated sunt Sediment*. Vol. \. R. A® Eater, ed., Am
Arbor Science Publisher*, Ins.* Ana Arbor, Ml. 1980b.
OrtLa, » V,, D. V. 2adUMua&, 0. R. Kctfhorter, and D. K. Sunada. Effects of
In-Transit Water on Groundwater Hounds Beneath Circular end Rectangular
Racharge Arena. Hater Besourcea Raaeaxxh 13 C JI 377-587, 197f»
Oscer, C. A. Review of Ground-Water Flow and Traoeport Model* in the
Unaaturated Zone. Bactelle, Pacific Mortlmest Laboratories. PKL-4427,
NUREG/Ci-2917, U.S. Nuclear Regulatory Cosesiesioa, Uaahington, DC.
1982.
ft-47
-------
OTA« technologies and W«n«ge*«t Stricagi-ta for Bezardous Wattes Control.
Office of Technology AiMiwoit, V«shingcati« DC. 1983.
Pscerson, S. Mods Is tar cb* Iaicial Integration of Physical «od
Cheaucal Properties. pp. 217-111 in laviraciiiwtJi Exposure Crow Chemi*
c«ii» Vol. I. U. 1. Neely and C» E. BlAu, eds.( CRC Press, toes fcacao,
ftT 1963.
Patterson, M, R., T. J• Sworski, A. L. 5|ar«wf H, G. Krovoun, C* C. Coutant,
D. M. Hetrick, 8. 0. Murphy, and R* Jr Raridoo. A Hum's Kanual for UTO-
TDX, tb« Unified transport Kodel• 0(UflL/XM-6064. Oak tidge National
Laboratory, Oik lidgc, TM. 1984.
Pettyjohn, 9. A,, I* A. PrickeCC, 0. C. K«oc and H. 6. LeCrand. Prediction of
Leachata flume Migration. pp« H-il »i Proceedings of thai Seventh
Annual ftesearch Sysvpoviun on Land Dispooei « Hazardous Vasts. EPA-60Q/9-
81*0026* U.S. gnvironaencal Protection Agency, Washington, DC. 1981.
Potest, E- Soluta Transport/Croundwecar Plov Modal. Colder Assoc.« Seattle,
VA. ifftl•
PriciMtt, T. A®, T< G. Haymk, and C. C. Lonaquist. A fcandoy, H* £., and ft. A. Criffin. Catinating Threshold Valoes for the Land
Oispotel or Orfaaie SolveAC-Contaniaatcd Wastes. J. HmearioM Materials
IS 36S-37i, 19t7. ' ""~—•
Schnoor, J. L. Hodaiing Cbaaucal Treesporc in Lakes, Siver*, and Esc^aruic
Systess. pp. 55-73 ia toyiro^Mital la#o»amg fra« Cheaiceis. Vol- II®
S. Rarely and G* i. Bins, ada.y CftC Press, toca &aco*t PT.« 1983#
Vl-48
-------
Sposito, C., iad S* V. Mattigod. CEQCiEH: A Computer Program for tha
Ca l.culacion of ChaaicaI Equllibtim I5~SiTl §olaiciottir~*nd<3niIr~Marural
tfater Syiuai. University of Call fornix( iivtr«id«, CA. i960.
$yfcesT J- F*» S, Souypak, ami C. J. Farquhar. Model log of Leachet* Organic
Migration and Attinuation in Cfou»dw«t«r Selov Sanitary Landfills* tf#Cec
Resources 8«i«arch 11(0 135-145, 1982.
TAC8. User's Guide Co the Texas CI ideological Model. EP/t/OP-Si-OOlB, Teus
iir Control Board, Austin* TX* August 1980*
Thais, T. L., D, J. Kirtaer, and A# A- Jennings* Httlc.i-SoLntst Subsurface
Ttntuiport Modeling far gnKray John Wiley
1 Sons. Chichester. 1980. 157 pp.
Uilson* J. L* and P. J. Killer. Two-Dieansional Pluse in Uniform Groundwater
Plow* JcHMtMil of tlm Hydraulic* Piv. An- Soc. of Civil BojuD—riag Paper
00* 13665/9*1* pp. 503-514, 1978.
ViscJaseier, V. B., and D. D. Saith. Predicting ftaiafsll Erosion Losses - A
Guide to Conaervecion Planning. Agriculture Handbook tfo. 537, U.S.
Departaieat of AftUulture, Washington, DC. If78.
VT-4»
-------
Wood, V. ?., I. A. Moid, J. Let'lar, and J. 0. Pilocta. Environmental Parti-
tioning Model; Do tumeric it loo (Draft >. Cenerai S«fiw»re Corpora* ion,
Landover, MO. Frepircd for U.S. EavlroniMncal Protect ion Agency. 1982
(ci ted Ln Travis, 1965).
ftfc. G. T. AT1230: Analytical trimiffat One-, T«o- and Three-Dimension*!
Simulation of U«»ce Transport ia the Aquifer System. ORffL-5M]l, Envi-
ronmental Se iencei Division, Publication No. 1^39, Oak Ridge National
Laboratory, Oak Ridge, TX. 1981 ( ex. ted in EPA, if82tK
tmh, G. T. and 0. W. Ward. FEMWATER? A finit«— ElefHnoc Model oS Water flaw
Through 5atarated-Unsaturated Port>u.« Media. ORKL-5567 , Environment all
Sciences Division, PublicaCion Mo« 1370, Oak Ridge National Laboratory,
Oak Kidft, TN- I960 (cited in EPA, 1912*),
¥«h, C. T". tnd 0. V. Ward. FOfMASTEi A Finice-Eleeienc Xodel of Waste Trans-
port Through Set urated-lTn saturated Porous Media. OWL-56UI. Enwiroo-
¦ental Sciences Division, Fublicacion No. 1462, Oak Ridge national
Laboratory, Oak. Ridge r TM. 1981 (ci ted in CPA, 1982a).
VI-50
-------
VII. OCTOTO1I FtfOICTlOi
prediction involves oaiag ii1k(«1 chaan'cat «ss« «s;i»i;??A:afeI«
i>(om(iea m which and bum amy paop(« «*tfetn«iw» aiitiianta of chn effecta of anvi routes eal contaer-
iuQti cao r*|«irn consideration of Mfiiurti of ••vcrai Mads of populations
md •efcpopuldtload* Populations cm ba coniilaiai conveniently in five gen-
erally distinct claanea aa
• tMittn at sicca >Acr« «MCau«aati originate or are being
handled la large
• IliiwlMiifis of the fablie nboa tba enntiaiaaat aty
• lateral papulations of (tan aatd fauna.
• Plants and jui Inula of agricultural, ccnnnerciel, and aesthetic
value (e.g., crops, hratta, fliberies, omejeentaLs).
• Other objecta of econeane or eeetbecie value {*«g», corrodahle
peinti, structural natariela, and werka of art).
The lollafio| diteuaiion briefly addraaaas oecupetieoal exposures, but focusaa
on eapoaurea of the poml public. Resources did oot peruit further disevn"
eion af other kinds of enposurer f«u«atH above, bat aaojr of the principles
noted aodar general ^optUttot npatnrM alto apply to %hm» In moat deci-
•i*M at DA, aaflfeavra* mi tnnama to bmmtimm mikmmtm »rm lifcely co ba tte
ptimuy tmmrnm*,
I. §>irmmtimm'k Workers at {aeiiitiaa where heeerdo«e
mIiwik ara peaannc ac hi||i» MKantrMlini and La large <|eai»ticie* can s«»
ti|>i(tciatly anpaea4 to aock idkitiMM if prtfir oafaguarda are not Mia-*
tataed. Many at«'lias ipt teas aaada mi BccmpaltHrtmt ixpMrvs aC planes that
produce, f omul ate, ar mc any of hundreds of ceaanarcial chunicals. Cai4«-
liMt ml v«|ulati«u far taortacr aKfNisttraa kin tawm aacabliab^ by iAduaory
~Kg (ivinMK ffff i*i«f cl*iisiii«*ls,( fvr «saa^lar ilia Vatiattal tnacicuea i(
Oeoifdt&«ML tdlacy m4 Kadleh» iadoaeriai firM or trada aad*cia-
tieai, tiitl iismti mIhu kmm data m RffiesS. csposmcM*
-------
The teericao Conference of Governmental Industrial Ifygienist *, che Occupa-
tional 5*fecy in4 HsaJLch Administration + and la sane ciwi »tac« or local
•fancies Mw published guidelines or ataadards.
Workers involved in creating, handling, transporting or disposing ol
hazardous wi«t«i at TSDFt generally wo«i4 be covered also by regulations op
»pecifLc chemical a 4 although potential exposures wiy not bst vail documented,
particularly when coisplax fixtures of chemicals are present. (#0rfc«rs invplved
in *piU eLeanup or corrective actions can also be exposed* Worker exposures
tend to be sics specific and arc difficult to «stta»t« quantitatively. They
can vary trtch che chemical mod physical a#t«r€ of the materials being handled*
che technology bsing used, and control Mature a and crtipmi prainu w»
place. Hiey can vary with the particular Jab or writ station «c a ®ivmn
facility and Che individual'* us* of avaiLabia control measure* and prescribed
practices (e-*«, use of dust natlcs). Th* prisuiry route of exposure is I Italy
to be inhalation, but absorption through the skin, and hand-to-siouth tranafers
are also possible.
The assessment nusc identify tha types of workers exposed. the nuwt-
ber of eaoh type, and the frequency* intensity and duration of exposure co all
relevanc taxicanta. La addition, secondary exposures of worker's families can
occur occasionally whan toxicants ara carried hone tm worker*s clothing at
bodies. While auch exposures are probably small for workers in well-run
modern manufacturing plants, this possibility should not be neglected in an
assesamenc of hasardoua waste disposal facilities-
2« CiJtftl population eiposureii Mtttbtra of cKt public
can be exposed to&axardtwis substances isleveral ways. The present discus*
slon will focus on exposures co oontamioanta moving through environmental
routes and eill oct include exposures to chemicals used directly in che pro-
duce loo of foods* medicines, or consumer products. Assessments of exposures
to environmental concaminanta must consider three factors: where people iivei
where they work! and other lifestyle factor* that Slight increase exposures.
Expoaura routes thac Bust be consideredL include inhalation ol contaminated
dusta, niata and vapors, ingestion of contaminatad water, food* beverages or
othar suhatancea * and dermal absorption follow lag contact with contaminated
substances.
lo asaesanents involving hazardous mstsi, attention usually will be
directed to the TSDPa as potential! poioC aourcas of contamination* In sane
cases attention must also be given to transportation accidents that would
cause additional point (or possibly not) point) sources« In either case the
eflvitopsMCal transport and fate teddies or modeling runs will have provided
information oa th« «pr««4 Cloth gtttfrJifMemliy and entttv time) of the pol-
lutants from the source through the air* aurface water, groundwaters, and
ochar aovirouttntal comparcmancs.
For air nmiasloo** dispersion of the pi oat oveur residential areas
would lead to inhalation of pollutants. Inhalation ac site-speoili-c eommer-
cial or institutional fasilltie* s*y require consideration in «oae eases, and
alternate exposure routes (euch as dapoaitloa of pollutants on foods or ser-
faces) s»f repair* attention 1a ochmrs, For pollutantf catering surface or
¥11-2
-------
irouahiUrf, contain nation #1 witar lapplias could lf*4 ll iageicioo of co«r»
(IMbiki in drinkiog aatar and foods or brm4|«t» frivitt walls, public
»*t«r ami batfc feaatt ui coaMtcial l#®4 procaasiag abould bm,
sidarad. fat iom pollutanta absorption through tba tkia during batbing or
othar contact and inhalation of cbamicals volatiliaing in cba showar nty
ra^uiro conaidaratioa. It mm «u«« lifcacyla (ictsri mth i> daily cocaaut-
Ing through s diapartIon plum or fraqueat racraaclonaL tctivicict in potan-
ctally contan£n«cad Ukaa oc parka nay aaed to ba conaldtrad. A praliaiaary
dMlytU of tba geographical diatribueioo m4 population dWCributioos tbm
4t(tniMi tba aajor ufoiura routaa and populations cbii (biuil be analysed
4d*aclcaciv«ly.
Ci»« tba iwiriptieat bouadaria* of cba pallutaat'* di**ar«laa
(coordinates of latitude aatd !aa|iCMi«) froa tho «our«« sica* a dacailad popu-
lation profLta can be dt*»lopai for tha study ifMi Hw U.S. BarMio of tba
teiw caioad from local govomoNiBca or planning coraii-
lioni* Local pLaaoing eo«iitsio«ft a«y ba abla co provido corront and pro-
i«cc«d eOHBuoity growth U tha tcudy area. Tabid fll-I liata aourcas of
cwtMa ayacific 4mim m poynlaciaaif.
tkrwrai 0.t« g»^rw»«nc at*aci«* and EW divifiona ara curranciy
c«yiot «• «•« Om Pi'« ICAOI Pill a* tM *ub4**4 |«o|T«fkie co^&bb a«chaaiaA
for data •Mcafe* Tba CTA't fodar-ai. Data lyataai (nu») U«u
- I0»000 coaMniity
-------
1A1LE 711-1
SPECIAL QA?A SOWCgS TO ASSlSt III PROFILING A POPULATION
Data. Bases
PMcriptioa &f Iafccantjop
Contact
Census
Drtokiaj Water
Supply file
Gron«ct»«t«r Site
Inventory file
Inventory ol Public
Vster Supply
REACH Fil«
STAR ied Grand
Condition*
Be«egr.®§Me characteristics; In-
cludes coawttins patterns and
*auaer*tioa by school dlstict
Provide* locations of viter
utilities, xat#Jte®, and sources
Lists location* of public tad
private well* ajidl spring*
Lifts detailed inlorauticn about
60,000 com unity vater tysteai
and 160,000 tnmiimt vitcr
sryvteis
Lilts the intersactioos of
river# with, tributaries, lakes,
•ad otber water bodies
Coat bum wind distribution
(rase) data, daai»aeat wind
direction, precipitation and
Cooperatare
U.S. Bureau of
Census
301-763-7315
EPA
202-352-7046
uses
701-460-603I
EPA
202-382-5551
Monitoring Branch
EPA
202-382-7074
National CJUniti
Cnnte*, *C
MOM
704-258-2850
¥11-4
-------
LieicatloDs on profiling the population effectively m: (a) the
quality m4 ooopleteae** the data available! and (b) the ability to specify
pertinent afrbpopulacieo* and iq obtain §co§r«^hic«iIy cod«4 information.
Extrapolated data fr«a other areas aiy be useful, at tims, In supplementing
the developoeat of the population profile. Daily hwoas activity pattern* and
short^cr* population fluctuation* *»ch as consutiog patterns, sehool hours,
etc., my also need to b® weighed, but are difficult to quantify confidently*
8inc« the intensity of haalc h effects could be compounded for indi-
vidual* in the general population who are unusually lanslciv* to the pollutant
of concern, eh* praaanca of such individual* should be noted foe coo aide rat ion
in the Health iopact analysis. Background health data, sum of »hich is
refionelly and locally specific, are available through the Metiooal Center for
health Statistics, Center for disease Control, and atate and county health
agencies.
8. Estimated Exposure*
The eatitaacion of exposure* involves two phases; (a) development of
exposure profiles for population groups; and (b) integration aero** population
groupa.
1. Exposure profile: Ac exposure profile nu*t be developed far
each of the Identified exposed population group*. This profile should idetr-
tify three characteristic* of the expoeuret
• Route of expoaore: air, water¦ foods, soil, or other routes
Tie* over which exposure will occur I cLbw to onset and dura-
tion
» Concentration of eootaolnant received: concentration by route
and variation over tine
Average rices for Intake of air, nc«r, and food for hususna are
available (sec Table VUI-2>» EacliSace* of the intake of contaminated toil in
different activities by children and adult* have also been published (Kawley,
IfM)#
2. jjiipsfiigs Integration: The exposure integration process
involve* aggregating exposure to the extent possible over routes, eoncencra-
tioas, cine* and population group*. The degree of aggregation that i* reaaon-
able can depend on several aspects of the coauueioumt*. The toxlcological
properties of the cheaicals of concern trill be particularly important, but
pbysical-chesdcsl properties and route of expoeure ¦*? alto require consider*
ation.
For exjwsple, if the chaasical is nonvolatile and the health effect of
priaary concern is lung cancer after prolonged inhalatioa of it as a dust*
Chen exposures via drinking water probably should not be routinely igtrtiatad
with chose from air emission* to calculate the total exposure. Similarly,
VII-5
-------
dermal exposures to a chesaxal ttot can I* readily absorbed froa ««t§r would
be diasimlar co diraiL exposures If it is tightly edsorfeed on dun particles*
*nd the two should not he routinely iwk4. ft* toxicologic*! licer*cur«
doeamenc* many is which ttui health effects diffar depending on eh*
io«*fe route- («t other hoMm, response is similar for different dot*
rDucsi, For l should be identified* They (tMrtllf1 should oot be
aggregated wiefa all other group* to caLculate an Average or neaA expoaura
level. The reason I* chat isic of clue adverse health iopacts could ba
incurred by a mulLI ouaber of highly or fre^ueacly exposed persons. The
validity of this reasoning i» moat essily seen For m ehaoiical with a well-
defined threshold exposure level (i.e., only exposures above threshold cause
adverse effects}. It also holds true, fcwvever, far effects such a* cancer
that are assvxsed not to have threshold tvfonuces, because thi uncertainty
range of estimated health affects usually increase* rapidly as exposures
decrease from thosa reported In Coxicological Or epideskio logical studies used
as reference points.
Aggregation across tine el so requires care in that some health
effects are associated primerlly with short-term exposures at relatively high
concentrationsr while ochara ara associated with long~tecm (even lifetime)
*xpoturea «t loner leveLs. Hence, population groups »ith substantially dif-
ferent temporal aspects of exposure shouLd be treated separately in the health
affects estimation ami should not be aggregated m the exposure prediction.
In sumsry, the integrated exposure tuttsaim tabulates all signif-
icant population groups according to the relevant environmental doae chat each
is estimated to receive.
fll-t
-------
Reference* *o Chapter VII
DOC. Councy and Cicy Data 3dok. Statistical Abstracts SuppLeaenc.
Table B. Counties - Vital Scjcitcici and fcftlith Car*. U.S. Bureau of
C«nsua, U.S. Department of Co
-------
-------
Vtll. MALTH EFFECTS PKEPICTIQM METHODS
list iimsaMCB involving roxic ttjbrtafte#§ require predictions of
the type end d«tr«t of adverse health effects that would, likely result from
given exposures. Such predictions riquirt line a review of .appropriate epi-
demiological , tosicologixal, biochemical, and othsr health-related literature
lor the cheaicals involved in order to- Identify their effects on humans, other
species, or other test systeos under known conditions of exposure. They
require« secondly, an evaluation of ss»thads for extrapolating health eff«oc<
from the known conditions to Chose of the problea *t hand. The «»< appro-
priate predictive methods art then applied to the information frooi rh* Lit era-
cure to estimate the probable effacte under espeoted conditions of environ-
mental exposures *
This chapter characterises the health effects data base typically
available for chemicals of Interest, and the models that have been proposed
for health effects predictions. The discussion is broad, bat not exhaustive*
tc is iatended to five the reader an overview of the strengths and limitations
of Che field. Good recent; reviews of most of the topics covered are available
elsewhere, e.g., N&S/tfRC Clf?5)| OTA (1979, 1981); Richmond et al. (19S1); CMA
(19.84); Clayton ec el. (1935); licci (IMS); Boodhead et *1. (1985)I OSTP
(1985). It provides a basis for the general nethodoiogy described in Chap-
cer t*
km Characteristics of Health Effects Ikta
Virtually every chemical bae the poteacial to produce adverse health
effects ia biological organisms if administered in sufficient quantity over a
sufficient period. The nature and degree of toxic response depend on the
chemical, thai conditions of exposure (e.g., route, concentration, duration)r
and the nature and condition of the receptor organism. The literature con-
tains information on many kinds of adverse health effects from a large ou«ber
of chemicals under a wide range of study conditions, including different
routes sod temporal patterns of exposure, different degrees of control over
exposure, and differenc test rpeaies or biological systems. A risk giseisnsec
of potential environmental exposure tc a given chemicalj ideally would use aa
a reference the results of oae or oor« carefully controlled studie* mad* under
CesC conditions very similar to the enviro«aeau»l conditions. Practically,
however, such ideal reference studies seldom are available, and the analyst
nast draw oa a more extensive literature to predict potential adverse health
iftpeccs*
Servers! approaches are oj«4 im toxicity studies, depending on the
properties of Che chenical of isterete, the health effect of concern* the type
of information desired, end the re>Our«es Of the reeearchers. For example,
rapid effeets of inhalation nay be of primer? concern for a new gaseous induur*
trial chemical, while long-term effects of ingestion would need to be known
for m food Additive, and tit mechanism* of UNA interaction night be of
interest for a known mutagen. Since information developed by different
approeches is co«pl^wc*rf# information frosi one or store specific studies on
f111-1
-------
a given chwiicil and the substantial body of background health effects valor*
macitta often will be needed for predictions.
i. Trw q£ wpoiorm Exposure to a tonic substance occur* when
tht ittbsCiwce ia present at the interface between a biological organist# or
test material and its environment, c.|., wiwn the substance is In air m
breathe, in food, mtcc, or other liquids «e in§«*t, or in neterials we use or
contact. C*po*«rc subjects nay be humans, laboratory and domestic anieals,
microorgaaisws, cellular preparations, or pertinent biochemical test systems*
Exposures can be controlled and Unotm, es is usually the case in phanraco-
logical ©r coxicological studies vith bimni (e.g. , of medicinal or consumer
products) and la coxicological studies wich pprthuman species and other blolof
ioal component systnasi or they ai|k be uncontrolled or uncertain, afi is the
esse o£ clinical reports or epidesslologioal studies of humans exposed to con-
taminants in the workplace or through foods or the environment*
In controlled studies, ch* substance nay be administered in air,
food, or water, by lavage (by tuba into the stomach), by injection under the
skin (subcutaneous)» or into a body cavity (e.g., intraperitoneal). The
aditnis gered. _dofe is the aorounc of substance that actually contacts or enters
the orgenisn through bodily oietsbrancs or portals. The total administered dose
in a study is 1 product of the concentration of the substance in its carrier,
the volume of carrier crossing the interface per unit tine, and the frequency
or duration of dosage* la nwy caii). however* the total administered dose is
not too useful a measure for making comparisons, and response will be reported
*s a function of sveapler variables such as concentration mass or tim. In
other «a»e#» particularly if chc substance is rapidly excreted or metabolised
by the organisa, chn concentration of the substance in the blood or at the
affected organ site nay be a more meaningful nauurs, aod sill be reported as
Che delivered dose or the effective dose.* The delivered or effective dose
can vary "depend inig on the route. carrier, and tiaing used for the administered
dose, and also with the test specie* ac a giveo dose. The concentration of
toxicant at the site of action on a cellular level most often is not linearly
relaced to the administered dosel a siganidal (S-shaped) relationship is com-
mon.
For environmental contaminants, the tern environmental dose is
analog*** to the administered dose la controlled studies. (Unfortunately, the
tera "exposure* often is used also In a quantitative senae for enviroomeatal
dose.) At In controlled studies, the effective eDvironcteatel dose and hence
the response* can depend on the conditions of exposure.
The tmporil characteristics of exposure can have significant etfect
on the (magnitude aod avem the type of response* for example, the affects of a
large single dose of alcohol say be readily apparent (inebriation and poe-
eibly death), whereas che same quantity administered over sMUiy dayi amy have
little effect. Oa the other hand, aoete substance* administered frequently at
levels without obvious effect nay aecusaulate in certain organs and cause
* The term "affective dose" occasionally is used to aiaan *minisa» dose
causing the effect," which is sure often called a threshold dose.
TOM
-------
subsequent adverse if facts. Far hunans both §fa®*t-t®r« and long- tor* eapcr-
turn can be important.* The traditional uriif of to*ic®Lagic»i studies with
laboratory uiuli includes acute, rubchronic And chronic exposures, mu4
ftpseiil studies which can address effaces ac certain developmental iU|i) or
in subsequent generations following exposure (sec Aniauil Studies below for
further discussion).
2- IMS- of effectst Toxic effaces have batft qualitative and
quantitative aspects. An Increase In the number tod severity of offset® can
occur with increasing exposures; an increase in tbe prevalence eE a given
efface (or response) with dose can be of prJjaary tnterea t ac tines.
a. Effects of coocorc in risk assessment* All significant
adverse human lute I eh effaces (end poioxs) should be considered in risk assess-
ments. These include effects usually teen aa a reeulc of lonf-tece (chronic)
exposures* and those seen after shorter exponnrti. The health effects of coo-
cam include general toxicity! effecca an digestive, respiratory* or cardio-
vascular systems; effects on central nervous ay*ten (neurotoxicity}; effects
on che liver (hepatatoaieity); effects on the kidney (renal or nephro-
toxicity); and growth end development rates* Other health effects of coocem
will also include; oncogenicity (causes tumors of aoy kio4)» carcinogenicity
(cautea caocer or leuktaia), otutageiticity (causes mutations). embryotoxiaity
or fecotcxicity, teratogenicity (causes defoneed fetuses), induced scarilicy
or decreased reproductive success, adverse behavioral effects, and tellular or
subcellular affects. Table Vtll-l sumaerlxes the rang* of effects of poten-
tial concern.
The literature will caat«ta a variety of information and data,
including observations sad experiasentel results involving exposures of humans,
animals, or other test organisms or substances, and conclusions regarding the
range of adverse heelch-related effects. Different chemicals may produce dif-
ferent characteristic adverse effects f but a given chemical amy alio produce
several different effects (out or rare nonlechal effects and perhaps death) in
different dosages and exposure situations . EPA hea proposed guide I Lnes for
assessing the risks ef carciaogena, nutagena, and developmental toxicants
(Anderson et al« I981| EPa 1986a, Wife, and 1986c). The literature can be
inconclusive about whether a chanical causes a given effect, particularly
cancer. The International agency for research on cancer established guide*
lines for evaluacing evidence of carcinogenicity, a«*d has reviewed available
evidence for many cheeiicals and classes of chemcals Ln a aeries of reports
(IA2C, 197?).
b. Hmigti _ of response: Biological response data are
reported in one of three ways! qu«nt*l, graded, and continuous.
QuantaI (dichotonous) data are based on a "yes-no," "all-or-
oone" detertaination of a specified eftd point. A test subject has the effect
# for example, both l>-i*ia peak exposures and continuous* lower-level
exposure* my be addressed in setting workplace standards.
niw
-------
TA1LI M1I-I
jwvcasc kalii nncts wkt cm k cjubm m §mm
g sag qlilCTI "
Itebilif Ltm Effect!
AUargfet
Aattaa
Arteritis
Bebftirionl disorders
Cirrhosis
Oeraatolojicsl disorders
fLodaerlaolofic«l disordsrs
iMruDologtcal disorder*
Neurological disorders
Reftel disorders
Smre veiffct loss
lip,.n^wcttye Genetic tfftctl
Sterility or fecrsaaed fertility
ftt»e*rr§i»|pe» stpoatweM* «tortieiM
A»d«OMi iwigkt or Ifi§« St birtfc
Teratogenic effects
Itottipiic tffecu
Freqneatlv Fatal Effects
Caticer and leukesuss
Central nervous system disorders
Seeere rsspirstsry or (sstroUtestiiul dlitress
Ktart sad circulatory disesse
liver function loss
fill-#
-------
Or It <10# i not » Data art rsporced, for example, as percanc of lubjacCs dead ,
p#rc«nc with cj.nc#r» or percent with siiia irritation. Qua&tal data ir« used
extensively for toxicologic*! aanparisons, pertiaulerly as indices of potency.
Best teowo of thai* Is the lethality 1 tides, the LBSiP which is the quantity of
substance which kill* one half of eferti tent subjects the aedlan dose).
The LDj0 ii usually stated m milligrams of substance {Mr kilcgrasi of body
of subject (og/kg) but can bit in other unit*. The LDS0 ia usually on*
of the iirie toxicologies! paraiicin dtcatnincd lor a chetaical, bat its v»Lu«
often varies with the «peci«i» strain« and cha a ax of tha test mimIii and
bcl»«ea laboratories* Orel LD#a'« arc usually I over than dermal LD*0's, but
higher Chan intraperitoneal U)90''< Tha U314 is one of the most reproducible
of coxicologiceL parameter*, but a Cwt»-fald variation between species (or «c
umi iMCveBo test groups of che jam ipecies) is doc uncommon* For *oom pur—
poaet| knowledge of other toxic dose lev*Is amy be dtiirabl*, tuoh ai cha LD|V
or ID5fl (causing SOX cuaorigecie rtipMii above background dose). Quancal
data art used is most of cha cmmqq risk extrapolation nodal a.
Graded data are based oa a seep response concept of severity of
the response, «.g., absent, minor, aoderate, $«v«rtr wy severe. The graded
format ii doc used executively to report toxicologic*! data (pathological
reports on lesions am one exception). Graded daca are considered to be a
subset of qu*ntal data by soxm authorities (Klaassen and DouLl, I9fi0)» In
quantitative risk, assessment, graded data usually would be transformed into
quant*! data for uaa in extrapolation aiodala.
Continuous data are often based on che decree of response
within individual test subjects, as wfll as art>og individuals. Typical daca
ia the continue?* response forsst eigbc include I percanc loss of weight I
changea in performance or behavior} percent cholinescerasa inhibition or
cartooxyherooglobin in the bloody or percent reduction Lo respiratory function
or a perm levels. Although such end paints a an be quantitative, they are not
necessarily direct neasures of tosiciey, or may cot reflecc basic toxicity
mechanisms.
3. Trmm of teit-wwuie relationships? If exposure to a sub-
stance causes a given adverse effect• then the relationship between the dose
and the measured response will usually take one of a feu faoiliar patterns* A
graphical presentation of quantei data usually will show a nearly linear rela-
tionship between response and dose (or logaritha of dose) in the tbidrange of
the plot,* If data are taken at sufficiently low dosesj the lies usually is
found to have convex curvature.The upper portion also May be oppositely
curved to give a slightly sigmoid ( a ** shaped) curve (che typical integrated
oormai distribution curve) as shown in figure VTtl-lA, Two good eidraoge data
~ Dose-res pond eae data are somotiiies specified as dose-incidence for quantaI
data and dasar-eJfcce for graded or continuous data*
~* Convex curvature often ia celled upward or subliaeer ourv*ture and concave
curvature called downward or supralinear (OTA, 1OTU Slckis and Krewski,
19§5). One can find, however, an extrapolation to very low (or high)
doses thee eotbibies co«iceve curvature described as sublioear {VRC/NA3,
19&3), end eves the co«bicietiOA tens "eo«c*ve ufMard" (IIC/1M, 1975).
VIII-3
-------
Typical Dow-R«sponw fcefatSoncMp
Ob*rv#4 for Many Effect*
too
r
^ r Region of
Experiment!
~ Data
w
•r
Ajymptotic
Extrapolation
to Origin
<«!»' w. " ¦ > 4»'W
0 Ckrte
Smooth Curve Extrapolation
to Low Dots
100
mtb
Bodcgiaund
ThmhoU
Lx/ifi
Typfeal Threshold Effactt
Without
lackgrtund
Do*
Family of Oe^-'fttipcrita Co/vm
for Multiple Effects of a Ote«ical
100
*
i
i
£
. D
Background
Incidence
y |nBlW)V«
r
Ea.—.1- j a.
jl
1 t I 1 L
Q Dots
Typcaol Background Effect
?o Do'
Do*»
fromfocwuiMMi of Cvrwt Shop# wittt
Background end Thru hold Levels
fifwf* VIII-I - fctlatioashipi for
laalcH Effects of Chemicals
Soercet Hidveec fcwtmrch iMtitvte
VIII-*
-------
points (I.e., bttveect iMui 16 and <842 raaponae) art usually adequate for
eseiiaaciag LDso- Th« appearance of a given plot of data will d*p«fid on the
dosage scale uaad and alio on factora apacific eo the t«tt( such as presence
of a threshold or backgroundt *¦ will be discussed below. An essentially
linear relationship passing through the origin i« one posaibikity.
lo general, txpaaufc to a chenical will cause not just one but
several types of response as the exposure lev«l ia increased. A family of
dose~respouse curve* could be developed, as shown graphically i-o Fig-
ure VTIl-l*. Ac lov doieit tile response ti|hc he b(tn«fi
-------
risk jfiiiiuiic of environmental enfowces co chestical&J (a) sensitivity. co
dose; Cb) b«ckirown4 effects: and (c) threshold In addition, tem-
poral effects cm influence Che nature of che obaarved dose-response relation-
ship. Each of these effects it discussed brittLy below and Uitn ilU$trtC4d
by che results of a major study of 4 carcinogen, Lira CD0l study,
• Seaaiciyity of risk ta dose: In cases where the asymptotic low
doae nlationship tMlds, the reapoos* (risk) changes too rapidly fro« about II
or W*"*1 to 10~l® ov«r • 4ma 11 doit range co panic quantitative «stiaacion at
a giv«a tow dose with tonfideflce. tfethane?icai. expressions that fit the uper-
iaental data have proposed for such extrapolations, but they also have
weaknesses f as mil he discussed in Section VIII.S.
• Be,t the
background is subtracted out of tba data* Practically* howeverr high back-
ground incidence for a given response in either the study population ( parti c\r»
larly nonhuman subjects) or the potentially expoaed husas population will
substantially Increase the uncertainty lo estimates 0£ increased risk at low
dose. In addition, mathaaatical extrapolations to low dose can depend on
whether the background ia considered to be independent of or additive Co tit*
response of the test substance.
threshold effect! Ia aeny cases cbe relationship apparently
does not hold because a ¦viniaum
-------
threshold* for cancer «nd so®e other responses.* E*p«rtnwnc«liy» the exis-
tence of a threshold cm be neither proved nor disproved in toxicologic*!
studies of population* vich a distribution of s given response. The number of
test anlauLl* needad to acquire each significant new data peine simply become*
too Urge *c decreasingly law doses (e.g., thousands or tana of thousands of
animals) to d»t«r*Lat if « few wjr sensitive individuals exist. Whether or
not a crae threshold exists io a heterogeneous population for a g£v®n effect
is a Meter of filch based on chut rationale used.**
Belief ehae thresholds should hoc exist for carcinogeos becjuie
aebedded in the so-called Delaney Clause of tha tfSl Amendments to the Food«
Drug and Cosmetics Act. Recently, reconsideration of tha range of biological
origins of cancer b«ve led Co suggestions that wtil.it threshold doses might not
exist for stwte carcinogens, thresholds might exist for otbers« in particular.
Chode associated with bladder or Chyroid tumors. Carcinogens that act through
"epigenetic" neehanisns (e.g.* vj* foanacion of bledder stones) were viewed in
one scad? as more likely to have thresholds then those causing somatic e»ut ac-
tions (genotonic mechanisas), although the data were not conclusive (OTA,
1981). In 1955, however, an expert review (OSTP, 198$) noted that a chanical
chac ottlj causes cancer secondarily to a gross physiologies! effect is 1Ikely
to have « threshold «c jama dose level below chat which cmusei the physio-
logical affect.
Froa a risk «aaagfnc viewpoint, belief that a threshold dose
exiici for a given chaaucal greatly aioplifies regulation: the threshold is
divided by s safety factor (e.g., 10, 100, or 1,000) that reflects the confi~
denes on a has ia the data base and the quotient is set as tha standard of
acceptable exposure. Thus occupational exposures to a great many chemicals
arc regulated under threshold limit values (TL7s) (ACGIK, i960. 1983). Levels
of many chemicals in food products are regulated under the concept of accept-
able daily intake (ADD, introduced by Che U.S. toed and Drug Administ rat ion
in 1954 using a 100-fold margin of safety (Lehaan and Fltzhugkt, 1954). With
faw exceptiona, regulation of noncArcinogenic chenicals to date havts bew
based on risk assessment that assuase thresholds existed, whereas regulation of
carcinogens has not (Moreau and Anderson, 1980)
* if the biological rationale precludes beliaf in a threshold dote for
response, a chreahold of regulatory condemn can exist, and lead to Che
setting of a "virtually sale dose" (YSD) or tolerance level, e.g.,
aflatoxJLa in natural fooda. VSDa obtained by siaeheatatical extrapolation
¦odels can be aa little aa ana eiilllonth of the no (or lowest) observed
affect level ia a study.
** Brown (1976) discussed the Mthaaacical aspects of the threshold concepC ia
dot*T«tpone* stadias of carcinogens. Veiaberg (1983) has noted the
trens-seienci f i c nature of regulating under conditions "beyond
demonstrable affect.*1
~~~Hjk assessments for regulating carcinogens have usually involved
extrapolations to very low doeer on a nonintercept lag~log scale, as
discussed in Subsection VI.g.2.
W1M
-------
Fro»s « ri*k assesswaac viewpoint4 belief in a threshold dose >»•
plifies the *n«iy*ii m dosit clearly belov threshold (i.e., the risk is
asstused to be eero), but it does not eliminate the sejisltivicj probLea cited
ebove for ainple extrapolation to the origin. Since chee»hol4« art usually
determined by extrapolation fron low doM data vich sub«t*nct*l uncertainty,
the threshold dose itself is uncertain, flightly above threshold, the asymp-
totic oicure of the cur-re makes retponse relatively iniMiicivi so dtiti and
r|»k prediction difficult at 4 given dose. Ac che threshold, cbe risk Changs¦
rapidly over a snail dots froa about 10~* co 10 In addition» thresholds
c«& very substantially between species end also between individuals, depending
en their genetic ulKup end ebteir general health at tines of exposure.
Threshold* could be sensitive to synergistic effects. These factors all
Increase the uncertainty 1a estimating risk, when environmental doses are in
the threshold region.
4s indicated above, both a background incidence and a threshold
effect for Increase above bee kg round could be observed in che sajse data. In
practice, a threshold may be eliminated II the toxic eechanisa of the teat
substance is similar to that causing Ike background, or a threshold may be
revealed only if studies are made on populetloos with negligible background
incidence. In essence, che basic shape of the carve is fairly constanc t but
the background and threshold levels of the particular chemical and population
detertdnes cbe low-end cutoff point and thus determine* how Mich of che loir-
end texl is observable (see Figure VIZI-IP).
• Temporal tlfiaw Three kind* of temporal effects cae icJTlu-
sncfl che rttulci observed m a toiltltf stuiy* Tba first U that tone
responses occur to a significant ascent only after nearly continuoua long-tern
exposures. Such responses nay be missed In shorter studies, and possibly even
in chronic studies at doses so high that the subjects die early pf other
effects. Results of good chronic studies (see subsection 4 J are essential for
risk assessment. The second effect involves latent response. So* effects,
notably cencer, can occur long after exposure has occurred or even ceased.
Latency periods of 10 co 2J years or even longer have been suggested for socte
human carcinogens. Hence, results of iwd chronic studies are again critical
to assessing risks. However, the observed slope of the curve» background
incidenaa, and minion effect levels, cma very with the study period.*1 The
third affect Involves heritable genetic change in the exposed population,
i.e., mutagenic effects* Studies of at least two generations following expo-
sure are usually needed to assess swsealian mutagenic risks confidently. The
effects an subsequent* generations ere generally e fraction of the effect for
the first generation.
• Illustrative data! The £Dai study is the largest ever made of
e carcinogen. The chemical «as 2-ee-etylestioof luorena (2~AAP) , a potent
bledder and liver carcinogen. Over 24,000 nice allocated to 81 different
treatment groups were dosedi in feed at seven exposure levels (30 to 150 ppn
pluj an undosed control) of 2-AAF until aacrifioed and examined. Croups were
* Incidence of so®e naturally occurring cancers is reported to Increase at
approximately the 4th power of age.
¥111-10
-------
i«cri£ic*d at etcher 111, 24, or 33 months, Subgroup* of flics were doj«d only
9, 12, or 15 nonths and then sacrificed at either 10 or 24 month*. Tha study
vat designed co etciuic precisely ch« effective dose Cue) producing a It
cuawr rate In the met (bancs, the ium E0a1). Tha study, Its results, tod
analysis virt published in a collected series of papers (Staffa and Hehl.man,
IW0)S and in other publications.
IlluDCreciv* results Cron che fpfll study (titclefield et a\., 1980)
irt thova in Figure and VIII-2B for bladder |dd liver Q«OpLj»ffl». Id
brief, etas bladder cancer results shoved essentially oo background,
increasing incidence wich dose and Exposure period (age at sacrifice). and a
claar Kininu* observed effete level of about 45 pps. The anchors say, hOv-
tvar, that the cocal reaultg ara cooaiscant vich a "ao-cbrashold concepc" of
cancer. The liver cancer data, in concrest, showed greater variation# At 18-
aonth exposures, bacground we* negligible, and incidence increased slowly
with dose; but at 24- or 33-«cocb exposures, tbe background and the Incidence
ratea increased drsmatiullf. The 33-flonch exposure daca illustrate che
uncertainties in aJhrouic studies at Low doaa irith high background response:
tha icatctr in the 30-, 35-, and 45-ppo doaa data point* occurred even, though
ovar 40 Bica wera in aach group.
-------
33 Mo.
Sacrifice
24 Mo.
Sacrifice
IB Mo.
Sacrifice
303545 60 75 100
OOM (pfHti)
A - Bladder Meaplotffli
33 Mo,
Sacrifice
C«^pn\!l« |>M«
OuM f»U*\
[
24 Mo.
Sacrifi ce
IB Mo.
Sacrifice
303545 60 75 100
Dote (ppm)
ft - Uver NeopJovn*
Source; Adepted by MM from da la o« the ED^j *tudy in Llttlef felti et »I. (i960)
(ciuryei fitted by i&etp«ctioti),
fill-? - Dasc-Reeponee Data lor Carcinogenic Effect* of 2-Ac«tyl4U»lnofluoreae
la Mice
-------
A. Data tomrcM and quality: The data needed foe a health risk
analysis vill depend on the specific natfcure of the chemicals present, th* type
of «dvex»e responses tb#y cm»ef and the n#curce of the exposure conditions.
One would like to have descriptive information and quantitative d<»8«~r«.»pottii
daca oo the range of health effects Ln of the subject chemicals ueder
controlled condition* closely approximating those predicted to occur. In
practice, however, one rarely ha* «utH information, and must resort co ilear-
oative approaches baaed oo the data chac are available and the Methods or
models for using tuck data in heslth-eifects estimation. Animal seudlm* are
usually *a acceptable first altsrmative for toxicity daca. II adequate daca
are available from neither huaum studima nor animal studies, one must look for
other, ten § desirable alternative*. Each of these three kinds of daca coerce*
1« described briefly belou.
a. Human studies: Determination of riik factor® for many
kind* of technologically related thrmats is greatly Aided If he«lttr-*ff«ccs
information on humans is available. Data on humans are obtained by three
approaches: direct experiment, clinic*! observation, and epidemiological
studies.
Human experimentation; The use of human subjects in con-
trolled studies la licxited by ethical considerations to tests in which one can
be confident that no serious or irreversible effects irill occur. For example,
human subject* eould oot be used to test the potential carcinogenicity of a
substance. A substantial number of situations do require human-subjert test-
ing liter Qch^r i;escs ha vet demonstret ed the general safety of the procedures
(e.g., pharmaceutical testing, biomedical eogiemering applications, and the
ttational Aeronautics and Space Administration a aimed space prof ram). These
are controlled clinical studies which include laboratory analyses of physio-
logical endpoiots. Experimental laboratory studies on humatvs are more likely
to be for less toxic substances or lover exposure levels; they usually address
Less severe effects such as skin, eye or bronchial irritation, and organo-
leptic effects *
Data from human experiments may be used in risk assessment sim-
ilarly to data from animal toxicology experiments. However, human experiments
generally provide "Ho Observed Adverse Effect LeveLs" (ifOACLi), "Lowest
Observed Adverse Effect Levels" (10AIS4), or "Frepk Effect Levels" (fELe)r
rather than fall dossr-xeepease date. The major limitation in Crying t-o use
human experimentation data in health risk analysis of haaardous waste is that
few of the chemicals of intereat have been tested in a controlled relevant
maimer*
• Cliiilcst eu» reports» la addition to controlled experi-
ments with husaan subjects, much usefuL information for determining human risks
is developed through clinical inveetltation and observation of persons who
have bean unintentionally-—and oft an exce s j ively—exposed to a health haaard*
An anaLysis of case reports on such observations can yield useful qualitative
information, such M identification of eodpoiots ac high axposuras. The liir*
ications of chase data* however, are chat exposures were uncontrolled, usually
unknown, and seldom of the lenf-t«rnt low level necure that are of most con-
cern with hazardous wastes. When animal studies oisc, clinical case reports
TZZZ-L3
-------
provide verification of the adverse endpoiot la humeos. Reports of 4 series
of cases, often found in medical journals, provide stronger qualitative
support of the type end degree of adverse effect. bat again usually no quanti-
tative. Lnfonmtian, OecMi
-------
ratio of cbe race Ln the exposed population co the rate in che otiwsposwd
population* This rat* ratio csn consist of Incidence or prevalence r«cea
(r«l#civ« risk), adjusced aurtalicy rates (e.g., Standardized Mortality
Ratio), or the ric« of exposure £a case* relative to that of controls (odds
ratio). A ratio of unity suggests no between cha exposure and the
effaces. When the ratio* do show an effect from exposure* the Incidence rates
fro* the study are suitable for oac in quantitative risk assessment. Preva-
lence ar mortality rates can be transformed to incidence rates when necessary.
Epidemiological studies have several inherent liieitations. One
is th»c increases of coaxial? occurring heelck effects (such ti cancer of the
lutt* or colon) usually arc not detected unless che change is very great.
Another limitation is that exposures were hoc cortCTotte4, a ad are usually
poorly known. In fact, many of cha huaran data are obtained from industrial
occupational studies is which sample sixe it liaiced by the number exposed,
and exposure levels are determined by the circuascaoces of the industrial *«c-
ting. Because of these ILai cations, a study uy yield only a siogle or m few
risk ratios that provide *a isolated effect level, Such data do not yield a
dgse~respon.se curve, and probebly do not indicate how far the effect level is
abo^e the threshold level, if any. A further inherent limitation of epideai"*
oIogy studies is chat chary nay fail co reveal a true adverse effect due to
insufficient data for son* reason. Uhen the data collection procedures aod
analysis are judged co be adequate, the determination of the validity of a
negative study is largely statistical. CuidelLnes exist for distinguishing
between valid and equivocal negative uuiiei and- for using isolaced effect
levels. Finally epidemiological studies stay establish correlations, but may
net be able to demonstrate a causa and effect relationship,
Despice these limitations, epidemiological resale* can b« use-
ful la quantitative risk isaisnent. Basulca of a conclusive epidemiology
study are likely to be the best date available. Even if the results of an
epidemiological study do not provide a dose-response relationship or are not
conclusive in demonstrating che risk or absence of risk posed by a given
agent, they can complement Che results of other studies (e.g., animal toxicc—
logical Ceiti, Clinical case report*, actuarial analysis) and any be the
deciding fecCor (particularly if they confirm other evidence) in reaching a
regulatory decision for potential sources of a disease. See for example the
review by Crouch and Wilson (1979) and specific studies on ethylene dihromide
(testacy et al», 19?S> and viayl chloride tCAring et al., 1979). Macti*
(1986) suggest* chat "molecular epidemiology techniques "—combinations of
analytical epidemiology with advanced biochemical methods—* hold promise for
quantitative assessment of a broad space run of human healch risks.
b* Imlaal studies: The traditional tteriea of tomicological
¦cudias with laboratory aniaals classified by langch of exposure eonsiscs of
acute (sometimes called single doae) studieat sabchronic (sometimes called
repeaced dose, subacute, or short-tens,* studies), chronic (often called long-
cons studies), and special studies (Doull et al, 1980).
* Ghfortumately, Che microbiological and biochemical in viero tests developed
In recent years are also often called short-tens studies.
VXXX-1S
-------
Acute studies jsmlly involve a single dose Qr, for mhjlaitont
a iingl« exposure of up to i4 nr. These studies use relaciveLy high doics.
Effects addressed in acute studies incl-j4« death, imtatiflfl, And other rela-
tively gross consequences. Acute toxicity bias traditionally been about Che
first tuxicalofical property to be studi ed for a chemical, and the LD^a is
usdilly reported in Che literature for studied compound!»
Subcbronic studies range f ron chose uiing « few refuted doses
over a few days tft those lasting 50 days (or soout 102 of the aubject ' s life-
time). Subchronic scudi.es usually look foe more subtle effects than do acute
studies. such as chinges in clinical chemistry values and microscopic tissue
pathology( racher than lethality. Subchconic studies can yield quant itat ive
dose-response relationships, but often arc used to dutermin« the range of
appropriate exposures f or chronic studies -
Chronic studies involve repeated and prolonged dosing for peri-
ods approaching lifetinwi (typical Iy 2 years to rodents and 7 years in dogs).
Chronic studies detect effects that are cumulative or have a Latency period,
and effects of bioaccuaulated toxicants • The end points of chronic studies
can be nominally the same at those ot acute studies, but very of ten are dif-
ferent. Multiple dose lev«lj irt used; a study should ident ify target organs
and tissues, the range of effects, dose levels uhore g wen effects are not
observed and first observed, and the frequency and leverity at increasing
dose»
Ehling (1918) recently discussed Che quantification of genetic
riik of •nvironaimttl outi|iBi, including the "direct," doubling dos# and
genetic number nethods for Mendelian mutants, chromosotne aberration methods,
and methods for irrtgularLy inherited disorders.
Special scudi es are designed to look at particular endpomts.
metabolic systems or unusual situations. The most comon are teratojenesia
studies (looking for malformed fetuses) and reproductive studies [looking at.
fetotoxieicy, reproduction and survival races. and other developmental
effects). Effects nay be 90qitored for a lifetime, or in the second and third
generation following exposure.
In coxicologicaL studies, an effect will not be observed unless
there it a receptive subject (e.g., pregnant feaale for teratogenesis*) and
the effect it syatatnatically sought (e.g., appropriate exposure period for
carcinogenesis). One can deoonstrate a specific adverse effect, but it is not
possible Co deoonscrate the coaplece abaance of adverse effects (i.e., estab-
lish coitplete safety), since true effecte may not have been observed in sta-
tistically or biologically significant nutabers, or say have b-een overlooked.
* Teratogenic effects occur only uhet) the female is exposed to the teratogen
during a critical period during the gestation period—usually only a few
days to a few week*, depending on the species and the effect. The
critical. period it utually early in the gestation (first trimester for
bunans).
¥111-14
-------
to tdt^uCt <•( of experim&ecal data will consist of one
or t*K> riLiible studies each of icuti, subchrcnic, and chronic qutncified
exposure* to the eheadcal ia en appropriate ioiiag nod* (e.g., by inhalation
for air exposure, tor by savage, In feed» or ia wacn* for ingestion. A conic
respotvaa will have bean studied tc a minimum of tbrM doss levels plus j cem-
troi group. Ideally, the Lowest dose tuttd «ill have negligible effects, one
oc aore doses will have definite toxicity but no lethality, And one or nor®
high doses will product the substance's full array of toxic tfficti including
substantial lethality. It i< helpful if the total dat* bia« [or the a harm ca I
also includes evidence chat the efftcd would occur in hustana. Inadequate
anistei data §
-------
• Acwmta knowledge of the dosage utsad foe each te»t is
extreaely inporeant, but ii often difficult to «cWew»
Tha ectual itocdgi dilivarid to an animal nay not occii-
¦arily be the iotttd*d dosaga, in inhalation
sc,u4i*« of eoagacea boc alto in other ceiti with tJva
coxicuat dispersed in. f««d, water. or air. for exampLe,
• cm* caat Mtifiil My decompose or be otherwise I oat
during chu« study. Equally liriportanc:, at tioe* the afcr-
sorbed dose »ay be 9 igjuficancly lees than the exposure
dose.
Negative dec* (no adverse effects observed) arc not with-
out imcecteiftCyJ 4 particular adverse effect may be sc«-
tlscieally unobserved under the case conditions, or if ic
v«« hoc loaned for, ic was cot likaly eo be reported, even
if present* This ia. especially true for «fface* requiring
special ceacs for detection* »uch at clloieai che»v»cry or
hiatopathology.
c. Other sources: If adequate data a re unavailable from
slcliBr huAuo or flfllsiaL <§turdi«i»other data sources *ay still provide uj®£uI
information for risk iiiMMNGt.
Such information could include to*icokift^cie data on tha cheat-
iceta, toxieologicel data £m humane or tnioeli on minced kinds of chaaicals
(cognate*), or data on tha cheaieal of concert fro* one of the variety of
larf« in vitro btdaeiaj studies* chat hew* bet® developed over the laat two
docee years. The** era short-cera tenca that uia sxcraorgacisma, call oil*
Carat or biodHtfltlcal ay)ceats; they provide information, about a chanical's
effeet*, particularly smtegenesis end, by extrapolation, carcinogenesis* Tha
dose-response fuoccioao provided by these casts are difficult co extrapolate
Co effect* in aaenals. Juch studies are widely used in prioritizing chesucels
tor cbronict casting, bat have cot been accepted ma « general aubatituca For
iMoouilian studies* Tha potential use of atudiea of these kinds it discussed
further undar Section 1, "Predictive Models*"
d. Judging evidence of carcinogenicity: Carcinogenesis it an
effect of particuLar uvortance, but atudiaa ot a |iven chaaical often report
different results depending on the specie*, aex, and exposure conditions* The
tnceraeciooel Agency for Research on Cancer haa developed a weight-of-evidence
ayace* for judging information on a chenicai's carcinogenicity (IMC, 1983).
Evidence from both hiwean and animal studies wei characterised as either suffi-
cient, lUictd, or inad«quater and the chaaical waa clasaified overall in one
of five categoriesi
1* Sufficient evidence*-'Hal ignaoe t users (a) in nulciple
apeciea or strains, (b) in nultiple experiaeata; to ao unusual degree
regarding incldaacei lici, or type.
~ LitereUy "In gl<*«" c«ic».
fni-i8
-------
2. Limited evidcncc~~SQggcstive date (i) from a single
species, atraid, or experiment! (b) from studies »itli liaitations in editing
procedures; and (c) b*$e4 cm oeopLesna. that tend CO occur often *pont*neo«»ly
or are hard to classify e* nuiignaat..
3. lo*4«^tt«ce evideaa*r*-Reported studies have amjor liaica-
cioos that precLude confident interpretations,
4® Htgtcivt *vidense—SCudlei report tb*c within lioici of
(•ict uied the chemical i« noc a aarcisiagea*
S. la data—Ko studies of the chsnical art available.
E?A propoiM tLight aio4ific«ei
-------
buaen MitlUi «ffecCi usually cu be identified with reasonable confidence if
toxicological test date in aaiaals or other orsaaiina «r> for cfc#
(•it cbetical or for evpiaii cbefticals (i.e., other cbanicels with ii«ilir
cbeadesl iiwctw* aad phyticoebeaical properties), or If covicoklaeTlc
{pta*mc^ia«cic> information it available on the c heart calt* absorbability,
nobility, and biatrantforautiona within the tuoaan body Or Other appropriate
t pedes.
The primary need ia to prHice Che d«|r«i of the of fact at cooctm
at eap««te4 exposure ccndUions based on the'IIffereac MtU) of exposure com"
dltiona reported ia tba etediet found in the 1 iterator*. Theae predictions
tlaest inevitably will r^uirc tba extrapolation of a doteretponte function
aeroas t«e or aore tat a of eMiUiooa. Tba to include extrapolations acrata
exposure routes, freqaenciee, ceatiaaitiaa* duratioes, and dosages; extrapo-
lations froai tatt ipaeiai to honest I aad extrapolations froai tonic® lexical
toac daea on individual chaadealft to tba exposures to coaplex fixtures of
chemicals frequently found ia tuurtoui Mcitu. A considerable literature
diacusaet and coaptrai methods for aaking Mck extrepolet ions ander various
circumstances mtxt tha uncart a int lee inharant ia thea. Although critical
review of thia literature La beyond tba scope of tha praaant discussion, tba
overview below mmy bm helpful ttt tha rudar.
«• ****** routes» Tha toxicity of a
given chemical aury vary fvbtcaacUily~ of tba acaling
factors or convenient rules of tbiuab that toxicolofiatt havo davalopad for
eo^iriai i<»M by 4iffaraait meal. Furttar laiprnvaawtt tbould ba poaaibla
by c«ati4arlfl| m tba natora of tka toxic of fact tba pvopar*
tla« of tba ehaadcai.
b- . caxic rwpme to
a given efc^tcal vall^mvTwltK^Eazra^^vyrtaaniaUy, mM leval of espo-
aura. Miki# aNttabolita eaa differ «c different date leveia, as discwiaad
by Cabriat at al. Clftll «a<| a'lTlahcrt.y (1JI5K a certvastclonal concrollad
eoxicolofy or epidaaioiefy a(i*dy asttdlXy^ yial4a a dttfnreaponta raiationahip
Uka cqb of thoaa ahoim prevlottsly lit Figure Vlll-l. Cxtrepolatlona and
interpolatloox are relatively fUtple it exteative ceotrollad doae-reaponae
data are availaJble aod tbt aaviroea«ncai expaeura U ia tba mm general
range* The caavaraioD mf be a*4e fray^ieeUy ar by fitting a curve to tba
data aatluBatlcallr, aad tkan MtevtacUic aa^MCatf response at kJm dose of
iAeavaec. T«o apaeiai freblaaa mf «co#ri (a> cba aapectt of tba
VU1-30
-------
expected tzfmntm mf be different from thorn uaedl La the controlled studies;
and (b) zhm 1avals evpested Mr tot mil below dotes used in the con-
trolled study.
(1) Temporal Mtrapolitiam.' If the frequency, contino-
icy, o* duration of the sxpected exposure differs from the frequencyr continu-
ity, or duretion used Ln the controlled studies, the health effects mey be
quit* different ~ even if the total doses seen to be cvufhlj equivalent- For
relatively small differences the tiar-veigbCed #vej?af« of clue expoaur* may be
used as * first approximation to estimate the e*f«BC«4 response frots the test
data. Time-weighted averages have been used with good results La many appli-
cation*. An i«pro»f^ estimate uy ba possible by considering information on
the specific chemical and che affect of concern. For exaisple, the chenical
my be aetabolixed r«fl4ly or may tend to accumulate in certein organs; the
effect say be readily repeired by tfe* body, or injury may be cumulative. tn
such cases* one ftay defenaibly adjust the simple prorating method Co five an
"effective dose."
Extrapolation of che results from short-term studies to
long-term expo turret is a difficult step because, among other pcoblexis . aoine
effects (e.g., carcinogenesis) are seen only in longer t«m studies (see
Schneiderttan, 1981, Griffin at al., 1981 and Bertsberg end Dfivrson, 1983; and
Mertxberg, 1584), Conversely, certain responses sometime* can be determined
better with acute or subchrontc rather then chronic exposures. For example,
teratogenicity ii usually best determined with e few repeated doses during
early pregnancy. Some research resales suggest thac che chronic toxicity of a
chemical can be partially estimated from aubehronic and acute tonicity tests
on it, and Croat general toxieelotieal principles. These results are worth
noting. This subject Is discus*** further later in this section.
(2) High dnse to lov do«a extrapolations: As noted la
Section A, quantitative extrapolation at low doses can be difficult, both for
relationships that exhibit threshold and those that do not. A. particularly
difficult problem occurs if the expected exposure is below an obaervad "mini-
mus effect level'* in the controlled studies of a carcinogen. Argumeaca have
been made on biological and statistiaal grounds that threshold level* do not
exist for exposures to chemical carcinogens,* that any exposure, no auartar how
small, poses some small ri«fc of cancer. Similar argument a could be (but gen-
eral If have not been) sutde in regard to the occurrence of mutagenic, terato-
genic sad assy other affects at very lov doses. glological rationales have
been suggested in support of a threshold level for tone effects (Cornfield,
1977). Aa noted previously, however, a conclusion on a threshold's existence
for most morbidity and mortality iffi«:9 is largely a matter of faith (or
probabilistic degree of belief) in tfce rationale.
~ This eheory was first based on a similar theory developed la the 1950s
regardins the carcinoge&ici ty of ionising radiations, particularly
radioactive nuclides* Sxees* exposure to X-rays use considered to cause
cancer as early as 1902 (CAQ, 1982).
YltZ-ll
-------
for carcinogens, mathematical models have been excanslvmly
studied ta aid* In predicting cancer cislw at v«ry low doses, as will be
described in Subsection 3.2. These noijels «ey be useful for och«r non-
threshold if(«cd. for effects foe which « threshold is assumed» risks down
Co threshold can be estimated by grspbica-l or i*ath*aac.icai model* incorporat-
ing a specified threshold, to addition, several other predictive approaches
have been reported, including) toxleokin-etic nodels, quantitative icmcturt-
activity relationships (QSAA), prototype relative patency methods, oonpsra-
ster ic methods, and extrapolations from in vitro test data with isic coo rgan is/is
and cell cultures. These other predictive methods are also discussed later in
this section.
e. Predictions across ipicifff So rigorous scientific basis
e*i»ts for general quantitative extrapolation of aninai test dace results to
turnout, and mmy problem exist in *Mklng predictions. Such extrapolations
require naking two kinds of assusptionsk (a) thet a method is Available for
determining equivalent doses in two species, and (hi that « eethod is avail-
able for determining the response in the two speciej at equivalent doses.
Cl) Posajre conversions* The dose units used should
enhance interspecies comparison* and be readily calculated frots available
data* if not used explicitly in the ociginei study. Two type* of does units
ars cosBonly used in the literature!
• Concentrati«o-in-»ediuni (e.g., parte per esilllon in
feed, eg/a* air)*
• Quantity per anLsel on either a weight basis (e.g.,
¦g/kg, assoies/lqi), or surface area basis (e.g.,
«g/n2, wtoles/**).
Ail are used in varloua studies, and each can be converted relatively easily
to the others, with proper conversion factors. Therefore* the units of choice
should be baaed on their biologic*! usefulness. The basic biological phenowe-
non involved is the reaction of the toxicant molecule with a biological woLe-
culs {usually called a receptor) located at an active site, initiating a
series of reactions, ultimately reaultiag in an observable effect. As
described in no re detail by Gilsan at ai. (1980), this sequence seems to Fol~
Lev cbesiical ness~actioo laws, se choice should be based on the characterise
tics of the receptor reactions.
The c«ftcencr*ci6a**iir'aaditM nee sure he a been found aesc
useful in toxicology for situations iavolviivg a direct contact between the
nediiUB and the receptor tissun, «.gM irritation of skin by liquids or of
respiratory organs by gasea. In risk eaa«4S
-------
With «ost bioioficmi »f»tensr Che it taction ii wore com-
plicated, tinea the cheaicel must enter clue body
-------
TABLE VI11-2
USEFUl FACTORS Mi INTERSPECIES COMPARISON
I
M
Species
Weight
m
¦f/k«
Surface Area
(.«)
M/«3
Food/Wi(ht
Fcartioo*
Valer/tfotftht
FracUoa
Reaplration
i«LC
l.i lei iiae
_ ifkfcSL
Mouse
0.030
0.34©
0.0102
2.94
o.n
0.1?
0.04
550
liiiCAC
i. 121
0.214
0.026*
4.12
0.11
0.48
910
fat.
o.as
0.150
0,0525
6.61
0.05
0.076
0.24
lt«
Cuinca pig
®.40
0.114
0.057S
1.0
0.028
0.34
0.07
730
l«W»it
2.00
0.0B4
0.168
11.9
0.049
8.ill
1.6
2,000
Cut
3.00
0.073
§.220
13. 6
0.030
0.09?
l. 5
5,100
(imk-m
4.00
0.067
0.267
15.0
0.042
0.13
1,3
5,500
B«t
12.0
§,©46
§.55S
21.6
1.021
0.025
1.5
1,500
llu»an
70 §
0,026
1 800
31
0.021
0.030
20
25,600
Sources'; Dttrkla H9B2); EM (1910*) (1963); *ad HRI c#lcwla^iow. Th* litter ttsunti
•* = 0.106 (wt in fcf)2^3.
£ Wet vmigkl of feed continent each day dlvldei by body weight.
Drinking <#«ter cooiwetf each day divided by body might.
Cubic meter* of *ir per «§§y.
-------
high u^iurei, but Uttlo if mf qoaoticociv* ioforttatioft m doaerespoaoe
function ac Moderate and l*« #•••».
7h« npnriMDt* Btcmwiri co obtain huMa 4*t« are usually
precluded by ethical considerations and l«t*L *tttrictlOfts> ftasuits of is**
cpidamola|ie*L studies «re available Uaclodiaj eecupotiondl, eovirttcuRontal,
and csnuatr exposures) on the tflteti ef iom cbanicals an huaant, hue these
o»ttraa.
A fev recent ttudioa eoesidcr tb* probl** §f mtmurm in a one
iotail* Tha Vatlooal taaaarch Caitacil CBA$/*tCt 1910a) attoa^itad to asaasa
tha hasarda af exaosora to sultlplt «haadeaLa U c Btriiia anviroaMot, In a
later report OC/lil, I9t2)» the €oua»eil publiahod a sya^nslaai on the state
af tlM art ia aaaeaaaHMit of ¦ftlticka^lcal cmUiiMtiM, tiilcli f«s»ii«a a |Ml
sugary of ch« LimtM of aaitho4a for •inty of Mumttwmmmt aipa*
itirti.
VXXX-2S
-------
The EPAf* |a*irooaiiflC4l Crincia aoi Assessment Office (ECA0)
In Cincinnati has bees developing « amleicLhtDQticil heaLth ri#k incsnuut «eth-
odology which cm be ufe4 In conduce ins sice-spatcific risk atsessawmcs, on
hazardous ««it« disposal facilities* Proceedings of to ECAO workshop held In
Ciociimaci, Ohio, la 1982, are available (CPA* 1984®), » The U.3. Department of tnergy's
Health aad Environmental Risk Mgmly%is Pcagraa aotvductad a workshop oa riska
from mixture* of chemicals. A suMaery of the workshop (DOE, tfS3) concluded
that the relative potency of « ixj rariou* animal sysceaj w#s useful
io estuaacLtig huaan risk* It. recoaaaended the ua« of Multiple short-Cerm cescs
of nuny cheadcals Mid mixtures to estimate thair relative potencies »s cArclft-
oijftEiii* in lieu of conducting the «ar« lengthy and costly loag-terai etic*
Mcetury Co determine potencies of each directly.
Chriscensen m4 Chen < 1985) have 4eriv*d and CMttd preliminar-
ily ooainteraotive multiple toxicity nodals for qu«ne«l response of organises
to two toxicants, using probit, logit and VeibulL crsaefftitutions for the tcl~
•runci distributions of each. Only Che oonnomally distributed Weibull Model
gave m acceptable fit Co eirp«riaN»taL data*
Fat «v«n t!» liarpleftt aixcura (oaXy two ch wall
aa baing cLarcinogenic if itaelf. In antagonist, ehc coabioed affect it leas
than cha turn of the iodividoai effects. Thia phcAoneaotx is often aaughc in
tha developaamt of aatidotas and other therapy AgainaC toxic effect#. With
• Caacar proaiotara arc soiaetuBea refenred Co u epigenetic agents. Sub-
stances which cauae irreversible ciuuigas theamelves» thersby iniciatiog a
carcinogenic process (initiators), are often referred to as gaciotoxic
agents. Sooe authorities have protested ttoc cha uaa of this terminology
should be li«ieed to dlacirvguishiag eK>de« of action, and should not bo
m«4 to etstslfy chiMaicali, aiaca « given chaatiaal nay act ac tioea by
titter acelcniM*
VI11-2I
-------
both nonaddLtive cases, the parameter of interest is che 4«|ree of interac-
tion; chit «ay be a constant or eay b« a function of itose.
If interactions csist» the health effects of sequential «a4
sieiultaneoui exposures are liltely Co differ. In addicioo, different {xntuca-
tions qf sequential exposure asy h»#vw different effects. Further, che
response variance «ay bo trtice* be eve en individuals for exposure to fixtures
then for exposure to a tingle; cheanceL.
In the absence of specific st®4£e»s one cannot predict which
-------
have ism tefia in early chaory of radiation effects. Recent discussions have
noted the iubAcaoetal differences Cas Mil as the sinilarlties) in urcino-
genesis by radiation and chacsicala, both in molecular mechanises (Borf, 1983)
and la cellular or nuadjajil models (Fry, 1985).
lis Bcchanisms of: carcinogenesis have been difficult to eiuci-
data and this difficulty accounts for thte sugfantiart and •vmluatloo of nafty
increasingly coesplex aeilels over cbe peat 10 y-Amt* (Uhiccanora and Keller,
1978} OTA, Wfi| OSTf, IWJi Irova, 19851 »«rg, WI5; and Fry, IMS). Currant
consensus is chat at least three nejor pb*ae* occur in tha development of can-
cart initiation; pnuantion or expression.; and progression. A given cturaical
M7 be an initiator, a promoter, or a costites carcinogen.
An initiator appeers to causa a rapid* irreversible heritable
change i_n a target DMA noleeule or Mil is a target organ. Tha reaction
appears to ta first order IdUtetleally oithout threshold. The chance can lie
latent (even long afcar cha chenical h*a baan eliminated) until promotion
occurs, unless cha call dias or is destroyed by bodily safety leechanisa*.
Pronation is lass wall understood, and nay hav« multiple forma or staffs
involving possibly both direct and indirect mechanisms. A promoter (of which
there are many*) appears Co alter th* differentiae ion capability of m
initiated ceil* possibly through adduce formation, amplifieatiod o£ daxaaged or
nocvel genes, or activation of repressed gefie»« A promoter' i effects are
susceptible to bodily repair aechanisms, and tha prostoter oust b# present on
en extended basis (repeated or ehronie exposure) to result in tumors. (Tumors
mj regress if exposure ends)* Ceiit thet have been Loitiated/prosiotad nay
9(ill or be coastrtiatd Iff the lisiui'i cell lyitw* In tat progreiAion
i(4|t tbe ceil apparently undergoes sufficienc genomic change (e.g., perhaps
through chromosomal translocations) thai; it largely escapes control by tha
surrounding tissue and proliferates unrestrained. Promoters siay at tLaes be
required to Auuncaia progressIon, lince tuoors will occasionally regress when
exposure to a given carcinogen is stopped~
la addition to onoercaintlee about the Bechaaisnis of a given
carcinogen, the subject orgajiisa ia also invariably beiag subjected to many
other canccr~Telated agents* including initiators or promoters (possible
cocarcinegeas) and also aoticarcioogeas. Ames <1983) has described a plethora
of natural Mutagens and carcinogens in the norewl diet. Many of these cHenr*
icAli act through generation of oxygen radical a which nay play a degenerative
roil In («acer, heart disease, and aging* The intake of these agents is cos-
pounded by the lifetime exposure (o naturally occurring radioisotopes and
iaceate cosnie rays* The diet alio containa chemicals believed to act as
«euc lewreiaogcae.
Considerations of cbe uncertainty in eha effective doae of the
e*rciiu>genv coupled with uncertainty in the carcinogenesis and bodily repair,
as partially summarised in Figure Vtl£«3, reveal wtoy apideeuological and
* Tha coanoo amino acids (*u«h as Crytoptoaa, leucine, and isoleucine),
present ia atcekeriAi
-------
a
a
i
e
ClMaiial li
ff-*- 1*-«t niY
1^1*\» •JZJJtfl"
©MA
»««!»
teach
OMA
-------
coaicologicaL studies often yield dose-response data that are difficult to
interpret with general mathematical models.
b. Doee-respoqae ^etapoiatigg modelsi Mathematical extra-
polation from high dose level* to low dos® levels is ueually required to
escieate human health risk. Typical. laboratory tonicity readies involve 10 ta
100 animals per dose level because of logistical limitations. The sensitivity
of the test I® arithmeticalIf Iindeed c» an efface of I co 101 incidence in
test an is*Is dos
-------
P(d) » a«d
where the cansnant, a, it the il@p€ of th
P(d) ¦ P(dQ) ~ *(d-dv)
where the effective or exesis* probabLlicy, Pe{d) in
Pe(d) » P(d> - ?(d0).
Thii nechod of carreccing the response for background is
satisfactory [or tiacy purposes, part icularly if the background incidence n
snail. latter companions can be: «iad« between response probabilities or risks
found in dit fereinc icy4iei, where differing (and s emetines high) background
Itviif ikiit, by eonv« - Hd )
!(„) . pfi j° ¦
Q
A scraighc line can be fit co a data, sec by regression
analysis and used for predictive purpose! , ificludittg extrapolat ion to estimate
the threshold dose or background incidence. Ptw data *et* Are linear over the
enci re ranger however, and nora coetpla* functions rmquirtd consideration. For
exanple, ch« four model® below have been taaced on the effects of iooixin&
radiation (S1C/NAS, 1980b; CAO, 1981}l
Square Root Kodei
F(d)
m
F(d0>
~
.4*
Quadratic Kodel
Hd)
m
H49)
~
Cubic Kodel
p(d)
m
H4J
~
#|dl
Linear Quadratic Nodal
P
-------
The lioitf no-threahoId model gave a reasonably good fit
to aoae data sets involving ctoctr incidence following radiation, particularly
for so-cAlie4 high~UTT radiation* suck a« elph* particles (helium nuclei),
prolans and fast neutrons twi cotaic r#y« (Heavy nuclei), but a poorer fit lor
law-UCt radiation Mta r«yj (tle«nms), gamaa rays., and X-raya..** These data
in away cases were based
-------
h*s 4 tolerance level which., if exceeded, «l««ya causes the effect, and on the
assumption th*c a discribucion of Mkriaou exists in the teat population.
Different distributions lead to U» probit end legistic ¦od«ls.* Tolerance
distribution models suggest easumpcicm of * threshold dose for ttch teat tub-
jecC, bat do not preclude vanishingIy a nail thresholds for s-pecific subjects*
Pre bit BOdel—The basic probit** model assumes « oooutl
distribution of ivmti. for toxicologiesI doae-response Application*, « iv%m
normal dietributioc of individual cal«r»ftc - tlw)-"' f & A
J n 7 du
vhere P(d) ¦ probability of effect at dose d
u ¦ log t, where * is the veriace of the classical bell-thaped
oor*ei distribution, and
a ¦ parameter CO be estimated, end
§ ¦ parameter to be estimeced (called the slope of the probi c
I La*), where B > Q*
Originally developed la drug development research to fit
research data for acute exposures, thus probit node! la very useful in obtain'*
l«| che L0JO by interpolation, since it ceo be applied with only we dose
level plus a control group. The model i« not highly flexible in fitting data,
but che log-probit model adequately fit the observed deta for both liver end
bladder canters in Che COqL study (Farmer ec el., 1980).
The probit model curve is convex in the 10 to 502 response
range, but the response approaches zero rapidly at low dose. lc is inherently
o« threshold, but is concave in eztrapclation to very loe dose oo a log-log
scale. It almost always yield* lowc estimates ef risk at low dose than do
other models.
Hancel and ftryen (1961) adapted Che probit model co aati-
ucb "$if« doaee" of n*tciIn chie n*c!u4r the parameter i waa see
estimated from the date, bat wee sat arbitrarily to unity (presumed to be a
conservative procedure). Other paruscett are then varied to §i»e an upper
99X confidence leveL on the risk at a given doe*. The safe dose is then
defined aa the dose expected at the §91 confidence level to give no more than
* The Wei bull model la MBKim grouped with these models, but is better
classified with mechanistic models as will be seen in Section B.2«a.3«
«* Probit is an acraoyn for the tarsi probability unit. The probit scale is
based on deviation* frocs the Man of nonsal distribution; the scale is
adjuated to avoid negative numbers.
fX11-13
-------
m assigned very low response, such as 10~* The U.S. food and &rug
Moiniscration Che log-probit model for • time around 1970 in support
of 1CJ regulation of carcinogens. This procedure was SubsequentLj improved
(Kane*I «C al.« 1975), and was specified by £PA in 1976 as one of two model*
(the alter being the one-hit node!) In ici interim guidelines for issessiflf
health risks of «>uptct«4 carcinogens. It has beer criticised, however, on
the theoretical grounds Chat it ruled out linearity et law do»« (a feature
viewed as ea®ential by Pete, 1974), give a poor lit CO data becnuee of its
concave curvature, end eventually underestimated tljln st«fl extrapolated co
vry low dopes {Cramp and Nesterment XfTf). la particular. It was not suf-
ficiently "conservative." In Edition, ch* version ef the prrobit modal chat
incorporated a background level of cancer iitplied that the mechanisms of back-
ground and dose-induced tuttors use* independent (Cranp ct al.. 197$; Crump,
1977 and 1979 J Hartley and Sialken, 1977; and Sal a burg, 19?9). This node I ia
no longer highly regarded for regulation ol carcinogens by the FDA or £?A, but
the Mancel-lryan concept of estrapolating upper confidence limits has been
used with other dose-response models.
In addition, Kattis (1987) notes that probit risk assess-
ment formulas uted for chlorine gas releases by different groups differed by
over a factor ef 10 in the UǤ for humaos, which Chen (because of the highly
nonlinear nature of this function) lad to a billion-fold difference in papula-
tion mortality risk in the region of the more conservative LD40,
Logistic "logit" model (developed about 1944)
uaat the log logistic distribution rather than the log normal diitritiution of
the log-probit model | otherwise the ctodels are similar, and the logit jieidi a
sijpsoid-shapmd curve alto. The equation ii:
P( 0
la lo» dose extrapolations, this modal can be linear |i ¦ I), sublioear (s >
1) or supralinear (B < 1). Thar logic nodal usually gives risk estimate* at
low dose somewhat higher than those of the probit model. The logic model is
used in mathematical models of many growth processes, but has not been applled
as catch as some of the other models in iealth risk assessment of chemicals.
In contrast, the eulcipl* logistic model of Truett at al. (1967) has become
the dominant model for analyeio ®£ cardiovascular disease. The equation is:
i n —,—JL_—__—__
~ Mi * ••• W
where 1 is the risk of developing a particular cardiovaacular disease over
tiati the is ara constants, and the Xa are the raw leveils of such risk factors
as aget bLood pwiitifi, cholesterol level, «tc.
VT 12*14
-------
fl) Stochastic mechanistic model*: Model* la this class
*11 Assume « decree a£ randomness in event* leading to rsjpooM, tent also have
some basis In toxicological ch«ocj. These models *11 tssuae that a certain
ftuabtr of retccinu, evencs, or "kits" (¦ tern used in radi»t ion carcino-
genesis theory) ate necessary between molecules Cor fragments of molecules) of
a toxic substance and * cell or molecule within • ceil of cite victim to pro-
duce the effects One s ho aid «wce that in the origin*! developsMuit of All
these aodels, response va» considered as a fu»ctioa of tineT but asit are now
¦ora familiar in their eticbotomoua dome-responae forms.
Onm-hit ani linear aatcrapolation mfrdmls—Th* onr^hit model
of carcinogenesis was propoaed by Iverson aad Ariey in 1952 CWIC, 19??a). It
«ssuaes that a single biologically effective dose reacting vxtb one receptor
site within a cell is adequate to cause a transition to a cancer eel I, which
then multiplies at a rate independent $£ Che initiating dose until a tumor li
detectable. A "bit* can be considered to be one Or more of a variety of pos-
sible fundamental biological events* within a specified Interval of time,
including, in Che extreme the reaction of i single toaicanc molecule with the
0®U of a single ceil in the organism*. If the number of hits in the interval
follows a generalised homogeneous Poissen process« then the aquation for the
probability of response lor an individual 1st
9(d) - I - « "m t
uhare 0d ¦ «*paet«i aumbar of hit* le dose d, (i > 0)
This la the one-bit modal, sometimes called the simple exponential model. Ac
very low doaaa, tha relationship becomes!
P<4> = id
Background response 1 avals can be taken into account in this and most other
models fry assuming that they are either in4tf«o4«4t or additive to the
response to the cast substance (Boel, 1980). The celeulatione differ^ but in
general:
MO • 1 - . "Ca * M)
* If the response to etiaaiai is assumed to be indepeadettt of the background*
Abbott's correction is used. l£ tha stimulus is assumed co have a mech-
anism similar co that causing thm background. thea it is added to an
assumed effective background dose. Extra risk calculated at Low dose by
tha two mathods can differ by several orders of magnitude.
VXXI-J5
-------
where • r«fleets the background. A version of tha one-hit vodel that in-
corponut a threshold cm also be written (Hie/IAS 1-977*3:
Kd) ¦ l-« where dt it the threshold dose.
Unfortunately, the oce-hic modal often is called tha
IInter nodal. In tact, the lis* fitted to doit-ftiponn data with this •odel
Is altghcly concave, although It a extrapolation to low doses becwwss o«ariy
linear. Because the aodeL ha* only ooe p^rmmzm*t it I* trot £te*ible in fit-
ting the typical *igDoid~shepe of s rick data aac. The parameter caat how-
ever, be obtained if only a single positive response dose point is available
by using the of the dose-response graph (cr background) a« « second
point. In practice, If tJha lit of the curve Co the lull data aat i« uaiacia^
factory (e.g., by che chl-sqtiare autistic test), high dose daca points are
dropped successively until * fit results, ereo if coif one positive point
revtioj' The slope of eha curve below che Lowest data poiot and an assumption
of linearity can ba used with a nonintereepc log-Log scale to estimate risks
at very low doaaa or a "virtually safe dose" CVSO) where tha risk. w»» say
< 10~* (Croup at al,# lf?TJ» I.e., a &* minimis ri«k. Confidence limits of
the Una alio cm bm extrapolated. The choiee at confidence liarits is arbi-
trary but 991, 95X, or possibly III are comon, Because of iti desirable fea-
ture* , intensive efforts were made during the 1970'a to juatify this rtodel and
use it in regulations of carcinogens.
The U.S. Food m4 Drug Adsriniscracion adopted a version of
chia model in 1971 to make conservative eetlaacea of risk at low dose* fDA
chBti 60 W% liM»r exCriptUtii") $f thf upper 99* confidence limits. Follow-
ing the SEtft cenaictee's 1972 cocclusion chat radiation~induce4 cancer wt< a
linear function (me/IAS, 1972) and feto's insistence thet the dona-response
curve must be linear with positive slope a® it approaches zero dose
-------
co statistical instability of the low dose ax trapelacionr* iod the conflict
between some pre«ticte4 response i«w»is mA chose detendned in epidenioLogical
studies. ttfcdle some mutters suggested cbec Cte nod^l night., with bom data
sats, actually underestimate lotr «to$c risks, others Leutisted that the high
risk* uimUy prod cue «d urme«e«*ari 1 y «4vcne jocitul inpicc* when
translated into regulations (see for exs*pl* Cohen, 1981).** Most troubling,
ptrbipi, was Che problem of justifying um of only the iovtie data point, and
disregarding sic re reprhdueibla fflidraoge poind «h«o the model provided an
unsatisfactory fit co the data hc, ai frequently occurs (Van ly»iaf 1980).
Wwb the om-fcit model was applied to the results of the itaJy (see
figure fltMK a satisfactory fit could be obtained far only the lowest three
data peine a lor the liver tvuaors, and the fit co the bladder Super data «»
not as good as that obtained crith the Weibull nodal (Carlborg< 1991c; OTA,
1981)® Because at the substantial credibility and practical problems with the
one-hit no del. the £M'» Carcinogea Assessment Croup replaced It with the
auiei~»tege model in its water equality - criteria development (EPA, 1980a,
19B0b, 1990c) and la other risk assessments.
The flar'kit nodeI «*& be r«g*rd«4 as a special case of the
ieulti-h.it, Multistage, and Weibull aad«l• (sea below) that results when appro-
priate parameter values- are used tc each (e.g., unity foe k -in Che tsultistage
or for is in the Ueibull). Therefore, thoee oodels ara acre flexible in fit-
ting data chaA the one-hit and battar suited (or regulatory use.
Rai and Van Ryvia-(19&S) have applied the one-hit stodel co
teratological data, although this respomse it often considered to exhibit e
threshold effect.
Multi-hit aodel—this etodel (First proposed by Cornfield
in 1954) assumes that severe! events or "hies" oust occur to cause response
and Chat they follow a game distribution function. The equation is:
, d -9u
p(d) * JM i9 (ic> " —
wtere
r(k> • the gaaoa function, J* tt~^e~u du
e
u ¦ variable of gan*a function,
k ¦ nuober of bits (but not necessarily m integer)*
* The upper 9§X or 9SX confidence limit is a»re stable for a given data set
for such models, aven if the Magi's basis is disputed•
** Because of the poor fit that the oee-hit Model provides co good deca seta,
it is ooe sorpriaing that it nan andereatijnnta as well m overeatimte
risks*
¥111-37
-------
This model* frequently called che Gams Muiti-hic mcdel,
can also be considered a tolerance distribution nodel, but mi d«riv«d from
caftaidantian* very siarllar to those of ch* aulcistage nodal. It can ba
duriv^d o 4 special case of the meltistage modal (Crump, 1983a) m4 contain*
tha oaa-fcic siodel as • special case* Rai and Van ly*in have developed ki use
(Vam Ryxia, 1980). the aulti~hit model is much sora flexible In fitting data
sets than the cae-hit modal, tat requires such more data to define the param-
atara. la fitting dace, the resevbles a blend of the probit Podel
at high dose end tha iofic model at low dose. A computer program, Mulct 80¦
was developed by !Ui and ¥«* ftjein (I960) Ear model fitting. Ral and Van
Ryain (1981) diacusaed applicability of at generalized «ulci*»hit oodel for low
dot# extrapolations.
In 1978 che Pood Safety Council (1978) and Sal and Ven
Ryzin (1981) suggested use of the oulti-hic nodel (or low dose extrapolation
of cancer riaka. Crump and Kastarman <1979) contended that nulti-hit model*
that arc linear at low dose are necessarily curved downward at high doses,
chet confidence Huts based on it can he either supralinaar Ck « I) or sub-
linear (k > I), and that lover confidence limits on VSDa Iron the modal with
h • 1 ceo differ substantially Iron VSDs calculated bf other tAtiifaceorlly
fitted curves on several data seta• Becauae the original version of the nodeI
assumes complete independence of background incidence, ic yields with some
data sees confounded estimates of background, and even of response at noder-
ately low dos«.* Hasenan ec el (1981) have reviewed the practical problems in
using the aailti-hit aodel. tn general, the multi-bit nodel does noc apt>ear to
be viewed favorably es the primary basis of cancer risk extrapolation.
Multistage ->4cl°'-i'Tt*e multistage model, first proposed in
1953 and described by Ami t eg* and Ooll (1961), after whom it is often named,
is baaed on considerations similar to those of the orulti-hit model. This
aodel assumes that cancer begins La a ¦ ingle cell (or cell line)• but only
after It he* undergone a number of random biological events or stages* {Mote:
in the oilcl'Kic the events muat occur In a one noorandcan sequence.) The
¦cages are independentl the tioe spent in each stage is exponentially distrib-
ucedi the effects at so«e stages are additive vith background effects! and che
ag*-epecifU race of occurrence of each event is linearly related to dose.
The multistage was not derived on the baais of stages of initiation, prooio-
tion, as new identified, did aoc consider the possibility of repair
or tumor regression and did aoc between bcnig« and
malignant tumors.
The generalised oultietage tsodel, m developed by Crump et
*1. (Ifli, 1977), Steess and Crump (1971, 1978), and Curse et al, (1977),
as sums that background carcinoganmsia ia presane, and that exposure Co a new
carcinogen acta additively. tk« probability of response, P(d) froei continuous
Iifetine exposure at dose, d, given by the aquation:
~ A version lb which background ie additive is essentially equivalaat to the
multistage ihM.
VXIX—3t
-------
p * 1 - «•"« * M*V ~ 8,f ~ •••
Nu«ert>u« publications ia the late 1970'•s supported tha
malciitage «*>dels' biological and §c#t£ii«le«l bases, demo tut if At #4 how isaximjis
Liteiitoa4 uiiMCej of risk could be calculated «M d«gcribtd mcliods oi
using confidence lieic* to estimate virtually *efe doae* (VIDa). S#«: Crump
et el.» 19?61 J>77| Cuess and Crump, 1976, 1978| Cueaa •€ «1 », 1977; Bartley
m4 Slelkeft, 1977; HtC/iAS, 1977a* Brv
-------
avoid this problem, the number of terms permitted is arbitrarily limited* eo
the number of nonzero dose levels in che data set. This convention slightly
weakens the biological biiii of tha model; che number of scagas of cancer for
a giv«n carcinogen it nominally dictated by the experimental design. (For
soma data sets, only two or three piraoiciri art needed to *ck£eve « satisfac-
tory fit). la addition, che linear hypodMiii is satisfied only if it > 0«
Therefore, m arbitrary decision li usually mede that a positive 8, parameter
¦use be included in Cite refrcstino analysis* although satisfactory fits can t>«
obtained at cine* without it (Guess cod Crue>p» 1978). Because of these con-
wmocioftif. havener, a possibility usually regains that a beccar-fittiog solu-
tion ves ujiidentiIied~*one that cnuld have |ivea a different escitiace of low-
dose risk for regulatory considerations.
Carlberg h«s evaluated the auLcictage model on the basis
of theory «nd experimental results, and discussed several "defects" in. the
application as a atodel for carcinogenesis (Carlborg, 1981b). Sielken (l98Sc)
found that for foreuildehyde cercinoeu (Uu, 4 five-stage multistage modal gave
a much better fit to the dat* than did a three-scage version with the conven-
tional restrictions (see above)~ and also gave substantially lower estimates
of risk at low doses (being aiailar to chose of Che Veibult model). In con-
trast P the 95X upper confidence lisiit of tha five-stage model ac low dose was
substantially higher than chat limit for the Veibull because of differences Lit
the model families and their upper confidence limit procedures* loterestiagly
enough, the 95J lover confidence 1 tm.it carve for the five-stage model fell so
far beK)1*! the maxiewn likelihood estimate that it indicated the possibility of
accessible through tha National Institutes of Health, Bethesd*, MD.
Whenever the multiatage modal does aot fit the date sufficiently well, data at
che highest dose are deleted, and the modal it refitted to che resc of the
date. this is continued until an acoapcable fit to the data ia obtained
(i.e. r the clu-square staeistic). GLOBAL also readily coitputea the extra risk
ovar background, confidence limits on the risk, and the virtually safe doses
at specified low risks.
Despite its flexibility, che conventional multistage model
doea not gives a satisfactory fit to data seta in which the doee-response func-
tion rise* steeply* then plateaus (i.e., stroogly concave),** Io addition, a
few data sets are know for which the fitted multistage ourve ia concave at
moderate to low doses.*** la eatrapolacion to very low dose, the tsultlstage
* The li are alao coos trained to nonnegative value a ao that only monotonia
relationships result.
** Chemicals giving audi date sets incliide DOT^ DBS, ethylene dibrexside, and
vinyl chloride (Carlborg, 1981b).
***t*awplea include 2-MF, VTA, md heKeeJUoroberutene (greeeki and Van Kyain,
1981K
ftll-M
-------
¦wdet and Ui upper confidence limit becone essentially linear. Ulch s»«c
adequate mono conic dec* sets, it give* estimated risks higher than chose of the
probi t, logic, aiulci-hic, ud UeLbuLl nodela, and of can nearly as high as
those of tha oat-Mc «odel.
Cibb mud Chen (lf®6) recently fropoa«4 a variation of eh«
s&iltiscage nodal that could address auUi^Hcitive carcinogenic effect* as
well " i - «d0»
Cearputatlon Cor the liaacriaed eultiscage aodel art mda
using the CLOftAL progress of Cramp Weceon (1979) as updated. The oulti-
atage nodal ia fine fitted to tha ddca* u|ls| a nunber of earns I the poly*»
noauel equal to the etabw of doaed groups in che etudy beside the control
group. The fit of che nodal to the data can be teetetd if desired by the cht-
square statistic.
• The symbol used for this extra risk varies among CPA documents end also
e*ong literature publications, rich A
-------
„.f u»:v*1!
* 2. - p.)
>i * v 1
where I| i# tha nuaber of enistai* In the 1^ dot* group, *| Is ct*« ounber of
aniMls in the i dgse iroup with. • turaor response« f ¦ ii tfce probability of a
rcspoaae in the ich dose group eatiaacedl by ficting the nulcistAg* model Co
th« data* end h is ch* QUMbtr of rameiniiiv groups. The fit ii decaminad co
b« unacceptable whenever it larger than the cumulative 991 point cf the
chi-iquare distribution vith f dtgrm o£ fre*do*. where f equals the nueiber
of lose groups etiau* tha Gaaber of nonzero Multistage eoeffici«aet» II eh*
f|i is tmeecepcable, data At eh* highest dose ire deleted «ed the model it
refitred to the rise of the dae*. This Is continued until *& acceptable fit
to the data £• obesined* A fit will ilv«7i be obtained far the legate dosed '
group, «vt« if eb* response appears anomalously high. coap*r*d to higher dosed
group®.
The upper 9SZ confidence Untie of tha best utiract,
value is c*leul*t*d by reeuutiniaiaf the l^g-likeLihood function (Li) for q |.
The value of qt is incraurecd t* • value q| eueh chat the rsexistum value of che
lot iiln*tiho«4 function Lt satisfi«t th* aquation;
2
-------
"potency factor" of the carcinogen. Tha converted in written a* in
iMi EPA publication* hat not In other*, and qf L* «oeeti«ia* used for tfc«
converted values alto. CM has published values for lv and VSDx for over 50
carciftttftai (Aad#rton at *1., 198 J; CPA, 19A&d). In addition, CAC hi* fevel-
oped j "potency inde*" for thaaa carcinogens by multiplying the nalecuLar
weight of each by Lea and also an "order of magnitude index" by caking the
logarithm (base 10) of the potency index and rounding to cbe nearest Mho la
ouober (£fA, 1,9844).
Mace chat the confidence Unite in the linearized «iulti~
• Cage are statistically correct only If the nodal need to ttoaipuce tha limit*
1$ ib accurate representation of the underlying doae-respoaee function; they
do doc provide tny neasure of cm ex teat to which ch« node I it correct or
Incorrect« In fact, the use of tbt upper confidence Halt la tha oultistage
model e-an refillc id a nonzero estitaete of risk for data sets chat do not ahov
carcinogenicity {Wbicceaore, 1900)®* Tha Chairman of EPA* a Carcinogen Aimd'
meat Croup recently noted th«cv overall, a feeling exiacs that Chi biological
foundation ia fiiasy for £PA' * curttct acthod of lou Laval risk, estimation
based on the linearized inulti stage tnodatl (Albert, 1986).
Vaibull model: Ute Veibull aodel ha a b on Che basis of a aedunlttic lx isodeis.
** A geoaral fom of the tfeibull diaCriimtioR iai
-(« ~ §4®Mt -
- 1 - a
where t is ch* tim after doaing atxrts^ w is Ctuaor gro«ch Ciiiflr 4m4 k ix
a axiaber of ilacrece ehaagea leading to tuaori.
W21-4J
-------
where P{d) It ett« probaJBil icy «£ rerponte at 4gm 4 and e. I, » are piriMiiri
CO be «*ttttated (S,n > ®}» Alpha (a) is dictmiMd by Ch« beckgro\u»d c«*or
incidence; I, * $e«le p«rittac«r, depend* on the units cf dose; ¦» a shape
piriMKr. £« Mutllf in the range I co S (not necessarily aft integer),
although a few data sets are known wick at of 0.5 or less. Alternatively,
PCdJ » I - «
if thtt background Incidence f» negligible (¦ = 0) e* If on* %»i»ho co express
the extra probability over tacfcgrow4» The relationship at til* low end of eh*
dose range is linear If * ¦ l» conm (subllnear? if m » 1, ®o4 tdeea*#»t
if a < 1, n in a few case* ttbmtm the curve rises rapidly and
ck«e plateaus.* At very low doses, Che titfi risk over background is
l(4> 3 The virtually safe dote «1 a| • UN, md is aa*etin*» described as a generalisation of It
-------
nine data point* va.s cnade available by the Fmd Safety Council In 1990. non-
linear weighted luic squares Mgwtiion computer progress were alto in use at
that tine co ociutt tit* paraseters directly, and a nonlinear mxiiua likeli*
hood ratthod prograo was available €ro« Krewski at the Directorate of Health
and Welfare. Canada {?oo4 Safety Council, i960). SimiLar programs have been
developed by Crump and Howe and p«rtmp>« by others (Crump m4 Howe, 1985).
Because Che exponent of the dose (w) is allowed to tjk#
fractional values (rather Chan being constrained to integers at ua the Multi-
state), an excellent lit can usually be obtained if the observed data exhibit
a conveneioo*l spread and stupe. The standard form of the Weibull generally
e«m give a good fit Co data that exhibit threshold-Iike appearance (e.g. , cha
CO#i bladder data), or It can be modified CO eecoMsotUce * threshold dose.
Overall, in oorve fitting the Ueibull aotftbines the better (secure* of stech-
aaittic nodels such as the aultistage and ftultihlc with those of the tolerance
distribution model* such Am the probit and Logic.
Carlborf C1981a) applied the Ueibull nodeI to II cancer
data tecs for IS chcaiicals (plus a toady In which protein and calorie intake*
were varied).* He reported generally satisfactory fits Cor the Wei bull to
these data (2? sets), and noted that -none yielded « best fit corresponding to
the one~hit special case (i.e., m - I.0)* The Food Safety Council (1580) and
Krewski and Van lysin (1981) compared the fit and low etoi« extrapolation prop-
ertiea of several models (including the Wieibull) to data seta for 14 sub-
stances, the response being cancer in nine ceset and ocher than caacar In live
cases. The nuitihic, multistage* aiM Ueibull models all gave fits consis-
tently superior t» chic q! the oae*fcic aod«l, which fraqtuncly giva quite
inadequate fits. The Multistage model, as usually constrained, did not give
quite as good a fit in general ai the taultihit or Weihul 1, the Veibull being
Che choice of displaying the data graphically. Carlborg (1981c) tested three
versions oi the Wei bull nodel (varying dose, time to tusior, and duration of
exposure) agaLaat data fton the CD0| study and found s4od fits in all cases
for both liter and bladder cumri. Christensen (lfiA| found the Weibull to
have considerable pronbLse in analysis of aquaci-c toxicology, end Christensen
aod Chen I If 85) have found it promising in predicting the combined toxicity of
two or aori chnicals.
The low-dose extrapolation characteristics of the Ueibull
are generally similar to those of the onltihit model* The estrapolation is
nearly 1 ioear, and the risks calculated at wmty low dose are usually lower
than those of the multistage nodal and near the Middle of thoae for al I the
common Models (Krewski and Van Kyi in, 1981) • for a few date sees, however,
the Mai bull gave higher estimates of risk, then the suit la cage sodal, although
aot as high as the ooe-hit vodel (EPA, 1989a)» Extrapolated risks with Che
Veibull appear to be more sensitive to low dose data points than those ex-
trapolated by the multistage (EPA, 1983a; Brown, 198S).
Van lysis end ftai (1987) introduced a vmri«t!oa of the
Veibull model which Incorporates the concept of the effective dose. The
* Carlborg (198La) used the computer program flMDFli, for curw-ficting*
VII1-45
-------
administered dose It cramfarMd fey MicJsaelis-fleacor noal Laear kinetics
give An equation vith lew pAranacert,
. i. * *• ('~V') ]
ea
Mill)
where 4 i« the administered dot*, and «l# m>a ij arc ail » 0. f( and t3
arc the i«nir respectivelyi *• the a And m ia the Vetbull* end t} La a produce
of che Ueibitll 1 and a constant raised to th* » power. Substituting these
fives Che form
" « ~
P(d) ¦ i - a I
Vstfl
vfaere a| end t] are constant® related to Che race of change and the
Michael i«-Mencor coaittnt, with *1 > -»l/Kt where M is the mtziaauB dose
adeiaiacer^d vo a study* A I. the authors applaed the ©o*-feic version of chair non-
linear Hiaectci model co dace sets for vinyl chloride. DDT, and saccharic*
which showed concave , linear, and cod vex dose-response curvti, respectively*
The Clt was tetter than for the conventional ona-hi t, and the nodal was judged
to he "reasonably adequate" for these three caxc Inogena. The node I gave a
virtually saie dose for saccharin intermediate between those calculated by the
cotmmtiQoal «ae-hic «ad Ucibull models. Thus authori did not apply tin
k±netici-*djusted Weibull Co ch«se data sets.
As noted in. the opening paragraph, reservations about tha
biological basis of the Weibull model for cancer (i.e., the eultitellular
hypothesis) has 1united its application in low dose extrapolation of risks.
Because of its other desirable features, strong Interest has developed re-
cently ia reevaluating its foundations. Carlborg < 1991a, L981b) has noted in
fitting the Weibull to £7 data sets that the ihapa ptrimcer, «» tends to have
values that cm be nt^rnM as a fraction, I IZ, where I is an odd positive
Integer, i.e., m typically has such veluesi as i/2» 3/2r 7/2, etc. The physio-
logical implications of this observations #re uncertain, but toxica*inetic
factors nay be involved. In addition, Carlborg (198la, 19811) note* that the
extended (t£«e-to-tumor) tyeibull model can be used dlrently to derive a (967
empirical observation by Drucfcrey that tha dose multiplied by the median tinr
co-tumor raised Co a power £i a constant,1** i.». t
due*" constant
~ §i go versa tbe shape of tha curve in Che low dose range in the sasia way as
u io the Ueibull.
** lecent resales indicate that Druckrmy' s rale holds for gemotoxic
carcinogen*, such aa NitroiAaioai, but not for cbe noageaocoxic
carcinogen*, OieldriA, a chlorinated pesticide (Pereire, 1989),
VIII-44
-------
Cerlborg also noted that tha extended multistage model eould not be used in
the general case to derive Druckrey* a mi*.
Cfcristensen and Chen (1985) recently found that certain
mechanistic •vstusptioqs about the reaction of toxicant molecules with key
re**ptor9 of etc organiM and probabilistic aaeumpcian* regarding the coneen-
tree io* of blocked receptors at any time Lad directly to the Wei bull modalr
The shape parameter, ¦» can be interpreted at the aviriga nuraber of toxicant
molecules p«r receptor® These authors wipstad chair conceptual nodal
particularly for cues where response occurred miiaiy in a single or|ao sice
in that organism (i.e., sices with vulnerable key receptors), and concluded
efuie tha parameters of the Weibull model are noc simple results of fire in* a
curve to data, but chemical-toxicalogical significance.
Alternative biological rationales occur co chat praMoc
cutter that may provide an Lsiproved theoretical basis for applying thij modal
to lo« dose cancer extrapolation. For exaatple, in plscc of the multicellular
hypothesis, Che Veibull nay be applicable Co multiple events within a cell,
suck as attacks by Che carcinogen moLecules on multiple chromosomes, on Mitt*
pie genes, on multiple nucleotides along a DNA strand, or on multiple parts o£
a cell (including membranes). The Veibull distribution may be applicable to
situations where multiple agents interact in the a arc (or nearby) ceils to
cause cancer—the agents possibly including carcinogens naturally occurring in
foodsr trace environmental contend nantSj natural radiation products, *nd
viruses, to addition Id the test carcinegesu Many carcinogens are known to
form adducts with OMAr the stedftfth bf ch* sdduct reflecting the patency.
9iJMtt4g pceut* et four fitt# prLm^rily (gttgnijae, tyrosine, adenine, and
tbyakine), buc different carcinogens preiar different sitae. Sach considere-
ciona may lie comfortably into a, Veibull model, together with other inter*
actions such as multiple encyue effects and cell killing.
Time-to-effect made!a: Because m€ the latency periods
observed for cancer, che 1i»e~to-a ppea ranc e of tutors may he an important
consideration in developing regulations for carcinogens. The cnri^fiaal ver-
sions of several of the cooaon dose-extrapolation model a incorporated cime-te-
effeet a* well as dose. These include variations of the probit (iog-noeael
and log-time)« log-logistic, ganaaa muLtihit, multistage, and Veibull, aad alio
others such as the general product, Sartley~Sielkea and Oaf fer-Cruatp-Meac«r*wan
modela (Daffar ec al., IfSOl Creweki at el, 1983) Crump end Sow, 1984; Crafty,
1915b; licbia and Crewski, 19B3; and Sielken, 1985). One example of an e
-------
This atodel «ms found to fit very well clue deca (rr did not fit the £D01 drncn as well 11 did tiu»
Hartley-Si*I ken aodel {not did the one-hif dr extreae velue models). Salzburg
(19H1J reported food fit o£ the chree-poraencer Ueibull to che ED0i bladder
data and £®c a modified Weibull Co the liver decs.
The Mo«lf*vk#r*V»j»»«a-*ftu4*om (KVIC) tvo-^stafc mathematical
nodal (Hoaltevktr and Yawn, 19791 HoaLssviur and Knudson, IM1) i« wll
ragjtdtd for cancer risk. assessment. Ic hat a biological b«sij ia cellular
dyoaadca and transformation* and Incorporation ol tiare-co-cunor« The equation
has die £ocn
i{t).
where 1(c) is the canctr incidence at cwum t» XCa) i* the oueber of suacepti-
bin cells at my ciaa (aiauned to bt decerministic m4 known), uv and ** are
rates of call chants* lo reaching the first event and foil eialigaancy, respec*
civ«ty, md * and i ara call formation m4 death. r«tei, respectively.
Thorsluod «t al. (ifUl) and Chen at al. (1956} have pro-
poaad doaer* and t u»e~dep«ndeac and ege-epeeif le cancer risk function* based on
the KVK model. Moolgavfcer e&d Dewan]i CUiil caution that the derivation of
Thorsluad et el* contains aa Approximation that my be adequate for hunaa
e*ncer Incidence rates, hut ia uaLikeiy to be valid foe animal experiments
vich. vary high, cuaor rates.
Siallten (199?) Haa described an Individualized Response
Model 4 ftctealftfy Policy (OSTP, 1983).
VX1X-48
-------
A major probiaa Ls tliac the uncertainty increases as one
moves ivtf !roa the experimental doaa-g* range to lower mart Lower doees in cb«
extrapolations. The problem is iuggescai Mheiftiticaliy in PLgur« Vni-4. The
divergence arises frod uncertainty stti whether ch# model is applicable at
very low doses« And alio Iron statistical aociirtaiotj ia applying 4 modal to a
particular data set. Thai tnost likmly sseiiMtt at vmrj lav doses becomi
increasingly unstable with a imLI change in the response at tt|MnaMui
doses* the use of the upper confidence Units for the multistage or Llneer-
ijeed multistage reduces tie instability problem in •tcioatlag maximum rieki
ao.** The flexible oultistage
* Interestingly enough, firowo (1915) found that for the good data sets for
bladder aftd liver cju&eer in tt&e EDaj acudy, dropping the lowest 441*
points had negligible effect ofi the extrapolated VSDs (I0^c risk) by
either the mulcistjige or probit nodeLs. The effect on the best estimates
was not reported.
** thia exas^le is Atypifl4il of VGI teat results. In fact, 10QI cancer
reapoeise at tolerated doses is unusual.
VCtX«4!
-------
X
X =0
1
2
3
4
S 5
6
7
5
9
10
Experimental longr—^
**'* S ''
^ * M
^ / /
^ /
t-UPPIT ' /
95% Confidence Untff /
/' /
/ A .
s ^*X— ifimm
if / 95% Confidence Urolf
~^v-» Low Do## 4
blfwpebfiwi
J L L JL L I L_ I _L
Dote (log scale)
Source: La»lea« (1986).
Figure VXXX«4 - XLlufttretive Increase in iac«rt»i«Cf for LovHDoee
Extrapolation
vziz-so
-------
SODIUM SACCHARIN
2-Mr
ft 1 .
CMS
1 J 1
30 100 tpfii
Q m w -j
m
Qfi*
QP3
«mu
m
mm
W«ifeuM «Jo*c-te*pOA*e tachkJ Ikied to ihc obccwnl Jala
10*
w'iiy
lO'lppal
DOSE
E • IJmm (jtlMHUllift t- Lifl Had*
6- IMU-Mil
• • WnM Mil t - PmMI MoM
Sourer; Adapted fro® Kuaro and Xr«v«kl, 1911b.
Pljgiuri VIII-5 - Pafle-)te«pofis« D*t* m6 Law Doie Extrapolations
with Various Model* for Four Carcinogen*
-------
model can be easily lit to cite data, but rive systematic uncertainty at low
do it extrapolation it inherently vdry tu^h.
In suswery, wetbeaeCical models have been used extensive!.?
for estimating cancer ritkj at low doses! they have been used (either in che
form of nsxiaudl likelihood • ¦tusates, upper 951 confidence lisits, or VSO
estisMtes) for reiuUtary (wrjioiti. Hatheaac ical modeli k*ve sufficimtly
serious problems, however, thee sever el alternatives to the use of a single
mdil have beta suggested. So* authorities iuues(ed extrapolation along *
fitted nodei to, *ay» 4 11 riik# and thin linear extrapolation at lower doses.
Others suggested ID-' to lO-* risk as the twitch, point (Van Byzia, 1980).
Ocher authorities have endorsed reporting sone combination of best judgment
peine estie^tes and emxitmus plausible risk estimate or b««c actioates with
both upper and lower confidence liaits (Park and Snee, 19*4). and still others
have suggested taking into account the best estieetes calculated by three or
lour model* (Food Safety Council, 1930).
If aathaaaticai extrapolation is used, the modeUs)
selected should iseet at least three criteria: (1) it should be capable of
fitting observed dose-response dace for a wide range of cheaicels if it la
expected to have much credibility Lb extrapolations below the observed dc«»«
range; (2) it should be id egretasumc vich (or at least not in dimtcettwt
with) our uoderscending of the nechanisns of carcinogenesis! and (3) it should
be useful with the kind, of data sets likely to be available for chemical*
typically found La hazardous wastes. These criteria appear to rule Out use of
the Heacel-ftrysa and oa*-kie moduli® The nodeIs of choice seem cleerly to he
cfae Multistage «od che ititelLt loch fmvc good fluibiLity in being tit co
diverse data nets, and usually beoosui essentially linear in low dose extrapo-
latidu The multistage has been wall regarded because of its rationale*
utility, and "cocaervativenessIts linearized version gives linear upper
confidence Units on risk In extrapolation. On the other hand* opinions have
been expressed that thai srutci stage node Is estimate risks chat are too high at
low dose to serve as the priaary basis of regulation {particularly when sub-
stantiating observations on bwars sre lacking). The fact that the. linearized
noltistage model gives nearly the i
-------
substantial liaitaciona* Th.it data ef models ea bodies mere basic and costpre-
b«niiv« cDBCftfci than cha cixed mathenutlcal aodelt discussed in the pfactdxag
subsections.
Cornfield (If??) noted chat the probit mn4 Logic mods I a gen-
erally gave similar results to toxicokinetic models ia the 5 to §5X response
range. and developed a method of estimating parameters, particularly the doaes
saturating metabolic mechanisms. Ramsey and Gehring (1961) described metltods
for applying pharmacokinetic principles to improve risk assessments end noted
p^rtiwtJUixlf U>e need co estimate Che retained dose in aoimal subject* and
exposed humans. Withey (IMS) recently reviewed the pharmacokinetic differ-
ences between species, mi Boel C19853 reviewed the incorporation of pharmaco-
kinetics into low-dose extrapolations. Pipers ia Voodhead et ml* (1?85) dts-
cuu many biological and extrapolations* chandler (IfSS) baa
compared biocheerical aecheniscic models with other models proposed for quanti-
tative risk assessment, ami urged greater use of the former,
Toxicokinetic models require a good deal of information. abouc
cbe absorption, distribution, storage» metabolism. and excretion of a chemical
in the organism, including the concentration of the chemtcal or its tone
metabolites as a function of tiac in the various body compartments relevant to
the effect produced (e.g.r concentration in the blood, liver, bile, adipose
tissues, urine, and the target organ). Thus information i» analyzed mathe-
matically in a scries of kinetic equations {first order kinetics usually can
be assumed), and then an Appropriate mathematical expression ia developed for
the dose-response function in the test species.
Accurate extrapolation* ecx6** dxpdAurd routed uiudlly CJfl be
made with these models, and they c«a be combined with other models for low
dose extrapolation** Accurate extrapolations across species can be made with
toxicokinetic models if sufficient data are available for any critical differ-
ences (e.g.. metabolism, pathways and rates) for the second species. Unfor-
tunately , the use of toxicokinetic models is severely ILmited by the lack of
date. In particular, most data on hunaas have been colleated on chemicals of
pharmaceutical interest rather than on environmental contaminants*
d. Other predictive —tho4ai Severe! other approaches are
aveilable to help predict the kind and degree of adverse health effects of a
chemical. In general. these methods woald be most useful when the aveilable
ccrxicity data baae is inadequate to parade use of a coxieokinetic model or one
of the better lew dose extrapolation methods described above. These
approaches very substantially in qualitative xxd quantitative characteri s t ics,
and in the uncertainty of cite results they yield- They could be useful in
some cases' (possibly in combination trick each other and with low-dose extrapo-
lation methods) ia assessing risks of hazardous waste disposal, end are
therefore saanerixed below* Approaches included are* Potency indices;
extrapolation from noocbrooic datal comparison with cognate ehefeiaaisf uae of
thart-tern edcrobiological and biochemical tests ( use of ADXi and TLVat and
nonperaMtri* methods.
mi- 53
-------
(1) Potoney Ifxjkus S«vorol ouchoriciaa h*vw iu|tiitid
chac an Ic4*» of can«ar potency. batod on do)r>riipoo«« r«uUi to cba «iwr*
taaoneol ri>Mi ihould bo *iod «i a bull of rcgulacioa i& plaoa of wry of cba
proposed «•€>*¦¦» tica I Mi«U for txcrapoUciog riilui to vory Lorn dom.
Sowtrol Approaches CO developing owed irtdieea luvo Imsm proposed, toan of
«fcich incorporate iMChaaetical See for eu«i>lat Meaelaoo and
lasitl |lf?ll| Crevcfe and 1JiUo« 1912)| lltC (iWd)t Sqwltm C Willi OTA
(IfH)l et #1* ClflJ); «ad Perk and Sato (i960.
Aa by ?ato it al. (If§4 J is of parcl
-------
While pOCMCj tndicas lo (aneraL a an be vaLuabLe tft eow-
pariog the relative carcinogenicities of a large number of chemicals, they
have ttac bMB ihowo to date co be helpful in predicting response At **ry low
en v£rorae&ceL is{N»urii for a gives ahenical' They do aac appear to be suf-
ficiently developed at present Co be useful a§ dose-response function! in
predicting the nuaber of cit<« of adverse tfftcci la a given population at a
predicted wriftHMfltil exposure. They could be useful, howivtr, la asm coir-
perativ* risk assessments wherein the alternative wast a disposal technologies
produced different carcinogen*. Potency indices hava not been used to date
for regulatory purposes, and it appears unlikely they will b« wtii a defini~
cion has been accepted lor s ' safe" Index for soi»e *cendard reference
carcinogen,
(2) ExtrepoLaeLpus frae nonaJurentc or other data: Tha
response of ®n orjwtisii to a cheoical ac one Level of exposure soflMtines cm
ba inferred frxxa epideaiologicel or animal data at another level «l exposure.
Tha uncertainty will vary with tha nacura of the cluriicaL and the response.
la tame cases the nature of the responses «ay be similar,
and, information on rsetaboLisis, accusalation, etc., can be considered in ex-
trapolating effects between different intensities and durations of exposure»
Extrapolation of subchronic exposure data to chronic response will moat often
be of use in hazardous waste disposed assessment. Dourson end Scare (1903)
proposed that chrooic N0EL8* HGAELS. or MAILS could be escuaated froei their
subchronic counterparts by dividing by a factor. Belated extrapolation
aethods have been used to estiomce upper ri»k Units of some end points,
ftigri* (1961) auaured Che Im4iict decreet* ia reeplrgcory rata (« measure
of irritacion) as a function of dose ef Labeled |ases and vapors, and calcu-
lated the W>se (dose which halved the respiratory rate), Although che t%Bt
varied by aver five orders of (tsgQitude4 11 of the was a good estimator
of the TLV for permissible occupational exposure for humans. Similarly,
ftenaga (.1962) calculated the ratio, of acuta to chronic eoxiaity af toxicants
for various aquaeic species. Acuta LC#f (median lethal conceit era cion) divided
by che mmximm eaccpceble concentration (a chronic NOEL) gave a range of over
five orders of oagxticude, but nest ratios ware lass than 100 (i.e., exposures
of about IX of the tXlft posed a relatively low risk).
la other cases thai naeure of the response my be different
under e*fOiar« conditions substantially different £ro® thnwa used in available
studies. Prediction of cercinogenic response is percioularly difficult, but
night be possible. Zeiae ec el. Hlf§4) found an efspirical relationship
between che acute tozicicies (10*#*) of e«ay eheaicals and their careino-
geoicitiea following chronic exposures. Although a biological rationale »ai
not suggested for such, e relationship, they proposed that it could be used to
¦uke preliminary eseinwees of the carcinogenic potency of an unstudied cheat-
cal f and to give an idea of tilt atteertaiflty of che estimate. Such. • rel*-
tiooship seetss quite reaioeeble if the chenioal killt ceils in rough propor-
tion to doac, and if che priflMiry target aells are cepeble of rapid renewel
(i.e., rapid cell proUferaciofl ia Maeeieted with ca&cer). In che future,
VIII-S3
-------
eheoretical advances and experioental data bases mj pa-rait other correlations
of biological «£(tctt.*
(3) Comrarison with cognate ch—icalt? The result* of
tkeaoie ciici on cognate l»iiiuler) ^hemic*Ia can losietijaes* used to cBtiuct
ch< slope of the'doie*Te9ponse curve, artd than comparable lubchrMic expoiunt
data from the ¦ abject and cognrnt* n«^du»4t tap be used to locate a ^oint on
the curve. Standard mechoda «re not jret available, but two approaches hja««
been described; the quantitative s troc ture-act i v i ty relationship (QSAR)
•plfoaeh, and the prototype relative pocancy approach. the literature is
•till inadequately developed And integrated, bat a recent review provides a
good bibliography (Kt£f 1912).
QSAft approachi If sufficient cognate* can be founds Q5AS
can be used. The baaic concept u Co calculate regression equationa for equi-
effective doaes of various clwnuls at a function of parameters of cbenical
structure. One then estimates the corresponding dose for ehe cpapouod of
interest. For calculating risks, it la aaauoed that the dose- response curve*
of the subject cheeiical and a a«U-icudicd cognate are parallel} the doaes of
the cognate arc adjusted by the ratio of the equi- effective doses (i.e., by
relative potency), and effects are calculated. There are two types of QSAft,
differing in the hind of structural parameters used for the regression.
The Hjutach itathod, an older and »re conenonly use*§
approach, is based on physical organic c heads try (Lyman et al., 1982). The
paraawters My be the usual frta energy-related terns (Hanmctt' s o constant,
Taft's scerie eonet«nct etc.), newly derived free eoargyteiated terms (such
a* If ana ch'a a for lipophillcity), or other paraaieters that hav* been suggested
(e.g.T certain chroautographic If'* which are proportional to *, some quantua
eachanical paraaratars, or infrared stretching frequencies). Various conbina-
tions of the paranyeters to the first or seoond power are correlated until one
finds the simplest regression equation wtch a good correlation coefficient.
This method requires a close chemical relationship e»ong the toxicants being
considered: for instance, Banach's original study considered Che effects of
various substitutions oa the rl#j< of a series of phenoryacatic acids on their
efficacies as plant growth stlmianCs.
The fcce-WilsOO oachodr a rec«ac development, nay be use-
ful ia cases ehe re the structural dlmiitf i) too great for the Hansch a»th-
od« la it the regression terse are e series of arbitrary peresseters, one par
structural feature, having the value one if the feature it present, and aero
U it ia absent. Thin out hod require* (as well aa allows) store cognate com-
pounds for equally good suitability of fit. This Method has been generalised
recently (NIC, 1982).
Prototype relatjy... pttacy aptretcht If a sufficient
nuaiber of cognate coaqtounde with suriiaibiaLaglcai
-------
or4tir to prepare a reasonable QSAft, the nice motive Is to use a Mthod reqsir-
Log f®wer copjattj or prototypesr orach «; the relative potency approach.
Using cMs method. one deteraines % chesnesl class or a scries of cUsiti uch
coacelaing che inadequately studied subject chemitai and one or more rela-
tively vell-studied prototype che®ie»L« tar which chronic test data are avei1-
4bl». If cooperable icttcc tut datj are available * one can uj« cho relative
potencies of the subject and prototyped s) to estiaate the chrtmic tenticiry of
the subject oheedcal~
Tha Inherent pro-biene of this approach ara that tha
assumptions ara a van ffMtit than vith QSAE (hence, the uncertainties are
greater), and that often the mott $cm41«4 corrpounds arc tha eose potent* the
risk estimates are likely, but not certfla, to be too hi|h. The application
o£ this nethod to risk aiiaautanc it very recent avd ill defined. One of the
fee studies of this typa has been af ay&futle emission products by Health
Scientists at Oak Bidga National Laboratory (Dudney ec al., 19821 Cell*
et al», undated).
(4) Short-tern ¦i.erobioaaaay testa: A vide variety of
oirrobioasiay and biochevixal testa haa been developed and are being used to
acudy certain biological affects* Their comon denominators ara the viae of
cultures of BAjamuiies cells* unicelluLar organises (bacteria, yeast» etc.)» or
viruses, usually with in vitro techniques. They often use a outagenic end
point* i.e.* induction of «n ieheri cable chan|e in the gens plasst of the test
apaciMsa. Kollatain at al. (1979) have reviewed the subject* Sooe of the
¦ore coanoo teats ara ttetftd below.
Thu Amis teat use a in vitro cultures of several nutated
strains of 8*iaon«il,.i typhiiairitaa, all oT~wSTch"require histidine for grovth
(Anas at a I.. 1962-, 19731 Haroo and Asms, 1983). Aliquot* of a bacterial sus-
pension are incubated, in a histidLn»-deflatent eiediuur, with the test cheni-
cal; teow platet also contain * portion of hoBAgemzed rat liver which will
Metabolite sooe cbanicale In a mmmt resemblIng chat ia che whole amsai, ao
that the amtagenicity of ratabolltes of the test substance can be studied*
The muober of observable colonies will be equel to tha number of bacteria
which revert (nutate beck) to the original far»» which does not require
extmal histidine*
Kefiy variants of Che tei test are available, particularly
the ose of different test species. A rocneon veriaec it che "host-radiated
assay" in which a suitable bacterial euspeasioa i» incubated in tha peritoneal
cavity of a csouae ar ret, vith the co*lcent given to the whole aninal. Thus,
the Haeterie are exposed to the toxicant whatever nateboWtea are circu-
lating ia ctm blood of chft hasc ro-deat# K«*ag.li,iis calls also any be grown in
culture* exposed to a tovioant, and eeaminad for atutagaaio affects^ although
deteratinatioo of an end point auiy be difficult.
k related procedure is cytrogeaatios testing* which is
uaually done on anfiaal 1 already being tested in an ordinary toxicity study.
Tiasue saosplea ara taken (blood for white cells during the test; bone narrow
and kidney at termination) and grown in cell culture. Uhea the eel Cures ere
rapidly dividing* tha cells are killed and the chronaososws in the nuclei of
VI11-57
-------
actively divi4ing cells are aiaiMd autroscopic«ily for abnormalities inch a*
breaks, f«pa» trisomies (tripled racher than faired), Balfonaatians, etc.
Sttbitmti*! aoc«f(»ifiC7 is inherent in »itn)biQ«s>ty rcmU) ctt i$ecu huiin risks co ttfoiuni to-n chniul. The
doaere^pans* reUci^ajhlp a«y be tfuch diffarioc—tew lin«<( cilitiontbips
for autjgeniaity In vic.ro «r» not in vivo because of tht bady'» def enaes*~*or a
tteoaal auir not reTpotwTTrt tbe sans* way. In addition, Mutagens arc not neces-
sarily caroinogena.
(5) Ua« of APIs. TLVst and ljpci»»: Acceptable exposure
level* have been tic for nany chemicals on the basis'of toxicologic*! reeuJLts
and human experience. The U.S. Food and Druf Administration has ton§ esteb-
Ushed Acceptable Daily Intakes (aOI»> of c«;c«io contaminants in foods,
drug*, and cosmetics. Hie American Conference of tovernmenceL Industrial
Hygieftiscs and the Occupational Sefecy and Health Adminisc ration have adapted
Threshold Limit Values (TLVs> for a targe number of chemicals for controlling
inhalation exposure in the workplace® The EPA he a published Recocnerded Maxi-
mum Concentretion Limits (KHCLs) for aeny cheed&al contastlnants In water.
The claaaiasl nachod of estimating an AM involves identi-
fying from dose-response data a "Ko Observed (Adverse) Effect Level" (UOEL or
HCAEL), I.e., a dose level at vhich no (adveree) affects have been observed in
appropriate studies. That dose La then divided by s sefecy factor audi as 10,
iOO, or 1*000 (depending on the nature and quality of the data available) to
produce an ADI.* Risks o( adverse effects are assumed co be negligible for
daily exposures at or below Che ADI, even far susceptible persons. One should
ute, however, that NOCLs ami HOAJELs can vary vith the speciea tested, number
of teat animals, and cast conditions, including doae lev*la, duration, and
effects looked for. Generally( AOIs have been ect4blished for chemicals
believed to be systemic toxicants, but ooe for carcinogens*
TLVs are based on a similar rationale except that they are
estimated to bm negligible-effeet 4o$es for healthy workers exposed *0 hr/ve«k
to the substance during the 168 br TI»Vs have been established both for
systemic toxicants and carcinogens by the ACGIH. As a €ir*c approximation, a
published TLV could be ixilciplied by 40/168 (0,24) co calculate an epproxiauttn
(time-weighted average) mfe continuous exposure* The derived number if pos-
sibly leas accurate in esteklishing en upper risk limit than is, the ADI
because of che uncertainty of converting fro* intermittent to continuous expa-
aure. In addition, workers are generally healthier, and therefore probably
less auacepcLble to many adverse effects then is the general population. *Ms
are sometimes calculated front a TLV (tfith appropriate Assumptions and conver-
sion of units), but would then h«ve similar uncertainties. ftMCLs are beaed on
a rationale similer to those for ADI eod TLVs (EPA, 1993b). Oourson and Star*
* The sefecy factor is- unfortunately scmetinii also referred to as an un-
nercainty factor. Bote that an established ADI might be subsequently
increased m more reliable data become Available that permit a decrease
la che sefety factor used. Conversely, an ADI isighe be decreased if data
on a oew response becoee available.
VIII-5I
-------
(1913) And Dowrton (1183) have reviewed the AM approach, and Moreeu end
Anderson (1M0) and MIC (1980) have discussed its u«# in rick iteaegeineaC» T&e
need for better uethods of establishing reliable and defensible AOls t»# been
di»cus»e4 recently by Cramp (1984b) and Dourson ec al. (11851, and alternative
stethodB k«ve b€«« suggested to ioprovc chw» process.
AOls end TLVs ere of quite limited use in quantitative
risk 4 che ADI
cannot be easily esciauiced. The risk oty actually still be negligible ac
exposures Much greacer then the AOL. The major limcacion in che use of
either of these procedures is cMc exposures «ay bo frueir than che estimated
ADI or TLV so thac actual response* (or risks) cannot be directly predicted.
Dourson (2983) tot proposed a Modified approach CO che
identification of an ADI chat may be useful ia e*cr*polscir\g the dose-response
daca far chenical* believed to have * threshold* Pirsc, using dose-response
data (converted if necessary to human equivalents) and an appropriate regres-
sion procedure» calculate a dose-response Enaction. Most, calculate che dose
estimated Co produce a 101 response (d 1#J and ics lower confidence limit. The
choice of dose extrapolation model is relatively afLuoporcant, since the coooatoc
nodels produce essentially the saM curve Id che 1Q to 9QZ response region.
Then, calculate a modified ADI as the lower confidence limit (of d(0) divided
by Che appropriate safety factor. Hexc, to MiiMie effects from a given
exposure, use one of three different functions, depending on what Che dose
is# If che dose is Che modified ADI or less* the estimated response is
aero. If che dose is d (ft or greater, the dose is that estimated by the dose-
response function* Between these two poincs (modified ADI of 0 end d|0 of
0.10)i che estimate is the peine on che scraight line a£ che Log dots-respaoee
ploc connecting the two poincs* 0e« limitation on che use of che method in
the present application is that if actual exposures are marginally above che
¦odified ADI, che uncertainty in the risk, is very high* If the actual expo-
sures are beloe the threshold, all effects are sero, and comparisons between
waste management technologies are redticed to a oenhenlch basis such as cost.
(i> Bonparsueecric tthodai If deta are too limited to
permit ptrantrit methods of assessing health effects, rank-order (nonpara-
nee ric) methods mj still be poasibLe. Soae gross variants have been used by
the Office of Technology Assessment for cLeasiiying waste into hazard cate-
goriee (OTA, 1963). A review of sore refined methods ia appended in Calls et
al. (undated). One of the options (nunber 4} considered by OTA (1903) r con-
sists of the development and use of an overall hasard classification system m
a cool Itir piidias the regulatory precisf on liasmricmi wattes.
As approach used in the Toxic Integration Program (EPA
1981), was applied to exposures specific to 41 chanicaLs associated with
-------
certain Lnduatn •*. Th« nsechod alio applied :o another irrnv of chemicals
tor the iron and Ǥ#'. industry (Clement Associates, 1962.>- A different
approach, mare general, in tccpe And coarser in classiricacion, is being used
irt the UETT risk/cost *»deL , developed by ICF, Inc.. for tne EPA Office o£
Solid Vase* (SPA, 19&J, 1584b).
A related jpproncb. is being developed for the National
Toxicology Program (SfRC, 19&2}. The system e*.anun«J4 avai I able infcrraac ion on
the subject chemicals (e.g., production, expOiures, chtnicai propert ici> %od
biological effects)f analyzes cbeir quanc icy and quality, and identitiea what
casta should be done on which chenicala, based on hazard potent tal **>d lack of
acceptable data. This scheme is expecced CO be useful in setting priorities
for toxicity testing. Ct (Ray alio b« adapced ai a mechod for ranking risk,
union data chat are inadequaca for quincifLcacicn.
The baa ic drawback af cha nonparamctrvc aystetns is thac
quaocicattve conclusions canooc b# drawn; only compari sons can be wade. If
the assessment is to compare courses of action to alleviate » particular
probLeo, ch«n th«§# methods nay be uaeful.
VIII-60
-------
Rafartmcas co Chapeer fill
ACCIIf. Pocuaaootiao of tJba Threshold Male Valo. 4oa-«it w«.
ACCIH. TLVf - Threaheld Lime Values for Cbeaical Subteancaa in the Work
Environment Moptid by ACCIH for L913-84. AMrie,m Conference of Govern-
naneal Induecrial Bygia&iets, Inc. Cincinnati, OH* 1983.
Alarie, Y. Doae-fteaponae Anal/tit Lb Animal Studiet: Prediction of 8ubm
Reapoaaea. Environ. Health Perapecc. 42 9-13, 1981.
Albert, R. E. (Chairman). Method for Oeceraininf Che Doit list E*cif»»c« for
Air Pollutants. Carcinogen Ajjcstoenii; Group, U.5. Environ®encal Protec-
tion Agency, UaittinfCOD, DC* JuLf 11, 1980*
Albert, R. E, Dose Rejponae and Extrapolation in Carcinogen Asseatneat.
AbtfCraet* of papers pr«9eara4 «c the Air Pollution Control Asaeciatioa's
International Specialty Conference» * Environmental Kiak Manageaent ** Is
Analgia Ujeful?" Chieego, IL. April 8-10, 1981-
Alcshuler, ft* KacBueauticel Overview of Doae-Reeponae Kxtrapolacion Model*~
pp. 349-3SQ ia lnnltA tlah Analyaii, C. t. Rictueond, P. J. W«l»hs and E.
D„ Copenhaver, edJ* FtmUiU Institute Fnit. Philadelphia, PA- 1911.
Aara, I. If. Dietary Carcinogen* «o4 A«ti-€«rciiwgwi# (Q*y$&n ftvtieala and
Degenerative Diaeaaea). Science .111 1236-1244, Will Lac ten and
reaponte: Cancer and Diet. Science 224 €5.f-§7Q» 757-760* 1984.
Amea. B. N.« R. Hagav, and i» $• Cold- Raakiog Poaaible Carcinogenic Seearda.
Science 2M 271-279, April 17, 1987.
teat, g. N., J. McCaan, and S» YeauMalci* Hethoda for Detecting Carcinogena
and Mutagen* with the SabaonelLa/HeiaMTlan-Kicroao«e Mutagenicity Taac.
Huut. Mb, W 53-71, 1*2.
Aaee, B. l.p W. E« Ouraton, ft. Yaasaaaki, and P. D. Lac. Carcinogen* are Muta-
gen*: a Sinpla Test Systea Coaabiniog Liver Konogenatea for Activation
and Bacteria lot Detection* IPtoc* Katl. Acad* 8ci 70 228i-£2S5, 1973,
Anderson, 1® l®» aad the Carcinogen Auatiaast Croup of the U.S. Environmental
Protection A|
-------
linuCtiBt L«# Sa Cold* t. Aa®aw ML C* Nk(| a&d &• C- Hod• S<
Tcuctlofoat Aspect* »f ili C®mpari»»«i of GarciaofMic fucumcy io %• fcmtf
mad Hies. W*mMm A—l. Toaicol. 5 Tf-*i§t 1M).
L., L. S. Cold, I. 1. Ahi( K. C. Flte« and 0* C. feel* To>icitf
•ad Carcinogenic htucjr (lattar). tlik Analyaia 1(4J IW-2W, 19S5-
lickiip II., and D. Kravaki. Seatiieical Das£|o afidi Analyaia of Che Loog-Tam
Caraiaogcaicicy SiMi(ax« pp. 125-147 In Tonic8logical Hiafc Aaiaaaaant*
Vol. 1. Biological ^ >tatlacUal Criteria. P. f. Clayton, D. KrawateTT
and !• Maare. *4«. CXC Prt«», Im« Boealacoa, n. IMS.
J.# P. Korriaoa, and *. foloam. aiak-Coet jfcthotfolotty for
*•«* Carpor«i« Kporc
•org, !• C. bpoft ol Kacloaal Cuietr Iaaeieata lyipoiiu*: Cowpariyoo •(
Mackaalaa* of CaraiMfaaaaii kf tadiacioa aai Cfewi««l Agaocs. l» Cot*
¦on Jfelacuiajr Machaai•«•• pp. 17-42 in
-------
Carlborg. F, V. 2-Ac«tylA»iao£iuorene sn4 the Weibull Model. Food Cocnee.
Toxicol. 19 367-371r 1961c.
Chandler, J. L. Maw Mechanistic Models for liik Aittstoiai. Fund. AppL.
Toxicol. 5 634-151 (L985).
ChriiCeasea, E. 1® DoaeHteapooi* Pu&cCioee in Aquatic Toxicity Teicing and
ch« Wei bull Nodal, tfatar Easearch 18 213-211, 1984.
Chriaciaftaen, £» R, and C. ?. Cheo. A Carta rat tfoaiatersetivei Multiple Tox-
icity Mo^el Including Probit» Legit, and Weibuli Transformation*.
Biowecflc* 41 111-715, 1985.
Clayson, D. B. Probleas In Interspecies Extrapolation. pp. 10S-122 in Tosi-
cological Riak Aaiassweat. Vol. I. Bioioiical And Statistical Criteria,
0. B. Clayvon, ol Krcnrak*, end Monro, eds. C& Preti, Inc. loca
latcm, FL. 1985.
Clayson, D. B«, D. Crevaki, and I. C* Kuaxo. The Power and Interpretation of
the Carcinogenicity Bioasaay. feegulet. Toxicol- Pharmacol. 3 329-348
<1983). "
Clayton, D. t., 0. Kravaki, and I® M«oro. ToxicologiesI fcisk Assessment
Vol, I» 272 ppl Vol. Ilt Z8S pp. CtC Press, Iocs taton, Ft. 1985.
CI extant Associates, toe. Health Sf feces Scoring for Industry Str P. i.t «ad K. EnsleLn. Scruccurt-Accivity in Hesard AiiiiineitCi
pp. 389-420 in liisnri AnttwmE of Chealcalt: Current Dey«lopwiitat
Vol. 1* J. Saxon* and P. Pither. eds* ^ Inc. tfew York.
WTTki.
Crouch, E. A. C. Uncertainties In Interspecies Extrapolation of Carcino-
genicity. ith. Perawrct, 50 321-31?t 1983.
Crouch, E., and E. Wilson. Interspecies Comparison of Carcinogenic Potency.
J., .of .Tweiepl. Environ. Health 5 1095-1118, 1979.
Crouchf E. 4. C., and K. Wilson. lit It,/ Btsssig j c Itniil yiii. fcelliager Publishing
Cowpeny. 1982. 218 pp.
VI EMI
-------
Crm^,, K. S. Kupout ca 0§i§® 9*tf7> Thcoratical !¦ tha Nodifi«t
Haaeal-ftrraa Pc«c*4a*«* §ttncl€» 33 1M-?SS, If If.
Crvap, Km S» Dona laapaaot Prablaaa in tta—Kricy 33 IJ7-
^ ^ l^^^~
Crump, K. 1. An taorwtd Prodiuri I«r Lt*rH3o»e Careisotonic Ktik Aisoimat
fron leiMl Doca. J» Iftvlrwu Pathol. TonittU 5(2) 473-4*6, tMli.
(Praquaneijr citad iaeofroec ~rofor«aa«; ieaCrvSp, "if 84® god Crump «r\4
Crocteutt, 1915}.
Crwtp* 1. S. SueiuicAl of Liour Eitrip«l«(i9A> pp. Ml*-192 ia
M^llbjm%h Analyfit. C. ft. Richmond, P. J. U*l«fct iM E. I>. Capankavar,
*40. ~ Franklin last i t uta >rau. Pkitatelphia, PA. Ifftifc.
Cwif, K. St Sfe"»';|jfKE I#* Diocfiaiaaeint Ictucn iaaary teas ftaapeaoa Model*
with Application* ca AaiMbL Carcinogenicicy SxpariMBtt. Coaawnicationa
km ltwi»fici« Pare A: Pwoyf tN Hexh&i* U<4)i|?J-JfJt llUT"
Crimp, K. 3. in teprova* Procaturo Car lov-Dota CarciftOfaiiie ftitk Attesiatnc
from Antral Daca. J, tnvtrca* Pathol. Toxicol. Oncol. 5(4/3) 334- J4if
Crunp, If. 1. * Maw Mach»4 Cor Oocomiaic* Ailowabla Oaily Intake*, fund
Anal. Taaicol. * 154-«?|( IH46. ™*
CrtHf, I. 8. Kacfcaaita* Landing ca Oaoa fteaponoa Jladako* pf, 135-211 ia
FrUclataa af ¦—lt> Unit tonmn. P. P. fticci* H. Prentic* Bail
I«. U|tiiaa4 Cliffs, iJ. IfftS*.
Cruap, K. 9. Ha thai* Cor Carciaagmic liak Ajaattraac. ff# 2?t-JI§ ia
PriacUloo of Kealch lt«k AiUtHmt. f. F. ticci. ad. frantic# Hall.
Crvmp, X. S., ma. M. 0. Mootorftaa. («vi«v and EvaIuncion of Machod* of Dr-
eorwining Kiilu fro® Chronic Low Laval Carcinogaaic Tntulc* pp. 154- 115
ia lavjgooiaatai In Pood, Of fica of Toclutolagy &i««ssaiBC
!• I.p and V. H. Uatoaa* CL09AL79 - A fOliH.UI Progra* co fjtcrapalaca
NdMMm Jialisad CarciMfMitity ismtm to Lot Daau. R*pa*t ca i»ct«wal
ImiIucm af bvifaaawtAl Maaicfe Sci«K«a on Caatracc Ml- BS-212S.
Augvoc Itlf. U pp. Cm» CLOtAl pragrnn baa baaa apiAtM pariodically.)
Craap, 1, S.t *a«| H» Vac too. Uacar Qulicy Cricaria Galculacadl (ran
Haitiatago ami Oor-Iie HodaU. laport prrparad for cha Environnantal
Procaetioo Agency, Cioainaaei, Ofl. 1980v 41 pp.
-------
Crump, K. S., mi 1# Bo we, A Review of Hathoda for Calculating Statistical
Confidence Haiti in La«"D©»» E*tra^ol«tio«, pp. 187-203 in Toxico-
loaical litfc. AMaatwant» Vol* 1» tiological mi Sgacittical Criteria,
D. §. ft^yfon» r'o. "fcrawskl. and I. Muavo, Mil * CSC Press. Inc. Qoca
ft«coa, ft. IW5.
Cruatp, K. S«, tad P* M. CrOduetc. bsproved Confidence toauLts for Ltrw Do**
Carcicogaeic Risk Assessment f rota Aoiouil D«C«- J. Hazard lUoriili
10(2/3) 41f-*4IIf IMS.
Crusp* X. S., D. C. Uo«l» C. i« Laagley, and t. Peto. PuAd«MQial Carcino-
genic Pracesaea and Their loplicatioe for Low Doae Risk Assessment.
Cancer Satearoh 36 2973-29T9, 1976.
Crump, K. 3., H. A. Casta, m.d K. L. Deal. Confidence Zncervala «nd Tests of
Hypotheses Concerning Dots Response Relations laferrsd from Animal
Carcinogenicity Data. Biometrics 31 &3?-4jl# 1917.
Crustp. X* S.» <4* Silvera, P. F. Ricci, and t. Vyzga. Intsrspecies Co"i?4f*i»oo
for Carcinogenic Potency to Hiaim. pp. 321-312 in Principles of Meslch
fcijk Asset a
-------
Dourson* H. L#, and J- F. Scare* ItftaUtory Itscorf and Enperioencj! Supperc
of Uncertainty (Safety) factors* togatut. Toxicol. Phanaacol* 3 224-238,
1513,
Dudney, C. S., P. J. Welsh, T. D. Mai, E. E. Calls, and G. 0. Griffin. On
UK Uae of Relative Toxicity for liik 1st inatioa. Paper Presented ac EPA
Synrpofiua, "Application of Short-Tene Biconyi itk the Aoalyais of Cow
plea Environmental Hi*tuXMtw January 2S-tl» 1982. ChapeL Hill, NC.
Durkin, P. 1. Site Specific liak Aisesanenc of Haiardoue Waste Dispose!
Facilities. Environmental Criteria and Aasesseent Office, 0.3. 8n»i-
ronaencal Protection Agency, Cincinnati r OH. 1982.
Ehlin|t V. 8® Quantification of the Genetic Btiak of Environmental Huta|«fiii.
lisk^Aiialji^i |(l) 45*'$?, 1988.
CPA. tACerin Procedures «a4 Cuidelines for Assessing Health Risfca at Sus-
pected Cereioogen*, U.S. EcyrooaMncal Protection Agency. Fed¦ Keg.
21402; 42 10A12, 1976.
EPA® Watar Quality Criteria Docweent*} Bequest for Canneries. Environmental
Protection Agency. Fad. *«f. 44 15926-13981 (1915).
CPA. Guidelines and Methwtolofy Used in ch* Preparation of Health Effect
Afimnnt Chapter* of th« Com est Decree Aablenc W*tar Quality CrLceria.
Environmental Criteria and Aiiexatdc Office, U.S. Environmental Protect
tion Agency, Cincinnati, CM* September 1910* and Fed. Reg. 45(231)
79347-7*357. Moveaber 1960a.
CPA. The CartinogtA Aasessauenc Croup's Procedure for Calculating Water
Quality Critera. (Typescript), Carcinogen Aaaeatoent Group. Cnviron-
mencal Protection Agency, Waahixigcen, DC« 1980b.
CPA. Water Quality Criteria Decunente. Environmental Protection Agency.
fed, fce». 45(231) 79317-79379, Koveaber 28, 1980c.
SPA. An Xnduatry Approach for Che legulation of Toxic Pollutants* Putnam,
tteyea, and Sartlett, Inc., Caabridgo, HA. Prepared foe the Tonic* inte-
gration Project* U.S. SnvironsMmtel Protection Agency, Uethiftgtan, DC.
Auf«et 15, 1481«
t?A. KCBA tlefc/Coet Policy Hodel Project* Phase I fteport, Draft. TCP, Inc.,
prepared for Office of Solid WasteT 0,3. Snvironaentai Protection Agency,
Uaekiftgtoe, DC. Jmm 15, 1982.
EPA. Approaches to Biak Aauuseient for Multiple Chemical Sxposurea. Susasary
of a Verkabop Held 1a Cincinnati, OH, July 12-13, 1983, Cnviroiuaan taL
Criteria and AuaiaMSt Office, lf»S. Urn! i nneemr >1 Protection Agency,
Cincinnati, 01. September 1983.
VXIMS
-------
CPA- Approach** to Risk /Uftinwic £«t Nnltifli Cfcaatical Ispoiur«». Pro-
coodingf of <¦ Workshop Hold ta Cincinnati, 01, Sopeonbor 29-30, 1581.
I. f. Sur« aad U I. Srdraicfc, ads. bnfliptiml Criteria aftd Aa-
ittMni Offico, f#f» Invlrtawitll hr«uctiM A|M<7, Ciftciaaaci, Ofl#
CPA-6O0/9-a4-00«, March 19*6«.
CPA. tlM i€AA lialt-Coac Ho4alt Hum# 111 t®p®n* ICP, Inc. Pn?fitt«4 fat
l ha Officii of Solid Vajca, U.S. Eavironaanc*! frotaction Afacury.
March i» 11841.
CPA* bviiid latirii Cui4«Li^u f«( tha Kaalch A«*a«M»nt of fuapaae Car-
cino|«i>, Draft. U.I. KtwlronaMuotal Protaccioa Agaoey, Motbiottoft* DC.
August 20, HNMk.
CfH« MaaLch Assaaananc DocusaoC lor Cthylana Ox ids. Offico of tfaaltk «a4
CavirQouMncoi iUiKMnt, U.S. CttviroaaiMtal hicectlos AgMcy, t>t§<§»rtk
Trvaogla tmA9 K. t»*-«00/«-«4-4aor. 19«6d.
CPA. Tachnlquaa lor cbm Aiiumdic of C«rci»of«aic li«k to cho U.S. Populs-
tioa Duo eo E*yo«ara froa Ml«cta4 Voiaclla Coapo«di4a ft* Drink-
ing Uacar wig ch* logostton. Inhalation and Daraal toutaa. C, I.
Coettarn, V. A. Con!(Ho, oat V, L. Ksrcus. CPA 570/9-6S-001, Environ-
Natal Prococtlon Ag«»cy» ttasbiagcoo, D.C.c 1963a.
CPA. BacooMndad K*»ioaua Cortcaatratioa LinHet. U.S. EaviroBmacal Procactioe
Agaocy. fad. 1m, 50 *6660, tfovaabar II, lM)b.
CPA. Guidalinas for Caroinogon li«k Assaastsenc. Cnwirooacntal frQKMCioa
Agency. fad. In. II 33992-34003, 1986a•
CPA. Guidelines far Hatagenicity fciak Assettawoc* Eoviroaerascal Protaccion
Afancy. fe4. Um, U 34006-340U, 19060•
CPA. Guideline* for OaalUt Aiiniaant of Suspecc Developmental Tozicants*
BnvirofUMiQcaL Protection A|Mey. Pad, lee. 51 Mill-1404o, 1966c.
CPA. Guidelines for the Health Halt Aaaessaaaic of Cheatical Matures.
ronwrotal Protection Agency. Ped. »ng. 31 34014-3*025, January 9,
I9i6d.
Krirtieht L, f., aad C. Survtect. Improving thm Uaa of Cpidoniolegteol Data in
Kaalch lifk Aatassnaat. Toaicol. Iadust. Isgiltli i(%) 63-61, 1983*
faraar, I. I.t t. L. KodalL, D. L. Gritwao, and C. W. She*. Dose and TUaa
1stpooas Hodaia Coy eba Incidenca #f Bladder and Uw Keep lean* Lb Nice
Pad MtttTtAbDnMrtM Conttaieedly«
3(3) SS» 1980.
FS4, Pood m«4 firvg AdaiAitcraCloa CoM.tt«« aa Protocol a lor Safacy
CaftlwclM* f«»«t m Cwrc |a«i«M«la an Cmuem Taacint in tht
tafocy evaluation of Pood Add&civaa and Poacicido*. toaicol. AppU
Phan—col. W 610-4Mt 1971.
VItI-47
-------
'uher, A. The Scientific Stsei for Raiatint Health Effects to Enpoaure
Environ. lapact Awen. R«v» 1(1} 27—41, ItSJ.
fmi Safety Council (Scientific Coe®ictae)« Propoaad Syaten for Feed fa£«ty
Asaatscant. rood COiHt. ToiticoU 16 (Suppiawnt 2) I, 1978.
-fu| Action and tha Relationship Between Prug Conenntration Ef-
face. pp. 28-39 is A. C. Cilaur S» Goo dawn, and A* Cilaun, ads.,
tmdman and CtltMg'i Tim thMrsmmlmitiml of Thggapactica, Sixth
19B0.
Cold, L. S». C. B. Savyer, ft. Magav, G. N. IacfcMua., N. d« Vaciana, ft.
La^ixuion, M. K. lo»p«r» U* t* Mavaodar* L. Barnatain, R. Peco, H. C.
PLka, and B. V. Ana. A Carcisogaoic Potancy Oacabaae of cha Standard-
iaad taauica of Aniaal ftioaaaaya. teviron* Bcalcb 51 9-319,
19B&.
Gold, L. J« i at at. Emwlwtm,. "H'mllk il 161, 1986.
V1II-68
-------
Griffin, 1. I,, M» Scbaeideraac, P. I. Eoterlioa, I* P. fcuilord, t, J.
Waxwilor, mi 1. k» T«Jt. Itiewnieni Session 2~~Dat«ceioo of (tea lib
t£f«et< «f Exposure to Low Doim of Agnfttt—Spidmialogie Prot>l««e.
Bwlron. Health ftrttxicc. 42 SI-ID» 1181.
Cut**, 8. A., and i. S. Cim*#. Uw-Oqk extrapolation of Date (rt*
Aai«ei C«ri»scto 8. A., aad K. S» Crap. Kaxiau* lifcelikood Estimation of Doi«HUlffsl«
Fanctloos lub]«ci co Abeolmaly Koootonlc Constraints. ten. Scat. § XOI-
ill, "*71.
Cuess, H. A. , K. S. Crunp, «xtiab Asseaswant, Vol. I» llo-
logical aod. Statistic*! CjjnTiriiiMFT]*^ 5. St CtayaonT Pi Kr—itrat CapOCur**, tinni roaaMncal Kaaaarch ConCar ( Infirm hi icai fro^^c-
cioa Afoacy. « ««ti, ON. July 12-13, 1913.
Ko«l r i. C. Ineorperatvcm of S*ek«roucva in Doaa taaponaa Model*. Eadartti'**
froc. (f»4. ia. Boo, to. Biol,) Jf ~7, 1980.
dealp 0. C. tftoorporAcioa of PhanMcOklr.acics ia L*«rH>ot« tiak KaciaMCion.
pp. 2«>-Il* im Tj»icolo«lc4l ti>b A»ta
-------
-------
KmI, D. «,» 0, U. 6ftyl»r, t. L. KirxMtlia, U, lafflotti, «»4 M« A#
ftehooidonaui. Eaclaatio* #f ftiatu In Xrrovoraiblo, Otl«]M Toxicity* J.
ttmltml* fcmr—. ¦—lth 1 IH-ISl, 197).
Boll And, «,, •( ml. Baalth E!Ueti of Particulate Air Poltutiooi fctapprsit-
ing cM Svid«oce, Attar, J. Ipldon. U0<>? 531-651, AfTf.
Sollatoin, i»» j„ McConsi* f» te|«loiaa(0t a»d V. Riceli. tbort-Tocw Tattt for
C«rciaos«na ami >KiC»t. fal» W I13-22*•
"•*'»« !.» B» i and If. I. Crwftp. CLOfiAt. 12* A CM^atvr Pr+ST«a to li(r«poU(|
Quanta! AnLntl Tcmlcicy 0*t« |i Lov Dosoa* Preptrid toy Science Kycia——ic tia* of Cheau-cala
Associated with Cancer ia Hwibi. Internee I oka* Agency for teeeerch on
Cancer, LyOa. Create* 1102 •
ItLC. Scientific bin for of Potential c*rcinotene «»<
Eatinntion of tieks. U.S. lotoregoacy fcrgktlator liaison Brevp* £«, C. !>•» «b< J. Otmllm (valuation o( Safetyt Toxicotogical Ceele*
ation, pp. 11-27 in J- Do-ull, C. o. XIsateen, «n< *. 0, 4adur, cds.,
CwMl; ti>4 Doull'a Toxicology. Second Edition* Nacxull»rit How York,
W. IfM.
Kodall, 1. L*« J. K. r«nHrr D. L. Creesmea, and C- 1. frith, CaclaetLon of
Distribution* of Tina to Appearance* fl Tuner and Tine to Death froai
Timor After Appearance la Mice Pod 2~Acetytaininof luorenc . J» taviron.
Pathol. To»icoi. 3CJ) «9-102, 19«0.
Kraybill. i. F. AaseasaMnt of Hunan Bzpetwre to EevlronnaAtal Contaminants
with Ipacial MftrftAcs to Cnctr. pp. IT-*2 In Tp»icolotlcol lilj.
AiHimUC. Vol. n. C«^r»i m4 c*%9 Hn4ioat 0. ». Clayton, 0. Krc«»kir
and t7 NtMror adlT CftC Priita, Inc. >oca &acon» FL« 1985.
Krevaki, D., and C. Irovo. CorcinOfooic ftiak 4s&««»Mot i A Culdo to tho
Littrattirt* tiaaatrici ST JJ>36$, ltll«
VtTI-70
-------
-------
Krowikl» ft,, and J. Van Bjrtin. Boto Batpona* HM«Ii lor QumiiI Boapouaa
Tosiclt? Poto* p. 11}i in Cmtmn% To»iea U> Probability o»d Itatlaciea*
M. Ci8r|£, D. Bowooq, J. V, K. ftoo, «ri E. Saloti, od«., 1fertfe~lfell«i»3»
tollo**. IMI.
Erjuftki, 0« > K. 1. Cn%, J. Faittr, D« W. Ctylftt, |» • * itovc, C*
t. talabwrt* l, L. fiolkon, mi J« Vu tjrzla. I Ccmpariaoa of lutli"
tfctl Method* far Low-toso BKCciftlaiicn Dtilitltf Tl»o-to-T«»or tat*..
fMf Appl. TokIcoI. I U9. 1913.
Lm4, C. z. Extrapolation frm Ur§«~So«U tadUtio* Cipoiumt Coocor.
if. JM-MI U Mt——t #f litlt tttm tmr*biFmi imwwt tm Mmdiatima
AM ChMHcoia: ACrUi<«l O^r»low. a7 0, WoodWadV C. J« ffcollalarwr.
V. Hmi, «od A. aolliwitf, tdirTlMi Prtit. tat ft A, «Y. 19K3.
Uawlota, E. W. fccitrtiittty la Conparaciva lltk AtiaiiMnc of K«s*r4ew« Woaco
ttenegaomcit Altaraaciv#a« pp. 11-fl In tnv».CTni^til^ii»te. MiauaiaE - I>
UffMlf Procoo4iitfa, Atr Pollution Control jUftoeiacioo
Specialty Coafaroooa! AKA, riteatn»r|fe. FA. If 14.
Lohauo, A. *ad 0. J. Wert®#!. 0m ¦uariratfold ftarfin of Safety. »ir»n.
Pathol. Toricol. 3(3) 17-24. 1940*
LyMut, W. J., V, r. Becbl, and ». ¦* Boa«*fcl«cc. Maailwok of Oamcal Feor»
orty taciamtioo. McGraw-Hill Book Coanpeny,
Mantel, N. Aapecta of tfca item Approach for Met to* 5*lt Doaea of
Carcinogens. p. 1397 la Oriaift* Of Mumm Cancar Book C» Mom II >h
Ayjmwwt, I. H. Hiatt, J- D.Wacaoo, and J. A. Vinatea, ada.Cold
tprlag Harbor Laboratory. M«w 19*7,
Muitat, i. Mmmim likelihood Yeraua tUaiav* Chi-Square. Sia—trlca *1
4??"»A#3, 1915.
Kancei, N* and W. ». Bryan. Safety Teating oC Carcinogenic Agenca. J. lUtl*
SlfflELlJIlii II «»-*«. mu
jUncal, i« and M. A* Ictuioidacvwut. EatiMt£n« "iJifo Levels," * Uaaardoua
UMtrtakiot. Caooor* l«s^ 35 Ulf» IfU.
Naacol* i.f i- Bohidar, C. Bnm, J. J. Tuksy. to ta^rowd
Hui{tlair7«B froe«4ura for ToatLag of C»icer m. M III,
If II,
Haroo, 0. N>, and B. f. Afiaa. M^iaod Mothoda for cha Balmotuilta Kutnitniclty
roac. Mtttac, m§. 113 173-215, 19*3*
flll-fl
-------
N., and K. CMkMriicmi of Carciaogaaic and Kueag&aic
M K. 1• Unci,
J. 0* Uitioo, and J. A# Viuen, *d*. Cold Sfrtag Karfcor LtboratnriM.
Cold Ifriaf Harbor, St. 1977,
ttaalgarfcar, S. 8.» «nd 0« J. Veoaaa. Hadali far Carcioogaftasis*
taaidance Ctir*a« far Childbaod mart Adalc Twaora. tuckaBatieal
giaitllftces 47 55—77, IfIf#
*oolga«%ac, 1. i.» aad A® C* Koudaoo, Hutacloo and CdAea*: I Hod*I for Uuaan
CareJ. Mac. Cmactf fascicle* ~> 103)-1092, 191L.
Hoolgavkar, f* H»* ab4 a» D«wa*jl. Ilologlcally Um4 Kodala for Cuctr «i»k
Asi«iit«nt; A Cautioaary Hoco, ' ii»k AMlytii £(i| 5-6, 1988.
D. I., sad N* E. AiMtaraoo. Qaaacitativai AiMtMHti af Baallb liskji
by Salactftd r«4»nl RaguLatory i|nci«H ft of fratant tractica*
with Sp*tial Atcancion to IwCarciMiaAlc Subatancaa* for
Office of Mr Quality Flans lag asd Standards, U.S. SmLraaattntal Procac-
cioa Agnatey. hmfeh Triangle MC. Octabar If*.
Morrit» I. H. C. Thoda, Jr., J. I. BaraAClkf I« fltcbar, P. D. Mcikouics,
J. Magy, «»d L* 0. Maftilcoa* Mathoda for AacaMLOg Cancar Kiaka Uaiag
Aaiaal and Imb Baca, freliainary Brafc. Beaakhavan National Labora-
tory, K«w York* WSJ#
Horrli, ft, C., I. fliahar, f 6. Haakavitc, K. Rybieka» ud U. C, Thoda. Biak
Aaaaaftautac: C4MM. MacA. Eai« Willi 40-45» 1914•
Nunro, l.t ami i# Kravaki. Tha Kola af tiak /UmkmhK Lb Bagulacory Daei«ion
Making. pp. 443-440 in loalth tiak Aoaivaia. C. R. BiahiKrnd, P. I.
fetltb, and t. P. Coyaahavor, ada. Praokl la laatitut* Praaa.
pfciladalptu*, ftu 191 la*
Mute®, I. C. r and B. Kravaki. tiak Mttstatot aod tagulatory Decision Making.
ft»«4 Cowit. Yoaicol. IJ 349, l9Blb.
iAi/^C* Friociplaa far fr—lttatiaa Cbaaticdlo km tb< tgnrmmt, Caaaritioa
fa* for traluaeiag
ChaBlcala Ia cba Eaviroonauit, National Aead««y of Sci«aev*r and Conatittoa
«>• <*• Chiinufi), Maclao«l Kataarcii Cauocil. JUtUmal
of ktaacu, UaabiBctM, DC. If75. 4)4 pf.
iCl* fttoa»«*y of TacraohlorooUifIao« far ^ottiibla C«r«iaofnicityt CAS
UMIri, (Socat Carboe t«tradiUrt4« v«» at 1 Poii(iv« Cor"
tzml ia chit StwCyh Oau Mo. CiW! Il-fll, Hacumal Caocar
taaticiita. Baclwa
-------
II Uhlan r«, K.» and f« NiyaflKXO* T«r*Kog«aic fffecci of Sodium ChLoridt la
rtiee. Acta Amanio 74 Hi, ifif.
NBC/MAS. The.. Effects on Fopulaci«na_ ot Exposure to Low La.velo of loeiaing
Kadiacion (BEltt IK Adviaory CQcaaictac on ch# Bvological Effecla m
loaning lUdiatiooa, Wacioaai l«#«Me€k Council. tfaeioMl ikidi| of
$cl«K«l« VaahiaftM, B.C. 1972.
MC/MS Fett Caocr*l: Aft 4si«fwo( of Frascac and Alternative T« HatbLaiCoo,
O.C. 19?Tb.
KtC/RAS. Tha Effects ott PvfOaciaM ft toptfucf to Uhv Ltvtlt of Ioeiftiof
tadiaciea d—| trail, HaahisttM. DC* INOb. 324 pp.
VftC/iAI. Drinkin* Vacar and, Health. Vol • 3® {particularly Froblane of H#k
UtimtSTwT^lr^rr^UkU% ttilfi? OmUim I J. iiri.ll, Clmi r-
ma) 1 Satioa*L laaMvcfc Ca—cil* VaCiM*! Acadta? fmi, flaablagtM, DC*
1910c.
WC/MAS. BAc»lAtL«i VcsCioidMt CMlctH m Aaalyoas fo*
Fattlcidao <1. Corfia«B» Ctuieiauin)» ttatiaoail Cou&ell. Kaeional
of Se4^«««» DC* IMOd.
tftC/WAl. to Dacaraina Maada and hitriciu f« Ttwlolty
ftlwiii II Pif^liipmnu RacidwO. Camell* HmtiinMii
Praia, HukiactM, 9C, lftS*
flll-fl
-------
ttC/NAS. ¦iak.Aaaoa—ant itt the federal Covem—wt; Sanaa Lay the Procasa.
Health Cl. A. Stallonea, Chairman), Cetnai a«loft oo Li It Sci«nc«i > national
taaaarcb Council. Statioaal >c«dwy Preai, UaehiagcMi, DC. IKS,
Itl »*
0f FlaJttarcy, B. J. Differences It Metabolic 4C Oiffareot Dose Level*.
pp. 53-9® Ib To»lcoloiical liate AateMPeac, Vol. I. Itolonca L^SS*
li< 11 at i cal Cr i ^
Praai, Inc. iocs Raton, PL. 1985.
Of TV* OMB&cfll CatciMceaa: A ft««t«« »f cbe Select entf tta 4iMct«tt4
Principle!• Office mi Science mA Technology Policy. Fed. Reg. |l(S0)
10371-IQ44, 1983.
OTA* NcctuENi* fer Assessing Health Risks. Chapter V (pf» in BairiLrgo-
natal Coat— iamatl- i? foodi. Offvee of Technology Amos anon t, U.I. Con-
mil. OTA-r-l03i U.S. CP© (neck Ifo* Q32-003-00724-0), Washington* DC*
December Itlf. 21$ pp.
OTA. AfitflftiMic of Taohnolagiea for DeCtrtiitung CiBflar Riaka Iron the Envi-
roiWMi. Office of Technology Asmmnoc, U.S. CoR|rni. OTA-«-l3g,
O.f. Govenwest Printing Office, HuUa|iN, DC. Iffl. 2*0 pp*
OTA# TechnoI9gtea «a4 Hanageeent leraca|in far Itasar4*iu Milit* Control.
Office of Technology iaiaineot, Uuhia|(m» OC. 1913.
OTA® identifying Ml Scfulacing Car*in©g«B». Itekftound P*^r« Officii of
Technology AisnuMti U.t* C®a§w«s» 0TA>"IP-8*42 - 0»l» CiwifMiMt Print-
ing Office, V«ikiA|(N, 0C. Iff* 14f fp,
Park., C. N., and R. S. fnea. Qurnatlcatiwn Risk AitosBaaeftci $tete-of-the-Art
far Carciaa|«B«>ii» pp. in fci** Hiniiiiiiiftt of Uiulai Ch—icalo*
ChMical Manufacturers Association* Uaahiagten, DC. ~J«ee 1904.
Percira, N* A. House Liver Tuner Data* of Carcinogenic A-etlvicy.
TonlcnU Iflaiiifc> Itoaith Jf41 31I-JJ3. 1913.
Pato, X. Paynes pniwiai at Conference on Lm Doae Risk Eatlaatioo, spen-
sornd try the ttatieoal Xsatitete »{ InviraiMB(«l Baalth kineu^
Veightsviltn B«w«Ji» K# fctflff lf?4»
PitOi ft. The Carcinognaic Kffecta of CkfOttic Kipwe to Very ton LeveLs el
Toain Substancea* hviw. Beatth Pera poet. 12 tSS-itf, Iff?.
Peto, t< Epi4aa£oleglaal Reservations About ftiik Aaseaanent. pp. 3*16 ia
Aiiiiawnt of Risk fron Lo»-Lcvsl Eagajure to_. Radiation and Qiamcalfli A
Critical Overview* A. 0. tfeodbead. C. J. Sballabars"* v« Poodi, &nd A^
Pleaua Preaa. Vev Yorii, IT. Iff5,
N(o, t#v mvt f» i. Um« Hiibll OiatrilMiCieaa far Qmtim—i-Cagcioajeaaiia
bfiriaMU. >ic—erica If 4f?-4f® (1973).
mi-74
-------
Poco. I., M. C. film, L. Barnieaia, U !, Coll, a*d B. i. Ami. THa TDs#i A
Propoaod Gaaoral Ceowtatittfli for ch« IfaBerical Doocripcioa of th* Coroiao*
l«i« ><*«•€? of ChwLcalt ia teiatl Espori*oaca.
Environ*. ilaalch Portpact. 38 1-8, If 84.
Porciar, C. J., aftd 9. C. 0001. Xiaua* Coacomiag tha Esciaatloo of cho
TDf#. lilt Amlyma y<4| 4JI-MI, OteMbir 1987-
Aai, E.f aod J. Vafi *yain. KlfLTIMi k Coa^vtor htpis lor Risk JUtMiMmt
of Toxic Substances. Technical tapcrc No, i^WIJ-IISiS, Raiui Corpora-*
cioo, Imat* Mofliea, California. 1980.
Ui, K» aiMi J. Vaa lyain. k C*ner«liaad Ifeltiktit Doct-MifoOM NMtl far U«
Boaa Extrapolation. 8toa»crica 37 341-392 (Itiil.
tai, K. and J. Van iftia. A Ooaa-toapoafa Modal Car Yoratolofical Expariaaac*
ta*al»i«if Qwmtal UifMMa. 8ioa»tri.c» 41 M ClfIS)*
Matty, 4• C., C» If. hrk, M. C. Oct, ao* 9, J» CmiMfNic Riik
Aiitiiweot i Dibraaido. Toxicol. Awl. Pharmacol, 47 41 1-414»
1971, ""
taaaay, 1» C.„ im4 f, J. Gafariaf. 7h« lAeat*«tioa «mI A^Uutiat if Phartta-
cofciaatic fritttlplei la Raaliseic Cscimattt of titk, fp* IU-M3 lo
8filS.lt Aftalyaia, C. 1# Ricbnond, P. J. ttalab, and B. 0. Capaftha«or,
•<».Fraakli* Xascicuto h«s«» Philadelphia, PA, IfII.
ticci, P. P., H. Intact tit# at ¦tltii. fciah >i»t»iant. ProneUi lull,
Inc. Raglovood cn77»7~SK 199$. 417 pp.
ftictaacMui, C. I., P. J. Malak, aad B. 0. Coftoahavar, ada. tealtfc litk Aoaly-
*i». fwet^fiiii of tha 3r* UN Sci««c«« Sjopaalua, C*(iiob«r|, t*,
Octafcar 27 tbroafk *t IW®- Praaklia (aaticuco Ptaa*, Philadelphia,
pa. ifii.
tow*, y, 0. , with oebara. Cyal^tii«c»ia»l«rtt of eta
Statistical Aaatfaia it Mjaacidf f*r tiM on Scwdy. FyJ. 4ool. T«k. I
81-17, 1581.
Schnoidonaaa* M. A. CxtrtfoUtlo* fro* Incoa^Loto Qati to Tocal or UfoclM
Mito ac Low Bb«4m. Peraxcc, 42 ll-l®, 1981.
Siolkaa, A. L., Jr. Ao-Vx«aiMtioo of tkm tOB| Itwdy • Alik 4»iaM«taC Dalnf
Tt». TW. 1 M-m. net.
VIII-™
-------
SieLkan* I. U , Jr, Some C*p*bi Li ait At loot, tuki Pitfall* km cba
Qu«Aiiticlv« ftiak of F*nnLMii^af pp. 74-122 Lb li*k AaAlwai*
ta th< Chgatxcal Ia4mccy. CfcovicAl MAsufActiurer* Aitociacion.
w«ftuotcon, K- 5apceob«rT$95a.
Sielkaa, A. L«» Jr. $m Ihvmi in tha QuaotitAtiv* Modalinf Portion of Gaa~
car fit* A»$et«MK(. ftc»ui4t. ^fftarn* To«. 5 175-181, 1905b (also
pp. 232-261 ia Awalvii-a tft cbo Chiwucal Industry. ChaaicAl
KamifaeCMrir* Association* WaBmogCOa, DC« jUptembor 19fi5b.
SioHura* ft. t. in# ®l cJms Barelay-Slalkaa Itoditl I* Uwftiie (xtripoUUla.
pp. liS-lli La TaaUoloaical ftlsk. Vol* l» ftiolottcal toi
jltAtvseicAl Crleaela. 0. I. Clayton. JCrawaki. -and t. Hitnro. ad*. OC
Pyam, Zm. Boca tAton, Ft, 1915a.
Stalk**, ft* L*. Jr. Caabiniag Pfcjraioloiy, Carcinogenic Nachanin, *4 Intar-
individual Variacitfi i« Cioe«r Doia-fcaspaft«a &xtrapelaclona. Pratantad
4( cka SocieK f far || Ik Analy*is Aftftual Maatlat* Kouatcn, IX,
Bovoator l*» Ifil,
Squirt, t. A. t#BkJ»®§ Aoiaal Carctnofan*: A hopcsH tagulacory Approaok.
feianca J14 177-IW, 198U
Staffa, J. A.# md H. A. KaAdUaaR• loaavacicma la Caacor ll»* Aminmoc (EDat
J. emw. Park, TakLcoI* |(3) 1-244, 1980.
Start, T. ft, Tho tola of NachanUtic Data la Do«a-lta*p*a*a K°4«UQ|.
ft* 101*124 in Uifntct of Ki»fc fn«B Loj^Layal Ifpmmrtft co ton
And OxNtiolai T Critical 0»crvia». A. P.^toodhaaA.C. i« Shollabartar.
v7 Pond, and A. MaUaondar, »4;», Planum Proa a. Nov fork, OT. 1985,
Starr, I. ft. la^rovinf Quaacilaliv* 11* luiMttt vUli HadumUcic tt*c«i
Tbo Caaa of For*al4afcyd«. pp. 141-133 la ftlak AaAlyan La ska cha*ucal
Codwicrr* Ckaaieal Maaufaccgrar* Aatocimtion. Waabingcon, DC. Sepctabor
ft. C. , aod |* i. Jaffa. Nacbaiaac ical Medals isi Safac? Faecara-
Lifa SfocaaM, lac.» Clarvalaod, Oti* tfovaaabor 1962.
Thor«iua4» t. W. , C. C. Browo, sisim fl. Cfeacvlay. Tho Uoa of BiaiAgically
KocUata* NachaaacicaL IMaU cm Predict eh# Actual Cmcmp liik Aaaacl-
atart wltfc BanrirottMBCml bvoaiiM to • tisk Aaulvsia 7 10f-
ilf, iff?, —— -
Trwiet, J.» J. Cornfa14» and tf. KcomL. k Moitiv«rijic« taolyiii »C tlui ftiak
of Cor«Miry temrc Msmm im r*m*wluum. J. (»na. Oia. I# 511-524.
1967. ~^ —
V®a Camrt, L. J-» Aftd A# i. Uct>ab«Aij*r, «4a. CaayilACion of Odowr
Tk««A»i4 filaw l» Air ttetor. U*ticiK« o( Utttr
*Hrtw|, *KbmeUm4». ami Caaeral ZA«t&cueA for Ihcrieioa And
iMwtriAi TWO, ZAitt, IfochtrlABdi. IfT?.
I
-------
Van RyiiOj J. QaancicaClv« Risk Assessaecx. J. Occupat. Mgd. 22(5) J21-326,
1980.
Van Ryzin* J. &«-ExAai&4tiot) of the £Dol Scut*? - levies 8taCi9Ci.cs! The
N««d Sar Realistic Statistical Models for Bisk Aiiijiiesti fundi. Appl.
Tq«7. I 124-126, 1981.
Van &y*in, J.p »nd K, EUi. A Doi«HlespocLJ« Nodal Incorporating Honlinaar
iCLoecics. Biowetrics *3 94-1051 Harch i987.
Uainburg, A. Science and Its Limits: the Regulators' Dilensna. national
Acadasy of Engineering Symposium on Technology and P«imes«.
tfaehiageon, DC- Jime 1-4, 1585.
UfcictaBora, A. t md i- 6« Keller. Quantitative Theories of CarQinogeaesia.
HAM laview 20(1) 1-29, 1978.
Wicteroore, A. S. M#ctuta«CL<«l Models of C«n««r and Their Use in, ti»k Atien-
nant. J. Environ* Pathol» To»lcol» 3. 153-362, 1980*
Wilson, J. C. Use of Bhasua Hocfcry* La Tautological Studies. Federation
Prcc. (F«d. Am. Soc. E»p. Biol. JO 104-109, 1971.
Withey, J. ft, Pharmacokinetic Differences Between Species. pp. 41-52 In
Toxicolo«ical >isk liitimnt. Vol.. I, Bittio«ieal m»4 Statistical
CrtE^uT «b5 lT~^»rft» ct<<» CfcC Prast,
Iftc« fcoce tscoo, ft. 1963*
Uoodhesd, A- 0~, C. J» $h«llib>rttc, V, Poadp sort A, KoLl«ead«r, ®di» Aimi-
miaac af Risk from Lon^tftwal ftmpojurn; to Radiati-oa and ChaauLcalst A~CfIt^
ical O^r^igw. Plemias Press. Hew forks MY, 1963, 329 pp.
Mrxgbti &. H« The Sanse of Swell. CK Press, Boca. Raton, Ft. 1982. 248 pp.
2elsa, L. , &« Wilson, «n4 C« Crouch. Use of Acute Toxicity to Estimate Car-
cioogenic Bisk, tipk Attml. 4 187-199, 1984. (With response to letters,
ibid. 5(4).265-270, 1915. ~
VIIX-77
-------
-------
ix. ggtmmm mm swurotf or mmnsE, mmcts
Tha ob]tcti«ti of loii riifc ***a«uMBta my ba Mt If determining a
givea health risk to the mic upm4 individual (i.t>« lifetime cancer rlak).
Tttw May «thar MWUMBti, aewrrer, and particularly for CMHriioa t(
alternative technologist* an as t in* t iota of tha total dvirM L*pact» cm to#
essential. Alter the popadaciont*) lifcely t« bo espoeed hav« been identified
imi the envirorneeotal doae(a) haw been eaeiaated » than
for tautay tociaecoaaakie mA ecologieal lapaeo.
A. Sunan Health Ia»ict»
- "" mir
The exposed hunan papulation could toel^« aaaatort of the general
public aod individuals exposed occupet ionally la actlvitiea directly related
to fcazardnts* «ait< diapoaal. C«nc*yc«nllyt eh* ntchotfa of predicting idveria
health effect* are mch tha iw» (or both group* or for subgroupa within
aithar. Tha following diseuasion (aou*i prlntrlly m publU health effecta
reaultiog from ia*oloncary, often uaideatifi«4( dnfMUFft, ind rotes turtleb
effacta froa occupational expoaufea only b#i«fly>* In a conprehaatUc hum*
mmnt health riake would b« lBC«§e»co cootaia tuh-
populitiooi that are parciaularly aeaeitivt h«auao of «gat aaxf general
health, ganatle deficieaey, act., I.e., thay hdva high caxiealogical mpii
at a given ezpoaure to the pollutdaca*
Ml 'sa|siifmpmmaem «« fine ti«acifia4 mi
^ua0Cifi«d« Ml aifKifieaac «gfa«iailr voberlmf ara ideati-
fied and if paoaihle If the dlatrihecioea of tha two auhgroupa
are tab*tancially dilfavest, aod if data are aoailahla, it mj he daairehle to
~ OocsLpatiofial aapoauraa ara goaarally battar «o&itorad.
a-i
-------
divida tba fetal papulation loe« dlacreta expoaura/aeatitivicy aabpopulaciona
bifen proceeding vich tft* aMtfiU* Probablj «on fra^uaatly, cha Laicial
aatlytU auuic It* uiit primarily ea Cte basis of the Ur|«it aapoaure *ro«ps»
aad fit itftittivc aubgroupa «r« ebaa contidYr«4 qtuaticacivaly, aa de-
acribed la eta next subtree ion.
ta Baking eh* health tlftcci **euaati©e» m ivirifi enviroa-
aencal doaa for cha individuala la each upoivrt graup ia Idaneifiad for *«eh
ctumical of concern. 4 riak factor (ft) Cpro^nbiLlty of tdverae effect) at
Chi a dot* (d> ia then ntrie(«4 fro® the doic^iiponai raUcioaship fot th«
cheaical, uaiog aicher a fr«phlc«l or aathenatical protestation of cha rele-
tieoahip aa appropriate* Thia la CM tvtra|« risk for individuals
wpoaai at ike lewil. Tit® risk factor ia than Multiplied If tin m»afcer af
iodividaala (•) m iKa asfaiura groupa to calculate the ftuaber of caaaa of the
•ffatt far tlut irou|.* If chare ir« aauLtiplc exposure graopa, than « riak
(KCar l» identified ic the 4oia level for cadi |mf.w The risk factor-
iitkyapululM prfMlucci art chea tuBaaed across all axfoiara groupa to eatiawte
the total ftvaber of eatai for thmt particular affaact
Prudicced nuisbar of caaaa • a ~ l|ft ~ ... » lRNft
o
or Predicted total €••«• ¦ I
•Mm 1| mme. the riak fwtari at it* doaai Ci|), «Ad
are nmaabara of people exposed at dp
A slightly diffaraot variioa of thia calculation ha* boon used
by the CPA Carcinogen Aases sweet Croup (CAG) baaed oa a veriation of eh#
Multistage nodal for cancar Cor othar appropriate Malih *ifeces). Tha
excrapoiatad probablii ty-doaa lino is as suited Co approach liaearvcy ac vary
low doaaa (l.a.t In tha probability range arou&d 1§"4)» to chac cha slope
la oaarly a conatane (k) .++* Atauaiog linearity,I
• A eiaiaaeai thgc tha pt^^c StVt ia the mpMlMl vaiat of risk
and id oftosi treated ia CM llterecare aa a dafiaitiM of risk rathar
thaa oaing pNbafcillty aa riak or a riak factor.
•* A ntl«W turn assart chat liafta applicatioa of riah factara to a give*
ptpmliitlna* ebacistee the tigsUiieanc diffaraac* ia ujaat oa
tocitl fatalitlta (a.g., »cci-
d«RCt) ««ul thoaa tJut tra d«l«y«d Ci.a«( lataat affaota).
actually oaaa tha alopa (K) of cha uppar f$l coofidanca
liait racker thaa cha k«»t aatiaata of riak. Tha Uetar ia uaad harain
It* raaaoaa daaeribad In Quptar I#
| A laviavtr kaa ooead tk»e aaay raaaarckars baliava tha asauatption of
liaaaritf ia ovariy "ooaaarvativa" in viaw of frowiag avidaoca of greater
•Ifaceiweoess of biological rapaulr ¦achani—a at vary lew doaea*
&®is alapa sssiy ifpiMCb s«*.
tl-2
-------
> • k i Or * * *
tba mabor mi MK< of etiMfr la oa expvsore gw«f U cban calculated bf
anltiplyiai th« tadividoai 4o»« (iccor, k, by tbt liu, laaoalljr as m cow-
camcratioai Cc) and by tba mteber of iMiwikali in cho group. The total
m«6«r of cancer case* can Chan bo utimc«4 coaNMhat «or« easily than above,
ainca the aunasation ratrrtnfif to*
n
Pr«diet«d total cam* ¦ 1 I O,
1 4 1
If tho 40Mafispwu( cvlaciddvfcip *®» M ceres ef rMi»«4 mistier will also be for extra c«m».
"ft* aiaqplofC COM fot aaelyais i« that Of * fepviacicKi uaai-
feraly asfeaed to J fixed level ef • »ia§l« ehaaical ilut produce* ow health
effect. for esaaple, « kiUf^Mi waste disposal aite mf eascatiiuc* the
drinkiag «K«r aapply for « imU city vich a duwical which can cause nervout
¦ fitt* atltcti apoD prolonged exposures The cast becomes mm complex as
•uLeipU axposcrre condition*. asiltipla chaeicala, or multiple health efface*
and Multiple tubpopulacieoa tsuat b« caoairfered.
If thai chaaaical of coocerti under a given axponrra condition
produces aultiple nfitcei, tbaia auat aLl be considered! calculate «acb effect
with cba «pfc®pciat« risk factors Mi aiabpopelatioias* Kot«f , that one
•r im afiacta mmj preioddlneCS «« CfcAt tfe* otteri CM he Mflactei w la®
adalydii behave affacciac let nlUitf ii|aifica0(ly« TIm 4adai«a nrittria
coaaoc bd •imflj seated* decisions mill nvfoire prefteolOBel jad^e*t aod
cenaideretUft of available resources aod tia*. If tint ahead eel uoiar dlf-
ftrwt rmtii of espesure produces the sua affect but different ratponia
rates, tlaply sua the estimated effects of each exposure for eech
tubpopulation.
If the cheaUcai under different levels ef exposure has multiple
tffieci, then each Of these combination* should be considered aaparately for
•id) tubpopeietloau If aulclpld chaatieals that cause dif farant alfacti (as
•left «lat«4 vaitai) *r* fmmmt, thm affocc of «§ttk on «ach
UfiM mf Im TIm of iimta c^«c*l» wmy bm twfciit«ci«llr
•islldr* (MM«air«r, «hick tioptlfiM ell* li^Ltarly, affact b«
of irMtiic (ieiiciwti hiicaii## of itt ucan m tocMM of cb* l®»5l ef evpHMM
ta CIm cfc^«ai caaaBiag itv and tba aakalysia is aiM^lifiod*
Dm hftUfttiac df dlafdvat* sffocta frwa oltarnaeiva! eactaMl*
ogiat ia «U>tiaMBt«lf • ptrt of tkm rift, auusafooanc procaaa, but cho riak
aaaaaaranc cab ba excondod ac tiatf to halp put tWa provorbUl M*ppl«i and
orattfoa** problan io uadful constat* f«t »«aplo« tb« nuaibor of cacaa of «oeh
zjpm of affoct rcanLtlftf f*e« adcb option c«n bm tabul«ead for eoroptriaoni th«
toeol mabbar of projodtdd laUlUiii of Mch can bo coaparodf and the total
ouabor of La4in4wla offoetod Htfitallr can bm cooptrod* Io tern cases tho
n-j
-------
dtcisieft am iters oight roquo»t chat *11 projected tffocct be rtone tired ie4
itircpiMI for mth apcioa for still uacMr booia of eoap«rtsoii.
b. 3ea»UIvo a»b»omlo ci«»»» If tbo czpotod population coar
taJLca aubgroupa chat differ grmatiy ia characteriitica, then thaao auiat alio
bo cettti4«r«4» particularIf if chay n«y tew o bighor roepooao mem (l.o.«
chay mw bo ooro aosceptiblc). C4Aarallr, the goograpbical iococioa of fisad
icracurta («.(., hospital*, schoola, mt§log boawa, rocrootlocal md
private voll*) within tha ir«« t( «uf Eloloitt to identify aaay apacific low*
tiona vhoro the geuorol popalatlooi deaoitiaa can bo nodifLad for tha
tobpopvlatkcm.
II teratogenic offacta aro possible fro* iht expottircer then
prtgnut faaaiea aro aivoya a aubpopulaciea of cetetTo. In this e»§« « stair"
da rd birtfe raco (currontly 15.f livo MrUu p«*r 1,000 population i« tbo USA)
can bo opplfcod co the ospoaed ftoorol popular Lao to oathaeto cba awabor of
eatcs.
A >«nplo work*boot fir surveyl** the potential haalth effects
on a nuabor of svbpopolAtiOttt It fivtn ia Flfur* H-|. Tbo varfcabaat
addresses health effect! fro* f§*r tt^mra roucoa (iaUltcioa, drialuog
voter, dinal, and food) and allow* raAklbf of effocca »e Low, OMUiuai, and
high, whore "I®*" would be above « di__«ialalt I oval tad "high" night bo
easigaad «a m fatal efface fcy the aaalyec* fkocormiflotioai of the ozposuro af
cB*t a^acial mmf to coaaidarebly morm ia|tmiai thaa that of eha
ganorally oxpoNi population; til» iapvaciaion would resole In a correspond-
ingly greater level of uncertainty ia eta estimated health effects*
I. Octwocioml health offacta* Tha «echod of eat location of
occnol health effect* in n^««< wortarT t» conceptually tbo mm m for
«»cu»aclng public health ifftett, escape cb«t the place* and oondieiooa of
oapoaura are difftreae and tha nuftber o< warber* to bo «oaaidored will uauoUf
bo roloci««ly MNill. SiaUlorly, •ubpapulatioaa of expoaod worfcora could bo
ooolytod aoporotoly ms obovo if 4tair«d In o rigoroot trmimmt,
ifOMiator of occu|totioMil boalcb fiJku la « quoncLtaciva mooaer can
ctsfiiiiri «ubocMci«X tffort to iewtlt# tit«- o»rfc^c«tiwi-«peeifl<
aeonorioa to oaclatato tbo «u^oai«rta iacwrr«i if mmf dliffernt kiiula of
varfcora aotaf«4 in. tbo voriotta octirlclot io«olvod £o troouoiot, atoroft,
diapoaol, or trooaporeotion of th« bocardoua viitia afld in cerrorelvo actions
or cleanup oporoeiona La tuo of abcci4oaeat apiila* firaa, or lookogo es
3*iSEBf':S| «t(cri« A sla«wfS#B of Mcu^cltaal risks sc«is» alcaw
s!»t 4l§pi*iim codmolof lot sis/tssil alao iacl^s est risers ia coaatnc*
tloa or HiDctmci octlvicioo ualqwe CO oajr of Cho a,ltamativca« Specific
aaacaaptiocui Mtluc bo ro^uitod rofordlnt cbo kinds oad officioocioa of proeac*
civo eltdkiat tad «qwipaiAt wtllltM bj tha varia«o tad otbor *rk«r
|wig«sl«&t. fl.;| alfOrt M(NMic4 could ta bitit for sack hasariMs waata.
Moro ^iuiliCBti«0 tec Itftl coidf aatiouataa of occupational, risks my
bo mti§tmzmf for ¦mmtm ia tin oajor
eoacm« For asaaplo, tbo trootMratv iitngt. dtioposai, traaaporcoeioo, «ad
corroceiro «cel^ tceaariot Mm Mcb, atreauai coo b« co oatiatto
11-4
-------
M I i li n j j , | . 1
11i!iI!iIf !U 11 i JIJ
M
u
o
3
(i
o
fcj
M
3
£
3
I
I
B
II
£
I
o
i
*
-------
cho ivartfi Bumiiir of vorkara cxpotad. Ttvc likely ramos tod rdlocivo lovol*
#1 ufaturi cab bm considered for etch, M4 Um rfUtiM tocal occupation
f|$fci of *U«rnativ* diiposol technology** (boa eon bm «ici«i(td. These
rcnlci are conaidored along *ieb thi relative public health riiki nM my
ether ftecori eoed is reaching the deeitiaa*
i. Eealoticil o»f Otfc#r papacta
Cc«U|iul, other oovirena«at«l and aoeioaconoaiic vaLuat night Incur
ulnrit effects «id«r mm technological #|t««nmtiwei» possibly |W«f than
potential htaieh lifitd. Only brief nota mii be wad* hart for eddraaaing
ecological awt lecioocooaaic «(fa«ii. A further brief review It given In
Appendix C* b«l a fvnkw discuaaion of environmental and nchnr iapwCi ia
kyttMl the teopa of chit report.
Conceptually, thai approach outlined for vuaaticetivelr estiaatiog
health inpacta on huaana could be appi.ted m other iptc)ai. froctle«lly»
Ikovmr, cMprtfeauif* quantification of awch iapacts mil be lafeasible in
moat cms. The vmtimzj Of doanatic an tea I* and cropa, wildlife «*d ece-
aystoae are often siaply coo gmit and tha data biia i» usually inconplect.
la « f«« ut«s, considerable dec* ney be available for quaatltative e*ti
feroat rocipieatt* CotwiaeoaQy In mitad and in proa«ntatioa of
{Lmliilfll mmmtttmlMiim} for sll of ete i«feas»l t«smi ope ions £• oaatncial
for iofotaod docitiOA MkiQBb*
a-#
-------
x. AiuiYin or agmiMnEi in 1111 Mteatttptt m
Dacidiaf hotwaaa «ltfcrn«ci*e dwicai if actioai rtq«ir«s c®w«i4cr«-
cio«, aot m,tf of tba anticipated banMtflt* «ad otto of each •lunwtlve, bat
•lM Of kow citliiaat on* is in cbe tetc mUttH of tbaoe benefits «ad
cfl«ct. la tha ragwlatioa of techooloty» a decisioa mir It porttcuiarly ccm-
caroad Ely*f, actual health, B«iroowie«l, or aoaatar? eoici aay prow* co be
oaaxfoccadly high |i.«.t riafcs arc ouicrtfeiMtH) or that the banefLta
Cp*rti6mi«rly raduacioas in risk.) Mf ha wdi lower thaa MieieipAtcd. ta
analysis of (ho uncertainties is inch cstioatea la therefore highly daair*
*W«. A acaiamont of tba degree of confidence eho analysts have io
of cha potential health acid envirsnaootal effecta and tba coata of alternative
baiardom milt manaiaoent practicea should bo An ieportant input inco cha
decision aaking preeaa*.
Concern over uncertainty enters lac# tha decision oaking process In
ac lawi two mftt
-------
only Indicates * quality or trace of being uncertain, but cen also imply a
scitemeat concerning the qoalicf or quantity of supporting objective evidence.
It Is la chit lair area tkae uiJLft differs most greatly in the
statistical, biological. social, pfc.fii«ai» and engineering science*; in busi-
ness economics\ in policy, legislative, and regulatory actions; and in
judicial proceedings. Uu|i may differ ureal those describing experimental or
historical fjctuM data# those describing most likniy future events, and those
describing che future as a spectrum of xteejrioi or probability distribution*
of possible future oupeowts. Souse sorters describe uncertainty in quice
qualitative cerrss, meny describe it us statistical terms, and some use uncer-
taincy synoaosously for "risk."* la. chit study the following definitions *re
used.
"Uncertainty" is a scatesutnc of the degree to which a sysceoi,
process* or measurement, or the components thereof are clearly
identified or defined, or che likelihood that an event will
actually occur»**
"Uncertainty analysis'* is J procedure tar attestpcing to
quantify this statement of uncertainty. In particular, we Mart
co quantify che uncertainty associated »ttf| estimates of che
number of cases of specified a
-------
i. AttftLTsia or naniTAiup it ti« Atsggitw or
0eci4is| between el tentative cfcoicea of ictUa rnquirea conaidera-
tioci. 00( mif of the anticipated benefit* and costs of eecli eltarnative* bat
ti 19 of tew c«afidcac an* ia La the btit eatinneea of the a a benefita and
eetc»» la iht regulation of technolaiy, s decia ion ukir ia particularly con-
cerned Chat actual health, anviromaental, or mooaeary coata way prove to bt
uneapeceadly high (i.e«» ritke are unarmstlftitedJ or that the benefits
(particularly reductiona ia risk) nay be mtch lower chaa anticipated. Aa
eaelyals of the BEucertaiAciaa in inch eiciaetai li ibirtfori highly desir-
able. A ic*ca**At of the degree of confidence the «iulyt«» heve in estin»ces
of the potential health u4 envtroonentel tffeeci *4 the Hill of altenutti«e
baaar^Ms waace nniiageaMnt friecieii ahosU be «» l^mTKmt, Input uiio the
deeiaien ariuoi process.
Concern ovar uaoreaiaty enter* iota the decision Matin* process in
at least tno ways* (a) asauating chut eta anticipated benefits of «a actio*
ara achieved, ntut ara the chances that they will ba eere thee offset by un-
•spt«ied incarnalized and axtarnallaed coses? l.a« > what ara the rlsksfi And
(b) *aawning tkuiL ona haa aaaaaaad theee risk#, raliabla ara the methods
and daca used! That ia * b(« applicable, reproducible, and dafensible L» the
uncertainty analysis ©I the rilk aKiaacii? This chapter presents na overview
if uncertainty astalysia, discusses sources of uncertainty in eaalyclag risks
free bnMr^®«» wiu, ai»l deecrlbe* slsa f®pi~»»ci*s ci •§§###•«tn§ and compar-
M| tiMaruinties. The r«*«ttCaeiM focuses en the ia health
risks* The reso«wces mlUM« ler thin itudy did aet pemit laveactiatiM of
necbedc fer tMm enelyels ef UBecrcaiaeina in estisvtes ml enviresMntal
effects, cescs sod laatfic* er ef riak-cesc-baaMtf it analysis of mtigecion
ack1m§. inch coaLd to ?«pair«wl ie aasy actual decisions. 0®e«ir>-
eaintiee In as dentate of coat* ara seen itude hi|btr then laitUi eetinncea
Im caaea of oneaoei pnbllc or refalatery coctcern*
X-l
-------
only ifidlcata* a quality or ic«u of b*La| ancttl«iaf but etn also iaiply *
lUCwm coacarniaf ch* quality a* cuoaelcy of tup port lag objective evidence*
Ic if if) tbia late iru eke ueage di{(«ri wit (reecLy I* the
ttatiaticol, biol•§!<*!• soeUl, pkyaicel. md engineering sciences; m buei-
mii «cMomicsi m f»Licy» Ugieiative, and reguletory action*; *n< In
judicial proceedings* Uugi i*ay differ among thoae deacribing t*f«ria®nc«I of
hiatoricaX factual dtt*« those describing «ose likely hiturt ivwti, and that*
describing the future as » i^ctna of kcmtIm or probability diatribes!on*
of pottlble future oocconoa. S«c Authors 4licrib« «neertji*ty I® quite
cifM, aeoy describe it in Kftllari&tl («ru, and im mm uncer-
tainty *ynonoa»u*Ly tor "rlik»"* 1» cbl* i;«4jr the following defiQitiofti in
used.
• "Oacertoiaty" !• a ic«cm«r( of tbe di|rcc te ufcicli
Md ftaftace (1981) «i®# eniparti of wevruiecy
ajuLyaoi for uao la proba^iliatic riak as»eaeaieai« aod Ldoncified five
¦atlioflMciedl aethodat aAalytiaol ttdteiiqueai Moat* Carlo aiduiatloBa;
rAsfeM® mtrf»m differoatioi aoftaitieity ui ovoU-
ocioa of coofiAoaec iatorwoia, fs€k tm§ It* o^eaatofto mi di*-
mdMimmgm9 to ch«o« «ockoro. V«a*Ly m4 (I9B4) ataco ch«e
in oatoloor prokobiliatic riak laiiiiiiflii one aoed* to difforootiate betvoeo
oBoceruUtlaa of two autior typaai (I) ptoy«lc*l etrlAbillcyI and (2) lack of
• Tkm "awamlaty foe tor" hot aIm om4 ijwnoo—aly fmr
(•etor*
-------
knowledge. In cootra&t, Fieriag ec i»l.» 1984, distinguished two ma)cr types
of uncertainty at "analogy" aftd wc«cjnfe#" In addition, there is uncertainty
over what baa, iti the paaC, been crue re|irdin| cause-effect relationships vs.
what chafigioft relations will oecw in Che future. Forecasting acthods and
• icidtatioe raftthodology e«m differ.
Several approaches to the analysis of uncertainty can be identi-
Hid. These approaches mve used co varying degrees let different fields of
analysis, but duty ara hoc Encuaiiy ftsaluttive «nd Mre than one approach e«n
be used in a givto analysis. Tba type of approach chat is mtd and the level
«C which it is applied will ba limited by the coatpleca«e«e in ability co
ideacify all of the causal factors important to outcoset and by cha quantity
and quality of date. Ic will also depend on the time and resources available
for compiling and evaluating eke information. A partial lift of approaches is
shown in Table t-l in generally increasing order of data and ti« require-
ments; she list is intended to be illustrative, net necessarily complete
(Lawless, 1984)# these approaches are discussed briefly below, followed by
several considerations chat eey affect the choice of approach in a specific
application.
TABLE 1-1
SOME APPROACHES TO l/NCERTAINTY ANALYSIS
Approach
Qualitative discussion
Expert judgment analysis
Sensitivity (parametric) analysis
Statistical analysis
(¦any It in da)
Propagation or cascading of
errors analysis
Information tcoolrepeats
Useful when causal understanding, data,
time to compile inloinnetian, perform
detailed analyses* or aeak expert
judgment is limited.
Requires eppropriata cross section of
technical experts! re«juiree techniques
for eliciting and combining chair
opinions based on subjective evalua-
tion at bath direct and indirect data
on cau*e-»e££eec: relationships.
Hachaoatiaal (or esperiaantal) model
of relationships; reasonable estimates
of likaly variance in key data/
information sources*
Extensive e*peria*«tel or hiatorioal
data or resulcsI (orml statistical
sMCtwda.
Mathematical formulation of pro bleat
analysis! aeaaure or estimate of un-
certainty of each fecj component.
-------
• • Qualitative I qualitative ipfrmci cm be
rlpful in a«alu*tia| uncertainty, particularly U iiu of «4a^eata quality
•m iiaitad m II tIm i# uaavailabia to eoapila information* parfona rigorous
talyaaa, or tank aspart jirfpsanc. Qnalitativa >udf«Mts a*y aloo t» baocrf oo
xtsidarabl* ~¦©Mac of data that *r« indicative »f cauoa-affact relet iaaushipa
tc %rnmzltml^mlf and scientifically demonstrable I ft « ri«oro*a hon.
A iftt«MCic ida&tificacion an
-------
events. Th* analyst mute be ivart of such poasibilitiaa an4» ii necaatary,
not* Chan appropriately in tha iiiuiBrnt raperc. A seoaitivity analysis aver
a wider range of coaclusioaa Bay be htlpful.
Monte Carlo sioulatioo approaches to (Mini expert-estimated
roitj in environmental icsii has been 4«»crib«d (Goddard, 1981, 1913 5 *1*
(1983)-. Publications wre not found that iLLusCreted the application of che
technique to environmental health risk#, wbere tile urncerteiacy of variables
tiny b« order* of magnitude greater ch*a in eo«c Or engineering reliability
applications. The general ?tlM mad limitation* of using. experc judgtMint ia
risk have bean diacuased by Hsfwad «c *1. C1984) and Morgan it ml.
C1984).
o. i
-------
oi Konte Carlo simulation. The relationships of the ptr®«ec«ees are »peelfi*4,
C9|(th
-------
• rron ot cascading of errors mechod.:* Considerable litencure Is available
on error analyiiit,
Two cattt nay exists the variables my bs richer independent
or dependent. Analyses with substantial Interdependence among variables are
more difficult. Dependence uy operate co either increase or reduce the error
ia the final result. The variance (v«r) of the suot of two variables, x and y,
is given by Che equation
var (x, ~ y) * var(x) ~ var(y) ~ 2 covar(x,y)
where the cov«c|«oce of x and j (covar(x,y)) may be positive or negative. If
* and f are independent, the cover Lance it tero. The covariance may be aero
without x and j bvinj necessarily independent.
If one ausm J variables, due variance of the sua b^cowsi
var<**y*i) » var(jr) ~ var(y) ~ vir(z) * 2 cov(
-------
TABLE X-2
CONSIDERATIONS AfFECTIMC CHOICE Of APPROACH
TO UNCERTAINTY ANALYSIS
InfomnC i oa ted Dica Natdftd for Productive Application, e.f. , quality,
qu^nCity, difficulty of C9li«CCion» L«v«L of detail required
Host Applicable ftange of tiik* e.(. . lou probability tvenci of grtater
fIJllificanca than de minimis risk
Most Applicable A ansa of CJucurcaiocy
Levclt of DiiniragJiion and A&gregacion Desired or Permitted
Ability CO Handlt Cornpltx Interactions of fitsk, Corapanants
Predictive V#|ue inch Incraaaing Tine Horizons
Modes Av»i UbU for ExpttSiing Jesuits CO Experts and Laypersons
Ctedibility/Defeasibility
Applicability CO Riak Hamienenc Cancimj
Ability Co Addresa Effaces of Corrective Actions tnd Reversibility
Ability CO Appraiji Vllu« of AddiCloruil InformtCion
Ability to Identify R««earch Needa
Difficulty of Bounding the Analysis
Ability to Complement Ochar Machods
LLkaiihood Utility o£ Approach Will B« Improved
i-a
-------
Assumptions ruade in structuring the problem so ic cac be ana-
lysed
• Limitations in the accessibility, quality, And quantity of the
available dtca, and in the comparability of da c* for the
alternatives bmiag assessed
• Applicability of available methods and models for performing
Chi risk iittitMcc
Applicability of methods for accounting for temporal dispar-
ities. quantification mi, discounting of noomonetiied value®,
«n4 value lyicu differences
• faliability of interpolation and extrapolation ttchniquec ut«4
» Allowance for corrective actions
The sources of uncertainty kn aasassiog the risfc» of hasardcua waste
t»ngi» treatment, acid disposal ara nuuseroua. Uncertainty arises from most
of tha activities of the laaiamai. These aources may be categorised inco
seven broad functional areas (factors) chat reflect major activities in a
comprehensive CMpAfiii^i risk assessment. Each factor it cceiposed o£ several
subfsctors or variables. Some 9! these arm discussed below# A checklist wes
provided in Chapter IV.
1. B* suits of the rial*,
assessment will inevitably be *f factoid by the way ch* particular scenario is
structured Cot analysis* Host problm in decision fnaking can be reduced to a
format of alternative scenarios* Tha uncertainty associated with applying the
results rill be * refleetion of the datail built into tha scenario, which way
range Iron quite staple to quite detailed. Tha fewer and wore similar tha
number af »ce®*rlos9 probably cba smaller tha relative uncertainty across
scenarios. UkeviM, cte smaller the mga of affect* considered, ctie swiilef
«b4 iAte homogeneous the population (or environmental values) at riak, and the
more elearly defined the |ci#ae« m4 cedhnology, ceteris paribus, the less the
uncertainty* finally, the mm atariy • scenario addresses real world policy
questions* the aK>re velum tha uncertainty Analysis should have ia swking deci-
sions* While tha various conditions chat are de£inmd in developing a scenario
ara of can subsequently taken aa fixed values and excluded from tha enalyala,
one ahoald not forget than io tha final assessment beesuae cfcmy may have
greatly shaped tha outcome*
2. Pollutant release* This tar* rafara to the source strength of
tbm pollutants Umc is, ic includes Information on tha probability end magni-
tude of release of given pollutants to the environment over tine. "The term
covers several factors Chat are iflweei of uncusrtainey. Tbese include:
general characteristics and quantities of waste being treated and disposed!
knowledge of the chosical coastituaqts of the waste; koowledte of the prop-
«rties of tbe constituentsf farticol«rly thoaa affeeting release to tha
environment| geofraphiesil, geological# neceorolcgical settings flf tha
treatswnt and dispOMl site) parrfonaanca of the treacaant disposal ctchnalogy;
1-f
-------
and identity, quantity, And rat# of hazardous matirial released Co environ-
mental media o£f*site. These rtluia factor# mmj Include several subfactors,
such tsi Cine to failure of a liner, volume of leachaCe, and quantity of con-
seitwent leached for a landfill; de»tructiim and removal efficiency for 4
particular chemical Im an incinerator* tnd frequency «nd Magnitude of «cci-
dwial spills in transporting hasardou* wastes* A»iiy»i$ of only che air
emissions fro* a treatment* storage, and disposal facility still require# con-
sideration of man/ uncertaiatiea {Wallace ec ml.® 1987).
], Cnviroaaseatal c.rio»parc »ad fata: Th# quanciey and rati of
hazardous constituents moving through the *ir» eater, or soil frtaai the point
of releaae to himma r«c«^cor« depend oa 4»v»ral factors that nay be soarces of
unaarcainty. These iaolude: applicability and cofltpletaness of environmental
monitoring dataj applicability of nodal a that simulate movement through che
media; d«e«ipo»itiqn» degradation, or stabilization processes that render che
constituent harmless or iammbilel and tendency of Che constituent co bio-
acQusulaCe in the flora and fauna or to biomagnify la food ehaifi*» The more
complete cha data base aod the more verified tie modelsr the mailer che
unocrtainty.
4, Exposure prediction* this involves estli&atioii of the ejAge in
apace and cum that the pollutant Interacts eith environmental receptors (4a
determined by analysis of pollutant release and snvirawventel transport) and
the auMber of receptors at choae points (as detersuned by data 00 population
distribution and lifestyle*)~ A soure« of uncertainty io maoy exposure
atteasauots it the range of exposures chat amy exist for even a narrowly
defined population. Expoauraa are 00c oaoesaarily cha «
-------
level. The effect muc Chen be integrated over all fwioas exposed mt that
level and ic all other levels- Htnc«« uncertainties exist aver Che numbers of
p«nons in each exposure group end the levels of their exposures. tn •44L-*
ciont <11 especially sensitive subpepulations , then the total standard error is given by an equation for swr-
ming Che standard deviations as follovst
aX H Of ••«!~... e*
This relations hip does not depend oo the fona mi the probability distributions
for the events or variables* 90 long as they have finite variances. If the
composite error is approximately nor—1 iy distributed, then multiplying o by a
factor allows one Co express the uncertainty as a confidence interval i.e»»
±0 has Ml confidence limic»j ±2
-------
The standard error is not the only way o£ expressing »o aspected or
error, A key concept is the distinction between in «xp«et«dl or svur-
igt error in a set of statistical data ud an actual error (a prediccinj *
future event» for example, if « risk analyst est Laws tea that « given eccidenc
would l«md co a spill -of hazardous waste of size i, bat tubeequently ma, actual
accidmnt gave * spill of tlx# t than the «ctu«l absolute #rror e of che esti-
mate was
e ¦ x - s
At che time of nuking the estimate, however, both ¦ and • vere unknot. The
analyst mi give calculate rsany reasonable values of s co estimate an average
error, i.e. , express the iv«ra|c error
Tha ftcaadard deviation it an averaging process.
Aaocbar consideration. la aggregating uncertainties involves that
"worst case" ic«ario# which mut fro® concern chat events or variables vill
combine ia s vmy that produce* a near aaxiau* r4(bir than average (rrar, i.e. ,
the passible error* are strongly biased ia oae direction. The total error is
then the sua of the absolute vaiu«i of che component errors. Although such
coabmat logs arc highly unllfcaly, thiy ar» possible. Th« rftguUcory decision
maker is frequently faced with Che difficult choice of whether or not to con-
sider such combination* and* If so, bow. In preccice, the issue is resolved
on a ceee-by-cese bests. this tvpie li discussed further in che Mie seccioo.
Geometrical aggreteclon: If ail of several events in a series muse
occur ligtulcaneoasly or in sequence for total ad;verve effect# to occur, thea
che risk factors are oroltiplicacive according to che principles used in pro pa-*
gstioa or cascading of errors analysis <«s 4i?ou*sed in Section A.2.d)l i.e.,
che overall risk is a produce of the separate risk, factors.
Ia assessing the risk of hazardous waste aunagement technologies*
ch* overall risk CD can be ¦athaaatiiialiy structured, as a product of a series
of risk~related factors (P), each, representing owe of che several najor fac-
tors discussed in Section B. The ¦egai&kte of the risk* can then be calcu-
lated fro« the estisated values of' the factors, assuming chat each is positive
and ell are probabilistically independent*** the relationship oay be seated
as a staple product or as a logarichoic sue;
• This ia an absolute rather than relative risk between alternatives.
•* If che factor* are not Independent, Chen che multiplication of probebil"
itiea is condi. tioeaX, i.e., in general, P^) ¦ P(Pj)P(Pf/Pi>j if Pj
end rj are independent, Chen ¦ P(r|)f(P2). Ia the absence of
independence* the factors are viewed as conditional probabilities*
1-12
-------
i • f% » rt * ... rn
or lot ® ~ log Fj • ... log F#
Tho «ev«r«l Uctari oa? is turn depacd or on* or cars other tub-
factor* tbuic reflect cha vifiiblia (v). If ail tha variable* within 4 factor
in aaaantial* ctu* f«cC9r li a produce of tha abaolue* valuea of cha verl-
ablas, each variable being poflclva and independent. In ganaral, howaver, All
of cha variables will mt be dfstncl&l* for exasple* a factor, couLd be of
tht for® i»tv, * few|. Tho contribution of all tht ««ri«bi» to the risk «M*t
zbm ba calculated by * coafciaatioa of additive «a4 auUipllcacive operation*.
HM iMljrfii (becclvrt bacoaa* problaa* *pacifie 4c tfcli point.
If the uncertainty i® «*ch fj terai ebova if aspr**«ed » 4 url"
ihm the mctfUiatf La log 1 can be *n^*®§g#4 At cfet *u» of cine m»e«r-
ceiotia* of the component fictsn, If chu F, factor* sm iMapudtat, thee
tho 0404Cioa 1*2
mr Clog 1) « v«r (log fj) ~ var (log Ft) ~ ... vtr (log F&)
or var Clog ft) - Uf • Uf ~ ... Uj
where Uf • «r (log f^}
If the f| «r« not iMapaadoac, tbea
vartlog 1) - 1 t «ev
-------
( ii 4 mi* fro* a probability di»tribuci«n talli t0rt |