February, 1977
             Joyce Salg
     Department of Epidemiology-
    University of North Carolina
      Purchase Order 5-03-4S23J
           Project Officer
          Leiand J. McCabe
       Water Quality Division
 Health Effects Research Laboratory
 Office of Research and Development
U. S. Environmental'Protection Agency
       Cincinnati, Ohio  45263'


JOYCE SALG.  Cancer Mortality Rates and DrinkisjJ 5fat;er Cwolity

             in the Ohio River Valley Basin

A series of unweighted and weighted regression analyses were used

to dateraina if an ecological association exists between drinking

water in 346 counties lying within the Ohio River Valley Basin and

Cancer Mortality Rates of individual and grouped organ sitas.

Using twenty year, (1950-1969) age-adjusted county cancer mortality

rates, I960 census denographic data and ntinicipal public watar

data (1963), tvo water variables were successively studied:  percent

surface water usage  (percentage cf a county's population whose

drinking water source was from a aajor or siinor river) and percent

prechlorination  (percentage of a county's population served whose

Orinking water underwent prechlorination).  Ecological association

between colorectal and bladder cancer mortality rates and the water

variables  suggests further research employing these specific sites

will be most productive for hypothesis testing.

r ’ ‘v’i
Cbjectives . 6
Ex er enta1 v! e ce ‘ C ic i Lca1
Carciic€enic ty. . . . . . . . . . . . . . . . . 20
Ed.e o1cEjca1 Evld.ence fo Cr a 1c Chen’ical
CaI1C1rc8eri C1t7 . . . . . . . . . . . . 22
ter a1s. .. 3
... . . 14.Q
P tc t S ace Water Jsa... •
Pe ei ± ? c cr± at oi. .. 69
i . ct ssIc .
S S 107

1 Trihiogenated—methate Content of Various Munici a1
Water Supplies .
2 Trihalogenatad—nethana Content of Water from Water—
Treatment Plant
3 Organoch3.orina Compounds in. Water from Sewage Treatment
4 Regression CoeffIcients for Variable Representing ProportIon
of Drinking Water from Mississippi 16
5 Results: Presence (*) or absence (NS) of Statistical Signi-
ficance at 5% Level for the Variable Representing Proportion
of Drinking Water from Mississippi River 26
6 Su .ry of Significant Results Bet ;een. Percent Surface Water
Usage and SLte, Race, Sex—specific Cancer Mortality Rates,
1950—1969, in 346 Study Counties
7 Estimated Percent Surface Racer Usage Regression Coefficients
and socIated p—values from Un.weigh ed and Weighted Regres-
sion Analyses of White Males, White Females nd Site—specific
Cancer Mortality Rates Versus Selected Dernographic Risk
Factors for Malignant Neoplasn Mortality, 1950—1969, in 346
8 Estimated Percent Surface Water Usage Ragrassion Ccefficer.ts
and Associated p—values from Unweighted and Weighted Regres-
sion Analyses of Non—white Males and Non—white Females and
Site—specific Cancer Mortality Rates Versus Selected Demo-
graphic Risk Factors for Malignant Nsoplasm :.iortalitv. 1930—
1969, in 346 Study Counties
9 Estimated Percent Surface Water Usage Regression Coefficients
and Associated p—values from Unweighted and Weighted ? nalysas
of the Gastrointestinal and Urinary Tract Systems Cancer Mor-
tality Rates Versus Selected Demcgraphic Risk Factors for
Malignant Neoplasm Mortality, 1930—1969, n 346 Counties 52
10 Estimated Percent Surface Water Usage Regression Coefficients
and Associated n—values from Unweighted and Rei2hted Analyses
of the Gastrointestinal and Urinary Tract Systems Cancer Mor—
tality Rates Versus Salectad Demogra hic Risk Factors for
Malignant Nec?1as Mcrtali , 19 0—1969, in f 6 Ccunt .es

11 sti ated Percent Surface Water Usage Regression Coefficants
and Associated p—values fron Weighted na1yses (reduced nodal)
of Selected Sex-race, Site—specific Cancer Mortality Rates
Versus Selected Dertogreohic Risk Pactors for Malignant sec—
plasm Mortality, 1950—1969, in 346 Counties 57
12 Unweightad and Weighted Regression analyses of Percent Sur-
face Water Usage and Co binad Sites: Large Intestine and
Ractun for All Race—sex Groups 59
13 Estimated Percent Surface Water Coefficients and p—values
for Cancer Site by Race—sex for Weighted Regression Analyses
(reduced Model) Including Lung Cancer Mortality as a Pra—
dictor Variable
14 Zatinated Percent Surface Water Usage Coefficients with p—
values fro WeIghted Regression Analysis Stratified by Popu—
laticn for 346 Study Counties 63
15 Zsti atad Percent Surface Water Coefficients with —va1ues
from Wei htad Regression Analyses Stratified by Population for
Counties with 0 Percent or More nown Water Source 66
16 Su arv of Significant Results Zetween Percent Prechloronation
and Site—, Race—, and Sex—s ecific Cancer Mortality Rates.
1950—1969, in 346 Study Counties 71
17 Estinated Percent ?ra lorinacion Regression Cceffici3nts and
Associated p—values fron Un eightad and Weighted Ragrassicn
Analyses of White Males and White Fetialas and Site—s ecific
Cancer MortalIty Rates Versus Selected Darograohic Risk Fac-
tors for Ma1i ant Neo 1asn MortalIty, 1950—1969, in 346
Counties 72
18 Estinatad Percent Prechlorination Ragression Ccefficianrs and
Associated a—values fron t.Tnweighted and Weighted Regression
Analyses for Non—white Males and on— rhi:e Pansies and Sits—
specific Cancer Mortality Rates Versus Selected Denograhic
P sk Factors for Malignant Neoplasn Mortality, 1950—1969, in
346 Counties
19 Zstj ated Percent ?rachlorina:ion Regression Coefficients and
Associated —va1ues fron Weighted Analyses (reduced nodal) of
Selected Sex—race, Sita—s ecific Cancer Mortality Razes 7arsk.s
Selected Denogra hic Risk Factors for Malignant Nec?lasn Mar-
taiitv, 1950—1969, in 346 Counties
20 Unei htad and Wei hted Regression Analyses of Percent Pre—
chlorination and Can ined Sites: Large Intestine and Ract
for All Race—sex G:cu:s

21 Estimated Percent Prachiorination Coefficients and p—values
for Cancer Site by Race—sex for Weighted Regression Analyses
(reduced modal) Including Lung Cancer Mortality as a Predictor
22 Estimated Percent ?rechlcrination Coaffic ents with p—values
from Weighted Regression Analyses Stratified by Population
for 346 Study Counties
23 Estimated Percent Prechiorination Coefficients with p—values
from Weighted Regress cn Analyses Stratified by Population
for Counties with 50 Percent or More Known. Water Source 85
24 Significant RegressIon CoefficIents nd p—values of Unweightad
Regressi3n Analysis for Percent Surface Water Usage and Se-
lected Soda—economic Variables Organ Site—specific for T hite
Males, Malignant Taop1asm Mortality, 1950—1969, in 346 Study
Counties 1C8
25 SIgnificant Regression Coefficients and p—values of Unweighted
Ragressiou Analysis for Percent Surface Water Usage and Se-
lected Sccio—economic Variables Organ Site—specific for tfliite
Females, MalIgnant Neoplas Mortality, 1950—1969, in 346 Study
Counties 109
26 Significant Regression CoeffIcients and p—values of Unt-zeighted
Regression Analysis for Percent Surface Water Usage and Se—
lected Socia—econonic Variables Organ Sice—specific for Non—
white Males,. Malignant Neoplasm Mortality, 1950—1969, in 346
Study Counties 110
27 Significant Regression Coefficients and p—va1ue of Unwaightad
Regression Analysis for Percent Surface Water Usage and Se—
lected Socio—econonic Variables Organ Site—specific for Non—
white Females, Malignant Neoplasm Mortality, 1950—1969, in 346
Study Counties
28 Significant Regression Coefficients and p—values of Weighted
Regression Analysis for Percent Surface Water Usage and Se-
lected Socio—econctic Variables, Organ Site—specific for White
Males, Malignant Neoplasm Mortality, 1950—1969, in 346 Study
Counties 112
29 Significant Regression CoeffIcients and p—values of WeIghted
Regression Analysis for Percent Surface Water Usage and Se-
lected Socic—economic Variables, Organ Sita—soecifi: for t hita
Fenale.s, Malignant Neoolasn Mortality, 1950—1969, in 346 Study

30 Significant ?.egression Coefficients and p—values of Weighted
Regression Analysis for Percar.t Surface Water Usage and
Selected Sccio—econonic Variables, Organ Site—specific for
Non—white Males, Malignant Naoolasci Mortality, 1930—1969, ±n
346 Study Counties 11
31 Significant Regression Coefficients and p—values of Weighted
Regression Analysis or Percent Surface Water Usage and
galeeted Socic—econo ic Variables, Organ Site—specific for
Non—white Fer.ales, Malignant Neopla.sn Mortality, 1950-1969,
in 346 Study Counties 115
32 Significant Regression Coefficients and p—values of Weighted
Regression Analyses (reduced model) for Percent urfaca Water
Usage and Selected Socio—econonic Variables, Organ Site—
Soecific for Sa: -race Groups, Malignant Naoolasn Mortality,
1950—1969, In 346 Study Counties 110
33 Significant Percent Surface Water Usage Rearession Coef-
ficients and p—values fron Weighted Regression nalvses (re-
duced rnodal), Sex—race—sita—s?ecific Cancer Mortality Rates
Versu: Selected Socio—econonic Variables Stratified by Po u—
lation fcr Malignant Naoniasn Mortality, 1950—1969, in 346
Study Counties 117
34 Significant Regression Coefflolenta -id p—values of Unweighted
Regression Analysis for Percent ?rechlorinatiort and Selected
Soclo—econoric Variables, Organ Size—specific for White Males,
Malignant eop1aan Moriality, 1950—1969, in 346 Study
Counties 12C
35 Significant Coefficients and —va1uas of Unweightad Regression
Analysis for Percent Prechiorination and Selected Scdic—
econonic Variables, Organ Sita—s ecific for White Penales,
Malignant Neoclean Mortality, 1950—1969, in 346 Study
36 SignifIcant Ragresslcn Cceff cients and p—values of Unwaighted
Regression Ana1y is for Percent Prechlorination and S l cted
Socio—econo ic Variables, Organ Site—specific for Non—white
Males, Malignant Nao lasn Mortality, 1950—1969, in 346 Study
37 Significant Coefficients and p—values of U.-eightad grassi:n
Analysis for Percent Prechiorination and elacted Socic—
econc c Variobles, Organ Si —scecifi: for ycn—white Fenales,
Malianan: Neco1a n Mortality, 1930—15, in 346 S udv

Chemical carcinogens in th environment have been
Lndict d as the pni ary ca ise of the ajor!ty of human
cance:s Hcwever, until recently, little attention ‘az
directed to the possibility that carcinoqens in d:in cinq
vater ay be causally related to human cancers
Potentially harmful chemical agents which may be
present in public water supplies include heavy metals,
pesticides, and a variety of organic and inorganic
che icais——aqants ‘nich nay be potential carcinogens.
Urban and dc estic sewage and dustrial ias es
discharging into water sources, atmospheric carcinogenic
contaminants present in rain water, pesticides nd
fertilizer residues conta.ned in agricultural runoff, and
inorganic substances incorporated into the soil and
leached into the ground-—all may pollute surface and
ground raters. In additicn, the icrcbial and aquatic
populaticn of both virgin and polluted streams contributes
metabolic products (Gregcrmpoulaus, 19 L ) - lso, prducts
of decomposition nay present new problems of contam aticn
and chenical and biological reactions may form mc:e
hazardous compounds.

Consequently, there is a potential for ctan ation
within the foreseeable future because of both the
increased urbanization and industrialization and the
conseqa ent qrsat increase in dszand placed on lakes,
rivers and underground sservoirs ( ueper, 1960)
Several studies have shown cheticals having
carcinogenic ctantial present in river vater tostel et
al ,.,1965; sites and ienan/972) and in finished unicipa.l
drinking water epe: and. payne, 1963; ndeleaan and
Suess, 1970: un c and Stanley,1975)_ Pollo7ing a 1972
nvircnzeutal rctacticn geacy study C! industrial
pollutants of the Lover ississippi iiver in Louisiana,
ti e .2.A. water Su?ply Research la o:atory, Cincinnati 4
Ohio identified 66 different chezicals in the drinking
qatar of av Orleans. Shortly thereaftsr, the consuoption
of nunicipal dninicinq iater in Louisiana counties as
found to be pcsit ve17 associated iith over—all cancer
nortalit7 in ‘ihite zales (Page and E rris, 197 )
Although the adequacy of tho statizticzl nethcds enployed
and the validity o the hunan health inte prataticus drain
frc this study ‘ie:a challenged Ta:cne and Gart, 1575),
continued scientific investicaticus (3dlla:, 197 ; !ock 9
157n) established that the chlorination p:ccess as
ccnnonj..y practiced can result in the fcration of a
of chlorinated o cc.np.!——sc:e of .ch
are : c cc s s:ac-ted c :cge s

WL h the continued. dccu entation of the chenical
ccnta iuaticn of drinkir g water in the U_S. md the
possible adverse hunan health effects of chernically
polluted drinking iater, the U. 2. Congress responded..
The Safe Drinking ater ict (PLS3—523) was signed into 1a
by the President in December 1975. Section 1Z 2 of the
Ict charged that a conprehensive study of public water
supplies and. drinking water sources be und.erta en by the
nvironoental Protection Igency to deternine the nature
and. extent of ccnta ination by carcinogenic substances.
Continued. studies by the nviron ental otection
Agency have since identified. organIc coopounds in the
drinking watar snp lies in the U.S .. (1976). Also, in a
survey of 80 selected cities, the National Organics
3econnaissancs Survey eazured. -the concentrations of six
halogenated cc pounds in raw and unfinished. iater.. :n
these studies, sc e ccnpotLnds, particularly chioroforn,
were ubiquitous in unici;al drinking suppl±es.. Under
l .boatory experinenta3. conditions these coupounds have
denonstrated carcinogenic activity in ani a1s..
owever, several issues renaln unresolved.: whether
these sane agents are carcI oqens for hunans; what doses
are within ac e tabla units; the shapes d. para eterz c
the dcse—res;crse curves; concentration levels; the
effects of rolo: ed e:pcszre to 7er7 lc arounts; and
wh the iatar supt l’ , as one of nany pc sihl ens

.s o: a y porta c 9..
this stud7 was ui de ta1ce to i.vestigat
ibethe watar supply, as one of rnany possibla exposure
sources, is ceoqraphically related to cancer oztality.
Two çeneral objectives a d thei: corespondiq specific
objectives are conz dered..

General Ob ctLves:
To determine Lf
exists betveen
source of public
To deterxine if
exists between
chlcxination of
sti;plie s.
an ecolcgicai. assoc.ia.tion
cancer mortality rates and
drinkinq water..
an ecological association
cancen o:talit rates and
public d.rin king vat er
Specif±c Cbjectives:
To deterxine if an ec loqica]. associaticn
exists betvaen race—sex-site—specific age—
adlusted cancer orta.lity rates, 1950—1969,
and erce:t of a county’s pçpulation vhose
public dninkinq water is frou nalor and
minor rivers in 2 6 connties of seven
contiguous states: IllInois, nd.iana,
entuc y, Cbio, ?e nsyivania, ‘en:essee and
ies-t 7i: inia.
To dets:nine if an co1cqica1 asscciation
e:ists bet-ieen : c —se te—z ecific age—

adjusted cancer aota lity rates, 1950e1 969,
and percent of a county’s population whose
public drinking nt*r undergoes
pnchlorination teataent in 3fl counties of
seven contiguous states: fllinois, Indiana,
Kentucky . Ohio, Pennsylvania, Tennessee and
Vest virqiuia.

32T237 07 Tfl Lzfln :u32

The qnestion of the potential for conta inaticn of
drinking iater scu ces with carcinogens was g ’ en i etu
and direction consequent to the identification of €5
different chenical conpounds in the dninkinq wate: of 1e z
Cricans y the Environnental Protection genc Water
Stip ly ?esea:ch Laboratory, Cincinnati, Ohio in 197LL I
nn ber of u;dat s no place the figure at 313 or ’anic
co pcunds identified in the drin2 inq water in the C. .
One class of organics, pol uciear aro ati:
h7d:ccarbcns (?A ),imclndes many cheoicals that are !crc .in
to be ca:cincqenic to anirnals and nay be carcinogenic to
nan A:d lnan and Suess 197O) :evie ied carcinogenic and
ncn-carcino enic P in the water en’,ircnnent with
enphasis on 3,L —benzypyr2n (3P). 3, -3enzyprene, a
hiqh2.’r potent carcinogen, is ubiguitous — tbough
z viou ly thought to be forned only by in p1ete
c ust±cn, evide cs points to the e oq nous foroation cf
n niants . Thile thei: sciubility in p :e
er i. T 7 lo , they O 7 be lubnl zod by such
ta nas ns fe :;ente, or na ct : ise occ ir in

solution associated with or absorbed into a variety of
colloidal naterials or biota and thereby be transported
through the water environ ,ne.nt..
PIEs have been fouxd in Lndustrial and municipal
waste effluents and occur in the sedi2ent and biota of
and surface waters. ith. the
or absorption by actIvated carbon,
processes do not remove P? Es;
been found in d.o eztic water
(fresh) groundwater contained
icrograrn per liter of P1 .
water has beer. as high as 0006
Waste effluents have ranged fron
per liter of 3,L — enz pyrene
soils, ground waters
excepticn of filtration
other water treat ent
consequently PIEs have
supplies. ncontaninated
fron 0.0O i to 0010
Carcinogenic PIE in ground
aicrcqrans ‘per liter.
0.5 to 7.5 nicrcqra
(Indelua ,1970).
wedquocd and Cooper {!ueper: 1960) , who studied
sludge from sewage for pclynucrear aronatio hydrocarbons
in !ngland, de:onst ted the presence anong various PIEs
of two carcinogens, 2, —Ben:ypy:ane and 1,2 -
Benzanthracene. They suspected the carcinogenic P H was
introduced into the s vaqe through effluents discharged
frog gas works, washings fron acadan roads and
atncspheri soot ashed fron the air by rain.
Euee: (19S 1,i960,1963) did e ;eri:ental studies
ac 7atad carbon absorbate.s of :n and finished
pollutsd with dtrial—cnica2. effl:ents. A

fraction of the a scrbates were ad inistar d to mica by
ea s of cutaneous appicaticn, bcuta.eo s injection,
and oral ad inis-tra .on_ Results shoved that carci oqenic
chemicals ars released. into bodies of water zeniinq as
sources of d. nkinq wz.ter for the gener 1 population. The
data incr i inates not only absorbates of raw water, but
also those of f nizhe water, and tbns ind c te that
carcinogenic che icals penetrate the filters used fcr
akin rai water actsniolcq.calIy safe fcr hunan
ccnsunpticn This study raised the question of whether
the: daily consurnption of d kinq water containing i nte
a o ints of chenical car incqenic pollutants o e any
years iqht play a direct pninar7 or contributary role in.
the production of cancers of internal crq s asong the
general pcpula-ticn
Rook (197L ) , f oo an. an. a17Z3i3 of the chlorine,
inferred that the or an.ohaljdes found in ater after
chlorination were not introduced by chlorine, but n st
have originated in the water u on chlorination.; thus
water—borne hu ic :aterials in the iater supply ‘iere
chetical pr curscrz of several tnihalogenated oetban.e
s;ecies.. urtber ivestiqaticns ba e so cn that
haJ..ogenatad cc c n.ds a:e p inced durin chic mn.ati:ii in.
iatens cc ninc hun.ic or othe: o:oa:Lc tan.c s .
C ie7er, :; i.e iat n— bcn:e h:n.ic eniai ien
ve::L :a a2- cies, t .e c .Lc 1

nature of aquatic hu ic substances is not sufficiently
established to say with csrta nty that hese structures
are the o t pcrtant part c±pa tz the hlo et ane
production reactions. This point is pivota’ to an
increase understanding of these reactions (Chriztnan et
al..,1 76)
3ellar et al . (197Lfl postulated the widespread
production of chlorinated organic naterial due to the rnany
organic cc pounds cc nonly found in natural and waste
waters that ay react with free chlorine. Thus the
extensive use of chlorine in water and sewage-treatn.ent
processes, in household and co nercial laundering, in
paper-pulp bleaching, and in related processes was
i plica ted.
Subsequently, Eellar found organochioride co:pcunds
in tap ‘iater while raw river water, source of tue tap
water, contained none or rnuch low-er concentrations of the
crqauohalide. where the water source w low in total
onqanics, i.e., well water, resulting concentrations of
haloqenated products were low. where the water source ‘ias
high in total orqan.ics, i.e., surface waters, the
:9sult :g concentration of halogenated produ s were high
(see table 1) -

Table 1. Trihaloge ated— atha e Content of Various unicipaJ. Water
Raw Water
Concentration, ,igI1
Bro o—
dich loro—
chior —
• 37.3
from Bellar et al., 1974.
* Sam 1e age <4 hours
Sa 1e age unknc :; > 24 hours
A;prox. value + 20 ?etceflt
§ o data

Similar resultz were round in exaaination of
chlorinated ‘iate:s in Ohio (Befl.ar, Liclitenbeng and
Kroner, 197L ) - The predc inant crgazoha.lide species was
chloroform, which had a ccncentraticn range a high of
5 microgram per liter in Rook’s study(197D ) to 150
micrograms per liter in the work of Ee12.ar, et aL.,(1 74) .
Is shown in table 2, trihalomethane prcd icticn
originates at the water treatment plant.. !ach time
additional chlorine is added to m ntaim or increase the
frae chlorine concentration in water, a si ificant
increase in the chloroform level results Table 3 shows
the increase in concentration of organochlorine compounds
in water from a sewage treatment plant following

Table 2. Triha1ogenated— ethane Content of Water from Watar—Treat ent
Concentration, ‘ g/1
D bro o—
p m
Raw un-
water 1 0.0 0.9 * *
Raw water
treated with
chlorine and
a1 chlc— 6 22.1 6.3 0.7
rine. Con-
tact tire 80
settled 3 60.8
Water flow -.
from settled
area to 4 2.2 127 21.9 2.4
filters ÷
effluent ** 83.9 13.0 1.7
water 6 1.75 94.0 20.3 2.0
from el1ar et al., 1974.
* none detected. If present, the concentration is <0.1 pg/i.
÷ Carbon siur added a: : is cint.
*— T,._,,__.

I ’
Table 3. Organochiorine Compounds in Water from Sewage Treatment Plant
Concentration, ugh
I f1uant
Methylene chloride
l2. l
l,l,2—tr ich loroethylene
t Dichlorobenzene
£ Trichlorobenzene
from Bellar et al., 1974.
* 1i. confirmed by GC—MS (gas chromatography)

1 rhe TationaJ. 0:qanic 3accnnajssance Sur7e7 (Sy2ons et
aL,1975) of 80 selected cities initiated by the u. S.
Znviron ental 2rotecticn qency, eas ired the
concentrticnz of six halogenated cornpounds in ra and
finished water. The six compounds were chicroforn,
bro odichlcro ethane, dibronochloro ethane, bro ofor ,
d bro c hlcro ethane, carbon tetrachlo:ide and 1,2—
dichiorcetliane. The NQ S study re3ea.led the widespread
cc erce of trihalo ethanes in finished drin inq water
as a a rsct zesult of the cnloz nat cu process
(Sy ons,1975).. Carbon tetrachioride and
dichiorcethane were not for d in th chlorination
reac ti on.
Trihalorethane production is prinarily influenced by
the nature and concentration of the organic substrate, the
nature of the aqueous ChlOrine compound, pE, tenperature
and Cl. IC ratio.. Trihiooethanes result frcn a reaction
or series of reacticas of chlorine with preczrscr naheria?
through the classical organohalide reaction or sone
echanis that includes an o idatiie cleavage step. The
precursor itself is not a single conpound but probably a
cor p2.ex i t of hunic substances and sinple lc
9olecu.lar weiq t ccapcunds ccntaininq the acetyl nci 3t7
with differing reactiv±ti s at 7ar7 :q p 7alu s,
sc lnbilities and other physical and cEe ica2

The rate of trihalcmethane formation is pa dependent;
an increase (toward alkalinity) in pa results in increased
txihalo ethane produc-tion possibly due to an increase in
the-. husic acid. reaction rate (via the classic 3echani.sin).
Or possibly, co:pou.ads present in the settled water, i..e .,
acetone, which are relatively inactive at pH 6..5, become
significant contributors to the overall reaction rate at a
higher pE (?airless et aL.,7975) . Therefore, the effect
of pH becomes a critical factor Linder treatment systems
where chlorination is executed at high pa, especially.
vhare lime- softening or excess lime softening is
Conseg ently, where chlorination follo inq
clarification is carried out at pH. values near pH 7,
effective coagulation and sediaentat on may be sufficient
to reduce the precursor concentration to levels where
ultimate trihalomethane concentrations are below the yet
undefined adverse health effect levels. Where
chlorination is ca ed out at high pH, i.e., lime—
softening plants, treatment for precursor remove?, is more
complicated. Thus, the point of chlorination in the
tea .tnent process, a signifIcant fa .ctor in trIhalo ethana
production, represents an important variable to be
considered for change in attempts tr reduce ultir ate
trihalomethane-ccnc ntraticns in finished drinking iater
2P.A.,1 4 75). Cu=ent esti at€s are that a plant can

af ect its triha1c ethane output by plus or rninus 15
pe caut by ad ust g the point 0: cn1cr e ap 1 ca:.on.

!xperi me ntaJ. ! vi dance of Organic Chemicai. C ar cino gen icity
Chloroform’s pctentia.l for tumorigenic activity: was
studied under erperimental conditions using a graded
series cf necroti .ng and non-necrotizing -doses. of
chioraf arm on le and female mice to determine the
relation between do , presence or a.bsence of. liv r and
kidney necrosis, and the occurrence of hepatomas .foUowinq
repeated administration of chloroform (Zschenbrenfler and.
!iile:.15L4). xidney necrosis in male sice was extensive
in contrast to the absence of kidney necrosis in .fe a1e
mice.. !epatomas ware observed only in female mice..
vidantly, chloroform doses sufficient to produce liver
necrosis caused lethal kidney necrosis in the male rats;
thus only the sur7ivinq female rats produced hepatomas..
The greater sensitivity of male mice to the toxicity of
chloroform was also demonstrated by Shubik and Pitchie
(195.3). iiho found under specified laboratory conditions
gross tubular necrosis in males only..
In light of determining the signi ic nce of these
compounds in lower doses, t e paper by ttahcni and
Greenbet (1q70) was tnte:asti:g. n eiperinent vaz
perfcrmed in which the filta:-sterii ed effluent cf an

activated sludge plant was used as the sole source of
drinking water for 13 rats. en natched rats whose
drinking water was from the locally treated supply acted
as controls. Du ing the course of the exper ent t o nale
rats exposed to the effluent developed assi7e tuucrs,
which was unusual for the strain of animal used.
uierous alkylatinq cheaicals als.o possess tu crgenic
properties. :eonq, ac ariand and Peese fl967j induced
lung. adeno as i aice by chronic inhalation of
Bis(chlcrcuethyl) ather . Induction of lung adena as in
new—born nice by is(chi.cronatbyl)ether following asingle
subcutaneous inlaction of the co poand was de oustrated by
Jargus, et al (19€S)

!pide2iolcqical !vidence for Organic Che ica.1
Carcinogenic t y
2pideaioloqical evidence subs antiating actual cancer
hazards to the human population using water .ccntasinated
with raccgnized snspected, and potential. :
carcinogens is slight. Thorough epida io1.cgic:studies on
the inpact of organic nattar in water upon hunan health do
not exist. The few studies conducted relate this qua.1 ty
parameter to disease by use of surrogate measures.
In lo-udon, !tock (19 7) noted lower cancer
crtalities i.n co aunitias supplied with well water than
those receiving surface water In Eolland, Dtihl. (1S53),
as part of a larger study, lookad at the ralaticu between
number of cancer deaths and the water system.. ais data
s gqes-ted that over the whole country there is a
pronounced, though not very great difference between
municipalities with or v thout water systens: those
without aunicipal water zyste s possess the highest cancer
death rates. Differences in cancer death rates becorne
even acre pronounced if the largest nunicipa.l ties with
water sys-te ..s ars e2.ininated fro the statistical
analysaz.. The average cancer rate for all 3un1ci alities

w±th water systems decreases fto 585 to 543.
Duh.l (1953) divided the main sources of water supply
in the 1etherlandz into four çroups river water, heath
water, dune— zatsr, and we.U. water. The average cancer
death rate varied frog 606 (river water) to 594 (heath
water) to 585 (dune water) to 568 (well water) per 100,000
population. In !ugland, Davies and Wynne—Grtffith. (1954)
invest qated the distribution of salignant neoplasns of
various body sites with relation to the kinds :cf soils at
the hoses of persons who died of cancer i the County of
Anqlesey during the 10 years 1 3—1952 Th firzt st idy
(195Z -.) found that cancer of the stouach nd breast
.a3sociated with gro 2p soil series, water supply and social
class. The secend study (1954 ) fcund that the
association with soil, in the case of sto acb and breast
cancer, was independent of social group and water supply.
- Of particular interest s- whether any of the
haloqenated coapounds found in drin iag water are also
present in hunans ingestin; such water over long periods
of tine In a recent investiqation Dowty, Carlisle,
Lazeter, and Store; (1975) studied halogenated
h7d;ocarbcus in ew Orleans drind nq water and in blood
plasnas cf area residents. Thirteen haloqenatad
hydrocarbons - er identified in the driking water; fiie
were identified in tne piasna. tchl :oethylane and
carbon tetrachlcti a were found in both piasoa and

drin kinq water.. Considerable daily variation - —in the-
concentration of halcqenated hydrocarbons in the d.rinking
-water was noted. The a1or halogenated hydrocarbons
isolated fron ew Orleans drinkinq vatec -vera: 1-
:chloroproper.e, chiorofora, carbon tetrachioride,
d.j ch .loretha ne, tric hioro ethylene, dichioro propane,
dichloropropene 1 brodichloro ethane, dichioropropene,
tetrachicroethylene, d bro cchlorcaethane, . :1 a
ctichlcrcpropene. Prelirninary studies showed carbon
tstrach.lcride- and tetrachiozoethylene present in :the
pooled plasma of two test 4ronps with concentration of
carbon tatrachicride substances higher in the-plasma than
in drinking water.. In view of the lipophilic nature of
cblorofor , Dovty sugqests that a bioaccuau.].aticn
aechanisi ay be in operation, if drinking iater is the
only source of such 3aterials.
I study by paqe and EaQrjs (197Z ) involved an
investigation of county (parish) mortality rates for
Louisiana, 1950-69, f or tota.l cancer and selected cancer
sites in white males.. Counties were categorized as to
fraction of the county population drinking from the
mississippi iiver . . I regression analysis (unwe ghted) was
used to terni e t a cancer risk attributable to use of
the rnississippi ‘iver water cr drinkinq. ajor cancer
sites included in the stud7 were: total cancers ninas
lung cancer, urmnar7 tract, gastrointestinal tract, ll7er,

and lung. The analytical procedore controlled for rural—
urban characteristics, median income, population density,
and proportion of employed population in pet:oleu
industry, chealcal industry and aining industry.
eqressicn results for total cancer shoved three
s qnificant va .ables with the “expacted signs:
water(+), ercent rural(—) and inco e(-). The addition
of the occupaticual iariables to the total cancer
regression had little inpact..
In the regression eguation for urinary organs i he
water and urbanizaticn variable as significant, while
income was not.. With the inclusion of the occupational
variables, only the che2ical industry 7arible as
zig nif ica nt.
Begressions of gastrointestinal cancer mortality
(sto ach, large intestine, and rectum) exhibit vater(±),
urbanization (—) and inco e —) 7ariables as significaat
Aqain the chenicaJ. industry variable was significant
!egrassicns of liver cancer showed no relationship with
any of the wariables. Peqressions of lung cancer
ortal1t7 showed ‘iatar(+) and the urbanizaticn(—) variable
as significant Cccupaticnal variables, when added to the
equat on, were not gn .ficant
second st d7 (Pace and Eanris,1973 was perfc:ned
with inde end :t ia.niables sinilan to the fist fo: total
cancer, cancer of .nnrv organs, and cancar of t

qastrci testjnaJ. tract for both sex-rac groups.:
total 2c tality rates, the drinking water variable was
significant for aU groups except white fenales; the
urbanization variable was siqn.Uicant for aU :fow :qrcups;
the. income variable was significant only for white ales..-
cancer of the gastrointestinal tract, the drint i ng
water variable .was statistically significant :in all f ur
pop lat±on groups. Ccnsidered separately, dr in g-. 2vat
significant for cancer of the rectum and stoaacb,- : t
ot the large intestine, in nonwhite na.].es and feaal:es;
large intestine and rectum, but not stoaach,; i whtta
fa ales; and rectun, but not large intestine and stoaach,:
in white mai. .es. For cancer of the rinary -Or a .nSr
drinking water was a significant variable on in white
sales and nonwhite fe ales . For total, cancer mortality
rates , the three occupational variables (petro.leua,
chenical, :ining do not add to the variance explained ncr
are they significant. Page and Earns (1975) snggest that
the lack of significance nay exist because only a fraction
of tha..work force is in any of• the three industries and
thas the occupational effects are buried in the aggregate
data.. owerer, for both cancer of the ux±aazy organs and
qasto 4 ntastinal tract the chen.ical industry var .able is
significant (table !1).
The investiqators cnnc.l ded that ca nogens in
drinking water ara in suffic±e t] y high cuncentrati3ns t

e danqer human health..
Taro e and Gart (1975) reviewed “ he I plicaticas of
Cancer-Causing Substances in !ississippi iiver- ‘ater..”
Tarone and Gart’s stud7 included n additicnal variáble:
:aleyatjofl above sea level, refined the reqress ion ei:b
usiing weighted regression and expanded the. an.alysLs::tø
r ce—se groups not inittally studied by aar is-(tabla 5)

Table 4. Regression Coefficients 4 for Variable Representing Proportion
of Drinking Water from Mississippi
Cancer Sites
Al]. sites
All other th ].ung
Urinary tract
Gstrointes e(n l
‘73 3*
from Page and irr s:
* Coefficient significantly different fran 0 at 3% level.
Di ensions of deaths per 100,000 per year percent of ‘ atar :f ran the
Table 5 , Results: Presence (*) or i bsenca (NS) of Statistical Signifi—
cance at 5% Level for the Variable Representing Proportion of
Drinking Water fran Mississippi River.
Significance of Regression Coefficients
* NS *
Cancer Sites
All sites
Ganitour i nary
fran Throne and Gart, 1973.

Tarone and Gart do not recognize a : causal
relationship between water source and cancer 2ortaIity for
the follcwinq reasons: 1) the hypothesis that the liver,
havinq been de2onstated as a prime target:; organ in
experimental ani l s-tudies for organic chemically- is duced
:hepatomas, would be principally affected in populations:
beinq served wi surface water has - not been
substantiated., 2) the time sequence of cause !ississipp i
liver water) preceding the hypothesized effect: (canter)
has not been established (data is not available:
accurately datersine when, in time, the suspect
carcinogens were present in water nor the quantities and
nature of the specific chemicals themselves) ;3) :tha study
was ecological: the variables used are -; descriptive
properties of qroups and not descriptive properties of
individuals, i.e., proportion rural, median incona and
population density. The anato y of the ecological
correlation is such that tbere need be no correspondence
between the individual correlation and the ecological
correlation; 1 ) the relationshIps between the various
cancer sites and water variable lac c consistenc7 across
the. four sex-race goups here is no adequate
explanation for the significant results obtained far
gastrointestinal tract nor why, far total cancers, white
females do not show a significant asscciation with the
iiater var .a2le. If the latter results are valid and if

drinking water is to be considered a causal. factor-
cancer mortality, then an explanation for deceazed water
consumpticn or less susceptibility to potential water
carcinogens by white feaales aast be given.
• DeRcuen and Diea (1975) did another analysis - of
cancer in Louisiana with the. izsissippi Pi er: Las
drinking water source. 3ecause of known sicecono ic-
ethnic differences in Louisiana, the in4estiqators
included an additional. variable • latitude, thus dividing
Louisiana into northern and southern counties.. . rban-
rural characteristics, sedian inc e, : : eaplo ynent
characteristics and elevation above sea level,: v iabIeS
found to be statistically associated with cancer ortalit’
!n the studies of Page and Harris (1974.1975) and Tar ona
and Gart (1975) were not included in the a.nalysis by
De3ouen and Dies.. The water variable was treated as a
sisple dichotony, i.e., none of the water obtained from
the mississippi Piver versus all or some fros the Hiver.
This handling of the water variable was in contrast to the
previous studies focusing on proportion of population
using the !iszissippi !iver as drinking water source..
DePonen and Dies concluded that those counties in
a. ‘—
Louisiana being r’ ed drinking water from the !ississippi
Biver do have slightly higher cancer rates than those
receiving water fran other sources.. Eowever, there nay he
a aultiple of other causes than that c water guality

contributing to the differential.
Bu rher (1976) in a study of Cincinnati drinking
water constructed a regrass ou analysis of the cøunties of
ohio in the an.ner that the regre.ssion for t.Ouisiana
v constructed by the niro nmental efense
rates were regressed on variables: propOZtiOn tnra2
-ae4ian inco e in S1000’s, density of population- p na e-
gil., and occupation. Occupation variables p .ayed
no role in the regression equations.
The residuals were inspected to dater ine fT::
patterns were present. The residuals were iew d- asthe
cancer rates after adjustnent had beefl : de fcr
differences in proportion rural, edian.:ift and.
population density . The urban/rural variable enta ed the
regression results most frequently: 1.11 Cancer-W , W Y:
.Lu’nq—-W!; Sto ach-- ; 3ladder—- ! and W?; Iidney-—W ;
Large Intastine-—W . Density of population per square
aile entered the regression results for All - Cancer-—WZ ;
-Pancraas——W ; Lung —Wy; Large Intestine — —Wi!. . For the
variable aedian income in $1000’s the following sites were
of interest: Pa.ncreas—WF; !ladder—-W ; Sto ach— —’WL
When the cit7 of Cincinnati was co parad to Colu *is
— 4.
and Cleveland and anilton Courty was cc pared. to Yrank.lin
and Cuyahcqa Counties, the death rates sai.iqnazt
d seasa did not seefl to have any :elaticnsh!p to any of
the findings concerning drinking iater.

In conclusion, drintinq water from the Ohio iver:was
not: associated with cancer death rates higher than cenc r
death rates frcn other sources of drinidn g water
Other investigations studied the relationship :between
chloroform concentrations in d.rinking water :afl :canc
acrtality rates. cCabe in 1975 chose 50 of :t a S0 cities
.participatinq in the ICES study for further i vt1qatioxr.
Eaquirements for each of the chosen cities were a : :t9 50
population of over 25,000 and 70 perceat ormore of the
city’s population receiving water comparahi. e .t that
sampled by the E.P.A. !cCabe found. a. .ztatistica11y
siqnifica ’tt correlation between the -chlorofo:
concentration In drinking water and the :aqsez—race-
adjusted cancer mortality by city for all cancers
combined. -
Similar results were obtaine d by 3uncher (7975) for
aLl cancers in white males and. chloro orn concentratior.z
in 23 cities having populations of 25,000 or more in 1970.
Lisa pancreatic cancer death rates in white males and
chieroforn concentration had s-tatisticaLLy signifIcant
correlations for 77 cities- specificaLly for 59 cities
vith surface water supplies and for cities that contained
more than O% of the o ty s popuiation Per cities with
greater than 70% of the county’s ;opulaticn. there was a
significant correlation between choroforn concentration

cai TH3

and bladder ca:cer for white males and white females.,
To sunmariza: studies by the 3nvironmental
Protection gency have affirmed the large number of
arq nic chemicals present in public water supplies . Whjle-
th ,; potential health hazards of many of these che icals
are .-; unknown, others have demonstrated carcinoqenic
activity in laboratory animals under experimental
con4iticns; still others have, in limited ses,: been
implicated in causinq cancer in hunans..
The lar e scale a pidemioloqical studies, to data,
iave—-at best- -pcintad to an association betweea sourca of.
drinkinq water and cancer nortality. Rowaver, the. data
has,. in general, lacked consistency and coherence., rose--
response curves have not been ascertained nor has the
question of temporality been adequately addressed..

The qeoqraphical area of the present - study
incorporated 3 6 counties selected fom -seven states
(Illinois, Indiana, Kentucky, Ohio, syl’rania,
Tennessee and est Virginia) having a total pcpulation in
1960 of 18,721,791. Selection criteria for counties was
location vtthin the boundaries of the Ohio 2iver Valley
Dazin as deta:ni .ned from state maps and vater drainage
maps. ?cur counties in Virginia, initially selected for
study, were deleted since two of the four counties had
aeen merged in census data..
Y tal Statistics crtality ata nsed was the race,
sex-specific direct aqe—adjisted. death rates for the 3056
counties of the ccnt qucus fl. .S.1.. over a 20 year period,
1950—1969, available from the National Center for ealth
Statistics, 3ock villa, Maryland in D e publication no..
31! 7ti 615 . Corresponding county populations were
provided by the 1950, 1960, and 1970 censuses with
interrensal estimates derived by linear inter;olation.
!or thi ty—f ve can r tes, age—standardi:ed crtality
rates by race and sai in each county had been caic .latsd;
the standard. beinq the age t. uticn of the 1960

pQpuj.aticn of the 7.5. Cancer sites were classified
according to the Sixth evisjon of the nternatjonal
Classification of Disease. he cancer ortai.1ty data was
available on tape from the ationa1. Cancer -Institute .
J acb tape record has six variables; the nunber -of reccrds.
for any one:county may vary from 1 to a aaxi aum of 140:
(the number of possible ccmb nations of sax, race, and:
cancer sites is 1Z 0).
Cancer sites (Internat cuai. Classtfication : cf
Disease,6th ed.). Chosen for study were: esophaqu.s
(150), larqe intestine, except rectun (151), rectun (15L ),
biliary passages and liver (155), pancreas (157), laryn:
(167), trachea, bronchus and lung specified az prisary;
and lung and bronchus unspecified as to whether prinary or:
secondary (162,163), breast (170), prostate (177), kidney
(180). bladder and other urinary organs (181), Eodgkins
Disease (201), lynphosarccma and- reticulosarcoma (202),
multiple myeloma (203), leukemia and aleukemia (204) and
all. malignant neoplasns (140—205). A file of the cancer
sites was created from the original, tape with 68 of the
possible 1 0 combinations (due to the elininaticu of sorne
cancer sites). iflle of al ]. couties in the seven study
states vith selected. socic-deuograpI .ic data was eated..
he zccio-de r pbic data was obtained fron the City-
County census books ( S) available on .tape . aria.blez
selected included: ercant urban pcp J.at±on, :uu er

eapl.oyed in aqiculture, percent employed in inera1
industry, numbex enployed in the anufact’ure of non-
durable goods, total population, population per squa.re
mile, percent non-white, median age, median number of
yee.rs of schoolinq completed by the population. iover the-.
of 25. median income, population change (tótaLand
percent), percent foreiqn stock..
Public water source data was obtained f oz- the
T.S.P.E.5. Public ‘water Supply Survey Publication, no.
775 1963 which provided water in rmation oü’ communities
(listed alphabetically for each state) tbroughout the 11.5.
I formae.on includes population (1.960), source -cf the:
public water supply, number of people servedand vatar
treatment data.
!ach community in the study state of the Public iatar
Supply Survey was identified with the appropriate county
by referring to the 1960 census p bl.ications .A county card.
was developed with the f i1cwiaq information: state,
county, 1960 county population, source of water supply,
estimated coamunity population served by this source and
type of wa -tsr traat ent with the estinated ccmmun t7
population served by such vat r treatment. Community
pcpulatian figures were validated against 1960 census
data; the estimated. population served by a specific water
source was made using individual state water data obtained
from t e !nvj:cnnental protection ) gency, Cincinnati.

check was 2ade of all connuaities not ‘atched in both
water publication and census sources--the 3ajority of
these coa unit±es had populations of less than 1000. Of
the; 3 6 counties initially selected, eight had no water
data. Of these eight counties, six had no nnun ities
yLt1 populations greater than 1000. Of the re .aining two,
one county had a single - coasunity of a size exceeding 3000
people and the second county had one coa.aunity with a
population of slightly over 5000 people. ecaase of the
larger ccunnity population it is assused that the water
data is aissinq rather than being totally dependent on
private wells. Thus, of the counties with some proportion
of hair population unaccounted for with respect .t - water
supply data, nforaation frorn census zoo.rces on, pcpulatLcn
size and private coanunication ( cCabe,i 976) would
indicate that most are served by private wells.
Initially, for each county, public vate supply
sources ‘were cataqo .zed ±uto six groups: a or rivers
(defined as ajc: t .butaries of the O ia 9iver) • ±nor
rivers, wells, lakes, springs, and unknown. The total and
percent population being served under each of the
categories was calculated. iqa.in, for each county, public
water supply sources were categorized by the co2annity’s
type of water reat en L ?our qrcups were eated: o
chlorination, p:ech lc ation, poztchlcri ation and both
ec1lorinatioi and poztchlorinat .cn . The total number of

:people and percent of population served by each treatment ”
process was computed. and recorded.. The operational water
variables were 1) percent s f ace water usage (nunber of
:peop].e receiving tbeLr public water supply fro3 a1or and
:iin Or rivers divided by county population tines 100) 2)’.
percent prechiorination (unaber of people drinking water
that has undergone precklarlnation and - 1 /cr both
prechlorinatjou and postohioninatien bined divided’ ‘by
county population tines 100). In assu ption regarding the-
f crier is that the percent of the county population whose-.
source of water was unknown was, in fact, servedby:water:
froa sources other than rnajor and amen rivers..-
The three original data sets coapnised:—
(7) water data file on cards for 3 6 counties
(2) aggregate cancer file on tape
(3) census file of socio—deaographic varia les
The water records were ‘sorted in ascending order
according to population si ze. 12 state codes oi atar
records and census records were inconsistent, all record
IDs of the cansu file were printed and the corresponding
state ccde frea the water files iere then aerged to
prad e a coaposite wcrk ng f±1e

The study focused on two different predictor
tariables (percent surface water usaqe and percent
prechlorination) using thee as test variahies in
successive order. To analyze the r2lattonship ‘between
age—adjusted cancer rates (site, race, sez-spe .fic and.-
the. water - vairabla cf interest, unweiqhted ul tple
regression analysis was performed . The county was the -
unit of study. The denoqraphi’ variables selected for
inclusion were: percent ncn-white population, edian
inco te, median number of years of schooling completed by
the population over the age of 25, number employed in
aqriculttxre, number employed in non—durable manufacturing,
population per square sue, percent urban pop L1ation,
percent foreign stock and percent employed, in
aanufacturi ug.
3ecause of the extreme variability of the size of the
individual county populations, weighted regression
a— .
analyses were used on the same data mats as in the
nveiqhted regression procedures.. For each coun-ty, the
weighting factor was the square rOot of the tot county
pooulation which, under specific aszu pti ns, will be

inversely proportional to the standard error of the
2crta.Lity rate (Rcover, 1975). While it is preferred to
use the square root of the appropriate sex-race strata of
the total 1960 population, the data file being ezployed
did not contain a sex—race breakdown for- the cOunty
population and total population figures were substituted.
However, this substitution would not be a ajor problem
given a sax-ace ratio constant across the study: count es 1 .:
The-usa of the total population in this study :2ay be
reasonable qiven a recent report ( ogan et al . ,1 76) where
regression analyses co ;aring weights using total county
population and sex-race county popui.ation figures for
- Region V counties showed no conclusions were : altered by
the choice of total population over sex-race spad.fic
Pollcwinq the prelininary screening procaduzes by
unweighted and weighted ultip1e regression procedures,
the reqress cn cdel was re ducad Those cance.r sites
having a. p—value of .10 or less (pcsittve direction) based
on the standard t-test were selected. or the individual
cancer sites selected, socio—econosic variables having a
two sided p—value of .20 or less based o the standard t-
test were chosen. llso, ste wise regression results were
Re;eated analyses were done relating cancer orta1ity
of s;ecific cancer sites to percent surface ‘later and t e

yarious explanatory variables in the reduced aodeL.
mother set of analyses iere perfor2ed for the sane cancer
sites and. de2ograpb.ic ;a. .ables but substituting percent
pechiorinaton for percent surface water usage.
Treating percent surface water usage and perc ent
prechlorination successively, sone groupings of cancer
sites were carried. ont. Por each race—sex g p, w.eiqhted
reqressicn van dcne on the coabined cancer act lity at2s-
for la:qe intestine and. rectus, for the urinary tra
(kidney and, bladder and other urinary organs) and
qastrointesti.nal tract (esophagus, stoaach, large
intestine ar4 ractn ) use, because of possible bias in
the cancer aortality data of large intestine and rectum
due.: to differences in diagnostic reporting, slaba ling
and raq onal differences, large intestine and rectua were
combined. Pollowinq these procedures, as a ude
surrogate aeasure of the possible.. inpact of air pcllut on
and sacking, lung cancer ao taIity rates were included in
a series of regressions as an independent 7ariable..
Given an expected iapact of population size on cancer
iortality rates, the study counties were grouped into
three -trata based on population size: less than 50,000,
50,000—250,000 and over 250,000. weighted regression
analyses using the reduced node.]. was again eucc szi7ely
sad fez percent surface water usage and percent
prec hiorina ti on

Thee to increase precision, these ana3.yses were
repeated select±nq only those counties for inclusion in
which 50 percent or nore of the county’s water sou rce w s


The results of the esticated percent surfaca water
coefficients from unweighted and weighted re.qressjcn
aalysez of the four race—sex groups are su arized as
follows tables 7,8);
1.. except for lymphosarcoma S reticu.lcsacco a ‘in
white males and. multiple myelcma in white females, whetber
the presumed percent surface water usage is significant or
even su gesti7e for any given site depended cn whether
unweighted or weighted regression anal yses vere ernployed;
2. with un eiqhtad regression procedures, percent
suface water usage showed weakly siqnificant or
significant positive statistical association with the
follcwinq cance r sites (t bl s 7,$3): in ifl .ite males,
lymphosarccma & reticulosarcona {p.096); in white
females, multiple myeloma (p.077); in non—white males,
rectu3 (p=.077) and pancreas (p=.025). In non-white
females, no significant asscciat on was found.
3.. with weighted :eqrsssicn procedures, percent
Surface •iater isage shovad veakly significant er
sigmi.ficant positive statistical asscciat cn itb the
following (tables 7,8): in white males, esophaq

(p=.O 1) , large intestine (p=.053), recturn (p=.015),
larynx (p=.003), trachea, bronchus & lung (p..O15) *
bladder S other urinary organs (p=.013) , lymphosarcoma &
reticulosarcoma (p .Q72) and all malignant neoplasas
(p 4 .OOO); in vhite females, rectum (p...OO7}, breast
(p. ..1OO) and multiple ayeloaa (p=.068); in. non—white
aa.les, no significant positive associations ‘were -obser ad;-
in : non—white females, esophagus (p .=.350) and: : l ary x
4. except for rectal cancer, prelininay regression
procedures aanifested no consistency across r ce-sez
groups ncr was there a pattern by sex or race (ta.ble :6).
5 . the sixty—six regressions (unweighted and
‘we qhted) evidenced a greater number of cancer sitas as
significant than would be expected by chance alone only
for white males (weighted regressicu).
The weight, w , asscciated with the ith observation,
, is a measure of the. relative importance of the
observation in the final rasult Usually the weight is
taken as the reciprocal of the variance so the
observations with the smallest scatter are given the
greatest weight. lssuxinq the àbservaticns are
unccrrelatsd, this procedure gives the best estinate of
the pcpu..lation rnean, i .e., an unbiased esti.uate with
jz±mu: variance (naxi un precision).

In this stndy, because of the great variability-
the size of the population at ris , each county
observation vas veighted by the square root of the county
popnlaticn:— - every variable in the observation—; was
2ultiplied by the weiqht.

Table 6:
Siir ry of Significant Results between Percent Surface Water Usage
and Site, Race, Sex—specific Cancer Mortality Rates, 1950—1969,
in 346 Study Counties. -:
Cancer Site Male Female Male Female
Esophagus S+ MS MS S+
Stomach MS NS MS MS
Large Intestine MS MS MS
Rectum S+ S+ S’ -MS
Biliary P 2 ssages MS NS MS -- MS
& Liver
Pancreas MS MS S’ MS
Larynx 5+ MS MS 5+
Trachea, Bronchus S+ MS MS MS
& Lung
Breast NS MS NS
Prostate MS MS MS 1 MS
Kidney MS MS MS MS
Bladder & Other S+ MS NS MS
Urinary Organs
Rodgkins Disease MS NS. MS MS
Ly phosarcoma & S + MS MS MS
Multiple Nyeloma MS S +• MS MS
Leukemia & MS MS MS MS
Aleuk ’i
AU. Malignant S+ . MS MS MS
Neoplas s
p—value based on standard t—test
* significant p = <.10 C or e-tai1ed test), unweighted
+ significant p = <.10 ( on tai1ad test), weighted

¶able 7: £ tienc.d Perc. nt Surface Water Usage Regression Coefficients and Associated p—values froa Ut—
weighted sad Weighted Rsgr.esion Analyses of Whit. Males, White ?enzles and Site—specific Cancer
Mortality Rates Versus Selected enographic Risk ractors for Malignant eoplasa Mortality, 1930—
1969, in 346 Study Countiu.
TInv.Lghced Weight.d
R.gr.s sio Regression
• 0407 • 0006 - .6923 .0003 .6350
.3830 —.0037 .308T —.0043 .1532
.0526 .0001 .9272 —.0020 .7329
.0153 .0011 .7984 .0067 .0068
.4892 —.0045 .478]. —. 0003 .8100
.5143 .0007 .3675 .0032 .2213
.0033 —.0001 .9392 .0008 .1100
.0134 —.0023 .6190 .0025 .3269
.3027 —.0034 .7260 .0095 .0993
.3621 — — — —
T nw.ighted
3 p*
3 p
3 p
Ridney .0024 .5063 —.0008 .6845 .0030 .3052 .0006 .6598
Bladder 4 Other Urinary Organs .0054 .1774 .0062 .0133 — .0007 .7333 .0018 .2206
Sad girins 3iseua .0009 .7473 .0022 .1683 — .0019 .3468 .0017 .1077
Lyaphosarcoaa & R.Uculosarcoss .0062 .0962 .0038 .0720 .0015 .6279 .0004 .8301
Multiple Xytlans .0000 .9901 .0016 .2171 .0036 .0773 .0020 .0679
Leukeeia & Aleukesia —.0013 .8333 .0017 .5856 —.0033 .4972 .0018 .4493
All Msi.igaanr M.ap1a .0200 .3556 .0784 .0004 —.0218 .4504 .0203 .2533
rtvo aided p—value based on sta945x4 t—test.
I p
erg. Intestine
Biliary Passages & Liver
Trachea, 3r nchus & lung
.0015 .6432 .0039
.0039 .6381 .0007
—.0061 —.7219 .0106
.0027 .6263 .0079
.0004 .9425 .0019
—.0015 .3022 .0023
.0021 .4063 .0042
—.0069 .6377 .0234
.0006 .8925 .0003
.0035 .7188 .0010

Table I Esifzatsd Percent Surface Water Usage Regression C3efficicnts and Associated p—values from
aveighced and weighted Regression Analysis of on—vhi:e a1es and on-vftit. Fenalas and Sits—
Specific Cancer Mortality Rates Versus S*lected D ographic Risk Factors for Ma i; snt ‘ ecp1aas
Mortality, 1950—1969, i 346 Study Counties.
misigttt.d Weighted Unveightad WeiZhted
Regression Regression B.;rasaion Regress1o
a a
Epitava —.0007 .7833 .0071 .6758 ..0016 .8187 .0107 .0499
—.0019 .3468 —.0297 .3063 —.0219 .3028 .0122 .5481
Large Intestina .0015 .6279 .0063 .9121 0644- 3855 - 0471 .4854
.0036 .0773 .0336 .1527 —.0227 .3912 —.3114 .4324
3i2.iary Passages 4 Liver —.0033 .4792 .0031 .8754 —.0230 .2659 —.0249 .0672
Pancreas .1145 .3246 .0477 .1629 —.0201. .5572 —.0092 .61.86
Larynx —.0123 .3633 .0063 .6203 .0087 .1123 .3071. .0385
Trachas, 3roachus S Lung —.1444 .2873 —.1070 .1327 —.0924 .2312 —.3305 .2979
3reasc .0004 .3624 .0019 .1697 .3544 .3823 .0239 .6742
Prostate .0129 .8764 — .0073 .8903 — — —
Ucw.ight.d Weighted Unveighted Weighted
Regression Regression Regression Recess i on
3 p 5 3 p 3 p 3 p
Ridney —.0256 .6063 —.0330 .1799 .0134 .2284 .0027 .6679
3Ladder & Other Urinary Orans .0360 .4177 .3078 - .6697 —.0182 .3698 —.0288 .0503
ffodkisa Otsess. .0054 .5867 .0007 .9060 .0083 .2626 .0016 .7170
Lymphosartoes 4 leticuloearcama.0337 .1137 .0236 .0387 .0009 .9794 .0029 .3791
Moltipi. Myelcna .0068 .7796 .0068 .3244 —.0010 .9384 . .0069 .4139
tsuk. a 4 A1auk a —.0369 .4863 —.0011 .9680 .0128 .7855 .0117 .7333
All. Malignant ¶scplaans —.1102 .7216 - .0348 .8316 —. 0684 .8112 -. 322.4 .3840
‘Tvo— i4ed p—valu based on sraj datd :—vsst.

PoU.owinq preli inary n’veighted and. weighted
a aj.yses, several. individ a.]. cancer sites ware combined
since relatively by site—specific cancer mortality could
result in bo numbers.., Also, combining several individua.l
cancer sites-allowed, us to analysis related organs, i.e., -
o qan systems.. Since experimental and epidealoloqica.t.
studies (!hthik S Pichie, 1953; Page arris, 1975k -
indict the gastrointestinal and urinary tract syst9 s as -
potential prime ‘targets for ca.rcinogenic activity, for
each sax—race group, weighted and we qhtad regression
analyses were pezfoz ed on several sites combined into two
systems; esophagus, stomach, large intestina and rectnm
comprised the qastrcintestinal tract system while id.nay
and:-- bladder & other urinary orq ns comprised the -mrina ’y’
tract system..
In whites, the gastrointestinal tract shoved a
s±qnifLcant statistical association (p.,013), table 9) for
males ‘when ‘we±ghting was used——a result consistent with
data f:c the ind.i7idua2. sites... urinary tract showed no
positive association in ‘vh te males,. Neither system
shoved a positive siqni f cant association for other race—
sax groups (table 9,10).

Table 9: Est -m- tad Percent Surface Water Usage Regression
Coefficients and Associated P—values fron Unweighted and
Weighted Analyses of the Gastrointestinal and Urinary
Tract Systems Cancer Mortality Rates Versus Selected
Demographic Risk Pactors for Malignant Neoplasn Mcrtality,
1950—1969, in 346 Counties.
unweighted weighted unweighted weighted
regress ion. regress ion regression regres sion
CancerSites p a p a
Sta ach
ç .0013 .9224 .0230 .0.29- —.0059 .6431 .0007 .9211
Large Intestine
Bladder & —.0010 .8981 .0043 .3414 —.0038 .6120 .0024 .5615
*t ...sided p—value based on standard t— test

Ta 1a 10: Estina.ted Percent Surface Water Usage Regression -
Coefficients and Associated P—values from Unwaighted andy
Weighted Analyses of the Gastrointestinal and Urinay
Tract System Cancer Mortality Rates Versus Selected
Demagra hic Risk Factors for Malignant Neoplasm Mortality
1950—1969, in 346 Counties..
unveighted - weighted imweighted weighted
regress .on regression regression regression -
CancerSites p p
/ Stomach
—.0548 .7093 .0173 .8242 .0185 .8824 .0585 .4173
Large Intestine
Recti .
Bladder & .1047 .1004 .0508 .2922 —.0429 .2887 —.0341 .1589
*two_sidad p—value based on standard t-tast

Having completed the preLisinary analyses (tables
7.8). specific cancer sites and associated indepandent
variables were - selected for f -ther analysism I reduced
cdel unique to each site was tasted since.independeat
varIables were not consistently significant :a os3 all
specific cancer sites. The “reduced cdal” includad, for
each race—sat group, only those cancer sites associated:
vith percent surface usage at a level of p<. 10 tone :sided):
based on the standard t-tast.. The criterion for inclusion
of de graphic variables into the equation was a two—sided
p—value of p<.20. Stepwise regression procedures :were.
also perforaed as a means to sale ct the appropriate
yaniables. Thus, for each selected cancer:.site, - the
selected denoq:aphic variables would be specific to that
cancer site.
Overall weighted regression analyses on the reduced
model had sin .rnal impact or. the regression coefficients
and associated p—values. The size of the regress icu
cceffic ents and associated p-values for the esophagus,
large intestine, rectum, larynx; 1 q, bladder and all
aliqnant neoplasas In the whIte 3ale underwent lIttle
change. The only site significantly affected was
lynphcsarccra & :e .culosarco a which changed fron wea cl7
siqnif cant in the full regression :odel to n-cr.—
significance In the reduced cdel: p=.O72— p.l 8 2 (tables
7, 1 1)

3zcept for breast cancer, little difference
observed in the respective sites in white females between
the full and reduced modeL. In the reduced model, breast
cancer mortality had a statistically siqnif cant
association with percent surface water usage of p+.000
while in the full regression model, the association was
only weakly significant. The respective regressicr
coefficients more than doubled (tables 7,11). xanining
the variables that had been present in the fuil model. but
removed from the reduced model, four of the five
($employed in non—durable manufacturing, population per
square mile, % urban population and % foreign stock) were
significantly correlated, p=+ .00 0 , with both east cancer
and percent surface water usage. hns, it is possible
that percent surface water usage is acting as a partial
surrogate for the deleted variables.
In, non—white males, the panceas which had been
significant in the unweiqh ted (but not in the weiqhted
reqrassicn) in .the full model was now significant (p.032)
in the weighted re.gression with the reduced set of
variables- In the reduced model, rectal cancer mortality
re*a red unchanged from that of the full model (tables
8,11) No apprec ble change in regression c efficiantz
and associated p—values were seen betzeen the full and
reduced weighted regression rodel in noa—white enales
(tables 9,11).

hi1e u1ticcUinearit7 was a p ob1e i a eng the
indepa dent va iab2.es, the overall lack o cha qe in the
reqrassicn cce fficientz when the 2odei was rednced
indicated that the variables Gxcllidsd fto the f iU. od.el
had little effect on the water csage factor.

Table U : Est atad Percent Surface Water Usage Regression Coe ffi—
dents and Associated p—values* fron Weighted Analyses
(reduced model)+ of Selected Sex—race, Zita—spec fic Cancer
Mort T{ty Rates Versus Selected Demographic Risk Factors
for 1 a t Neoplasm Mortality, 1950—1969, ±n 346
White Males White P ’m 1 es Non—white Males F 1es
Cancer Site p p p
Esophagus .0038 .0412 .0103 .0267
Large .0092 .0808
Rectun .0071 .0269 .0084 .0006 .0213 .2270
Pancreas .0464 .0380
Larynx .0047 .0006 .0047 .0818
Trachea, .0238 .0034
Breast .0250 .0001
Bladder & .0053 .0148
Other Un—
naiy Organs
Ly phosar— .0020 .1313
cona & Reti—
cu losarcoma
Multiple .0019 .0562
AU Malig— .0810 .0002
nant Necplasns
* jc sided — alua based at standard t-tesr
+ inclusive criteria: p <.10 in full cde1, posi:.ve

Differences in diagnostic raport ng, islabeling and
regional differences nay have caused bias in the cancer
2ortality data of large intestine and rectun. And, since
rectal cancer mortality was siqnificantly associated with
percent surface water usage for three of the four race-sex
groups and since large intestine was also positively
associated with percent surface water usage in white males
it saeiied necessary to run another set of analyses with.
large intestine and rectal cancer mortality combined..
Rasu.lts frog coabining the two sites were relatively.
consistent with outcc es from separate site analysis., o
statistically significant results were obser7ed with
unweightad rea:ession procedures for any race-se: group.
With weig.hted regressIon procedures, large intestine and
rectun together produce a positive statistically
significant association (p-=.O11) with percent surface
water in white gales. ll other race—sex groups shoved no
significant association (tabte 12) -

ab1s 12: w.thtM and vsi bc.d R* ession a.1yi.s of ?* ent Surfaca ats Usa. s and C02—
bin.d sires: t. ,. Intu Un. and ac f r a.U. ac .—s .x gra
C. TC! SI! J.i,titsd
3 P
nit. a.te — .. 32 .7541 . t:3
t.hL:. T .L. —.0007 .9451 .0047 .4ô 8
ou hi . a1 . —.0349 .7761 .0399 . 645
!on Whits Yenala .0419 .7290 .0356 .6074
Tvo—sLdsd p—va1 as based on standa rd t-t.sc

‘ost cancers have a aultiple etiology. hr pollution
and s:oki.nq are potential portant confoinders in the
present stndy. !ince an adequate index of air poi.lut cn
on a ccu ty level vas not. available and since data on
stoking were alac lacking, a surrogate ceasure for both
air pollution and saoking was used. This surrogate
aeasure, aqe—ad usted lung cancer mortality rate, was
included in the reduced weighted reqressicn model as an
independent variable and regression analyses were
repeated. V
With arcent surface water as the var abla of
interest, it is important to iicte that the overall pattern
of relationships remained stable (tables 11,13). ?cr
white V males and white females, the direction, size and
associated p—values of the regression coefficients
re3ained relatively constant.. For non—vh tes, female
esophagus and larynx and male rectum did. not shcv a malor
change. Only in non-vh.ite aJ.e pancreas was there a siqn
reversal (3:.O 6,- _ ’:.O59) in the estimated regression

Table 13 : Lsti ated Percent Surface Water Coefficients and p—values
for Cancer Site by Race—sex for Weighted Regression
At yses (reduced model) Including Lung Cancer Mortality
as a Predictor Variable.
Race—sex Site p_ralue*
White Male
Esophagus .0033 .0632
Large Intestine .0119 .0276
.0063 .C447
Larycz .0044 .0011
Bladder & Other .0065 .0042
Urinpry Organs
Ly pbosarco a & .0005 .7820
- Reticu1osarcc a
All Ma1i tt nt .0389 .0236
Iceoplas s
White 1 e
.0072 .0032
Breast .0187 .0001
Multiple Mye1o a .0018 .0737
Non—white Male
Racti.ua .0283 .1052
Pancreas -.0589 .0184
Non—white a
Esophagus .0114 .0176
Larycz .0056 .0601
*ided fl—value based on standard t-test

Ovfrall, inclti .on of lung cancer mortality rate as a
independent variabi.e did not altar t e siqnif cant results
observed betvsen percent surface water and the selected
cancer site s..
The regrassion nodel e plo7ed not be consistent
across all population strata. Therefore another set of
analyses ware perforned with thq counttes stratified by
three population strata: counties wtth less than 50,000
population, 50.000—250,000, and sore than 250,000
populatIon.. The results are suanar zed in table iLL
Generally, the individual cancer sites showed a
pattern for regression coefficients which in eased in
size across the population strata (smaller to larger) with
the largest population stratun (>250,000) 5howing the
greatest nunber of significant associations between the
cancer site and percent surface water usage variable
(table 1 ) .

Table 14: !sr w ted Percent Surface Water Usage Coefficients tth
.Value$* from Weighted Regression M 1ysis Stratified by
Pcpui.ationr for 346 Study Coimeies.
<50,000 230,000 >230,000
Sex-race Ceneer Sits B p-vai.zie p—value a p—value
hits Males
—.0011 .7422 .0005 .8905 .0100 .3153.
Large Intestine —.0159 .0611 .0029 .8168 .0669 .0743
—.0024 .6101 -.0005 .9519 .0333 .2516
Larynx .0012 .6426 .0026 .3858 .0107 .0895
Trachea, —.0233 .1686 .0363 .1136 .0391 .3640
Bronchus & Lung
Bladder & Other .0026 .4803 .0128 .0001 .0012 .9258
Uri ’7 0rg s
AU Mi-i t —.0581 .1076 .1039 .0453 .3317 .003.6
White Fe a1es
—.0108 .2861 —.0040 .7231 .0197 .0912
Breast .0091 .321 .0107 .2432 .0512. .0089
i1tip1.e fye1oa .0043 .0331. .0002 .9369 .0018 .1312
Non-white a1ea
Rect m .0187 .8577 —.0074 .9113 .0142 .5437
Pancreas .2447 .0001 —.0418 .2973 .0013 .9193
Non—wt2.ita 7enalea
soDhagus —.0059 . a.089 .0060 .6951 .0208 .0035
.0136 .0378 .3070 .1689 .3036 .1949
* ,c_ jded ?—‘Talue ase cn z a dard -tast
<50,000 ( 272); 30,300—250,300 ( — 62); >230,300 ( — 12)

populations of <50,000..
For non—white feiales, e ophaqeal cancer mortality
was associated (p= . .00 ) with the water variable in the
largest population stzatu (>250,000) with the regessian
cceff cients showi q an increasing qradient from the
smallest to the largest population stra a. The reverse
was true for laryngeal cancer.. The size of the regress±on
cceffic±ents decreased vit increasing ccnnty popn.lation.
Counties with populat±ons of less than 50,000 bad a
siqniftcant azscciae.cn of p=.038 between laryngeal cancer
mortality and percent surface water usage.
To increase praci on, only co ties having 50
percent or more known water source were chosen for f t her
analysis (n1 9). These counties were then stratified by
county population size: <50,000 (n=9 ), 50,000—250,000
(n 5), >250,000 (n=10) . .
8cth the reduced set of counties (rt=1L49) and the full
county set (n=3 6) showed marked consistency in the
regression results when stratified. by county populat ion
size (table 1Z,15).

Table 15: Estinatad Percent Surface Water Coefficients with p—values
from Weighted Regression Analyses Stratified by Pcpulat on
for Counties with 50 Percent or More Known Water Sourca.i
<50,000 250,000 >250,000
Sex-race Canner Site B p_value* B p—value p-value
White Male
Esophagus —.0042 .3141 —.0019 .6905 .0114 .1611
Large Intestine —.0182 .1016 .0108 .4507 .0803 .0234
Ractun —.0068 .3091 —.0024 .7945 .0387 .li020
La iy x .0010 .7824 .0029 .4319 .0104 .1647
Trachea, —.0315 .2096 .0269 .3046 .0422 .6020
Bronchus & Lung
Bladder & Other .0017 .7442 .0163 .0016 .0011 .9287
Urinary Organs
fl I4aliguant —.1168 .0258 .1091 .0582 .3352 .0835
Neop lasms
White Fensla
Ractun —.0029 .6171 .0066 .2292 .0281 .0691
Breast .0110 .3019 .0034 .7324 .0502 .5303
x1tip1e Myeloma .0042 .1145 .0010 .6546 .0024 .2091
Non-white Male
Bactun .0241 .7664 .0214 .6004 .0189 .2635
Pancreas .3483 .0011 —.0590 .2166 .001.5 .9341
Non—white Pale
Esophagus -.0138 .1671 .0069 .7090 .0217 .0095
Larynx .0238 .0366 .0053 . C55 .O0h 3 .1.590
,—value based cn acandar - ast
—<50,C00 ( 94); 30,000—250,000 ( 45); >250,000 ( 10)

As with the full model in white ales, esophagus,
large intestine, rectus, larynx, trachea, bronchus and
lung and all aliqnant aeoplaz s showed regression
coefficients increasing in size across population strata.
Uc ever, while in the full cael large intestine and
larynx had significant association between the respective
cancer sites and the water variable in the largest
population stratn ( >250,000), when the nu2ber of counties
were reduced, only large intestine shoved an association
with the water variable (p=..023). 7iaally, in white ntles-
all aal gnant necplaszs prodnced regression coefficients
increasing in size over the population strata and had. a
päsitive statistically significant association between the
co2bined sites and percent surface water usage in two
strata: 50,000—250,000 and >250,000-—as had the ode1
with the full set of counties.
Par white fe a1es, reduction of the nunber of
counties esployed resulted m only rectun shoving both an
increasing regression coefficient gradient acoss
population strata and an association (p=.069) between the
site and percent surface vatar usaqe in the counties with
the largest population size, >250,000
3on—vhite rnales showed no significant asscciat cn
between rectun and the water 7ar±able ut there was a sire
gradient in its rsc ress cn coefficients ‘with the lar;est
coefficient ‘being in the county population size of

<50,000.. While in pancreatic cancer mortality, the
patte:n was si i.1ar to that of the full ode1 (n=3 6): a.
statistically significant association between the site and.
water variablein the s a1lars±zed counties (<50,000),
gon—whita fesales had sinilar results whether the
full nuaber o counties (n=346) were used or only those
counties with 50 percent or sore known water source
(n 1 9) Zsophegeal cancer crtality expressed regression
coefiic±ents that increased in size across population
strata with the cancer site and water variable hayinc a
positive significant association (p=.010) in the largest
stratun (>250,000). Is was also true in the full model,
lar7nqeal cancer nertality had regression coefficients
decreasing .n size froa smallest populated counties to
largest with a positive statistically significant
asscc ation (p .03 7 ) seen in those counties with
popuJ.at±cns of less than 50,000.

pc r
Since recent studies have suggested a posz ble
relationship between the t7pe of water treatnent process,
trihalo ethane production and cancer potentia]. (Syncnq
ai.,1S75; ifoqan et al..,1976) , the process of treatnent cf
pablic ater supplies by chlorination nay be of -
i portance. In this study, the process of prechlorination
was con .dered a crude surrogate oeasure of possible
trihalonethane levels in finished drinking iatar..
Consequently, data was collected for each conannity in
each county studied on whether the water supply underwe.nt
chlorination and if so, at what phase of the treatnent
process the chlorination was added.. n tiaUy the four
groups created were: no chlorination, prechicrinaticu
(defined as chlor naticn prior to filtration step), post-
chlorinatio n (chlorination after filtrat en) and both
prechiorination and pcstchlorination .. For analysis
purposes, ‘ater undergoing both prechicrination and
postchlcrmnation was added to the prechi.ormnaticn qroup
he nurber of people ser7ed y public water supplies
undergoing rsch:jnatcn ias then &i7ided by the
ccunt7 po lation to q ’7e a percent o the t ta county

population served by water undergoing prechiorination
treatnent. !scause of the possible link of trihalo ethane
production with cancer and the relationship between the
- organic content in water, chlorination process and
tlihalo2ethane prcductian, it is assused that those
counties having a larqer proportion of its population
served by prechiorirated water would have higher death
rates than those served by water undergoing no
chlorination process or undergoing poetchionination only.
Site—specific cancer iorta.lity rates ware regressed -
against percent prachlorinarion and a number o socic-
acoso3 c variables for each of the race—se: groups. I
su ary of the results are as follows (tables 16,17,18):

Table 16: S’i ry of Significant Results 3et een Percent Prech2.oro—
nation and Site—, Race-, and Se —specifjc Cancer tort 1ity
Rates, 1950—1969, in 346 Stndy Counties.
•C cer Site Male P le Male ‘e a.1a
Esophagus NS MS MS S i-
Stomach MS MS MS MS
Large Intestine NS MS MS MS
Rectun S*t si. S* MS
Liver & Bii.iaiy Passages MS MS MS MS
Pancreas MS NS Si MS
Larynx MS MS MS
Trachea, Bronchus & Lung MS MS MS MS
Breast MS Si- N! NS
Prostate MS MS MS MS
Xidney MS MS MS MS
Bladder & Other Urinary Organs S*t MS MS MS
Eodgk(rts Disease MS MS MS MS
Ly hosarcona & Raticu1osarco a S t N ! MS MS
1tip1e Mye1a a MS N ! N! MS
Leukenia & Alaukaria N! N! MS
All Malignant Neoplas MS N! MS
p— raiue based on standard t-cast
*sjgnj fj t p — <.10 (one—tailed test), un eighted regression
ts gnificant p” <.10 (one—tailed test), e ghced regression

a 1. 11: !aci u.d Parcanc ?rsth .1cr nacion 3.p.s.Lou Co. ficiencs and Associued P—values froa .ej c.d
aM J.i 4 hcad Rs raasioa Ana .Lys of hi a a1es aM Fs a1.a and SLca—ep.c.tftc Cancer
orca1.tcy Races Versus 5.L.c cad Ds rspbic Risic racora for a1ignanc sopLa r:aJ.iy,
1930—1969, tn 346 Co c es..
ita Fenalie
3a r aiaa Regression Z. v. ssiot Ragression
3 p 5 3 p 3 p 1
.0043 .1.72.5 —.0029 .0973 .0003 .3398 .0008 .3047
—.3054 .5009 —. 31.00 .0168 .002.5 .6.457 —.0004
.args tnc.scics .0039 .6334 .0022 .6495 .3052 .5760 .0034 .4924
Racc .0072 .0900 .3141 .0001 .0041. . 3U.2 .3062 .0052
a .0092 - . 3 .1. 7 2 .0020 .4309 .0033 .5961 .0023. .4889
3i1.iaxy Passage.
—.0034 .3723 .0032 .3167 —.0002. .9764 —.0000 .9961
t.azyux —.0004 .3722 —.0033 .0028 .0004 .6523 —.3002 .6483
7racnda. .0238 .1.709 .3019 .3419 .0037 .41.38 —.0064 .0060
3ronc ua I
—.0005 .4337 —.0002 .3977 .0081 .3972 .02.30 .01.1.9
? usc.sc. .0059 .3320 .0050 .3223 — — — —
.002.3. .7692. .0018 .3134 .0302 .9540 .0002 .3573
•3lsddar & 0tt er .0067 .0393 .0089 .3001 .0023 .3306 .0017 .2131.
rinec7 Organs
3od a 3is.aa. —.3032 .2377 —.0019 .1.753 —.0013 .62 .9 —.0025 .0090
ty i oaare a & .001.9 .0998 .3036 .C033 —.0025 .3532 —.3047 .0013
ia gassrco.a
3 1cip1a : .1oza —.002.3 .975 -.3026 .3239 .0005 .3074 .3113 .0973
T. ke.i 4 .0046 .2131 —.0082 .0035 —.0063 .2939 .0035 .i078
A1. eMa
*3.1 .0412 .0974 .0032. .3794 .0001 .9958 .0196 .2212
* o_ j4a4 ;— *Lue bassd on acandard :—cssc

7 3.* 18: !.d nacs4 ?*r c P .c 1oTin*Cica !ar.sgjo C .fficL.nu and Aaaoc .aM ?—rajuai from 0 .ig e.d
W.i hrnd 8agrmton An,Lys.a of 1a.1a. and o -iih1.ee T.nai.i and Sta—sp.eLf Caac.r
1s Licy 2ac V. *os S.lsc:ad 0. O8T*pbic Zisli E ac:ors fo ta.Li ane 3.oplaan or a3.it7, 1950-
1969, in 346 Cn&nci . -
3on—iihi:. Ma.L.. 3—v *i . T. a.1 .s
W.imcM .igf t.d
3a a.aion •Sagr,ission R.g ansi o m
3 5 p 5 p 5 p
E.apb..ua .0202 .11.92 .0023 .377 .0029 .6631. .0121 .0872
—.1020 .2387 —.0336 .3786 —.0093 .7703 —.0197 .2512
tax8 tne . .0984 .2104 .0239 .5762 .0050 .96f4 .0136 .7599
.001.7 .0835 .0083 .5964 .0127 .5231 .0052 .6941.
tivsi 5 —.6148 .6493 —.0032 .776 —.9173. .3982 —.0166 .1.766
3L1i .ry Pa.zs.
.0630 .1740 .0003 .0782 .30 0 .3125 .0035 .3461
Lary —.027.6 .2964 —.0028 .3096 .0130 .3950 .0003 .5933.
.3197 .3824 —.0013 .9775 —.0600 .4231 —.0.06 .1236
3r.ut .0021 .3610 .0009 .5605 .0738 .4459 .0323 .3290
hosca&a —.0162 .8627 —.0521 .2768 — — — —
.0555 .2350 .3196 .3786 .0049 .6948 .0021. .7123
31.add.r S 0th. .3204 .6391 .0066 .6872 .0231 .2035 .0140 .2399
U .xy Oranz
!odi i a Dis.a.. —.01.14 .2373 —.3060 .2933 .0027 .7061 .0076 .1337
3 .0046 .3276 .0133 .1 969 .0103 .7324 .0023 .393.3
9 . icu1osar oma - -
.1oma .0265 .2505 .0084 .3822 —.0090 .4717 —.0049 .5237
Lavkand.a 5 .0540 .2983 .0222 .3312 .0159 .7302 .01.49 .6302
A1 k a
All Mll an .0061 .9339 .1432 .3267 .1366 .5769 .0312 .3138
o—,id.d p — %Lu. aa.d n scandard t— ea

1, for white fe a.].es, nan—white ae.les and non-white
fe alas, whether percent prechicrination was significant
or snggesti e for any specif±c site depended on whether
unweighted or weighted regression analysis was used.,
However, for white ales, three of the fo*ir sites were
sigsificant under both nnveighted and weighted conditions.
2. with nnveighted regression procedures, percent
precblo .naticn showed weakly significant or significant
positive statistical association with the following cancer
sites: in white males, rectun (p.0 O), bladder & other-
urinary organs (p=+.000), ly phosarccna & reticulosarcoma
(-p .100) and all malignant neoplas s (p=.100); in non-
white sales, rectum (p=.08 ); in non—whitafemales, larynx
(p=.100). !o positive significant asscci t cn was found
in white females.
3,. with weighted regression procedures, percent
pr chlarination showed weakly significant or significant
positive statistical associations in the folloving sites:
white males, rectum (p+. 000), bladder and other urinary
organs p.f.0OO), and lymphosarccma & reticulosarcoma
(p= .O03): in white females, rectum (p=.005) and breast
(p=..01 2 ); in non—white ma.les, pancreas (p=. 07 8) and in
non-white females, esophagus {p.0e7).
the number of cancer sites that. were
significantly associated iith percent prec or natjou were
fewer in nunber than as the case with percant surface

water usaqe. he substitution of percent prechlorinaticn
for percent surface water did not add any new cancer sites
to those suspect with percent surface water usage.
3ectu , bladder, lymphosarco a & reticulcsarco a and, a..U
aai.ignant neoplazms in white males were positively
associated in the successive analyses with both percent
surface water usage and. percent prechlorinat1on . Eowever,
in white !alas, esophagus, large intestine, larynx and
trachea, bronchus & lung were not significantly associated
with the predictor variable (percent prechiorination) as
they were with the surface water regrassion series.
In white fanales, rectum and breast were positively
associated with both percent surface water usage nd
percent prechlorination when tested successively but with
the latter water variable, u.ltiple nyelc a was no longer
positively associated with the cancer site. yor non-white
ai.es and non—white fe:ales, the sane cancer s±tes showed
significance for both water -variables (non—white nales:
rectum, pancreas; non-’whitefe ales: esophagus, larynx).
5. the sixty—six regressions (unweighted and
weighted) evidenced a greater number of cancer sites as
s±guiflcant than could be ex ected by chance alone only
for white na.les and white fenales (weighted regressions).
6. only rectal cancer nortality showed any
consiztenc7 acrcss race—sal groups. As was also true fc:
percent su rface water usage, rectal cancer nortality was

significant for ail groups except non-white fe2ales in
regression analyses with percent prechicrination as the
water variable of interest.
The prcc dure used to select cancer sites and secic-
econoeic variables for percent surface water usage was
repeated to choose the reduced nodel vith percent
prechiorination as the water variable of interest.
Generally, overall weighted regression analyses on
the reduced 20d21 had little i2pact o the regression
cceffic .ents and their associated p—va.luc.s as seen with
the full cdel (table 19). The exceptions were ultLple
eyelc a in white fei ales (B:..0112,p=.10 0 4. !+. 000 ,p..207)
and esophagus in non-’ hite females (3:.O1 3 ,p..O87 3-
OO2,p 656)

Table 19: Es ’ ted Percent Prechiorination Regression Coefficients
and Associated P-values fron Weighted na1ysas (reduced
de1) of Selected Sex—race, Site—specific Cancer Mortality
Rates Versus Selected De ograph c Risk Pactors for Malignant
Neoplasn Mortality, 1950—1969, in 346 Coonties.
White White Non—white Non-white
Males es Males e a1es
Cancer Site P P
Esophagus —.0020 .6569
.0124 .0001 .0080 .0002 .0065 .7281
Pancreas —.0092 .7138
Breast .0249 .0001
Bladder & .0092 .0001
Other Urinary
Ly hosarccna & .0037 .0258
Reticulos arcona
Multiple .0012 .2072
All Malignant .0026 •.8934
*two_sided p—value based on standard t—test
inclusion criteria: p < .10 in full model, positive direction

Por reasons as suggested with percent surface water
usage, lar e intestine and rectal cancer 2orta.l ty rates
were combined and unwe qhted. and weighted regression
analyses of percent prechiorination were performed for all
sex—race groups. Only in white males was there a weakly
significant positive: association (p=.099) between the
water variable and the coabined cancer site (table 20)
When lung cancer mortality rates vera included as an
independent variable in the reduced, weighted regress ion
model, there were no malor change in the results. The
addition of the variable into the model did cause sorne
change in ly phosarc a& reticulcsarco:a (p=.0267p=.081)
for rhite nales (table 21).

Table 2 : T vei tcad and Waisftt.d !sgression a1ysss o Psrc.n: Prec 1 rinaci n and C tbined
Sites: LarZs I tesc a and Rsctu for All ace-s.ic G a*ps.
CA 1C StT ! _________
Whiti fa1. .0117 .2466 . .0l0 .09?4
his Pann Ia —.0010 .g245 .0fl94 .1U
:Ton Whit. aLa .0730 . 439 .03Z2 .547
: m Whit. Zana.1a —. 22S .8486 . i .16 .353w
*Two_sid.d p—value based a scandaxd t-t.st

Table 21: Estinated Percent Prechlorination Coefficients and P—values
for Cancer S ta by Race—sex for Weighted Regression Analyses
(reduced nodal) Including L g Cancer Mortality as a
Predictor Variable.
White Male
Bladder & Other
Urinary Organs
ty hosarcoi a &
Reticulos arcoma
LU lig .ant
Teop 1as s
Rec turn
Es ouhagus
p _value*
• 8323
White Paai.:
Non— hita Male
Non—white re a1e
* j _ jded p—value based on standard t-test

When selected organ sites were stratified by size of
COllnt7 pcpulation (<50,000, 50,000—250,000, >250.000)
there were differences in estimated regress ion
coefficients and their associated p—values from the
results when the study coun-t es were not stratified by
populatict size (table 22).
In .vh±te males,, the esophagus had net been positi Tely
associated with percent prechlor naticn. owever, when
the counties were stratified by population, esophagus was
zignif cantly associated (p=.0066) with the water variable
in the lowest population stratum (<50,000). also, the
estimated regression coefficients had a gradient going
from a negative sign in the largest population stratum to
a positive sign in the si allest stratum. L±kewise,
laryngeal cancer mortality now was positively azsoc ated
in those counties w thpopulat ons of <50,000 (p=..015).
in estimated regression coefficient gradient was present
going from a positi’7e in the s a.Uest sized counties to a
negative in the >250,000 population stratum.. Kectum,
which had bean positively associated with percent
prechi.orinatiofl (unve!qbted and weighted regressicn
analyses), showed no statistical asscciation with the
water variable with stratification by study county
pcpnlat .cn siZe. :n bladder, estimated grassicr.
coefficients increased in sine from 3:_002-,B:.018 with a
significant po .t7e azsoc .3 .Ct bet’ean the s te and the

water variable for counties with populations of 50,000—
250,000 (p=.055) and >250,000 (p=.022),.. AU. 2ali ant
neoplas s which had been weakly significant with
unweightéd regression (p= . .097) did notdenonstrate any
association when the study counties were stratified by
popui.atjon. ifowever, the size of the regression
ccefficjents did decrease in size fron =.051 in the
saallezt stratu to 3:—..009 in the largest population
In white fe .les multiple yelo a and rectun were no
longer positively associated with cent prechlorinat on
when stratjf ed by population though the latter did show
estinated regression coefficients increasing in size
across the strata in a positive diecticn . !reast cancer
iortality, previously associated with weighted arzalysis ,
was now sig nificantly associated with the water variable
in counties having a population of 50,000-250,000.
There was no consistency between unstratified and
stratified resu.Lts in non—white males and fe a.les...
neither rectum nor pancreas in non—white a.ales nor
esophagus and laryni in non—white fenales shoved any
association with peroent prechiorination in any of the
population strata..

Table 22: Estimated Percent Pr hiorination Coefficients with P—values*
from Weighted Regression analyses Stratified by Population
for 346 Study Counties.
<50,000 2.50,000 >250,000
Sex-race Cancer Site p—value p—value p—value
White Males
Esophagus. .0096 .0066 .0036 .3225 —.0087 .0918
Large Intestine .0032 .7349 .0098 .3791 -.0006 .9896
Rectum .0079 .1271 .0082 .2700 .0150 .4346
I.axy z —.0002 .9410 .0009 .7622 —.0050 .3300
Trachea, .0442 .0148 .0102 .6869 .0004 .9892
Bronchus & Lu ;
Bladder & Other .0024 .3586 .0110 .0551 .0183 .0219
Ur{nary Organs
All Ma1ig nt .0512 .1890 .0269 .5921 —.0087 .9413
Neup 1as
White Fe a1es
Rectum .0023 .4616 .0107 .41.51 .0213 .2317
Breast . 3l7 .1110 .0200 .0582 .0220 .3756
1tip1a ye1ona —.0017 .4446 .0018 .3605 —.0005 .6770
Non—white a1es
.0149 .6715 .0097 .6312 .0076 .6437
Pancreas —.0307 - .6480 —.0752 .1082 —.0064 .6079
Non—white ?emales
Esophagus ‘.0254 .0014 —.0291 .0592 .0013 .8584
Larynx .0024 .7 41 .0041 .1889 —.0016 . 531
* J s 4 dad —val e based on standard t-tast
<50,300 (n 273); 50,000—250,000 ( 62); >250,000 (n II)

1qai , for the purpose of i creasing precision, o 1y
those ,co ties with 50 per cent or more known iater source
vere selected for further analyses. these counties were
stratified into three population stratum dependinq on sie
and weighted reqIes .cn analyses were done (table 23).

Ta Le 23: Escimacad Per e.nt Prachiorination Coefficients with p—values
from Weighted Regression Analyses Stratified by Populat±onf
far Co tias with 50 Percent or More nown Water Source.
<50,000 230,000 >230,000
Sex-race C car Site p value* p—value p—value
White Male
Esophagus .0109 .0203 .0029 .5015 —.0093 .0889
Large Intestine . .0008 .9433 .0134 .2699 .0185 .0855
Rec u .0083 .2747 .0067 .4122 .0190 .4736
Lary .0007 .8341 .0004 .9065 —.0048 .2371
Trachea, .04.48 .0839. .0099 .,258 .0066 .8987
3ronchus & Lung
Bladder & Other .0012 .8388 .0086 .0872 .0107 .2376
Urinary Organs
All Mali ant .0616 .0547 .0236 .6366 .0164 .6418
Neop 1as
tThite Pe a1a
Rectun .0059 .3180 .0064 .2044 .0103- .1667
3reast .0189 .0896 .0152 .2079 .0102 .5303
Multiple Nyelona —.0048 .0974 .0005 .7991 —.0001 .9347
Tan—’ jhite Male
Recti. .0161 .8588 .0159 .7433 .0024 .3549
Pancreas —.1309 .2330 —.0900 .0974 —.0089 .3546
on— hita ta
Esophagus .0306 .0453 —.0333 .0734 —.0021 .6222
La —.0159 .2061 .0018 .3097 —.0039 .1243
* 4 dad —va1ue as ad c standard t—tas z
<5o,o0O ( — 4); 30,000—150,000 Cc — 43); >150,000 Cc 10)

In white sales, reduction of the number of study
counties resulted in large intestine shoving a eakly
signi.ficant positive association (p=. 086) with percent
prechloriratjon in the larqest population stratun
(>250,000). Large intestine had not previously shown any
positive association with the water variable. trachea,
bronchus & l nq went froa a significant association in the
smallest populat cn stratus to a weakly significant
association in the stratun (p.015- p=.08 $) when the
nusber of counties were reduced.. For bladder, with -
reduction of -county nurnbers, the site was now associated
with the ‘ater variable only for the stratum (50,000—
250,000, p= .D87). The association with those counties with
population-s larger than 250,000 no longer held.. Feduction
of county numbers for all. alIgu ant neoplasms did result
in the sIte now shoving a positive association of p=.055
in the lowest population ztratu (<50,000)..
In white females, results from analyses whether with
the full co plement of countIes or the reduced number were
consistent. Eovever, breast showed a shift In population
stratus.. While under the f nil coapleient of counties the
site was assocIated with percent prechiorinatioc in the
21&dle stratum (50,000—250,000), when the number of
counties were red iced. the asscciation was f nnd i the
sailest strat n (<50,000).

cElP 4 r!B ? 17

i ncn—vhjta a1es, .o pcsiti e azzcciatio ias found
bet een the zite and the vater v ria 1e in any strata .. In
tton—vhit females, a pcsiti e azsociation between
esophagus and percent p:echlo:ination was again observed
in those counties vith the smallest pop zlaticns.

In assessing the Lnfo ative value of the sti dy’s
results, two major problans inherent in an ecological
study have to be evaluated: - the quality of the data and
the adequacy of the statistical methods
The series of unwq .qhtad and eiqhted reg.rassicn
analyses and secondary analyses resulted in a nu bar of
siqn ficant and sug 3stive asscciatjons between percent
surface water usage (public drinking water supplies) and
site—s ec fjc cancer ortality rates (1 5O-1 69) as
re crted for the 3L 6 study ccunties located in the Ohio
Piver Yalley 3asin This was also true when the water
va able, perce t surface ats.r usage, was replaced by the
siicond variable of int :est, pencent p:echlonination
Eoveve:; while the nurnbsr o signt icant associations
exceeded the number expected by chance alone (for weighted
reqression with percent surface water isaqe) soie positive
sites were questionable in terrns of plausibilit7, i.e,
trachea, bronchus lung, 17 phosa:co a and
:et..cuiosarcc a and nult p1e 7elc a.
?zvicus ex; nta. r s azc , epi e cloqica2. data
and ClCC CZ u .-.t7 ZCC i 5GC i tiai att t r. c:

five specific sites: livers kidney, bladder, large
intestine and :ectu .
Liver and kidney showed no significant association
with either unweiqht d or weightzd regression procedures
icr any race—sex group. When kidney was grouped wi’nh
bladder and other iirinary organs, all race—sex groups
except white males showed no statistically s qnificant
associations with respect to both pp rc nt surface water
usage and percent prechiorinatict.
!ladd.er was. highly siqnif cant in white sales for
both percent surface water usage ( :.O06,p.Ol3) and
percent pchi.orinaticn (B:.009,p=+.000) when weighted
regression procedures were used... On stratification by
population si:e (3 groups), bladder cancer ncrtality was
positively as3cc ated with both percent surface water
usage and percent pr,chlorinatiot for counties with
pcpulaticns of 50,000—250,000 (3:.OIS,p+..000 and
3:..O11,p=.0551!. n addit on, bladder was also associated
wjth the latter ‘dater variable in counties with
populations over 250,000 (!:.018.p . 0 29).
estrtctinq the analises to counties with 50 percent
or gore kncwn water source did not appreciably change the
results. or both percent surface water usage and percent
prachlorinatiCfl, the asscc at.On t’wee each of the water
variables and bladder •ja fc .ind n those stud7 ccun es
b .avir.q ;c u .latio: sizes of 5O,30C2 0,O00.

These findinqs a:e not unique ecent
epidesicloqical studies have also shown a positive
aszoc atncn between bladder cancer and a watar variable
thotiqh results are not totally consistent across sex and
race. Pace and Harris reported siquificant regression
ccefftcients for drinking water p<.05) frog the
iz3issippi Piver for uhit males and non vhite females
(1975); !uncher (1976) for white nales in counties using
the Chic Rtver as thein source of drinking water; and
acqan €t al ,(1976) found an association .b tween
chlcrcfor levels and bladder cancer c:tality which was
weakly suggestive for white females in the egicn V survey
(Illinois, :ndiana, !ichigaan, !innasota, Ohio, and
Wisconsin) and suggestive or significant for both sexes in
tha Nat cnal Organics !eccnr.aisance Etudy. Cantor et
al., (1977) looked at the correlatiOn - between cancer
3ortality rates, by site andsex in whites with leyels of
ha lc enated eethanes in drin .ing water supplies after
ccrrect nq the rates for known or suspected risk factors.
9l dder cancer was posit vel7 correlated in females wtth
levels of ncn—chlcrofcrn t ihalcnethaneS in three :eq ons:
north, west and scuth. In xales, the cc elation was
strong in the northern :ecion an eszent.aliY :src in the
other regions.
eqa’ ’e5S o whathet perCent urfacs t r usage on
:ercent ;rechlcninati i. ‘ a3 useC a tn at9n 7arna .9 c

interest, each of these water f .ctors were ocst
consistently associat d with large intestine-rectal cancer
mortality. Considering percent surface water usage, in
white males cancer of the large intastine was weakly
asso ated with the w .ter variable wh3n using the weighted
regression for the full 31€ connties and after
stratif caticn by population, (B:. 067,p=.07 ) for
counties with populations exceeding 250,000. educticn of
the number of co mties considered (50 percent or sore
known water source) again resulted in counties with the
greatest pcpu.lations showing a significant association
between percent surface water usageand intestinal cancer
orta2.ity rates (3:.OeO,p=..023).
Although large intestine cancer ncrtality rates
showed neither statistically significant association with
;ercent prechlorinaticn using the full coaple ent of study
counties, ncr an associatio affer stratification of the
entire study s,t by pcpulation, reduction of the nunber of
counties analyzed (50 percent or acre : noin water source)
d d show a significa:t relationship between the water
variable and the cancar site in the cst heavily populated
counties. Also, the :eqression coefficients for both
percent surface water and percent prechlorination increase
in sire as one cves across tratu : lowest population
o h qbest ccpulaticn size. A ain, previcus stzdjez
lend suprort to a possibl link betwaan large nes ine

cancer mortality and the water var ab1e.. Page and Harris
(197!) and 3uncher’s (1976) investigations resulted in
significant results for white feaa.les. Hogan’s study
(1976) uzinq the OP5 data showed a siqnif cant increase
for white fanales Also the site specific cancer
iortal ty rats for - white males was significant when
weighted regression procedures were uSadm sqicn V data
showed siqniftcance fcr white zales when weighting was
used in the mere heavily populated counties.
Pectal cancer was the site 3ost consistently
associated with the water variable in this study:
significant for both whits oales and females (weiqhted
regression) and non-white nales (unw iqhtad) when percent
surface water usage was included in the model. with
percent prechlor at on as the water variable of interest,
rectal cancer mortality was positively associated with
white males (unweighted and us ghtad), white fernales
(wa qhtad) and non—white males (unweightsd).
Stratif±caticn by county population size showed sc e
consistent patterns and in whites implicated the cra
heavily pcpula.ted. ccunties(>2 50 , 000 ). !cr either cf the
water variables in white males and white fa ales,
est ated : zion c i ientS inca as pcpulaticn
stratun size i: eaEed. Ic: white fenales the cancer site
was weakly sscciatcd with percent su Ce wa a: usage in
the cp aticn stratum >250,000 when both the full

complement of study cc’ nties were a 1yzad (B.02 0 ,p=.0S1)
and when the number of count±es wer reduced to increase
pr3c sicn (B:. 02S,p= 069).
ks there is with bladder, so is there some
consistency in linkinc r cta2. cancer mortality and surface
vater as a result of associations found in various
studies. Eoqan et aL.,(197E) in the eqion V Survey,
fcund an assoc attcn between ctal cancer crtality and
chioroforn leve’s when a weighted analysis was used,
thouqh rest ictad to white alez when counties were acre
heav i populated.- ectal cancer was statistica.ly
sjqnif±cantly associated with chioroforu levels ( FS
data) for whIte sales and females. owever, results did
not seen to vary accordinq to population stratu3. In
Cantor’s study (1 77) correlations with lcq—tota.l
trihaio athanes of residual values for male and fenale
colon cancer rates net study criteria. owever, these
statistically significant asscciaticns ‘vera elj inated b7
‘he •jnclusicn of 10 foreign stock predictor variables in
the regression nodal. -
pjdenjcloqicallY, colorsctal canca: exhibits a
north—south, urban—rural qradient (!e:g,1 7 ), thus
suppcrtinc in part, data f:cn this study. Eowever, while
the results f: n t i and other jnv s iqaticns p7icusl7
noted sugceet a link hetween the water ia:i ble and
colorectal cancer, there a:9 ee ea1 qua1i7 nc factors,

sc e of which aay be qeneraljzed to other sites.
Because of the possibility of jsdiagncsjs,
:isclassification and cther sources data, bias favorable
to rectal cancer deaths over colonic cancer deaths is
present on death certif caticn (9.rq,19711). This bias
increases as the number of true colcaic cancers are found
and is greater for wc en than for man. mother factor
important to the interpretation of colorectal data is nct
only how the daf iticn of colon and rectal cancer is
stated but the definition itself is sub act tc change over
time. With regard to different definitions of rectal and
colonic cancer, the area of ambiguity is from 3 cm.to 19
cm, from the teni nation of the bowel with one—tweith of
afl bowel cancers f linq into this region of uncertainty
Those beyond 8 cn are epidimiolcg call7 related to colon
cancer. Consequently, while izcide:ca rates for the cclcn
and rectum are highly ±ntercorrelated and. whIle
etiological factors are common to both colon and rectal
cancer, there is a second set of rectal cancers of
different etioloqy. While combining large jntistjne and
rectal cancers mininied the prcbab.l2.ty o: oiases, there
are still substantial differences among populations.
nother conside: .tiOn is that the isolation f
i:di7idual ects o pcsure vania las ‘with the ti 1

qress cn approach i an ncertaj: proc ss. !stinatin
reqression coefficients and evaluating statistical
significance for individual variables ay depend solely on
the other variables included in the equatton.
Illustratively, for cclorectal cancers, Eerg and 3urbank
(197 ) 1 ave reported that many, if not all, aetals are
potential cocarcinocens being capable of depressing the
en ynic activitiqs involved in the netabcis of organic
carcinogens. Their study showed a significant positive
correlation between intestinal cancer and average cad iuii
and lead levels . The results showed, at best, only an
association between the independent and dependent
variables without ccnsideration of the heavy ietals in the
cdsl . . Also, cancer of the colon has been found to be
highly correlated with ndicatonicf affluence, i.e.., high
fat• diet rich in anisal protein and specifically with beef
consusptjon {3erq ,197 &;Dressar md :rvinq,1973). Cther
vartables inplicated in b wel cancer etiology have been
alcohol (s aU iztest ne: a.les and fe a2 .es); cigarettes
(colon: females). alcohol, ciqaratte consumption and
beer-dr±nkinq data wer nct included in the aodal because
no data were available to character .ze county a ffe:ences
a onq the variables..
Breast cancer in white fe al s hcwed a ieak.y
sianificant asscciatiC: with ;errent iater usage
(!:.O1O, . . TQO) and a significant aesociation ijt . pa:cs t

prechiorination (B:..013,p=.012) e Eowever, when stratified
by population size he results were variable.
Stratification of a.Ll 3 counties resulted in a positive
assoctaticn in the most h av ly populated counties
(3: • 051 ,p=. 009) with percent sur face water usage as the
water variable of Interest. 3ut, when stratification on
countIes with 50 percent or acre known water source was
carried cut,, no association was found for any of the
respoctive strata.. When precent prechiorluation was
analyzed as the water varIable of interest, breast cancer
3ortal ty showed a weakly siqnificant association in the
population strati m, 50,000 —250,000 !: . .020,p.05S) when
3i 6 counties were considered.. Peduction of the number of
study counties analyzad showed breast cancer mortality
weakly associated with percent prechiorination in the
lowest population stratum, <50,000 ( :.019,p.090). The
validity of th.is findinq is questionable. Aside from thc
possibili T of a sj nificant result occurring from chance
alone, breast cancer is known to be correlated with colon
cancer; it Is associated also v th beer drinking’ and with
fat and animal prota ccnsumpticn. he last is possibly
ex,lained by greater trcgen synthesis both by the body
and the ineestinal flora when a rich diet is consu ed
Thus beer drirkir , alcohol. ci arett9 s okiug, ±at and
animal protain i.nt ks, colon cancer , br aat ca:c : and
bladder cancar forn a set 0± hi hly correlated variables

whose effects are diff±cnlt to isolate and as potential
confoundars, were not controlled in the present stud’ .
with aultiple regression technjqu s, it is difficult to
adequately isolate the specific effects of intarcorrelated
exposure variables -
!sophageal cancer mortality rates were s qnificantly
associated with percent surface water usage for both white
males fp. ,OLL1) and non—white females (p=.050).
Stratification by population resulted in similar patterns
for both groups.. whether the full complement of c.ounties
or the reduced number were used, estimated regression
coefficients increased as county population size
Lncreased. ?or non-whLte females. the cancer site and
water iariable was po tively associated. in the largest
populat cn stratum (>250,000) for both the full number
(3:.021,p.0O ) and the r ducsd number (E:..022,p_O10) of
counties. The guestion may be r i.sed as to the validity
of this finding. In white nales, esophageal cancer
mortality is correlated with lunc, larynx, colon, rectum
and bladder mortality forming an urban zation” complex.
sophaqeal and. bladder mortality are also correlated with
smoking but to a lesser degree than with urbanization. :n
white males, a significant a . sociatiOr. was found between
er ent surface watqr usage and all of the highly
correlated cancer sites C prising the “ur anizatiCn-
smoking’ complex. a. c acl wh c: s c

predispcs tc eso haqeal cazc and thus may possibly be a
confounding factor in this study, was not controlled. I
similar problem exists in interpreting significant results
obtained for laryugeal , and trachea, bronchus and lung
data: aside on chance, the urbanization complex,
possibly a confounding factor in this study was net
controlled. flso, the biological plausibility of surface
water qualtty as a :is.k factor fcr t. e mentioned cancer
sites is not clear. !owever, the lung is the primary
excretory organ for a number of the volatile halo.genated
compounds. The above cancer sites did not shcw
consistency across race-set groupings. Given that the
data available on non-whites are not reliable, ‘why only
white males and not white females also? Possibly they
have different exposure to the water variable and/cr to
other factors that may have an interactive/synerqjstjc
effect; or there may be a differing biolcg cal rasponse to
the agents in question. -.
wo other canccr sites were significant: multiple
myelcma in white females and lymphcsarccma &
reticulosarccra in white males. Cancer mortality : tes
for both sites were significantly associated with pencant
precblo:i:atic. Cancer mortality rates for multiple
myelcma and lymphosarcoma & : ticulc arccma ie:e
significant with percer t u:face u aqe ind3r both
g ted. ar. Jeiqrt3d :e i analy E ca ce

rate for zltiple iyelo a was vea3cly associated with
percent prechicrination with weiqhted analysis while the
cancer rate for lyaphcsarco a & reticulosarcoma was weakly
associated with percent prachlor nat±on under unweighted
cond .tions and strongly asscciated ‘with percent
prechiorinaticu vtth weighted analysis. These results
v re not expected ..nd cannot be explained at present.
acvev r, since they were relatively persistent, it sse s
wise to include the two sites in further studies.
3riefly, the results f:c this study and ether
studies previously mentioned, show significant
asscciat onS between percent surface water and/c:
p:echlor nation practiceS for sc e cancer sites.. while
causality cannot be deterained, an association is present
and further studies with part±cular e phasison intestine,
rectum and bladder are warranted. owever, interpretaticn
of results is made difficult because of p:cbla s already
e.ntioned as well as additional prcble s co cn to the
type f data available.
he question of intracounty variability was not
addressed:— even greatly ho o eniouS regions include
i t —reqicn yariabiity. I question a.y be raised as to
the degree to which areal traits characterize each of the
elements comp:iSin the :e iCn. :: Cu: unit of stud.7, the
county, :e a:dlesS of in ercCU t7 size vazia ility,
contains a numbe: of elated factors. Pach of these
— * —

factors, not totally indape .dent of th other, nay vary
from point to point in the area in a number of ways or ay
vay for su.bpcpulations within the county. however, to
have some understandinq of the phenomena in a specific
place (i.e., county), £t is necessary to ignore these
var aticus within the unit of study and proceed on the
basis of arsal hcmc eneity. he extant to which the
assumption of ntraccunty homoqene±ty is violated and the
posstble effect on the conclusions drawn from the data are
not known.
1c ain, in ecological s-tudias, the sampling anit is a
population or group ratter than an individual. Since
ascer tainment of individual risk as determined by
individua.l exposure is central to many epidamioligical
studies, ecological s’ adLes could yield individual risk
only conditional on. each individual within a population
having the same exposure. !oweve , in actuality, there is
a heterogeneity of exposure. Consequently, the
relationship based on average exposure need not reflect
the exposure of any individual; populations with identical
average exposures may differ markedly in the distribution
of exposure levels nd, ignorinq the variation n
exposures results in a loss or the deta led ..nfor at.cn
necessary to zo o t thE effects of different
envircr.:enta.l aqents..

Another problem, specific to this present study was
the use of mcrtaljty data rather than incidence data; this
use could prcduca spurious correlations if mortality data
does not reflect incidence in all areas. lisa, an issue
is the adjustment of mortality (or exposure as may be the
case) for other etiolc ic aqents:—for most cancers many
environmental and non— environmental factors are known or
suspected to be etioloqically significant. Adjustment for
such variables in eco.loq±cal studies is ha pe.rad by the
hiqh deqree of ccnfoundinq, by the fe study units
availa 1e for analysis, and by use of data with different
frames of reference with respect to time, place, or
persons. for example: data availability may be the
deciding zsue in determining th unit of study as was the
situation in this study. ata collected for purposes
other than the specific intent of the researcher may
differ not only in the cbaracteri tic of inter st but also
in other characteristics; these data aay introduce
confounding; other data items may have to be aggregated so
that the data set for the unit of study may possess
properties that are not matched on individual items in the
population. and, different sets c: neede ata are often
available with differing eç:ees of areal br a dow:.
often then, the situation arises that as the research
design increases in conp1e it7, the probability decreases
that the individual data sets’ w .ll have suz: c er .y

detailed information on the variable of interest..
Of special importance in cancer studies is the latent
period between a carcjno enjc stimulus and the development
of the disease. The water data are for 1960. Given a
general latency period for cancer of 10 to 30 years, the
critical time factor is not adequately taken into account.
There is in addition the difficulty of temporal
variability even within the jndiv daal data sets. The
pattern of areal variability may have, and most probably
has. underçon chanqe percent snr aca water usage itself
has not remained stable nor have the quality and
constituents of the water rernained constant. The unit of
study, i.e., county, and/or individual data sets may have
undergone administrative changes or chanqes in definition.
or example, the source of water for each community in
this study has not remained static over the 20 year period
for which the cancer rates wire calculated and do not
reflect earlier (or later) levels ev3n in a relative
sense. use, the data were not collected in a
standardized manner over differa:t t :e periods, thus
peventinq examination cf time trends. Cther questions
concern temporal variability with respect to iqratory
patterns .it is not known how long the individual resided
in the area in which the death was renorted.. Tao, cancer
in some :.cns nay be caused by agents having no
:elati3nship to any cf t e gza:ti:i€s neasu:ed.

Statistically, sultipj.e regression permits a rather
sjip listjc type of ad1ust ent for the s ltiple var ab1as.
Ccrrelat on and regression tests are oit3n dependent cn
the areai. unit chosen,. s Pobinson de cnstrated (1950),
cc relatjcn generally increases with the aggregation of
data into larger units.. Size of correlations, and even
their direction, say change with a change in the unit of
s was discussed earlier, with any hypothesis
eneratjn prccedur s, the probability of Obtaining
significant assoc atians by chance alone are hiqh.
Turthersore, the nunber of sigrificant results in this
study was strongly dc endent on whether or not we ghtinq
was used Sose possible explanations for this occurrence
have been advanced by Eoaan et al . (1976). Weighted
analysis aay tend to give undue esphasis to the nost
populated counties.. However, further testing de onstrata
that the difference betwefin unweighted and eiqhted
analysis can not be attrIbuted solely to the effect of the
scat heavily populated counties.. Given the large
variability i the weights used, the correlation patterns
a onq the var a bles icladed i the nndenly g cdel qht
have been significantly altered by the qht nq factor,
thereby cha gina the degree of cc1 iea ±t7 present in the

Thus, given the nature of ecolcq cal observations,
the choice anon alternative in erpretaticns can not be
‘ade solely on the basis of the observational data. The
choice of interpretation also depends en the assumptions
ada by the researcher and Instified by analogy or the
quality of infozmat on available.
, thou gh
Two other points are i portan ‘ ‘
pravtotis epid iolcgical studies (Page & a:ris,1975;
3unchar,197 ), eiperimental data ( schenbrunmer 8
iller,1945), and theoretical expectatjons inplicate sites
which were not found to be significant in this study,
kidney and liv ra The finding of no association is
not considered suf ic ent eviden for reiectinq an
associatic bet’aeefl th predictor variable and kidney and
liver site—specific c:tality. Secondly, given t .at a
positive statistical association b tween surface water
‘asaqe and sIte—specifiC ortalit7 is not spurious, an
association is not quivalant to a causal relationship.
lnexplanatiofl does not consist er ly in suggestIng the
factors involved in, or even the general fcr of a
relationship: It must also include an estinate of the
actual paraueterz of an enpiical relationship and. a
de cnstrat.Cn that these paraneterS satisfac ri17 account
for the v rjaticn in bcdy of act .ia1 d ta SpecificitY
is :egui:ed. I z- ati tiCall7 siqnifica:t association can

cn. .7 sugges....
Lud yet, associations derived f:o ecological studies
have ep de ioloqjca2. val ie by provtd ng a basis f or
further rezearcb, and understanding o the factors involved
that ay isad to hypothesis testing studies regardless of
the aany limitations and interpretative difficulties.
In conclustcn, results ind.icate the relationship of
cancer of the large intestin3, rectun and bladder with use
of surf ace water and/or prachiorination practices warrants
f ther research. The preliminary results do not .p:ovide
a basis for drawinq conclusions concernin causal


2abl,u 21 $t ni tcanC c .ff4 j- .’ .. i d p.. alu.a* of v.j$hC.d 2. rua.io. 2yIia ror ?.rg.nc
Sc. W.c.r Usa. a a .S ScL.ct.4 Soc Vutabtts O sn 544. p,ci ic for b44e taJ. .s.
i( a r .op1a ortai.iy, 1930-1969, in 346 S 4 Co.. ’4- .
!a 1.y’i4 ?o9 s2a—
i ’- - . . I in s -ez - :
‘ r Sorfac. 3o — YA4. in A ri- a 3.. a- Squirt 1ot.tg in f*ui-
Vicar btta 1.n tiui i1car . fac zrtn t1. SCock faccuxta
.009 .o715
4.05 4.01.
—.3390 —.0012 .3863
4.10 4.01. 4.01.
.0008 .0334 .08*1. .0424
‘.05 4.01 ‘.05 4.10
.0001 .3244 .1 331.
4.01. 4.Q5 4.01.
LLv .4247
.0252 —.0355
4.02. 4.05
—.0001. .002.6 .0092
‘.10 c.05 ‘.023
.0029 —.0039 .0018 .1628
3ro- es ‘.01 ‘.02. 4.10 ‘.01 4.02.
Prsscaz. .001 0 —.0238
11*ddtr6 .0005 .0536 .0174
0 sr Ott—
—.3003 .0096
.01 ‘.05
.3062 .0004
6 4J .0 4.025
*.ticn le-
s .rc
J.ttpLm .0002 -
‘.3 .5
.0032 .0168
4.10 .10
.0179 .:63.5 .96L2
.10 ‘.01
p..4&L *S tied n standard :-CSIC

.ab1a 25: 5Li ificaan !. wiQa C .f!i ients and p raLu.I* 01 isiibtsd !.8rtamio Ma .Lyiis f r Pt c e
Surfac. Watan Uza . and 5.1.c ed Socio—.conc in Varta LlI Orpa Si .ci1i f r it .
7 1—, a1.i nsnc I.o i.a .sa or aUty, 2950—1969, in 346 S ndy C.,unci.s.
# 1oy .4 Popula—
_____ $ p1nyad in Snu-dur- tioc Pan Orban
‘ ——— — Sunfac. o — Madian !dura— in t— abi. Maai— Squax . Popu— ‘arsi in
5t. Wat.z bits Inco .. ian ci4n rt facurmn Mil. 1a ion 5 cc fac urin
1.0041 .0295
<.10 <.32.
—.6672. —.0004 . 1.626
‘.01 ‘.3.0 ‘.01.
.2391 .0670
<.03 <.31
Ltv.r —.3873 —.0006
‘.10 ‘.325
Lary —. 0001 —.0069
<.01 <.31
.0007 —.3575 —.0003 .0024 —.0330
.01 ‘.05 <.10 ‘. 3.0 <.0 1.
.001.0 1.1.571 .0010
‘.323 ‘.01. <.01.
31 .ddar 6
Othan Ort—
nary 0r8an*
od ki .0001
1 L9 .01.Lt.
J .cL 1.. .0036 31 <35
ii—— ’ —
7 — 00” .1254 .3903
<01 <.05 <.01 <.325
mann t.o—
* 0 — .j .d a *V : - .$

1 (0
Table 26: 5i ifi anc 3a*rssston Co.Ui i t.. and p.v*Jjaaa* f i.i td ?aç .ssia MaL7ata f r
Surface W*c.r aage and SeLected Socio-eco E.c Varlablee Organ St sSp.cific for ao-vbits
91.aza ortaLiy, 1950—L969, in 346 Stu4 Counti.s.
• yed Po ,.L.— T
in ie urban Z : !npLaynd
Cancer Surf ace 3oa- “ dian Educe- in L. ri- abLe i— Square !op a— 7or.in in aan-
Sit. W .t*t Thi:s Iac . eion cuiu . factirfa iii. lacioe Stock factu .n
i.ct . 036 .1.969 .02 1 .8
4.1 .0 .01.
?ancr.la .1.1.45
.3532 —.3957
51ad ’ 8
other Ur .—
tat, Orlan.
________ S
3724 .;319
£LL a.Lig— __ •,t <.03
tent : —
,o—sidM p— ni a a .ec sta24ar i :- eit

Table 27; Significenc Reres io Costftci t .s and p—vaLu.a f 3nuai cad sErsasiøe AnaL .in for ?erc.at
rf*e. Wat.t Usa , and Selected Socio—econo ic Vsrieb1 Organ Sit. -Specif in for Mow-i,hits
la.1.s, MaUgnaz*c ‘.opLaa. MortaLity, 1950—1969, in 346 Study Counties.
S !ploy.d PopuLa— 2
Z 2 Median S Enploy.d in Non-dut— tion ?.r Orban 2 2 !nployed
Cancer Surface son— Median duca— in Agri— able Mann— Squat. Popu— Por.ign in Menu-
Sit . atsr Vain, Intone Uon cultur. fatatiu$ Mile 1atia Stod i f*cturing
c .at
Sconecb .0311 .0026
<.10 <.3.0
Liver — 1.8068 .0677
Otber Uti-
nery Orgsne
Ly boear —
c 4
a1nip1a .0012 .0008
<.025 <.3.0
0012 —.0332 .1404 .5432 .2232
<.10 <.023 <.05 <.05 <.10
&L1 !alig-
oant Neo—
*tvo gided p—value b*aed n standard ;— eet

Table 2$: S1 nifie.ant .grssaiou Co.fficieuta and p —valusa* f .ig t.d ft.gr...ion Melyci.. for Percint
Surface ‘ ist.r Usa a and Selected Socic—econosic Variables, 0r an Site—Specific fot Whit. Moles,
Malignant N.opiaza Mortality, 1950—1969, in 346 Study Counties.
S !aploy .d Po7uLe— 2
2 2 Median S Enplnyed in Mon-dur— tion per Urban 2 2 Enployed
Surface Mo.— fl.dia Eduea— in Agri- able Man.— square Popu— For.ign in Mine—
Sit. Vater Whit. Incoes tion culture facturing eu. lation Stock facturing
!.opba$us .0039 .0003 12106 .0010 .0141 .0653
<.03 < .01. <.025 <.01 <.01 <.01
—.4963 —.0007 .4020
<.03 <.01 <.01
Large .0106 .0006 —.0001 .0034 .0211 .0464 .0882
t .uadne <.10 c.10 <.10 <.01 <.03 <.05 <.01
*ectos .0079 .0008 • —.0003 .0017 —.03.74 .0773 .0356
<.023 <.01 <.023 - <.01 4.01. <.01 <.01.
lii .tar7 —.3228 —.0003
6 <.025 <.01
—.0004 .0421 —.0354
<.01 <.01 <.01
.0042 00”t —.2238 —.0001 .0014 .0139 .0085
‘.01 <.10 ‘.01 <.01. <.01 <.01 4.10
Trachea, .0234 .0038 —.1.1210 —.0037 .0002 .1335 .1419
3ros s & ‘.023 <.01 <.025 ‘.01 4.10 <.01 <.01
t o n i
IrasaC .0009
.0009 —.0001 .0021 —.0206 .0331
<.01 ‘.10 ‘.05 <.10 <.10
—.0001 .0071
- ‘.10 ‘.1.0
31add .r 6 .0062 .0008 #10000 .0503
Other Un— ‘.025 ‘.01 ‘.05 <.01
ta y Ot ani - -
Lyapbosar .0038 .0005 —.1.794
& lati— <.10 ‘.01. ‘.10
Multiple .1769 —.0004
‘.10 ‘.05
#.0000 -.0011
Leakasia 6 <.10 <.03
AU Malig— .0754 .0098 3.7716 .0057 .0096 .2360 .7683
ant3e <.01. .01. <.01 <.01. <.023 <.01 <.01
?Wo-4i4ed p— ’ a1u* a.sed ou standard t cast.

Table 29: Significant Regression Coeffici.nts and p—valu.s* of Weighted Re, rssaion nsly.t* for Percent
Surface Water Usage and Selected Socio—,conoaic Variables, Organ Site—Specific for .hir.
Y ales, Halignant eopLasa Hortality, 1950—1969, in 346 S rudy Counties.
• Epl.oyed Popula—
2 2 Hedian S Eaployed in Non-dur— tion per Urban 2 2 Eaployed
Cancer Surface Non— (.dian Educe- in Agri— able Mann— square Popu— Foreign La Menu—
Sit. Water Whit. m a cne . tine culture facturing tile lation Stock facturiD4
—.0710 .0050 —.0050
<.10 <.01 <.10
Stoe.ch —.0004 —.3688 —.0003 .1380
<.05 4.01 <.05 4.01
tdg. .0007 —.0001 .0036 —.0176 .0817
Intestine <.023 ‘.03 <.01 4.10 4.01
.0047 .0003 .2690 — .0171 .0326 .0232
<.01 <.10 <.023 <.01 <.01 4.01.
3iliary —.3340 —. 0004 .0310
Passages 4.05 <.01 ‘.023
4 Liver
Pancreas -
Laryan — .0001 —.0045
<.01 <.01
2rachua, .0005 -. 2315 -. 0004 -.0000 .0011. -.0227 -.0212
Iroac s & ‘.01 <.10 <.01 <.10 ‘.01 4.025 <.02.5
Ireast .0093 .0011 .7798 .0005 .0507
‘.10 <.01 <.01 <.05 4.01
8ladder 8
Other On— -
nary Organe
godgkin e .0001 —.0000 .0005
‘.10 <.01 ‘.01
Lynpbnsar- .0002 .0002 — .0007 .0075
c & Rati— <.025 <.01 ‘.025 4.025
Multiple .0020 .1003 . +.0000 - .0004 .0065
Hyelo <.10 <.05 <.05 - <.10 <.1.0
Lauk .a 4 .2460
______ <.05
All Mallg— .0O 3 -.0022 .0081 .2153
usat Nea— <.31 <.01 4.31 <.0 1.
r o_tai1.d rvaluu Sa ed on scaulard :—test.

Tabia 30 Si ifiGaut Reg . .sioO Co.Ui ienr.* .ad p .L *i* of WeightM Rigruatoft Analysi. for Perceot
Siarf.c. Wat. Usa . a*d Sel.ctad Socio—,cono ic Variabi.., Organ Sita—Specific for )01t-iJI%itt
Male., Malignant N.eplas* Mortality, 1950—1969, in 346 Stu4y Counties.
• p1oy.d ?op i1a- 2
2 M.dian f.ployed in Mon-dur— tion per urban Eapley ed
Surfac. Moit— Median !duca— in A zi— able l4snu— square ?rpu— Foreign in Manu—
Sit. lar Whit. Incous eton culture facturing nil. lation Stock facturing
Esophagna .0024 —2.9799 .1040
<.025 ‘.03. 4.01
ft o.acb .1550 .2533
c.i0 <.10
.0023 .0011
<.10 <.10
Passage. £
<. 10
.0001 —10000 .0006 —.0140 .0112
Sron_cbua 6 ‘.023 <.03 ‘.023 <.01 ‘.023
Sledder 6
Oth.r On—
nary Organs
Lysphosar- -.0236 <.05
R.ci— <.05
r n1os.rc
ltip1s 0012 —1.1320 ‘.10
<.10 ‘.02.5
All Malig— .0167 - 16.3793 .9117
<. .0 ‘.05 <.31
nant .o—
p las n s
‘r-io—cail.d p va1ue Sa.ssd on standard c—east.

rabi. 31: Si8nift .axat ReUession Couffici.nts and p values1 of Weighted Mgr*ssion ‘gi*lysis for ?srtent
Surface Water aase end S.l.cted Socio—econo ic Variablan, Organ Site-Sp.cilic ?or ‘On—%lhita
Tansies, Maii8nanc :eopLa.as Mortality, 1950—1969, in 346 Study Counties.
f Eploysd Pop il .a-
: 2 ‘edian # j N() -diaL — ‘ion per Orban 2 2 Enployed
Cant.r Surface Son— Median Educe- in Agri— able Menu— square Popu . Foreign in Sean-
Sit. Water uhits Intone tion culture facturing nil. letion Stock facturing
aopha ua .0107 .0337 —.0763
(.05 ‘.01. ‘.01.
7.3370 —.2442
<.025 4.10
.0012 .1012
‘.10 <.10
Siliuy —.0249 — 1.3500 0576
?aaa*$S & 4.10 <.05 ‘.05
Pancreaa .0023 —2.3238
‘.05 ‘.025
tarynu .0071.
Sroncl*us &
3reaat 5.5203
3laddur & —. 0268 .0462
Other 0r — 4.10
nary Organs
Eod$kins .0078
Disease <.10
c 4 leti—
Sye ioan
______ ‘.10
All MeLig—
oa.nt Neo—
.rwo— ai1ed ? —uaiue based on st ndiVd t— eSt.

Table 32 : Significant Regression Coefficients and p—values 5 of Weighted Regression Analyses
(Reduced Model) for Percent Surface Wat.r Usage and Selected Socio-economic Variables,
Organ Sits—specific for San—race Groups, Malignant Neoplasa Mortality, 1950—1969, in
346 Study Counties.
Median Educa—
Income tion
.0047 .0002 —.2434
<.3 ]. <.05 <.01
.0053 .0007
<.02.5 <.01
.0005 —.1476
<.01 <.10
.0084 .0004 .1758
<.01 <.01 <.10
8reast .0250 .0018
<.01 <.01
Employed Popula—
Employed in Mon—dur— tion Per
in Agri— able Mann— Square
culture facturing Mile
.0005 .0148 .0712
<.01 <.01 <.01
—.0001 .0035 .0228 .0046 .0833
<.05 ‘.01 <.05 <.05 <.01
—.0001 .0016 .0133
<.01 <.01 <.01
esophagus .0103
jo—sided p —uulue based on standard z—tast
.0316 —.0684
<.01 <.01
2 Employed
in Menu—
S — / Cancer
Raca/ Sits
White Males
Esophagus .0038 .0004 —.2562
<.05 <.01 <.0].
Large .000* .0092
Intestine ‘.01 .t0
.0071 .0009
<.05 <.01
a4der 4
Other Un-
nary Organa
t.y hosercoma
4 Raticulo—
s oan
All .1(gnant
8eopl —
White Females
.0288 .0038 —1.1955
<.01 <.0]. <.025
—.0167. .0839
4.01 <.01
.0810 .0092 —3.387].
<.01 <.01 <.01
.2.333 .7711
<.0]. <.0].
— .0003
.6101 .0010
<.025 <.01
• . .oooO .0004
< .025 <.025
My e l
Non—vhite Males
<• 10

Table 33: Sig iftcant Percent 5ur fac. Water Usage Regression Cost ficiants a d p .3.nas* fran Waihce4
Regression Mtalysss (reduced nodul) • Sex—race—sits—specific Cancer Mortality Rates .rsus
Selected Soc o-econoed.c Variables Stracifted by Papulatioia 1 for Malignant Meoplas3 Mortality,
1950—1969, in 346 Study Cot ties.
# S E 1oyed Papula— 2
2 N.dian Rap loyed in on—dur— tion Per Urban 2 £nployed
Surface Median Educa— in Airi— able Ma u— Square Popu— Foreign in Mane-
Water logan. tion culture facturing t1a Lation Stock facturing
<50,000 —.0042 .0203 .0630
<.05 <.31 <.01
30.000— .0005 .0039 .0987
230,000 <.023 ‘.01 ‘.01
,z50•000 .0550
Large totes tins
<50,000 .0012 —.0246 .0777 .1351 .0398
<.01 <.01 <.01 ‘.35 ‘.025
se.ooo— .1133 .0962
230,000 ‘.023 ‘.325
.230,000 .0669 .2264
<.10 ‘.10
<50,000 .0008 .2121
‘.01 ‘.01
.0043 .0919 .0483
<.05 ‘.01 ‘.10
<50,000 —. 1646 .0093
‘.10 <.05
‘250.000 .0107 —.0001 .0016
‘.10 ‘.025 <.025
Branches 4 Lung
‘50,000 .0026 —.3024 .1703 —.1860
<.3 1. <.01 <.01 <.01
50,000— .0061 —.0044 —.1341
250,000 <.03. <.01 <.10

Tabi. 33: (cont .n .d)
I I Ep1a .d ?ap ga— : Z
fl.dian ! joy.d in Son—dur— tion Per Urban : £nplcy.4
S &rfans *dian E4uca— in Ari- abl. anu— Square Popu- Ior,i$n in u
W.c.r Inc u cian culture facturin Mile latiou Stack facturtn$
M*LZ (cane’f)
lla4dsr & Oth..r
Urinary Qrgai
‘ 30,000 .0007
30,000— .0185 .0366
230,000 ‘.0 1 ‘.05
All MAli ant
<50,000 .0044 —.0031 .0366 .3522 .7126
<.01 ‘.01 ‘.01 ‘.31 ‘.31
50,000— .1059 .0116 —5.3330 —.0086 .7204
250,000 ‘.05 <.01 <.025 ‘.01 ‘.01
‘250,000 .3317 .0326 —10.4942 .0045 .6664
<.01 <.01 <.01 <.10 <.01
<50,000 .0571
50,000- .0873
250,000 ‘.025
‘250,000 .0197 .0041
<.10 <.10
3r .at
.0011 1.2398 .0013
<.0 ] . <.01 <.01
50,000— .3019 .0610
250,000 <.01 .10
‘zso,ooo .03U .0066 .0027
‘.01 ‘.01 ‘.03
)taltiplm .1oaa
‘50,000 .0043 .1681
<.03 <.01
50,000— .0001 —.0003
250,000 -
.0000 —.0003
‘250,000 <.025 <.05

33: (con tinu.d)
t I i — :
Mudian Ep1oy in • on—th z ‘ion P .r Crba 2
Surfac. Msdta n duca— in Agri— able a u— Square Fopu— ?or.i*n in M u-
1nco tion cultur, facuring il. 1*tton Stock fscturLng
‘50,000 .2447
Egoplt ag
‘50000 .0276 —.0779
‘.01 ‘.10
‘230,000 .0208 —.0921
<.01 ‘.31
‘50,300 .0136
tvosjd.d pulu. Sea.d on standard t—Cest
‘<50,000 Co — 27 ); 50.000—230,000 Cu — 62); 250,000 Co — 11)

I2 o
Table 34: Signif canc Regression Coeifici. ts and p-vslues* of Ua’.ight.d Regression Analysis for Percent
Prechloriuacjon and Selected Socio -econo c Vases • Organ Site—specific for S hice aleu,
Nalignanc eaplas. ortality, 1950-1.969, in 346 Study Counties.
# S E p1oyed Populs- 2 2
2 2 fladtan E 1oy.d in toa—dur— clan Per Urbao 2 Replayed
Prechian— lløn— edian Educa— in Agri— able Menti— Square Papu— Foreign is Mann-
maRion vhLt. tncoee clan cvlture facturiag tie latian Stock facttarins
Esaphagos .0009 .0019 .0719
‘.01 ‘.10 ‘.01
—. 3871 —.0012 .3907
(.05 .0l (.01
Large Intestine .0007 —.0016 .0023 .0312 .0804 .0453
‘.05 ‘.10 ‘.023 (.023 (.10 ‘.05
.0072 .0007 —.0006 .0014 -.0213 .1288 .0402
‘.10 ‘.01 - (.10 ‘.023 ‘.025 ‘.01 c.01
liver 4 Other —. 3890
Itiiaxy Passages ‘.05
Pancreas .0265 —.0330
‘.01 ‘.05
taxyoz .0005 —.1570 —. 0003 .0017 .0100
‘.10 ‘.10 c.05 ‘.05 ‘.01
Traab.a. .0058 —1.8231 .0017 .0001 .1731 .141l
Irunck . ,a 4 tang ‘.01 ‘.023 ‘.01 ‘.01 ‘.01 ‘.025
Prostate .0010 —.0263
.023 .10
lladd.r 8 Other .0067 .0005 .0007 .0576
Vrinesy Organs .10 .01 .10 .01
ledkios Disease .0003 .0112
.01 .01
Lyepbasarcosa & .0002 —.0032 .0201
lacian ia_rcaea .01 .10 .023
Lsukeeia 8 .0019 .0004
A1suk a .10 .025
aji a3 .ignant .0482 .0077 .0053 .0187 .233.3 .9667
1 eopLasas .io .01 .01 .10 .01 .03.
,— -,.
*tvo sided p—value based on standard t—test

‘able 35: Significant Coefficients aM p—values 5 of Ua v.ighted Ragression Ma slysis for Percent
Pranhioriution and Selected Socio—ecooonic Variables, Organ Sits—specilic for White
Ts el.s • Malignant eop1asa Mortality, 1950-1969 • in 346 SrMy Counties.
I 0 E Loyed Popula—
2 2 Msdiaia £nploy.d in on-dur- tion p.r Urban 2 Epioi.d
Pr.chlor- Moe— Median Educa— in Agri— able Mann— square Pop.— Poreign in Menu—
mattes White Incoes ties culture facturing mu. litton Stock faccuriug
.0044 .0091
4.10 ‘.05
—.6038 —.0004 .1552
<.01 <.10 <.01
Large Intestine .0010
‘ .01
.3482 .0676
‘.023 c.01
Liver & Other -.3006
Siliary Passages i .025
Pancreas .0119
—.0001 -.0069
‘.01 4.01
Trachea, .0006 .0003 .0024 .0319
Srosthus 4 Lungs ‘.01 ‘.10 ‘.10 ‘.01
Sre.az .0001 1.1665 .0001
<.315 ‘.01 ‘.01
3ladd.r & Orher
Urinary Organs
aodgkins ta.eae .0001
Ly bo.ercoas &
tmticul.oearcoe* -
MeiXipl.e Mps3.oss .2 -579 .0096
Lonkesia & Aleukania
.0036 —.0023 - .1059 .3676
<.01 <.05 ‘.023 ‘.023
o—jid.d p—’ra3 .u . hued on standard t—ceat

Table 36: Significant .re.sicn Coefficients and p—va1ue. of Unv.ight.d Regr.uion Malysia for Percent
Prch1or .nstien and Selected Socio-.conosic Variables • Organ Site-specific for on .-vhite Males,
Malignant Maoplaa Mortality. 1950—1969, in 346 Stud’, Co tte..
I I Esplayed Pea—
Z Median Eaploy.d in Nco—d*ar— non Per Urban 2 E loy*d
Precblor Moe- Medi Educe- in Agri- able Mann- Square Papa— Toreign in Mann-
Lnar.ian uttjta Incose cion culture factaring 1. 3 ,aticn Stock factu4ng
.0023 —2.0832 —.1.582
<.10 ‘.10 ‘.10
Large Intestine .2436
Rectus .0017 .0049
<.10 ‘.10
Liver & Other
3Lliary Passages
L ar nz
3ronchus 4 Lung
Pruscat . .0057
32.adder Other
Urinary Organe
Ly ho.arcusa 4
L.uks.ia &
—18.8606 .8856
‘.10 ‘.10
*tvc iided p— a1u* aaed O, scandard t-teaC

Table 37: Si .if ca.nc Coefficientj and p —v*Lusi 5 of t3nv.ight.d Rs reasi Analysis foc Pereest
Prechlørinacion and Selected Socio—econenic lariab les, Organ SL:e—specific for Non—
whit. Tenai..i, a1ignan: ‘.opLaa Nortality, 1950—1969, in 366 Study Coimnies.
# # !nploy.d POpula— 2
2 Median Eçloy.d ja on-dur- tion p .r ban 2 2 £npl.oyed
Pr .chlor- . ?on— Median Educe— in Agri— able tenu Square Popi*— Foreign in Mann-
insUan vtLite Into .. tion culture facturing ) ls litton Stock factur5.ng
Esophagus .0274
.0306 .0023
‘.10 ‘.10
Large tntestLAs 8.2762
Liver & Ocher 1.67 1 3 .0642
3iliary Passages <.023 403
taxynx .0130 —.1031 .0012
tra .a,
3roncbus S Lung
3rea.st 6.2308
3ladd.r & Other .3515
Urinary Organs <.03
EodgkLni Disease .0083
ty boearco &
aticuLos arc on*
Meltiple My.lona .0013 —.8109 .0008
<.02.5 ‘.10 <.10
.0012 —.0326 .1399 .5515 .2176
L.uka a .1e <. <.023 <.05 ‘.023
All ai.ignant Neaplaao.
*t iaiidad p—value based n stand-axd t—celt

ab1 . 31: Significant 2. ression Co.f2ici nta and p_v.lu .a* of Weighted Regression Analysts for ?*rc.nt
?r.chlottn.tion d S.1.ct .d Socio—acononic Variables. Organ—Sits—specific for White Hal..,
‘ aii .ast .optas* Hortality, 1950—1969, in 346 Study Cøimci .s.
• 9 £ploy.d ?o u1a— 2
2 2 Hedisa Eploy.d in Hen—dur— tio Per 0rb 2 tnploy.d
Pr.cblor— Non— ,dian Educa- in A4ri- able Hasu— Square Po - Foreign in u-
macme vhits Inco.. tine cutter. facrerto 14i1. 1atic Stock f.cturing
.0003 —.2484 —.0002 .0009 .0179 .0712
<.03. <.01. <.3.0 <.01 c.O1 <.01
—.3242 —. 0009 .4096 —.0267
<.025 <.01 <.01 <.10
Large lacasein. .0006 —.0001 .0034 .0299 .0589 .0804
<.10 <.10 <.01 <.01 <.01 <.02
.0140 .0008 -0000 .0021 —.0156 .0759 .0377
<.01 <.01 ‘.05 <.01 <.02 <.02 <.01
Liv*t & Other —. 3341 —.0003 .0081
3ilisry Psasagn. <.01 <.01 <.10
—.0005 .0117 .0423 —.0353
<.01 <.10 <.01 <.01
larynx . 002 —.2665 —.0001 .0000 .0014 .0182 .0132
‘.10 ‘.01 <.10 <.01 ‘.01 <.01 <.02.5
Trachea, .0037 —L1318 —.0038 .0002 .1555 .0854 —.1358
8ronchns 4 Uang <.01 ‘.02 <.03. <.10 <.03 . <.05 <.01
3rwt .0009
Prnstac .0009 —.0001 .0022 —.0215 .0337
<.01 ‘.10 ‘.025 <.05 <.03
NLadder & Other .0089 .0076 .0507
Ortnary Organs <.01 <.01 . <.01.
Rodgkins Disease
Lyho.arcoea 8 .0056 .0004 —.1982 .0002
!.ticniOUrcOSa <.01 <.01 <.03 <.10
.1382 .0000 -.0003 .0040 —. 0088
<.01 <.10 <.03 <.10 . <.05
.0000 —.013 .0126
Lneka .a 8
A1auk & <.025 <.023 ‘ .025
All. a1 .ignant .3097 4.4036 3061 .0104 .2951 .8474
R.opl.ag <.01 ‘.01 <.01 <.01 <.31 <.31
.—.— -. -
*r.o...,jdd p—va.Luea based on standard :—teat

ab1. 39: Si iftcanc P-e rea.ian Coefficients and p_,alues* of W.ignt.d agreseion Analysis for Percent
Pr.chlortnation and Selected Socio—.cono c Variable, Organ—Site—specific for Whit. ?enali, ,
‘ ali ast !eapl.asa Mortality, 1930—1969, in 346 Study Cow ciea.
# I Eaplay.d Popula— 2 2
2 2 Median 1oy.d in oa-dur- cion Per Urban 2 £nploy.d
Prechlor- So .- ‘ .dias Educe- in Ari- able Xa *- . Square Popu- Tor,i a in Menu-
ination whit, taco.. io. culture facturing Uia lation Stock Lecturing
—.0736 .0035 —.0056
‘.10 <.01 <.05
—.0004 —.33 —.0003 .1837
‘.03 ‘.025 <.03 <.01
Lar u Intesrias .0007 —.0001 .0037 —.0201 .0847
<.02.5 <.025 <.31 <.05 <.31
.0062 .0002 .2286 —.0138 .0352 .0222
‘.01 <.10 <.05 <.01 ‘.01 ‘.01
Liver & Other —. 3322 —.0005 .0318
3iliary Pasanges ‘.05 <.31 <.023
Pancreas .0373 —.0145
<.01 <.10
—.0001 —.0051
<.01 <.01
rath. .. .0005 —.26â4 —. 0005 .0012 .0100 —.0260
Sroncbo. & Lung <.01 <.025 <.01 <.01 ‘.05 <.01
Sr.a.t .0130 .0011 .7319 .0006
<.023 • <.01 <.01 <.325
$ladd.r & Other . oo
Urin*ry Org
Rodgkio. Disease —.0000 .0004
<.03 - <.05
Ly ho.arco$ & .2232
eticu1csarco5a <.03
Melcip I. frelone .0868 .0000 —.0008
‘.10 <.05 <.01
L.uk a & .0002 .0002 .0000 —.0008 .0092
A1.uk a ‘.023 ‘.05 <.05 ‘.31 ‘.01
jj .0043 — .0021 .0002 .0087 .2230
<.31 <.01 <.10 <.01 ‘.31
tjo—sided —vaju* based on a candard t—cesc

Table 40: Significant Regreesion Coefficients and p—values* of Weighted R.gr.seton Ana1ysi for Percent
Preeblartascion and Selected Socin—econoede Variables, Qrgan—ei:e—ap.ciftc for Non—white tales,
Malignant N.oplaes Morta3.lcy, 1950—1969, in 366 Studp Counties.
# Employed ula— 2 2
2 2 Median Employed in Non-dur— tion Per Urban I E loy.d
Pr.chlor Non— Median Educa- in Jgri- able Mann- Square Popu- Foreign in Manu-
inatio white Imc. tion cultvre facturing Mile latton Stock facturing
Esophagus .0024 —3.0423 .1102
(.023 ‘.01 ‘.01
St...ch .1429 .2769
<.1.0 <.10
Large Intestine
.0024 .0017
<.10 <.10
Liver 4 Other
Siliarp ?weges
Pancreas .0003
Trachea, —. 0055 .3139
Sronchus & twig <.10 ‘.023
3rwr .0001 -.0000 .0006 —.0114 .0096
<.05 <.03 <.025 ‘.05 ‘.05
Zladdar 4 Other
Urinary Org -
Hodgkins Disease .0203
Lymphosarcnna &
Racicuioearco ma /
Ma. ltip le .0013. —I. 1884 .0710
eiusa .10 .025 .10
L ik a &
A l.uke . a
All Malignant .0170 —16.9206 .9886
.10 .025 .01
*t ,o ided p—valu, based on standard t—cest

Table 41: 5Lnific t !.resston Coefficients and p—valu .s of V .tghtad egresiton Analysts for Percinc
Pr.cblorinatioe and Selected Socjo—.cononj Variable,, Organ—sita—ipecific for ‘4on- tte
Final... ‘tslignanc Neop1as ? rtaUcy. 1950—1969, in 346 Cozmcies.
S iEployed Popula— Z Z
tadian En 1cv.d in 1o—dur— tion Psr Urban £ployed
Prechior— you— edim Educe— in Agri- able anu— Square Papu- Foreign in tanu—
tuition uhite tncese tion culture facturing M l . taiion Stock facturing
.0131 .0429 —.0629
‘.1.0 ‘.01 ‘.01
Seoascb .1960
Large Ingeptin . 6.9951 —.2137
‘.023 ‘.10
Rsctt .0012 .0929
‘.10 ‘.10
Liver S Other —1.1345 .0435
3tljarv Passages ‘.10 ‘.10
?ancrea. .3024 2.2409
‘.35 ‘.02.5
Zroncbus 6 Luag
3mc 5.3979
3ladd.r 6 Other
Urinary Organs
odgkinu Disease .0076
ty hc.arcoea 4 .1945
leticulosarcoan .10
9z1ttp1 . .01.2.2
T.auk a S
£1* uks *
All !alignznt
*t,o . .ejded —r Lu . based on standard c—test

I 3
Tabi. 42: Significant agr.a.ion Coefficients and p_vatu*s* of ¶Jetghtsd Regression Analyses
(l .duc.d Model) for Percent Pr.chloriaatjon and Sei.ct.d SocLo—.concuic Variables,
Organ Sits—specific for Sex-race Croups. Malignant M.opla Mortality. 1950—1969,
in 346 Study Counties.
# # Eploy.d PopuLa- 1.
.dian Eplayed in ilon-dur- tion per Urban 1. Enploysd
S.x-rzcsl Pruchiur- Median Educa- in Agri- able Mann- Square Popu- Foreign to Mann-
Carter Site inazioe men .. tin, culture facturiog Mile latin. Stock facturiog
Whit. Melsa
.0006 -.3C53 .oooo .0182 .0737
4.01 <.01 ‘.01 4.01. cOt
Lule .0007 —.0001 .0035 .0303 .0533 .0777
Inte gia . <.01 <.33 <.01 <.01 <.01. <.01
.0124 .0009 -.0002 .0010 -.0146 .0877 .0309
‘.01 <.01 ‘.10 <.01 <.01 <.01 <.01
.0002 -.3179 -.O 01 .0017 .0174 .0083
<.05 ‘.01 <.31 <.01 <.01 4.35
Trachea., .0038 -1.4635 -.0035 .0001 .1313 -.l25 )
3roncbua, 4.01 ‘.01 ‘.01 ‘.01 ‘.01. <.01.
Iladd.r 6 .0092 .0007 +.0000 .0503
Other Un— 4.01 4.01 <.10 <.01
any Organs
Lp phoaareu . & .0037 - .0005
icaticulosarcOsa 4.35 ‘.01
All Malig— .0087 -.3887 -.0064 .0087 .2933 .8294
nant ice.— ‘.01 ‘.01 4.01. <.31 <.01 <.01
p la.i
Whit. T..mL.s
l.cti .0080 .0003 —.0069 .0433 .01.25
<.03 ‘.023 <.01 <.01 4.01
.0249 .0020 .4462 .0012
4.01 4.01 ‘.01 ‘.01.
.0665 .,..0000 -.0003 .0043
4.Q5 ‘.03 ‘.05 <.10
$o ,-.hits Males
.0019 .0017
Xoa- it . Fe.ale a
Esophagus .0390 -.591
*tuo..gided p alue based an standard t -test

Tabl. 43: Si nificanr Percsnt ?rechlortnatton a .s.ion Co.fficianca and p_v*lu.s* fro* Weighted
*sgr.eeien ‘ ,nalya.* (reduced nodel) of S.z—Race’-Stte—specifit Cancer corra1ity aces
‘ersia Selected Socio—.cono c Variables Stratified by Popuiaciont for 4alignanc eup1as
rta1ity, 1950—1969, in 31.6 Scndy Couutiea.
S S ! p1oye4 PopuLa— 2 2
2 .4i !ploy.d in cn—dur— ti*n Per Orban 2 Eploy.d
Prechior- (edian !duca- in Art- able A$ri- Square Popu— Foreign in tanu-
macion Inc ti culture culture 1. lation Stock faccurin
<50,000 .0096 —.0056 .0160 .0593
<.01 <.02 <.0 ]. ‘.02
50,000— .0005 .0041 .0971
250,000 <.0 1. <.01. <.01
‘250,000 —.0087 .0011
‘ .31 ‘.10
Laz. lnt tin.
<50,000 .0013 —.0269 .0708 .1091 .0663
<.01 ‘.01 <.01 <.10 ‘.01
50,300— .0036 .1133 .0950
250,000 <.10 <.01. <.025
‘50,000 .0008 —.0051 .2063
<.01 ‘.10 ‘.01
‘0 000— .0046 .0907 .0482
20:300 <.023 ‘.01 ‘.10
‘50,000 —. 1781 .0098
‘.05 t.323
an .0031
Z50 000 .0013
3r cku. 4 Lung
‘30,000 .0442 .0022 —.0043 . 81 -. 9 6
<.025 <.31 ‘.01 .01 . 1.
50 000— .0061 —2.5817 —.0043 .0838
250,000 <.31 ‘.05 <.01 ‘.01

TabLa4S: (conrinu.d)
I I Eplcyed ?ap La— 2
I Median Enployed La on—thar- tion Per Urban 1 Enploy.4
?tecbior— Median Educt— La A i— able Agri— Square Popu— Foreign in Mania—
1nca . rto eu1tu e c .4r . . re Mile mUon Scock faccuring
U1fl MALZS (cont’d)
Iladd.r 8 Other
Urinary Organs
d O.000 .0007
< .01
30,000— .0110 .0371
250,000 ‘.10 ‘.10
p250 ,000 .0183 .0012 .0650
‘.025 ‘.05 ‘.025
All Ma.Lignan
‘50,000 .0042 —.0042 .3148 .5118
‘.01 ‘.01 ‘.01 ‘.025
50,000— - .0103 —5.3939 —.0085 .0317 .3711
250.300 <.01 ‘.025 <.01 ‘.023 <.01
W Z7Z F tAL25
50,300— .0123
250,000 ‘.10
3 reast
.0010 L.Zt.07 .0013
‘.01 ‘.01 ‘.01
50,000— .0200 .3020
230,000 ‘.10 ‘.01
<50.000 .1347
30,300— .0001
250,000 < i D

Tai 1.+ (conciuu.4)
S I E 1a .d ?optal,— Z Z
2 Median E p1oyed in oa—d ar iou Per Urban 2 E.nploy.d
Prechior- Median Educe-’ tn Mn- able Agni— Square Papu— Foreign in u—
inetien Inc tioa culture culture tie latien Stock Cacuuning
‘50,000 .0416
50.000— —.041 3
‘2.50,0 0 0 .0241
‘.0 5
50 Z
‘50,000 0234 —. 1 050
<.01 <.025
50,000— —.0291 .0492
Z50, <.10 <.023
p2 30 300 .0612 —. 3336.
<.10 <.01
5 t’ o_ajde4 p- a1uu based an standard t—tesc
‘c50,O00 Co —273); 50,000—250,000 (a 62); ‘250,300 (a U)

• 1 1 y ,•,••
I .. I ..I,
I •.. r.
•ip • ,.,,“ ., •I.. .• • i I,, •
I , .:.,, . /
• 1”• . . . .. .. ..,.
— -“ .-
, /‘

Andeli i,, J. 3 ., and Suess, M. 3. 1970. Polynuc3.ear Aromatic Hydra—
carbons in the Water Environment. Bull. W. 3. 0 . 3: 479.
anonymous. 1960. Census of Population. Governent Printing Of fi ce,
Washington, D. C.
Anonymous. 1964. 1963 Inventory Municioal Water Facilities . Publi:
Health Sexy. Pubi. No. 775,. Vol. 5. Government Printing Office,
Washington, D. C.
Anonymous. 1970. Water Quality Criteria Data Book , Vol. .1.. Organic
emicai. Pollution. of Freshwater. Gov rnment Printing Office,
Washington, D. C.
Anonymous. 1974. Draft Analytical Renort: New Orleans Area Wate.
Sunuly Study . EPA, C ncinnatj.
Anonymous. 1975. A Report: ssessnent of Health Risk from Organics
in Drinking Water by an d Hoc ttidv Grout, to the Hazardous Ma-
terials Advisory Coittee, Science Advisory Board . EPA,
Washington, ID. C.
Attohoni, A. and Greenberg, A. K. 1970. Toricologica.L Aspects of
Waste Water Reclamation: A Preliminary Report. Jour. WPCP 42:
Bellar, T. A., Lichtenberg, 3. 3., and Kramer, R. C. 1974. The
Occurrence of Organohalides in Q 1orinated Drinking Water.
3. Am. Water Works Assoc . 66: 703.
Berg, 3. W. and Burbank, F. 1972. Correlations Between Carcinogenic
Trace Metals in Water Surp lies and Cancer Mortality. Ann. NY
Acad. Sd . 199: 249—262.
Berg, 3. W. and Ens tram, 3. K. 1974. Geographic CorrelatIons Between
Cancer Mortality Rates and Alcohol-Tobacco Cons ption in the
United States. Jour. N. C. 1 . 53: . 3.
Berg, 3. W. and Howell, M. A. 1974. The Geographic Pathology of
Bowel Cancer. Cancer 34: 307—814. -
3*.u cher, C. P.. 1975. Cincinnati Drinking Water: An Epid emiolcgic
Study of Cancer Rates. Univ. of CincInnati Med. Ctr., , CincInnatI.
Davies, P.. I. and yme—Griff±th. 1954. Cancer and Soils in. the Cc mty
of Anglasey: A Revised Method of Connarisons. British 3. Cancer
3: 594.
DeRouen, T. A. and Diem, 3. E. 1975. The New Orleans Drinking qatar
Controversy: A Statistical P active. An. 3. ? bi. ea..ith 65:

DeRnuen, T. A. and Diem, 3. E. 1975. Ethnic, Geographic Differences
in Cancer Mortality in Louisiana. Unpublished.
Dostal, K. A., Pierson, R. C.., Eager, D. C., and Bobeck, G. C. 1975.
Carbon Bed Desi Criteria Study at iitro, W. Va. 3. i m. Water
Works Assoc . 57: 663.
Dowty, 3., Carlisle, D., Laseter, 3. L., and Storer, 3. 1975.
flalogenaced Rydrc arbon.s in New Orleans Drinking Water and Blood
Plasma. Science 137: 75.
Dowty, 3. J., Carlisle, D. R., and Laseter, 3.1. 1975. Ne’i Orleans
Drinking Water Sources Tested by Gas Chromatography—Mass
Spectrcmetx7. Envir. Sci. Tech . 9: 762.
Dras ar, B • S., Irvihg, D. 1973. Environmental Factors and. Cancer of
the Colon and Breast. Br. S. Cancer 27: 167.
Duhi, 3. C. and T uup, S. W. 1953. First Report on Geographical and
Geological DistributIon of Carcinogens in the Netherlands. Laiden
Foundation for the Study of Psychophysics. -
Eschenbrunner, A. 3. and Mifler, E. 1945. Induction of Hepatanas
in Mice by Repeated Oral. Adninistration of Chloroform with
Observations on Se Differences. S. Nat. Cancer Inst . 5: 251.
Gregoropoulous, S. C. and Smith, S. W. 1963. Trace Organics in
Missouri Subsurface Waters. Jour. AWWA 60(3): 586.
Rites, R. A. and Bier r , K. 1972. Water PoLlution: Organic
Compounds in the Charles River. Science 178: 158.
Hogan et al. 1976. Association Between Chloroform Levels in Finished
Drinking Supplies and Various Site—specific Cancer Mortality
Rates. N.I.LH.S., submitted for publication.
Hueper, W. C. 1960. Cancer Hazards from Natural and ArtifIcial
WaterPollutants. Proc. Ccnf. on Physiblozlcal Aspects of Water
Qualit’r . U.S.P.A.S., Washington, 13. C., p. 181.
Hueper, W. C. and Payne, W. W. 1960. Carcinogenic StudIes an Petro—
laun Asphalt, Cooling Oil, and Coal Tar. Arch. Path . 70: 372—384.
Hueper, W. C. and Payne, W. W. 1963. Carcinogenic Effects of Ad.sorbatas
of R and Finisiaed Water Supplies. amer. S. Clin. Path . 39: 473.
Huaper, W. C. and Ruchhoft, C. C. 1934. Carcinogeoic Studies on
Adsorbates of IndustriaLly Polluted Raw and Ftnished Water
Supplies. A.M.A. Arch. Indust. Ev . 9: 495—499.

Iargus, 3., Reese, W. K., Jr., and Butter, H. A. 1969. Induction of
Lung Adenomas in Newborn Mice by Bis (chloronethvl) Ether. To .c.
A o1. Phar aco1 . ii: 92—96.
Jolley, B. L. 1973. 1orinatton Effects on Organic Constituents in
Effluents fron Do eric Sanitary Sewage Treatment Plants. Pub 1.
No. 565. Envir. Sci. Div., Oak Ridge National Laboratory.
Junk, G. A. and Staaley, S. E. 1973. Organics in. Drinking Water.
• Part I, Listing of Identified Chei ica1s. NIlS, Springf!el.d.
Kleopfer, B. 1). and Fairless, B. 3. 1972. Q aracterization of Organic
Co iponents in a Municipal. Water Supply. Envir. Sd. Tech . 6: 1036.
1assen, C. D. and Plaa, G. 1.. 1967. Relative Effects of Various
( i1orinatad Eydrocarbo s on Liver and Kidney Pun.ctian in Dogs.
Tozicol. Anni. Phar aco1 . 10: 119.
Leong, B., MacParland, K., and Reese, W., Jr. .971. Induction of
Lung Adanonsa by Chronic Inhalation of Bis (chioronethyl) Ether.
Arch. Environ Health . 22: 663—6,
McCabe, L. 3. 1975. Association Between Trihalomethanes in Drinking
Water (NOBS Data) and Mortality- ‘(Draft Report). EPA, Cincinnati.
Page, 1. and Harris, B. K. November 1974. The Implications of Cancer-
causing Substances in Mississippi River Water. A Report by the
Eniro mantaL Defense Euud, Washington., 1). C. Unpublished.
Page, ±. and Harris, B. K. 1973. Relation Between Cancer Mortality and
Dr4 ’king Water in Louisiana. Unpublished.
Page, 1., Harris, B. 3., and Epstein, S. S. 1976. Drinking Water and
Cancer Mortality in Louisiana. Science 193: 55.
Reck, 3. 3. 1974. Por ation of Ha.l oforns During Chlorination of a—
turai. Waters. Water Treat. and Exan . 23: 234.
Science Advisory Board, Enviroental Protection Agency. 1975. A
Report Assessment of Health Risk from Organ cs in Drinking Water
by an Ad Roe Study Group to the Hazardous Materials Advisory
Coittee. April 30.
Stavrakv, K. 1976. The. Role of Ecologic Analysis in Studies of the
Etiola ’ of Disease: A Discusston vith Reference to Lar;e 3cwel
Cancer. 3. Chezt. Dis . 29: 14.33—444.
Stocks, P. 1947. Ragiona .1 and Local DIfferences in Cancer Death Rates:
Studies on Medical and Polation. Subjects. General Registry
Office, No. 1, London.

Subik, P. arid itchie, A. C. 1953. Sensitivity of Male dba Nice to the
Toxicity of Chloroform as a Laboratory ifazard. Science 117: 285.
Symons, S. A. 4. 1975. ational Or ariics Racou issance Survey for
Halogenated Organics. Jour. WA 67 (November): 634—647
ardiff, 1. G., CarLson, C. P., and S wn, V. 1976. Ealogenated
Org ics in Tap Water: A Toxicological Evaluation. In The
• Environmental Imoact of Water Chlorination . National. Tec ic.a1
Information Service, Springfield, Va.
Tarone, 3.. 3.. and Cart, S. 5. 1975. Revie i of “The Implications of
Cancer e ’ti ing Substances in Mississippi Rain Water” by 3.. 3..
Harris. Unpublished.