PB-243 009
ESTIMATING LIMITING  RISK LEVELS FROM ORALLY  INGESTED
DDT AND DIELDRIN  USING AN UP-DATED VERSION OF THE
MANTEL-BRYAN PROCEDURE
GEORGE WASHINGTON  UNIVERSITY
PREPARED FOR:
ENVIRONMENTAL  PROTECTION AGENCY
9 APRIL 1974
                            DISTRIBUTED BY:
                            Nations! Technical Information Service
                            U. S. DEPARTMENT  OF COMMERCE

-------

-------
                              TECHNICAL REPORT DATA
                        (Please read futtruclions on the reverse before completing)
 1. REPORT NO.
 EPA560/5-75-003
                         2.
                                                  3. R
 4. TITLE AND SUBTITLE Estimating Limiting Risk  Levels
  From Orally Ingested  DDT and Dieldrin  Using
  Up-dated Version of The Mantel-Bryan Procedure
           5. REPORT DATE
             April  9,  1974
           6. PERFORMING ORGANIZATION CODE
                     243
 7. AUTHOR(S)
           8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
 Nathan Mantel
 Research Professor of Biostatistics
 George Washington University
 7979  Old Georgetown Rd.,  Bethesda, MD
           10. PROGRAM ELEMENT NO.
             2LA328
           11. CONTRACT/GRANT NO.
    20014
P4-01-02962
 12. SPONSORING AGENCY NAME AND ADDRESS
 Office  of  Toxic Substances  (WH-557)
 Environmental Protection  Agency
 401  M St.,  SW
 Washington, DC  20460
            13. TYPE OF REPORT AND PERIOD COVERED
             FINAL
           14. SPONSORING AGENCY CODE
 15. SUPPLEMENTARY NOTES
16. ABSTRACT
 Mathematical extrapolations  of the upper  limits on risk of  cancer
 at various low levels  of  exposure to dieldrin and DDT are presented.
 The  statistical model  used is that described by Mantel, ejt  al. ,
 Cancer  Research, 3j5, 865-872, 1975, the so-called "updated"~llantel-
 Bryan procedure.  The  data upon which the  extrapolations are  based
 are  derived from the studies by Tomatis,  e_t  al. ,  International'
 Journal of Cancer, 10,  489-506, 1972 for  DDT and  by Walker, ejt al.
 Food and Cosmetics Toxicology,, 11, 415-432,  1972  for dieldrin.

 Several alternative methods  of treating the  data  are presented,  and
 pesticide levels associated  with various  levels of risk are estimated
 Certain precautions which  must be observed in applying the  "updated"
 Mantel-Bryan technique  are discussed.
 7.
                           KEY WORDS AND DOCUMENT ANALYSIS
               DESCRIPTORS
                                       I).IDENTIFIERS/OPEN ENDED TERMS
                       c. COSATI Field/Group
 Dieldrin
 DDT
 Mantel/Bryan
 Carcinogens
 Risk Estimates
 Statistical  Extrapolation
 8. DISTRIBUTION STATEMENT
 Unlimited Distribution
                                       19. SECURITY CLASS (ThisReport!
                                        Unclassified
20. SECURITY CLASS (This page)

 Unclassified
EPA Form 2220-1 (9-73)

-------
ESTIMATING LIMITING RISK LEVELS FROM ORALLY INGESTED




         DDT AND DIELDRIN USING AN UP-DATED




             VERSION OF THE MANTEL-BRYAN




                      PROCEDURE

-------
ESTIMATING LIMITING RISK LEVELS FROM ORALLY INGESTED DDT AND DIELDRIN

      USING AN UP-DATED VERSION OF THE MANTEL-BRYAN PROCEDURE
                                 by
                            Nathan Mantel
                 Research Professor of Biostatistics
                    George Washington University
                      7979 Old Georgetown Road
                      Bethesda, Maryland  20014
                          prepared for the
                     Office of Toxic Substances
                   Environmental Protection Agency
                       Washington, D.C.  20460
                           Project Officer
                          Michael J. Prival
                      Contract No.  P4-01-02962

                            April 9, 1974


                             ..'   3<

-------
                    NOTICE
This report has been reviewed by the Office of
Toxic Substances, EPA, and approved for publication.
Approval does not signify that the contents neces-
sarily reflect the views and policies of the
Environmental Protection Agency, nor does mention
of trade names or commercial products constitute
endorsement or recommendation for use.

-------
     I will give below a description of the up-dated Mantel-Bryan


procedure and the rationale for such up-dating.  Let me begin


however, with a description of the data to which the Mantel-Bryan


method was applied.


DATA EVALUATED


  DDT


     The basic DDT data came from a report by Tomatis, Torusov, Day,


and Charles appearing in the International Journal £f Cancer (The.


Effect of Long-Term Exposure to DDT on CF-1 Mice, Vol. 10, A89-506,


1972).  In this work 6-to 7-week old CF-1 mice received dietary DDT at


levels up to 250 ppm for their remaining life span.  One-third of the


females in each group were mated when 9 to 10 weeks old, and their


offspring were given similar dietary levels of DDT.  Only sufficient


first generation offspring were added to the treatment group as


approximately to double the total— there would thus be something on


the order of 120 mice of a sex on a treatment level, and the authors


report in many instances only combined results for both parents and

                                    t

offspring.  (Results are also given for a positive control group re-


ceiving urethane in their drinking water.)  Except for a few mice


with tumors killed early for transplantation tests and two mice per
                        >                                       •

sex-generation-DDT level-group killed for study of DDT levels in their


tissues, mice were Kept until natural death or killed when moribund.


At experiment termination all survivors were killed and autopsied, the


parent generation having been carried to 140 weeks of age, the offspring


to 130 weeks.  The data reported include:  number of survivors at the end
                                     5<

-------
                                     -2-


of each ten-week period; the total number of tumors and of tumor-bearing

animals; the total number of tumors at a specific site.  For liver tumors,

lung tumors and lymphomas data are given separately for three time

periods of tumor occurrence:  between 0 and 69 weeks of age; between

70 and 94 weeks; from age 100 weeks on.  Data are not given on the          ,

instances where mice fell into more than one of these tumor groups, and

the total of such tumors frequently exceeds the total number of mice

under study.

     I note that the total number of tumor-bearing animals is likely

unsuitable for analysis since many non-malignant tumors are included

in any case this is such a broad catch-all category that upwards of

80% of control animals developed some kind of tumor.  The main tumor of

interest seems to be those of the liver which the authors report to be

metastatic and dose-dependent in frequency.  (Transplantation of such tumors

was unsuccessful in 2 cases with young adult recipients, successful in 1 of

2 newborn recipients.)  For these reasons, I have made a "safe"* dose analysis

using liver tumor data.  In this analysis, I have alternatively included

or excluded tumors arising from 100 weeks on, with the exclusion giving

less weight to the late-appearing tumors and minimizing any spontaneous tumor

effects.  I have made a similar analysis relative to lung tumors, but indicate

the "safe" doses resulting as not reliable.  The cause of unreliability is

that at high dose levels of DDT liver tumors arise so early that the mice do

not have a chance to develop lung tumors.
*The word "safe" as used in this report is not necessarily meant to imply
 complete absence of hazard or risk.  The "safe" dosages obtained are
 ones for which, by the procedures used and under the assumptions made,
 the true risk does not exceed some pre-specified low level.
                                      6<

-------
                                  -3-
This competitive effect from liver tumors is largely responsible for

the failure of the authors to detect any dose-progressiveness in their

occurrence.

     The consequent data going into my analysis are shown in table 1.

     I have additionally examined a report by Tarja'n and Kemeny in Food

and Cosmetics Toxicology (Multigeneration Studies on DDT in Mice,  vol. 7,

215-222, 1969) on the effects of dietary DDT.  The multigeneration study

in BALB/c mice was unsuitable for ''safety" analysis for a variety of

reasons.  For one thing, DDT was fed for only 6 months, and at only

comparatively low levels, about 3 pptn.  (But the data do seem suggestive

of tumorigenic and leukemogenic effects at this time-  and dose-restricted

treatment.  The authors attribute this to the generational method of their

study which could have permitted fetal exposure to DDT.)  Nextly, while

the authors do give separate data on tumors by sex, they do not give a

corresponding sex breakdown on the number of animals examined.  The real

curiosity about the data, however, is that no data on tumors arising

during the course of the study are provided-^—what the investigators

report are their findings on an examination of surviving animals 26 months

after the study began,'at which time the fifth-generation mice were

typically 4-6 months younger than the first generation mice.  The data

to me seem to show about the same tumor rates for all generations (which

might indirectly indicate higher rates for the later generations since

they are younger) so I am surprised to see the authors claiming  almost

logarithmic increase in tumor incidence with tumor generation.  The number

of animals with tumors does increase with generation, but so do the number
                                             i
of mice examined so that incidence rates remain stable, in which case the

authors' remarks on fetal exposure become less meaningful.

                                    7<

-------
Dieldrin                 ^   '    •'<"-••   •  :




     These data came from a report by Walker, Thorpe, and Stevenson




appearing in Fpod and Cosmetics Toxicology  (The Toxicology of Dieldrin




(HEOD).    1.  Long-Term Oral Toxicity Studies in Mice,  vol.  11,  415-432,




1972).  Here in various experiments and subexperiments CF-1 mice received




dietary dieldrin at levels ranging as high as 20 ppm, for variously




104, 128, or 132 weeks, ADAB being used as a positive control treatment.




Treatment was initiated in the week following the period in which the




mice as 3-week-old weanings were housed together so as to accustom them




to the change from the breeding to the experimental unit.  In the first1




study, experiment 1, dose levels used varied by 10-fold  increments;



experiment 2.1, of study 2, directed to elucidating dose response




relationships involved 2-fold dose increments of dieldrin, the diet    t




being sterilized by use of-ethylene oxide; in experiment 2.2, a




single dose level of dieldrin was used, interest focussing on the



influence of radiation-sterilization of'diet and use of  litter.




For the present analysis I have ignored manipulations of sterilization



in experiment 2.2 and of litter^ combining all control groups into



one, and all dieldrin<-treated groups into one". •»     ;;:



     The report gives the percentage of animals with tumors of.



various sites but does not permit identification*of instances of



mice with tumors of several-rsltes. v •Here^l' have chosen liver tumor



data as the basis for "safety"''aiValysis^ making alternate determinations



according to whether all liver tumors (a-tb) or only those showing areas




of papilliform or adenoid ^growth' (b)r are- considered.  '1  do not use



the data from two additional experiments "reported,: one directed: to

-------
getting at the combined effect of dieldrin and DDT, the other to the


effect of changing the duration of dleidrin treatment.


     I note that the data published in this report were incomplete in


that they represent percentages of animals with tumors, not actual numbers as


I would have  needed.  In table 2, I give my best efforts reconstruction


of the actual data, which reconstruction I used in my analysis.  An uncer-


tainty results from situations like that in which 20% of 288 mice are

                                                    /
shown to have liver tumors — I took this as 58 tumorous mice, but 57


or 59 are alternative values which would have rounded off at 20%.  My


analysis is probably not critically dependent on which of these


alternative values I used.  I note th°.t in experiment 2.1, liver tumor


frequency at the two highest dose levels was reduced suggesting that


lethal toxicity of treatment had interfered with tumor appearance.  Such


lethality is further indicated by the reduced numbers of animals shown


as at risk which are less here (and also elsewhere in the table) than


those the authors describe in their text as being allocated to each


group.  I would judge that the authors reasonably reduced the numbers of


animals considered at risk so as to allow for mortality occurring


before any tumors were likely to occur.  But they nowhere make clear


just how they did this.  It may be that the reduction in tumor


incidence at high dose levels would not have occurred if the author


used a more effective method of reducing the numbers considered


at risk.  For example, at 20 ppm, the report shows that nearly 50% of


female mice died within 3 months  (all were dead within a year), yet


the number of "such mice for which tumor incidence was calculated is


21, or 70% of the original number.  While the authors accept that their
                                     9<

-------
data demonstrate the induction of liver  lesions, it seems to me that




they are concerned with the failure to induce other lesions like




sarcomas, which ADAB(but tested at 600 ppm) seems to induce.  They in-




dicate that their results support the idea that the mouse may be




Unsatisfactory  for  testing  compounds which induce hepatic microsomal



enzymes.




THE UP-DATED MANTEL-BRYAN "SAFETY" PROCEDURE




     In the original publication by Mantel and Bryan in the Journal




of the National Cancer Institute ("Safety" Testing of Carcinogens, vol.27,




455-470, 1961) two particular qualifications are noted.  One point is




that the procedure is overconservative in the handling of spontaneous




tumor rates, a minor point when control tumor rates are low.  In more




recent long-term testing of carcinogens the occurrence of high spontaneous




rates, say on the order of 40% is not uncommon, and overconservatism




by the original Mantel-Bryan procedure can become extreme.




     An up-dating of the procedure to remove this conservatism has been




prepared by a cooperative effort of investigators at several institutions




and will in the near future be prepared for publication and it is this




up-dated method that I applied to the instant data.   The method formally




takes care of spontaneous responses by formulating a model, jnplicitly




suggested in the Mantel-Bryan paper, that the response at a dose level




reflects the independent addition of probabilities—observed proportion *»




Induced proportion + spontaneous proportion - product of induced and




spontaneous proportions.




          The up-dated procedure further incorporates a second modifi-




cation, the need for which Mantel and Bryan indicate in an appendix




ii.'i superior to the method of "justifiable" combined results described

-------
                                  7                   ;          .

in the text— that modification was suggested in part by Cornfield.

The "justifiable" combined results approach has a kind of bending-over-

backward aspect in that it permits treating any outcome observed at a   :

higher dose as though it had in fact occurred at a lower dose-—this sometimes

could result in a higher estimate of the "safe" level by the Mantel-Bryan

method since that method could be used only for the outcome at a single dose

level.  What this second modification does is to permit the simultaneous

handling of data at several dose levels so that the outcome at a higher

dosage is not artificially treated as an outcome at a lower dosage.  In

applying the method a principle suggested by Mantel and Patwary at the 32d

session of the International Statistical Institute (Interval Estimation of

Single Parametric Functions, Tokyo, Japan, 1960) is employed.  A maximum

likelihood estimate is made of all the parameters of the system, that is

the spontaneous tumor rate or rates, and of the "safe" dose, or equivalently,

the intercept for the induced tumor rate since the probit slope is conser-

vatively fixed at unity, or some other prespecified value.  (The Intercept

parameter is the normal deviate corresponding to the induced rate at unit

dose, I.e. log dose equal to zero.)  The next step is to find an alternate

"safe" dose or intercept value such that when a conditionally maximum

likelihood estimate of the remaining unspecified parameters is made, the fit

is just significantly worsened.  A consequenqe here however, is that use
                  ,'••',     •'••''        1 '  '<  !'
of exact binomial probabilities as Mantel and Bryan espouse is no longer
                      i   ••             '  .• -1
feasible, and the up-dated Mantel-Bryan procedure instead relies on use of
                                          ':, f'
-2 loglikelihood ratio taken as following an asymptotic chi-square
                 "           ' .             t
distribution for setting limits.  Of interest, it is not essential under the

modifications incorporating the handling both of spontaneous response and
        •«             •       .         i
multiple dose levels that control data be available.  If the data available

over several non-zero dose levels suggest that there is spontaneous response

-------
                                  8 .•,£-'/   ;




this will be automatically weighed into the analysis.




     There is one basic way' in which both standard" and' up-dated; Mantel^




Bryan procedures remain the same relative to the use"of data at several




dose levels.  Suppose we have determined the "safe" level using all the




data up to a certain dose level, and then add the'data for the next  ,




higher dose level, calculating a new "safe" level.  If the new "safe"



level is higher we accept 'it, otherwise not.The  logic here is that the




assumed probit slope of unity is conservatively shallow, the true slope




in the observable region somewhat steeper.  A decrease in the "safe"  •




level can accordingly be explained away-—but an increase is accepted




because this means that the added information relative to safety has




outweighed the tendency for use of higher dose level data to" reduce the



calculated "safe" dose.       '                              .   • - ;    t"




     this rule for using higher "safe" levels when they arise, discard-



ing lower ones must be used with judgement,  the rule should not be" used



where the increase has occurred because the response at higher doses has



flattened out or turned down—it should be reserved: for the case where



the increase in "safe" dose has occurred despite a rapidly rising response



rate with dose.  As I noted above, and will refer "to again below, the




case of DDT  elicited lung tumors is one whe're responses may turn down




at high levels because of earlier-appearing liver  tumors.  For  the



liver tumors elicited by high dosesi of 'dleldr in '"intone "experiment   ;




there also was a reduction,"'but this perhaps could have been "avoided by




reducing further the number of mice considered as  at'risk.



     In general the "safe" dose procedure"has to be used with care and '




with thoughtful interpretation.5" it"'mayF'be7" for1 example", if'an



extended time study is conducted with  late-appearirtg tumors  getting



full weight that an agent may appear tumorigenic by virtue of its life-

-------
prolonging qualities.  This is the reverse of the situation where tumor-

igenicity is masked by early lethal toxicity.  It is thus important

to check whether higher tumor frequencies in treated animals might not

be due to increased longevity, whether only a chance increase or one

actually attributable to treatment.  This is an added reason for not

giving great emphasis to late-appearing tumors	the other reasons

are that they magnify the spontaneous-response problem and that anyway

they are less important.

     Up-dating of the Mantel-Bryan procedure is characterized by a

novel feature—it permits handling independent sets of data.  If an

agent has been tested independently on two or even more occasions,

how do you get the "safe" dose?  If you calculate a "safe" dose for

each separate test, which do you use or how do you use them?  If you

calculate one "safe" dose for males and one for females using data
                                    i

obtained at the same occasion, how use or combine them?  It would

seem improper to use the higher of the separate "safe" doses, because

this may be too high for the other sex—also biases could result

if we permitted repeated testing, then selecting the highest

individual "safe" dose.  Yet if individually calculated "safe"

doses are about equal, somehow the appropriate pooled "safe" dose

should be higher than either one alone because of the reduced
                                           l
statistical variability and increased  Information available.

     Our new feature accomplishes this kind of thing.  If we want

to come out with a single number for the I'safe" dose albeit we

have data from several experiments, we should postulate that a

single common '"safe" dose exists, then proceed to estimate it,
                   >                1      »,

then finally putting a statistical lower limit on that estimate.

-------
                                10


This we do by fitting a joint model to all the data under which the

intercept parameter is common to the various data sets but each

data set has its own spontaneous response parameter — the slope

parameter remains fixed for all the data sets at its specified  *•

conservatively low value.

     In application to the DDT and dieldrin data described above,

the combined "safe" dose was consistently higher than the (geometric)

average of the separate "safe" doses, and substantially higher than

either "safe" dose when the individual "safe" doses were about equal.

As between sexes, note that the model postulates males and females

to differ only in their spontaneous tumor rates but to be identical

in their "safe" dose values	if this premise is unacceptable, then
  *
separate "safe" doses would be required and in practice the lower of

the two would have to be used.  An incidental consequence of the

added feature is that it can permit combining results, but with

caution, from several test species if we are willing to commit

ourselves to a particular way of scaling dosages, e.g. mg/m^	

milligrams of agent per square meter of surface area of the test
                                                                   i
animal.

RESULTS OF ANALYSIS

     Tables 3 and 4 give certain summary analyses of the DDT and diel-

drin tumorigenicity data.  These summary analyses are premised on use

of a unit probit slope, an assurance level of 99% and a "safety"

level of 10~* as initially illustrated by Mautel and Bryan.   Results

of additional computations and manipulations will be given in a

supplementary report rather than in an appendix to the present one,*

-------
                                11
so that it can be self-contained.


     "Safe" levels shown in Tables 3 and 4 are expressed in parts per


trillion  (parts per lO^ or ppt),  These can be converted to "10   or 10"


"safe" levels by multiplying respectively by either 2.59 or 7.23.


Use of a higher probit slope of 1.5 would likely result in "safe"

levels upwards of 20 times greater, the respective multipilication


factors for 10~7 or 10~6 risk then being 1.88 and 3.74.  I would


recommend against, use of such a steeper slope as being inadequately

                                         Q
conservative.  With a slope of 1, 'the 10   risk dose level is


l/1930th the 1% risk dose level.  If we should have used the one-hit

                                       _Q         ' .
model for extrapolation our apparent 10- ° risk dose would represent
                     ' •                     I  ;'"•.-

a risk of 1/193,000. With a change in slope to 1.5, the 10"8 risk


dose level is l/l$5th the 1% risk dose so that the true risk if the


one-hit model obtained would be 1/15,500.  (For a slope of 2 the


fraction and one-hit model risk would be 1/44 and 1/4400 respec-


tively.  Note that use of a slope somewhere between 1.5 and 2 gives
                                            1  i
results on the order of the standard practice of using 1% of the
                                           ' ••,    )   *

apparent no risk dose.)  Another modification which could lead to
               %       '                    ..  » •

an increase in the"saf"e," dose would be that of -reducing the


assurance level, say from 99% to 95%.


     Inspection of Table 3 shows that the combined data "safe" dose
                                              1    '

of DDT relative to' liver tumors ranges from 5Q6-2390 ppt depending
                     < * ;           '' '       •   ' !  *'

on whether all tumors or only tumors within 99 weeks are considered
                  <.
               .            '                  •* '  .   •
and on whether the risks for all mice or only for those surviving

70 weeks are considered.  As illustration of  how the procedure


combines independent data I show in Table 3 the separate results


for male and female mice.  The combined "safe" level is much closer
                  |J  .  i                     '
to the higher female values of 670-3270 ppt than they are to those

-------
                               12

          1
lor male mice, 50-420 ppt. The lower "safe" level values for male mice

do not   necessarly presage greater vulnerability of males'to DDT;—'—It,

could also reflect that in some way the data for male mice are less

informative.  A possible way for such reduced information to come, about

is through a high spontaneous rate, a situation true for males but not

for females, as then the effect of treatment cannot be measured as

precisely. Calculated estimates of the spontaneous rates associated with

each set of data analyzed are also shown as a matter of interest—these

should be given limited credence since they are based on use of all,  the

cUta and with the acceptance of a unit probit slope.  Inclusion of the   ••  -

data at the dose level of 250 ppm would have resulted in overestimates

of the spontaneous rates as the responses at that dose level were higher

than the unit slope would have called for. In support of this, in 11  of

12 instances related to liver tumors the highest "safe" dose, level of

DDT resulted with the analysis of data up to 50 ppm; the dose level  at .-.„..,

which this maximum occurred is designated in Table 3 as the '-highest.'. ,

influential dose", signifying that incorporation of data at the next

higher dose level would have resulted in a  reduction in the calculated

"safe";,dose.                                               , ,  ......  ,

     In the lower half of Table 3 I show  the "safe" levels for lung  tumors,   TV

combined "safe" levels range from 1820 -  19,100 ppt and I  show them  in

parentheses in order to convey that they  cannot be reliably accepted.

This is due to the fact that at high DDT  levels many mice-getwearly

liver tumors and so fail to display lung  tumors	the  fact that  the

highest influential dose is the highest dose employed  reflects this  ,in

part.  An interesting feature In the lung tumor analysis,  however,.is

that it shows the combined "safe" dose to be higher, sometime much,,  so,

than the separate sex "safe" doses where  the separate  "safe" levels

-------
differ only moderately.

     The dieldrin "safe"  levelg  in Table A relate only, to, the combined

analysis, individual  aqalysep  by sex and/or experiment being reserved
                               .       '        '*,''•,     i
for the supplemental  record,, •  These  "safe" levels are 11 ppt relative

to all tumors, 87 ppt re.lat:J.ve to }.iver tumors showing areas of

papilliform or adenoid growth.   In both cases the "safe" dose occurred
                                       .       ','"•"  ,  -
at an intermediate value  for the highest influential dose,  1.25 ppm
                                                *' • i       i
and 5 ppm respectively, notwithstanding  the reversal  at  high .dose

levels in experiment  2.1,   Th,e shajp increase in response at .10 ppm in the

other two experiments apparently more than compensated for the reversal.

Recall, also, that the authors apparently dropped early deaths from

the number of mice at r:f.sk  and so avoided extre'me reductions in response

rates at high dose levels.  As a poiqt of interest I include' two sets

of estimated spontaneous  tumor rat;e^ in Table 4?  one based on all the
                                             I'*-'.
available data, the other based  only pn data at dose levels up to and
                                  , >          .  ,     •
                                              i
including the highest influential dqse.        '  , .
                       *   • -      *     *'         t ' *
     Let me now come  to question of  hpw the "safe" dose values can be
                       i' V  , ' .' :l  '' \ '  '     •'•.','.   \ • :•         v  • .
                                             '  f '..-''      ,'.".••
used.  First, note th,3t th^y are in  this case expressed in parts per
                    "';••' :•''''   ',   i      '"* '••<'""       •     ^" "
trillion of diet, Tfie dietary  4-nJ;a^e of a mammal is, to an approximations,

proportional to its surface area.  Th,us no further adjustment need be
1            •         ''         '              •:'}'.'-•
made if we extrapolate tQ another species,, as to nan.- There could be some

point, if we wished,  tp use an 4d.d.itipnal safety factor (as distinct

from an extrapolation factor)  to ppyey the possibility that even after

species size adjustment.man is pfi}4 more sensitive than mouse.  (There

is a curiosity about  use  of one-hit  !!safety'' determination methods.   By

its underlying logic  the  absolute s$fe dope in man should be the same

-------
                               14






as for mouse with no further adjustment permitted for species size.  In




fact the absolute daily "safe" dose would be  lower  for a man           ,




than for a mouse because of his greater longevity	lifetime "safe" doses




would remain the same.)




     Daily food and water intakes of a mammal are, in many cases,




 approximately  equal.  Thus  the  dietary "safe" levels  can be interpreted



as drinking water "safe" levels if drinking water is anticipated to be




the source of human exposure to DDT or dieldrin.  But a "safe" level in




water actually drunk, which is limited in quantity, should not be loosely




interpreted as a permitted level in water available for drinking, as in



water supply systems,which  is a much larger quantity.  Much less should




it be interpreted as a permitted level in water supply sources which




vastly exceed the amount of water actually drunk.  The adverse effects




of permitting in all fresh water bodies in the United States chemical levels



seemingly safe for humans is not immediately predictable.  One likely




consequence could be extreme harm for fish and other animals inhabiting




the water since their total exposure at a given water concentration




would be much higher than for man or other animals which consume only




limited amounts of water each day.

-------
Table 1. Data from Tomatis, et al., on t|u, occurrence of liver and of lung tumors in CF-1 mice receiving vatving levels
of dietary DDT, parent and first noneration offspring combined. Results are shown separately for all mice
and for those surviving to the 70th week of age.
All Mice

DDT Dietary
Dose Level
pom

- o
2
10
50
250

0
2
10
50
: 250
A





Number at .
Risk


125
126
111
135
1 117

117
110
126
109
T.05





With Liver Tumors
1
Within 99
Weeks

7
22
25
31
72

0
1
1
3
43





At Any
Time
With Lune Tumors

Within 99
Weeks
Male Data Only
24 18
57 34
50 33
66 36
81 35
Female Data Only
4.
4
11
13
60





12
11
23
' 21
,20






At Any
Time

42
70
58
64
42

35
30
50
40
30
A




Mice surviving to week 70 only

Number at
Risk


94
104
85
110 '
72

94
'Jl
107
-' 85
76





With Liver Tumors
1
Within 99
Weeks

7
20
25
30
60

0
1
1
2
39


•



At Any
Time

24
55
50
65
69

4
4
11
12
56




•
With Lune Tumors •
Within
99 | At Any
Weeks

16
27
25
34
29

8
10
21
18 -
18
•




Time

40
63
50
62
36

31
29 - '
48
-37
28 i
1
1
:
j
1


-------
Table 2. Reconstructed data frota Walker et a.1. on the occurrence of liver tumors in CF-1 nice receiving varying Bevels
of dietary dieldrin. Results are shown separately for three exper liticn.cs.

Dieldrin
Dietary Dose
Level, ppm






0
0.1
1
: 1.25
• 2.5
5
10
20
£
7(-
0
0.1
1
1.25
2.5
Experiment 1

Number at
Risk
,/ '





288 [
124 i'
111
__ • .
__ '• ' :
~»—
176 r
• 	 '•'' .


297
90
87
___
___
S «-T— —
10 i 148
20


-^—



Liver Tumors

Total





58
32
34
	
__
- _
165
; 	


39
I 25
.! 32
!\ 	
— — -
^ _-«~»
? 136
„ —-*-»*•
:1
i
With
Papilli-
form and
Adenoid
Growth

Experiment 2.1

Number at
Risk





Male Data Only
12
5
9
	 .
	 	
	
100
- —

Fema
0
4
5
	
	
__«.
81
*^""^^

i
78
—
—
30
30
30
11
17

le Data Only
78
	 ' '
' 	 *
30
28
30
17
21


; i
:

Liver Tumors

Total





9
—
	
6
13 .
26
5
12


8
—
	
5
12
.5 ,18 : rj f
•'• •; 9
8




With
Papillifonn
and Adenoid
Growth



0
	
	
2
1
3
1
9


0
— —
	
0
1
.5. ' •
•• >. • 2 ;
3





Experiment 2.2


Number at
Risk





77
—
—
	
: 	
-. '• 	
56
"• '


74


-
— — ;
'; ; ; ..• . ••
is. ^ 65






Liver Tumors .

Total





25
,— —
— —
—
• . — —
— —
45
•"•• ^

' . : -
12
• .•
• ' ,'
~
". ' T • ' •
' / '
36

r.



With
Papilli-
fonn and
Adenoid
Growth
i
i

2 ' i
— — ]
•™— • ;
• '
•
.
13
( ^"^^
:;• ;; ' >• ,
;-, I \ >
';. ,;0.::.;': '
• ' ' r ~7" '. - '
	 •r ;'.'; !
' -* ;• " ' : *
1 ' '~ - [ '
•i— — .
12

••
- '. J

«

-------
Table 3. "Safety" analysis of. the data from Tomatis et al. shown in Table 1. A combined analysis is shown tor males
and females as well as separate sex analyses. Analyses use a slope of unity, an assurance level of 99%,
and a "safety" level of 10~8. . •


Data Analyzed




Liver Tumors
All Mice
Tumors Within 99 Weeks
All Tumors
70 Week Survivors Only
Tumors Within 99 Weeks
All Tumors
"-Lung Tumors
..All Mice
Tumors Within 99 Weeks
All Tumors
70 Week Survivors Only
Tumors Within 99 Weeks
-All Tumors
'
^
*

Combined Sex Analysis

DDT
"Safe"
Dose
ppt '




2390
697

'2130
506


(4540)
:i9,100)

(1820)
(5080)
figures
.ndicate
Highest
Influential
Dose
PPt





50
50

50
-50

-
V 250
250
. •
250
250
.n parenthesi
I in the tex
Estimated Spontaneous
Response Using All
Data %



Males



14.3
34.7

17.5
44.4


23".7
45.0
,
24.5
54.0
is represent
, because c


Females



0
2.4

0
2.3


13.5
32.6

12.8
38.2
"safe" dc
f the higt
Separate Sex Analyses

Males

DDT
"Safe"
Dose
ppt


420
191

271
50


(2920)
(11,600)

(966)
(2690)
se estima
frequenc
Highest
Influen-
tial
Dose
ppm .


50
250

50
50


, 250
250

250
250
:ea which
f of early
Estimated
Spontane-
ous
Rate %



10.3
31.7

10.6
33.2


23.6
45.0

23.5
54.0
ire cons id
liver tun

Females

DDT
"Safe"
Dose
PPt


3270
891

3110
670


(3590)
(7400)

(1800)
(2650)
•red unre
ors.
Highest
Influential
Dose
ppm



50
50

50
50


.250
250

250
250
Liable, as .

Estimated
Spontane-
ous
Rate %

•

0
2.7

0
2.9
•

13.6
32.6.

13.4
38.2
^


-------
Table 4.  "Safety" analysis of the data from Walker et al.shown in
          table 2.  Only the combined analysis for both sexes, all
          three experiments, is shown.  Analysis uses a slope of unity,
          assurance level of 99%, and a "safety" level of 10"^.
                                    All Liver      Tumors With Papilliform
                                     Tumors        and Adenoid Growth
Dieldrin "Safe" Dose, ppt               11                     87
Highest Influential Dose, ppm            1.25                   5
Estimated Spontaneous Rates, %
     Using All Data
Expt. 1, Males                          20.3                    4.4
         Females                        15.6                    0.3
Expt. 2.1, Males                         9.6                    0
         Females                         6.9                    0
Expt 2.2, Males                         32.6                    2.0
          Females                        13.3                    0
Up To Influential Dose Only
Expt. 1, Males                          20.7                    4.2
         Females                        16.7                    0.9
Expt. 2.1, Males                        10.5                    0
         Females                         9.0                    0
Expt. 2.2, Males                        32.5*                   2.6*
         Females                        16.2*                   0*
*Represents tumor rates among controls.  There were no actual  test  groups
in this experiment below the highest  influential dose.

-------
ESTIMATING LIMITING RISK LEVELS FROM ORALLY INGESTED



    DDT AND DIELDRIN USING AN UP-DATED VERSION



           OF  THE MANTEL-BRYAN PROCEPURE



              (SUPPLEMENTARY REPORT)
                        23<

-------
ESTIMATING LIMITING RISK LEVELS FROM ORALLY INGESTED DOT AND DIELDRIN

       USING AN UP-DATED VERSION OF THE MANTEL-BRYAN PROCEDURE
                            Nathan Mantel
                 Research Professor of Biostatisties
                    George Washington University
                      7979 Old Georgetown  Road
                      Bethesda, Maryland  200T4
                          prepared for the
                     Office of Toxic Substances
                   Environmental  Protection Agency
                       Washington,.D.C.   20460
                           Project Officer
                          Michael  J.  Prival
                      Contract # P5-01-2598-J

                          December 27,,  1974

-------
                     NOTICE
This report has been reviewed  by  the Office of
Toxic Substances,  EPA,  and  approved for publication.
Approval  does not  signify that the contents neces-
sarily reflect the views  and policies of the
Environmental Protection  Agency,  nor does mention
of trade names or  commercial products constitute
endorsement or recommendation  for use.
                            25<

-------
   the risk level, specifically 10"8, 10~7, 10~6, 10~5.  For each of the
     The accompanying tables show the results of application of the "safe"*

dose determination procedure with varying specification of the assurance

level (95% or 99%) and varying specification of the postulated shallow

slope (1.0 or 1.5 normal deviates per 10-fold dose increase).  Yet another

variation, but one readily obtainable, was that of alternative limitations

on

two agents overall "safe" levels were determined using all available data

and separate "safe" levels were determined for certain data subsets.

Relative to DDT there were only two data subsets, male or female.  But

for dieldrin there were eleven such data subsets—two for sexes,  three for
            ^-,
experiments, six for each sex in each experiment.

     Other variants in the analyses made have been indicated in the

initial report.  For DDT the variants included whether data for all

animals or only for 70-week survivors should be considered, and whether

all tumors or only those occurring within 99 weeks should be considered.

In the present report relative to DDT, I am showing "safe" level deter-

minations only for liver tumors, since I do not consider the lung tumor

determinations to be reliable, for reasons indicated in my initial report—at

increasing dose levels lung tumor rates fail to increase because of the

earlier occurrence of liver tumors.  The added variant relative to

dieldrin was only that of alternatively considering all liver tumors

or only those liver tumors displaying areas of papilliform or adenoid

growth.
*The word '-'safe" as used in this report is not necessarily meant 1.0 imply
 complete absence of hazard or risk.  The "safe" dosages obtained are
 ones for which, by the procedures used and under the assumptions riiac.;;,
 the true risk does not exceed some pre-specified low level.

-------
                                    -2-






     The several variant analyses of the DDT liver tumor data are given




in tables 1A, IB, 1C, and ID, the tables varying in the postulated




assurance levels and slopes.  The column headings in the tables indicate




-./hether results relate to combined data or to sex-specific data, whether




all tumors or only early tumors are considered, and whether data for all




animals or only for 70-week survivors are considered.  Tables 2A, 2B, 2C,




and 2D relate similarly to the dieldrin data on all liver tumors, the




column headings indicating to which subset of the data the results shown




relate.  This pattern is repeated in tables 3A, 3B, 3C, and 3D which




related to the dieldrin data on liver tumors showing papilliform and




adenoid growth.




     Specifically what is shown in the various tables is this.  All the




non-zero dose levels, in parts per million (ppm) in the experiment or




experiments are arranged in ascending order at the left end of the




table.  The row entry in a column represents the computed upper limit on




the intercept parameter (see initial report) using data obtaining for




dose levels at or below the dose level shown for that row—the entry,




however, relates to the analysis indicated by the column heading.  Where




no entry would be appropriate because of unavailability of data a dash




is used.  The dash is also used to indicate instances where the entry




should be the same as at a preceding lower dose level, since no new data




have been introduced which could influence the analysis.  In each column,




the lowest of the upper limits on the intercept parameter has been




routinely (but see discussion below) underlined. Those lowest upper




limits on the intercept parameter are decoded into "safe" level estimates.



expressed in parts per trillion (ppt) which are shown in the lower portion

-------
                                    —3—





of each table, for the alternative limitations on the risk level as

                 \

designated in the left most column of the table.  No attempt is made in



this report to give estimates of the spontaneous tumor rate for each



experiment, though such estimates are available if desired.  My reason



for this is the multiplicity of such estimates which can be made.  A



different estimate can be made for each analysis into which a data set



i3 entered, for each different progressively increasing dose level,



and for each of the two specified slope values.



DISCUSSION OF DDT RESULTS. TABLES 1A. IB. 1C. ID.



     Table 1A duplicates in large part the liver tumor results given in



table three of the Initial report.  Overall "safe" doses at the 10~8 risk



level range from 500-2390ppt depending on which of the four alternative sets



of data is considered.  The overall "safe" doses are heavily weighted



toward the higher female "safe" doses which range between 670-3270 ppt, while



the lower male "safe" levels range between 50-420 ppt.  (Note that in



only 1 of 12 instances does the "safe" dose shown result from the



inclusion of data at the highest dose level of 250 ppm.)  As the limiting


                                         —8      —5
risk level is increased in stages from 10   to 10  , the "safe" level can



l.e seen to increase by a factor of approximately 3 for each 10-foid



Increase in risk.



     The altered assurance level of 95% used in Table IB produces no



great changes from the results of Table 1A—typically the "safe" levels



of Table IB are about 30% larger than those of Table 1A.  Much more



profound effects on the "spfa" dose are evidenced in Table 1C in which



the slope parameter has been increased from 1 to 1.5.  "Safe" doses are



i-.ow 20 to 100 times greater than those of Table 1A at the 10~?  risk level.
                                   28<

-------
                                    —3—





of each table, for the alternative limitations on the risk level as

                 \

designated in the left most column of the table.  No attempt is made in



this report to give estimates of the spontaneous tumor rate for each



experiment, though such estimates are available if desired.  My reason



for this is the multiplicity of such estimates which can be made.  A



different estimate can be made for each analysis into which a data set



is entered, for each different progressively increasing dose level,



and for each of the two specified slope values.



DISCUSSION OF DDT RESULTS. TABLES 1A. IB, 1C. ID.



     Table 1A duplicates in large part the liver tumor results given in



table thre€ of the initial report.  Overall "safe" doses at the 10"8 risk



level range from 500-2390ppt depending on which of the four alternatiye sets



of data is considered.  The overall "safe" doses are heavily weighted



toward the higher female "safe" doses which range between 670-3270 ppt,  while



the lower male "safe" levels range between 50-420 ppt.  (Note that in



only 1 of 12 instances does the "safe" dose shown result from the



inclusion of data at the highest dose level of 250 ppm.)  As the limiting


                                         —8      —5
risk level is increased in stages from 10   to 10  , the "safe" level can



le seen to increase by a factor of approximately 3 for each 10-foid



increase in risk.



     The altered assurance level of 95% used in Table IB produces no



great changes from the results of Table 1A—typically the "safe" levels



of Table IB are about 30% larger than those of Table 1A.  Much wore



t""found effects on the "s?fa" dose are evidenced in Table 1C ir. which



the slope parameter has been increased from 1 to 1.5.  "Safe" clones are



iv.iv 20 to 100 times greater than those of Table 1A at the 10"  risk level.

-------
                                    -4-






At higher risk levels the multiplicative factor is somewhat reduced since




at the slope of 1.5, each 10-fold increase in the animal risk level results




in only about a 2-fold increase in the "safe" dose.  (In these analyses




with a slope of 1.5 there are 2 added instances in which the "safe" dose




results from inclusion of the data at 250 ppm, but these are marginal cases,




and, in fact, disappear in the Table ID analyses in which the assurance




level is changed to 95%.)  When the assurance level is altered to 95%




using the 1.5 slope, the effect, as shown in Table ID, is even more moderate




than before, the "safe" dose level being increased by only about 20%, or




even less.




DISCUSSION OF DIELDR1N RESULTS. ALL LiVER TUMORS. TABLES 2A, 2B, 2C. 2D.




     The overall "safe" dose at the 10~8 risk level shown in Table 2A is




11 ppt  with only a moderate difference between 13 ppt for males, 6 ppt



for females.  The range of the variously determined "safe" levels is from




2-16 ppt. In only one instance is the "safe" level associated with the




inclusion of data at the highest of several dose levels employed.  This




instance relates to the data of experiment 2.1 which, as described in




my first report, show  a questionable inversion at the two highest dose




levels.  Since in experiment 2.2 only a single non-zero dose level was




employed, any "safe" levels calculated specifically for that experiment




must employ data at the highest dose level therein.  And as the overall




calculated "safe" dose results from data at dose levels of 1.25 ppm. or less,




there can be no ambiguous effects due to data inversion.




     As in the case of the DDT data analysis (Table 1A), each 10-fold




increase in the risk level results in about a 3-fold increase in the "safe"




level.  Table 2B, shows, as before, that change in the assurance level to




95% produces only a moderate increase in the "safe" level,  again about  30%.

-------
                                    -5-


Changing of the slope parameter, Table 2C, again produces dramatic Increases

in the "safe" dose, in one instance by a factor in excess of 100.  The in-

crease in the "safe" dose per 10-fold increase in the risk level is again

about 2 at the higher slope of 1.5.  A particular effect of use of the

higher slope is that it has allowed the data inversions of experiment 2,1

to dominate certain analyses.  Thus, all "safe" dose determinations for

experiment 2.1 and the overall female "safe" dose are based on the use

of all data including that at the highest dose level of 20 ppm.  An

anomalous consequence is that the overall "safe" dose of 580 ppt  is lower

than that of either 590 ppt  for males or 620 ppt  for females.  The

anomaly disappears, however, if we discount the high "safe" level for females

as attributable in part to the data inversion at the highest dose levels of

experiment 2.1.   With such discounting the overall female "safe" dose would

be reduced to 450 ppt.  The same kind of pattern occurs in Table 2D in

which the assurance level has been altered to 95%, the effect of the change

in assurance level being only to increase the "safe" levels by a moderate

10-20%.

DISCUSSION OF D1ELDR1N RESULTS. LIVER TUMORS SHOWING AREAS  OF PAPILLIFORM
AND ADENOID GROWTH. TABLES 3A. 3B. 3C. 3D.

     The results of these analyses parallel those for all liver tumors

obtained with dietary dieldrin, but at somewhat increased "safe" levels.

This  at  the  10   risk level with slope of one the overall "safe" dose

is raised from 11 to 87 ppt,  while that for males is  raised from 13

to 78 ppt,  and that for females from 6 to 57 ppt.  There is again a

pattern  of moderately increased "safe" doses as the assurance level is

changed  to 95%,  of sharply increased "safe" doses as the  slope is  in-

creased  to 1.5 and of 2 or 3-fold increases in the "safe" dose with

10-fold  increases in the risk level.  Again the data Inversions of


                                   30<

-------
                             -6-






experiment 2.1 give rise co some instances of estimated "safe" doses




which should perhaps be discounted, but these are not so pronounced as




in the preceding analyses.




GENERAL DISCUSSION AND SOME INTERPRETATIONS




     From the foregoing the pattern of effect of altered assurance




levels, slope parameters, or risk levels seems somewhat evident and I




will not dwell on these here.  A particularly interesting point however,




to which I specifically alluded in my first report, was the large




difference between males and females in their "safe" DDT doses, with the




combined "safe" dose somewhat closer to the higher of the two.  I suggested




that this could reflect a difference in the relative amounts of information




available for the two sexes, rather than any true difference in their
          «



susceptibilities to DDT.




     I give next a simple hypothetical example which brings out my point.




It consists of five experiments in whicli the control spontaneous rate,




based on 50 animals, increases progressively from 0 to 50%, while the




apparent induced rate (measured by  (Pt - PC)/(I - PC)) remains constant




at 10% .  The following table shows the "safe" doses for the separate




experiments and the combined "safe" dose as determined by the method




I have used, employing an assurance level of 99%, .a slope of unity, and a




limiting risk level of 10~8.

-------
-.7-
i
Experiment Controls
1 0/50
2 10/50
3 20/50
4 30/50
5 40/50

Treated
lOOppm
5/50
14/50
23/50
32/50
41/50

"Safe"
Dose, ppt
1380
700
400
220
90
Combined                           -                ;                   1700




     What is curious about this result is that the highest "safe" dose




attaches to the first experiment which is the one exhibiting the most




clear evidence for a carcinogenic effect.  Yet only a low "safe" dose




attends the results for experiment 5  which displays questionable evidence




of carcinogenicity.  The explanation is that in the first experiment, though




the data suggest a carcinogenic effect to exist, the upper limit on that




effect is probably limited to something under about 20%, say as due to a




zero control rate and a 20% treated rate.  The upper limit on the tumor




induction rate for experiment 5 is considerably higher—the data, for




example, are not too inconsistent with true rates of 70% for controls,




90% for treated, for an induction rate of 67%.  The high spontaneous rate




in experiment 5 led to such an imprecise estimate of the induced rate that




the data in it are consistent with the possibility either of a zero   ,,




induction rate and of a rather high induction rate.  Interestingly, the




combined "safe" dose is higher than the individual experiment "safe" doses-




in fact the data for the last few experiments, while not so informative




In themselves, did add to the total information available and so did serve




to raise the "safe" dose estimate.

-------
                                     -8-



     The lesson here is the need to avoid un in formative experiments in


establishing "safe" doses.  If one's interest is in obscuring carcinogenic
                     <*

effects, an experiment which is less informative because of the obscuring


effect of a high spontaneous rate can be ideal	but by the Mantel-Bryan


approach such uninformative experiments will be penalized by the assignment


of a low "safe" dose.  In a mixture of experiments, some highly informative


others with only minimal information, the up-dated Mantel-Bryan procedure


will be much more influenced by the high "safe" doses of the informative


experiment than by the low ones of the uninformative experiment.  (However


if a low "safe" dose arises in a relatively informative experiment, it


will be given high weight.  This could signify a situation in which there


is a true difference in "safe" doses in the several experiments.)  To


the extent that high spontaneous rates lead to less clear indications of


induced tumor rates, certain kinds of experiments would be best avoided.


These include experiments with animal strains with high spontaneous rates.


Also it would be better not to count tumors which occur very late in


animal  life  (though we can be free to observe- them if we wish).  Such tumors


are, in principle, less important; they give rise to higher spontaneous rates,


hence less information about induced rates and so to reduced "safe" doses;


they could lead to misinterpreting a life-prolonging effect  as  a tumor-

inducing effect.


     An added point of interest I wish to make is the improper high "safe"


dose estimates which could result when data at excessively high dose levels


are included.  I deliberately permitted these to be shown routinely as they


occurred  in  the foregoing analyses so that I might point them put.  If tumor
                                          33<

-------
                                   -9-
rates go down at high doses because of early lethal toxicity,  these reduced



rates should not be given full faith and credit.   Apparently the investigators



in the dieldrin study did make some allowance for early toxicity by excluding



from animals considered at risk those dying early—but their adjustment did



not seem to have been fully adequate.  Note that  in dieldrin experiment 2.2



there was only a single non-zero dose level.  Suppose that in such an



experiment we knew only the final outcome and had no clue as to whether an



observed low tumor rate were due to competing lethal toxicity or whether it



resulted from the test agent being only weakly carcinogenic.  Any "safe"



dose we might calculate would then be necessarily suspect.  The remedy here



is that our experiment should involve a series of dose levels in order to



permit differentiating between true and only apparent low tumor rates.



     Let me note that the feature of the up-dated Mantel-Bryan procedure which



allows combining the data of several experiments  also permits making up-dated



estimates of "safe" doses as the results of new experiments are added to the



old.





                                      34<

-------
Tnblo 1.   Influence of va.yliis HIP Hipp-,  i'.u- nHHc.nini-p Icvpl, mil lli« rink  Irvi'l on tin- "milYly" nnnlyHln of the ilntn from Tomnttn Pt nl., HVM  .
          tiiMir d:it;i inily,  illrMry HOT.
DUT DOSE DATA FOK Al.l, MICK DATA FOK 7(1 WliliK SURVIVORS
!'!!oi>?r " tumors vll'hln 70 wfcku only nil llvrr iiimini ' luimirs wlihln <>') wi'rkH only oil liver tumors

2
10
50
250


10-;
W?
io"u
io°-

2
10
50
2 SO

10-8
10-'
10"*


2
!.0
SO
iO
10-8
10"?
10~5


2
10
50
250

ID*8
10-'
10"*
10-5
both ilexes
combined
-1.719
-2.473
-2.990
-2.587

. l
2390
6180
17.300
53,100

-1.874
-2.606
-3.108
-2.636

3130
8110
22.600
it. TOO

-1.870
-2.967
-3.862
-3.635
68,100
128,000
255,000
539,000

-2.024
-3.124
-3.985
-3.688

82,300
155,000
308, QUO
651,000
nuil en onljj females
. only
-1.059
-1.470
-2.235
-2.131

-1.965
-2.663
-3.127
-2.734

"safe" doses, p
420
1090
.1010
9330
IB upf
-1.163
-1.560
-2.350
-2.200
"81
550
1420
3950
12,200

-1.209
- 1 . UH5
-1.222
-3.260
27,000
50, '100
101,000
214,000

-1. 113
-2.104
-3.348
-3.135

30,900
58.300
116,000
245,000
3270
8470
2 3, .7 00
72,800
er limits on
-2.143
-2.774
-3.265
-2.795
fe'* doses, p
4500
11,600
32.500
100,000
It
-2.116
-1.200
-3.970
-3.762
80,400
152,000
301,000
636,000
ID l
-2.293
-3.405
-4.115
-3.830
"s«f
100,000
189,000
376,000
794,000
both SOXUH] ninliin nnly
' cnmblnril
-1.449
-2.000
-2.455
-2.286

it. DDT, corr
700
1800
5030
15,500
Intercept p
-1.597
-2.101
-2.544
-2.337
it DDT, corr
850
2210
6180
19,000
iipprr 1 Imll
-1.599
-2.49')
-3.110
-3.348
31,000
58,100 .
116,000
245,000
ipper limits
-1.747
-2.608
-3.433
-3.407
;" doses (pp
35,300
66,500
132,000
279,000
-0.400
-0.914
-1.671
-1.894

espondlng t<
190
490
1380
4260
aramcter, a]
-0.499
-1.007
-1.790
-1.986
espondlng to
240
610
' 1710
5260
< tin Inlrrrr
-0.550
-1.450
-2.711
-LA"?
21,400
40,400
80,200
170,000
on Inlorcop
-0.650
-1.578
-2.842
-3.207
t). DDT cotro
24,900
47,000
93,200
197,000
femnlvH
nnly
-1.790
-2.116
-2.562
-2.328

minimum, ut
890
2310
6440
19,800
ope • 1.0,
-1.978
-2.250
-2.668
-2.387
ninimum, utt
1140
2940
8220
23.300
)l irir-imrUT
-1.940
-2.M2
-3.427
-3.346
34,900
65,9111)
111,000
276,000
parameter.
-2.129
-2.734
-3.553
-3.415
^ponding' to
42,400
79,900
158,000
335,000
liotll HUXUH
-1.689
-2.381
-2.941
-2.387

tnnli
-------
NOT REPRODUCIBLE
Table }. Inflnvn 	 r »;irjlnx llu- ulnix-. tin- nnimriiiirv l.-vrl, HIM) Ito rink li'vi'l >m ito "iwtMy" iwiilinlit nf Ito ilolo (ran Vnlkvr I>1 nil '}£?*'
dntit hir nil llv.T UuurH, dlrlnry illHilrlii • • OO
OltUattt
DOSC, py
" or HCfl
Mm

0.1
1.0
1.11
1.3
1
10
10

10-8
w*
w*
10-3
S.I
1.0
1.11
I.I

10
to

in-e
10-7
10-6
10-'


O.I
1.0
1.11
1.1

10
to

io-e
W~6
10-'
10-'

0.1
1.0
1.11
1.1
J
10
:o

jo-»
10 "'
to",
10''
1 OVCRAI.I.
> COHtlKKD
*- -BAtA-— •> .

0.108
-0.411
-0.664
-0.641
-0.111
-0.181
-0.346
*
11
29
•1
IM
0.101
-0.141
-o. ivi
•TO.) 11
-O.MO
-O.)16
-0.1A2
"nnfi.*1* don
' 14
W
99
'wo

2C uppvr 1
0.701
-0.119
-0.741
-0.714
.-0.666
-0.611
-0. 707

160
1010
1160
4)60
20 upper
fl.MH
-0.1,15
-O.HJ1
-O.R28
"0.72)
-O.A61
-0.746

(,50
1220
7410
SHU
OMniNKI) P.XI'KHmnrrs
SBPAHATK »EXKS
7IAI.K
. _A upper
0.1)1
-o.Mi
-0.711
-0.668
-0.411
-0.171
-0.197
"•arc" do
1)
3)
9)
190
O.OJ4
-11. 7MI
-O.MI
-0.111
-0.246
• , ppt dleli
17
4}
114
MO

• UN ml Illtl
0.671 •
-0.611
-0.771
-0.757
-0.171
-0.4W
-0.10)
"..(."
. 110
1110
1110
4680
•II » im li
0.124
TO. 7f>'l
-0.900
-0.860
-0.616
-0.111
-0.117
"»»fe" di
710
1)60
1700
1)10

rtMAU! .
iltl on In
0.418
-O.I06
-0.174 •
•$.4 16
-b.4bi"
-O.M6
-0.394
i. ppt die
t
U
46
140
0. 1)1
-0.2IW
-0. 4Hll
-O.VW
-Q.i'lft
-0. »8
-0.441
rln, rnrreii
• 8
10
17
180

rcrpl p.lr.iiM
o.nn
-0.141 '
-0.444
-0.5*2
-0.1M1
-0.671
— --
•M, ppl 1
610
1160
1)00
4680
rcopt pnr<
O.H71
-0.24



-O.'l'l
„.,,
270
510
1020
2110

KALK



-0.144
"07490
-0.177
-0.281
-O.J4)

9
11
' 61
190
-ojii
-o!l78
-0. 182
-0.4)4

14
16
100
110
1



-O7l9)
-0.619
-0.168
-0.121
-0.641

490
110
1820
)B40



-o7so2
-O."7"l4
-0.471
-0.610
-0.7)6

610
1170
1)20
4110
two*



-07644
-0.118
-0.417
-0.1)0
-0.7)9

1)
M
97
WO
. -0.664
-0.161
-0.616
. -0.8)2

16
46
110
190




-0.612
-0.614
-0.661
-0.718
-1.111

1020

11123
6070



-O.'wn
-O.-'H!
-O.I).'


uon
MAU



-""


07691
—

1
. I
14
44
-
"~
-O^OM
-

1
7
10
6)







-0.40*1
-

140
640
(260
3670 .


"
•—•
— •—
"~"
-07T6)

410
li'j'.l 810
6» J.l 1610
'!4tO S400
1
rnw.1 .


--
*~"
^
"*
-07649
—

11
18
79
• 140
—


Qr7rl
— •

14
17
too
110



"


"•^
r|."Y4"^


1060
1000
)960
8110


— .,

' 	
-.,.
«i7V7;-
-•


.' :. ( n
il'. '••


-------

-------