PROBLEMS IN THE USE OF TOXICOLOGICAL DATA FOR HUMAN RISK ASSESSMENT

Martyn T Smith Ph D

Environmental Health Sciences Program School of Public Health

University of California Berkeley California 94720

Introduction

My goal in this paper is to discuss some of the problems associated

with using toxicological data obtained in animals and microorganisms as

described in the Pesticide Assessment Guidelines Subdivision F

(EPA-5409-82-025 October 1982) for human risk assessment It is not my

purpose to give a complete critical review of the assessment guidelines but

rather to highlight some of the difficulties one faces when trying to use

data obtained from these protocols in assessing risk to humans I propose

to be critical of certain areas in the guidelines but do not wish to insult

anyone associated with their development I wish only to draw attention to

problem areas so that the limitations of the data will be clear to persons

not directly t rained in toxicology It is obvious to me that the guidelines

represent a compromise between rigorous highly specified regulations and

some freedom to perform toxicological experiments The guidelines also

represent state of the art toxicology and it is clear that it is the preshy

sent toxicological procedures that are lacking rather than the guidelines

It should be stated from the outset that the toxicologist is in a

no-win situation with regard to human risk assessment He or she can pershy

1

form a study with only a limited number of animals 1000 being the normal

upper limit Thus the lowest incidence rate for a toxic effect that can be

detected is in the 1-10 range This rate is obviously far too high to be

acceptable to humans if the compound is to be widely distributed in the

environment The toxicologist must therefore use high doses in order to

achieve sensitivity and extrapolate back to determine the effects of low

doses Several mathematical models have been suggested as aids in making

these extrapolations to low doses These models are discussed elsewhere in

this report by Kenneth Bogen Although these models still have their advoshy

cates they are coming increasingly into disrepute largely because the

validity of the assumptions made in their development is highly

questionable I will discuss this in further detail later in this report

The second problem that the toxicologist faces is that the group of

animals he or she uses for the toxicity test tends to be for both experimenshy

tal and financial purposes a genetically inbred homogeneous strain of

small rodent The relevance of toxicity data generated in such a test to

the outbred extremely heterogeneous human population is of course highly

dubious The toxicologist is therefore faced wi th two major extrapolation

problems i e from high doses to low doses and from animal species to man

Moreover the toxicologists problems do not end there One is also

faced with the problem that many chronic effects can take years to manifest

themselves eg cancer The toxicologist therefore resorts to short-term

tests that can give answers within days as to the ability of the compound to

initiate genetic damage Once again however the validity of such tests to

the human situation is very controversial and the subject of increasing

2

debate Much of course has been written in these areas and the reader is

directed to references 1-5 for further discussion I therefore propose to

discuss these problems only in the context of the Pesticide Assessment

Guidelines and will especially concentrate on oncogenicity studies I will

basically list the present difficulties and then discuss how we might overshy

come them with the aim being to develop a streamlined consistent approach

to human risk assessment To start with however I will make some criti shy

cism of the guidelines on acute toxicity and irritation studies (series 81)

and then go onto the problems associated with chronic long term studies

Criticism of the Present Guidelines for Acute Toxicity and Irritation

Studies - Their Relevance to Man

The median lethal dose (LD50) test is advocated as being a useful acute

toxicity study in the present guidelines This test is of course not

intended to provide a predictive value for man but rather to produce a

ballpark figure which enables the toxicologist to rank the toxicity of the

substance Anyone who has ever done this test will however tell you that

they are not readily reproducible Moreover since there are many proceshy

dural differences among laboratories in the manner in which these tests are

performed comparisons of LD50 values are usually very difficult In a

recent collaborative study among laboratories in Europe it was shown that

the LDso can vary up to 8-fold despite identical protocols being used in

each laboratory (6) The LD50 is also of little use for determining dosages

for more long term experiments It therefore seems unjustified to submit

lOOs of ani mals to the rigors of this test without useful information being

obtained Moreover the guidelines require both oral and dermal LD50 tests

3

This seems pointless to me especially in light of the fact that shaved

rodent skin is different from human skin in so many aspects and the applicashy

tion method does not reflect the human exposure situation An acute toxishy

city range- finding study could be performed on relatively few animals using

i p or sc injection so that some knowledge of the chemicals toxicity

could be gained before proceeding to a 7-day test of the type described by

Weil al (7) These tests are also far more reliable than the LD50 for

selecting the critical dosage range for use in subchronic 90-day tests on

the same material I therefore advocate the removal of the LD50 test protoshy

cols and their replacement by a range-finding study (with no doses gt 2

gkg) followed by a 7-day test as described above

It is quite clear that the present methods for testing skin and eye

irritancy in animals are brutal and crude In vitro replacements for these

tests should be sought and incorporated into the guidelines as soon as

possible Their lack of relevance to man is also of deep concern because

of the differences in the skin and eyes of humans and animals (8) New

techniques now allow us to maintain human skin in non-proliferative culture

for several days and techniques for the quantitation of cutaneous toxicity

in vitro are available (9) Extensive validation of these techniques is now

required followed by their incorporation into the guidelines The replaceshy

ment of eye irritancy tests is of course more problematic and the reader

is referred to (8) for further discussion

In conclusion then acute toxicity testing is usually cruel and of

little relevance to man Techniques used in this area must be refined and

applied more sparingly

4

Problems Associated with Chronic and Long- Term Studies

Because of the obvious importance of cancer I propose in this section

to discuss problems chiefly associated with oncogenicity studies and will

split this into two parts namely problems in protocol and problems in

extrapolation

(a) Problems in the Guidelines Protocol

There seems to me to be six problem areas in the protocol described

for oncogenicity studies These areas are

1) Determination of the high dose

2) What dosage levels other than the high dose to use

3) Which animal species and strain to use

4) Standardization of the animal diet and its relevance to the human

diet

5) Housing of the animals

6) Interpretation of the histopathological results

The present guidelines are either very unclear or lack specification in

each of these problem areas What therefore should be done to correct

this in each of the problem areas

1) Determination of the High Dose

The selection of the high test dose is perhaps the most important and

most controversial aspect of carcinogenicity testing This stems primarily

from the use of the maximally tolerated dose in carcinogenic bioassays

5

Very high doses have therefore often been used in this type of bioassay

Doses which in prechronic studies reduce body weight gain by 10-20 but

induce no mortality have been considered appropriate by the National Cancer

Institute This often results in very high and sometimes very toxic exposhy

sure levels which are not compatible with normal lifespan eg the ethyshy

lene dibromide gavage study performed by NCI (10) Overtly toxic test doses

are of little value in quantitatively assessing human risk from low level

exposure if there are differences in metabolism and pharmacokinetics at

various exposure levels Killing of the target cells or animals may also

actually mask a positive effect I am therefore somewhat heartened to read

that the guidelines recommend (p119) that the highest dose level should be

sufficiently high to elicit signs of minimal toxicity without substantially

altering the normal life span This statement is however a little

unclear e g exactly what is minimal toxicity I propose the guidelines

be rewritten so that they are far more specific as follows

The high dose in chronic testing should be defined as one which in a

subchronic study

a) Induces no overt toxicity ie cell death or organ dysfunction

b) Induces no toxic manifestations which could shorten lifespan except

neoplasms

c) Is not detrimental to conception rates or neonatal survival

d) Does not retard weight gain by gt 10

e) Takes into consideration metabolic and pharmacokinetic data

2) Dose Levels

6

The guidelines recommend that at least thre e dose levels should be

used What they do not stipulate is how these doses are related to the high

dose I would recommend the use of a high dose (HD) (determined as

described above) half t he HD and one-quarter the HD at a minimum A

fourth dose at one-eighth the HD is also to be recommended for human risk

extrapolation This relationship should be specifically stated in the

guidelines

3) Diet

No reference is made in the guidelines as to what diet the animals

should be fed during chronic toxicityoncogenicity studies This is an

omission which should be corrected immediately given the strong influence

of diet and especially dietary fat on carcinogenesis (11) Moreover a

diet which reflects the human dietary make-up would also seem appropriate as

would a restriction on cal oric intake since rodents are known to overeat

and reduce their own lifespan when given food ad libitum (12)

4) Animal Species and Strain

The guideline s are very unspecific in this area indicating only that

commonly used strains of rats and mice be used It is essential to use

strains on which good historical data is available e g Fisher 344 rat and

B6C3F mice The guidelines must be specific The use of outbred animals

may more realistically reflect the human population than the above inbred

ones but a high variation in disease incidence among control populations

tends to occur in random bred animals Their use is therefore not recomshy

mended The possibility t hat using mice as well as rats may be redundant

(13) should be investigated further

7

5) Animal Care

Specific instructions regarding animal care should be provided given

the powerful effects of marginal malnourishment and infections on many

bodily functions (14)

6) Pathological Interpretation

I am very concerned at the use of rodent hepatic cell tumors for the

identification of active carcinogens which supposedly work at very very

low levels Major support for the carcinogenicity of PBBs chlorinated

hydrocarbons azo dyes and other chemicals arises from the identification of

mouse hepatomas which show a very high incidence in controls It is now

not deemed sufficient to call something a carcinogen if it induces lung adeshy

nomas or adenocarcinomas in Strain A mice because of the high incidence of

these tumors in controls It therefore also seems unjustified to use the

mouse hepatoma to claim something is carcinogenic when the mice are inishy

tiated promoted or transformed without application of the chemical Nonshy

toxicologists and toxicologists alike must be aware of exactly what tumors

are formed in the experimental animals and their relevance to both

background levels and man The simple plugging in of numbers of tumors

(unspecified) into risk assessment calculations which I have seen so often

in the past is totally unjustified

(b) Difficulties in Extrapolation

As briefly stated in the Introduction several mathematical models are

available for making extrapolations from high doses to low doses It is my

8

opinion however that none of the models presently available have any

scientific validity I base this opinion on the fact that many of the

assumptions made in their development are highly questionable For example

in the determination of the equation for the multistage model (15) it has

been assumed that the relationship between the dose and the number of hits

is always first order regardless of dose and that there is no individual

differences among animals in a given experimental group It is of course

laughable to compare this to humans who are remarkably heterogeneous in

their response to foreign compounds Some models for example the one hit

model also have no bearing on the reality shown by experimental data

Results of the ED01 experiment of the National Center for Toxicological

Research (16) showed a highly non-linear dose response incidence of bladder

tumors in mice sacrificed at 24 months and an analysis of that data (17)

showed a lack of fit p value of less than 0001 for the one-hit model

could go on and on criticizing the models used and modelling in general but

there is no point We simply have to model there is not other way How

then can we do it more reasonably

A better way may be to use several models and compare their goodness of

fit This approach has been advocated by the Food Safety Council (18) It

involves fitting four models the one-hit model Armitage-Doll model

Weibull model and the gamma multihit model to the dose-response data and

determining goodness of fit Each model can then be used to determine virshy

tual safe doses at the risk level of interest The regulator could then

determine from the goodness of fit values which model gives the most

reasonable estimate of risk and determine a safe level also taking into

9

I

account societal usefulness of the chemical We may therefore be willing to

tolerate relatively high levels of a weak carcinogen which appears to have

curvature in its dose-response curve but may wish to completely remove a

strong carcinogen having a more l inear dose-response For further details

see (18)

We are still of course faced with the problem of extrapolation from

animals to man My only comment here is that we must in some way take into

account metabolic and pharmacokinetic data as well as extrapolating on the

basis of body weight rather than surface area or some other parameter The

justifications for this are amply described in (18)

Future Directions - Possible Usefulness of Inter-Risk Comparison

In order to put health risks from chemicals into perspective Crouch and

Wilson (19) have taken a new approach by comparing some measure of risk for

various aspects of everyday life with similar measures of risk from chemishy

cals There are obviously still uncertainties in this inter-risk comparison

procedure but it does provide a badly needed focus as to significant and

insignificant health risks associated with chemicals The procedure

outlined in (19) is still very crude a fact which the authors are only too

aware of and could be much improved Briefly Crouch and Wilson apply one

model (a linear model for relating dose and effect) to a variety of chemishy

cals using the formula

R = 8d

where R = lifetime risk

d average dose rate and

S = carcinogenic potency of the chemical

10

The carcinogenic po t ency (8) is derived from the animal dose-response

data

They then introduce two other factors an interspecies conversion factor

K and a dispersion factor I which is the proportion of total production of

a chemical which is finally absorbed into humans The expected annual

cancer death rate (n) due to an annual production P of a chemical is

therefore

n = $KIP x 0 254

(0254 is the factor used to convert from lifetime risk to annual risk)

This equation does not of course tell us the actual number of likely

cancer deaths but does provide what Crouch and Wilson call a Hazard Index

For piperonyl butuxide for example this Hazard Index is 40 whereas for

endosulfan it is 50000 despite their annual production levels being

approximately the same Differences in carcinogenic potency certainly make

a difference here and would help regulatory agencies focus on potentially

significant hazards to human health from pesticides

The incorporation of much more toxicological data eg mutagenic

potency pharmacokinetics etc into Crouch and Wilsons equations along with

the fact that we are now able to take into account variability in human

exposure using Professqr Spears calculations could mean that excellent

inter-risk comparisons couldmiddot be made This is our present goal and we will

report on it at our earliest opportunity

11

form a study with only a limited number of animals 1000 being the normal

upper limit Thus the lowest incidence rate for a toxic effect that can be

detected is in the 1-10 range This rate is obviously far too high to be

acceptable to humans if the compound is to be widely distributed in the

environment The toxicologist must therefore use high doses in order to

achieve sensitivity and extrapolate back to determine the effects of low

doses Several mathematical models have been suggested as aids in making

these extrapolations to low doses These models are discussed elsewhere in

this report by Kenneth Bogen Although these models still have their advoshy

cates they are coming increasingly into disrepute largely because the

validity of the assumptions made in their development is highly

questionable I will discuss this in further detail later in this report

The second problem that the toxicologist faces is that the group of

animals he or she uses for the toxicity test tends to be for both experimenshy

tal and financial purposes a genetically inbred homogeneous strain of

small rodent The relevance of toxicity data generated in such a test to

the outbred extremely heterogeneous human population is of course highly

dubious The toxicologist is therefore faced wi th two major extrapolation

problems i e from high doses to low doses and from animal species to man

Moreover the toxicologists problems do not end there One is also

faced with the problem that many chronic effects can take years to manifest

themselves eg cancer The toxicologist therefore resorts to short-term

tests that can give answers within days as to the ability of the compound to

initiate genetic damage Once again however the validity of such tests to

the human situation is very controversial and the subject of increasing

2

debate Much of course has been written in these areas and the reader is

directed to references 1-5 for further discussion I therefore propose to

discuss these problems only in the context of the Pesticide Assessment

Guidelines and will especially concentrate on oncogenicity studies I will

basically list the present difficulties and then discuss how we might overshy

come them with the aim being to develop a streamlined consistent approach

to human risk assessment To start with however I will make some criti shy

cism of the guidelines on acute toxicity and irritation studies (series 81)

and then go onto the problems associated with chronic long term studies

Criticism of the Present Guidelines for Acute Toxicity and Irritation

Studies - Their Relevance to Man

The median lethal dose (LD50) test is advocated as being a useful acute

toxicity study in the present guidelines This test is of course not

intended to provide a predictive value for man but rather to produce a

ballpark figure which enables the toxicologist to rank the toxicity of the

substance Anyone who has ever done this test will however tell you that

they are not readily reproducible Moreover since there are many proceshy

dural differences among laboratories in the manner in which these tests are

performed comparisons of LD50 values are usually very difficult In a

recent collaborative study among laboratories in Europe it was shown that

the LDso can vary up to 8-fold despite identical protocols being used in

each laboratory (6) The LD50 is also of little use for determining dosages

for more long term experiments It therefore seems unjustified to submit

lOOs of ani mals to the rigors of this test without useful information being

obtained Moreover the guidelines require both oral and dermal LD50 tests

3

This seems pointless to me especially in light of the fact that shaved

rodent skin is different from human skin in so many aspects and the applicashy

tion method does not reflect the human exposure situation An acute toxishy

city range- finding study could be performed on relatively few animals using

i p or sc injection so that some knowledge of the chemicals toxicity

could be gained before proceeding to a 7-day test of the type described by

Weil al (7) These tests are also far more reliable than the LD50 for

selecting the critical dosage range for use in subchronic 90-day tests on

the same material I therefore advocate the removal of the LD50 test protoshy

cols and their replacement by a range-finding study (with no doses gt 2

gkg) followed by a 7-day test as described above

It is quite clear that the present methods for testing skin and eye

irritancy in animals are brutal and crude In vitro replacements for these

tests should be sought and incorporated into the guidelines as soon as

possible Their lack of relevance to man is also of deep concern because

of the differences in the skin and eyes of humans and animals (8) New

techniques now allow us to maintain human skin in non-proliferative culture

for several days and techniques for the quantitation of cutaneous toxicity

in vitro are available (9) Extensive validation of these techniques is now

required followed by their incorporation into the guidelines The replaceshy

ment of eye irritancy tests is of course more problematic and the reader

is referred to (8) for further discussion

In conclusion then acute toxicity testing is usually cruel and of

little relevance to man Techniques used in this area must be refined and

applied more sparingly

4

Problems Associated with Chronic and Long- Term Studies

Because of the obvious importance of cancer I propose in this section

to discuss problems chiefly associated with oncogenicity studies and will

split this into two parts namely problems in protocol and problems in

extrapolation

(a) Problems in the Guidelines Protocol

There seems to me to be six problem areas in the protocol described

for oncogenicity studies These areas are

1) Determination of the high dose

2) What dosage levels other than the high dose to use

3) Which animal species and strain to use

4) Standardization of the animal diet and its relevance to the human

diet

5) Housing of the animals

6) Interpretation of the histopathological results

The present guidelines are either very unclear or lack specification in

each of these problem areas What therefore should be done to correct

this in each of the problem areas

1) Determination of the High Dose

The selection of the high test dose is perhaps the most important and

most controversial aspect of carcinogenicity testing This stems primarily

from the use of the maximally tolerated dose in carcinogenic bioassays

5

Very high doses have therefore often been used in this type of bioassay

Doses which in prechronic studies reduce body weight gain by 10-20 but

induce no mortality have been considered appropriate by the National Cancer

Institute This often results in very high and sometimes very toxic exposhy

sure levels which are not compatible with normal lifespan eg the ethyshy

lene dibromide gavage study performed by NCI (10) Overtly toxic test doses

are of little value in quantitatively assessing human risk from low level

exposure if there are differences in metabolism and pharmacokinetics at

various exposure levels Killing of the target cells or animals may also

actually mask a positive effect I am therefore somewhat heartened to read

that the guidelines recommend (p119) that the highest dose level should be

sufficiently high to elicit signs of minimal toxicity without substantially

altering the normal life span This statement is however a little

unclear e g exactly what is minimal toxicity I propose the guidelines

be rewritten so that they are far more specific as follows

The high dose in chronic testing should be defined as one which in a

subchronic study

a) Induces no overt toxicity ie cell death or organ dysfunction

b) Induces no toxic manifestations which could shorten lifespan except

neoplasms

c) Is not detrimental to conception rates or neonatal survival

d) Does not retard weight gain by gt 10

e) Takes into consideration metabolic and pharmacokinetic data

2) Dose Levels

6

The guidelines recommend that at least thre e dose levels should be

used What they do not stipulate is how these doses are related to the high

dose I would recommend the use of a high dose (HD) (determined as

described above) half t he HD and one-quarter the HD at a minimum A

fourth dose at one-eighth the HD is also to be recommended for human risk

extrapolation This relationship should be specifically stated in the

guidelines

3) Diet

No reference is made in the guidelines as to what diet the animals

should be fed during chronic toxicityoncogenicity studies This is an

omission which should be corrected immediately given the strong influence

of diet and especially dietary fat on carcinogenesis (11) Moreover a

diet which reflects the human dietary make-up would also seem appropriate as

would a restriction on cal oric intake since rodents are known to overeat

and reduce their own lifespan when given food ad libitum (12)

4) Animal Species and Strain

The guideline s are very unspecific in this area indicating only that

commonly used strains of rats and mice be used It is essential to use

strains on which good historical data is available e g Fisher 344 rat and

B6C3F mice The guidelines must be specific The use of outbred animals

may more realistically reflect the human population than the above inbred

ones but a high variation in disease incidence among control populations

tends to occur in random bred animals Their use is therefore not recomshy

mended The possibility t hat using mice as well as rats may be redundant

(13) should be investigated further

7

5) Animal Care

Specific instructions regarding animal care should be provided given

the powerful effects of marginal malnourishment and infections on many

bodily functions (14)

6) Pathological Interpretation

I am very concerned at the use of rodent hepatic cell tumors for the

identification of active carcinogens which supposedly work at very very

low levels Major support for the carcinogenicity of PBBs chlorinated

hydrocarbons azo dyes and other chemicals arises from the identification of

mouse hepatomas which show a very high incidence in controls It is now

not deemed sufficient to call something a carcinogen if it induces lung adeshy

nomas or adenocarcinomas in Strain A mice because of the high incidence of

these tumors in controls It therefore also seems unjustified to use the

mouse hepatoma to claim something is carcinogenic when the mice are inishy

tiated promoted or transformed without application of the chemical Nonshy

toxicologists and toxicologists alike must be aware of exactly what tumors

are formed in the experimental animals and their relevance to both

background levels and man The simple plugging in of numbers of tumors

(unspecified) into risk assessment calculations which I have seen so often

in the past is totally unjustified

(b) Difficulties in Extrapolation

As briefly stated in the Introduction several mathematical models are

available for making extrapolations from high doses to low doses It is my

8

opinion however that none of the models presently available have any

scientific validity I base this opinion on the fact that many of the

assumptions made in their development are highly questionable For example

in the determination of the equation for the multistage model (15) it has

been assumed that the relationship between the dose and the number of hits

is always first order regardless of dose and that there is no individual

differences among animals in a given experimental group It is of course

laughable to compare this to humans who are remarkably heterogeneous in

their response to foreign compounds Some models for example the one hit

model also have no bearing on the reality shown by experimental data

Results of the ED01 experiment of the National Center for Toxicological

Research (16) showed a highly non-linear dose response incidence of bladder

tumors in mice sacrificed at 24 months and an analysis of that data (17)

showed a lack of fit p value of less than 0001 for the one-hit model

could go on and on criticizing the models used and modelling in general but

there is no point We simply have to model there is not other way How

then can we do it more reasonably

A better way may be to use several models and compare their goodness of

fit This approach has been advocated by the Food Safety Council (18) It

involves fitting four models the one-hit model Armitage-Doll model

Weibull model and the gamma multihit model to the dose-response data and

determining goodness of fit Each model can then be used to determine virshy

tual safe doses at the risk level of interest The regulator could then

determine from the goodness of fit values which model gives the most

reasonable estimate of risk and determine a safe level also taking into

9

I

account societal usefulness of the chemical We may therefore be willing to

tolerate relatively high levels of a weak carcinogen which appears to have

curvature in its dose-response curve but may wish to completely remove a

strong carcinogen having a more l inear dose-response For further details

see (18)

We are still of course faced with the problem of extrapolation from

animals to man My only comment here is that we must in some way take into

account metabolic and pharmacokinetic data as well as extrapolating on the

basis of body weight rather than surface area or some other parameter The

justifications for this are amply described in (18)

Future Directions - Possible Usefulness of Inter-Risk Comparison

In order to put health risks from chemicals into perspective Crouch and

Wilson (19) have taken a new approach by comparing some measure of risk for

various aspects of everyday life with similar measures of risk from chemishy

cals There are obviously still uncertainties in this inter-risk comparison

procedure but it does provide a badly needed focus as to significant and

insignificant health risks associated with chemicals The procedure

outlined in (19) is still very crude a fact which the authors are only too

aware of and could be much improved Briefly Crouch and Wilson apply one

model (a linear model for relating dose and effect) to a variety of chemishy

cals using the formula

R = 8d

where R = lifetime risk

d average dose rate and

S = carcinogenic potency of the chemical

10

The carcinogenic po t ency (8) is derived from the animal dose-response

data

They then introduce two other factors an interspecies conversion factor

K and a dispersion factor I which is the proportion of total production of

a chemical which is finally absorbed into humans The expected annual

cancer death rate (n) due to an annual production P of a chemical is

therefore

n = $KIP x 0 254

(0254 is the factor used to convert from lifetime risk to annual risk)

This equation does not of course tell us the actual number of likely

cancer deaths but does provide what Crouch and Wilson call a Hazard Index

For piperonyl butuxide for example this Hazard Index is 40 whereas for

endosulfan it is 50000 despite their annual production levels being

approximately the same Differences in carcinogenic potency certainly make

a difference here and would help regulatory agencies focus on potentially

significant hazards to human health from pesticides

The incorporation of much more toxicological data eg mutagenic

potency pharmacokinetics etc into Crouch and Wilsons equations along with

the fact that we are now able to take into account variability in human

exposure using Professqr Spears calculations could mean that excellent

inter-risk comparisons couldmiddot be made This is our present goal and we will

report on it at our earliest opportunity

11

debate Much of course has been written in these areas and the reader is

directed to references 1-5 for further discussion I therefore propose to

discuss these problems only in the context of the Pesticide Assessment

Guidelines and will especially concentrate on oncogenicity studies I will

basically list the present difficulties and then discuss how we might overshy

come them with the aim being to develop a streamlined consistent approach

to human risk assessment To start with however I will make some criti shy

cism of the guidelines on acute toxicity and irritation studies (series 81)

and then go onto the problems associated with chronic long term studies

Criticism of the Present Guidelines for Acute Toxicity and Irritation

Studies - Their Relevance to Man

The median lethal dose (LD50) test is advocated as being a useful acute

toxicity study in the present guidelines This test is of course not

intended to provide a predictive value for man but rather to produce a

ballpark figure which enables the toxicologist to rank the toxicity of the

substance Anyone who has ever done this test will however tell you that

they are not readily reproducible Moreover since there are many proceshy

dural differences among laboratories in the manner in which these tests are

performed comparisons of LD50 values are usually very difficult In a

recent collaborative study among laboratories in Europe it was shown that

the LDso can vary up to 8-fold despite identical protocols being used in

each laboratory (6) The LD50 is also of little use for determining dosages

for more long term experiments It therefore seems unjustified to submit

lOOs of ani mals to the rigors of this test without useful information being

obtained Moreover the guidelines require both oral and dermal LD50 tests

3

This seems pointless to me especially in light of the fact that shaved

rodent skin is different from human skin in so many aspects and the applicashy

tion method does not reflect the human exposure situation An acute toxishy

city range- finding study could be performed on relatively few animals using

i p or sc injection so that some knowledge of the chemicals toxicity

could be gained before proceeding to a 7-day test of the type described by

Weil al (7) These tests are also far more reliable than the LD50 for

selecting the critical dosage range for use in subchronic 90-day tests on

the same material I therefore advocate the removal of the LD50 test protoshy

cols and their replacement by a range-finding study (with no doses gt 2

gkg) followed by a 7-day test as described above

It is quite clear that the present methods for testing skin and eye

irritancy in animals are brutal and crude In vitro replacements for these

tests should be sought and incorporated into the guidelines as soon as

possible Their lack of relevance to man is also of deep concern because

of the differences in the skin and eyes of humans and animals (8) New

techniques now allow us to maintain human skin in non-proliferative culture

for several days and techniques for the quantitation of cutaneous toxicity

in vitro are available (9) Extensive validation of these techniques is now

required followed by their incorporation into the guidelines The replaceshy

ment of eye irritancy tests is of course more problematic and the reader

is referred to (8) for further discussion

In conclusion then acute toxicity testing is usually cruel and of

little relevance to man Techniques used in this area must be refined and

applied more sparingly

4

Problems Associated with Chronic and Long- Term Studies

Because of the obvious importance of cancer I propose in this section

to discuss problems chiefly associated with oncogenicity studies and will

split this into two parts namely problems in protocol and problems in

extrapolation

(a) Problems in the Guidelines Protocol

There seems to me to be six problem areas in the protocol described

for oncogenicity studies These areas are

1) Determination of the high dose

2) What dosage levels other than the high dose to use

3) Which animal species and strain to use

4) Standardization of the animal diet and its relevance to the human

diet

5) Housing of the animals

6) Interpretation of the histopathological results

The present guidelines are either very unclear or lack specification in

each of these problem areas What therefore should be done to correct

this in each of the problem areas

1) Determination of the High Dose

The selection of the high test dose is perhaps the most important and

most controversial aspect of carcinogenicity testing This stems primarily

from the use of the maximally tolerated dose in carcinogenic bioassays

5

Very high doses have therefore often been used in this type of bioassay

Doses which in prechronic studies reduce body weight gain by 10-20 but

induce no mortality have been considered appropriate by the National Cancer

Institute This often results in very high and sometimes very toxic exposhy

sure levels which are not compatible with normal lifespan eg the ethyshy

lene dibromide gavage study performed by NCI (10) Overtly toxic test doses

are of little value in quantitatively assessing human risk from low level

exposure if there are differences in metabolism and pharmacokinetics at

various exposure levels Killing of the target cells or animals may also

actually mask a positive effect I am therefore somewhat heartened to read

that the guidelines recommend (p119) that the highest dose level should be

sufficiently high to elicit signs of minimal toxicity without substantially

altering the normal life span This statement is however a little

unclear e g exactly what is minimal toxicity I propose the guidelines

be rewritten so that they are far more specific as follows

The high dose in chronic testing should be defined as one which in a

subchronic study

a) Induces no overt toxicity ie cell death or organ dysfunction

b) Induces no toxic manifestations which could shorten lifespan except

neoplasms

c) Is not detrimental to conception rates or neonatal survival

d) Does not retard weight gain by gt 10

e) Takes into consideration metabolic and pharmacokinetic data

2) Dose Levels

6

The guidelines recommend that at least thre e dose levels should be

used What they do not stipulate is how these doses are related to the high

dose I would recommend the use of a high dose (HD) (determined as

described above) half t he HD and one-quarter the HD at a minimum A

fourth dose at one-eighth the HD is also to be recommended for human risk

extrapolation This relationship should be specifically stated in the

guidelines

3) Diet

No reference is made in the guidelines as to what diet the animals

should be fed during chronic toxicityoncogenicity studies This is an

omission which should be corrected immediately given the strong influence

of diet and especially dietary fat on carcinogenesis (11) Moreover a

diet which reflects the human dietary make-up would also seem appropriate as

would a restriction on cal oric intake since rodents are known to overeat

and reduce their own lifespan when given food ad libitum (12)

4) Animal Species and Strain

The guideline s are very unspecific in this area indicating only that

commonly used strains of rats and mice be used It is essential to use

strains on which good historical data is available e g Fisher 344 rat and

B6C3F mice The guidelines must be specific The use of outbred animals

may more realistically reflect the human population than the above inbred

ones but a high variation in disease incidence among control populations

tends to occur in random bred animals Their use is therefore not recomshy

mended The possibility t hat using mice as well as rats may be redundant

(13) should be investigated further

7

5) Animal Care

Specific instructions regarding animal care should be provided given

the powerful effects of marginal malnourishment and infections on many

bodily functions (14)

6) Pathological Interpretation

I am very concerned at the use of rodent hepatic cell tumors for the

identification of active carcinogens which supposedly work at very very

low levels Major support for the carcinogenicity of PBBs chlorinated

hydrocarbons azo dyes and other chemicals arises from the identification of

mouse hepatomas which show a very high incidence in controls It is now

not deemed sufficient to call something a carcinogen if it induces lung adeshy

nomas or adenocarcinomas in Strain A mice because of the high incidence of

these tumors in controls It therefore also seems unjustified to use the

mouse hepatoma to claim something is carcinogenic when the mice are inishy

tiated promoted or transformed without application of the chemical Nonshy

toxicologists and toxicologists alike must be aware of exactly what tumors

are formed in the experimental animals and their relevance to both

background levels and man The simple plugging in of numbers of tumors

(unspecified) into risk assessment calculations which I have seen so often

in the past is totally unjustified

(b) Difficulties in Extrapolation

As briefly stated in the Introduction several mathematical models are

available for making extrapolations from high doses to low doses It is my

8

opinion however that none of the models presently available have any

scientific validity I base this opinion on the fact that many of the

assumptions made in their development are highly questionable For example

in the determination of the equation for the multistage model (15) it has

been assumed that the relationship between the dose and the number of hits

is always first order regardless of dose and that there is no individual

differences among animals in a given experimental group It is of course

laughable to compare this to humans who are remarkably heterogeneous in

their response to foreign compounds Some models for example the one hit

model also have no bearing on the reality shown by experimental data

Results of the ED01 experiment of the National Center for Toxicological

Research (16) showed a highly non-linear dose response incidence of bladder

tumors in mice sacrificed at 24 months and an analysis of that data (17)

showed a lack of fit p value of less than 0001 for the one-hit model

could go on and on criticizing the models used and modelling in general but

there is no point We simply have to model there is not other way How

then can we do it more reasonably

A better way may be to use several models and compare their goodness of

fit This approach has been advocated by the Food Safety Council (18) It

involves fitting four models the one-hit model Armitage-Doll model

Weibull model and the gamma multihit model to the dose-response data and

determining goodness of fit Each model can then be used to determine virshy

tual safe doses at the risk level of interest The regulator could then

determine from the goodness of fit values which model gives the most

reasonable estimate of risk and determine a safe level also taking into

9

I

account societal usefulness of the chemical We may therefore be willing to

tolerate relatively high levels of a weak carcinogen which appears to have

curvature in its dose-response curve but may wish to completely remove a

strong carcinogen having a more l inear dose-response For further details

see (18)

We are still of course faced with the problem of extrapolation from

animals to man My only comment here is that we must in some way take into

account metabolic and pharmacokinetic data as well as extrapolating on the

basis of body weight rather than surface area or some other parameter The

justifications for this are amply described in (18)

Future Directions - Possible Usefulness of Inter-Risk Comparison

In order to put health risks from chemicals into perspective Crouch and

Wilson (19) have taken a new approach by comparing some measure of risk for

various aspects of everyday life with similar measures of risk from chemishy

cals There are obviously still uncertainties in this inter-risk comparison

procedure but it does provide a badly needed focus as to significant and

insignificant health risks associated with chemicals The procedure

outlined in (19) is still very crude a fact which the authors are only too

aware of and could be much improved Briefly Crouch and Wilson apply one

model (a linear model for relating dose and effect) to a variety of chemishy

cals using the formula

R = 8d

where R = lifetime risk

d average dose rate and

S = carcinogenic potency of the chemical

10

The carcinogenic po t ency (8) is derived from the animal dose-response

data

They then introduce two other factors an interspecies conversion factor

K and a dispersion factor I which is the proportion of total production of

a chemical which is finally absorbed into humans The expected annual

cancer death rate (n) due to an annual production P of a chemical is

therefore

n = $KIP x 0 254

(0254 is the factor used to convert from lifetime risk to annual risk)

This equation does not of course tell us the actual number of likely

cancer deaths but does provide what Crouch and Wilson call a Hazard Index

For piperonyl butuxide for example this Hazard Index is 40 whereas for

endosulfan it is 50000 despite their annual production levels being

approximately the same Differences in carcinogenic potency certainly make

a difference here and would help regulatory agencies focus on potentially

significant hazards to human health from pesticides

The incorporation of much more toxicological data eg mutagenic

potency pharmacokinetics etc into Crouch and Wilsons equations along with

the fact that we are now able to take into account variability in human

exposure using Professqr Spears calculations could mean that excellent

inter-risk comparisons couldmiddot be made This is our present goal and we will

report on it at our earliest opportunity

11

This seems pointless to me especially in light of the fact that shaved

rodent skin is different from human skin in so many aspects and the applicashy

tion method does not reflect the human exposure situation An acute toxishy

city range- finding study could be performed on relatively few animals using

i p or sc injection so that some knowledge of the chemicals toxicity

could be gained before proceeding to a 7-day test of the type described by

Weil al (7) These tests are also far more reliable than the LD50 for

selecting the critical dosage range for use in subchronic 90-day tests on

the same material I therefore advocate the removal of the LD50 test protoshy

cols and their replacement by a range-finding study (with no doses gt 2

gkg) followed by a 7-day test as described above

It is quite clear that the present methods for testing skin and eye

irritancy in animals are brutal and crude In vitro replacements for these

tests should be sought and incorporated into the guidelines as soon as

possible Their lack of relevance to man is also of deep concern because

of the differences in the skin and eyes of humans and animals (8) New

techniques now allow us to maintain human skin in non-proliferative culture

for several days and techniques for the quantitation of cutaneous toxicity

in vitro are available (9) Extensive validation of these techniques is now

required followed by their incorporation into the guidelines The replaceshy

ment of eye irritancy tests is of course more problematic and the reader

is referred to (8) for further discussion

In conclusion then acute toxicity testing is usually cruel and of

little relevance to man Techniques used in this area must be refined and

applied more sparingly

4

Problems Associated with Chronic and Long- Term Studies

Because of the obvious importance of cancer I propose in this section

to discuss problems chiefly associated with oncogenicity studies and will

split this into two parts namely problems in protocol and problems in

extrapolation

(a) Problems in the Guidelines Protocol

There seems to me to be six problem areas in the protocol described

for oncogenicity studies These areas are

1) Determination of the high dose

2) What dosage levels other than the high dose to use

3) Which animal species and strain to use

4) Standardization of the animal diet and its relevance to the human

diet

5) Housing of the animals

6) Interpretation of the histopathological results

The present guidelines are either very unclear or lack specification in

each of these problem areas What therefore should be done to correct

this in each of the problem areas

1) Determination of the High Dose

The selection of the high test dose is perhaps the most important and

most controversial aspect of carcinogenicity testing This stems primarily

from the use of the maximally tolerated dose in carcinogenic bioassays

5

Very high doses have therefore often been used in this type of bioassay

Doses which in prechronic studies reduce body weight gain by 10-20 but

induce no mortality have been considered appropriate by the National Cancer

Institute This often results in very high and sometimes very toxic exposhy

sure levels which are not compatible with normal lifespan eg the ethyshy

lene dibromide gavage study performed by NCI (10) Overtly toxic test doses

are of little value in quantitatively assessing human risk from low level

exposure if there are differences in metabolism and pharmacokinetics at

various exposure levels Killing of the target cells or animals may also

actually mask a positive effect I am therefore somewhat heartened to read

that the guidelines recommend (p119) that the highest dose level should be

sufficiently high to elicit signs of minimal toxicity without substantially

altering the normal life span This statement is however a little

unclear e g exactly what is minimal toxicity I propose the guidelines

be rewritten so that they are far more specific as follows

The high dose in chronic testing should be defined as one which in a

subchronic study

a) Induces no overt toxicity ie cell death or organ dysfunction

b) Induces no toxic manifestations which could shorten lifespan except

neoplasms

c) Is not detrimental to conception rates or neonatal survival

d) Does not retard weight gain by gt 10

e) Takes into consideration metabolic and pharmacokinetic data

2) Dose Levels

6

The guidelines recommend that at least thre e dose levels should be

used What they do not stipulate is how these doses are related to the high

dose I would recommend the use of a high dose (HD) (determined as

described above) half t he HD and one-quarter the HD at a minimum A

fourth dose at one-eighth the HD is also to be recommended for human risk

extrapolation This relationship should be specifically stated in the

guidelines

3) Diet

No reference is made in the guidelines as to what diet the animals

should be fed during chronic toxicityoncogenicity studies This is an

omission which should be corrected immediately given the strong influence

of diet and especially dietary fat on carcinogenesis (11) Moreover a

diet which reflects the human dietary make-up would also seem appropriate as

would a restriction on cal oric intake since rodents are known to overeat

and reduce their own lifespan when given food ad libitum (12)

4) Animal Species and Strain

The guideline s are very unspecific in this area indicating only that

commonly used strains of rats and mice be used It is essential to use

strains on which good historical data is available e g Fisher 344 rat and

B6C3F mice The guidelines must be specific The use of outbred animals

may more realistically reflect the human population than the above inbred

ones but a high variation in disease incidence among control populations

tends to occur in random bred animals Their use is therefore not recomshy

mended The possibility t hat using mice as well as rats may be redundant

(13) should be investigated further

7

5) Animal Care

Specific instructions regarding animal care should be provided given

the powerful effects of marginal malnourishment and infections on many

bodily functions (14)

6) Pathological Interpretation

I am very concerned at the use of rodent hepatic cell tumors for the

identification of active carcinogens which supposedly work at very very

low levels Major support for the carcinogenicity of PBBs chlorinated

hydrocarbons azo dyes and other chemicals arises from the identification of

mouse hepatomas which show a very high incidence in controls It is now

not deemed sufficient to call something a carcinogen if it induces lung adeshy

nomas or adenocarcinomas in Strain A mice because of the high incidence of

these tumors in controls It therefore also seems unjustified to use the

mouse hepatoma to claim something is carcinogenic when the mice are inishy

tiated promoted or transformed without application of the chemical Nonshy

toxicologists and toxicologists alike must be aware of exactly what tumors

are formed in the experimental animals and their relevance to both

background levels and man The simple plugging in of numbers of tumors

(unspecified) into risk assessment calculations which I have seen so often

in the past is totally unjustified

(b) Difficulties in Extrapolation

As briefly stated in the Introduction several mathematical models are

available for making extrapolations from high doses to low doses It is my

8

opinion however that none of the models presently available have any

scientific validity I base this opinion on the fact that many of the

assumptions made in their development are highly questionable For example

in the determination of the equation for the multistage model (15) it has

been assumed that the relationship between the dose and the number of hits

is always first order regardless of dose and that there is no individual

differences among animals in a given experimental group It is of course

laughable to compare this to humans who are remarkably heterogeneous in

their response to foreign compounds Some models for example the one hit

model also have no bearing on the reality shown by experimental data

Results of the ED01 experiment of the National Center for Toxicological

Research (16) showed a highly non-linear dose response incidence of bladder

tumors in mice sacrificed at 24 months and an analysis of that data (17)

showed a lack of fit p value of less than 0001 for the one-hit model

could go on and on criticizing the models used and modelling in general but

there is no point We simply have to model there is not other way How

then can we do it more reasonably

A better way may be to use several models and compare their goodness of

fit This approach has been advocated by the Food Safety Council (18) It

involves fitting four models the one-hit model Armitage-Doll model

Weibull model and the gamma multihit model to the dose-response data and

determining goodness of fit Each model can then be used to determine virshy

tual safe doses at the risk level of interest The regulator could then

determine from the goodness of fit values which model gives the most

reasonable estimate of risk and determine a safe level also taking into

9

I

account societal usefulness of the chemical We may therefore be willing to

tolerate relatively high levels of a weak carcinogen which appears to have

curvature in its dose-response curve but may wish to completely remove a

strong carcinogen having a more l inear dose-response For further details

see (18)

We are still of course faced with the problem of extrapolation from

animals to man My only comment here is that we must in some way take into

account metabolic and pharmacokinetic data as well as extrapolating on the

basis of body weight rather than surface area or some other parameter The

justifications for this are amply described in (18)

Future Directions - Possible Usefulness of Inter-Risk Comparison

In order to put health risks from chemicals into perspective Crouch and

Wilson (19) have taken a new approach by comparing some measure of risk for

various aspects of everyday life with similar measures of risk from chemishy

cals There are obviously still uncertainties in this inter-risk comparison

procedure but it does provide a badly needed focus as to significant and

insignificant health risks associated with chemicals The procedure

outlined in (19) is still very crude a fact which the authors are only too

aware of and could be much improved Briefly Crouch and Wilson apply one

model (a linear model for relating dose and effect) to a variety of chemishy

cals using the formula

R = 8d

where R = lifetime risk

d average dose rate and

S = carcinogenic potency of the chemical

10

The carcinogenic po t ency (8) is derived from the animal dose-response

data

They then introduce two other factors an interspecies conversion factor

K and a dispersion factor I which is the proportion of total production of

a chemical which is finally absorbed into humans The expected annual

cancer death rate (n) due to an annual production P of a chemical is

therefore

n = $KIP x 0 254

(0254 is the factor used to convert from lifetime risk to annual risk)

This equation does not of course tell us the actual number of likely

cancer deaths but does provide what Crouch and Wilson call a Hazard Index

For piperonyl butuxide for example this Hazard Index is 40 whereas for

endosulfan it is 50000 despite their annual production levels being

approximately the same Differences in carcinogenic potency certainly make

a difference here and would help regulatory agencies focus on potentially

significant hazards to human health from pesticides

The incorporation of much more toxicological data eg mutagenic

potency pharmacokinetics etc into Crouch and Wilsons equations along with

the fact that we are now able to take into account variability in human

exposure using Professqr Spears calculations could mean that excellent

inter-risk comparisons couldmiddot be made This is our present goal and we will

report on it at our earliest opportunity

11

Problems Associated with Chronic and Long- Term Studies

Because of the obvious importance of cancer I propose in this section

to discuss problems chiefly associated with oncogenicity studies and will

split this into two parts namely problems in protocol and problems in

extrapolation

(a) Problems in the Guidelines Protocol

There seems to me to be six problem areas in the protocol described

for oncogenicity studies These areas are

1) Determination of the high dose

2) What dosage levels other than the high dose to use

3) Which animal species and strain to use

4) Standardization of the animal diet and its relevance to the human

diet

5) Housing of the animals

6) Interpretation of the histopathological results

The present guidelines are either very unclear or lack specification in

each of these problem areas What therefore should be done to correct

this in each of the problem areas

1) Determination of the High Dose

The selection of the high test dose is perhaps the most important and

most controversial aspect of carcinogenicity testing This stems primarily

from the use of the maximally tolerated dose in carcinogenic bioassays

5

Very high doses have therefore often been used in this type of bioassay

Doses which in prechronic studies reduce body weight gain by 10-20 but

induce no mortality have been considered appropriate by the National Cancer

Institute This often results in very high and sometimes very toxic exposhy

sure levels which are not compatible with normal lifespan eg the ethyshy

lene dibromide gavage study performed by NCI (10) Overtly toxic test doses

are of little value in quantitatively assessing human risk from low level

exposure if there are differences in metabolism and pharmacokinetics at

various exposure levels Killing of the target cells or animals may also

actually mask a positive effect I am therefore somewhat heartened to read

that the guidelines recommend (p119) that the highest dose level should be

sufficiently high to elicit signs of minimal toxicity without substantially

altering the normal life span This statement is however a little

unclear e g exactly what is minimal toxicity I propose the guidelines

be rewritten so that they are far more specific as follows

The high dose in chronic testing should be defined as one which in a

subchronic study

a) Induces no overt toxicity ie cell death or organ dysfunction

b) Induces no toxic manifestations which could shorten lifespan except

neoplasms

c) Is not detrimental to conception rates or neonatal survival

d) Does not retard weight gain by gt 10

e) Takes into consideration metabolic and pharmacokinetic data

2) Dose Levels

6

The guidelines recommend that at least thre e dose levels should be

used What they do not stipulate is how these doses are related to the high

dose I would recommend the use of a high dose (HD) (determined as

described above) half t he HD and one-quarter the HD at a minimum A

fourth dose at one-eighth the HD is also to be recommended for human risk

extrapolation This relationship should be specifically stated in the

guidelines

3) Diet

No reference is made in the guidelines as to what diet the animals

should be fed during chronic toxicityoncogenicity studies This is an

omission which should be corrected immediately given the strong influence

of diet and especially dietary fat on carcinogenesis (11) Moreover a

diet which reflects the human dietary make-up would also seem appropriate as

would a restriction on cal oric intake since rodents are known to overeat

and reduce their own lifespan when given food ad libitum (12)

4) Animal Species and Strain

The guideline s are very unspecific in this area indicating only that

commonly used strains of rats and mice be used It is essential to use

strains on which good historical data is available e g Fisher 344 rat and

B6C3F mice The guidelines must be specific The use of outbred animals

may more realistically reflect the human population than the above inbred

ones but a high variation in disease incidence among control populations

tends to occur in random bred animals Their use is therefore not recomshy

mended The possibility t hat using mice as well as rats may be redundant

(13) should be investigated further

7

5) Animal Care

Specific instructions regarding animal care should be provided given

the powerful effects of marginal malnourishment and infections on many

bodily functions (14)

6) Pathological Interpretation

I am very concerned at the use of rodent hepatic cell tumors for the

identification of active carcinogens which supposedly work at very very

low levels Major support for the carcinogenicity of PBBs chlorinated

hydrocarbons azo dyes and other chemicals arises from the identification of

mouse hepatomas which show a very high incidence in controls It is now

not deemed sufficient to call something a carcinogen if it induces lung adeshy

nomas or adenocarcinomas in Strain A mice because of the high incidence of

these tumors in controls It therefore also seems unjustified to use the

mouse hepatoma to claim something is carcinogenic when the mice are inishy

tiated promoted or transformed without application of the chemical Nonshy

toxicologists and toxicologists alike must be aware of exactly what tumors

are formed in the experimental animals and their relevance to both

background levels and man The simple plugging in of numbers of tumors

(unspecified) into risk assessment calculations which I have seen so often

in the past is totally unjustified

(b) Difficulties in Extrapolation

As briefly stated in the Introduction several mathematical models are

available for making extrapolations from high doses to low doses It is my

8

opinion however that none of the models presently available have any

scientific validity I base this opinion on the fact that many of the

assumptions made in their development are highly questionable For example

in the determination of the equation for the multistage model (15) it has

been assumed that the relationship between the dose and the number of hits

is always first order regardless of dose and that there is no individual

differences among animals in a given experimental group It is of course

laughable to compare this to humans who are remarkably heterogeneous in

their response to foreign compounds Some models for example the one hit

model also have no bearing on the reality shown by experimental data

Results of the ED01 experiment of the National Center for Toxicological

Research (16) showed a highly non-linear dose response incidence of bladder

tumors in mice sacrificed at 24 months and an analysis of that data (17)

showed a lack of fit p value of less than 0001 for the one-hit model

could go on and on criticizing the models used and modelling in general but

there is no point We simply have to model there is not other way How

then can we do it more reasonably

A better way may be to use several models and compare their goodness of

fit This approach has been advocated by the Food Safety Council (18) It

involves fitting four models the one-hit model Armitage-Doll model

Weibull model and the gamma multihit model to the dose-response data and

determining goodness of fit Each model can then be used to determine virshy

tual safe doses at the risk level of interest The regulator could then

determine from the goodness of fit values which model gives the most

reasonable estimate of risk and determine a safe level also taking into

9

I

account societal usefulness of the chemical We may therefore be willing to

tolerate relatively high levels of a weak carcinogen which appears to have

curvature in its dose-response curve but may wish to completely remove a

strong carcinogen having a more l inear dose-response For further details

see (18)

We are still of course faced with the problem of extrapolation from

animals to man My only comment here is that we must in some way take into

account metabolic and pharmacokinetic data as well as extrapolating on the

basis of body weight rather than surface area or some other parameter The

justifications for this are amply described in (18)

Future Directions - Possible Usefulness of Inter-Risk Comparison

In order to put health risks from chemicals into perspective Crouch and

Wilson (19) have taken a new approach by comparing some measure of risk for

various aspects of everyday life with similar measures of risk from chemishy

cals There are obviously still uncertainties in this inter-risk comparison

procedure but it does provide a badly needed focus as to significant and

insignificant health risks associated with chemicals The procedure

outlined in (19) is still very crude a fact which the authors are only too

aware of and could be much improved Briefly Crouch and Wilson apply one

model (a linear model for relating dose and effect) to a variety of chemishy

cals using the formula

R = 8d

where R = lifetime risk

d average dose rate and

S = carcinogenic potency of the chemical

10

The carcinogenic po t ency (8) is derived from the animal dose-response

data

They then introduce two other factors an interspecies conversion factor

K and a dispersion factor I which is the proportion of total production of

a chemical which is finally absorbed into humans The expected annual

cancer death rate (n) due to an annual production P of a chemical is

therefore

n = $KIP x 0 254

(0254 is the factor used to convert from lifetime risk to annual risk)

This equation does not of course tell us the actual number of likely

cancer deaths but does provide what Crouch and Wilson call a Hazard Index

For piperonyl butuxide for example this Hazard Index is 40 whereas for

endosulfan it is 50000 despite their annual production levels being

approximately the same Differences in carcinogenic potency certainly make

a difference here and would help regulatory agencies focus on potentially

significant hazards to human health from pesticides

The incorporation of much more toxicological data eg mutagenic

potency pharmacokinetics etc into Crouch and Wilsons equations along with

the fact that we are now able to take into account variability in human

exposure using Professqr Spears calculations could mean that excellent

inter-risk comparisons couldmiddot be made This is our present goal and we will

report on it at our earliest opportunity

11

Very high doses have therefore often been used in this type of bioassay

Doses which in prechronic studies reduce body weight gain by 10-20 but

induce no mortality have been considered appropriate by the National Cancer

Institute This often results in very high and sometimes very toxic exposhy

sure levels which are not compatible with normal lifespan eg the ethyshy

lene dibromide gavage study performed by NCI (10) Overtly toxic test doses

are of little value in quantitatively assessing human risk from low level

exposure if there are differences in metabolism and pharmacokinetics at

various exposure levels Killing of the target cells or animals may also

actually mask a positive effect I am therefore somewhat heartened to read

that the guidelines recommend (p119) that the highest dose level should be

sufficiently high to elicit signs of minimal toxicity without substantially

altering the normal life span This statement is however a little

unclear e g exactly what is minimal toxicity I propose the guidelines

be rewritten so that they are far more specific as follows

The high dose in chronic testing should be defined as one which in a

subchronic study

a) Induces no overt toxicity ie cell death or organ dysfunction

b) Induces no toxic manifestations which could shorten lifespan except

neoplasms

c) Is not detrimental to conception rates or neonatal survival

d) Does not retard weight gain by gt 10

e) Takes into consideration metabolic and pharmacokinetic data

2) Dose Levels

6

The guidelines recommend that at least thre e dose levels should be

used What they do not stipulate is how these doses are related to the high

dose I would recommend the use of a high dose (HD) (determined as

described above) half t he HD and one-quarter the HD at a minimum A

fourth dose at one-eighth the HD is also to be recommended for human risk

extrapolation This relationship should be specifically stated in the

guidelines

3) Diet

No reference is made in the guidelines as to what diet the animals

should be fed during chronic toxicityoncogenicity studies This is an

omission which should be corrected immediately given the strong influence

of diet and especially dietary fat on carcinogenesis (11) Moreover a

diet which reflects the human dietary make-up would also seem appropriate as

would a restriction on cal oric intake since rodents are known to overeat

and reduce their own lifespan when given food ad libitum (12)

4) Animal Species and Strain

The guideline s are very unspecific in this area indicating only that

commonly used strains of rats and mice be used It is essential to use

strains on which good historical data is available e g Fisher 344 rat and

B6C3F mice The guidelines must be specific The use of outbred animals

may more realistically reflect the human population than the above inbred

ones but a high variation in disease incidence among control populations

tends to occur in random bred animals Their use is therefore not recomshy

mended The possibility t hat using mice as well as rats may be redundant

(13) should be investigated further

7

5) Animal Care

Specific instructions regarding animal care should be provided given

the powerful effects of marginal malnourishment and infections on many

bodily functions (14)

6) Pathological Interpretation

I am very concerned at the use of rodent hepatic cell tumors for the

identification of active carcinogens which supposedly work at very very

low levels Major support for the carcinogenicity of PBBs chlorinated

hydrocarbons azo dyes and other chemicals arises from the identification of

mouse hepatomas which show a very high incidence in controls It is now

not deemed sufficient to call something a carcinogen if it induces lung adeshy

nomas or adenocarcinomas in Strain A mice because of the high incidence of

these tumors in controls It therefore also seems unjustified to use the

mouse hepatoma to claim something is carcinogenic when the mice are inishy

tiated promoted or transformed without application of the chemical Nonshy

toxicologists and toxicologists alike must be aware of exactly what tumors

are formed in the experimental animals and their relevance to both

background levels and man The simple plugging in of numbers of tumors

(unspecified) into risk assessment calculations which I have seen so often

in the past is totally unjustified

(b) Difficulties in Extrapolation

As briefly stated in the Introduction several mathematical models are

available for making extrapolations from high doses to low doses It is my

8

opinion however that none of the models presently available have any

scientific validity I base this opinion on the fact that many of the

assumptions made in their development are highly questionable For example

in the determination of the equation for the multistage model (15) it has

been assumed that the relationship between the dose and the number of hits

is always first order regardless of dose and that there is no individual

differences among animals in a given experimental group It is of course

laughable to compare this to humans who are remarkably heterogeneous in

their response to foreign compounds Some models for example the one hit

model also have no bearing on the reality shown by experimental data

Results of the ED01 experiment of the National Center for Toxicological

Research (16) showed a highly non-linear dose response incidence of bladder

tumors in mice sacrificed at 24 months and an analysis of that data (17)

showed a lack of fit p value of less than 0001 for the one-hit model

could go on and on criticizing the models used and modelling in general but

there is no point We simply have to model there is not other way How

then can we do it more reasonably

A better way may be to use several models and compare their goodness of

fit This approach has been advocated by the Food Safety Council (18) It

involves fitting four models the one-hit model Armitage-Doll model

Weibull model and the gamma multihit model to the dose-response data and

determining goodness of fit Each model can then be used to determine virshy

tual safe doses at the risk level of interest The regulator could then

determine from the goodness of fit values which model gives the most

reasonable estimate of risk and determine a safe level also taking into

9

I

account societal usefulness of the chemical We may therefore be willing to

tolerate relatively high levels of a weak carcinogen which appears to have

curvature in its dose-response curve but may wish to completely remove a

strong carcinogen having a more l inear dose-response For further details

see (18)

We are still of course faced with the problem of extrapolation from

animals to man My only comment here is that we must in some way take into

account metabolic and pharmacokinetic data as well as extrapolating on the

basis of body weight rather than surface area or some other parameter The

justifications for this are amply described in (18)

Future Directions - Possible Usefulness of Inter-Risk Comparison

In order to put health risks from chemicals into perspective Crouch and

Wilson (19) have taken a new approach by comparing some measure of risk for

various aspects of everyday life with similar measures of risk from chemishy

cals There are obviously still uncertainties in this inter-risk comparison

procedure but it does provide a badly needed focus as to significant and

insignificant health risks associated with chemicals The procedure

outlined in (19) is still very crude a fact which the authors are only too

aware of and could be much improved Briefly Crouch and Wilson apply one

model (a linear model for relating dose and effect) to a variety of chemishy

cals using the formula

R = 8d

where R = lifetime risk

d average dose rate and

S = carcinogenic potency of the chemical

10

The carcinogenic po t ency (8) is derived from the animal dose-response

data

They then introduce two other factors an interspecies conversion factor

K and a dispersion factor I which is the proportion of total production of

a chemical which is finally absorbed into humans The expected annual

cancer death rate (n) due to an annual production P of a chemical is

therefore

n = $KIP x 0 254

(0254 is the factor used to convert from lifetime risk to annual risk)

This equation does not of course tell us the actual number of likely

cancer deaths but does provide what Crouch and Wilson call a Hazard Index

For piperonyl butuxide for example this Hazard Index is 40 whereas for

endosulfan it is 50000 despite their annual production levels being

approximately the same Differences in carcinogenic potency certainly make

a difference here and would help regulatory agencies focus on potentially

significant hazards to human health from pesticides

The incorporation of much more toxicological data eg mutagenic

potency pharmacokinetics etc into Crouch and Wilsons equations along with

the fact that we are now able to take into account variability in human

exposure using Professqr Spears calculations could mean that excellent

inter-risk comparisons couldmiddot be made This is our present goal and we will

report on it at our earliest opportunity

11

The guidelines recommend that at least thre e dose levels should be

used What they do not stipulate is how these doses are related to the high

dose I would recommend the use of a high dose (HD) (determined as

described above) half t he HD and one-quarter the HD at a minimum A

fourth dose at one-eighth the HD is also to be recommended for human risk

extrapolation This relationship should be specifically stated in the

guidelines

3) Diet

No reference is made in the guidelines as to what diet the animals

should be fed during chronic toxicityoncogenicity studies This is an

omission which should be corrected immediately given the strong influence

of diet and especially dietary fat on carcinogenesis (11) Moreover a

diet which reflects the human dietary make-up would also seem appropriate as

would a restriction on cal oric intake since rodents are known to overeat

and reduce their own lifespan when given food ad libitum (12)

4) Animal Species and Strain

The guideline s are very unspecific in this area indicating only that

commonly used strains of rats and mice be used It is essential to use

strains on which good historical data is available e g Fisher 344 rat and

B6C3F mice The guidelines must be specific The use of outbred animals

may more realistically reflect the human population than the above inbred

ones but a high variation in disease incidence among control populations

tends to occur in random bred animals Their use is therefore not recomshy

mended The possibility t hat using mice as well as rats may be redundant

(13) should be investigated further

7

5) Animal Care

Specific instructions regarding animal care should be provided given

the powerful effects of marginal malnourishment and infections on many

bodily functions (14)

6) Pathological Interpretation

I am very concerned at the use of rodent hepatic cell tumors for the

identification of active carcinogens which supposedly work at very very

low levels Major support for the carcinogenicity of PBBs chlorinated

hydrocarbons azo dyes and other chemicals arises from the identification of

mouse hepatomas which show a very high incidence in controls It is now

not deemed sufficient to call something a carcinogen if it induces lung adeshy

nomas or adenocarcinomas in Strain A mice because of the high incidence of

these tumors in controls It therefore also seems unjustified to use the

mouse hepatoma to claim something is carcinogenic when the mice are inishy

tiated promoted or transformed without application of the chemical Nonshy

toxicologists and toxicologists alike must be aware of exactly what tumors

are formed in the experimental animals and their relevance to both

background levels and man The simple plugging in of numbers of tumors

(unspecified) into risk assessment calculations which I have seen so often

in the past is totally unjustified

(b) Difficulties in Extrapolation

As briefly stated in the Introduction several mathematical models are

available for making extrapolations from high doses to low doses It is my

8

opinion however that none of the models presently available have any

scientific validity I base this opinion on the fact that many of the

assumptions made in their development are highly questionable For example

in the determination of the equation for the multistage model (15) it has

been assumed that the relationship between the dose and the number of hits

is always first order regardless of dose and that there is no individual

differences among animals in a given experimental group It is of course

laughable to compare this to humans who are remarkably heterogeneous in

their response to foreign compounds Some models for example the one hit

model also have no bearing on the reality shown by experimental data

Results of the ED01 experiment of the National Center for Toxicological

Research (16) showed a highly non-linear dose response incidence of bladder

tumors in mice sacrificed at 24 months and an analysis of that data (17)

showed a lack of fit p value of less than 0001 for the one-hit model

could go on and on criticizing the models used and modelling in general but

there is no point We simply have to model there is not other way How

then can we do it more reasonably

A better way may be to use several models and compare their goodness of

fit This approach has been advocated by the Food Safety Council (18) It

involves fitting four models the one-hit model Armitage-Doll model

Weibull model and the gamma multihit model to the dose-response data and

determining goodness of fit Each model can then be used to determine virshy

tual safe doses at the risk level of interest The regulator could then

determine from the goodness of fit values which model gives the most

reasonable estimate of risk and determine a safe level also taking into

9

I

account societal usefulness of the chemical We may therefore be willing to

tolerate relatively high levels of a weak carcinogen which appears to have

curvature in its dose-response curve but may wish to completely remove a

strong carcinogen having a more l inear dose-response For further details

see (18)

We are still of course faced with the problem of extrapolation from

animals to man My only comment here is that we must in some way take into

account metabolic and pharmacokinetic data as well as extrapolating on the

basis of body weight rather than surface area or some other parameter The

justifications for this are amply described in (18)

Future Directions - Possible Usefulness of Inter-Risk Comparison

In order to put health risks from chemicals into perspective Crouch and

Wilson (19) have taken a new approach by comparing some measure of risk for

various aspects of everyday life with similar measures of risk from chemishy

cals There are obviously still uncertainties in this inter-risk comparison

procedure but it does provide a badly needed focus as to significant and

insignificant health risks associated with chemicals The procedure

outlined in (19) is still very crude a fact which the authors are only too

aware of and could be much improved Briefly Crouch and Wilson apply one

model (a linear model for relating dose and effect) to a variety of chemishy

cals using the formula

R = 8d

where R = lifetime risk

d average dose rate and

S = carcinogenic potency of the chemical

10

The carcinogenic po t ency (8) is derived from the animal dose-response

data

They then introduce two other factors an interspecies conversion factor

K and a dispersion factor I which is the proportion of total production of

a chemical which is finally absorbed into humans The expected annual

cancer death rate (n) due to an annual production P of a chemical is

therefore

n = $KIP x 0 254

(0254 is the factor used to convert from lifetime risk to annual risk)

This equation does not of course tell us the actual number of likely

cancer deaths but does provide what Crouch and Wilson call a Hazard Index

For piperonyl butuxide for example this Hazard Index is 40 whereas for

endosulfan it is 50000 despite their annual production levels being

approximately the same Differences in carcinogenic potency certainly make

a difference here and would help regulatory agencies focus on potentially

significant hazards to human health from pesticides

The incorporation of much more toxicological data eg mutagenic

potency pharmacokinetics etc into Crouch and Wilsons equations along with

the fact that we are now able to take into account variability in human

exposure using Professqr Spears calculations could mean that excellent

inter-risk comparisons couldmiddot be made This is our present goal and we will

report on it at our earliest opportunity

11

5) Animal Care

Specific instructions regarding animal care should be provided given

the powerful effects of marginal malnourishment and infections on many

bodily functions (14)

6) Pathological Interpretation

I am very concerned at the use of rodent hepatic cell tumors for the

identification of active carcinogens which supposedly work at very very

low levels Major support for the carcinogenicity of PBBs chlorinated

hydrocarbons azo dyes and other chemicals arises from the identification of

mouse hepatomas which show a very high incidence in controls It is now

not deemed sufficient to call something a carcinogen if it induces lung adeshy

nomas or adenocarcinomas in Strain A mice because of the high incidence of

these tumors in controls It therefore also seems unjustified to use the

mouse hepatoma to claim something is carcinogenic when the mice are inishy

tiated promoted or transformed without application of the chemical Nonshy

toxicologists and toxicologists alike must be aware of exactly what tumors

are formed in the experimental animals and their relevance to both

background levels and man The simple plugging in of numbers of tumors

(unspecified) into risk assessment calculations which I have seen so often

in the past is totally unjustified

(b) Difficulties in Extrapolation

As briefly stated in the Introduction several mathematical models are

available for making extrapolations from high doses to low doses It is my

8

opinion however that none of the models presently available have any

scientific validity I base this opinion on the fact that many of the

assumptions made in their development are highly questionable For example

in the determination of the equation for the multistage model (15) it has

been assumed that the relationship between the dose and the number of hits

is always first order regardless of dose and that there is no individual

differences among animals in a given experimental group It is of course

laughable to compare this to humans who are remarkably heterogeneous in

their response to foreign compounds Some models for example the one hit

model also have no bearing on the reality shown by experimental data

Results of the ED01 experiment of the National Center for Toxicological

Research (16) showed a highly non-linear dose response incidence of bladder

tumors in mice sacrificed at 24 months and an analysis of that data (17)

showed a lack of fit p value of less than 0001 for the one-hit model

could go on and on criticizing the models used and modelling in general but

there is no point We simply have to model there is not other way How

then can we do it more reasonably

A better way may be to use several models and compare their goodness of

fit This approach has been advocated by the Food Safety Council (18) It

involves fitting four models the one-hit model Armitage-Doll model

Weibull model and the gamma multihit model to the dose-response data and

determining goodness of fit Each model can then be used to determine virshy

tual safe doses at the risk level of interest The regulator could then

determine from the goodness of fit values which model gives the most

reasonable estimate of risk and determine a safe level also taking into

9

I

account societal usefulness of the chemical We may therefore be willing to

tolerate relatively high levels of a weak carcinogen which appears to have

curvature in its dose-response curve but may wish to completely remove a

strong carcinogen having a more l inear dose-response For further details

see (18)

We are still of course faced with the problem of extrapolation from

animals to man My only comment here is that we must in some way take into

account metabolic and pharmacokinetic data as well as extrapolating on the

basis of body weight rather than surface area or some other parameter The

justifications for this are amply described in (18)

Future Directions - Possible Usefulness of Inter-Risk Comparison

In order to put health risks from chemicals into perspective Crouch and

Wilson (19) have taken a new approach by comparing some measure of risk for

various aspects of everyday life with similar measures of risk from chemishy

cals There are obviously still uncertainties in this inter-risk comparison

procedure but it does provide a badly needed focus as to significant and

insignificant health risks associated with chemicals The procedure

outlined in (19) is still very crude a fact which the authors are only too

aware of and could be much improved Briefly Crouch and Wilson apply one

model (a linear model for relating dose and effect) to a variety of chemishy

cals using the formula

R = 8d

where R = lifetime risk

d average dose rate and

S = carcinogenic potency of the chemical

10

The carcinogenic po t ency (8) is derived from the animal dose-response

data

They then introduce two other factors an interspecies conversion factor

K and a dispersion factor I which is the proportion of total production of

a chemical which is finally absorbed into humans The expected annual

cancer death rate (n) due to an annual production P of a chemical is

therefore

n = $KIP x 0 254

(0254 is the factor used to convert from lifetime risk to annual risk)

This equation does not of course tell us the actual number of likely

cancer deaths but does provide what Crouch and Wilson call a Hazard Index

For piperonyl butuxide for example this Hazard Index is 40 whereas for

endosulfan it is 50000 despite their annual production levels being

approximately the same Differences in carcinogenic potency certainly make

a difference here and would help regulatory agencies focus on potentially

significant hazards to human health from pesticides

The incorporation of much more toxicological data eg mutagenic

potency pharmacokinetics etc into Crouch and Wilsons equations along with

the fact that we are now able to take into account variability in human

exposure using Professqr Spears calculations could mean that excellent

inter-risk comparisons couldmiddot be made This is our present goal and we will

report on it at our earliest opportunity

11

opinion however that none of the models presently available have any

scientific validity I base this opinion on the fact that many of the

assumptions made in their development are highly questionable For example

in the determination of the equation for the multistage model (15) it has

been assumed that the relationship between the dose and the number of hits

is always first order regardless of dose and that there is no individual

differences among animals in a given experimental group It is of course

laughable to compare this to humans who are remarkably heterogeneous in

their response to foreign compounds Some models for example the one hit

model also have no bearing on the reality shown by experimental data

Results of the ED01 experiment of the National Center for Toxicological

Research (16) showed a highly non-linear dose response incidence of bladder

tumors in mice sacrificed at 24 months and an analysis of that data (17)

showed a lack of fit p value of less than 0001 for the one-hit model

could go on and on criticizing the models used and modelling in general but

there is no point We simply have to model there is not other way How

then can we do it more reasonably

A better way may be to use several models and compare their goodness of

fit This approach has been advocated by the Food Safety Council (18) It

involves fitting four models the one-hit model Armitage-Doll model

Weibull model and the gamma multihit model to the dose-response data and

determining goodness of fit Each model can then be used to determine virshy

tual safe doses at the risk level of interest The regulator could then

determine from the goodness of fit values which model gives the most

reasonable estimate of risk and determine a safe level also taking into

9

I

account societal usefulness of the chemical We may therefore be willing to

tolerate relatively high levels of a weak carcinogen which appears to have

curvature in its dose-response curve but may wish to completely remove a

strong carcinogen having a more l inear dose-response For further details

see (18)

We are still of course faced with the problem of extrapolation from

animals to man My only comment here is that we must in some way take into

account metabolic and pharmacokinetic data as well as extrapolating on the

basis of body weight rather than surface area or some other parameter The

justifications for this are amply described in (18)

Future Directions - Possible Usefulness of Inter-Risk Comparison

In order to put health risks from chemicals into perspective Crouch and

Wilson (19) have taken a new approach by comparing some measure of risk for

various aspects of everyday life with similar measures of risk from chemishy

cals There are obviously still uncertainties in this inter-risk comparison

procedure but it does provide a badly needed focus as to significant and

insignificant health risks associated with chemicals The procedure

outlined in (19) is still very crude a fact which the authors are only too

aware of and could be much improved Briefly Crouch and Wilson apply one

model (a linear model for relating dose and effect) to a variety of chemishy

cals using the formula

R = 8d

where R = lifetime risk

d average dose rate and

S = carcinogenic potency of the chemical

10

The carcinogenic po t ency (8) is derived from the animal dose-response

data

They then introduce two other factors an interspecies conversion factor

K and a dispersion factor I which is the proportion of total production of

a chemical which is finally absorbed into humans The expected annual

cancer death rate (n) due to an annual production P of a chemical is

therefore

n = $KIP x 0 254

(0254 is the factor used to convert from lifetime risk to annual risk)

This equation does not of course tell us the actual number of likely

cancer deaths but does provide what Crouch and Wilson call a Hazard Index

For piperonyl butuxide for example this Hazard Index is 40 whereas for

endosulfan it is 50000 despite their annual production levels being

approximately the same Differences in carcinogenic potency certainly make

a difference here and would help regulatory agencies focus on potentially

significant hazards to human health from pesticides

The incorporation of much more toxicological data eg mutagenic

potency pharmacokinetics etc into Crouch and Wilsons equations along with

the fact that we are now able to take into account variability in human

exposure using Professqr Spears calculations could mean that excellent

inter-risk comparisons couldmiddot be made This is our present goal and we will

report on it at our earliest opportunity

11

account societal usefulness of the chemical We may therefore be willing to

tolerate relatively high levels of a weak carcinogen which appears to have

curvature in its dose-response curve but may wish to completely remove a

strong carcinogen having a more l inear dose-response For further details

see (18)

We are still of course faced with the problem of extrapolation from

animals to man My only comment here is that we must in some way take into

account metabolic and pharmacokinetic data as well as extrapolating on the

basis of body weight rather than surface area or some other parameter The

justifications for this are amply described in (18)

Future Directions - Possible Usefulness of Inter-Risk Comparison

In order to put health risks from chemicals into perspective Crouch and

Wilson (19) have taken a new approach by comparing some measure of risk for

various aspects of everyday life with similar measures of risk from chemishy

cals There are obviously still uncertainties in this inter-risk comparison

procedure but it does provide a badly needed focus as to significant and

insignificant health risks associated with chemicals The procedure

outlined in (19) is still very crude a fact which the authors are only too

aware of and could be much improved Briefly Crouch and Wilson apply one

model (a linear model for relating dose and effect) to a variety of chemishy

cals using the formula

R = 8d

where R = lifetime risk

d average dose rate and

S = carcinogenic potency of the chemical

10

The carcinogenic po t ency (8) is derived from the animal dose-response

data

They then introduce two other factors an interspecies conversion factor

K and a dispersion factor I which is the proportion of total production of

a chemical which is finally absorbed into humans The expected annual

cancer death rate (n) due to an annual production P of a chemical is

therefore

n = $KIP x 0 254

(0254 is the factor used to convert from lifetime risk to annual risk)

This equation does not of course tell us the actual number of likely

cancer deaths but does provide what Crouch and Wilson call a Hazard Index

For piperonyl butuxide for example this Hazard Index is 40 whereas for

endosulfan it is 50000 despite their annual production levels being

approximately the same Differences in carcinogenic potency certainly make

a difference here and would help regulatory agencies focus on potentially

significant hazards to human health from pesticides

The incorporation of much more toxicological data eg mutagenic

potency pharmacokinetics etc into Crouch and Wilsons equations along with

the fact that we are now able to take into account variability in human

exposure using Professqr Spears calculations could mean that excellent

inter-risk comparisons couldmiddot be made This is our present goal and we will

report on it at our earliest opportunity

11

The carcinogenic po t ency (8) is derived from the animal dose-response

data

They then introduce two other factors an interspecies conversion factor

K and a dispersion factor I which is the proportion of total production of

a chemical which is finally absorbed into humans The expected annual

cancer death rate (n) due to an annual production P of a chemical is

therefore

n = $KIP x 0 254

(0254 is the factor used to convert from lifetime risk to annual risk)

This equation does not of course tell us the actual number of likely

cancer deaths but does provide what Crouch and Wilson call a Hazard Index

For piperonyl butuxide for example this Hazard Index is 40 whereas for

endosulfan it is 50000 despite their annual production levels being

approximately the same Differences in carcinogenic potency certainly make

a difference here and would help regulatory agencies focus on potentially

significant hazards to human health from pesticides

The incorporation of much more toxicological data eg mutagenic

potency pharmacokinetics etc into Crouch and Wilsons equations along with

the fact that we are now able to take into account variability in human

exposure using Professqr Spears calculations could mean that excellent

inter-risk comparisons couldmiddot be made This is our present goal and we will

report on it at our earliest opportunity

11

References

1 Task Force of Past Presidents of the Socie ty of Toxicology (1982) Animal data in hazard evaluation Paths and Pitfall s Fund Appl Toxicol 2 101

2 Gil lette J R ( 1984) Solvable and unsolvable problems in extrapolating toxicolgoical data between animal species and strains In Drug Metabolism and Drug Toxicity (JR Mitchell and M G Horning eds ) Raven Press N Y

3 Clayson D B (1977) Relationships between l aboratory and human studies J Environ Pathol Tox 1 31

4 Cornfield J (1977) Carcinogenic Risk Assessment Science 198 693

5 Norpoth K H and Garner R C eds (1980) Short- Term Test Systems for Detecting Carcinogens Springer-Verlag Berlin

6 Zbinden G and Flury-Rovers M ( 1981) Significance of the LD50- Test for the Toxicological Evaluation of Chemical Substances Arch Toxicol 47 77

7 Weil cs Woodside MD Bernard J R and Carpenter CP (1969) Relationship between single- peroral one- week and ninety day rat feeding studies Toxicol Appl Pharmacol 14 426

8 Gilman M R (1982) Skin and Eye Testing in Animals In Principles and Method s of Toxicology (A W Hayes ed ) Raven Press N Y

9 Kao J Hall J and Holland J M ( 1983) Quantitation of cutaneous toxicity An in Vitro approach using skin organ culture

10 NCI (1978) Bioassay of 12-dibromoethane for possible carcinogenicity NCI Carcinogenesis Technical Report Series 86

11 Roe D A ed (1983) Diet Nutrition and Cancer From Basic Research to Policy Implications Alan R Liss Inc NY

12 Berg B N and Simms HS (1960) Nutrition and longevity in the rat II Longev ity and onset of disease with different levels of food intake J Nutr 71 255

13 Schach von Wittena~ M and Estes PC (1983) The r edundancy of mouse carcinogenicity bi oassays Fund Appl Toxicol 3 631

14 Por t er WP et al (1984) Toxicant- Disease-Environment Interactions associated with Suppression of Immune System Growth and Reproduction Science 224 1014

15 Whittemore A S (1978) Quantitative theories of oncogenesis Adv Cancer Res 27 55

12

16 Littlefield NA Farmer JH Gaylor DW and Sheldon WG (1979) Effects of dose and time in long-term low dose carcinogenic study In Innovations in Cancer Risk Assessments (ED01) Study) PPbull 17-34 Pathotox Press - shy

17 Carlborg FW (1981) Dose-response functions in carcinogenesis and the Weibull model Food Cosm Toxicol 19 261

18 Wodicka vo (1984) Use of risk assessment and safety evaluation In Assessment and Management of Chemical Risks (JV Rodricks and RG Tardiff eds) p 131 ACS Symposium Series 239

19 Crouch EAC and Wilson R (1984) Inter-Risk Comparisons ibid p 97

13

16 Littlefield NA Farmer JH Gaylor DW and Sheldon WG (1979) Effects of dose and time in long-term low dose carcinogenic study In Innovations in Cancer Risk Assessments (ED01) Study) PPbull 17-34 Pathotox Press - shy

17 Carlborg FW (1981) Dose-response functions in carcinogenesis and the Weibull model Food Cosm Toxicol 19 261

18 Wodicka vo (1984) Use of risk assessment and safety evaluation In Assessment and Management of Chemical Risks (JV Rodricks and RG Tardiff eds) p 131 ACS Symposium Series 239

19 Crouch EAC and Wilson R (1984) Inter-Risk Comparisons ibid p 97

13