NCRP report no 150 extrapolation of radiation induced cancer risks from nonhuman experimental systems to humans

This Report concentrates on life shortening and cancer, withsome comments on the use of data from mice, augmented by datafrom humans, in the estimation of the risk of radiation-inducedhe

National Council on Radiation Protection and Measurements

N C R P

Extrapolation of Induced Cancer Risks from Nonhuman Experimental Systems to Humans

Radiation-Recommendations of the

NATIONAL COUNCIL ON RADIATION

PROTECTION AND MEASUREMENTS

Issued November 18, 2005

National Council on Radiation Protection and Measurements

7910 Woodmont Avenue, Suite 400/Bethesda, MD 20814-3095

Measurements (NCRP) The Council strives to provide accurate, complete and ful information in its documents However, neither NCRP, the members of NCRP, other persons contributing to or assisting in the preparation of this Report, nor any person acting on the behalf of any of these parties: (a) makes any warranty or rep- resentation, express or implied, with respect to the accuracy, completeness or use- fulness of the information contained in this Report, or that the use of any information, method or process disclosed in this Report may not infringe on pri- vately owned rights; or (b) assumes any liability with respect to the use of, or for damages resulting from the use of any information, method or process disclosed in

use-this Report, under the Civil Rights Act of 1964, Section 701 et seq as amended 42

U.S.C Section 2000e et seq (Title VII) or any other statutory or common law theory governing liability.

Disclaimer

Any mention of commercial products within NCRP publications is for tion only; it does not imply recommendation or endorsement by NCRP.

informa-Library of Congress Cataloging-in-Publication Data

Extrapolation of radiation-induced cancer risks from nonhuman experiment systems to humans.

616.99’4071—dc22

2005031014

Copyright © National Council on Radiation Protection and Measurements 2005 All rights reserved This publication is protected by copyright No part of this publica- tion may be reproduced in any form or by any means, including photocopying, or utilized by any information storage and retrieval system without written permission from the copyright owner, except for brief quotation in critical articles or reviews.

[For detailed information on the availability of NCRP publications see page 251.]

This Report reviews the scientific issues associated with theextrapolation of radiation-induced cancer risks from nonhumanexperimental systems to humans The basic principles of radiationeffects at the molecular and cellular level are examined withemphasis on comparisons among various species includinghumans These comparisons among species are then continued forcancers of similar cell types in the same organ system Risk esti-mates are made from an observed level of effect as a function oforgan dose The major organ systems are individually considered.Extrapolation models are reviewed and include external and inter-nal radiation exposures

At the beginning of the nuclear age there was no idea what risksworkers would face in the handling of such substances as pluto-nium The only option was to rely on experimental animal data andextrapolation This effort, together with good health physics prac-tices and medical surveillance resulted in, by and large, a well-protected workforce

Many experimental animal studies were undertaken shortlyafter World War II One aim was to understand the biologicaleffects of radiation, and, hopefully, the mechanisms of the effects.Another aim was to determine the influence of such factors as doserate, radiation quality, gender, and age at exposure

Much has been learned and the general information has beenincorporated into efforts associated with risk estimations Morespecifically, the experimental data have been the basis for selectingdose and dose-rate effectiveness factors (DDREF) and radiationquality factors These factors are used to moderate risk estimateseither for patterns of irradiation or types of radiation for whichthere are inadequate data

The risk estimates used in radiation protection throughout theworld come almost entirely from the atomic-bomb survivors, whowere acutely exposed to high-dose-rate gamma rays, not at all like

the industrial (e.g., uranium miners’) exposures Hence, the

con-cern for the influence of dose rate and fractionation An additionalconcern is the lack of appropriate data for estimating risk to peopleexposed in space missions Again, we must rely on experimentalstudies

Much remains to be done It is believed that if more relevantdata could be obtained to develop acceptable methods of extrapola-tion across species, risk estimates could be improved With anunderstanding of what information is necessary to undertakeextrapolation, it would be possible to make better use of the consid-erable body of data on cancer induction by radiation

This Report includes a discussion of nontargeted radiationeffects that potentially influence dose-response characteristics ofcells and tissues at low absorbed doses These nontargeted effectsinclude bystander effects, genomic instability, and adaptive radia-tion responses, all of which are subjects of presently activeresearch It is anticipated that future NCRP reports will analyzethe influence of these factors on radiation dose-response character-istics and variations in radiation response(s) among species.This Report gives an account of the steps by which a group ofresearchers has advanced the pragmatic and theoreticalapproaches to extrapolation of estimates of risk from radionuclidesand external radiation The Report identifies the problems inextrapolating the current data from, for example, mice to humans

It provides examples of using Bayesian statistics to successfullyestimate DDREF values for humans from data for mice, and alsoprovides a measure of uncertainty for this estimate The Reportalso shows how a defensible quantitative estimate of radiationinjury can be determined for humans even when the exposure data

of interest are either lacking or of poor quality, and how mortalitydata for laboratory animals can be used to predict age-specific radi-ation-induced risks for humans for general endpoints like lifeshortening, all cancers, and selected subsets of cancers involvinghomologous tissues

Every example of an interspecies prediction of induced mortality contained within this Report was made in theabsence of an extensive understanding of cellular and moleculargenetic effects of radiation These latter types of effects are impor-tant for providing a degree of confidence to use of the so-called bio-logical models, but they were not critical, as stated above, for theempirical models, which take into account a great deal of what ispresently known

radiation-Perhaps the most encouraging aspect of the Report is theaccount of how the effect of life span can be extrapolated across spe-cies, and examples of doing so from mice to dogs to humans aregiven The differences of life span among species has long beenstudied, and the possibility of using life shortening not only as anintegrated index of radiation effects, but also for deriving a singlevalue of relative biological effectiveness and DDREF for radiation

protection purposes seemed worth examining Lastly, the Reportrecommends research that is required to advance this importantfield.

This Report was prepared by Scientific Committee 1-4 on theExtrapolation of Risks from Nonhuman Experimental Systems toMan Serving on Scientific Committee 1-4 were:

NCRP Secretariat

Morton W Miller, Consultant (2004–2005)

Bruce B Boecker, Consultant (2004–2005)

William M Beckner, Senior Staff Scientist (1992–1997, 2000–2004) Thomas M Koval, Senior Staff Scientist (1997–2000)

Cindy L O’Brien, Managing Editor David A Schauer, Executive Director

David G Hoel, Chairman

Medical University of South CarolinaCharleston, South Carolina

The Council wishes to express its appreciation to the Committeemembers for the time and effort devoted to the preparation of thisReport NCRP gratefully acknowledges the financial support pro-vided by the U.S Department of Energy, Office of Biological andEnvironmental Research.

Thomas S Tenforde

President

Preface iii

1 Executive Summary and Recommendations 1

1.1 Why Is Extrapolation Still Required? 2

1.2 Summary of Findings 3

1.2.1 Historical Aspects 3

1.2.2 Neoplastic Disease 4

1.2.2.1 Hematopoietic System 5

1.2.2.2 Lung 5

1.2.2.3 Breast 6

1.2.2.4 Thyroid 6

1.2.2.5 Skin 6

1.2.2.6 Gastrointestinal Track 6

1.2.2.7 Bone 7

1.2.3 Somatic Genetic Damage at Molecular and Cellular Levels 7

1.2.4 Extrapolation Models and Methods 8

1.2.4.1 Toxicity of Chemotherapeutic Drugs 8

1.2.4.2 Life Shortening 8

1.2.4.3 Interspecies Prediction of Life Shortening and Cancer from External Irradiation 9

1.2.4.4 Extrapolation of Dose-Rate Effectiveness Factors 10

1.2.4.5 Interspecies Prediction of Injury from Internally-Deposited Radionuclides 10

1.3 Conclusions 11

1.4 Recommendations 13

2 Introduction 15

3 History of Extrapolation: Nonhuman Experimental Systems to Humans 17

3.1 Introduction 17

3.2 Lessons Learned from Genetic Risks 21

3.2.1 Methods of Estimation 22

3.2.1.1 Doubling-Dose Method 22

3.2.1.2 Direct Method 23

3.2.1.3 Gene-Number Method 23

3.2.2 Discussion of Methods of Estimating Genetic Risk 23

3.2.3 Role of Genetics in the Estimation of Somatic Risks 25

3.3 Somatic Risks 26

4 Tissue and Organ Differences Among Species with Emphasis on the Cells of Origin of Cancers 38

4.1 Introduction 38

4.2 Hematopoietic System 40

4.2.1 Introduction: Leukemias and Lymphomas 40

4.2.2 Comparison of Radiation-Induced Leukemias Among Species 42

4.2.3 Pathology and Dose-Response Relationships 43

4.2.4 Comparison of Hematopoietic Systems 44

4.2.5 Target Cells 45

4.2.6 Comparison of Cytogenetic Processes: Common or Species-Specific Patterns 47

4.2.7 Leukemogenesis Resulting from Gene Rearrangements 49

4.2.8 Secondary Cytogenetic Lesions Associated with Leukemia Promotion and Progression 50

4.2.9 Hematopoietic Cell Origins of the Putative “Critical” Genic Lesions and the Nature of Induced Genic Dysfunctions 51

4.2.10 Cooperating Oncogenes in Lymphoid Neoplasias 52

4.2.11 Cooperating Oncogenes in Myeloid Neoplasias 53

4.2.12 Hematopoeitic Microenvironment 54

4.2.13 Summary 55

4.3 Lung 56

4.3.1 Introduction 56

4.3.2 Adenocarcinoma 57

4.3.3 Squamous-Cell Carcinoma 58

4.3.4 Small-Cell Lung Carcinoma 59

4.3.5 Large-Cell Carcinoma 62

4.3.6 Summary 62

4.4 Breast 63

4.4.1 Histogenesis of Mammary Glands and Mammary Cancer 63 4.4.2 Hormones and Mammary Carcinogenesis 65

4.4.3 Cellular Origins of Mammary Cancer 66

4.4.4 Summary 68

4.5 Thyroid 68

4.5.1 General Background 68

4.5.2 Histogenesis of the Thyroid Gland and Thyroid Cancer 69

4.5.3 Thyroid Function and its Control 70

4.5.4 Cellular Economy of the Thyroid Gland and the Origin of Cancer 71

4.5.5 Summary 72

4.6 Skin 72

4.6.1 Introduction 72

4.6.2 Epidermal Cancers 73

4.6.3 Melanoma 74

4.6.4 Tumors of the Dermis 74

4.6.5 Mechanisms of Epidermal Carcinogenesis 74

4.6.6 Importance of Interactions 75

4.6.7 Summary 75

4.7 Gastrointestinal Tract 75

4.7.1 Introduction 75

4.7.2 Stomach 75

4.7.3 Small Intestine 76

4.7.4 Colorectal Tumors 76

4.7.5 Summary 77

4.8 Bone 77

4.8.1 Humans 77

4.8.2 Mice 78

4.8.3 Rats 79

4.8.4 Dogs 80

4.8.5 Summary 80

5 Radiation Effects at the Molecular and Cellular Levels 84

5.1 Introduction 84

5.2 Effects of Ionizing Radiations at the Molecular Level 85

5.2.1 DNA Damage 85

5.2.2 Repair of DNA Damage 86

5.2.2.1 Single-Strand Breaks 87

5.2.2.2 Double-Strand Breaks 87

5.2.2.2.1 Nonhomologous End-Joining 88

5.2.2.2.2 Recombination Repair 90

5.2.2.3 Base Damage Repair 91

5.2.3 Characterization of Genes (Enzymes) Involved in DNA Repair 93

5.2.4 DNA Repair and Cell-Cycle Progression 95

5.2.5 Genetic Susceptibility to Ionizing Radiations 98

5.2.6 Conclusions 100

5.3 Effects of Ionizing Radiations at the Cellular Level 101

5.3.1 Point (or Gene) Mutations 101

5.3.2 Chromosome Aberrations and Deletion Mutations 102

5.3.3 Use of Mechanistic Data on Mutation and Chromosome Aberration Induction 107

5.3.4 Cell Killing 108

5.3.5 Potential Confounders of Dose-Response Curves 109

5.3.5.1 Bystander Effects 109

5.3.5.2 Genomic Instability 110

5.3.5.3 Adaptive Responses 110

5.3.6 Genetic Alterations in Tumors in Humans and Rodents 111

5.3.6.1 Oncogene Activation 111

5.3.6.2 Tumor-Suppressor Genes 117

5.3.7 Conclusions 120

6 Extrapolation Models 122

6.1 Interspecies Correlations of Chemical Toxicities 122

6.1.1 Introduction 122

6.1.2 Acute Toxicity 122

6.1.3 Chronic Toxicity 125

6.2 Interspecies Prediction of Summary Measures of Mortality: Relative Risk Models 127

6.3 Interspecies Correlations of Radiation Effects 130

6.3.1 Introduction 130

6.3.2 Predictions of Radiation-Induced Mortality 130

6.3.3 Example of Interspecies Prediction for Single

Exposure 135

6.3.4 Conclusion 138

6.4 Interspecies Prediction of Age-Specific Mortality 138

6.4.1 Introduction 138

6.4.2 Background and Justification for Interspecies Predictions 139

6.4.3 Continuous Exposure: Mice to Dogs 140

6.4.4 Single Exposure: Mice to Dogs and Humans 143

6.4.5 Conclusion 149

6.5 Extrapolation of Dose-Rate Effectiveness Factors 149

6.5.1 Requirements and Limitations 149

6.5.2 Conclusion 156

6.6 Extrapolation of Results for Internally-Deposited Radionuclides from Laboratory Animals to Humans 156 6.6.1 Temporal Pattern of Delivery of Radiation Dose 157

6.6.2 Spatial Pattern of Delivery of Dose 158

6.6.3 Linear-Energy Transfer (Radiation Quality) 159

6.6.4 Internally-Deposited Radionuclides for Which Human and Laboratory Animal Data are Available 160

6.6.4.1 Radium-226, 228 160

6.6.4.2 Radium-224 161

6.6.4.3 Thorotrast® (232Th) 163

6.6.4.4 Radon and Radon Progeny 164

6.6.5 Examples of Internally-Deposited Radionuclides for Which Laboratory Animal Data are Available and for Which Links Could be Made to Human Data 166

6.6.5.1 Bone Cancer 166

6.6.5.2 Liver Cancer 169

6.6.5.3 Lung Cancer 171

6.6.6 Examples of Linking Risks from Laboratory Animals to Human Data 175

6.6.6.1 Bone Cancer 175

6.6.6.2 Lung Cancer 181

7 Summary 183

7.1 Introduction 183

7.2 Summary 184

7.2.1 History of Extrapolation from Nonhuman

Experimental Systems to Humans 184

7.2.2 Cells of Origin of Cancer in Different Animal Species 185

7.2.3 Radiation Effects at the Molecular and Cellular Levels 185

7.2.4 Extrapolation Models 186

Glossary 188

Symbols and Acronyms 192

References 194

The NCRP 242

NCRP Publications 251

Index 262

Recommendations

The process of extrapolation, which involves projection from theknown to the unknown, can be a daunting task A quote from aleading biometrician defines what the task needs: “To be useful,extrapolation requires extensive knowledge and keen thinking”(Snedecor, 1946) In preparing this Report, the National Council onRadiation Protection and Measurements (NCRP) faced the task ofevaluating extrapolation of the risks of radiation-induced cancer tohumans from experimental data There is extensive informationfrom experimental studies at the animal, cellular, chromosomaland molecular levels that might be used in deriving approaches tothe problem of extrapolating risk estimates across species.But there are also gaps in the database Therefore, uneasiness per-sists in the acceptability of extrapolations from nonhuman data,with the exception of the use of data from mice in the derivation ofdose and dose-rate effectiveness factors (DDREF) Furthermore,although considered a choice of necessity rather than an ideal solu-tion, data obtained from experimental animals have been used toderive an estimate of the dose-rate effectiveness factor (DREF) andradiation weighting factors

The overall aim of this Report is to consider the possibilities, thedifficulties, and the attempts to extrapolate estimates of radiation-induced stochastic effects across species, especially laboratory ani-mals to humans This Report is neither a compendium of stochasticeffects studies, nor a detailed account of mechanisms of the induc-tion of stochastic effects in different species This Report does, how-ever, discuss some of the similarities and differences of responses

to radiation at the molecular level

This Report concentrates on life shortening and cancer, withsome comments on the use of data from mice, augmented by datafrom humans, in the estimation of the risk of radiation-inducedheritable diseases Unless it can be shown that the aspects of themechanisms of importance in extrapolation are similar in cancersthat arise from cells that differ in type, there must be concern aboutpooling the data for different types of tumors in a specific organ.The fact that extrapolations of risk estimates across strains of mice

are feasible (Goldman et al., 1973; Grahn, 1970; Norris et al., 1976; Sacher, 1966; Sacher and Grahn, 1964; Storer et al., 1988) may

have been because the cells of origin of the tumors were the same

in the same specific organs in the different strains

The goals of this Report were broad and included an evaluation

of the full range of somatic risks, the quality and quantity of thedata from which extrapolation could be projected, and existingand potential methods for the extrapolation process A number ofdifferent methods of extrapolation of risk estimates of radiation-induced cancers had previously been proposed but there has beenneither a systematic examination of the similarities and differ-ences among species at the molecular, cellular, tissue and whole-organism levels, nor of the appropriateness of the data available toattempt extrapolations

Data from experimental animals are used in the estimation ofgenetic risk and in the derivation of factors to account for the effect

of dose rate and radiation quality because of the lack of humandata However, direct estimates of risk of radiation-induced cancer

at specific sites in animals have not been used

1.1 Why Is Extrapolation Still Required?

Extrapolation is still required because the available data onirradiated human populations have several important limitations.The risk estimates for radiation protection are based largely onthe data from the atomic-bomb survivors, who were exposed to anacute, high dose of gamma rays, and radiotherapy patients whohave been treated with high-dose-rate fractionated exposures Alarge number of people have been exposed occupationally, someprotracted over long periods but in complex time patterns that havemade it difficult, if not impossible, to determine the effect of totaldoses of low-dose- rate radiation Hence, there is the need for datafrom experimental studies to determine values of DREF Most ofthe exposures of humans are to very small multiple exposures ofhigh-dose-rate radiation Diagnostic radiation is one example Inthe case of occupationally exposed individuals, much of the expo-sure occurs at ages of reduced susceptibility, and the reduction ineffectiveness is not solely due to the reduced dose rate

There are also inadequate data of the effects of high energy transfer (LET) radiations such as fission neutrons of rele-vant energies and heavy ions for the estimation of risk in humans.There are some experimental weighting factors, but more data areneeded

linear-It is the aim of this Report to examine the problems and tial of extrapolation of experimental findings in laboratory animals

poten-to risk estimates in humans It is necessary poten-to determine the ria on which the suitability of the data for extrapolation purposescan be decided For example, consideration is given to whether riskestimates of cancer induction should be based on cancers originat-ing in the same types of cells and not just the same organ

crite-1.2 Summary of Findings

This Report has five sections following a brief introduction.First, the history and existing methodologies of extrapolation arepresented Second, a discussion is presented of selected neoplasticdisease endpoints of particular importance to humans Third,radiation-induced damage at molecular and cellular levels isreviewed in detail as the underpinning of comparisons at the tissueand organ levels Fourth, extrapolation models and examplesare given and include a discussion of the complex fields of radionu-clide toxicology and chemical toxicology is included Fifth, is a sum-mary of the Report’s findings followed by a glossary of terms andreferences

1.2.1 Historical Aspects

A brief review of the extrapolation of genetic risks from animals

to humans revealed concerns about somatic effects, which are notreadily studied by traditional genetic processes There are substan-tial differences in approaches to studying genetic or somatic effects:the baselines, issues and methodologies of the two areas of investi-gation are different, with one important exception; it has becomeclear that risk analysis requires a biological commonality to linkthe different species For geneticists, the commonality is simply thedeoxyribonucleic acid (DNA) molecule and its associated metabolicmanagement For those dealing with somatic effects, there is, inaddition to DNA, the common process of “dying out” (the actuariallife table), the importance of which slowly became appreciatedbetween about 1925 and 1950 This Report recounts the attempts

to extrapolate risk estimates across species, critically examines thestrengths and weaknesses of these attempts, and supports andextends the contention that estimates of the effects of radiation onlife or neoplastic diseases can be extrapolated across species tohumans

1.2.2 Neoplastic Disease

In this Report, seven tissues or organs (hematopoietic system,lung, breast, thyroid, skin, gastrointestinal track, and bone) werestudied for the feasibility of extrapolation of cancer risks Thesewere chosen for their general importance in human cancer riskanalysis and because extensive data exist from animal studies.There are distinct differences in the problems and their solutionsthat are encountered in undertaking risk estimation and extrapo-lation of risks for external and internal radiation The studiesrelated to external and internal radiation are reported separately

In the case of solid cancers related to external radiation, a majorbarrier to success for extrapolation and the testing of methods ofextrapolation is the lack of data of cancer based on the specific type

of cancer or the type of cell or origin Most of the data that havebeen used in risk estimates, such as those from the studies of theatomic-bomb survivors used in the selection of radiation limits,are for specific organs, for example lung tumors, not squamous-cellcarcinoma (SCC) or small-cell cancer

Studies of animals, particularly dogs, have long been one of theprincipal sources of data for estimating the risk of the effects ofexposure to internally-deposited radionuclides in humans Suchstudies, for example, have been used to estimate risks of latenteffects of exposures to bone, lung, liver and bone marrow An impor-tant impetus for these studies was that there was either a completelack or a paucity of relevant data based on human experience Theneed for estimates of risk to humans resulted in the examination ofapproaches of how to extrapolate these results across species andthe adoption of the relative toxicity ratio method (reference can bemade to the glossary for a definition for this and other terms, acro-nyms and abbreviations) There are problems of dosimetry andcomplicating factors such as relocation of the radionuclide, that arespecific to internal emitters; thus, internal exposures to radionu-clides are discussed in Section 6.6

The similarities of aspects of the risk of induction of canceramong species encourage the search for methods of extrapolation.Success in this endeavor will not only improve the current values

of DREF and relative biological effectiveness (RBE), but also maymake it possible to use the considerable body of experimental data

in risk estimation Between humans and experimental animalsthere are some fundamental differences that are difficult to assess,such as lifestyle, longevity and environment Perhaps the mostimportant problem in being able to derive and test methods ofextrapolation is the fact that most epidemiological data for cancer

for humans that are used in radiation risk estimation are based oncancer site and not on cell type In the case of mice lungs, the factthat lung cancers in many strains are exclusively adenocarcino-

mas, and small-cell cancer is not found, limits the potential for

extrapolation There have been few attempts to extrapolate risks ofradiation-induced solid cancers, and in the case of hematopoieticdiseases, despite all the information for leukemias in humans, dogsand mice, there has been no success in doing so This Report alsoconsiders the problem of differences in host factors among species

In breast, there are important differences in the hormonal ence on tumors among species One of the encouraging features isthe demonstration of the similarities in the genes across the speciesthat are important in the initiation of cancers While this is obvi-ously important, is it sufficient to allow some method of extrapola-tion? Another problem is the fact that much of the suitable data forthe induction of solid cancers has been obtained after exposure atone age whereas human data, such as that from the atomic-bombsurvivors, are for exposures at all ages This, of course, is moreimportant for the tumors with a marked age dependency such asthyroid cancer Lastly, much of the data from experimental animals

influ-is restricted to a small number of strains Thinflu-is influ-is particularly true

in the case of the dog

Some of the characteristics of the organs discussed in thisReport are briefly noted in the following subsections

1.2.2.1 Hematopoietic System Leukemia has been considered to be

a major oncogenic effect in irradiated human populations Rodents,dogs and humans have nearly identical hematopoietic systems,similarities in the cell types of myelogenous and lymphocytic leu-kemias and reticular-cell sarcomas, and some common underlyinggenetic components

1.2.2.2 Lung There are significant differences between animals

and humans and their susceptibilities to lung cancer induction andthe predominant type of cancer (Section 4.3) Thus, simple extrap-olation of radiation-induced lung cancer from animals to humans isnot reasonable Selected risk analysis may be feasible for sometumors of the same cell types when there are appropriate experi-mental animal models, but this is not always the case For example,small-cell carcinomas in humans lack a counterpart in experimen-tal animal models The extrapolation of risks from radionuclideshas been reported and is discussed in Section 6.5 Further, the role

of smoking (which may interact with radiation) cannot generally beevaluated in animal models

1.2.2.3 Breast The cellular components and the major anatomic

and histologic features of mammary glands are similar amonghumans, dogs and rodents but there are important physiologic dif-ferences, for example, in the hormonal control of growth(Section 4.4) The marked strain-dependent differences for bothnaturally occurring and radiation-induced mammary cancer inthe mouse and rat makes these rodents very useful models for thestudy of molecular, cellular and tissue aspects of the mechanismsinvolved There has not been a systematic and critical study of how

to extrapolate the extensive data on risk estimates of mammarycancer in different strains of rats and mice

1.2.2.4 Thyroid The physiology, morphology and tumor cell of

ori-gin are comparable among different mammalian species lation of the estimate of risk of radiation-induced cancer appearsfeasible but no quantitative tests have been reported For the thy-roid (Section 4.5) as for the breast, the rat is considered to be therodent of choice for extrapolation studies

Extrapo-1.2.2.5 Skin Provided that the data for humans are not

con-founded by interactions, extrapolation is feasible, but care must betaken to restrict the effort to tumors of the same cell type It isimportant to appreciate that data from humans can be confounded

by interactions with other chemical and physical agents Factorssuch as ultraviolet (UV) exposures and how much of the skin isexposed are important in human skin cancer induction, and thusmay be difficult to address in animal models Furthermore, non-melanoma skin cancer such as basal-cell carcinoma (BCC) can also

be caused by skin exposures to UV light; this is a typical ing factor with the assessment of radiation risks Melanoma is con-sidered only briefly because of the lack of evidence that ionizingradiation is a major etiological factor These issues are discussed inSection 4.6

confound-1.2.2.6 Gastrointestinal Track There are many similarities in the

different gastrointestinal (GI) track tumors among mammalianspecies Therefore, these tumors provide an excellent resourcefor the study of the mechanisms of carcinogenesis However, thereare not adequate data for analysis of dose-response relationshipsfor the induction of tumors of the same type in either humans

or rodents to test the possibility of extrapolation The induction

of cancer of the GI tract requires relatively high doses in rodents(Section 4.7)

1.2.2.7 Bone Most of the data for the induction of bone tumors in

laboratory animals comes from studies of internally-deposited onuclides Permissible body burdens for a number of radionuclides

radi-in humans have been derived from studies radi-involvradi-ing dogs Whetherosteogenic tumors arise from the same cells or the same lineage indogs and humans is not clear, but the separate data for osteosarco-mas and for fibroblastic and fibrohistiocytic types of bone tumorsprovide an opportunity for testing methods of extrapolation Highdoses of external radiation are required to induce bone tumors inhumans and experimental animals, and no attempts to extrapolatethe risks have been reported (Section 4.8)

1.2.3 Somatic Genetic Damage at Molecular and Cellular Levels

The mechanism for the induction of chromosome aberrationsand mutations by ionizing radiations are currently best understoodfor human cells However, similar mechanisms of induction areknown to prevail across a range of species The processes that con-vert radiation-induced DNA damage into genetic alterations areerrors during DNA repair or replication; some damage is irrepara-ble The errors, as judged by radiation-induced mutation rates, arebroadly similar within a factor of two across mammalian species,with much of the differences in mutation rates accounted for by dif-ferences in DNA content Thus, on the assumption that sensitivity

to mutation induction is directly reflective of sensitivity to tumorinduction, an extrapolation for radiation-induced tumors thatallows for this factor of two is defensible

Multiple steps seem to lead to spontaneous and induced tumors in rodents and humans, mutations and chromo-some alterations (structural or numerical) being involved at eachstep But particular gene alterations involved for a specific tumortype tend to be different across species Whether this difference issignificant in terms of extrapolation is not clear The data on radi-ation-induced tumors are, however, limited Certainly a similaritywould strengthen the confidence in the extrapolation In addition,there are species-specific host factors that can alter the probabili-ties of tumor development from initiated cells These factors need

radiation-to be investigated further radiation-to establish how they influence tion-induced tumor dose-response models and extrapolation acrossspecies Additional data on the mechanism of tumor formation willimprove the level of confidence in extrapolating from data onrodent tumors to human tumors

radia-1.2.4 Extrapolation Models and Methods

This Report considers separately the species extrapolationmethods that have been used for external and internal exposures.For comparison purposes, a short review is first given of speciesextrapolation issues in chemical carcinogenesis For external expo-sures both life-shortening and cancer risk estimations are reviewedincluding Bayesian methods for DREF estimation

1.2.4.1 Toxicity of Chemotherapeutic Drugs The correlation

among mammalian species of the toxicity of chemotherapeuticsprovides encouragement to the radiation toxicologist for extrapola-tion methods A judicious combination of pragmatism and pharma-cologic chemistry has created feasible and practical approaches tothe preclinical and clinical trials of anticancer drugs The lessonhere is to take advantage of the animal data collected for this pur-pose and look for the underlying physiological commonalities

1.2.4.2 Life Shortening Two actuarial methods of extrapolation

were examined for the life-shortening endpoint The first method(specially designed for single exposure to low-LET radiation) relies

on two findings described in the Report First, Gompertz models

(i.e., linear equations on a semi-logarithmic scale) used to describe

age-specific death rates exhibit parallel displacements from the

control that are proportional to dose (i.e., can be described as a

function of dose) Second, at least for the species compared in thisReport (B6CF1 mouse, beagle dog, and humans as represented byatomic-bomb survivors), the dose-dependent displacements of theage-specific death rates can be described by the same equation.This finding produces the desirable effect that species with gooddose-response data can be used to predict dose-dependent lifeshortening in species for which information on radiation exposure

is either poor or lacking The second example presented in theReport describes a method that uses proportional hazard models(PHMs) to perform interspecies predictions of radiation-inducedmortality In this case, the endpoint examined is “intrinsic” mortal-ity, which refers to causes of death that arise from within the indi-vidual As such, intrinsic mortality and life shortening (anintegrated measure of damage) are closely related Simple PHMswere used to describe the dose response in a species chosen to bethe “predictor” species The resulting model was then used to pre-

dict cumulative survivorship [S(t)] curves at levels of dose observed

in a “target” species, in which S(t) is the value of the cumulative survivorship function at time “t,” where t is the age when the

deaths from whatever cause are being examined The ages

associ-ated with the predicted S(t) values were scaled by a constant

(the ratio of the predictor species’/target species’ median ages ofintrinsic death) When confidence intervals were calculated for the

empirically derived S(t) curves for each dose group observed in

the target species, the scaled predictions from the PHM fell withinthese confidence intervals In combination, these two examplesdemonstrate that not only can summary measures of life shorten-

ing (e.g., days lost per centigray) be predicted from one species to

another, but the entire schedule of age-specific death rates for lifeshortening can also be successfully predicted

1.2.4.3 Interspecies Prediction of Life Shortening and Cancer from

External Irradiation The methodology for interspecies prediction

of cancer from external irradiation relies upon the fact that the lifespan of all mammalian species can be described by the same math-

ematical formula that describes the species life table (i.e., the

exponential process of dying out) Intercepts and slopes are specific, but the equation is the same, and all species can be “cre-ated equal” by appropriate codification of the parameters

species-Two examples are presented In one test case, data from miceexposed to protracted daily gamma irradiation are used to predictthe survival of beagles subjected to comparable exposure In thesecond case, data from mice exposed to single doses of gamma radi-ation are used to predict observed survival of the atomic-bombexposed individuals in Hiroshima and Nagasaki Though the lifetables for the latter are not yet complete, a remarkably consistentrelation is apparent for the mouse to human extrapolation Thesurvival of unirradiated populations of mice, dogs and humans canall be described by a single cumulative function of median age atdeath

Extrapolation models involve comparisons among different cies or different populations within a species In either case, differ-ences among populations in mortality that are not related to the

spe-cause of interest (e.g., accidents, infectious disease, environmental

trauma) can conceal or distort a shared species response to tion, especially for an endpoint like life shortening, which is based

radia-on all deaths This mortality cradia-ontaminatiradia-on problem was solved inthe above extrapolations by coupling a biologically relevant parti-tioning of mortality, “intrinsic” mortality in the mouse to dogextrapolation, “solid-tissue tumors” in the mouse to human extrap-olation, with widely available models for survival analysis thatincorporate censoring

Host factor differences also limit extrapolation Although mice,dogs and humans often die from identical or nearly identicalcauses, these deaths need to occur at identical time points withinthe relative life span of the species to have utility in extrapolationprocesses Since background mortality risks vary by age, thesespecies-specific (host-factor) time shifts would probably cause anextrapolation based on a relative risk model to fail

Finally, it is necessary to account for the fact that humans areexposed to radiations at a wide range of ages, while the animal dataconsist overwhelmingly of exposures of young adults Radiation-induced mortality risks are known to vary by age at exposure Thisappears to raise a barrier to testing methods of extrapolationbecause Radiation Effects Research Foundation data on atomic-bomb survivors are largely determined by effects from dosesaround 1 Sv at high-dose rates When interspecies data involvelarge single doses or high-dose rates, the emergence of age-relatedeffects and species dependent pathology syndromes cause theextrapolations to fail In summary, interspecies extrapolations

of radiation-induced risk for external exposure to radiation arefeasible, reasonable and reliable when performed within fairlybroad levels of pathology and level of exposure It remains to beseen if extrapolations within defined levels of pathology detail arereasonable

1.2.4.4 Extrapolation of Dose-Rate Effectiveness Factors Since

human data for prolonged exposures to almost all types of ionizingradiation are insufficient for direct estimation of risks, extrapola-tion of dose-rate effectiveness factors (DREF) using animal datahas to be considered This Report presents one example for extrap-olation with data on female BALB/c mice exposed to 137Cs externalgamma radiation at low- and high-dose rates A probability densityfunction for a DREF in mice is estimated for mammary tumorswith Bayesian methods and combined with a probability densityfor the breast cancer risk coefficient in female atomic-bomb survi-vors This results in an estimated probability density function and

a risk coefficient for breast cancer in humans after prolonged sure This probability density also describes the remaining uncer-tainty about the breast cancer risk in humans after prolongedexposure

expo-1.2.4.5 Interspecies Prediction of Injury from Internally-Deposited

Radionuclides Interspecies extrapolation for radionuclide toxicity

has been conducted for many years, in particular for bone cancer

The metabolic behavior, internal distribution, and deposition of thetransuranics, radium and strontium, are predictable for mice,rats, dogs and humans, and the target organ, the skeleton, alsoresponds predictably This has permitted the use of available datafrom human exposures to radium, for example, to make predictionsfor dogs The approach can be extended to other nuclides for whichhuman data do not exist but animal data do exist and forwhich exposure or uptake patterns are different among species.The toxicity ratio, which is based upon the ratio of dose-response slopes for the response to two different radionuclidestested in the same species, has encouraged extrapolation of riskassessment when the internal distribution and deposition of thenuclide is similar for humans and the test species.

The availability of occupational and clinical data involvingexposures to radium, radon and/or thorium has provided baselinedata necessary to test the reasonableness and practicability of theapproach of estimating human cancer risks based on animal cancerstudies using internal emitters

Additional methods of analysis have also been explored ABayesian statistical model has been used to evaluate the extrapo-lation of bone cancer risks to humans from 239Pu A collection ofstudies with rats, dogs and humans on the effects of plutonium andradium was used in this effort

Proportional hazards modeling has been employed to analyzeintraspecies studies on the induction of bone and lung tumors byseveral different radionuclides The comparison of bone tumorrisks from 226Ra in mice, dogs and humans has also been examined

by the derivation of a power function relating a time-based eter to skeletal dose rate

param-Extrapolation methodology is obviously quite different forinternally-deposited radionuclides than for external radiationexposures For both situations, the availability of some reliablehuman data is critical to the development of reasonable extrapola-tion procedures The human database provides (1) an essentialfeedback to plan laboratory studies and (2) a target against whichnew methods can be evaluated

1.3 Conclusions

The extrapolation of some radiation-induced risks from tory animals to humans is feasible Extrapolations have been per-formed for many years for external radiation exposures andinternally-deposited radionuclides, though different methods must

labora-be used for the different patterns of radiation exposure

The success of interspecies extrapolation rests upon the mostbasic common factors, which entail somatic-genetic aspects of dam-age and repair of the DNA molecule This baseline then supportsthe consistent pattern of mortality seen among mammalian spe-cies The mortality pattern then provides a quantitative, analyticalbasis to compare and unify the responses of different species toradiation injury.

The analyses in this Report on how data for life shortening can

be used to extrapolate risks across species are very encouraging Ithas been shown how the ratio of the median survival in control pop-ulations, based on a Gompertz distribution, can be used to adjustfor the differences in life span among mammalian species Whenthese adjustments are used, the age patterns of radiation-inducedmortality are markedly similar, thus allowing the extrapolation ofrisk to be made There are, however, two differences that have to beconsidered First, the fact that the animal data have been derivedfrom animals that were exposed at the same age, whereas, the datafor humans are from a general population in which persons of allages were exposed Second, the animal data are from populationswith a restricted gene pool and with susceptibility to certaintumors, whereas, human populations are considered to be hetero-geneous There are suggestions as to how to overcome these differ-ences Even with these remaining issues it is suggested that itwould be better to use life-shortening data for selecting values forDREF and RBEs because they are more representative of the totalradiation effect than the data currently used that are restricted to

a small number of relevant tumors

It has been demonstrated that, at least for external exposures,

a life-table-based model will provide an accurate interspecies diction of death rates when summed across all causes of mortality

pre-or from all solid-tissue tumpre-ors However, extrapolation of specifictumor mortality still requires some development The findings forall-cause mortality support the recommendation that animal data

on the effects of high-LET radiation (e.g., neutrons) can be

reason-ably used to predict total mortality in human populations wherefew analytically useful data for humans exist

The approach of extrapolation based on life shortening, the ative toxicity ratio and one of the proposed approaches to extrapo-lation of risk of solid cancers induced by external radiation depend

metabolic parameters in animals has led to sensible extensions

to other radionuclides in test species with extrapolation back tohumans This type of extrapolation modeling seems effective inthose cases where uptake, distribution, deposition, and pathologicresponse need to be accommodated by the prediction model

1.4 Recommendations

1 There has been productive work on the comparison of tion-induced life shortening among animal species and itsextrapolation to humans The current values of DDREF andradiation weighting factors are based on experimental databut the animal tumor data are considered inadequate; it isrecommended that the use of the life-shortening data beconsidered The studies of life shortening suggest that thedata for this effect might be used with advantage to derivevalues for DREF and RBEs because they provide an inte-grated index of both cancer and noncancer effects Such datashould be obtained from appropriate regimens of protractedbut terminated exposures It is also important to focus onspecific cancer types In these analyses, the appropriate celltypes and cellular events should be linked with the animalpathologic response By incorporating molecular informa-tion, biodosimetry and radiation effects at the molecularlevel, uncertainties should be reduced and the quality ofanimal extrapolations enhanced

radia-2 It is important for risk extrapolation that the results fromanimal studies are archived, maintained and made avail-able to researchers This would involve good documentation

in electronic form for the individual studies and tion of possible problems with the studies This resourcethen could be used by more investigators concerned with theissue of animal extrapolation

identifica-3 Develop both human epidemiological and experimental mal data that are based on specific types of cancer and notjust cancer site

ani-4 Continue to develop and compile information about the parative aspects of the mechanisms of radiation-inducedcancer

com-5 Develop approaches to extrapolation that take into accountthe problem of age-dependent susceptibility and differences

in the heterogeneity of human and experimental animalpopulations

6 Test the hypothesis that susceptibility for induction of cer by radiation is related to the background rate.

can-7 Extrapolation of risk estimates from experimental systemsare still required for radiations such as mid-energy neu-trons and heavy ions and this entails obtaining relevantanimal data Furthermore, if values of DDREF and radia-tion weighting factors continue to be based on experimentaldata, additional data for life shortening and cancer should

be sought

8 To conduct space flights safely, it is important to obtaininformation about the potential adverse health effects fromheavy-ion exposures There are limited animal studies inthis area, and it is recommended that attempt to determinethe predictive value of neutron exposures for estimating therisk of heavy ions be continued It would be important tovalidate these predictions by conducting whole animal stud-ies using actual heavy-ion exposures

9 The use of high-dose diagnostic radiology and nuclear cine with both acute and chronic exposures places largenumbers of patients at potential risk It is recommendedthat animal and human data be used to estimate futurerisks of these procedures

Epidemiological studies have provided the basis for tive estimations of radiation-induced cancers in humans Theindividuals exposed in these studies received mainly acute orfractionated low-LET exposures of gamma or x rays For high-LET radiation, radon exposures have been studied in miningpopulations as well as in homes Also the effects of radium andthorium in patients and occupationally-exposed populations havebeen assessed

quantita-From these data, dose-response estimates have been developed.However, there is a continuing need to understand and estimatethe effects of continuous exposures and other dose-rate patterns aswell as the effects of particular radionuclides and ion particles Forexample, proposed space travel requires estimation of healtheffects from heavy-ion exposures Epidemiological studies do nothave sufficient numbers of exposed humans to provide risk esti-mates for these types of exposures What has been done in the past

is to use experimental animal systems to provide the needed riskestimations This has been particularly true for DREF, radiationweighting factors and exposures to materials such as plutonium.Previous NCRP reports either used the results of animal extrap-olations or directly carried out such extrapolations NCRP Report

No 128, Radionuclide Exposure of the Embryo/Fetus (NCRP,

1998), made use of the animal results but expressed concern with

the validity of the extrapolations, while Report No 64, Influence of

Dose and Its Distribution in Time on Dose-Response Relationships for Low-LET Radiations (NCRP, 1980), directly made calculations

based on animal data

This Report reviews the usefulness of experimental animal data

in the prediction of human oncogenesis from exposures to ionizingradiation Several sections address the relevant issues

Section 3 provides a review of the statistical approaches used inthe extrapolation of data from nonhuman experimental systems tohumans

Section 4 presents a consideration of tissue and organ ences among species with an emphasis on those cells that are theorigin of the radiation-induced cancer The section is furtherdivided by organ systems, with discussion of the significance of

differ-each particular organ in terms of radiation carcinogenesis and itsrelation to extrapolation to human risk.

Section 5 is concerned with radiation effects at the molecularand cellular level Specific types of DNA damage and repair as well

as genetic susceptibility are reviewed Mutations and chromosomalaberrations as well as specific tumor genes are considered withrespect to similarities in mammalian species This Section also con-siders the effect on the development of cancer by changes in geneexpression and the extracellular matrix

Section 6 deals with specific quantitative extrapolation models

It begins with a discussion of extrapolation for chemically-inducedcancers The section describes methods used for prediction amongspecies of the carcinogenic effects of external radiation Thisincludes the use of Bayesian methods for estimation of dose-rateeffects The section concludes with a discussion of extrapolation ofthe effects of internally-deposited radionuclides in laboratory ani-mals to humans

Nonhuman Experimental Systems to Humans

3.1 Introduction

The study of the mechanisms of biological effects of radiationhave involved research at the whole-animal, tissue, cellular, chro-mosomal and, lately, molecular levels These studies have eluci-dated many aspects of the effects of radiation but many importantquestions remain, especially some that are important in the deter-mination of risk estimates required for radiation protection Theestimate of risk of radiation induction of germ-cell mutations thatresult in inherited diseases, and of radiation induction of cancer arefundamental for setting appropriate radiation limits for the gen-eral and working populations Risk estimates of cancer induction

by acute low-LET radiation based on the study of humans havebecome increasingly more accurate However, the risk posed byexposure to low-dose-rate, fractionated, or protracted irradiationcannot be accurately determined from available data from studies

on humans, and estimates must rely on experimental systems.Also, estimates of the carcinogenic or genetic effects of radiationssuch as neutrons and heavy charged particles, cannot currently bebased on human experience Despite studies on large populations,such as the atomic-bomb survivors, estimates of genetic risksdepends in large part on studies of the mouse

Extrapolation of risk estimates across species has been used invarious aspects of protection against the effects of chemical agents

as well as radiation In the case of chemical and chemotherapeuticagents, the determination of the toxicity and threshold levels have

depended on animal studies Schneiderman et al (1975) wrote a

guide to a road map of how one might get “from mouse to man.”Despite the “detours, chuckholes, swamps, quagmires and deadends that may be encountered,” they concluded that it was possible

to extrapolate from mouse to humans, not with precision but atleast usefully Data from animal experiments remain the basis ofthe estimates of setting toxicity and threshold levels, and theseapproaches are discussed in this Report

The need for data on the effect of radionuclides became clearearly in the days of development, testing, deployment and use ofatomic weapons and nuclear power How were safety standards forworkers in the weapons and energy industries to be set withoutadequate data from human experience?

There was some knowledge of the effects of exposure to radiumdating back to the Curies and the technical and medical staffsinvolved in radium therapy An increasing amount of data, espe-cially, on the induction of bone tumors, was revealed from study ofradium dial painters that spanned many decades (Evans, 1969;Fry, 1998; Martland, 1929; Rowland, 1994; Stannard, 1988) Stud-ies of the exposure to radon in uranium miners has provided esti-mates of the induction lung cancer (IARC, 2001; NAS, 1999).The U.S Atomic Energy Commission set about solving the needfor data to assist in setting protection standards for nuclear work-ers It was in this endeavor that studies using dogs were added tothose on mice Claus (1976), the Special Assistant to Shields War-ren, Director of the Division of Biology and Medicine, recorded in aflamboyant style the events that led up to the beagle project at theRadiobiology Laboratory at the University of Utah to study radio-nuclides “ the Great White Father and his name was War-ren, and the Father turned to his noblest son John Bowers”(Director of the Radiobiology Laboratory in the early 1950s) And

he said, referring to plutonium and radium, “go thou hence Johnand bring light unto the benighted And John did and he calledupon his henchmen Evans, Brues, Eisenbud, Brandt and Clausand together with their Great White Father they gathered andenunciated the Dogma, that as radium in man shall be unto radium

in the dog, so shall plutonium in man be as plutonium in the dogand the name of the dog will be the beagle Now these mighty wordswere hailed abroad as one giant step for mankind—but they were

a very small step for the beagle.” This is the concept that is known

as the relative toxicity ratio and that was the first approach toquantitative extrapolation across species of risk estimates for radi-onuclides

Needless, to say, this attractive but somewhat fanciful historicalaccount left out some salient steps The first studies on the compar-ative carcinogenicity of 239Pu and 226Ra were carried out by Brues(1951) and Licso and Finkel (1946) at the Argonne National Labo-

ratory (ANL) using mice, rats and rabbits (see also Mays et al.,

1986a) In 1943, Evans introduced the concept of dosimetry ratiosthat subsequently became termed as “toxicity ratios.” The toxicity

ratio is analogous to the RBE introduced by Failla and Henshaw(1931) In 1950, the toxicity ratio of 239Pu/224Ra in rodents was used

to derive the maximum permissible body burden of 0.04 C of 239Pufor occupational exposures (Brues, 1951; Langham and Healy,1973) It might be noted that the follow-up studies of the workerssuggest that the limits set were wise ones The experimental stud-ies on radionuclides are discussed in this Report

In 1980, NCRP Report No 64 (NCRP, 1980) examined all theavailable data on the effect of dose rate on a wide range of biologicalsystems, including cancer and life shortening in experimental ani-mals Values for the DREF were determined as the ratio of theeffect per unit dose of radiation at a high-dose rate and the effect at

a low-dose rate based on linear regression coefficients fromdose-response curves for a small number of types of tumors in threestrains of mice Other data from rats and dogs were examined Itwas concluded that the values for DREF were between 2 and 10depending on the endpoint In 1991, the International Commission

on Radiological Protection (ICRP) included low dose and nated the factor as low DDREF, which appears to imply a curvilin-ear response ICRP chose a value of two for DDREF (ICRP, 1991)

desig-If the responses on which the DDREF is based are curvilinear, what

is the reason for the apparent linearity of total cancers as a tion of dose for the data from the atomic-bomb survivors (Preston

func-et al., 2003)? It should be noted that the value of DDREF was

con-sidered a major source of uncertainty in the estimate of risk forradiation protection purposes (NCRP, 1997)

These examples of the use of experimental data involve olation of risk estimates across species In the radiation-risk esti-mates used in radiation protection the estimates are transferredfrom the Japanese population (a form of extrapolation) to deriveinternational dose limits This has been performed using two mod-els, the additive and multiplicative (UNSCEAR, 1994) The addi-tive model assumes that the average excess risk, in any population,given the same distribution by dose, gender and age at exposurewill be the same for similar follow-up periods The multiplicativemodel assumes that the ratio of excess risk to baseline risk at anyage is invariant over populations with different baseline risks Thebaseline risks for various cancers varies among populations Studies of extrapolation across species are needed becausethere are insufficient data for humans to estimate: (1) hereditaryeffects; (2) the influence of dose rate, fractionation and protraction;(3) the effect of high-LET radiations, with the exception of alphaparticles; and (4) how to transport risk estimates across popula-tions as well as species

extrap-The influence of ionizing radiation on life shortening and thecomparative effects among strains of mice and between mice anddogs have provided insights into problems involved in extrapola-tion of estimates of radiation risks across species Studies of lifeshortening began early in the atomic age (Brues and Sacher, 1952;Sacher, 1950a; 1950b) At a symposium on the delayed effects ofwhole-body radiation in 1959, George Sacher suggested that theproblem of long-term effects from animals to humans could be sep-arated into two subsidiary problems The first is the determination

of the mechanism of the radiation response and the second is to findways to estimate the relevant species parameters (Sacher, 1960)

In the early days of studies of radiation-induced life shortening

it was considered that the effect was nonspecific but experimentalevidence indicated that the effect was due to increase in the sametypes of cancer and normal tissues that caused mortality in unex-posed populations Exposure to radiation increased the probability

of an excess of cancer and did so in a comparable manner in all cies studied The early studies identified a relatively invariant lifeshortening of ~28 d per 1 Gy in six genetically different strains ofmice (Grahn, 1959) If the dose rate of the protracted exposures wasless than ~0.10 Gy d–1 the life shortening was decreased to ~4 d per

spe-1 Gy (Grahn and Sacher, spe-1968) More recent studies have provided

a comparison of life shortening in mice, dogs and humans whichwill provide a guide to extrapolation of risks for cancer at specific

sites (Carnes et al., 2003).

The need to establish exposure standards for humans underrelevant exposure conditions is the principal motivation behindinterspecies modeling in the risk analysis framework At present,available data may be insufficient for some exposure conditions

(e.g., protracted occupational exposures) and nonexistent for

oth-ers For example, humans have been exposed to high-LET

radia-tions (e.g., neutrons and heavy ions); however, health effects have

been assessed except in a narrow set of circumstances Therefore,risk estimates of cancer incidence, prevalence and fatality cannot

be made based on human exposures For similar reasons, it isimpossible to estimate directly the influence of changes in dose rateand dose protraction on the induction of cancer in humans by radi-ation It is probable that the determination of the influence of thesefactors and estimates of the effects of high-LET radiations will con-tinue to be derived mainly from the results on experimental ani-mals It is essential, therefore, to establish how the estimates ofrisk based on data from experimental animals, may be extrapo-lated to humans

The extrapolation of biological effects across species can beaddressed at empirical, mechanistic and theoretical levels Thepurpose of this Section is to provide a historical context forthe quantitative methods and the biological endpoints usedfor interspecies comparisons In Section 5, a biological rationale forinterspecies extrapolation at the cellular and tissue levels of orga-nization is examined In the final section of the Report, the empha-sis will shift to discussions and demonstrations of methods thathave been used to estimate radiation-induced risks for humansderived from data for laboratory animals

The history, the progress, and the current success in tion of risks across mice, dogs and humans are described in thisReport It is hoped that this Report will stimulate thought andresearch that will result in acceptable methods of extrapolationacross species It is certain that such studies will help our under-standing of what underlies the probability of the induction of can-cer and noncancer effects by radiation

extrapola-3.2 Lessons Learned from Genetic Risks

This Report will not address the issue of interspecies tion of genetic risks, which is a topic beyond the scope of thisReport It is reasonable, however, to ask if anything might belearned (about interspecies extrapolation of somatic risk) by exam-ining the procedures and problems geneticists faced, since theyhave had to deal almost exclusively with data from laboratory ani-mals and with theoretical concepts

extrapola-Concern about the possible genetic consequences in humansexposed to ionizing radiation date to the late 1920s, when the cor-relation between radiation exposure and mutation frequency wasclearly established (Muller, 1927) This concern became a majorconsideration during World War II, and several large-scale labora-tory studies were initiated under the auspices of the Manhattan

Project (Charles, 1950; Deringer et al., 1954; Spencer and Stern,

1948) By the mid-1950s, concerns about somatic and germ-lineradiation injuries entered public discussions of political and mili-tary policy

The increasing frequency of the testing of nuclear weapons inthe atmosphere introduced the inescapable exposure of wholenations to low levels of ionizing radiation that resulted from thefallout of radionuclides produced by the fission of uranium and plu-tonium Accordingly, public concerns were addressed by the nearlysimultaneous establishment in 1955 of several expert committees

that were commissioned to evaluate existing knowledge and to putforth recommendations on genetic risks to governing bodies Thecommittees and/or reports were the Committee on the BiologicalEffects of Atomic Radiations, U.S National Academy of Sci-ences/National Research Council (NAS/NRC, 1956), the Medical

Research Council of the United Kingdom report on The Hazards to

Man of Nuclear and Allied Radiations (MRC, 1956), and the United

Nations Scientific Committee on the Effects of Atomic Radiation(UNSCEAR); the latter issued its first report in 1958 (UNSCEAR,1958)

Forty years ago, genetic effects were the primary concern, butthere were no simple methods for quantitative prediction of thenature and magnitude of genetic risks to humans following expo-sure to radiation The situation was somewhat better regardingsomatic effects, with an accruing body of knowledge of radiationeffects from the long-time use of radiation in medical diagnosis andtherapy, from industrial applications, and from the use of atomicbombs Geneticists, however, had only the satisfaction of knowingthat genetic mechanisms are essentially invariant across species

3.2.1 Methods of Estimation

Over the years, several procedures have evolved as a means ofestimating the genetic effects of radiation exposure in humans.These are briefly defined in the following paragraphs A detailedhistory of these methods and their application can be found inUNSCEAR (1988) In this Report, the primary concern is with themethods of estimation and how these may have contributed tothe development of extrapolation models for somatic risks

3.2.1.1 Doubling-Dose Method The doubling dose is that dose

con-sidered to double the average spontaneous mutation rate in thegenome of a species The reciprocal of the doubling dose is calledthe “relative mutation risk.” When the relative risk is multiplied bythe spontaneous genetic burden, the product is the expectedincrease in that burden induced by an absorbed dose of 1 Gy It

is assumed either that the spontaneous burden is all sustained byrecurrent mutation (as for fully expressed dominant genes) or thatsome known portion of the burden is sustained in this way.That portion is known as the “mutation component” of the sponta-neous burden and entails an additional multiplier The doubling-dose method is often called the “indirect method.”

The origin of the concept is not certain, but may have first beenidentified in a paper by Wright (1950), although the idea derivesfrom a number of considerations These include the assumption ofsimple proportionality between dose and response and the interest

in identifying the proportion of spontaneous mutations able to natural background radiation Discussions on the magni-tude of the doubling dose as a result of the assessment of clusteredmutations lead to the conclusion that the current estimate ofgenetic risk is too high (Russell and Russell, 1996; Selby, 1998a;1998b)

attribut-3.2.1.2 Direct Method Because the doubling-dose method was so

indirect, it was natural for geneticists to seek a means of lating risks more directly to humans from data obtained fromexperimental animals The direct method employs experimentallyderived mutation rates for specific classes of mutation that haveclinically defined counterparts in humans and can thus be usedmore or less directly to estimate the risk in humans The methodwas developed by Ehling (1976; 1991) and Selby and Selby (1977)

extrapo-3.2.1.3 Gene-Number Method This approach evolved from the

con-cept that if the total number of mutable loci were known for thehuman genome along with a reasonable estimate of the averagemutation rate per locus per gray, then one could predict the totalnumber of new mutations induced by a given radiation exposure.The approach reflected Muller’s concerns about the existing andpotentially added load of mutations in humans and the conse-quences thereof in terms of the frequency of “genetic deaths”(Muller, 1950)

3.2.2 Discussion of Methods of Estimating Genetic Risk

Of the three methods described, the third (the gene-numbermethod) has fallen into disuse There were too many unknowns,and the method could not produce useful quantitative values thatcould be related to the known genetic burdens borne by society.Present-day risk estimates use a mix of indirect and directmethods There are strong proponents and rationalizations for bothmethods An excellent history of their application for risk assess-ment is given in UNSCEAR (2001) An analysis of the strengthsand weaknesses of the two methods is in Sankaranarayanan(1991a) Both methods require matters of judgment rather thanmeasurement and thus lack estimates of variability The direct

method is not as direct as the name implies and requires complexgenetic and clinical matters (NAS/NRC, 1990) The doubling-dosemethod suffers from inadequate knowledge of radiation-inducedmutation rates for specific classes of human genetic diseases and,especially, of the magnitude of the mutation component of the vastclass of genetic conditions listed as congenital abnormalities andother disorders of complex etiology (NAS/NRC, 1990) As men-tioned above, there is also concern with regard to the magnitude ofthe doubling dose as a result of considerations of how to handleclustered mutations.

Gene expression profiling is a very new technique whereby onecan determine whether there is a propensity for a patient todevelop a neoplastic disease, or if the disease is already present,

to gain insights into its type and state of development (West et al.,

2001) The present genetic analyses indicate three possible nisms for the development of a disease There can be: (1) impair-

mecha-ment of a DNA repair pathway, (2) transformation (i.e., mutation)

of a normal gene into an oncogene (i.e., a cancer gene), or (3)

mal-formation of a tumor-suppressor gene

This new, emerging and growing technology has involved a cess known as gene profiling as undertaken in DNA microarraytesting sequences One can examine thousands of genes from a sin-gle tumor sample through analyses of messenger RNAs An indi-vidual messenger RNA is a direct complementary representation of

pro-a portion of encoding DNA (i.e., the gene encodes pro-a messenger, pro-and

by assaying the messenger one can decipher the gene) Suchmicroarray analyses allow for a sampling of literally thousands ofmessengers, and therefore, assay for genetic bases In this way,usually through statistical analyses, one can assay for gene expres-sions associated with a particular tumor type Once tumor-specificmessengers are identified, then that information can be used forscreening purposes to determine if noncancer patients have thegenetic potential for such a disease

Gene expression profiling is perhaps best known for breast

can-cer, where two human breast cancer genes (BRCA1 and BRCA2)

have been identified These genes are of the mutation type, andtheir individual presence indicates an increased potential for devel-

oping breast and/or ovarian cancer BRCA1 is a mutation on mosome 17 and BRCA2 is a mutation on chromosome 13.

chro-The field of gene expression profiling is a fast moving one chro-TheNational Center for Biotechnology Information is charged with pro-

viding daily updated information in this area through its Genes

and Disease website (NCBI, 2005) Presently, the website lists

breast cancer, Burkitt lymphoma, colon cancer, leukemia (chronicmyeloid), small-cell lung carcinoma (SCLC), multiple endocrineneoplasia, neurofibromatosis, pancreatic cancer, polycystic kidneydisease, and prostate cancer as having a genetic basis to its etiol-ogy There is also a wide range of similar information availablefrom commercial sites dealing with animal diseases and microar-rays by which to undertake specific analyses.

3.2.3 Role of Genetics in the Estimation of Somatic Risks

Somatic and genetic risk assessments progressed dently with little or no cross communication From the mid-1950s

indepen-to the mid-1980s, the data on sindepen-tochastic somatic effects in humanscontinued to accrue and become more compelling regarding ancil-lary variables such as age, sex, dose and cell type for neoplastic dis-eases Animal data seemed less important except for questionsabout the influence of dose rate and radiation quality and esti-mates of the risk of cancer from internal exposure to radionuclides.Great strides were also being accomplished in genetics but, iron-ically, as noted by Crow (1982), “ the magnificent accomplish-ments of molecular biology have diminished rather thanaugmented our confidence in quantitative assessment of thehuman genetic risk.” As Crow added, this new knowledge creatednew uncertainties Yet, in radiation genetics, most of the paramet-ric values needed for extrapolation were now available for themouse

Since molecular and cellular biology had not solved the lems involved in expressing human genetic risks, what was miss-ing? At least one major item was missing, namely, reliablemeasures of the radiation-induced mutation rates for representa-tive human genetic defects While estimates of radiation-inducedinduction rates were developed for several indicators of geneticdamage among the children of atomic-bomb survivors in Japan

prob-(Neel et al., 1990), no measures were statistically significant and

some were essentially zero or negative The absence of even a singlebaseline mutation rate has made it extremely difficult to reduce theuncertainty of genetic risk estimates for humans

The role of genetics then is that baseline values of inductionrates derived from human populations are absolutely essential toreasonable and quantitative extrapolation modeling Somaticistshave the needed parameters, geneticists do not, at least not to thesame degree of certainty as exists for induction rates for severalneoplastic diseases The present availability of transgenic mice

may provide an experimental system by which radiation-inducedgenetic risks of multilocus chronic diseases can be quantitated forsubsequent application in extrapolating to human genetic risk.

3.3 Somatic Risks

The history of species comparisons of mortality can be viewed as

an effort to acquire relevant data and as an evolution in the opment of endpoints used for comparison Although the availability

devel-of data is less devel-of an issue today than in the past, neither a sus on the appropriate biological endpoints nor agreement on thequantitative methods for extrapolation has emerged Pearl (1922),

consen-in his search for a fundamental law of mortality, may have been thefirst investigator to compare species formally His comparisonswere based on vital statistics (Pearl, 1923) arranged in standardactuarial tables

Potentially large differences in the length of life represent themost obvious obstacle to making comparisons between species.Pearl (1922) chose to superimpose “two biologically equivalentpoints” within the life cycles of the organisms being compared toadjust for differences in life span He used the elapsed time fromthe age where death rates reach a minimum and the age where sur-vivorship is reduced to one individual per 1,000 to convert (normal-ize) time into units expressed as “centiles of the life span.” Usingthis scaling approach, Pearl (1922) produced normalized survivor-

ship curves for Drosophila and humans that were sufficiently alike

to suggest that the laws of mortality were fundamentally the samefor the two species Quantitative differences between the survivalcurves for the two species were attributed to humans’ ability to con-trol their environment A year later, Pearl (1923) added the rotifer

(Proales) to his comparisons and issued a plea to the scientific

com-munity to collect mortality data on other organisms

Using data from the field of animal husbandry and nutrition,Brody (1924) also attempted to characterize quantitatively what hecalled the “kinetics of senescence.” Brody’s model derived from Loeband Northrop (1916; 1917) observations that a number of life phe-nomena, including duration of life, has a temperature coefficient,which is related to an organism’s body temperature, characteristic

of chemical reactions As such, duration of life was thought to have

a physiological basis determined by the time course required tocomplete a series of chemical reactions

Using a variety of biological endpoints and species of animals

(e.g., milk production of dairy cows, egg production of domestic fowl,

survival of fibroblasts in the serum of domestic fowl of different

ages, wound healing in humans, Pearl’s data on Drosophila

mortal-ity), Brody (1924) demonstrated that the behavior of these diverseendpoints through time could be described by a simple equation forexponential decline, which is termed “the law of monomolecularchange in chemistry.” Supplemented by examples drawn from

existing mortality data for humans (e.g., cancer, diseases of the

arteries, and diseases of several organ systems), Brody (1924)argued that age and disease-specific death rates represented aquantitative measure of senescence and the normalized reciprocal

of the death rates could be interpreted as the time course of thegeneral “vitality” of an organism

Greenwood (1928) attempted to provide a biological perspective

on the “order of dying-out” described by a life table Specifically, hefocused on the physiological implications of the actuarial modeldeveloped by Gompertz (1825) The lack of arithmometers and theprevailing attitude of actuaries that the life table was just a work-ing tool were cited by Greenwood as the reasons why Gompertz’sgeneralization of his simple graduation model to be more biologi-cally plausible (by incorporating individual heterogeneity) hadbeen forgotten (Section 6) By example, Greenwood also demon-strated that the extension of the Gompertz equation by the famousactuary Makeham (1867), who adjusted for factors affecting mor-

tality independent of age (e.g., environmental factors), made

nei-ther biological sense nor provided any significant improvements tothe graduations provided by the earlier, simpler Gompertz equa-tion Greenwood used the numerical uncertainties of life-table sta-tistics, estimated in the tail of the dying-out distribution, whereonly few individuals remain, to question the scaling approach ofPearl (1922) Instead, he used the expectation of life from the age

of minimum mortality as a scaling device for comparisons amongspecies Using this approach, Greenwood concluded that “we have

no sound reason for thinking that the force of mortality in miceincreases with age more nearly geometrically than the force of mor-tality in men” nor is there any “reason to think that any more com-plex formulation of a physiological law would describe the observedfacts better than Gompertz’s century-old simple formula.”

Over the next decade, mortality data for a variety of organisms

(e.g., saturniid moth, roach, domestic fowl, mice) began to

accumu-late In response to the biological and statistical criticisms byGreenwood (1928) of their original method of life-span adjustment,Pearl and Miner (1935) began expressing time as “percent devia-tions from the mean duration of life.” This method of adjusting for