Environmental Contaminants in

December 1979

NTIS order #PB80-153265 Library of Congress Catalog Card Number 79-600207

For sale by the Superintendent of Documents, U.S. Government Printing Office Washington, D.C. 20402 Stock No. 052-003-00724-0 FOREWORD

This report presents the major findings of an OTA assessment of Federal and State efforts to deal with the environmental of food. Undertaken at the request of the House Committee on Interstate and Foreign Commerce, the study examines both regulatory approaches and monitoring strategies for coping with contaminated food. The assessment is concerned with chemical and radioactive contaminants that inadvertently find their way into the human food supply. To bring the scope of inquiry within manageable bounds, we excluded naturally occurring such as fungal and microbial toxins. The Office of Technology Assessment was assisted by two advisory panels of scientists and representatives of public interest groups, agriculture, the chemical industry, fisheries, and State and foreign governments. The Food and Drug Ad- ministration, the Department of Agriculture, and the Environmental Protection Agency each designated staff members to attend panel meetings, provide back- ground information, and review draft reports, Background papers were commis- sioned concerning the scientific aspects of detecting and regulating environ- mental contaminants in food. The Congressional Research Service provided five analyses of previous food contamination episodes, Reviews of the draft report were provided by the advisory panels, Federal agencies, and a number of inter- ested individuals not previously involved with the assessment. Because this assessment addresses concerns of American citizens as well as policy makers and scientists, the summary of the report is also being published as a separate document. The summary provides the interested citizen with an in- formative and clear overview of this complex problem. Copies of the summary can he obtained free of charge from the office of Technology Assessment, U.S. Con- gress, Washington, D.C. 20510.

JOHN H. GIBBONS Director

/// .-

Tolerance Advisory Panel

James Anderson Robert Jackson Richard Merrill Manager for Material Safety Assistant State Health Commissioner School of Law Allied Chemical Corporation Director, Office of Health Protection and University of Virginia Environmental Management Louise Burke Virginia State Health Department Joseph Rosen Executive Director Kazuo K, Kimura Department of Consumer Council of Virginia Foundation School of Medicine Rutgers University Consumer Council of Virginia, Inc. Wright State University Kenneth Crump Marvin Legator John Rust Department of Mathematics and Statistics Director Professor Emeritus Louisiana Tech University Division of Environmental Toxicology Pharmacology/Radiology Deportment of Preventive Medicine and University of Chicago Donald Epp Community Health Department of Agricultural Economics and University of Texas Medical Branch Gary Van Gelder Rural Sociology David Lindsay College of Veterinary Medicine Pennsylvania Stale [University Technical Secretary University of Missouri Steering Group on Food Surveillance Frederick Halbert Ministry of Agriculture, Fisheries, and Dairy Farmer John Van Ryzin Food Delton, Mich. RAND Corporation United Kingdom Joseph Highland Roy Mart in George Whitehead Chairman Director Deputy Director Toxic Chemicals Program Science and Technology Office for Consumer Affairs Environmental Defense Fund National Fisheries Institute, Inc. Michigan Department of Agriculture

Monitoring Advisory Panel

Raymond Allred William Cooper S. Roy Koirtyohann Director Professor Environmental Trace Substances Center Environmental Conservation Research Department of Zoology University of Missouri Continental Oil Company College of Natural Science Michigan State University Charles Krister Richard Anderson Retired Manager Chairman John Davies Chairman Biochemicals Department Statistics Department E. I. DuPont de Nemours and Company University of Kentucky Department of Epidemiology and John Bourke School of Medicine John Laseter Director Universitv of Miami Director Analytical Division Blake Early Center for Bio-Organic Studies Food Science Department Environmental Action, Inc. University of New or leans Cornell University Edward Elkins Vaughan Bowen Director John Martin Woods Hole Oceanographic Institute Chemistry Division Moss Landing Marine Laboratories Washington Research Laboratory Shirley Briggs National Food Processors’ Association Rebecca Sharitz Executive Director Ronald J. Hingle Savannah River Ecology Laboratories Rachel Carson Trust for the Living University of Georgia Environment, Inc. Administrator Food and Drug Control Unit Leon Chesnin Bureau of Environmental Services Gerald Ward Department of Agronomy Louisiana Office of Health Services and Department of Animal Science University of Nebraska Environmental Quality Colorado State University

Iv Food and Renewable Resources Program Staff

Joyce C. Lashof, Assistant Director Health and Life Sciences Division J. B. Cordaro, Program Manager* Walter Parham, Program Manager Catherine Woteki, Project Director Robert Smith, Research Assistant Chris Elf ring, OTA Congressional Fellow

Administrative Staff Phyllis Balan Elizabeth Galloway Ogechee Koffler

Contractors Lloyd W. Hazleton, Technical Advisor to Tolerance Panel Robert Huggett, Technical Advisor to Monitoring Panel Samuel Iker, Editor

OTA Publishing Staff

John C. Holmes, Publishing Officer Kathie S. Boss Joanne Heming

Food Advisory Committee

Martin E. Abel, Chairmen Lorne Greene Robert 0. Nesheim Senior Vice President Chairman of the Board Vice President, Science and Schnittker Associates American Freedom From Technology Hunger Foundation The Quaker oats Company Johanna Dwyer, Co-Chairman Kathleen O’Reilly Director Richard L. Hall Frances Stern Nutrition Center Director Vice President, Science and Consumer Federation of America Technology David Call McCormick & Company, Inc. R. Dennis Rouse Dean, College of Agricultural Dean, School of Agriculture and Life Sciences Auburn University Cornell University Laura Heuser Member, Board of Directors Lauren Seth Cy Carpenter Agricultural Council of America Agricultural Consultant President Minnesota Farmers Union Arnold Mayer Thomas Sporleder Legislative Representative Professor of Agricultural Eliot Coleman Amalgamated Meat Cutters and Economics Director Butcher Workmen of North Texas A&M University Coolidge Center for the America Sylvan Wittwer Study of Agriculture Director and Assistant Dean Almeta Edwards Fleming Max Milner College of Agriculture and Social Program Coordinator Executive Director Natural Resources Florence County, S.C. American Institute of Nutrition Michigan State University

*Resigned June 16, 1979, and continued to serve under contract to assessment completion. Contents

Chapter Page

I. Summary ...... `“ “ “ “ ● ● ● “ ● 00 G o “ ● “ o ● ● o ● o ● “ o ● 00 “ ● “ o “ “ “ 3 Health Impacts of Environmental Contamination...... 4 Economic Impacts of Contamination...... ““ .“ . ““. ““. “ .“ 4 Major Problems in Identifying Environmental Contaminants ...... 5 Problems of Regulating Environmental Contaminants...... 5 Findings and Conclusions ...... “ “ “ “ . “ “ “ . m “ “ “ o “ o “ . “ . “ “ “ “ 6 Congressional Options...... ” . . “ “ “ “ .0. “ o ‘. “ c o “ o “ “ o “ “ “ ““ 9 Option l.--Maintain the Present System ...... ”””’ ...... ”.”””. 9 Option2.—Amend the Food, Drug, and Cosmetic Act ...... 9 Option 3 .—Establish an Investigatory Monitoring System...... 10 Option 4.— Improve Federal Response to New Contamination Incidents . . 10 15 II. Environmental Contamination of Food.”~””.”. .“””.’””ooooQo ● +00 ”00”0 How Food Becomes Contaminated ...... ””””” ....”””..”.... .“”” 16 Magnitude of the problem ...... “.”””””.”..”. “...” 19 Evidence of Human Illness Resulting From Consumption of Contaminated Food...... ””””””.””.” “o” ”””.. 19 Polychlorinated Biphenyls...... “ ~ . “ “ m “ “ ~ 0. ~ 0. “ “ o ~ . 19 Polybrominated Biphenyls ...... ”””” ..”””.””” .“.”.”.”””.” 20 and Methylmercury ...... “ . “ “ ~ ~ .“ “ “ ~ “ . “ “ 0. 21 Radioactivity ...... “ “ “O 00 “ ““ ~ s o “ ~ 000 s Q “ o “ “ s “ o “ 22 Number of Food Contamination Incidents...... “ . ““” ~ “ “ “.” 23 Economic Impact ...... ””””.”” “0”””00.0””””. ““s””” 25 Costs of Food Condemned...... ”.”. “. ““””.””””.”” 25 Health Costs...... ,..””””.” 0“0.o’0”””””00 ““”””””.. 26 Distributional Effects and Costs ...... “ ““ “ . “ . “ “ “ “ “ .0 “ ““ 27 Potential Food Contamination problems...... “ .“ “ . “. ‘“” ““ “ “ 29 Conclusions and Issues ...... “ . “ “ ~ “ . ~ ~ . ~ “ ~ . “ “ “ “ o “ “ o ~ o “ s “ “ 29 Chapter II References ...... “ “ “ ~ “ . . . “ 0.0. s o “ 0000. s “ “ . 30 35 111. Federal Laws, Regulations, and Programs” ...... “ “ “ “ “ 00 ● o “ o “ ● 0000 Federal Laws...... “ “ .“” “ “ ““ “O . “ “ “ 000 “ “.” ~ o ‘ . “ 00 “ .- e o 35 Action Levels and Tolerances...... “ ‘ “ “ “ “ “ ~ “ “ ~ “ “ “ ~ o “ o “ “ o “ “ ~ 37 Criteria forgetting Action Levels and Tolerances ...... 38 Determining Action Levels and Tolerances ...... “ . . “ . “ “ “ “ . . 39 Evaluation of the Scientific Data...... “” . ““ “ ““ ~ ““” “ “ “ 39 Evaluation of Decisionmaking Approach ...... ““ . .“ “.” . “ “ 41 Federal Monitoring programs...... “ . “ “ . “ “ 000 “ “ s m “ . “ “ “ “ ~ 42 Food and Drug Administration ...... “ “ . “ “ “ “ “ “ “ “ .0 “ o ‘ “ “ “ . 42 U.S. Department of Agriculture ...... “ . “ “ ‘ “ “ “ “ “ “ “ o “ “ 0.00 43 Environmental Protection Agency ...... ”..”.”.o ““” ”s”s.”s 44 Critique of Federal Monitoring programs...... “ . . . “ . “ 45 Chapter III References ...... ““” ““”” ~ “ “ “ .’. o “ .0 “ . .“ s . “ o .“ “ 46 49 49 50 52 52 55 V. Methods for Assessing Health Risks.....”“”””OCO””OOOO 0“00”00”0 59 Possible Toxic Effects ...... ””.”””.””””.” .“”..”””””. 59 Human Epidemiology...... ”.”.o.o’s ““0”””””.”.”00 60 Animal Tests and Other Methods for Toxicological Testing ...... 62 The Bioassay...... “ . ““ “ “ “ “ .0 “ “ 000. c “ .0..... 62 Short-Term Tests for Mutagenesis and Potential Carcinogenesis...... 63

vii Contents--continued

Chapter Page Interactions of Two or More Substances ...... 65 Extrapolating From High Doses to Low Doses...... 67 Chapter Preferences ...... 69

VI. Methods of Estimating and Applying Costs to Regulatory Decisionmaking . . 73 Cost Effectiveness ...... 73 Cost-Benefit ...... 76 Application of Methods for Regulation, ...... 77 Chapter VI References...... 78 VII. Monitoring Strategies ...... 81 Strategies...... 81 Regulatory Monitoring...... 81 Investigatory Monitoring ...... , , ...... 81 Monitoring for Suspected Environmental Contaminants ...... 82 Monitoring for Uncharacterized Environmental Contaminants ...... 84 Specimen Banking ...... 85 Sampling ...... 0 ...... 86 Quality Assurance. ,...... 88 Chapter VII References...... 89 VIII. Monitoring Instrumentation ...... 93 Organic Environmental Contaminants ...... 93 Analysis ., ..., ...... , ...... 93 Application to Investigatory Monitoring ...... 96 Detecting and Quantifying Trace Metal Contaminants ...... 97 Analysis ....., ...... 97 Application to Investigatory Monitoring ...... 98 Speciation of Trace Metals ...... 98 Analysis of for Radioactivity...... 99 Analysis ...... 99 Applications to Investigatory Monitoring...... , ...... 102 Estimated Cost to Equip a Laboratory to Conduct Investigatory Monitoring. . 103 Chapter VIII References ...... 105 Ix. Congressional Options...... $ ...... 109 Option 1—Maintain the Present System...... 109 Option 2—Amend the Food, Drug, and Cosmetic Act...... 110 Option 2A--Simplify Administrative Procedures...... 110 Option 2B-Requiret reestablishment of Tolerances ...... 110 Option 2C-Clarify the Role of Economic Criteria ...... 111 Option 2D-Establish Regional Tolerances ...... 111 Option 3-Establish an Investigatory Monitoring System ...... 112 Option 3A-Establisha National Monitoring System for Suspected Environmental Contaminants...... 112 Option 3B-Establisha National Monitoring System for Uncharacterized Environmental Contaminants...... 113 Option 4—Improve Federal Response to New Contamination Incidents . . . . . 113 Option 4A-Designate a Agency ...... 113 Option 4B-Establish a Center to Collect and Analyze Toxic Substances Data ...... 114 Appendixes A. Substances Whose Production or Environmental Release Are Likely to Increase in the Next Ten Years . . . . , . . , ...... , , . . . , . , . . , , . . . . . , . . . , . , ...... 119 B. FDA Derivation of PCB Tolerance—Federal Register. . ., , ...... 125

.,. Vlll Contents--continued

Appendix Page C. Methods for Toxicologic Testing ...... 137 D. Review and Evaluation of Methods of Determining Risks From Chronic Low-Level Carcinogenic Insult ...... 154 E. Measuring Benefits and Costs ...... 166 F. Priority Setting of Toxic Substances for Guiding Monitoring Programs ...... 177 G. Approaches to Monitoring Organic Environmental Contaminants in Food ...... 187 H. Analytical Systems for the Determination of Metals in Food and Water Supplies . 195 I. Analysis of Foods for Radioactivity ...... 215 J. Commissioned Papers...... 228 K. Executive Agency Observers at Advisory Panel Meetings...... `229

LIST OF TABLES

Table No. Page 1. 16 2. 17 3. 17 4. 18 5. 22 6. 26 7, 36 8. 50 9. 51 10. 51 11. 51 12.

52 13. 66 14. 68 15. 69 16. 74 17. 83 18. 95 l9. 98 20. 101 21, 104 A-]. 123 D-1.

164 G-1. 188 G-2.

G-3.

192 H-1. 199 H-2. 201 Contents--continued

Table No. Page H-3, Multielement Atomic Emission Determination Limits for Two Common Plasma Excitation Systems...... 204 H-4. Determination Limits for X-Ray Fluorescence Techniques...... 205 H-5. List of Sample Preparation Facilities Required ...... 209 H-6, Summary of Cost-Productivity Estimates for Various Analytical Techniques . . . 210 I-1. Costs for Space, Furnishings, Equipment, and Supplies .,,...... ,.. 225

LIST Of FIGURES

Figure No. Page l. Distribution of PCB Contamination ...... 8 2. Distribution of States Responding to Questionnaire ...... 24 3. State Food Laws...... 49 4. Scope of Investigation: Distribution of Materials ...... 53 5. Steps in the Ames Spot Test...... 65 6. A General Scheme for the Qualitative and Quantitative Analysis of the Organic EPA Priority Pollutants in Semisolid Foods ...... 94 7. Comparison of the Gas Chromatographic Analysis of a Standard Mixture of Polychlorinated Biphenyls on Two Types of Gas Chromatographic Columns . . . 96 D-I, Illustration of Why the Dose-Response Curve Should Be Linear at Low Dose in the Presence of Background Carcinogenesis ... , ... , ., ...... ,.. 156 D-2. Typical Fit of Mantel-Bryan Curve to Experimental Data , ...... 158 D-3. Comparisons of “Safe” Doses Computed From the Mantel-Bryan Procedure and From a Procedure Based on the Multistage Model ...... , . . . .,, 161 D-4a, Example of Data Exhibiting Downward Curvature ...... 162 D-4b, Example of Data Exhibiting Upward Curvature...... , , ...... ,. 162 E-1 . Analysis of a Shift in Supply .-. , ...... , ...... , ., , ., ...... 0. 170 G-I . A Simplified Diagram for the Qualitative and Quantitative Analysis of the Organic EPA Priority Pollutants ...... , . .,. 189 G-2. A General Scheme for the Qualitative and Quantitative Analysis of the EPA Organic Pollutants in Semisolid Foods ...... , ...... , . . . ,., 189 G-3. General Analytical Scheme To Detect and Monitor New Trace Organics and Priority Pollutants in Food and Water Samples ...... 192 H-1 . Flow Diagram of the General Operational Protocol of a Monitoring Laboratory 197 1-1. Comparison of Gamma Spectra Taken With a Sodium Iodide Detector and a Germanium Diode Detector...... ,..,...... 222 — . . .—

\ Summary

I — — — .—. J

. Chapter 1

Summary

The environmental contamination of food is a nationwide problem. A number of recent incidents dramatically illustrate the potential health haz- ards and economic harm that can be caused by such contamination-animal feeds in Michigan contaminated by polybrominated biphenyls (PBBs), the Hudson River contaminated by polychlorinated biphenyls (PCBs), and Vir- ginia's James River contaminated by kepone. These are some of the more serious of the 243 food contamination inci- dents identified in an OTA survey of the 50 States and 10 Federal agencies. These incidents have occurred in every region of the country. They have in- volved all categories of food. While the OTA survey clearly shows the national character of such contamination, the true extent of the problem is still unknown. The latest major food contamination incident—one not included in the OTA survey-graphically points up the ominous dimensions of the problem. PCBs from a damaged transformer contaminated animal at a packing plant in Billings, Mont. The plant used the adulterated fats to produce meat and bone meal that were sold both to feed manufacturers and directly to farmers. The contaminated feed spread through at least 10 States—polluting poultry, eggs, pork products, and a variety of processed foods (including cake). The result: contaminated food found in 17 States, and hum dreds of thousands of pounds of food products seized or destroyed. This assessment, undertaken at the request of the House Committee on Interstate and Foreign Commerce, examines the adequacy of current Federal and State efforts to deal with the environmental contamination of food. In particular, the study evaluates the effectiveness of 1) Federal and State monitoring systems in detecting contamination episodes before they reach crisis proportions, and 2) Federal efforts to regulate . The study explores alternative approaches to the problem and presents policy op- tions for Congress.

Environmental contaminants in food fall manufacturing practices, and that may make into three categories—synthetic or natural food injurious to health. organic chemicals, metals or their organic and inorganic derivatives, and natural or syn- Unlike food additives, environmental con- thetic radioactive substances. Such contam- taminants inadvertently find their way into inants are regulated under the Federal Food, the human food supply (including sports fish Drug, and Cosmetic Act. To regulate them un- and game). They can enter food directly or in- der the law, the Food and Drug Administra- directly as a result of such human activities tion (FDA) defines environmental contami- as agriculture, mining, industrial operations, nants as “added, poisonous, or deleterious” or energy production. In no instance is their substances that cannot be avoided by good presence in food ever intended.

3 Four factors determine whether and how tigue, abdominal pains, and liver and men- seriously the environmental contamination of strual disturbances, food will affect human health: the toxicity of the contaminant, the amount of the substance No such mass-poisoning episodes have oc- in the food, the amount of the contaminated curred in the United States. But there are food eaten and the physiological vulnerabili- studies indicating that present levels of some ty of the individual or individuals consuming environmental contaminants may cause phys- the food. iological changes. For example, the acciden- Based on other countries’ experiences, tal contamination of animal feed in 1973 ex- there is considerable evidence of human ill- posed most of the population of Michigan to ness caused by the consumption of food con- PBB in dairy products and other foods. Evi- taining various organic chemicals and dence on what impact this exposure has had on human health is conflicting, although some metals. In such cases, the level of the con- disparities in white blood cell function have taminant in food exceeded the levels usually been noted in farm families, The long-term found in the U.S. food supply. The effects of significance of these physiological changes is mercury poisoning are well-documented. The not yet known. best known case involved the consumption of mercury-contaminated fish from Japan’s Min- The clinically obvious harmful health ef- amata Bay. Some of the offspring of exposed fects of radiation are usually associated with mothers were born with birth defects, and massive, high-level exposures. Past cases of many victims suffered central nervous sys- radioactive contaminated foods have in- tem damage. volved relatively small amounts of radioac- Another incident in Japan stemmed from tive substances with low dose rates. General- the inadvertent contamination of rice oil by ly, the young are most sensitive to radiation PCBs. The consumption of food cooked with exposure. However, since any amount of radi- this oil resulted in 1,291 cases of so-called ation is potentially harmful, prudent public “Yusho disease”—a condition marked by policy must assume that any unnecessary ex- chloracne (a severe form of acne), eye dis- posure to high-energy radiation should be charges, skin discoloration, headaches, fa- avoided.

The economic impacts of a contamination medical expenses and lost workdays from ill- incident have traditionally been stated in ness resulting from food contamination in- terms of the estimated dollar value of the re- cidents. Since the health effects may not im- sulting food loss. Only limited data on such mediately be evident, the expected illnesses costs are available. Dollar value estimates for or deaths from an episode are usually esti- condemned food were available for less than mated on the basis of available toxicity data 30 percent of the contamination episodes for a particular contaminant. In other words, noted in the OTA survey. Thus, the real cost estimated health costs are more likely to be of environmental contamination of food dur- projections than actual figures based on ing the 1968-78 decade is at least several known cases of illness or death. times the $282 million reported to OTA. The “distributional” costs of environmen- The loss of food only partially reflects the tal contamination disclose the expenses or total economic impact of environmental con- losses incurred by affected businesses, indi- tamination, Health and “distributional” costs viduals, and government bodies. These might are also involved. The health costs include include farmers, fishermen, food processors, Ch. l—Summary ● 5 animal feed suppliers, chemical companies, fied, their actual dollar losses are not known. consumers, and local, State, and Federal To understand who is bearing the major eco- agencies. Although the individual organiza- nomic brunt of a contamination episode, ac- tions suffering such losses are usually identi- tual cost data are required.

MAJOR PROBLEMS IN IDENTIFYING ENVIRONMENTAL CONTAMINANTS

To determine whether an environmental substances are not released into the environ- contamination incident has occurred, it is ment. There are various Federal environmen- necessary to establish the presence of the tal laws that are designed to limit such re- contaminant in food. In some instances, peo- leases. But the laws and regulations are not ple or animals have become ill before the re- likely to prevent the deliberate or accidental sponsible contaminant was identified. No one misuse or disposal of the thousands of toxic knew or even suspected that the particular substances manufactured in the United substance was present in food. This has been States. the pattern in many major contamination in- The problem is compounded by disposal cidents—those involving PBBs, PCBs, and and handling practices that was accepted in mercury. the past but are now recognized as posing Our regulatory monitoring system has serious environmental hazards—hazards failed to detect such environmental contami- that will persist for many years to come. The toxic chemical waste dump at the Love Canal nants as they entered the food supply. Thus, near Niagara Falls, N. Y., clearly illustrates this assessment identifies and evaluates other approaches for monitoring either food the threat. According to Environmental Pro- or the environment for toxic substances that tection Agency (EPA) estimates, there are 1,200 to 2,OOO of these abandoned chemical may harm human health. The ultimate objec- and radioactive waste sites in the United tive of monitoring is to prevent or minimize States that pose an imminent danger to hu- human exposure to environmental contami- man health and will cost as much as $50 bil- nants in food. lion to clean up. As long as these substances The only sure way to prevent this kind of remain in the environment, the potential for contamination is to make certain that toxic food contamination exists.

Once an environmental contaminant is on its toxicity, the amount present in the food, found in food, limits are established to control and how much and what kinds of food are and restrict its presence. Such regulations contaminated. By monitoring the food supply are set up and enforced by FDA for food for the contaminant, regulators can deter- traded in interstate commerce, and by State mine the level and extent of the contamina- agencies for food produced and sold within a tion. With this information and necessary State. In either case, the aim is to limit the toxicity data, they can establish regulatory public’s exposure to a particular contami- limits for the contaminant in food. nant. However, this kind of information general- Key factors involved in such regulation are ly takes time to generate—usually longer time and information. After a contaminant is than the public is willing to wait in the event identified. authorities must have information of a food contamination incident. As a result, authorities are often pressed to set regula- develop information on the nature and extent tory limits before they have enough time to of the contamination.

This assessment has focused on two cen- fluence. Because the amount and quality of tral problems: regulating environmental con- information available when FDA encounters taminants and identifying environmental con- an environmental food contamination prob- taminants. Following are major findings and lem are inevitably unpredictable, it does not conclusions growing out of this assessment. predetermine the weighting of various fac- tors. However, FDA maintains that the public ● FDA relies on action levels rather than tol- health factor outweighs all others in its con- erances to regulate environmental con- taminants in food. siderations. Under the Food, Drug, and Cosmetic Act, The Food, Drug, and Cosmetic Act does not FDA is given authority to set tolerances for specify the role that the costs of a regulatory the amount of an unavoidable contaminant decision should play in setting a tolerance or permissible in food. However, the procedures action level. The Act does require that FDA required to set a tolerance are complex, cum- take into account the extent to which a sub- bersome, and time-consuming. Therefore, stance cannot be avoided in food production. FDA relies on action levels, informal judg- FDA interprets this requirement as justifica- ments about the level of a food contaminant to tion for weighing the costs of food condemned which consumers may safely be exposed. against the health benefits derived from a tol- Action levels are administrative guidelines erance. that can be developed and promulgated more easily and quickly than tolerances. Action ● To assess human risk from exposure to levels are used when scientific data are in- chemicals, FDA and EPA rely on already- complete. Public input is not required. They existing animal studies and epidemiologi- are used when new information is likely to be cal evidence derived from previous hu- forthcoming that might alter the level. FDA is man exposures, under no constraints to review action levels When a new environmental contaminant is or to replace action levels with formal toler- discovered in food, regulatory agencies are ances. FDA has sometimes lowered or raised under intense pressure to act to protect the action levels as new data became available. public. FDA and EPA (if the contaminant is a FDA is now in the process of lowering the ) review the available literature on PCB tolerance. the contaminant and calculate an action level ● No policy exists defining the relative based on that evidence. Rarely are new weights to be given to the evidence when studies commissioned—even when the data setting an action level or tolerance. are inadequate. In setting an action level or tolerance, FDA New human epidemiological studies and takes into account short- and long-term tox- conventional 2-year animal studies are of lit- icological data, available information on the tle immediate help because so much time is levels of the contaminant in food, the amount required to generate results. However, toxi- of contaminated food consumed by various cologists have developed a variety of tests population groups, the level that can be meas- that can evaluate a substance’s possible toxic ured, and the potential impact of various ac- effects in 90 days or less. Some of these short- tion levels or tolerances on the national food term tests measure the potential of a sub- supply. Generally, the more information stance to produce mutations and possible about a particular factor, the greater its in- , Ch. I—Summary ● 7

Short-term tests could be used more widely employed in Federal or State regulatory moni- in screening environmental contaminants to toring. determine whether they are mutagens or po- The goal of food monitoring is to protect tential . Although the results of such tests do not provide the data needed to consumers by determining short- and long- set an action level or tolerance, they still can term trends in the levels of various chemicals alert regulators to latent dangers that re- in food and the environment. Investigatory quire further investigation. monitoring could be designed to complement already-existing regulatory monitoring. Each Conventional 2-year animal studies (which of these approaches could be complemented usually entail an additional year for data by specimen banking-the regular collection analysis) would continue to serve an impor- and storage of samples that could be later tant role in the setting of tolerances. If data analyzed if a new contaminant is found in from a carcinogen bioassay are available at food. EPA and the National Bureau of Stand- the time an environmental contaminant is dis- ards are now working towards developing covered in food, the information can prove such a specimen-banking program. crucial in reaching a regulatory decision. If However, food sampling may not be the data were nonexistent or inadequate, a newly best approach to investigatory monitoring. To commissioned carcinogen bioassay could be discover a substance as it enters the environ- used to revise an initial action level. ment and before it gets into the human food Epidemiological studies would remain use- supply, it is necessary to monitor water, soil, ful for confirming suspected chronic effects air, river sediments, and nonfood organisms. of a toxic substance to which a population has unknowingly been exposed over a period ● Management of food contamination inci- of time. They can also confirm retrospectively dents is hindered by the complexity of the or refute the adequacy of regulatory actions. food system, the rapidity with which food is moved through the system, and failures No currently available toxicological testing by State and Federal agencies to coordi- methods or statistical interpretation tech- nate their information-gathering activ- niques are adequate for evaluating the com- ities. bined effects of low-level exposure to toxic Many food contamination incidents initial- substances. Indeed, there are no satisfactory ly fall under State jurisdiction. Technically, techniques for testing the interactions of more than two substances. the Federal Government does not become in- volved unless requested by a State or until contaminated food enters interstate com- ● Current monitoring at both Federal and merce. This country’s system State levels is regulatory, designed to en- is complex. Most food produced or processed sure that substances in food do not exceed within a particular State is distributed for prescribed limits. Little effort is made to consumption in other States. Thus, most envi- detect and identify substances in the food ronmental contamination incidents are likely supply for which no action levels or toler- to become interstate problems. Figure 1 illus- ances exist. trates the extent of food contamination that can occur from a single source of contamina- Technology now exists that would make tion, in this case PCB-contaminated animal possible a national investigatory monitoring feed from a meatpacking plant in Billings, system to detect unregulated chemicals as Mont. they enter the food chain. Such advanced technology is available in some Federal regu- The number of State and Federal agencies latory monitoring laboratories and in a lim- involved complicates the generation and dis- ited number of State labs. It is not routinely semination of scientific information on the

51+-515 ~ - 79 _ 2 8 • Environmental Contaminants in Food

II l II ' II

c:: ~ 't! ~c E '0 ~ 01 2 a '0 c t1I '---~ g u. W <.,) a:: o:::l (/) Ch. /—Summary ● 9

toxicological and chemical properties of the tential exists for breakdowns in communica- contaminant, the amount and type of food tion. This was the case in the recent PCB con- contaminated, and the concentration of the tamination of animal feeds in 10 Western substance in food. At least three Federal States. The Idaho Department of Agriculture agencies (EPA, FDA, and the U.S. Department did not inform the Idaho Department of of Agriculture (USDA)), each with different Health and Welfare of the PCB contamina- responsibilities, may provide technical assist- tion. USDA would report the results of its in- ance. At the State level, departments of vestigations only to the Idaho Department of health, agriculture, and the environment may Agriculture. EPA attempted to determine the share accountability for regulating environ- source of the PCBs by analyzing air and wa- mental contaminants in food. ter samples, but failed to report its negative results to the State. In the absence of a clear authority to coor- dinate activities of various agencies, the po-

CONGRESSIONAL OPTIONS

There are four basic options for Congress Cons: The time needed to identify an envi- to consider regarding the Federal response to ronmental contaminant in food and- take cor- the environmental contamination of food. rective action would not be shortened if the Each is discussed in greater detail in chapter current system were retained. Moreover, ac- IX. Congress can: tion levels and tolerances permit a certain Allow the present system to continue by level of contaminant to be present in food. If 1. tolerances or action levels are not reduced, taking no action. The present system little effort will be made to eliminate the con- consists of regulatory monitoring and the establishment of action levels (and taminant. There is no requirement for review occasionally tolerances) for environmen- of an action level once it is established. Thus, tal contaminants in food. FDA is under no pressure to actively seek out new data to verify the appropriateness of an 2. Amend the Food, Drug, and Cosmetic Act to specifically address the unique existing action level. Finally, States have no problems posed by environmental con- clearly defined authority to turn to when they tamination of food. suspect environmental contamination of food. 3. Establish a national investigatory moni- toring system. Option 2 4. Improve the Federal response to new Amend the Food, Drug, contamination problems by designating and Cosmetic Act a lead agency or establishing a center to orchestrate the delivery of Federal as- An amendment to the Food, Drug, and Cos- sistance to affected States. metic Act could contain one or more of the following changes. Each change is discussed in greater detail in chapter IX. Option 1 Maintain the present System ● Congress could amend the Food, Drug, and Cosmetic Act to simplify the admin- Pros: There are two principal advantages istrative procedure for setting toler- in maintaining this system. No additional ap- ances. The change could be modeled propriations or legislation are required. No after section 553 of the Administrative changes in existing regulations are neces- Procedures Act. This would encourage sary. FDA to move from action levels to toler- ances, thus bringing more public partici- humans if consumed in food. Uncharacter- pation into the process, ized environmental contaminants are sub- ● Congress could amend the Food, Drug, stances that may have entered the food sup and Cosmetic Act to require the estab- ply, but are not regulated or suspected food lishment of a tolerance within a speci- contaminants. A system that combines ele- fied time after the setting of an action ments of both approaches could also be set level. This would encourage the FDA to up. Because any of these monitoring ap- gather additional information on a con- proaches would require some research and taminant’s toxicity and the public’s ex- development before going into full operation, posure. It would result in a definitive tol- Congress could choose to establish a pilot pro- erance that FDA could enforce with less gram. Such a program would spur research concern over legal challenge. and development and assess the feasibility ● Congress could clarify to what extent and cost effectiveness of the various ap- economic criteria can be used in setting proaches. tolerances for environmental contami- nants in food, Pros: Investigatory monitoring would in- crease the probability of detecting unregu- ● FDA could be granted authority to set re- gional tolerances. This would provide lated substances in food. Present food-moni- FDA with flexibility to set different toring efforts are not designed to detect levels for different regions based on ex- unregulated environmental contaminants in pected levels of exposure, regional levels food. The limited amount of investigatory of contamination, and local eating pat- monitoring that does exist is primarily con- terns. cerned with trace metals. To identify new contaminants as they enter the food supply, Pros: Since its passage in 1938, the Food, more of this type of monitoring is needed. Drug, and Cosmetic Act has been amended several times to deal with new problems of Cons: The costs of setting up an investiga- food regulation. Congress has never directly tory monitoring program could be large, and addressed the environmental contamination there is no certainty that the sampling plan of food. There are several unique character- would identify all environmental contami- istics of this problem that could be clarified nants before they enter the food chain. Fur- through an amendment dealing specifically thermore, investigatory monitoring relies on with the environmental contamination of sophisticated instrumentation that is gener- food. ally not found in Federal or State monitoring laboratories, Cons: Even though the Food, Drug, and Cos- metic Act does not contain provisions on envi- Option 4 ronmental contaminants, FDA has been able improve Federai Response to to regulate them through interpretation of sections 402 and 406. New Contamination Incidents The Federal response to new contamina- Option 3 tion problems has been hampered by the mul- Establish an Investigatory tiplicity of agencies with regulatory or mon- Monitoring System itoring responsibilities for the environment and for food. Congress could designate a lead Congress could establish a national investi- agency or establish a center to orchestrate gatory monitoring system based on monitor- delivery of Federal assistance to affected ing for either suspected or uncharacterized States. environmental contaminants. Some chemi- cals are not regulated by action levels or tol- Pros: With a clearly delineated agency or erances but are suspected to be dangerous to center, States suspecting contamination of Ch. /—Summary ● 11 food would have one reliable Federal source ter would not ensure that information would for generating, evaluating, and disseminating be generated more quickly than is now the technical information. Response time might case. be shortened, duplication of effort reduced, * * * and effective management of the incident en- Options 2 through 4 are not mutually ex- hanced. clusive. If Congress wishes to put greater em- Cons: Better coordination among FDA, phasis on protecting consumers from contam- USDA, and EPA could accomplish the same inated food, one or more could be chosen. For goals without the expense of establishing a example, Congress could decide to simplify new research center. Historically, the major the administrative procedures for setting impediment to timely Federal response to tolerances (Option 2), require the setting of a chemical contamination of food was lack of tolerance at some specified time after an ac- awareness that food contamination had taken tion level is set (Option 2), establish a pilot place. When contamination became apparent program of investigatory monitoring for or- and one or more Federal agencies were ganic chemicals (Option 3), and designate alerted, response was rapid. Furthermore, FDA the lead agency to deal with new con- establishment of a lead agency or a new cen- tamination problems (Option 4). Chapter II

Environmental Contamination of Food Chapter II

Environmental Contamination of Food

Maintaining an adequate, safe food supply has been a major goal of the Fed- eral Government since 1906, when the first Federal food and drug law was signed into law. Historically, chemicals such as , , and wood smoke have been used to preserve foods. Modern relies extensively on the use of chemicals not only for preservation but also to produce appealing colors, , aromas, and textures. Most developed countries now have food laws designed to permit the use of such chemicals in food under conditions judged to be safe. These chemicals are not considered or contaminants and are classed as intentional addi- tives. Other chemicals may enter food as a result of their use in food production, handling, or processing. Such substances maybe legally permitted if they are un- avoidable under good manufacturing practices and if the amounts involved are considered safe. These chemicals are classed as incidental additives. The pres- ence of both these classes of chemicals in food is controlled by regulation.

Environmental contaminants include sub- chemicals (l). During the production, use, and stances from natural sources or from indus- disposal of these substances, there are oppor- try and agriculture. Many of the naturally oc- tunities for losses into the environment. For curring contaminants in food are of microbio- example, the Environmental Protection Agen- logical origin and consist of harmful bacteria, cy (EPA) estimates that there are more than bacterial toxins, and fungal toxins. (Aflatox- 30,000 chemical and radioactive waste dis- in, a contaminant of peanuts and grains, is an posal sites. Of these, 1,200 to 2,000 are con- example of a fungal or . ) The sidered threats to human health (2). second category of environmental contami- Environmental contamination of food takes nants includes organic chemicals, metals and two forms: long-term, low-level contamination their complexes, and radionuclides. Only resulting from gradual diffusion of persistent those environmental contaminants intro- chemicals through the environment, and rela- duced into food as a result of human activities tively shorter term, higher level contamina- such as agriculture, mining, and industry are tion stemming from industrial accidents and considered in this assessment. waste disposal. The environmental contamination of food is An example of low-level contamination is a result of our modern, high-technology soci- polychlorinated biphenyls (PCBs). This group ety. We produce and consume large volumes of substances was widely used in transform- of a wide variety of substances, some of ers and capacitors, as heat-transfer fluids, which are toxic. It is estimated that 70,000 and as an additive in dyes, carbon paper, pes- chemicals may currently be in commercial ticides, and plastics (3). Although production production in the United States and that 50 of was halted in 1977, PCBs remain an ubiqui- these chemicals are manufactured in quanti- tous, low-level contaminant of many foods, es- ties greater than 1.3 billion lbs per year. pecially freshwater fish. Seven percent of this country’s gross national product (GNP), $113 billion per year, is gener- An example of the second type of contami- ated by the manufacture and distribution of nation is polybrominated biphenyls (PBBs) in

15 16 . Environmental Contaminants in Food dairy products and meat. PBBs, a fire retard- level contaminant in Michigan because they ant, were accidentally mixed into animal are very stable and resistant to decay. Ani- feed. Dairy cattle that were fed the contami- mals raised on farms affected by the original nated feed produced contaminated milk. The feed contamination are now contaminated by distinctions between the two types of food the PBB residues remaining in the pastures contamination are not exclusive. For exam- and farm buildings. pie, PBBs have now become a long-term, low-

HOW FOOD BECOMES CONTAMINATED

Chemicals contaminate foods through dif- gated railroad cars, trucks, ships, or storage ferent routes depending on the chemical and buildings used for transport or storage of its physical properties, its use, and the source human food and animal feed are also sources or mechanism of contamination. of environmental contamination. The interi- ors are sprayed or fumigated with , Organic substances that have contami- iand if not sufficiently aired, contamination of nated food have been either industrial or 1 the food or feed occurs. agricultural chemicals. Pesticides are the only agricultural chemicals known to be en- The manufacture of organic chemicals pro- vironmental contaminants in food (see tables duces sludges, gases, and liquid effluents of 1-3). A pesticide becomes an environmental varying chemical complexities. The usual contaminant when it is present in foods for ,waste disposal methods (sewage systems, in- which the application or use of the substance cineration, landfill) are unable to prevent or- has not been approved. Livestock, poultry, ganic residues from entering the environment and fish can be contaminated when applica- in spite of Federal laws and corresponding tion or manufacturing of pesticides occurs in regulations governing disposal. The routes in- the vicinity or when residues are transported clude the atmosphere, soil, and surface or through the environment. Improperly fumi- ground water.

Table 1.— Reported Incidents of Food Contamination, 1968-78, by State and Class of Contaminant

State Pesticide Mercury PCB PBB Other Total New York ...... • ., . 1 1 1 — — 3 Idaho...... 1 2 — 3 South Carolina...... — 1-biphenyl 1 Minnesota ...... — 1 — 1 Louisiana ...... 8 5 4 — 17 Colorado...... 1 1 — 2 Georgia...... 15 — — — 15 Maryland...... — 1 — 2-petroleum 3 Texas ...... — 1 — 1 New Jersey...... 2 1 — 1-ß-methoxy napthalene 4 and tetraline Kansas . . . — (I)* — — (l)* Missouri ...... 4 — — — — 4 New Mexico. . . — 1 — — — Alaska...... (l)* — — — — (:)* California . 1 3 — — — 4 Indiana ., 3 well-documented — 1 — — 9 5 other incidents Michigan. 13 1 2 1 — 17 Virginia, ., ... 1 — — — — 1 Totals ...... 56 19 8 1 4 88 “Several conservatively estimated as one SOURCE Of fLce of Technology Assessment Ch. //— Environmental Contamination of Food ● 17

Table 2.— Reported Incidents of Food Contamination, 1968-78, by State and Food

Fruit/ Game/meat/ State Dairy Eggs Vegetable Fish/shellfish Grain poultry 3 — — 1 — 1 — 1 1 — 9 1 — — — — 3 — 1 — — 2 — — — 1 — — — — 1 2 — 1 — — 1 4 — 4 — — 1 — 27 3 8

Table 3.— Number of Incidents of Environmental Contaminants of Food Reported by Federal Agencies, 1968-78

Agency Pesticides Mercury PCB’s Other Food affected USDA/FSQS 39 — — — Chickens, turkeys. ducks. cattle, swine, lambs — 1 — — Swine — — 6 — Poultry — — — 1 (Phenol) Cattle FDA ... 21 — — — Fish. cheese, pasta — 84 — — Fish — — 3 — Fish, eggs, bakery products Total. . . . 60 85 9 1

SOURCE Off Ice of Technology Assessment

Metals can be released into the environ- operation of nuclear reactors and processing ment in several ways. The mining and refin- plants, and fallout from nuclear weapons ing processes produce dust and gases which tests. The primary route by which food be- enter the atmosphere. Metallic formed comes contaminated is the deposition of air- during recovery and refining processes can borne material on vegetation or soil. The sub- escape as waste products into surface and sequent fate of the radionuclide is deter- ground water. Sewage sludge used as fertiliz- mined by its chemical and physical nature er on agricultural land also poses a potential and whether it is absorbed and metabolized food contamination problem. Trace metals by plants or animals. Natural radioactivity present in the sludge can be taken up by may become a concern when ores containing crops grown on treated soil. is the radioactive substances are mined and proc- trace metal in sludge that currently gener- essed. The products or wastes may concen- ates the greatest concern. trate the radionuclides. Examples of this are uranium tailings, phosphate rock waste, or Radioactivity in food stems from three slags from phosphorus production. Radium sources: natural radioactivity. releases from may enter the food chain when it dissolves in 18 . Environmental Contaminants in Food ground water and is taken up through plant summarizes the radionuclide contaminants of roots. significance for foods. Nuclear reactors normally release radio- Nuclear waste-processing plants could active noble gases that do not contaminate also have either gaseous or aqueous releases. foods. Reactors do contain large inventories In this case, the fission products are aged of fission products, transuranics, and other before processing, and iodine and the gas- activation products. Accidental releases can eous precursor radionuclides are not re- contaminate vegetation by deposition of parti- leased. Tritium and carbon-14 are the major cles on leaves and soil, or through water. Gas- airborne products, while the waterborne ra- eous releases would most likely involve the dionuclides are the same as for reactors. volatile elements such as iodine and tritium, or those with volatile precursors, such as Atmospheric nuclear weapons tests dis- strontium-90 and cesium-137. Aqueous re- tribute their fission products globally. Local leases would follow failure of the onsite ion deposition depends on the size of the weapon exchange cleanup system. Any of the water- and the conditions of firing (high altitude, sur- soluble elements could be involved. Table 4 face, or underground).

Table 4.— Radionuclide Contaminants of Significance for Foods

Normally present in small amounts, Significant only when enhanced Normally present in small amounts Member of uranium series. Normally present, metabolized somewhat Iike calcium As radium-226, only a member of thorium series Members of uranium series Normally present

Product of nuclear reactions

Product of 24’ Pu decay

Low energy, usually in form of water or organic compounds Low energy, usually in form of organic compounds

Short half-life, so important only for fresh foods, e.g., milk and leafy vegetables Follows calcium somewhat in Follows potassium somewhat in metabolism

Most important are isotopes of zirconium, cerium, barium, rubidium, rhodium. Mostly surface contaminants

Follow stable elements in metabolism Ch. //— Environmental Contain/nation of Food ● 19

MAGNITUDE OF THE PROBLEM

There is little information available on the edge and the extent of uncertainties on the number of food contamination incidents, the health effects of these environmental con- amount and costs of food lost through regula- taminants. tory actions, or the effects of consumption of contaminated food on health. To obtain infor- Polychlorinated Biphenyls mation on the extent of the problem, OTA re- viewed the literature and sought information PCBs occur in food as the result of environ- from the States and Federal agencies. mental contamination leading to accumula- tion in the food chain, direct contact with Evidence of Human Illness food or animal feeds, or contact with food- packaging materials made from recycled pa- Resulting From Consumption of per containing PCBs (5). Several comprehen- Contaminated Food sive literature reviews have been published In evaluating the significance of environ- in the last 5 years detailing the acute and mental contaminants in food the key question chronic toxic effects of PCBs in animals and is whether consumption of contaminated humans (5-1 1). foods poses a health risk. Measurable health Human illness has been caused by expo- effects depend on the toxicity of the sub- sures to PCBs at much higher levels than stance, the level at which it is present in food, those that occur in the United States. In the the quantity of food consumed, and the vul- early part of 1968 the accidental contamina- nerability of the individual or population. In tion of edible rice-bran oil led to a poisoning Japan, foods contaminated with substances epidemic among the Japanese families who such as PCBs, mercury, and cadmium have consumed the oil. The disease became known produced human illness and death. No such as Yusho or rice-oil disease. Its chief symp- mass poisonings have occurred in the United toms were chloracne (a severe form of acne) States. However, in cases such as PBBs and eye discharge; other symptoms included where a large populace has been exposed, skin discoloration, headaches, fatigue, ab- some physiological changes have been noted. dominal pain, menstrual changes, and liver But no conclusions can as yet be drawn on the disturbances. Babies born to mothers who ultimate health effects. consumed the rice oil were abnormally small It is known from limited surveys that the and had temporary skin discoloration. The U.S. population is exposed to a wide variety first symptoms of Yusho disease were regis- of chemical contaminants through food, air, tered on June 7, 1968, and 1,291 cases had and water. The long-term health effects and been reported as of May 1975 (9). the implications of possible interactions Since the rice oil was also contaminated among these residues are unknown. A recent with polychlorinated dibenzofuran (PCDF), it literature review of over 600 published is difficult to determine from the Yusho data studies (4) found that nonoccupationally ex- exactly what effect(s) exposure to PCBs alone posed U.S. residents carry measurable resi- could have on humans. It has been calculated dues of 94 chemical contaminants. Twenty- that the PCDF made the rice oil 2 to 3.5 times six of these are organic substances, including more toxic than would have been expected twenty pesticides and pesticide metabolizes. from its PCB content alone (1 1). Careful rec- The remainder are inorganic substances. ords of the 1,291 Yusho patients have been Americans also have been exposed to low kept to determine possible long-term effects. levels of PCBs, PBBs, mercury, and ionizing At least 9 of 29 deaths that occurred as of radiation through their food. The following May 1975 were attributed to cancer (malig- sections briefly summarize current knowl- nant neoplasm), but a causal relationship be- 20 ● Environmental Contaminants in Food tween PCBs and cancer cannot necessarily peninsula excreted PBB in their breast milk be inferred because of the high concentration (20). of PCDF in the oil. The Yusho study, neverthe- less, had two important results: first, the in- Low concentrations of PBBs also have been formation established that PCBs can be trans- detected in animal feed in Indiana and Illi- ferred from mother to fetus and from mother nois. Unconfirmed surveys of food throughout to child through breast feeding, and second the country found extremely low levels below highly chlorinated PCB compounds are ex- the Food and Drug Administration (FDA) ac- creted more slowly from the body than less tion level in the following States (21): chlorinated ones (9). State Food Alabama. ., ...... Chicken More recent experiments in animals have Indiana...... Turkey demonstrated a variety of toxic effects. Iowa , ... , ...... Beef have been produced in mice and rats Mississippi...... Chicken fed PCBs (6, 12). Monkeys fed levels of PCBs New York ...... , . . . . Chicken equivalent to the amounts consumed by Yu- Texas ...... Chicken sho patients developed similar reproductive Wisconsin ...... Duck disorders (13-16). Young monkeys nursing on Wolff, et al. (22) reported that serum PBB mothers consuming feed containing PCB de- was higher for males than females. It was veloped toxic effects and behavioral abnor- suggested that the greater proportional body malities (15-1 7)0 in women may account for this difference, but exposure may also be important. Males Polybrominated Biphenyls may consume more contaminated food or have more direct contact with PBB than Practically every Michigan resident has females. been exposed to PBB-contaminated food prod- ucts. It is estimated that some 2,000 farm The same study found no consistent trends families who consumed products from their with respect to age. It was observed, how- own PBB-contaminated farms have received ever, that young males had greater concen- the heaviest exposure (18). trations of serum PBB than young females. Young females had greater concentrations Fries (19) studied the kinetics of PBB ab- than older males, and older males had sorption in dairy cattle and its elimination in greater concentrations than older females. It milk, If intake of contaminated milk alone is was also found that very young children and considered, those Michigan residents most individuals who had lived on farms less than severely exposed consumed from 5 to 15 1 year had lower serum PBB levels than other grams of PBB over the initial 230 days of the groups (22) exposure. Those residents that coincidentally 0 consumed contaminated meat and/or eggs Serum PBB concentration is related to the may have received higher total doses of PBB, intensity of exposure. Most studies indicate but the number of such cases is probably that consumers and residents of nonquaran- small, tined farms had significantly lower PBB lev- els than residents of quarantined farms; how- Geographically the residents of the lower ever, families on quarantined farms stopped peninsula, where the original accident oc- consuming meat and milk from their own ani- curred, were found to have the greatest levels mals (20) of exposure. In 1976, the Michigan Depart- 0 ment of Public Health conducted a study on In late 1974, the Michigan Department of PBB concentrations in breast milk. It was Public Health conducted a survey to deter- found that 96 percent of the 53 women mine if any adverse effects could be corre- selected from the lower peninsula and 43 per- lated with PBB levels in the body, A sample of cent of the 42 women selected from the upper 165 exposed persons (quarantined farms) Ch. Il—Environmental Contamination of Food ● 21

and 133 nonexposed (nonquarantined farms) Mercury and Methylmercury was studied. Medical history interviews and Foods are the major source of human expo- physical examinations were performed on sure to mercury. The mercury concentration each subject and blood specimens were in food is dependent on the type of food, the taken, Blood PBB levels as high as 2.26 parts environmental level of mercury in the area per million (ppm) were found in the exposed where the food is produced, and the use of individuals; about half exhibited levels mercury-containing compounds in the agri- greater than 0.02 ppm. Of the nonexposed in- cultural and industrial production of the food. dividuals, only two showed blood PBB levels All living organisms have the ability to con- greater than 0.02 ppm; 70 percent of the centrate mercury, Therefore, all animal and adults and 97 percent of the children exhib- vegetable tissues contain at least trace ited levels of 0.0002 to 0.019 ppm. Compar- amounts (26). Several recent reviews have ex- ison of a list of selected conditions and com- amined the health effects associated with plaints revealed no significant differences in consumption of mercury (26-28). The results the frequency of illness between the two of these reviews indicate that the effects of groups. Physical examinations and clinical methylmercury poisoning become detectable laboratory tests disclosed no effects attrib- in the most sensitive adults at blood levels of (24). utable to “chronic” PBB exposure mercury of 20 to 50 µg/100 ml, hair levels The effect of PBB exposure on white blood from 50 to 120 mg/kg, and body burdens be- cell (lymphocyte) function of Michigan dairy tween 0.5 and 0.8 mg/kg body weight (26). farmers who consumed contaminated farm products was examined by Bekesi, et al. (25). Since the Minamata Bay tragedy in Japan, Forty-five members of Michigan farm fami- the effects of chronic exposure to methylmer- lies who had eaten PBB-contaminated food cury have been well-documented. Mercury for periods of 3 months up to 4 years after the readily accumulates within the central nerv- original accident were compared for immuno- ous system (29-3 1), and clearance of mercury logical function to 46 Wisconsin farmers and back into the bloodstream is slow (32). Conse- 79 New York residents. All of the exposed in- quently, the central nervous system is consi- dividuals showed reduced lymphocyte func- dered to be the critical target in chronic mer- tion, and 40 percent showed abnormal pro- cury exposure. The clinical symptoms of cen- duction of lymphocytes. There were also sig- tral nervous system involvement include nificant increases in lymphocytes with no de- headache, vertigo, vasomotor disturbance, tectable surface markers (“null” cells). How- ataxia, and pain and numbness in the ex- ever, the short- and long-term health implica- tremities (30). The most prominent structural tions of these differences are not now known. changes of the central nervous system result- ing from chronic mercury exposure are dif- Lillis (20) examined Michigan farmers and fuse cellular degeneration (30). consumers of dairy products and found that the effect of PBB on humans was mainly neu- In evaluating the teratogenic hazards of rological in nature. He found marked fatigue, mercury exposure to man, the placental hypersomnia, and decreased capacity for transfer of mercury is particularly signifi- physical or mental work. Other symptoms in- cant. Levels that are not toxic to pregnant cluded headache: dizziness: irritability; and women are sufficient to produce birth defects in their offspring (33-35). Transfer of methyl- swelling of the joints with deformity, pain, mercury across the human placenta results and limitation of movement. Less severe gas- in slightly higher blood levels in the infant at trointestinal and dermatological complaints birth than in the mother (36). Table 5 com- were also encountered. pares fetal and maternal blood concentra- 22 ● Environmental Contaminants in Food

Table 5. — Methylmercury Concentrations in tion exposure on health are extremely diffi- Normally Exposed Populations — cult to evaluate. High dose rates (100 million Concentration (µg Hg/g) to 1 billion times background) are estimated Location Maternal blood Placenta Fetal blood to produce 2,600 ionization events per second Japan ...- – ‘ - 0.017 0.072 0.020 in cells. Background radiation levels are esti- Sweden ...... 0.006 — 0.008 mated to produce less than one ionization in Tennessee. . . 0.009 0.021 0.011 lowa. . . . . 0.001 0.002 0.001 the cell nucleus per day (37). Because cells SOURCE Adapted from B J Koos and L D Longo ” Mercury To; IcIty In the have the capacity to repair damage to their Pregnant W’oman, Fetus, and Newborn Infant “ A rner(can ,/ourrra/ of Obstetrms and Gynecology 126(3) 390, 1976 genetic material, repair of ionization damage may occur at low radiation exposure. Higher tions in normally exposed populations in exposures may overwhelm the cells’ repair Japan, Sweden, and the United States, capacity. Whether any effects are observed In humans, the most widely reported fetal in such cases depends on several factors. These include the dose delivered to the tis- risk associated with maternal exposure to sues, the nature of the emissions, and the me- mercury is brain damage. The placental tabolism of the cell. The following examples transfer of mercury and its effects on the illustrate these points: human fetus were first recognized in the 1950’s with the well-known outbreak of con- Strontium-90 in food arouses most con- genital in the towns of cern not only because of its long half-life Minamata and Niigata, Japan. By 1959, 23 in- but also because it behaves in the body fants suffering from mental retardation and in a manner somewhat similar to calci- motor disturbances had been born to mothers um. The replacement of bone calcium exposed to methylmercury during their preg- with strontium-go exposes tissues and nancies. The clinical symptoms of the infants cells covering the bone to radiation, In resembled those of severe cerebral palsy or addition, bone marrow is subject to the cerebral dysfunction syndrome. They in- ionizing radiation from the strontium-go. cluded disturbance of coordination, speech, Thus, cancer of the bone-forming and and hearing; constriction of visual field; im- bone-covering tissue as well as leuke- pairment of chewing and swallowing; en- mias of the bone marrow blood-forming hanced tendon reflex; pathological reflexes; cells can possibly result. involuntary movement; primitive reflexes; su- Iodine is concentrated by the thyroid perficial sensation; salivation; and forced gland. Radioiodines produced in atmos- laughing (30). Only 1 of the 23 mothers exhib- pheric nuclear detonations or released ited any symptoms of mercury poisoning (32). from nuclear power stations are also taken up and concentrated by the thy- Radioactivity roid, increasing the risk of thyroid cancer. Ionizing radiation (X-rays, gamma rays, or Tritium, or radioactive hydrogen, com- beta particles with sufficient energy to strip bines chemically with oxygen to form electrons from molecules and produce ions) water. Tritium derived from food would can produce birth defects, mutations, and be widely distributed throughout the cancers (37). These adverse health effects body exposing all tissues to radiation. are usually associated with high dose levels delivered at high dose rates. The uncertainties surrounding the repair Such a combination is not ordinarily en- capacities of cells and the irreversible nature of the possible health effects have led to the countered in food. Previous radioactive con- tamination of foods has involved relatively adoption in the United States of a prudent policy toward low-level ionizing radiation. small quantities of radioactive elements which have delivered low dose rates (38). Since any amount of radiation is potentially harmful, unnecessary exposure should be In these situations, the effects of the radia- avoided. Ch. //—Environmental Contamination 01 Food ● 23

Number of Food Contamination tal contaminant incident since 1968 for a Incidents total of 88 incidents. All food categories were involved and a variety of substances were im- Questionnaires were mailed to the commis- plicated (see tables 1-3). sioners of health in each of the 50 States and the District of Columbia as well as to Federal The data provided by States are comple- agencies. For the 10-year period 1968-78, mented by the Federal responses. The two each was asked to report on the number of in- Federal agencies responsible for regulating cidents of environmental contamination of the Nation’s food supply reported the number food that resulted in regulatory action. This of environmental contamination incidents survey has limitations. Some States did not that they had identified since 1968. FDA had answer all questions. The questions were 108 reported incidents, and the subject to interpretation and misunderstand- and Quality Service of the U.S. Department of ing. The accuracy and completeness of the Agriculture (USDA) had 47 reported inci- answers were dependent on the respondent. dents (see table 3). The combined Federal and The results presented are therefore prelimi- State total number of incidents comes to 243. nary and do not necessarily represent com- Neither State nor Federal responses indi- plete and comprehensive information on all cated any significant radionuclide contam- States responding. Nonetheless, these data ination episodes during the 1968-78 period. are the first to be developed on the extent of Extensive Government programs for monitor- environmental contamination of food. ing radionuclides in food exist. Thus far, ra- Responses were received from 32 States. dionuclide contamination of food has not been Seven of the top ten agricultural States and found to exceed the exposure limits recom- six of the top ten manufacturing States re- mended in the Federal Radiation Council Pro- sponded to the questionnaire, The agricultur- tective Action Guides. In most cases, the al States in the top 10 were California, Texas, amount of food contamination in the conti- Minnesota, Nebraska, Kansas, Indiana, and nental United States has never even ap- Missouri. The manufacturing States in the proached these limits (39), While atmospheric top 10 were California, New York, Michigan, nuclear testing is less a threat today than New Jersey, Texas, and Indiana. Three of before the signing of the 1963 Test Ban Trea- these States—California, Texas, and Indi- ty, radionuclide contamination of food is still ana—are in the top ten for both agricultural a concern of both Federal and State govern- and manufacturing production. A fairly rep- ments. resentative distribution of States responded from each region of the United States (figure The number of food contamination inci- 2). dents reported to OTA does not represent the total number that has occurred in the United In the following discussions, an incident is States, only those in which the Federal Gov- defined as a case in which a Federal or State ernment and 18 State governments have ta- agency has taken regulatory action against ken regulatory action. Many incidents never contaminated food, The Michigan PBB epi- come to the attention of State or Federal sode is reported as one incident because the authorities. This is because local government contamination stemmed from one source and officials can and do handle environmental was limited to one State. Mercury contami- contaminant incidents by warning offenders nation is reported as separate incidents be- or by condemning contaminated products cause the sources differed (environmental without informing the appropriate State offi- mercury v. industrial waste), the States in- cials. Also, the farmer whose livestock or volved are widely separated, and regulatory poultry has been environmentally contami- actions were taken at different times, Eight- nated may negotiate directly with the firm re- een States reported at least one environmen- sponsible for the contamination for financial 24 . Environmental Contaminants in Food

u) shaded (Jd VT~H The ... . t:I\ ~~o~~SS production. farm of value their to Questionnaire receipts. to cash 1974 proportional Responding by were States 001-Q19-OO325-1. ranked of No. States are of Stock GPO sizes States 1976, the The if 2.-Distribution this USDA/ERS, Figure like Farmer. look questionnaire. the American would to The () .. the - ~ from States . (~ HAWAII responded .~~ Adapted .~~~ -~ United ALASKA States The SOURCE: Ch. //— Environmental Contamination of Food ● 25

reimbursement without reporting the con- ence in the environment can mean food con- tamination to Federal or State officials (40). tamination above the action level or tolerance for many years after the source of the pollu- Economic Impact tion has been stopped. The James River in Vir- ginia, for example, is still closed to commer- The economic impact of an incident involv- cial fishing several years after kepone dis- ing the environmental contamination of food charges into the river have been eliminated. includes the cost of condemned food, health The relative influence for each of these fac- costs, and the corresponding distributional tors on the final cost will vary in each con- effects and costs. The magnitude of the eco- tamination incident. nomic impact is determined by: Estimates of the cost of food condemned the amount of food contaminated, through regulatory action are most often ex- the concentration of the contaminant in pressed in dollars. Consequently, this cost is food, usually (and incorrectly) cited as representa- the chemical and toxicological charac- tive of the total economic impact. Such costs teristics of the contaminant, and were collected in OTA’s State and Federal the corresponding regulatory action surveys. The data, however, only partially re- taken on the contaminated food. flect the total economic impact for environ- The initial regulatory action taken by Fed- mental contamination of food in the United eral and State authorities may be the issu- States. This is because the cost of condemned food is only one component of the total eco- ance of a warning or the establishment of nomic impact of an incident. In addition, few either an action level or a tolerance. A more of the incidents reported to OTA included detailed discussion of this regulatory action is presented in chapter III. Action levels and tol- data on the cost of food condemned. OTA esti- mates from the available data that the total erances establish a permissible level for the cost of condemned food as a result of environ- contaminant in food. Any food found to con- mental contamination in the United States tain concentrations of the substances above 1968 is over $282 million (table 6). this level is condemned and either destroyed since The or restricted from being marketed. only cost estimates used were those clearly stated for an incident by the reporting States or Federal agencies. Costs of Food Condemned State Estimates. —Of the 18 States report- In addition to the four factors listed above, ing contamination incidents, only 6 provided the cost of condemned food is also affected by data on the economic impact in dollar terms. its position in the food production and mar- Of those six, Michigan represents 99 percent keting process at the time of condemnation, of the total cost ($255 million) while reporting An action level or tolerance for a contam- only 19 percent of the number of incidents in inant is the most important of the five factors. the 18 States. Indeed, Michigan accounts for If no action level or tolerance is set, no food 90 percent of the total costs reported in the would be condemned and thus there would be United States while reporting only 7 percent no costs incurred. The impact of such a reg- of the incidents that occurred during the ulation will depend on the exact level of a 1968-78 period. It must be recognized, how- substance that is allowed to be present in ever, that 84 percent of Michigan’s costs are food, attributed to the PBB incident. Many inci- The chemical properties of a contaminant dents reported by State and Federal agencies are also important because of the potential are considerably smaller than the PBB epi- for long-term effects on the amount of food af- sode. Thus, the PBB episode is an indication fected. Since many contaminants biologically of how severe a contamination incident can and chemically degrade slowly, their pres- be. 26 . Environmental Contaminants in Food

Table 6.—Economic Impact of Food Contamination Food Safety and Quality Service (FSQS) re- —————— ————— — ported food condemnation cost estimates. Reported incidents Total estimated cost ($) These estimates, however, only cover live- State Idaho. . . $ 100,000 stock and poultry—the food products over PCP 3,000 which FSQS has regulatory authority. FDA, Colorado. . . . Dieldrin 100 which has regulatory authority over the re- Mercury 3,700 Maryland, Mercury 23,000 maining food commodities, did not estimate Texas. . . Mercury 85,000 costs for reported environmental contamina- Indiana . Dieldrin 25,027 tion incidents (70 percent of the Federal Dieldrin 250,000 Michigan. Mercury 10,000,000 total). Thus, a significant proportion of the PCB 30,000,000 total costs for environmental contamination PCNB 100,000 incidents requiring Federal action is un- PBB 215,000,000 Picloram 12,000 known. Comparison of the two agency re- Chlordane 2,500 sponses with the State responses reveals lit- DDT 2,000 tle duplication in the reporting of incidents. Toxaphene 2,000 Parathion 328 Diazinon 13,700 FSQS cost estimates were determined by Pentachlorophenol 28,468 the number of animals or pounds destroyed PCB 150,000 Dieldrin 12,500 multiplied by the market value at the time of confiscation. Since most of these animals $255,813,323-- Federal were taken at the farm or wholesale level, the USDA/FSQS Pesticides 18,900,000 market value was the farm or wholesale Mercury 63,000 PCB 7,450,000 price. Most of the losses resulting from FDA Phenol —350 —— actions would be based on a wholesale or re- 26,413,350 tail price because the seized products had ad- Total United States ...... $282,226,673 vanced further in the marketing system. Therefore, their estimated costs would be SOURCE Off Ice of Tec;n810gy Assessment greater than if they were seized at the pro- duction level (generally the case with FSQS Some States reported the amount of food seizures). destroyed without estimating the cost. Ken- tucky, for example, reported the destruction Summing up, the available data on the cost of 400,000 lbs of milk since 1968 because of of condemned food is limited; consequently pesticide contamination. While such informa- OTA’s $282 million condemned-food estimate tion can be converted into dollars, data on is likely to be a gross underestimation of the market position and price of product at time actual costs. The true cost would be impossi- of confiscation are not readily available. ble to estimate from this limited sample. Many States were unable to provide any esti- mates on either the cost or the amount of food condemned as a result of reported contami- Health Costs nation incidents. New York (with PCBs) and Health costs are also an important compo- Virginia (with kepone) are two States that nent of economic impact. These costs are in- could not provide cost estimates for food con- curred by the consumer whose health has or demned as a result of environmental contami- potentially can be affected adversely by a nation. Virginia, however, has initiated a contaminant present in food. These adverse study to determine the economic impact of the effects can cause illness and death, and the kepone incident. range of effects will vary depending on the Federal Estimates.—Of the two Federal toxicity of the contaminant, the concentration agencies reporting information to OTA on en- of the contaminant in food, and the amount of vironmental contaminant incidents, USDA’s food consumed. Ch II—Environmental Contamination of Food s 27

In this country, the concentration of con- Producers.— Food producers are affected taminants has been at levels that have not economically in different ways by contamina- produced immediate measurable and conclu- tion episodes. But all are affected directly sive effects in exposed populations. Estimates when the food they produce is condemned. are therefore made for the potential long- For example, food found contaminated at the term effects on exposed populations from farm level is confiscated and destroyed. This various contaminants in food. was the case for over 500 Michigan farmers whose dairy herds were partially or entirely Health costs can be estimated from such destroyed (41). In such cases, farmers either projected health effects. Costs would include replace their livestock, plant a new crop, or health care costs for treating illness and go out of business. burial expenses associated with death. Addi- tional costs would include estimated value of Farmers can be faced with severe econom- productive days or years lost from work due ic hardship, since they are not always reim- to the projected illness or death associated bursed financially for the animals or com- with the contaminant in food. All of these modities confiscated. While insurance pro- health-related costs, however, do not and grams such as the Federal Crop Insurance cannot include the emotional and psycholog- Corporation are available to cover natural ical impacts on those afflicted and their hazards which might destroy crops or live- friends and families. stock, such Federal assistance is not avail- able to farmers for losses from environmental Health costs are not available for previ- contamination. An injured farmer can obtain ous U.S. food contamination incidents. Ap- a loan at commercial rates or sue the respon- proaches and techniques for estimating sible firm for compensation. But the loan and health costs are discussed in chapter VI. the interest add to a farmer’s financial dif- ficulties, and suing for compensation can take Distributional Effects and Costs time that the farmer may not have. Distributional effects and costs involve the The commercial fisher is faced with a dif- various people, groups, and organizations ferent situation. If a river, lake, or species of who are economically affected by an environ- fish is restricted because of environmental mental contamination incident. Information contamination, the fisher whose source of in- on the extent and distribution of such effects come depends on this species or waterway and costs provides a clearer picture of the may have few employment alternatives. The total economic impact on society. This infor- alternatives depend to some degree on the ex- mation is usually couched in descriptive tent of the contamination. If the only water- terms. Those who are economically affected way available in a section of a State or a are identified but the extent of the impact is whole State is closed to commercial fishing seldom estimated in dollars. The exact distri- because of the contamination, the fisher’s bution of costs from an incident through soci- source of employment is eliminated until the ety is affected by the same five factors that restriction is ended. Since the restriction can influence the cost of condemned food. last for years (depending on the chemical sta- bility of the contaminant), the fisher either Many of the distributional effects and will have to move to other commercial fishing costs for various types of environmental con- areas or seek other employment. taminant incidents are discussed in the fol- lowing sections. The purpose of this discus- Food producers economically affected by sion is not to identify all the distributional the condemnation of contaminated food are costs but rather to demonstrate the variety of likely to incur health costs. This is because effects and costs that can result from an inci- many of the producers and their families reg- dent. ularly eat the food that they produce or har- 28 . Environmental Contaminants in Food vest. Consequently they are exposed to the handle the kind of long-term problems cre- contaminated food at greater concentrations ated by a PBB or kepone incident. The Michi- than the average consumer. This was the gan Department of Agriculture, for example, case for several farm families in Michigan. estimates it will spend $40 million to $60 mil- lion within the next 5 years to monitor and Firms Held Accountable for Environmen- test for PBBs in animals and animal byprod- tal Contamination.—In most instances blame ucts (44). This is money that could have been for a contamination incident can be estab- saved or spent for other programs if PBB con- lished. Those accountable are subject to fines tamination had not occurred. In order to re- and lawsuits. Firms admitting responsibility cover its expenses from the PBB incident, the often try to settle with producers out of court State of Michigan filed a lawsuit against both if possible. Most of the compensation is for the Michigan Chemical Corporation and the economic damages stemming from the de- Farm Bureau Services, Inc., claiming more struction of food or loss of employment. Com- than $100 million in damage (45). pensation for people whose health has been impaired as a result of eating contaminated Federal involvement is limited unless the food would be sought through civil litigation. contaminated food is part of interstate com- Such litigation, however, is rare in this coun- merce, Many of these incidents are not con- try, since the level of contamination in food is sidered by the Federal Government to involve so low that demonstrating the necessary interstate commerce, FDA may provide tech- cause and effect is difficult. nical assistance at the request of the State Fines or compensation paid by the firms government when a contamination incident is regarded to be a local problem (43). These held accountable for the contamination are, technical facilities and experts are available in fact, poor indicators of the true costs in- to all States through the Federal and regional curred by the producers. This is because the offices. Additional expenditures by the Fed- settlement costs which are frequently negoti- eral Government for contamination incidents ated or imposed bear little relationship to the are limited. Additional State expenditures, actual costs incurred. however, can be substantial. Federal expend- For example, compensation has been pro- itures are made when Federal regulations vided by Michigan Chemical Corporation and are developed and promulgated for particu- Farm Bureau Services, Inc., to many of the lar contaminants in food such as PCB. farmers whose livestock and poultry were de- stroyed following PBB contamination. Mich- Consumers. —Consumers can incur costs igan Chemical and Farm Bureau Services from an environmental contaminant incident have together paid more than $40 million in in several ways. The removal of food from compensation from a jointly established in- commerce could increase prices for that food surance pool (42). In another case involving product or other food products being sold. PCB-contaminated fish meal sold to poultry Thus, the consumer could pay more for food producers, Ralston Purina Company negoti- as a result of an environmental contamina- ated compensation for the 400,000 chickens tion incident. In order for this price increase destroyed. The cost of the compensation has in food to occur, however, a significant not been disclosed (43). amount of a food product or food products would have to be taken off the market. Such Governments.— Federal, State, and local prices of food might vary by State or region governments also incur costs from an envi- and affect certain socioeconomic classes dif- ronmental contamination incident. Although ferently. the Federal Government and most State gov- Health costs could increase as a result of ernments have agencies with programs to the consumption of contaminated food. This regulate or control food safety problems, would not affect all consumers but rather these programs usually are not funded to those who received the most exposure and/or Ch. II—Environmental Contamination of Food ● 29

those most susceptible to a contaminant, such ple, a bait and tackle store on a lake that is as children or senior citizens. While these closed to commercial and sport fishing be- costs would already be included in estimated cause of an environmental contamination is total health costs, the distributional effects likely to suffer economic hardship, Food proc- could indicate those consumers most likely to essors whose normal supply of food has been be affected. condemned because of environmental con- Indirect Costs. —Most of the costs men- tamination will also suffer economically un- tioned directly stem from an environmental less they find new sources of supply. These contamination incident. However, indirect or are just two examples of the many indirect secondary costs can and do occur. For exam- costs which might occur.

Because a limited number of substances tential hazard and 2) by surveying the uni- posing health problems already have been verse of industrial chemicals and ranking identified in food, concern exists that other them according to their potential for entering toxic substances are likely to contaminate the food supply in toxic amounts. These meth- food in the future. This concern arises from ods are discussed in more detail in chapter the number of substances presently being VII, “Monitoring Strategies. ” manufactured, used, and disposed of in the United States, and the difficulties in prevent- Of the three categories of environmental ing them from entering the environment. New contaminants considered in this report, or- substances developed to meet new needs or to ganic chemicals probably pose the greatest replace known toxic substances may create potential environmental and food contamina- unexpected environmental problems if not tion problems. This conclusion is based on the properly controlled. Byproducts of new tech- number, volume, and toxicity of the organics nologies such as synthetic fuels are also po- manufactured and used in this country (40). tential environmental contaminants. These Both trace metals and radioactive substances are described in appendix A. continue to warrant concern, but not as great There are two methods of objectively as- a concern as organic substances. The extent sessing possible future contaminants: 1 ) by of food contamination from these substances sampling the food supply for chemical con- depends on our success in preventing them taminants and ranking them according to po- from entering the environment.

CONCLUSIONS AND ISSUES

Data presented here indicate that environ- health risks were considered unacceptable. mental contamination of food is a nationwide These episodes have resulted in economic problem of unknown magnitude. Long-term, losses when contaminated food was removed low-level exposure to toxic substances in food from the market, poses health risks that are difficult to The following chapters analyze several evaluate given present techniques. Incidents issues related to the regulation of environ- of high-level contamination of food that cause mental contaminants in food. These are: human illness have not occurred in the United States, However, regulatory actions have Q Is our present regulatory system protect- been taken to restrict consumption of con- ing the public health? (Chapters 111 and taminated food in cases where the potential IV) 30 . Environmental Contaminants in Food

● Are methods used by the regulatory how should economic impacts be evalu- agencies for estimating health impacts ated? (Chapters III and VI) the most appropriate ones? (Chapters III and V) c Should regulatory monitoring be capable ● Should economic impacts be an explicit of detecting substances as they enter the part of regulatory decisionmaking? If so, food chain? (Chapters VII and VIII)

CHAPTER II REFERENCES

1. Storclz, William J. “C&EN’s Top Fifty Chemi- phenyls (Aroclor 1254), ” ]ournal of The Na- cal Products and Producers, Chemical and tional Cancer Institute 53:547, 1974. Engineering News 56(18):33, 1978, 13, Allen, J. R., et al. “Residual Effects of 2. Murray, Chris, “Chemical Waste Dispos- Short-Term, Low-Level Exposure of Non-Hu- al A Costly Problem, ” Chemica~ and Engineer- man Primates to Polychlorinated Biphenyls, ” ing News, p. 12, Mar, 12, 1979. Toxicology and Applied Pharmacology 3. National Academy of Sciences, National Re- 30:440, 1974. search Council, Commission on Natural Re- 14, McNulty, W. P. “Primate Study, ” Nation- sources, Environmental Studies Board, Com- al Conference on Polychlorinated Biphenyls, mittee on the Assessment of Polychlorinated Chicago, 1975. Biphenyls in the Environment. Polychlori- 15. Barsotti, D. A., et al. “Reproductive Dysfunc- natea’ Biphenyls (ISBN O-309-02885-X), 1979, tion in Rhesus Monkeys Exposed to Low Lev- 4. Hammons, A. “Levels of Chemical Contam- els of Polychlorinated 13iphenyls (Aroclor inants in Nonoccupationally Exposed U.S. 1248), ” Food and Cosmetics Toxicology 14:99, Residents,” draft manuscript prepared for 1976. Health Effects Research Laboratory, Environ- 16, Allen, J. R., and D. A. Barsotti. “The Ef- mental Protection Agency, 1978. fects of Transplacental and Mammary Move- 5. U.S. Food and Drug Administration. “Poly - ment of PCBS on Infant Rhesus Monkey s,” chlorinated Biphenyls (PCBS), Federal Regis- Toxicology 6:331, 1976. ter 42(63):17487, 1977, 17. Allen, J. R. “Response of the Non-Human Pri- 6. Kimbrough, R, D. “The Toxicity of Polychlori- mate to Polychlorinated Biphenyl Exposure, ” nated Polycyclic Compounds and Related Federation Proceedings 34:1675, 1975. Chemicals, ” CRC Critical Reviews in Tox- 18. icology 2:445, 1974, Meester, W. D. and D. J. McCoy, Sr. “Human Toxicology of Polybrominated Biphenyls, ” 7, Fishbein, L, “Toxicity of Chlorinated Biphen- Clinical Toxicology 1o(4):474, 1977. yls, ” Ann. Rev. Pharmacol, 14:139, 1974. 8. Peakall, D. B. “PCBS and Their Environ- 19. Fries, G. F., G. S, Marrow, and R. M. Cook. mental Effects, CRC Critical Reviews on En- “Distribution and Kinetics of PBB Residues in vironmental Contaminants 5:469, 1975. Cattle, ” Environmental Health Perspectives 9. National Institute of Occupational Safety and 23:43, 1978. Health. Criteria for a Recommended Stanci- 20, Lillis, R., H. A. Anderson, J. A. Valcinkas, S. ard. Occupational Exposure to Polychlori- Freedman, and I. J. Selikoff. “Comparison of nated Biphenyls (PCBS), 1977 (DHEW Publ. Findings Among Residents of Michigan Dairy No, 77-225). Farms and Consumers of Produce Purchased 10. Kimbrough, R, D., et al. “Animal Toxicity, ” From These Farms,” Environmental Health Environ, Health Persp. 24:173, 1978. Perspectives 23:105, 1978. 11, Cordle, F., et al. “Human Exposure to Poly- 21. Environmental Protection Agencv, Office of chlorinated Biphenyls and Polybrominated Toxic Substances, Special “Act;ons Group. Biphenyls, ” Environ. Health Persp. 24:157, “Assessment of the Hazards of Polybromi- 1978. nated Biphenyls (PBBs)” (draft), 1977. 12. Kimbrough, R. D,, and R. E. Linder. “Induc- 22. Wolff, M. S., B. Aubrey, F. Camper, and N. tion of Adenofibrosis and Hepatomas of the Hames. “Relation of DDE and PBB Serum Lev- Liver in BALB/oJ Mice by Polychlorinated Bi- els in Farm Residents, Consumers, and Michi- Ch. //— Environmental Contamination of Food ● 31

gan Chemical Corporation Employ ees, ” Envi- 35. Rizzo, A. M., and A. Furst. “Mercury Terato- ronment] Health Perspectives 23: 177, 1978. genesis in the Rat, ” Proceedings of the West- 23. Cook, H. D., R. Helland, B. H, Vander Weele, ern Pharmacology Society 15:52, 1972. and R. J, DeJong, “Histotoxic Effects of Poly- 36. Evans, H. L., R. H. Garman, and B. Weiss. brominated Biphenyls in Michigan Dairy Cat- “Methylmercury: Exposure Duration and Re- tle,”’ Environmental Research 15:82, 1978. gional Distribution as Determinants of 24. Kay, K. “Polybrominated Biphenyls (PBB) En- Neurotoxicity in Nonhuman Primates, ” Tox- vironmental Contaminants in Michigan, 1973- icology and Applied Pharmacology 41: 15, 1976, ” Environmental Research 13:74, 1977. 1977. 25, Bekesi, J. G., J. F. Holland, H. A. Anderson, A. 37. National Academy of Sciences, ‘ Advisory S. Fischbein, W, Rem, M. S. Wolff, and I. J. Committee on the Biological Effects of Ioniz- Selikoff. “Lymphocyte Function of Michigan ing Radiations. Effects on Populations of Ex- Dairy Farmers Exposed to Polybrominated Bi- posure to Low Levels of Ionizing Radiation, phenyls, ” Science 199:1207, 1978. Washington, D. C., 1972. 26, United Nations Environment Program and the 38. National Academy of Sciences, Food Protec- World Health organization. Environmental tion Committee. 13adionuclides in Foods, l-iec~lth Criteriu 1: Mercury, Geneva, p. 23, Washington, D. C., 1973. 1976, 39. Dodge, Chris. “Contamination of Food by Ra- 27. National Academy of Sciences, National Re- dioactive Nuclides, ” Congressional Research search Council, An Assessment of Mercury in Service, OTA Working Paper, August 1978. the Environment, Washington, D, C., 1978. 40. Monitoring Advisory Panel, Sept. 21, 1978. 28. U.S. Environmental Protection Agency, Ambi- 41. Michigan Department of Agriculture, Infor- en t Water Quulity Criteria: Mercury, 1979. mation and Education Division. “A Brief 29. Lu, F. C. “’Mercury as a Food Contaminant, ” Chronology of the PBB Incident” (no date). WHO ChronicJe 28:8, 1974. 42 Anonymous. “Two Firms Fined $4,000 Each 30, D’Itri, F, M. “The Environmental Mercury for PBB Contamination of Cattle Feed in Problem, ” CRC Press, p. 73, 1972. Michigan, ” Toxic Material News, p, 144, May 31. Friberg, L., and J, Vestal. “Mercury in the En- 24, 1978, vironment, ” Cl? Caress, 1972. 43. Jaroslovsky, Rich. “Contamination of Fish- 32. Gerstner, H, B,, and J. E. Huff. ‘*Clinical Tox- meal with PCBS Sparks FDA Study of Mishap icology of Mercury, ” Journal of Toxicology at a Ralston Purina Plant, ” The Wall Street and Environmental Health 2:49 1, 1977. Journal, p. 46, July 12, 1978. 33. Olson, F. C,, and E. J, Massaro. “Pharmacody - 44. Whitehead, George, Deputy Director for Con- namics and Tera tologic Effects of Transpla- sumer Affairs, Michigan Department of Agri- centally Transported Methyl Mercury in the culture, Jan. 16, 1978. Nlouse, ’” Jnternationul Conference on Heuvy 45 McNally, JoAnn. “Polybrominated Biphenyls: Met(lis in the Environment, pp. 32 and 184, Environmental Contamination of Food, ” Con- 1972. gressional Research Service, OTA Working 34. Weiss, B., and R. H, Doherty. “Methylmer- Paper, Sept. 21, 1978. cur~ Poisoning, “’ Teratology 12(3):311, 1976. Chapter III

Federal Laws, Regulations, and Programs Chapter III

Federal Laws, Regulations, and Programs

FEDERAL LAWS

Congress has enacted several laws that not only regulate but also attempt to limit or restrict the introduction of toxic substances into the environment. Table 7 summarizes the Federal laws affecting toxic substances control. Some of these laws give Federal agencies authority to prevent unsafe food from reaching consumers. Most important in terms of this assessment is the Federal Food, Drug, and Cosmetic (FD&C) Act (l). Broadly speaking, this statute prohibits the introduction of adulterated food into interstate commerce. The FD&C Act allows the Food and Drug Administration (FDA) to establish tolerances for toxic substances whose occurrence in food cannot be avoided. The Poultry and Poultry Products Inspection Act (2) and the Federal Meat Inspection Act (3) give the U.S. Department of Agriculture (USDA) authority to inspect meat, poultry, and their byproducts. The adulteration provisions of these Acts govern all environ- mental contaminants except for pesticides that may occur in meat and poultry. Under these laws, section 408 of the FD&C Act applies to such pesticide contami- nation. In practice, USDA uses the tolerances established under the FD&C Act and consults with FDA and the Environmental Protection Agency (EPA) to deter- mine an action level when no tolerance exists.

Following is a brief summary of the perti- ly because of their economic utility or the in- nent provisions of the FD&C Act: ability of existing, commonly used production methods to eliminate them. The legislative Section 402(a)(l) declares that any food history of section 406 is skimpy. But Con- that “bears or contains any poisonous or gress’ principal objective apparently was to deleterious substance which may render it in- permit continued use of pesticides on raw jurious to health” is adulterated. The single agricultural commodities while giving FDA an exception is if the substance is not, in the effective means of control—the power to de- language of the Act, “added” to the food. In clare illegal any food that contained any such cases, the presence of the substance amount of an added substance that exceeds does not imply that the food is adulterated FDA tolerance. Congress left the distinction unless it is present in sufficient quantity to between “added”’ and other constituents un- “ordinarily render it injurious to health. ” defined, and did not attempt to clarify the The FD&C Act recognizes that certain concepts of “necessary” or “unavoidable”’ in “added” toxic substances in foods require section 406(4). special attention, Section 406 empowers FDA to establish tolerances for “added’” poisonous Amendments to the 1938 FD&C Act deal substances whose occurrence in food cannot with specific categories within the broad be avoided or whose use is “necessary” to class of substances “added”’ to foods in- produce the food. Thus, Congress authorized cluding pesticides, food additives, vitamins FDA to “license” the presence of certain and minerals, and animal drugs. Each amend- potentially toxic substances in food, seeming- ment in effect establishes a system under 35 36 ● Environmental Contaminants in Food

Table 7.— Federal Laws and Agencies Affecting Toxic Substances Control

Statute Year enacted Responsible agency Sources covered Toxic Substances Control Act ...... 1976 EPA All new chemicals (other than food additives, drugs, pesticides, alcohol, and tobacco) and existing chemicals not covered by other toxic substances control laws Clean Air Act...... 1970, amended 1977 EPA Hazardous air pollutants Federal Water Control Act (now Clean Water Act) ...... 1972, amended 1977 EPA Toxic water pollutants Safe Drinking Water Act ...... 1974, amended 1977 EPA Drinking water contaminants Federal , Fungicide, and 1947, amended 1972, 1975, EPA Pesticides Act ...... 1978 Act of July 22, 1954, (codified as Section 346(a) of the Food, Drug, 1954, amended 1972 EPA Tolerances for pesticide residues in and Cosmetic Act) . food Resource Conservation and Recovery Act ...... 1976 EPA Hazardous wastes Marine Protection, Research, and Sanctuaries Act ...... 1972 EPA Ocean dumping Food, Drug, and Cosmetic Act ...... 1938 FDA Basic coverage of food. drugs, and cosmetics Food additives amendment ...... 1958 FDA Food additives Color additives amendments ... 1960 FDA Color additives New drug amendments ...... 1962 FDA Drugs New animal drug amendments. .. .1968 FDA Animal drugs and feed additives Medical device amendments .. 1976 FDA Medical devices Federal Meat Inspection Act...... 1967 USDA Food, feed, and color additives; Poultry Products Inspection Act .. ..1957 USDA pest icicle residues in meat and Egg Products Inspection Act ...... 1970 USDA poultry products Fair Packaging and Labeling Act .. ..1976 FDA Packaging and labeling of food and drugs for man or animals, cos- metics, and medical devices Public Health Service Act ...... 1944 FDA Sections relating to biological products Occupational Safety and Health Act .1970 OSHA, NIOSH Workplace toxic chemicals Federal Hazardous Substances Act. 1960 CPSC Hazardous (including toxic) household products (equivalent in many instances to consumer products) Consumer Product Safety Act 1972 CPSC Hazardous consumer products Prevention Packaging Act .. 1970 CPSC Packaging of hazardous household products Lead-Based Paint Poisoning 1973, amended 1976 CPSC, HEW, HUD Use of lead paint: on toys or furniture, Prevention Act ., on cooking, drinking, and eating utensils, in federally assisted housing Hazardous Materials Transportation 1975, amended 1976 DOT (Materials Transportation of toxic substances Act ., Transportation generally Bureau) Federal Railroad Safety Act . . . . . 1970 DOT (Federal Railroad safety Railroad Admin.) Ports and Waterways Safety Act 1972 DOT (Coast Guard) Shipment of toxic materials by water Dangerous Cargo Act .. ...1952 Federal Mine Safety and Health Act. .1977 Labor (Mine Safety Toxic substances and other harmful and Health Admin.), physical agents in coal or other NIOSH mines

CPSC = Consumer Product Safety Commlsslon HUD = Housing and Urban Development DOT = Department of Transportation NIOSH = National Institute for Occupational Safety and Health EPA = Environmental Protection Agency OSHA = Occupational Safety and Health Administration FDA = Food and Drug Admlnlstratlon USDA = United States Department of Agriculture HEW = Health Educatton and Welfare SOURCE Environmental Law Institute, An Andlys(s 01 ~asl federal Efforts To Contro/ Tox)c .Subslances Washington D C 1978 Ch. Ill—Federal Laws. Regulations, and Programs ● 37

which FDA is empowered to license, and could have regulated them all under the “may thereby limit, the use (or in the case of render injurious” language of section 402(a) pesticide residues, the occurrence) of poten- (1). But this provision would not have given tially toxic substances in or on food. FDA authority to determine administratively what levels of a contaminant could be toler- The Pesticide Chemicals Amendment of ated. FDA would have been required to prove 1954, now section 408 of the FD&C Act, pro- its claim of hazard each time it seized a con- vides that a raw agricultural commodity shall taminated product. be deemed to be adulterated if it bears or con- tains any residue of a pesticide that does not Accordingly, since the early 1970’s FDA conform to a tolerance established under sec- has classified environmental contaminants as tion 408, Pesticides that are unintentionally “added poisonous or deleterious substances’ present on commodities are usually consid- whose occurrence cannot entirely be ered environmental contaminants by FDA, avoided, thus avoiding the less rigorous “or- and are regulated under section 406. dinarily injurious” standard of section 402(a) At no point, in 1938 or subsequently, has (l). The tolerance-setting authority of section Congress specifically addressed the problem 406 can then be applied to environmental of environmental contaminants in food. FDA contaminants in food.

Relying on section 406, FDA prescribes the data available (which are often incomplete), level of a contaminant that, under section FDA will prescribe what level of contamina- 402(a)(2)(A), will render a food adulterated. tion will trigger enforcement action (5). Before FDA can ascertain this level, suffi- Regulatory procedures employed to control cient scientific data must be accumulated to environmental contaminants in food include answer several questions implicitly posed by the establishment of action levels or toler- sections 402 and 406. FDA must be able to ances. A formal tolerance is a regulation hav- determine that the environmental contami- ing the force of law. Tolerances are adopted nant in question is: through formal rulemaking procedures and 1. added, specify the level of a contaminant that will 2. poisonous or deleterious, render a food adulterated. If supported by 3. a substance unavoidable by good manu- substantial evidence in the rulemaking rec- facturing practice, and ord, FDA’s tolerance cannot be questioned by 4. one which may make the food injurious any court. An action level is an in.formed judg- to health. - ment about the level of a food contaminant to which consumers may safely be exposed. It is Furthermore, an analytical method that can a statement of FDA’s professional judgment reliably detect, measure, and confirm the and represents a commitment to initiate regu- identity of the contaminant in the food under latory enforcement action against any lots of scrutiny must be available (5). food discovered containing excess levels. Es- sentially the same criteria are considered in To determine whether these requirements establishing tolerances and setting action lev- can be met, FDA scientists explore and re- els. The principal differences between the view the scientific literature, consult FDA two approaches lie in the procedures for files, and draw on information available in their adoption, the strength of the scientific other agencies, in academia, or in private in- data supporting them, and the differing dustry. Then, based on the best scientific weight they carry in court (4). 38 . Environmental Contaminants in Food

FDA will set a tolerance when the follow- ments will be accepted, FDA makes no com- ing conditions exist: mitment to respond to any comments or, in- deed, to reconsider the action level within 1. The substance cannot be avoided by any specified period (6). The process does not good manufacturing practice. require a detailed public discussion of the 2. The tolerance established is sufficient for the protection of the public health, selected levels, nor does it trigger a public taking into account the extent to which debate about the correctness of FDA’s prem- the presence of the substance cannot be ises or its balancing of relevant factors. avoided and the other ways in which Finally, the Commissioner of FDA may ex- the consumer may be affected by the same or related poisonous or deleteri- empt from regulatory action any contam- ous substances. inated food if he determines “based upon all 3. No technological or other changes are available scientific evidence, that the food is foreseeable in the near future that safe for consumption and that destruction or might affect the appropriateness of the diversion of the food involved would result in tolerance established (6). a substantial adverse impact on the national food supply” (6). This has only happened To establish a tolerance, FDA first pub- once, and the action did not involve human lishes a proposal, accepts comments, and food, issues a “final’* regulation, Formal objections can be raised to this “final” tolerance. Such If the environmental contaminant is a pes- objections can, if they raise material issues of ticide for which no tolerance has been estab- fact, require a lengthy trial-type hearing lished by EPA, FDA relies on EPA to recom- before an FDA administrative law judge, who mend an action level (as it did in the case of then issues an initial decision based on the kepone). In other respects, the procedures formal hearing record. That decision, in turn, and criteria for regulating pesticides as envi- can be appealed to the FDA Commissioner ronmental contaminants are the same as for (who issued the original “final” tolerance). other contaminants. The Commissioner’s ultimate decision is sub- ject to review in a court of appeals (l). Criteria for Setting Action Levels and Tolerances Because the tolerance-setting procedure is cumbersome and time-consuming, FDA ini- For the setting of action levels or toler- tially relies on an action level when it regu- ances, neither the law nor regulations re- lates an environmental contaminant. An ac- quire FDA or EPA to follow a standardized tion level is an administrative guideline and set of toxicologic protocols to evaluate risk. the functional, though not legal, equivalent of No policy exists defining the relative weight a section 406 tolerance. It is established to be given to evidence. The burden of proving when “technological or other changes that there is a health hazard lies with FDA. When might affect the appropriateness of the toler- setting an action level or tolerance, FDA con- ance are foreseeable in the near future” (6). siders the following types of data: To set an action level, FDA simply an- 1. available acute and chronic toxicologi- nounces in the Federal Register that it is cal data, including information on the establishing an action level for a contami- biological half-life of the substance and nant, and states that the data supporting the its metabolic fate; designated level are available for public in- 2. available data on the levels and inci- spection. This announcement may briefly dis- dence of the contaminant in the overall cuss the pertinent factors that went into the food supply and specifically in the food decision, but any discussion is not likely to be commodity or commodities that are (nor is required to be) extensive. While the being considered for an action level or announcement also notes that public com- tolerance; Ch. IIl—Federal Laws, Regulations, and Programs ● 39

3. normal serving sizes of the concerned associated with consumption of a toxic sub- food(s) and frequency of ingestion; stance in food, Both agencies consider three 4. susceptibility of certain population areas when evaluating the scientific informa- groups, such as infants and the aged, to tion: 1) existing animal toxicity data, 2) ex- adverse effects from anticipated dietary isting human toxicity data, and 3) exposure exposure to the contaminant: data based on the level of the contaminant in 5. the level at which available analytical food and the average consumption of that techniques can detect, measure, and food. Both agencies also consider what effect confirm the identity of the contaminant; an action level or tolerance will have on the 6. capability of manufacturers to monitor availability of food. their food production to ensure that the products comply with the action level or Evaluation of the Scientific Data tolerance; and When a food contaminant is identified, the 7. the anticipated impact of various possi- first step in establishing an action level or ble levels of regulation on the national tolerance is to assemble and evaluate all food supply. available information on its toxicity. This in- In response to questions on how FDA eval- formation comes from articles published in uates these data, Commissioner Donald Ken- the scientific literature, information provided nedy wrote: by private industry, and data from other Gov- ernment agencies. From animal toxicity data Each factor is assessed individually (as- suming information on each is available) and and whatever human toxicity data may be then collectively brought into balance by a available, a no observed effect level (NOEL) is composite analysis in terms of the estimated calculated and expressed in milligrams per risk to the public health versus both the ex- kilogram of body weight per day. NOEL is the tent to which the substance is unavoidable level at which the substance had no observed and the quantity of food that would be unlaw- effects. ful under levels being considered (5), An (ADI) is then FDA has not fixed the weight to be given to calculated by dividing the NOEL by a safety each of the above factors. Each will, to some factor. The term “acceptable” does not imply degree, influence the final decision; general- absolute safety for all people in all cases, and ly, the more information about a particular the term “safety factor” implies more than it factor, the greater its influence, This is one seems. The safety factor reflects the uncer- reason that FDA offers for not prescribing a tainty of translating animal data to humans, predetermined quantifiable set of criteria for the variability of the human population, the each factor. The amount and quality of infor- insufficiency of the data available, and the mation available when FDA encounters an severity and reversibility of toxic effects. environmental food contamination problem are inevitably unpredictable, FDA maintains When 2-year chronic toxicity studies in that because of this uncertainty, it is imprac- animals are available, the safety factor used tical to state in advance the precise weight of is 100. When threshold levels have been ob- each factor in the final determination. How- served in humans, the safety factor employed ever, FDA maintains that the public health is 10. If long-term studies are available and factor outweighs all others in its considera- show no irreversible effects, a much smaller tions (5). safety factor might be selected. If evaluation of available toxicological data indicates that Determining Action Levels no threshold exists, a very large safety factor and Tolerances (on the order of 1,000 or more) maybe used. F. ‘I’. Arnold, chairman of the Kepone Action In general, EPA and FDA follow similar Level Hearings, stated that “the determina- procedures when evaluating the health risk tion of an appropriate safety factor is an art 40 . Environmental Contaminants in Food

rather than a science and is dictated by the requirements for toxicity data. Moreover, in chemical in question, its toxicological proper- most instances there are no petitioners or ties and surrounding circumstances” (7). sponsors of commercial uses of the material to which FDA can look for the necessary addi- The next step is to calculate the maximum tional tests. Available industry data are fre- permissible intake. The maximum permissi- quently used because of the lack of published ble intake is the product of ADI and the or publicly available information on the sub- average weight of an adult (this figure varies stance. While known toxic effects of metabol- from 60 to 70 kg). This figure is then com- ic products of the substance in question are pared to the maximum potential exposure considered, unknown metabolizes cannot be. through consumption of contaminated food. Finally, additive and synergistic effects be- Tolerances are set so that the amount of the tween the contaminant and other toxic sub- contaminant consumed in food is less than or stances in food are not considered in the tol- equal to the ADI. Mathematically, this can be erance-setting procedures. expressed as:

ADI x Average body weight Questions are frequently raised about Tolerance = of consumer safety factors and the extrapolation from ani- mal data to humans. Comparisons between Food factor x 1.5 kg Where ADI = the NOEL divided by an appropriate animals and humans are based on milligrams safety factor of the toxic substance per kilogram of body Average body weight, although many maintain that milli- weight of gram of toxic substance per square centi- consumer = 60 kg for EPA calculations, and 70 kg for FDA calculations meter of body surface area is a more appro- Food factor = percentage of the average daily diet priate comparison, made up by the food in question 1.5 kg = the average weight of food consumed in Little scientific evidence exists to support a day safety factors. Historically, a factor of 10 in extrapolating from animals to man and a fac- Appendix B provides a detailed example of tor of 10 in extrapolating from the least-sensi- how these concepts were applied by FDA in tive human to the most-sensitive have been the development of tolerances for polychlori- used although they have little theoretical or nated biphenyls (PCBs) in food, factual basis. Many believe that the use of These procedures have several limitations. different safety factors for different toxico- The data on toxicity of an environmental con- logical effects (a greater safety factor for ir- taminant for humans and animals may often reversible effects, a smaller factor for rever- be inadequate for setting a tolerance. Be- sible effects) or the use of mathematical mod- cause of time constraints and the necessity to els to extrapolate from animal to human risk make a regulatory decision, action levels for would be more appropriate. Finally, the safe- environmental contaminants in food may be ty factor approach may not make allowances based on incomplete, scanty toxicological in- for vulnerable groups such as infants, except formation. Once a level is set, however, no in those instances in which infants are con- law requires FDA to collect new evidence on sidered the primary population at risk. Indi- the toxicity of the substance in question even viduals with predisposing conditions or previ- though that issue may well continue to be of ous exposure may not be adequately covered critical importance to the population receiv- by safety factors. In truth, it may be impos- ing the highest exposure and to food pro- sible to protect every individual with allergies ducers. or predisposing physiological conditions. Because of the nature of the problem pre- The methods used for estimating dietary sented by environmental contaminants in exposure to toxic substances are limited by food and because so little data are usually lack of sound data. FDA bases some tolerance available, EPA and FDA cannot have formal decisions on their Total Diet Studies, Because Ch. III— Federal Laws, Regulations, and Programs ● 41

these are based on the diet of a teenage male, decisionmaking process under section 406 in they do not reflect the dietary patterns of a way Congress did not intend. Obviously, vulnerable groups, nor do they reflect ethnic consideration of the amount of food loss and regional preferences or vegetarian diets. caused by a tolerance helps to ensure that the direct economic consequences of the EPA and FDA rely on USDA’s Food Consump- tolerance (in this case, decreased sales and tion Survey, which was completed in 1965-66. employment in the commercial fishing in- It is believed that some shifts have occurred dustry) will not be disproportionate to the in- in consumption patterns since then, and creased degree of public health protection USDA is now conducting a new Food Con- accomplished by the tolerance: but FDA con- sumption Survey. siders secondary economic consequences, such as potential impact on the recreational Evaluation of Decisionmaking Approach fishing industry, totally beyond the scope of section 406 (8). As discussed earlier, FDA regulates en- vironmental contaminants in food under sec- The decisionmaking process used by FDA tion 406 of the FD&C Act. This section does is a form of cost-effectiveness analysis—a not specifically address environmental con- procedure to compare the change of health taminants, but authorizes FDA to regulate effects in biological terms with the change of food for potentially toxic substances that are the cost in dollars (a further analysis of this “added” and “unavoidable’” in the produc- procedure appears in chapter VI). In the PCB tion of food. When setting an action level or decision, FDA compared change in human tolerance, FDA considers the impacts on the risk data for 5 parts per million (ppm), 2 ppm, national food supply or, stated in another and 1 ppm levels in fish with the estimated manner, the impacts on the availability of amount of food that would be condemned in food to the American consumer. In quanti- order to arrive at a 2-ppm tolerance level. In fying this criterion, FDA estimates the FDA’s judgment, a 2-ppm tolerance was a amount of food that would be banned from proper balance between “providing an ade- commerce because of the action level or toler- quate degree of public health and avoiding ance. excessive losses of food to American con- The final rule reducing PCB tolerances (8) sumers. ” While other factors were consid- illustrates FDA’s interpretation of section ered by FDA in its decision, the estimated 406 of the FD&C Act for environmental con- human risk data and loss of food are the two taminants in food. FDA clearly states that for principal factors weighed in the decision. PCBs “(i)t has had to decide, in effect, where The language of section 406 provides FDA the proper balance lies between providing an with the flexibility to interpret the un- adequate degree of public health protection avoidability requirement as it sees fit. FDA and avoiding excessive losses of food to recognizes that its regulatory decision will American consumers. ” have an economic impact, but FDA considers only a component of the total economic im- FDA later goes on to state that: pact in its decision-i. e.. food condemned. (I)n establishing a tolerance for PCBs in FDA also realizes that the amount of food fish, FDA must take into account the amount condemned will have an effect on employ- of fish a given tolerance would remove from ment and commercial sales associated with commerce. Sect ion 406 of the Act, however, the contaminated food product. It must be neither requires nor authorizes FDA to recognized that even for such a widespread weigh secondary economic impacts when it considers the level at which a tolerance contaminant as PCBs, an action level for an should he set. Consideration of such impacts environmental contaminant is likely to have a would be inconsistent with the paramount more severe impact on local employment and concern of section 406, which is protection of commercial sales than on the amount of food the public health, and would complicate the available to the American consumer. 42 . Environmental Contaminants in Food

In addition, FDA’s attempt to measure the case for incidents that condemn significant impacts on availability of food by estimating amounts of food. A more accurate estimate of cost of food condemned has several flaws. the impact on the availability of food would First, FDA estimates the amount of food ex- be to consider the percentage or amount of a pected to be condemned for one year from a food product condemned out of the total proposed action level or tolerance. For amount of that particular product that is pro- substances such as PCBs in freshwater fish, duced in this country, Then an estimate of the estimates for the amount of food expected to relative importance of the affected food prod- be condemned should be for more than one uct to the American diet would need to be year. This is because PCBs are ubiquitous in made, FDA does not do this second step. This the environment and degrade very slowly. It type of analysis would also attempt to estimate is highly likely that freshwater fish will be the impact such a tolerance would have on contaminated with PCBs at levels above 2 the supply and price of other foods available ppm and consequently restricted from com- to and consumed by the American consumer. merce for several years. While the latter analysis may be theoreti- Second, estimating the cost of food banned cally sound, given the time constraints for set- by the tolerance does not necessarily reflect ting an action level for a newly discovered the impact on the availability of food. While contaminant, it may not be practical for an this might occur for small amounts of food initial regulatory decision. The more thor- condemned—a herd of cattle or a few hun- ough analysis is more applicable to tolerance- dred gallons of milk—this would not be the setting.

FEDERAL MONITORING PROGRAMS

On the Federal level, the responsibility of of origin or processing, and attempts are monitoring foods for environmental contami- made to collect the seafoods as close to the nants falls mainly on two agencies, FDA and point of origin as possible (9). USDA. USDA limits its monitoring activities to meat and poultry products, while FDA ana- Some agricultural products are analyzed lyzes samples of animal feeds, , vege- for the presence of trace metals—lead, zinc, tables, grain, eggs, milk, processed dairy and cadmium —as well as for synthetic or- products, and seafood. Most monitoring activ- ganic chemicals. All fish samples collected ities of these two agencies could be classified for determination of chlorinated pesticides as regulatory monitoring-analysis of food and PCBs are also analyzed for mercury. samples for known environmental contami- Some canned-tuna samples are analyzed for nants and for some suspected environmental lead, cadmium, , selenium, and mer- contaminants for enforcement purposes. cury. Because some containers have lead- soldered joints that may contaminate the food, an unspecified number of canned-food Food and Drug Administration samples are analyzed to determine lead con- tent (9). FDA collects approximately 8,000 food and feed samples a year and analyzes them for a FDA also determines the total dietary in- variety of chemical residues, mainly pesti- take and exposure trends of some known and cides. Domestic food commodities comprise suspected environmental contaminants in its about 70 percent of this total, imported com- Total Diet Studies. The contaminants in- modities make up around 30 percent. In addi- cluded in this program include some pesti- tion, approximately 1,300 seafood samples cides, PCBs, mercury, lead, cadmium, arse- are collected for analysis annually. Most food nic, selenium, and zinc, Other organics and and feed samples are collected at their point metals are excluded (9). Because it involves Ch. Ill—Federal Laws, Regulations, and Programs ● 43 not the analysis of individual raw food prod- EPA monitors milk, air, water, and soil for ucts but rather combinations of prepared radioactivity. The milk-monitoring program is cooked foods which approximate a total a joint effort with State and local agencies, dietary intake, this program differs from the FDA advises and monitors the milk-sampling regulatory monitoring activities of FDA. program, which is carried out by 63 State and local health agencies. Milk from each area is Monitoring for known environmental con- sampled once a month by the State and/or taminants is carried out through the use of local inspectors and submitted for analysis to specified “accepted” methods for extraction, the EPA laboratory in Montgomery, Ala. cleanup, and identification. The procedures, when applied to samples in which regulated U.S. Department of Agriculture synthetic organic chemicals such as pesti- cides are to be determined, will often indicate USDA is responsible for evaluating the the presence of other compounds that are quality of meat and poultry products and pro- chemically similar to those under analysis. viding the consumer with products that meet The FDA Compliance Program Guidance the criteria spelled out in the Meat and Manual (10) specifies that if, during the Poultry Inspection Acts. These criteria in- regulatory monitoring analyses, unidentified clude monitoring for environmental contami- analytical responses appear (thus indicating nants. The majority of compounds evaluated an uncharacterized chemical) with “signifi- are those that are approved for use in agri- cant” intensity, data from the sample collec- culture, either administered directly to food tion and analysis should be transmitted to the animals (such as growth promoters), or ap- Bureau of Foods Laboratory in Washington, plied to agricultural crops to which food ani- D. C., which presumably will identify the un- mals may eventually be exposed (such as pes- characterized chemicals. ticides). FDA may select a chemical for further Testing of meat and poultry products for study based on production volume, toxic by- residues by USDA falls into broad monitoring products, environmental stability, volubility, and surveillance categories. The monitoring behavior, toxicity, uses, and methods of dis- activity, called the National Residue Monitor- posal. After an analytical method is devel- ing Program, is designed to determine the oped, samples that have the highest probabil- frequency at which tolerance-exceeding ity of being contaminated with the particular amounts of monitored compounds are occur- (selected) compound are collected. These ring in the national meat supply. In effect, the samples are often fish, since rivers, lakes, monitoring program is designed to evaluate and estuaries receive chemicals not only from how effectively users and/or manufacturers direct discharges from municipalities and in- of the compounds are complying with the dustry but also from erosion and runoff. Fur- laws or use restrictions (12). ther research and monitoring activities on a given chemical depend on the results of these Under the monitoring program, animal tis- initial analyses (9). sues are collected from slaughterhouses un- der Federal inspection throughout the United The present radionuclide-monitoring pro- States at a rate that will detect violations if gram is a joint undertaking by FDA, EPA, and they are occurring in at least 1 percent of the the States, This program monitors: 1) foods animal population [13). Based on statistical grown near eight selected nuclear power fa- calculations, 300 samples per compound per cilities for tritium, gamma emitters, and species have to be collected annually to deter- strontium-90; 2) food samples from the total mine a l-percent incidence with 95-percent diet studies; 3) specified imported foods; and assurance. In effect this means that the same 4) milk, , vegetables, and water collected sample of tissues may be analyzed for more near phosphate mines in Florida (11). than one compound. This level of testing re- 44 ● Environmental Contaminants in Food quires sampling approximately 1 in 8,000 as a State or region, depending on the cir- head of livestock and 1 in 700,000 poultry. cumstances. Indications of problems come These collections are stratified according to from many sources: the National Residue geographic areas, with more samples being Monitoring Program, activities of USDA in- collected from areas where meat or poultry spectors, information from State or other are slaughtered. Federal officials, or from public news sources. Because these samples are biased USDA does not monitor all compounds for (i.e., collected in response to a given need), which tolerances have been established. they do not reflect the overall condition of the There are no suitable methods to analyze national meat supply. In most cases the sam- some of them within existing regulatory ples are used to determine either the extent monitoring laboratory capabilities. Available of a problem or to evaluate the acceptability resources may also limit the number and of a herd or product. Any product found to be variety of compounds tested, The selection of in violation may not be released into com- which compounds to monitor is based on fac- merce until subsequent samples show that it tors such as frequency and patterns of use, is in compliance. These followup samples are toxicity, previous testing results, and public considered surveillance samples, concern. Major groups of compounds in the monitoring program include synthetic organic Analyses of samples collected in USDA chemicals (mainly chlorinated hydrocarbon programs are performed in a manner similar ), trace metals, antibiotics, sulfon- to those in the FDA program—prescribed amides, and certain hormones and drugs analytical methods are used. USDA, like FDA, used for growth promotion and disease con- may try to identify an “uncharacterized” sub- trol. Organophosphate pesticides were moni- stance when “unknown” peaks appear in the tored for several years but this monitoring analysis of a sample. Depending on the identi- was discontinued because residues of the ty of the compound, further investigation and parent compound were not found in animal sampling may be carried out, tissue and suitable routine methods for de- tecting the metabolizes were not practical Environmental Protection Agency (12), EPA has no mandate to regularly analyze In 1978, about 150,000 tests were com- food commodities for chemical contamina- pleted on approximately 20,000 domestic tion. Some of its programs designed to deter- samples collected under the national residue mine the ecological impact of pollutants may monitoring program. An additional 2,000 include certain types of foods for analysis. samples were collected from imported prod- This is particularly true in the case of ucts (12). Many of these samples are analyzed seafood. Samples from aquatic food chains for potential contaminants other than trace are often selected for analysis to ascertain metals and synthetic organics. In 1977, for in- whether a pollutant is concentrated as it stance, around 2,300 out of 22,000 samples moves up the food chain. were analyzed for chlorinated hydrocarbon pesticides and 1,300 for trace metals. The re- In addition, the mobilization, degradation, mainder were analyzed for antibiotics, hor- and transfer of pollutants is often studied by mones, and drugs (13). Data generated by this EPA. For example, under the auspices of program are used not only for regulatory EPA’s Chesapeake Bay program, trace metals functions but also to help pinpoint problems, and synthetic organic chemicals are being assist in trend analysis, and indicate areas studied in water, living organisms, and sedi- that need more intensive sampling. ments of the Chesapeake Bay by various State agencies and academic institutions. The re- Surveillance samples are those collected to sults of this and similar studies are not de- evaluate a problem. The area of sampling signed to protect the public from consuming may be as small as a single farm or as large contaminated food and are seldom used as Ch. Ill—Federal Laws, Regu/at/ens, and Programs ● 45

such. Rather, the results are used in protect- toring if it has caused a problem at some time ing the environment from chemical insults in the past. In other words, a compound may and helping EPA better regulate the introduc- enter the food supply and be undetected for tion of toxic substances into the environment. years but categorized as a known contami- nant when discovered. An example would be Some data transfer between EPA and other kepone in seafood of the James River in Vir- Federal agencies exists. For example, in the ginia (16). This compound is an insecticide EPA-funded kepone studies in the Chesa- that had been entering the aquatic food chain peake Bay region, kepone concentrations in for at least 7 years before discovery. Its pres- edible fish, shellfish, and crabs are obtained ence in seafood was determined after work- which assist the Commonwealth of Virginia ers in the facility that manufactured it be- and FDA in their efforts to keep contaminated came ill from industrial exposure to this toxi- sea food off the market. This is true even cant (17)0 though the study is not designed specifically for this purpose. As soon as the sick workers were discov- Another program administered by the EPA ered, samples of fish and oysters from the ad- is the Mussel Watch (14). It is designed to jacent river were analyzed and shown to con- analyze shellfish collected from strategic tain kepone. Within a few months, action locations in the marine coastal zone of the levels were established for kepone in seafood. United States for selected organic chemicals, Because existing monitoring programs ana- trace metals, and radionuclides. This pro- lyze for known regulated environmental con- gram was started in 1976 and now involves taminants, the compound was not discovered the collection of oysters and/or mussels from because it was not sought. It is unlikely that 29 stations on the west coast, 34 stations on the presence of kepone would be known and the east coast, and 26 stations on the gulf regulated today had the workers not become coast, Collections from these stations make ill. In this case, kepone was placed in the up a total of 107 samples, known-environmental-contaminants category as a result of the illness of production The trace metals for which samples are workers, not chemical monitoring. analyzed in this program include lead, cad- mium, silver, zinc, , and nickel. The ra- Another example of a toxic compound en- dionuclides measured in the samples include tering the food supply and going undetected is plutonium-238 and -239, americum-241, ce- the fire-retardant polybrominated biphenyls sium-137, curium-242 and -244, and lead-210. (PBBs), PBBs entered the food supply in Mich- The synthetic organic chemicals included in igan in the late spring or early summer of the analysis are the halogenated hydrocar- 1973 (18,19), Bags of a fire retardant contain- bons p,p’-DDE and p,p’-DDD—two of the prin- ing these compounds were shipped to farm- cipal breakdown products of the pesticide ers’ cooperative units in Michigan in place of DDT—and PCBs. Samples are analyzed for the intended livestock feed supplement, The these synthetic organic chemicals as well as fire retardant was unknowingly mixed with petroleum hydrocarbons, whose presence cattle feed and fed to herds throughout the may indicate oil pollution. Recently, a number southern part of the State. Even though farm- of uncharacterized substances have been ers soon noted the symptoms of poisoning in found by Mussel Watch scientists. They are the cattle and reported the problem to State now designing the program to encompass and Federal officials, it was almost a year systems to track and identify them (15). before the causative agents, PBBs, were iden- tified. During this year dairy products and Critique of Federal Monitoring beef contaminated with these compounds Programs were consumed by the citizens of Michigan. Soon after the compound was identified in A chemical substance found in food be- food products, action levels were established comes the subject of Federal regulatory moni- to protect the consumer. In this case the trig- 46 . Environmental Contaminants in Food ger for designation of PBBs as known environ- check for known environmental contaminants mental contaminants was sick cattle, not are insufficient to detect toxicants in food for chemical monitoring. which there have been no action or tolerance From these two case studies it appears levels established. that monitoring programs designed mainly to

CHAPTER III REFERENCES

1. Federal Food, Drug and Cosmetic Act, 21 12. Clark, G. U.S. Department of Agriculture, U.S.C. 321 et seq. 1979. 2. Poultry and Poultry Products Inspection Act, 13. U.S. Department of Agriculture. The National 21 U.S.C$ 451 et seq. Residue Program, USDA-Food Safety and 3, Federal Meat Inspection Act, 21 U.S.C. 601 et Quality Service, May 1978. seq. 14. Goldberg, E, D., V. T. Barrington, J. W. Har- 4. Merrill, R, A, FDA Re,gu~ation of Food Safety, vey, G. H. Martin, J, H. Parker, P. L. Rise- unpublished manuscript, 1978. brough, R. W. Robertson, W. Schneider, E. S. 5. Kennedy, D., Commissioner, Food and Drug and E. Gamble. “The Mussel Watch, ” Envi- Administration, to J. B. Cordaro, Jan. 22, ronmental Conservation, vol. 2, Summer 1979, 1978, 6.21 CFR 109. 15. Risebrough, R. W. Bodega Marine Labora- 7. Arnold, F. T. Kepone Action Levels—Review, tory, University of California, Bodega Bay, Analysis and Recommendations, Office of Calif., personal communication, 1978.

Special Pesticide Reviews, Environmental 16. Huggett, R. J., M. M. Nichols$ and M. E. Protection Agency, Mar. 9, 1977. Binder. “Kepone Contamination of the James 8. “Polychlorinated Biphenyls (PCBS), Reduction River Estuary, ” Proceedings of Symposium on of Tolerances, ” Federal Register 44(127): Contaminants and Sediments: Fate and 38330, 1979. Transport, R, A. Baker (cd.), Honolulu, 9. Food and Drug Administration. Report in Re- Hawaii, April 1979 (in press), sponse to the Office of Technology Assess- 17. Food and Drug Administration. Compliance ment letter of June 5, 1978—’’Request for In- Program Evaluation (FY ‘77), Kepone and formation on Federal Capabilities for Envi- Mirex Contamination (7320.79A). ronmental Mcmitoring and Toxicity Testing, ” 18. McNally, J. M, “Pol~brominated Biphenvls: 1978. Environ-mental Conta-mination of Food, ” spe- 10. Food and Drug Administration. Compliance cial paper for OTA, Library of Congress, Con- Program Guidance Manual, Compliance Pro- gressional Research Service, 1978. gram 7305.007 -’’Pesticide and Metals in 19 Michigan Department of Agriculture, Infor- Fish Programs” (FY ‘78), Oct. 1, 1977. mation and Education Division. “A Brief 11. Food and Drug Administration, Compliance Chronology of the PBB Incident” (no date). Program Guidance Manual, Compliance Pro- gram 7305.006-’” Radionuclides in Foods” (FY “78), Oct. 1, 1977. Chapter IV State Laws, Regulations, and Programs

. . ,=- Chapter IV

State Laws, Regulations, and Programs

STATE LAWS

State food and drug laws, and the organizations that administer them, vary widely. Basic State food and drug statutes are based on the Federal food laws; however, not all States have adopted the model uniform State food, drug, and cos- metic bill of the Association of Food and Drug Officials. As shown in figure 3, 42 States have adopted the model statute, which is almost identical to the 1938 Fed- eral Food Drug and Cosmetic Act. Consequently, these 42 States have the same legislative authority for regulating food contaminants within their borders as the Federal Government does for food in interstate commerce. The 1906 Act, still re- tained by eight States, does not contain the tolerance-setting provisions of the 1938 Act under which environmental contaminants are regulated (l).

Authority for regulating environmental Figure 3.— State Food Laws contaminants in food rests with two or more agencies in most States. Usually the Depart- ment of Agriculture and the Department of 50 Health share responsibilities for food regula- 42 tion. In some States a variety of other agen- r cies and bodies are also involved in regula- 40 — tory or research activities: departments deal- 34 ing with commerce, fish and game, consumer protection, environmental improvement, pub- 30 — In u 25 lic administration, conservation, along with % 2L z university divisions, independent laboratory 20 agencies, and various independent boards 20 — and commissions (2). The Food and Drug Administration (FDA) 10 — 8 believes that this variability in laws and reg- — ulatory organizations makes it more difficult for the States to accomplish their goals: o — 4 The variability of organizational structure FD&C Act 1906 f Enrichment complicates the problems of many of the in- f Food Pest Req dividual State agencies in accomplishing Uniform FD&C Act Additives I Amendment (Food ) Color their program goals because of overlapping Additives responsibilities and the lack of a clear delin- eation of responsibilities. For example, it is not uncommon to find authority granted to

two agencies for some divided program seg- SOURCE Food and Drug Admtnlstratlon State Programs and Serwces m Food ments of a single program category (e.g., and Drug Control, 1978

49 milk, shellfish). Frequently, two or more in- stances, the State agency has an unclear role dependent agencies of relatively equal rank as an advisor or consultant to the local gov- are charged with enforcement of portions of ernment. However, the local agency may not the same general food and drug law. In still be legally bound to follow the advice and/or other States, there is no central State control direction that may be suggested by the State over the food and drug program. In these in- agency (1).

STATE Monitoring PROGRAMS

Non-Federal monitoring is limited almost Table 8.—Program Areas and Expenditures for exclusively to programs originated and car- States ried out by individual States. In some cases Expenditures there is close coordination between State and Millions of Percent of total Federal activities. Federal agencies may de- Program dollars (food and drug) crease their monitoring in a given area if Food ‘...... $64.9 72.3 State monitoring is considered sufficient. For Drugs, devices, cosmetics . . . 9.7 11.0 Feed ...... 6.4 7.0 example, FDA does not monitor seafood for Weights and measures (food) . 5.5 6.0 kepone (3) even though the concentrations in Pesticides ... , ... , ...... 3.3 3.7 finfish, crabs, or shellfish remain essentially Total ...... $89.8 100.0 unchanged since its discovery in 1975 (4). SOURCE Food and Drug Admlnlstratlon S/ate Progrdms and Serv/ces /rJ food FDA feels that the ongoing monitoring pro- and Drug Contro/, 1978 grams in Virginia are sufficient to protect the figures do not include the estimated $75 mil- consumer. lion expended by local governments. Analysis of the OTA State survey reveals The educational levels of the chief chemist, that State food-monitoring laboratories are supervisory chemist, and chemist and labora- equipped to analyze for those substances that tory technicians are presented in table 9 (l). are regulated in foods through action levels The salary ranges for the chemical personnel or tolerances. For instance, most had instru- are shown in table 10 (l). Only 1 percent of mentation (atomic absorption spectrophotom- the chemists working in State food and drug eters and gas chromatography) to analyze for programs earned more than $20,000 per year mercury and chlorinated hydrocarbon pesti- in 1975 and 1976. Approximately 64 percent cides. of them had annual salary ranges of $8,000 to FDA published a study entitled “State Pro- $15,000. grams and Services in Food and Drug Con- The available analytical equipment and trol” in September 1978 (l). The publication physical facilities are listed in table 11 (l). provides a compilation of States’ analytical These data confirm the OTA survey findings, capabilities during the years 1975 and 1976, since gas chromatography and atomic ab- listing numbers and educational levels of sorption spectrophotometers rank first and chemical analysts, types of analytical equip- third, respectively, in numbers. A breakdown ment, and expenditures for food inspection. of food commodities analyzed, samples col- In 1974 and 1975, the States spent annual- lected, and analyses performed is given in ly about $64.9 million for food inspection and table 12 (l). The FDA document urges caution analytical activities. This amounted to 72 per- in interpreting these data because “. . . some cent of their total inspection and analytical States do not maintain comprehensive ana- expenditures (l). Table 8 shows program lytical records on food analyses especially if areas and expenditures for States. These food is not the major laboratory workload.” Ch. IV—State Laws, Regulations, and Programs ● 51

Table 9 .—Number of Employees by Category and Education Level in Various State Food and Drug Programs

Personnel ‘- Number of employees Advanced degree College degree Some college No college Chief chemist...... 109 68 ‘ - 40-- 1 0 Supervisory chemist 232 62 167 3 0 Chemist...... 633 66 561 6 0 Laboratory technicians . . 833 0 127 375 332

SOURCE Food-and-Drug Adminlstratlon Slate Programs and Serv/ces (n food and Drug Conlro/ 1978

Table 10.—Salary Ranges of Various Chemical Personnel in State Food and Drug Programs

Salary-ranges Number of chief chemists N u m ber of s u pervisory chemists Number of chemists $ 6,000- 9,000- ...... o 3 o $ 7,000-10,000 ...... 0 0 26 $ 8,000-12,000 ...... 0 0 70 $ 8.000-14,000 ...... 2 17 153 $10,000-15,000 ...... 2 143 139 $12,000 -18000...... 17 53 5 $16,000-20,000 ...... 71 7 0 over $20,000 ...... 8 0 0

SOURCE Food and Drug Admlnlstratlon- Sfate Programs and Sew/ces In Food arrd Drug Cor?tro/ 1978 -

Table 11 .—Physical Facilities and Key Equipment in This probably accounts for the extensive ana- 114a State Food and Drug Laboratories lytical activity reported in the “other food” category, rather than in the categorical pro- Number of key Key equipment items equipment itemsb gram areas”’ (l). Even so, it is evident that Spectrophotometers ...... - 269 milk and milk products are the most common- Flame photometers ...... 90 ly sampled and analyzed food commodities. Atomic absorption spectrophotometers . . 125 Spectrofluorimeters . 88 The large number of samples and analyses Gas chromatography . . . . 348 indicate that the States perform extensive Polarographs. ... . 42 Polarimeters . . . . . 54 monitoring for regulated contaminants in Mass spectrometer. ., . . 19 food, But the low salary ranges, the lack of so- Electrophoresis...... , . . 86 phisticated analytical equipment such as Auto analyzers. ., ...... 120 Liquid chromatography ...... , . . . . . 18 mass spectrometers, the time spent per anal- Physical facilities ysis, and the sample-type distribution indi- Average floor space (ft.2 ) . . . . . 10,245 cate that State monitoring programs are as Average bench space (, ft.)...... 845 z inadequate as Federal programs in detecting Average storage (ft. ) 1,150 environmental contaminants in food for aTwo States not reporting lab equipment or space which no action or tolerance levels have been bshows only total number of equipment Items reported by the s!ate a9encles Does not show those labs that do not have one or more of the equipment items established, listed SOURCF Food and Drug Admlnistratlon Stale Programs and Services In Food and Drug Con Ire/ 1978 52 ● Environmental Contaminants in Food

Table 12.—Total Number of Samples, Sample Determinations, and Man-Hours by Food Commodity Categories With Number and Percentage of States Reporting Analytical Activity

Number of States Number of Food commodity reporting analytical Number of samples determinations Number of man- categories activities Percent of States (thousands) (thousands) hours (thousands) Bakery products. ., . . . . . 35 70 11.5 23.3 31.5 Soft drinks ...... 35 70 18.7 64.6 17,5 Candy ...... 29 58 7.0 18.1 13.6 Grade A milk (raw) ...... 50 100 342.0 1,157,1 319.0 Other milk products. 47 94 200,7 546.8 185.6 Canned foods. . . 35 70 21.3 870 48.5 Frozen foods ...... 27 54 7.3 27.6 19.3 Seafood...... 34 68 6.6 34.4 33.2 Shellfish ...... 24 48 37.6 106.3 67.6 Raw agricultural products 29 58 40,1 89.6 112,7 Other foods...... 40 80 103,0 326.3 220.1 Totals 7958 2,481,1 1,068.6

SOURCE Food and Drug Adnllnistrallon S/ate Programs and Serv/ces In Food and Drug Control 1978

FEDERAL/STATE LIAISON Many environmental contamination inci- Such information is used by State and Fed- dents are initially State problems. Theoreti- eral authorities to: 1) determine the appropri- cally, the Federal Government does not be- ate Government response for protecting the come involved until a contamination incident public health and 2) inform the general public is determined to be an interstate problem. about the incident and explain the Govern- Given the complexity of this country’s food- ment response. To contain an incident, scien- marketing system, most food produced or tific information needs to be accurate and im- processed within a particular State is distrib- mediately available. This is as true for an epi- uted for consumption in other States. Thus, sode involving a substance that has previ- most environmental contamination incidents ously contaminated food (1979 PCB contami- are likely to become interstate concerns. nation of animal feed in Billings, Mont. ) as for Figure 4 reveals the extent of food contamina- a substance which has not (1973 polybromi- tion that can occur from a single source of nated biphenyl (PBB) contamination of animal contamination, in this instance polychlori- feed in Michigan). nated biphenyls (PCBs) contaminated animal feed from a meatpacking plant in Billings, The generation and dissemination of scien- Mont. (5). This widespread contamination of tific information on an incident is hindered by food occurred during an estimated time peri- the number of State and Federal agencies in- od of 2 to 5 months. volved. As already noted, three Federal agen- cies (the Environmental Protection Agency The States and the Federal Government es- (EPA), FDA, and the U.S. Department of Agri- tablish liaison when contamination crosses culture (USDA)), each with different respon- State lines. Liaison is also established at the sibilities, can be involved along with various request of States. States often require Fed- State agencies. The PBB incident in Michigan eral assistance in investigating a contamina- and the PCB incident in Montana reflect this tion incident. The objective is to generate sci- particular problem. entific information on the nature and extent of the contamination. This information would include the toxicological and chemical prop- Generating Information erties of the contaminating substance, the amount and type of food contaminated, and Before the needed scientific information the concentration of the substances in food. for developing regulations is generated, the Ch. IV—State Laws, Regulations and Programs ● 53

.L_- 4 I .54 . Environmental Contaminants in Food

contaminating substance in food must be authority for regulating contaminants pres- identified. This identification process may re- ent in food. Examples of actions taken by quire lengthy investigative work which could State and Federal authorities in such in- be hindered by a multiplicity of involved Fed- stances include removing food from the mar- eral and State agencies. The PBB incident ket, setting action levels or tolerances for the provides an example of the extensive investi- contaminant in food, and disposing of the con- gations sometimes necessary. taminated food. Because little scientific in- formation on PBBs was available, it took time The Michigan Department of Agriculture (MDA) initially analyzed blood and feed sam- to generate the information and establish ples from a herd owned by Mr. Frederic Hal- final permissible levels (although contami- nated herds and food were identified as bert (at his urging). These samples proved to quickly as possible and removed from the be negative because MDA was not analyzing market). The PBB incident involved FDA, for PBB. Additional samples were analyzed USDA, MDA, and MDPH. Even more State by the USDA’s National Animal Disease and Federal agencies were involved in the Center in Ames, Iowa. While the Disease PCB contamination episode in the Western Center detected an unusually high peak of an United States. unidentified substance, it did not determine its identity. The substance was eventually The PCB contamination began in Montana identified by Dr. George Fries of USDA’s with animal feed and quickly spread to 16 Agriculture Research Center at Beltsville, other States. Idaho was particularly affected Md., who had previous experience in chemi- by the contamination of poultry and eggs. cal analysis of PBB (6). The discovery came This incident involved all three relevant Fed- nearly 8 months after adverse symptoms oc- eral agencies— FDA, EPA, and USDA—as curred in the herd at the Halbert farm. well as the Idaho Departments of Agriculture The laboratories and agencies involved in and Health and Welfare, and district health identifying PBB are part of the agriculture departments. USDA made the initial analysis system in the United States. They were the of poultry samples which proved positive for obvious institutions to which Mr. Halbert PCB; FDA was involved with the removal of would go for assistance. If he had initially PCB-contaminated food from the market; and contacted the Michigan Department of Public EPA with the proper disposal of the contami- Health (MDPH), it is likely that different nated food. At least 5 days elapsed from the Federal laboratories and agencies would time USDA was confident it had a PCB con- have analyzed the samples. There was and tamination incident to the time it notified FDA still is no systematic procedure at the Federal of its findings. It took FDA an additional 5 level for assisting States in identifying a days to begin its investigation of the contami- potential food contamination problem like nation incident (7). Such delays would be PBB. Were it not for the perseverance of Mr. unlikely if only one Federal agency were in- Halbert and the experience of Dr. Fries, the volved or communications between the two identity of PBB might have taken longer and agencies were better. the people of Michigan would have been ex- PCB is a substance whose chemical and posed to PBB-contaminated food even longer. toxicological properties are fairly well Once a contaminant has been identified in understood. It has contaminated food in the food, the necessary scientific information can past. Nevertheless, there was confusion be generated for either regulating or control- among the State agencies in Idaho as to the ling the contamination. In the PBB incident, proper response to the contamination. The FDA helped to develop and evaluate this in- confusion resulted from two conditions. First, formation and worked with MDA in control- some of the State officials involved were not ling the incident. FDA’s involvement was familiar with the chemical and toxicological based on the fact that it has the Federal properties of PCB. PCB was a new food con- Ch. IV—State Laws, Regulations. and Programs ● 55

taminant in Idaho, and the appropriate of- ment of Health and Welfare, which is con- ficials had little experience with this type of cerned with protecting the public health, at problem. Second, the involvement of three dif- first was not informed by either USDA or the ferent Federal agencies obstructed efficient State Department of Agriculture of the PCB communication between the State agencies contamination. Communication broke down and the Federal Government. at two levels, between the State and Federal While the Federal agencies had the most Governments and within the State govern- ment (8). The fact that there are several dif- expertise on PCBs, the sharing of that ex- ferent Federal and State agencies involved perience was hindered by the fact that it was with different aspects of controlling and available in and distributed by three sources regulating a contamination incident further instead of one. Consequently, State agencies had to go to different Federal sources, de- complicates an already complicated problem. pending on what information they needed. In The major environmental contamination in- addition, the Federal agencies did not always cidents that occurred in Idaho and Michigan communicate with the various State agencies. continue to be major issues of concern among EPA, for example, took air and water samples the residents of these States--a result of in the area surrounding Ritewood Farms to their fears over a potential health threat that determine whether PCBs that contaminated cannot be seen, smelled. or tasted. In Michi- Ritewood’s eggs and poultry came from either gan, for instance, the PBB episode remains a of these two sources. This was before the live and controversial political issue. Conse- PCB-contaminated animal feed was identi- quently, it becomes imperative that the in- fied. EPA, however. did not report their nega- formation generated by the State and/or Fed- tive findings to the State. In another instance, eral Government on an incident is accurate USDA initially would report its results only to and appropriately applied. This objective is the Idaho Department of Agriculture, the hindered by the variety of State and Federal State agency with which USDA has had a agencies that become involved. long-standing association. The Idaho Depart-

CHAPTER IV REFERENCES

1. U.S. Department of Health. Education, and fore Subcommittee on Oversight and Investi- Welfare, Food and Drug Administration. gation, House Committee on Interstate and State Programs and Services in Food and Foreign Commerce, Sept. 28, 1979. Drug ~ontrol, 1978. 6. Michigan Department of Agriculture, Infor- 2. OTA Survey of 31 States, 1978. mation and Education Division. ‘‘A Brief 3. Food and Drug Administration. Compliance Chronology oft he PBB Incident ,“ (no dat e). Program Evaluation (FY ‘77), Kepone and 7. Foreman, Carol Tucker, Assistant Secretary hlirex Contamination ( 7320. 79A). for Food and Consumer Services, U.S. Depart- 4. Huggett, R. J., hi. hf. Nichols, and M. E. ment of Agriculture. Testimony before the Binder. “Kepone Contamination of the James Subcommittee on Oversight and Investiga- River Estuary, ” Proceedings of Symposium on tion, House Committee on Interstate and For- Con tc~minan ts and Sediments: Fete and eign Commerce, Sept. 28, 1979. Transport. R. A. Baker (cd.), Honolulu, Ha- 8. Gallagher, Ed, State Health Officer, Idaho De- waii. April 1979 (in press). partment of Health and Welfare, personal 5. Gardner, Sherwin, Acting Commissioner of communication, Oct. 10, 1979. Food and Drug Administration. Testimony be- Chapter V

Methods for Assessing HeaIth Risks Chapter V

Methods for Assessing Health Risks

The availability and quality of data on the toxicity of contaminants in large part determine the Food and Drug Administration’s (FDA) ability to protect public health. Chapter III reviewed the strengths and weaknesses of FDA’s procedures for setting action levels and tolerances. This chapter addresses four issues re- lated to the methods used for assessing health risks: ● What is the role of human epidemiology in the setting of action levels and tolerances? ● What are the roles of animal and other toxicological tests? ● Are current testing procedures adequate for the assessment of interac- tions of toxic substances? ● Are methods for quantitative risk assessment sufficiently developed for application to environmental contaminants in food?

POSSIBLE TOXIC EFFECTS The possible toxic effects of an environ- tations (which generally affect a single gene) mental contaminant depend on its chemical or more extensive effects such as gross nature, its concentration in food, and the type changes in chromosome structure or changes of toxic action involved. If the substance is in chromosome number. Such genetic effects highly toxic and/or is consumed in large quan- in humans often cannot be immediately de- tity, acute toxic effects may occur and the tected. Indeed, damage to the human gene onset of the symptoms would be rapid and pool may not become apparent for many gen- severe. If a small amount of a highly toxic erations if the deleterious effect is due to a re- substance is consumed, or if the substance is cessive gene (2). However, a chemical’s muta- of low toxicity, the health effects may not be genic potential can be evaluated indirectly seen for many months or years (or may not be from various biological tests involving micro- observed at all). Potential effects of toxic en- organisms, mammalian cell cultures, insects, vironmental contaminants in food include sys- and whole mammals (l). temic toxicity, mutations, cancer, birth de- fects, and reproductive disorders. Some environmental contaminants in food may cause cancer. Direct cause-and-effect Systemic toxicity involves changes in the associations between exposure to a specific structure and function of organs and organ chemical and human cancer are difficult to systems: weight change, structural altera- establish because of the complex nature of tions, and changes in organ system or whole cancer and the vast number of potential car- animal function. Functional effects may in- cinogens to which humans are exposed. In clude changes in the lungs, liver, and kidneys, some cases, exposure to one toxic agent may in cardiovascular function, in brain and nerv- trigger a sequence of events leading to ous system activity and behavior, and in re- cancer. In others, carcinogenesis may depend sistance to disease (1). Systemic toxicity is on interactions of several factors, combina- studied in whole animal tests. tions of noxious agents, co-factors, and Some environmental contaminants have natural or acquired metabolic peculiarities been shown in experiments to cause point mu- (2). The cancer-causing potential of a sub-

59 stance is evaluated through animal tests and Environmental contaminants can have a variety of short-term tests employing bacte- other effects on the reproductive process in ria, insects, or cell cultures (1). both males and females. Toxic effects may in- clude reduced or altered sperm formation, in- Environmental contaminants may produce hibition of ovulation, increased spontaneous birth defects other than inherited mutations. abortion, fetal resorption, and increased These are called “teratogenic effects. ” Mal- numbers of stillborn infants, Teratogenic and formations may include gross, histological, reproductive effects are evaluated through molecular, and sometimes behavioral anoma- animal tests (l). lies. Human sensitivity to a teratogenic agent during pregnancy is determined by: 1) the Given the range of possible adverse health time at which the insult is received during impacts, it is clear that newly discovered en- gestation, 2) the dose of the compound, vironmental contaminants must be subjected 3) transfer of the teratogen to the developing to the best available toxicological testing fetus, 4) uterine factors, and 5) dietary fac- techniques so that any harmful effects can be tors. The teratogen may work its effect by uncovered. Furthermore, regulators must producing mutations or chromosomal aberra- have information on the possible toxic effects tions, interfering with cell division, altering of ingesting small amounts of a substance in nucleic acid synthesis, inhibiting specific food over an extended period of time, per- enzymes, or altering membrane characteris- haps over a lifetime. It would also be desi- tics. A seemingly small change may have far- rable to know what effects other toxic sub- -reaching effects, since the fetus is undergoing stances already present in our air, water, rapid biochemical, structural, and physiologi- and food may have on the metabolism and cal growth and change (1). toxicity of a new contaminant,

HUMAN EPIDEMIOLOGY

The science of epidemiology seeks to deter- can also be used to directly estimate human mine the distribution and causes of diseases risk, thus eliminating the need for extrapolat- and injury in humans, It focuses on groups ing from animal experiments. rather than individuals (3). Epidemiological evidence has sometimes There are several types of epidemiological weighed heavily in the setting of tolerances or studies. Each provides different levels of in- action levels. This was the case in FDA’s deci- formation on environmentally related disease sions on mercury and polychlorinated biphen- or injury. The cohort-study method is the best yls (PCBs). However, if epidemiological data way to develop information on potential toxic are lacking when an environmental contami- nant is discovered in food, the usefulness of effects in a population exposed to an environ- further epidemiological studies is restricted mental contaminant in food. In a cohort study, individuals are classified into groups by inherent limitations of the science. according to the levels of exposure, including Beyond the obvious ethical considerations, a control group with no previous exposure to such studies are of little use when a rapid the suspect substance. The groups are then regulatory decision must be made. At the low followed over a period of time and studied for levels that environmental contaminants usu- differences in disease incidence (3). Such epi- ally occur in food, pathological effects may demiological studies have an advantage over not occur for many years, Most cancers, for animal tests. There are some agents known to example, have a latency period of at least 5 to cause disease in humans that do not produce 10 years, As much as 40 years may elapse be- similar adverse effects in animals (, tween the onset of exposure and the develop- for example) (4). Human epidemiological data ment of the disease, Thus epidemiological Ch. V—Methods for Assessing Health Risks ● 61

methods can only be used retrospectively to statisticians. This would double or triple the evaluate the health status of individuals who costs (5). Furthermore, demonstrating that have been through this 5- to 40-year period. the incidence had increased would still leave The findings are compared to the health his- the vexing problem of correlating the in- tories of “control” individuals who presum- crease with the exposure to the specific ably have not experienced comparable ex- suspect substance. posures. Such limitations make cohort epidemiology Such retrospective studies are vulnerable studies inappropriate for determining the tox- to scientific criticism: a) it is very hard to find ic potential of a given contaminant. More- an adequate control group (unlike animals over, in a regulatory scheme designed to pre- whose entire lives can be in a “controlled” vent serious illness from developing in hu- environment) because of the variability in mans, the time required to generate epidemi- human susceptibility and personal behavior ological data would preclude its use in initial patterns, b) it is impossible to define potential regulatory decisions. The prospects for devel- synergisms and/or inhibitions by other sub- oping a human epidemiological method that stances to which the target group may have would meet such regulatory demands are been exposed, and c) it is difficult to quantify presently hard to imagine. previous levels of exposure (5). One area, however, where human epide- miological studies have and will continue to Retrospective studies suffer other handi- be useful is in the confirmation of suspected caps, They are extremely expensive, difficult, chronic-effect data when a population that and time-consuming to carry out, A given type has inadvertently been exposed to a sub- of cancer, for example, may occur in the stance or a hazard (such as radiation) over a “normal” population at a rate of 1 case for long period of time can be identified. In this each 100,000 people. An exposure to a toxic instance, carefully designed studies can substance that increases by a factor of 100 be the “clincher’” which finally provides the likelihood of that cancer occurring would thorough documentation of the suspected tox- raise the incidence to 1 case for each 1,000 icity. Such studies can confirm or refute the people. To ascertain in a statistically satis- adequacy of regulatory actions (in retrospect) factory and compelling way that such an in- (5), Again, however, these kinds of studies are crease had occurred, one would have to do extremely expensive and consume limited re- comprehensive medical studies on at least sources which might be better spent support- 3,000 people. Assuming an average cost of ing other types of evaluations using other $200 for each person studied (which is a low techniques. estimate considering the costs of physicians” time, laboratory analysis, and the technologi- People who fail to understand these limita- cal assessments necessary to show a person tions criticize reliance on animal tests when free of a given tumor or disease), the expense epidemiological information is negative or for such a study would be $600,000. It would lacking. This fundamental misunderstanding be necessary under most circumstances to may delay health-related regulatory action find more than one new case of a disease to and dilute its effectiveness in preventing convince a nonstatistician and even most illness. 62 . Environmental Contaminants in Food

ANIMAL TESTS AND OTHER METHODS FOR TOXICOLOGICAL TESTING

Toxicologists have developed a number of carefully selected battery of 90-day tests, techniques to assess the toxicity of a com- considerable data can be generated on a con- pound. Some tests can be conducted in 90 taminant’s potential toxic effects including its days or less. These include simple tests such potential for causing mutations and cancer. to determine the as 2-hour “range finding” Despite widespread scientific agreement dose of a substance that is lethal to 50 per- on the usefulness of animal tests, several con- cent of the animals (LD50). More complex troversies exist. These include the appropri- tests include 90-day continuous exposure or ateness of the carcinogen bioassay tech- paired feeding studies, and short-term tests niques, the potential of short-term tests for for mutagenesis and potential carcinogene- evaluating mutagenesis and carcinogenesis, sis. Tests requiring more than 90 days, such the lack of emphasis on testing for potential as lifetime exposure studies, are generally toxic interactions, and the usefulness of considered long-term. In addition to the time methods for extrapolating from high-dose ani- necessary for exposure and data-gathering, mal test results to low doses in humans. analysis of the results may take up to an addi- tional year, depending on the complexity of The Carcinogen Bioassay the experiment, the number of animals used, and the amount of data collected (l). The carcinogen bioassay (6) is a chronic Testing methods can be categorized by toxicity study requiring 2 years’ exposure to ‘‘endpoint” as well as duration. Some ex- the test substance and usually a third year periments are based on expected results such for evaluation of the results. The carcinogen as functional changes, birth defects, or can- bioassay is used to determine the cancer- cer, By the use of an appropriate experi- causing potential of a compound in males and mental design, several such endpoints can be females of two mammalian species, usually assessed in the same experimental period (l). the rat and mouse. The test animals are ex- Appendix C describes the range of toxicologi- posed to the test substance throughout their cal tests available for assessing the health entire lifespan. risks associated with consumption of contam- Different groups are exposed to different inated food. Each section describes the tests levels of the substance up to the maximum for assessing a given endpoint, and includes a tolerated dose (MTD). The MTD is the highest discussion of available long- and short-term dose given during a chronic study that predic- tests, tively will not alter the animals’ normal Animal tests provide valuable information longevity from effects other than cancer. In for the setting of action levels and tolerances. practice, the MTD is considered to be the The animals serve as proxies for humans in highest dose that causes no more than a 10- cases where data are lacking on the human percent loss in weight compared to control health effects of a contaminant and experi- animals. Throughout the study, animals are ments involving humans would be unethical, examined weekly for signs of toxicity. Ani- Because of shorter animal lifespans, the ef- mals may be killed at prearranged times and fects on several generations can be studied. their tissues examined for signs of cancer. At Furthermore, animals can be raised in con- the completion of the study, all remaining trolled environments, thus eliminating many animals are killed and detailed examinations of the factors that complicate human epide- are made of their tissues. miological studies, In many cases, animal Animal tests for carcinogenicity have been experiments are the only means by which questioned because the results are general- needed information can be obtained. Using a ized to humans while the animals are fed Ch. V—Methods for Assessing Health Risks ● 63 much larger doses of the suspected substance also to detect the serious hazard posed by mu- than would be consumed by people. However, tagens. In the long run, chemical compounds an earlier OTA report entitled Cancer Test- causing mutations and teratogenic effects ing Technology and (7) found that may create a greater burden on society by in- “animal tests are the best current methods creasing the incidence of genetic disease and for predicting the carcinogenic effect of sub- birth defects than by causing cancer (2,11). stances in humans. All substances demon- A number of short-term tests and batteries strated to be carcinogenic in animals are re- of tests have been developed to assay chem- garded as potential human carcinogens; no icals for mutagenicity (1,2). Many of the same clear distinctions exist between those that tests are also viewed as screens for carcino- cause cancer in laboratory animals and those gens on the assumption that cancer arises that cause it in humans. The empirical ev- following damage to the genetic material of idence overwhelmingly supports this cells (2). The ability of these tests to detect hypothesis. ” carcinogens is currently undergoing valida- The report also found that the rationale for tion (8). The Environmental Protection Agen- feeding high doses was sound. cy (EPA) currently uses some of these tests to identify mutagens (9, 10). Presently there is no The rationale for feeding large doses of a on a “best” set of tests. substance in animal tests is as follows. As universal agreement the dose of a substance that causes cancer is Environmental contaminants in food are increased, the number of exposed animals screened to identify those capable of causing that develop cancer also increases. To con- genetic damage, to determine the types of duct a valid experiment at high dose levels, only a small number of animals (perhaps sev- damage they cause, and to evaluate the risks eral hundred) is required. However, to con- they pose to the general population and cer- duct a valid experiment at low dose levels, a tain subpopulations. very large number of animals is required. The major difficulty in designing tests to (The smallest incidence rate detectable with detect mutagens and potential carcinogens is 10 animals is 10 percent or 1 animal. To detect a l-percent incidence rate, several the need to reflect the metabolic capabilities hundred animals would be required.) An- of man as nearly as possible while remaining other important variable is the strength of economically realistic. Whole animal tests the carcinogen. The stronger the carcinogen, more nearly reflect responses of man and are the greater will be the number of animals therefore useful in estimating risk to humans. getting cancer at a particular dose (7). Tests using lower organisms are less directly applicable to humans, but they are simpler If data from a carcinogen bioassay are and more economical to perform. As a conse- available at the time an environmental con- quence they are more popular for use in mu- taminant is discovered in food, the results tagen screening than animal tests. Risk esti- provide crucial information to guide the mates based on tests in lower organisms have regulatory decision. However, if no data or in- been proposed (12) but are not accepted as adequate data exist, a newly commissioned sufficient evidence for setting action levels carcinogen bioassay would require 2 to 3 and tolerances. years for completion. In this case, the car- cinogen bioassay could be used to revise an Many chemicals are not directly muta- initial action level. genic, but once ingested or absorbed are con- verted into mutagenically active derivatives. Short-Term Tests for Mutagenesis Activation therefore poses important practi- and potential Carcinogenesis cal and conceptual problems that must be dealt with in evaluating the mutagenic poten- Chemicals are evaluated for mutagenicity tial of environmental agents. The practical not only to detect potential carcinogens but problem is one of designing tests to identify 64 ● Environmental Contaminants in Food mutagens that adequately mimic the metabol- Several procedures that can be carried out ic capabilities of an intact animal. The con- directly in exposed human populations have ceptual problems focus on the fact that evi- been found to be indicators of the presence of dence of an agent’s mutagenicity is not suffi- mutagenic agents. These procedures include cient to evaluate the actual risk posed to an analysis of white blood cells for chromosome individual or a population by exposure to the aberrations (cytogenetic analysis) (13), anal- agent (2). ysis of blood and/or urine of the presence of mutagenic agents (14), analysis of semen for The ability of an agent to cause genetic abnormal sperm (15), and the detection of ex- damage in an intact animal depends not only tra chromosomes in sperm (16). on the mutagenic potential of the agent or its metabolizes but on a number of other factors Intact animal tests usually use as test as well. The fate of any chemical substance animals either mice or fruit flies [Drosophila entering the body is determined by a dynamic melanogaster). The dominant lethal test de- process involving its absorption, distribution, tects lethal damage resulting primarily from biochemical alteration, and excretion. Many chromosomal aberrations by treating male mutagenic agents are formed by the action of mice with test agents and mating them. The intestinal bacteria. Consequently, the route of uteri of the pregnant females are examined exposure to an agent can play a significant for fetal death and resorption. Because the role in determining its activity. Enzymatic endpoint is detected in the offspring of the processes in the body inactivate as well as treated animal, mutation in the germinal cells create mutagens. Many active mutagens may (the sperm) is detected by this method. The exist in the body as intermediates in the specific locus test involves the crossing of two metabolic pathways that process the parent genetically distinct strains of mice with visi- compounds. As a result, the active mutagen ble markers, such as spotted coats. Mutations may have a short lifespan and be distributed are detected by the appearance of recessive only in those tissues that possess high levels traits in the offspring. The utility of this testis of activating enzymes (2). restricted because of the number of animals needed. Cytogenetic screening of white blood There are four major factors determining cells or bone marrow cells from treated the ability of an environmental agent to pro- animals detects damage to chromosomes. duce a genetically adverse effect in an orga- The micronucleus test detects chromosome nism, The first is the metabolic response of breaks in bone marrow cells of treated ani- the organism to the agent. This response will mals (17). determine the distribution, lifespan, and fate of the mutagen in the body tissue. The second The use of cultured mammalian cells for is whether the mutagen damages the genetic mutagenesis testing is a great simplification material in cells. The third is the response to over the use of intact animals. At the same the genetic damage by the DNA repair ma- time, the cells retain much of the mamma- chinery in the cell. Fourth is the type of lesion lian metabolism and genetic characteristics. the agent is capable of producing in the genet- Moreover, human cell cultures can be used ic material. where exposure of people to toxic agents Short-term tests for detection of muta- could not be allowed. The major weakness of genesis and potential carcinogenesis fall into cultured mammalian cells as test systems is four categories: that they are isolated from the metabolic ac- tivity of the intact animal (2). ● procedures that can be carried out di- rectly in human populations, Several strategies have been developed to ● intact animal tests, try to approximate the metabolic capabilities ● tests employing cultured mammalian of the intact animal in mammalian cells, The cells, and simplest and most widely used activation sys- Q. tests employing micro-organisms. tem employs homogenates of mammalian Ch. V—Methods for Assessing Health Risks ● 65 liver. Either the liver homogenate is incu- The great appeal of these techniques is that bated with the test substance before it is ap- they are rapid, simple, and inexpensive. No plied to the cell culture, or the test substance animal colony is required, and the necessary and the homogenate are included in the test technical skills are those required for conven- plates. A major difficulty with the liver micro- tional bacteriology. some activation system is that it cannot be These tests have been criticized because prepared with reproducible enzyme levels the cell structure of bacteria is very different from one batch to the next. Storage condi- from the cell structure of mammals. Valida- tions also alter the activities of the prepara- tion studies now underway are attempting to tions, Furthermore, not all the activation evaluate 30 mutagenicity assay systems for mechanisms present in the intact animal are their ability to predict chemical carcinogeni- represented in the liver homogenate. The city (8). greatest advantage of the liver homogenate method is that it preserves the simplicity of in Bacteria also lack the enzyme systems that vitro testing (18). are the principal mammalian mutagen-acti- vating enzymes. This criticism is partially Methods for mutagen screening have been overcome by the use of liver homogenates de- developed using many micro-organisms, in- scribed above. Finally, each test is sensitive cluding fungi, bacteria, and viruses present to one particular type of genetic damage. in bacteria. The most widely known system is Therefore, the best approach is to use a bat- the “Ames test” which employs specially tery of short-term tests designed to test for bred strains of bacteria. Figure 5 the different types of genetic damage. illustrates the steps in the Ames spot test. The Ames plate test allows quantitative compari- Interactions of Two sons to be made of mutagenic potential (19). or More Substances Numerous other tests have been devised which are in varying stages of validation (20). Tests designed to detect effects from inter- Some of these are described in appendix C. actions of two or more substances are of lim-

Figure 5. —Steps in the Ames Spot Test

Suspected carcinogen on filter paper Rat-liver Tester 2 extract bacteria

MUTAGENESIS is detected in the Ames test by mixing an assay” a dose of the chemical to be tested is placed on a extract of rat liver (which supplies mammalian metabolic disk of filter paper on the tester bacteria. After 2 days most functions) with tester bacteria (which cannot grow because of the his- bacteria have died for lack of histidine (2), but a mutation makes them unable to manufacture histidine, a DNA damage caused by the chemical diffusing out from necessary nutrient) and plating the mixture on an agar the disk has given rise to mutations, some of which result medium so that a thin layer of bacteria covers the medium in reversion of the his- mutation. The histidine-making evenly, as is shown on a microscopic scale (1). In this “spot revertant bacteria proliferate, forming visible colonies (3).

SOURCE Redraw n v, II h perm[sslon Devoret R Bacterldl Tests for Polen!lal Carcinogens Scfentlffc Arner/can 241 (2} 40 August 1979 ited use for regulatory actions. This does not or by acting differently on different organs to mean that such information is not important. produce the same effect. Most of the interac- It rather reflects the rudimentary nature of tions are at the biochemical level and the our present understanding of interactions, mechanisms are still being studied. the complexities of the tests, and the difficul- The key consideration is whether present ties in interpreting the results in a fashion testing technologies are adequate to provide meaningful to regulators. data that are useful in making regulatory Six different types of interactions may oc- decisions. Techniques have been developed to cur when two or more chemicals interact in a study the interactions of drugs, other chemi- biological system. The effects produced may cals, and physical agents such as radiation. differ in magnitude from those caused by any Such techniques should be applicable to the one of the chemicals alone. study of the effects of environmental contam-

● inants. However, their scope is usually lim- If the interaction of two substances pro- ited to the study of two component interac- duces an effect equal to the sum of the tions because of the large number of test ani- individual effects, the response is called mals required and the difficulties in inter- summation or addition. ● preting the results. For example, a relatively If the interaction of two substances pro- simple test involving two agents in a factorial duces an effect greater than would be design would require about 100 animals. A anticipated from the sum of the individ- more recent study (22) of the combined ef- ual effects, the response is known as po- fects of cadmium, mercury, and lead used 50 tentiation or synergism. ● to 60 animals in each of 15 different treat- If the effect is less than the sum of the ment groups. The results indicated that a par- two would predict, the response is an- ticular combination of the three metals could tagonism. ● be synergistic, antagonistic, or additive, de- When an inert substance, having no pending on the relative doses employed (table observed effect at a given dose, en- 13). hances the effect of another simultane- ously administered chemically, the ef- fect is usually referred to as activation. ● If an inert substance decreases the ef- fect of another chemical administered si- multaneously, the effect is called inhibi- tion. ● Finally, there may be no interaction and each chemical would exert its own effect independently of the other (2 1). Once ingested, a chemical may exert an ac- tion locally in or on the stomach or intestines. While such effects possibly could be pro- duced by environmental contaminants, they are not likely to be observed at the low levels normally encountered in food. The more ser- ious concern is the systemic effects that may occur after absorption from the gastrointes- tinal tract. Following absorption from the gastrointes- tinal tract, additive chemicals may produce their effect by acting on a target organ, by acting on different target organs or systems, Ch. V— Methods for Assessing Health Risks ● 67

The testing for interactions among toxic in the number of extra cancers that would ac- substances is further complicated by the ne- company some projected increase or de- cessity to limit the number of substances crease in the level of human exposure occur- studied, Because of the large number of ani- ring either as a result of regulatory action or mals required, and the difficulties in inter- inaction, The technique is therefore used to preting results, the number of substances calculate benefits to be derived from re- tested are confined to those that may be most ducing human exposure to a substance. important, Deciding what is ‘‘important” to The typical carcinogen bioassay uses study out of the vast number of toxic sub- around 100 animals at each experimental stances to which we are exposed is a difficult dose. If a particular experimental dose pro- problem, Perhaps a reasonable approach duces evidence of a lifetime increase in would be to focus on those to which people cancer of 1 in 10, this increase can be meas- are known to be exposed (see chapter VII). ured using 100 animals. But if the increased The choice could also be based on structure- cancer risk is less than 1 in 100, this increase activity relationship or known mechanism of will often not be detected even with a 100- action, Chemical selection for testing is being animal feeding experiment. The extra human reviewed and evaluated in an ongoing OTA risk resulting from environmental exposure is study entitled “Assessment of Technologies usually much smaller than 1 in 100 (perhaps for Determining the Carcinogenic Risk from on the order of 1 in 100,000) for any given the Environ merit.” chemical over a lifetime exposure. It would The study of the interactions of two or not be practical to conduct an experiment more substances is an area in which far more with enough animals to directly measure this basic research is required before meaningful small an increase in risk (30). information can be generated for regulatory decisions. At the present time, no satisfactory For these reasons, carcinogen bioassays methods exist for testing the interactions of use (in addition to a control dose of zero) more than two chemicals. several doses at which the projected extra cancer incidence may be 1 in 10 or larger. Extrapolating From High Doses These high-dose data are then used to esti- mate the extra risk at a very low dose where tO Low Doses the extra risk may be no larger than 1 in Quantitative risk assessment. based on 1,000,000. These problems are often referred mathematical models, is often proposed as an to collectively as the “low close extrapolation alternative to the current use of safety fac- problem” (30). tors (chapter 111) (24,25). This approach is A low-dose extrapolation involves the now most extensively employed in assessing choice of a mathematical function to model carcinogenic risks (26). The technique is also the dose-response relationship and the choice being investigated as a means to evaluate of statistical procedures to apply to the math- other types of irreversible toxic effects (23). ematical function (24-29). The choice of math- The following discussion of mathematical ex- ematical function is crucial to the outcome of trapolation of risk involves only its uses in de- a low-dose risk estimate. If the assumed rela- termining risks from chronic low-level carcin- tionship between tumor occurrence and dose ogenic insults. does not apply in the low-dose regions to Mathematical models can be used to esti- which the extrapolation is being made, a seri- mate the number of extra cancers that are ous overestimate or underestimate of the likely to be caused by the presence of a car- “safe dose’” may result. Even though different cinogen in the food supply. Models may be dose-response models may agree in the ob- used to estimate a tolerance based on calcu- servable response range, they could differ by lations of a “safe dose’” for human consump- many orders of magnitude at low-dose levels. tion. They can also forecast the likely change Furthermore, there is no experimental way to 68 . Environmental Contaminants in Food

predict the shape of the dose-response curve response curve can be predicted with cer- at very low doses, tainty, As pointed out earlier, experiments of sufficient size to permit direct experimental Because of the possible disparity of dose- investigations of the dose-response curve at response functions when extrapolated to low low dose cannot be conducted. There are doses, the dose-response function must be se- plausible arguments that the dose-response lected not only on the basis of how well it can curve is linear at low dose for many carcino- be made to fit experimental data but also on gens, On the other hand, no mechanism that the basis of known (or at least plausible) in- would lead to a more cautious dose-response formation on the biological mechanisms relationship has been seriously proposed ex- through which a chemical induces or pro- cept for some dose-response relations in motes cancer. This is a major source of un- radiation-induced carcinogenesis. certainty in extrapolation procedures, since the exact biological mechanisms through In view of these uncertainties, it would which carcinogenesis may occur are un- seem reasonable to base estimates of added known (30). cancer risk on a mathematical model that en- compasses low-dose linearity until the mecha- Several theories on the nature of the proc- nisms through which the carcinogen operates ess by which cancer develops serve as the are understood sufficiently to conclusively bases for different low-dose extrapolation rule out low-dose linearity, models. Some “linear” models project that the relationship between the dose received Caution must be used in interpreting the and the cancer risk is a straight line at low results of low-dose extrapolations, partic- doses and that there is no threshold. Other ularly when they are applied to humans, models level off at low doses and therefore Table 14 demonstrates that different models predict that a threshold dose exists below applied to the same data sets yield differing which the agent has no carcinogenic effect estimates of virtually safe doses. Virtually (24-30). safe doses based on the multistage model are identical with those based on the one-hit Newer models take into account the effects model when there is only one experimental of metabolic activation and detoxification dose. This is illustrated with Data Sets I and upon carcinogenic dose response (26), These “kinetic” models encompass free toxic sub- stances, metabolizes, deactivators, and the Table 14.—Virtually Safe Doses Computed possible interactions of these substances, From Three Different Data Sets and Only a “steady-state” situation is studied in Three Different Models that the variation over time of the concentra- Virtually safe doses (lower tions of these agents is not considered. If Dose- 95-percent confidence bounds deactivation of the carcinogen is 100-percent Data response for dose) in ppb corresponding efficient, the model predicts a threshold dose set model to extra risk of 1/1 ,000,000 below which there is no carcinogenic risk. la ... Probit 14.2 However, in a naturally occurring process, it is likely that deactivation would be less than 100-percent effective. A number of modifica- tions to the model allow for imperfect deac- tivation. These situations rule out a threshold and would lead directly to a model for which carcinogenic response varies linearly with dose at low dose, The mechanisms through which most car- cinogens produce cancer are not sufficiently SOURCE Adapted from K S Crurnp and M D Masterman Assessment of Car understood so that the shape of the dose- c Inoqenlc R(sk From PC Bs In Food OTA Working Paper 1979 Ch. V—Methods for Assessing Health Risks ● 69

II. The multistage model also yields virtually Differences also arise when varying ani- the same value as the one-hit model for Data mal data serve as the basis for the risk as- Set III. However, as explained in appendix D, sessment, And, of course, the choice of the the multistage model can yield higher values model also affects the final outcome. In this for the virtually safe dose than the one-hit case, both extrapolations were based on mil- model whenever the data exhibit upward cur- ligrams of PCB per kilogram of body weight vature and are inconsistent with the one-hit per day. Both FDA and OTA applied a linear model (30). model: however, different assumptions were made on the amount of PCBs that would be in- Caution also must be exercised when com- gested and on the size of the exposed popula- paring calculations of extra lifetime risks. tion. FDA’s risk assessment therefore applies Table 15 summarizes the results of an FDA to the 15 percent of the total U.S. population risk assessment of PCBs (31) with a similar expected to consume the fish species known risk assessment commissioned by OTA (30). to be most highly contaminated with PCBs. Differences can arise in such calculations The OTA calculations are based on the aver- when different methods are used to extrapo- age daily intake of PCBs based on FDA's total late from animals to man. The extrapolation can be performed on the basis of milligrams estimates of the PCB intake per kilogram of body weight per day, parts consumed more than 24 lbs of per million of substance in the diet, milli- sport fish per year. grams of substance per square centimeter of body surface area per day, or milligrams of substance per kilogram of body weight per Although many different lifetime. models exist and can be used for low-dose risk assessment, it appears prudent at this Table 15. — Extra Lifetime Risks of Cancer Associated With Consumption of PCBs in Food time to use a model that does not arbitrarily rule out low-dose linearity. Models are still Extra lifetime Upper limit of new being developed and when” appropriate could Dose (g/day) risk/100,000 cancers/year be used to guide regulatory decisions. Dose- FDA’ response models for low-dose risk assessment 9 2’ 44 21 14 9’ 72 34 provide a useful technique for assessing the 201 [ 98 47 possible added risk attributable to environ- b OTA mental contaminants. Such models also might 3 3(! 013 4 035 11 well be used in place of safety factors in 127f 5 future procedures adopted for food contami- ‘]BaCI,j r,n NC I birldssd 7 io j tc. td mallqn,+n Ifs fr,! mdlf,: .in{j f[malfs nant regulations. The major impediment to : Ba+fI! ,rl K !mh rrju qh ~ 1 q 751 S(I jff Y and hppa tor~lrular care {ncm]ds ‘ Ba\Ifj c, n h [q h f~ f [ or SII MC rs ?f!lh ~If rc c r] t I If) of f I sh ~ per I(,S con tam I n ,itr?{j the widespread use of the models in environ- * I t h PC. F3< I f trd(. rar> r c I,> I dt I I ~ h~, t j ,it 1 2 r r ~ [~ [ r.) [j Ba>e(j , ,r, FDA T r,ld I D IPt St (1 j Y 1 q ?(J mental food contamination considerations is ~’ Ba\P j ‘n F E! A T 01+ I D I( t 5 I hIj ~ 1 Q 75 fBa~{, cj ,, ~1 ~+(.r~qf In tak ~ c f rf’flrlf ‘r n., II m IrTq mr)re than 24 I hs t par Lak c the lack of data on which to perform the anal- MlchiQan fish ysis.

CHAPTER V REFERENCES

1. SRI International. “Assessment of h4ethods Company, Boston, Mass., 1970. for Regulatin~ ‘Unavoidable” Contaminants in 4. Ray, V. “Are Benzene Effects Limited to the the E’ood Supply, ” OTA Working Paper, 1978. Chromosomal Level?, ” Banbury Report 1— 2, Legator, hf. S. “Short ‘1’erm Procedures for Assessing Chemical A4utqgens: The Risk t~~ Determining Nlut:ig~;niclCar cinogenic Activ- Humans, Cold Spring Harbor Laboratory, ist},’” OTA ~lrorking Paper, 1979. 1979. 3. Llachfahon, B., and T. F. Pugh. Epidemiolog}r 5. ]ackson, R, S. “The Limitations of Applying Principles (]nfl Mcthmls, Little, Brow-n and Iluman Epidemiology to the Regulator Pro- 70 ● Environmental Contaminants in Food

cess for Toxic Substances, Particularly Car- Sciences 70:2281, 1973. cinogens,” OTA Working Paper, 1979. 19. Ames, B. N. “A Bacterial System for Detect- 6. Sontag, J. M., N. P. Page, and U. Saffiotti. ing Mutagens and Carcinogens, ” Mutagenic Guide~ines for Carcinogen Bioassay in Sma~l Effects of Environmental Contaminants, H. E. Rodents, DHEW Publication No. (NIH) 76-801, Sutton and V. I. Harris (eds. ), Academic Government Printing Office, Washington, Press, New York, 1972. D. C., 1976. 20. Devoret, R. “Bacterial Tests for Potential 7, Office of Technology Assessment. Cancer Carcinogens, ” Scientific American 241(2):40, Testing Technology and Saccharin, Govern- 1979. ment Printing Office, stock no. 052-003- 21. Fingl, E., and D. M. Woodbury. “General Prin- 00471-2, Washington, D. C., 1977. ciples, ” The Ph~lrmacological Basis of Thera- 80 U.S. Department of Health, Education, and peutics, L. S. Goodman and A. Gilman (eds. ), Welfare. National Toxicology Program An- 5th edition, MacMillan Publishing Co., Inc., nual Plan Fiscal Year 1979, New York, 1975. 9. U.S. Environmental Protection Agency. FIFRA 22. Schubert, J., E. J. Riley, and S. A. Tyler. Testing Guidelines, 1978. “Combined Effects in Toxicology—A Rapid 10. U.S. Environmental Protection Agency. TSCA Systematic Testing Procedure: Cadmium, Test Standards, 1979. Mercury, and Lead, ” Journal of Toxicology 11, Neel, J. V. “Mutation and Disease in Hu- and Environmental Health 4:763, 1978. mans, ” Banbury Report l—Assessing Chemi- 23. Van Ryzin, J., personal communication, 1979. cal Mutagens: The Risk to Humans, Cold 24. Interagency Regulatory Liaison Group, Work Spring Harbor Laboratory, 1979. Group on Risk Assessment. “The Scientific 12. Committee 17, Council of the Environmental Bases for Identification of Potential Car- Mutagen Society, “Environmental Mutagen cinogens and Estimation of Risks,”’ Journal Hazards,” Science 187:503, 1975. National Cancer Institute 63:241, 1979. 13. Kilian, D. J., and D. Picciano. “Cytogenetic 25. Scientific Committee of the Food Safety Coun- Surveillance of Industrial Populations, ” cil. “A Proposed System for Food Safety As- Chemical Mutagens, Principles and Methods sessment ,“ Food and Cosmet. Toxicol. 16, for Their Detection, Vol. IV, A. Hollaender Supplement 2, 1978. (cd.), Plenum Press, New York, 1976. 26. Cornfield, J. “Carcinogenic Risk Assess- 14. Legator, M, S., L. Troung, and T. H, Connor. merit, ” Science 198:693, 1977. “Analysis of Body Fluid Including Alyklation 27. Crump, K. S., D. G. Heel, C. H. Langley, and R. of Macromolecules for Detection of Muta- Pete. “Fundamental Carcinogenic Processes genic Agents, ” Chemical Mutagens, Princi- and Their Implications for Low-Dose Risk As- ples and Methods for Their Detection, Vol. V, sessment, ” Cancer Research 36: 2973, 1976. A, Hollaender and J, DeSerres [eds.), Plenum 28. Guess, H. A., K. S. Crump, and R. Pete. “Un- Press, New York, 1978. certainty Estimates for Low-Dose-Rate Ex- 15. Heddle, J. A., and W. R. Bruce. “On the Use of trapolations of Animal Carcinogenicity Multiple Assays for Mutagenicity, Especially Data, ” Cancer Research 37: 3475, 1977, the Micronucleus, Salmonella, and Sperm 29. Rai, K., and J. Van Ryzin. “Risk Assessment of Abnormality, Assays in Progress, ” Genetic Toxic Environmental Substances Using a Toxicology, P. Scott, B. A, Bridges, and F. H. Generalized Multi-Hit Dose Response Mod- Sobels (eds.), 1977. el, ” Energy and Hecdth, N. Breslow and A. 16. Kapp, R. A. Mutation Research, in press, Whittemore (eds, ) SIAM Press, Philadelphia, 1979. Pa., p. 94, 1979. 17. Kilbey, B. J., M. Legator, W. Nichols, and C. 30. Crump, K. S., and M. D. Masterman. “Assess- Ramel. Handbook of Mutagenicity Test Pro- ment of Carcinogenic Risk from PCBS in cedures. Elsevier, Amsterdam, 1977. Food, ” OTA Working Paper, 1979. 18. Ames, B. N., W. E. Durston, E. Yamasaki, and 31. U.S. Food and Drug Administration. An As- F. O. Lee. “Carcinogens and Mutagens: A sessment of Risk Associated with the Human Simple Test Combining Liver Homogenates Consumption of Some Species of Fish Con- for Activation and Bacteria for Detection, ” taminated With Polych~orinated Biphenyls Proceedings of the National Academy of (PCB’S), 1979. Chapter VI

Methods of Estimating and Applying Costs to Regulatory Decisionmaking

*

54-515 0 - 79 - 6 Chapter VI

Methods of Estimating and Applying Costs to Regulatory Decisionmaking

In the previous chapter, various testing methods for assessing health risks were evaluated for their usefulness in regulating environmental contaminants— i.e., setting an action level or tolerance. The primary issue involved in assessing health risks is not whether the potential risks from an environmental contaminant should be evaluated for purposes of regulation but rather what testing methods are most appropriate for assessing potential risks? The situation is reversed, however, when the associated costs of an action level or tolerance for an environmental contaminant are assessed. The primary issue is whether the costs should be taken into account in the setting of a tolerance or action level. This is a policy issue which is addressed in chapter IX, “Congressional Captions. ” In this chapter, the various approaches and techniques for estimating the costs and benefits of a proposed tolerance or action level are assessed, along with two common methods for applying cost and human risk data in the regulation of environmental contaminants. A more detailed discussion of the approaches and techniques for estimating the costs and benefits is provided in appendix E. Two methods, the cost-effectiveness method and the cost-benefit method, are analyzed for their strengths and weaknesses as regulatory decision-assisting tools. This analysis provides a basis for the discussion in chapter IX on the role of economics in regulating environmental contaminants in food.

Cost-effectiveness is a regulatory decision- a reduction in medical costs as a result in the assisting tool that compares the estimated net reduction of human risk. These costs would costs of the proposed regulation in dollar need to be subtracted from the cost of re- terms with estimated reduction of human risk stricted food. expressed in scientific terms. In the regula- The cost in lost food would also have asso- tion of environmental contaminants, the cost- ciated distributional effects which would effectiveness method would compare the esti- need to be identified to fully weigh the impact mated change in costs of a proposed action of a regulation. These effects would include level or tolerance for a contaminant in food identifying the consequences for producers, with the associated reduction in human change in the cost of food to consumers, and health risk—i.e., the benefits of the proposed indirect impacts on such businesses as food tolerance or action level. The comparison of processors, animal feed manufacturers, or net costs and risk reduction is performed for bait and tackle shops, several alternative levels in order to select a tolerance or action level. The principal cost The cost of food losses and related effects of regulatory action is the cost of food held off would be compared with the estimated reduc- the market because it exceeds the proposed tion of human risk for the contaminant at the action level or tolerance. There would also be proposed action level or tolerance. The hu- 73 74 ● Environmental Contaminants in Food man risk data could include such information Thus, FDA has placed an implicit dollar value as the reduction in the estimated number of on life. A more thorough application of the new cancer cases, the potential mutagenic cost-effectiveness method would attempt to and teratogenic effects, and other toxicologi- identify all the net changes in the benefits cal effects described in chapters II and V. (the reduction of all known risks) and all The toxicological effects vary by substance known costs, and are stated in scientific terms. The esti- Judgment is an important aspect to the mated net cost and reduced human risk at cost-effectiveness method. It allows the deci- various proposed action levels or tolerances sionmaking body, in this case FDA, flexibility are estimated and compared. The cost-effec- in interpreting health data and evaluating the tiveness method then requires the regulatory implications of that data on the setting of a agency to make its own judgment on what is tolerance or action level. This flexibility can the proper balance of net cost and reduced be particularly important in the assessment human risk when setting a tolerance or action of insufficient or variable toxicological data. level. While an agency’s interpretation of toxicolog- ical data might differ from others and gener- The Food and Drug Administration (FDA), ate disagreement about the final decision, the for example, followed a similar procedure in cost-effectiveness method does allow the com- its decision to reduce the tolerance for poly- plexity and shades of gray to be factored into chlorinated biphenyls (PCBs). In deciding on a the decisionmaking process, 2 parts per million (ppm) tolerance for fish, FDA compared for 5 ppm, 2 ppm, and 1 ppm The health data used in the cost-effective- in fish the estimated reduction of human risk ness calculations are generated by the toxico- with the estimated increase in the amount logical and epidemiological procedures re- and cost of food expected to be taken off the viewed in the previous chapter. There are market. Table 16 shows the human risk from also various approaches and techniques used cancer and cost data on commercial fish for to estimate the costs of food likely to be taken each of the three levels and the net changes off or restricted from the market as a result in risk and cost for moving to a 2-ppm and 1- of a proposed tolerance or action level. The ppm tolerance. It is these net changes that cost estimates will vary depending on the ap- are evaluated in the cost-effectiveness meth- proaches or techniques employed. All these od, Though FDA reviewed other available approaches require chemical analysis of a toxicological data, it was FDA’s judgment statistically significant number of food prod- that the reduction in risk (13.3 cancers) for a ucts known or thought to be contaminated, l-ppm tolerance did not offset the increased With such data, the amount of contaminated cost ($10, 3 million) and that a proper balance food that exceeds the proposed action level or for the net changes was established at 2 ppm. tolerance can be estimated,

Table 16.— Impact of PCB Tolerance

One year cost (1974 Lifetime risk of cancer for Reduction dollars) of commercial Increasing Tolerance heavy consumers of fish in human risk fish (land value) costs 5 parts per million (ppm) 9.8 per 100,000 or 46.8 new cancers per year Starting level $ 06 million Starting level 2 ppm, 7.2 per 100,000 or 343 new cancers per year 125 new cancers $ 57 million $51 million 1 ppm 4.4 per 100,000 or 21 new cancers per year 13,3 new cancers $16 million $10.3 million

SOURCE Federa/ Reg/ster VOI 44 No 127 June 29 1979 p 38333 Ch. VI—Methods of Estimating and Apply/rig Costs to Regulatory Decisionmaking ● 75

FDA multiplies the estimated amount of same amounts and at the same price that pre- food removed from commerce by the appro- vailed before the establishment of the toler- priate price (production, wholesale, retail, ance or action level. etc. ) per unit (pound, bushel, ton, or animal, etc.). This method of estimating food costs is The modeling technique, however, includes not the most accurate way of estimating costs not only supply shifts but also resulting price and does not attempt to measure the indirect shifts in both contaminated food and other costs or distributional effects that result from food products. With this technique, mathe- an action level or tolerance. matical models of the relevant portion of the economy are employed to trace shifts in sup- There are two other approaches for esti- ply, changes in the amount and price of var- mating the costs from the reduction of the ious commodities, and other factors affected food supply: the alternative-cost approach by the restriction on the use of a contam- and the opportunity-cost approach. The alter- inated product. native-cost approach estimates the cost of the next best alternative method or methods, if These models can, if properly programed, available, for replacing food removed from project changes in the production location of the market. particular crops as well as changes in the use For example, assume that 164,000 lbs of of various production factors in each region lake trout in Lake Superior were restricted of the country. Thus modeling can more real- from commerce because of the 2-ppm toler- istically describe likely reactions in food pro- ance for PCBs. The alternative-cost approach duction to the change created by a tolerance would require cost estimates of each avail- level. The result is a better estimate of the po- able method as well as various combinations tential cost of lost or restricted food. of methods for producing (and thus replacing) In attempting to project replacement costs, the entire 164,000 lbs of lake trout, the the opportunity-cost approach (using either amount condemned by the tolerance. In es- budgeting or modeling techniques) more ac- tablishing the replacement cost, this ap- curately estimates the true costs of a toler- proach does not consider price or supply ance or action level then either the alterna- shifts in the restricted food product nor shifts tive-cost or the FDA approach. in the supply of other food products. Such shifts are likely to occur when large amounts The distributional effects of a tolerance or of food in relation to the total supply are action level can be incorporated in the cost- taken off the market. Consequently, this ap- effectiveness method. But the opportunity- proach can lead in some instances to an over- cost approach can better generate these dis- statement of costs. tributional effects. With either method these The opportunity-cost approach does in- effects are most easily identified when the clude shifts in supply. But what information is data are initially being collected. The infor- included in the analysis depends on which mation is important in determining who will techniques are used. The budgeting tech- bear the brunt of the costs, both directly and nique only includes shifts in supply for the indirectly. The distributional effects flowing contaminated commodity. This technique from human health risks can also be identi- uses data on the production of the contami- fied. The available risk data can be assem- nated product to calculate the costs of pro- bled to determine who bears most of the po- ducing and consequently replacing a particu- tential and actual risks. Thus the distribu- lar amount of that food product that has been tional effects of the net cost and the reduction restricted from commerce. It is assumed that of risk from a tolerance or action level can be other commodities will be produced in the compared (l). 76 . Environmental Contaminants in Food

COST-BENEFIT The cost-benefit method compares the esti- Two approaches are available for convert- mated dollar costs of a proposed regulation ing the risks into dollars: forgone earnings with the estimated dollar benefits of reducing and willingness to pay. In the forgone-earn- human health risks. The cost-benefit method ings approach the analyst places an explicit of setting tolerance or action levels also re- economic value on life in attempting to esti- quires the evaluation of several alternative mate the productivity lost as a result of illness levels. The differences in benefits and costs or premature death caused by a contaminant as one moves from one proposed regulatory in food. This approach, however, does not in- level to the next are then compared. The tol- clude associated health costs such as medical erance or action level would be lowered until expenses incurred from illness. Including the costs (impact from food condemned) are such costs is required in order to more ac- greater than the benefits (reduction of human curately represent the total reduction of risks). When this level is reached, the eco- health costs—i.e., the benefits. The willing- nomically efficient tolerance or action level ness-to-pay approach allows people affected will be the next higher level at which the ben- by the regulation (rather than the analyst) to efits exceed the costs. estimate how much they would be willing to pay to avoid a risk. By representing the costs and benefits for The forgone-earnings approach attempts to a proposed action level or tolerance in dol- estimate the lost earnings of those individuals lars, cost-benefit analysis attempts to make who are estimated to become ill or die prema- the two sides of the ratio more comparable. turely because of the contaminated food (l). This method recognizes, as does cost-effec- The estimated dollar value of the benefit will tiveness, that any proposed action level or be determined by various techniques chosen tolerance will incur costs in terms of food or assumptions made by the analyst in con- taken off the market and benefits reflected in verting the risk data to a dollar value. Dis- the reduction of either actual or potential counting and earnings are two areas in which health risks. As the level is lowered, costs in- the analyst can influence the value of the crease and human health risks decrease. benefit. Techniques of estimating costs are similar Since many of the benefits of a tolerance in the cost-benefit and cost-effectiveness are not likely to be realized until 10, 20, or 30 methods with the exception that the reduced years later, discounting is used to convert medical costs are included with the benefits. future dollar benefits into present dollars. There are some ways of valuing benefits, Discounting is also used for estimating future though, which are unique to the cost-benefit costs. Discounting is required even if the fu- method. ture dollars are adjusted for the rate of infla- tion because a dollar spent now can be in- Though both methods rely on the available vested productively to yield a larger number toxicological data, cost-benefit requires the of real dollars—i.e., inflation adjusted—in conversion of all data into dollar values. This the future, For example, $100 invested at 5- economic conversion is best accomplished percent interest becomes $105 in one year. when the data are expressed as the number Discounting is the reverse: $105 next year of premature deaths per year avoided, the has a present value of $100 when the dis- number of person-days lost to illness avoided, count rate is 5 percent. This means that or the probability that some percentage of the $10,000 worth of benefits that will occur 10 exposed population would die prematurely or years from now would actually have a pres- would lose a specified number of days from ent value of $6,135 if a 7-percent discount normal activity due to illness. rate is used, Ch. VI—Methods of Estimating and Applying Costs to Regulatory Decisionmaking ● 77

While most economists agree that future will be greater than the net earnings esti- costs or benefits need to be discounted, they mate. do not agree on the value of the discount rate. The willingness-to-pay approach is concep- The rate can vary from I percent to as high tually a more correct approach in that it asks as 15 percent (z). The rate, however, has to the individual to place a dollar value on the remain constant for both the costs and the reduction of associated risks from an environ- benefits. Obviously, the value of the rate will mental contaminant. This value is then used affect the estimate of benefits or costs, The for placing a value on life itself. While con- lower the discount rate, the greater the dollar ceptually correct, this approach does have estimate in present dollars. Consequently, some inherent problems, such as: 1) the capa- more weight would be given to the benefits bility of the questioned person to accurately and costs accrued in the future. understand the ramifications of the risk and 2) the individual’s economic position. For ex- The estimate of the benefits also depends ample, an economically disadvantaged per- on whether gross or net earnings are used. son might place a small value on risk not be- Gross earning estimates include an individ- cause the person feels the risk is of little or no ual’s or a group’s total wages or salaries. Net concern but because that person cannot af- earnings consist of total wages or salaries ford an increase in food prices. This ap- minus the individual’s or group’s consump- proach is affected by the assumptions made tion. Obviously the gross earnings estimate by the public being surveyed.

APPLICATION OF METHODS FOR REGULATION

The use of the cost-benefit and cost-effec- needed on the amount of food likely to be con- tiveness methods for regulatory purposes is taminated. Such information was available affected by the following factors: 1) FDA’s in- for PCBs in fish, but it is not available for terpretation of the Food, Drug, and Cosmetic PCBs in milk, poultry, and eggs. This is be- Act, 2) the approaches and techniques for es- cause contamination of milk, poultry, and timating the economic value of the costs and eggs is likely to occur as a result of industrial benefits, and 3) the inherent difference be- accidents. Such accidents are sporadic and tween the two methods. therefore difficult to predict (e.g., the July 1979 PCB contamination of poultry and eggs As noted earlier, FDA interprets the Act as in Idaho) (3). Consequently, the estimates for requiring it to weigh only the impact of the costs incurred because of food removed from cost and the amount of food removed from commerce cannot be determined for such commerce in the setting of the proposed tol- contamination incidents, and thus neither erance or action level. FDA’s approach for method can be employed. Both methods could estimating this cost can be applied in either be used to set a tolerance or action level for the cost-effectiveness or the cost-benefit mercury and kepone in fish. method, but both of these methods can rely on alternative-cost or opportunity-cost tech- The approach and techniques used by the niques which more accurately estimate the cost-effectiveness methods for generating cost incurred from condemned food. These necessary data can take considerable re- other techniques also more accurately esti- sources and time, 2 months to over a year to mate the tolerance or action level’s impact on gather the data just for the costs alone. The availability y of food than the approach used by amount of time needed depends on the ap- FDA. proaches and techniques used. The more ac- Whichever technique is employed to esti- curate the information being generated, the mate the costs, adequate information is more time and resources are required. 78 . Environmental Contaminants in Food

FDA, however, often has to make an initial cost-benefit method, which only requires a decision in the form of an action level in 2 comparison of the numbers for each side of months or less for a newly identified environ- the ratio to establish the appropriate toler- mental contaminant. A sufficient amount of ance. time is usually available for utilization of the various approaches or techniques if a follow- The cost-effectiveness method has the po- up decision is involved. For- example, 6 years tential to reveal more of an agency’s thinking expired from the time an initial PCB action in the decision than cost-benefit does. This level was proposed until a final tolerance for was demonstrated with the earlier discussion PCB was proposed this year. Either method is of FDA’s setting of a 2-ppm tolerance for more likely to be used in setting a formal PCBs in fish. In addition, because it recog- tolerance than an initial action level for an nizes the uncertainties inherent in the esti- environmental contaminant. mates of the health risks, the cost-effective- The substantive difference between the ness method allows FDA the flexibility to ad- just the weight given to the benefits or the cost-benefit method and the cost-effective- costs in its decision. For these reasons, the ness method is that the cost-benefit method cost-effectiveness method is the more appro- places an explicit value on life by converting priate method at this time for weighing the the health data to dollars while the cost-effec- costs in the setting of a tolerance. tiveness method places an implicit value on life by weighing the health data in its scien- Neither method can evaluate all the infor- tific form with the costs. As a result, a signifi- mation required by FDA in setting a toler- cant amount of judgment is exercised by the analyst using the cost-benefit method when ance. For example, FDA requires for enforce- ment purposes analytical methodologies that selecting the different approaches and tech- niques for estimating the benefits. As dis- can detect, measure, and confirm the identity of the contaminant at the level being pro- cussed earlier, the selection of these ap- posed in food. This means that the tolerance proaches and techniques has a strong bear- cannot be set at a level below the available ing on the outcome of the ratio and conse- analytical capabilities for detecting the con- quently the tolerance established by this taminant in food. While the analytical capa- method. bility is an important factor in setting a toler- The cost-effectiveness method places a ance, it cannot be evaluated within the deci- greater judgment burden on the agency and sionmaking framework by either method. less on the analyst. While the analyst does af- Consequently, these two methods should be fect the outcome, judgment is primarily exer- viewed as decision-assisting aids that allow cised by the agency in weighing the net cost the regulator the means to weigh many of the with the reduction in human risk (benefits). relevant costs and benefits in the setting of a The agency exercises less judgment in the tolerance.

CHAPTER VI REFERENCES 1. Epp, Donald J. “Economic Analysis of Alter- 3. Foreman, Carol Tucker, Assistant Secretary native Action Levels in the Regulation of Envi- for Food and Consumer Services, U.S. Depart- ronmental Contaminants of Food, ” OTA ment of Agriculture. Testimony before the Working Paper, April 197’9. Subcommittee on Oversight and Investiga- 2. Weinstein, Milton C., and William B. Stason. tions, House Interstate and Foreign Com- “Foundations of Cost-Effectiveness Analysis merce Committee, Sept. 28, 1979. for Health and Medical Practices, ” New England Journal of A4edicine, Mar. 31, 1977. \ Chapter VII Monitoring Strategies Chapter Vll Monitoring Strategies

Monitoring involves the systematic collection and chemical analysis of food samples or other samples from the environment. The aim is to protect consumers by determining short- and long-term trends in the levels of various chemicals in food and the environment.

STRATEGIES

Monitoring strategies can be shaped to Therefore, investigatory monitoring ap- achieve either of two objectives. The first is to proaches are vital. identify food lots that violate established tolerances and action levels. The second is to Investigatory Monitoring identify new environmental contaminants as Investigatory monitoring attempts to detect they enter the human food chain. unregulated chemicals as they enter the food The first objective is met by regulatory chain. This strategy involves the collection monitoring: the second through investigatory and analysis of samples which may or may monitoring. Each of these strategies could be not be foods, The analytical techniques em- complemented by specimen banking. Neither ployed for the detection of unregulated chem- is incompatible or mutually exclusive. icals may or may not be accepted as standard methods comparable to those used for regula- Regulatory Monitoring tory monitoring. The Federal agencies responsible for limit- Analytical methods for investigatory moni- ing consumer exposure to contaminated food toring include broad-spectrum determina- now conduct regulatory monitoring. Food tions that may sacrifice some quantitative in- samples are collected and analyzed for envi- formation (i. e., exactly how much of a given ronmental contaminants for which action substance is present in a sample) for more levels and tolerances have been established. qualitative information (i. e., better assess- Based on available information or agree- ment of what or how many foreign sub- ments with States, not all samples are ana- stances are in the sample). These analytical lyzed for all regulated substances. Regulatory methods are not necessarily designed for use monitoring employs standardized, accepted in litigation, They are designed primarily to analytica1 techniques. Because the proce- indicate the presence of a potentially hazard- dures are standard and can be verified by ous substance. If one is found, an accepted other laboratories, they generate data that analytical method to detect the substance can be presented in courts of law with little would have to be developed—a method com- probability of being successfully contested. patible with instrumentation existing in regu- latory-monitoring laboratories. Chapter III reviewed Federal monitoring programs and chapter IV reviewed State Investigatory monitoring includes two dis- monitoring programs. It is unlikely that these crete types of monitoring: monitoring for sus- monitoring programs will detect new envi- pected environmental contaminants, and ronmental contaminants, since both are re- monitoring for uncharacterized environmen- stricted to searching for regulated chemicals. tal contaminants. Each of these (as well as regulatory monitoring) can be complemented use, oil/water partition coefficients, bioac- by specimen banking. cumulation potential, known toxicity, and means of disposal. The FDA list of chemical Monitoring for Suspected Environmental contaminants in foods is shown in table 17. Contaminants This approach is limited by available Some chemicals that are not regulated by analytical methods. Most of the chemicals action levels or tolerances are suspected to recognized as food contaminants are those be dangerous to humans if consumed in foods. that are relatively easy to detect by gas This group includes chemicals that may be chromatography or atomic absorption spec- present in food because of their use, toxicity, trometry. Chemicals that cannot be easily production volume, and persistence. Exam- detected by these methods may, of course, re- ples of these chemicals can be found on the main unidentified and unrecognized as food Environmental Protection Agency (EPA)/ contaminants (l). Natural Resources Defense Council (NRDC) The factors used to identify a chemical’s priority pollutant list established in June potential for entering the food supply are 1977. These substances may be called “sus- based on knowledge of the properties and en- pected” or “potential” environmental con- vironmental behavior of other chemicals taminants. already known to be in food. This knowledge, Monitoring for suspected environmental in turn, is based on our information about the contaminants involves a different strategy extent of contamination information that than the one used in monitoring for regulated depends on our analytical capabilities. Thus, contaminants. Under the latter strategy, there is an inherent tendency to identify as foods are analyzed for compounds with speci- potential food contaminants those chemicals fied action levels and tolerances to provide in- that are similar to chemicals already iden- formation for regulatory enforcement. This is tified in food. Such a tendency can only be off- not required for investigatory monitoring. set by the use of good scientific judgment or Thus, the monitoring program for suspected the development of new data. This bias illus- environmental contaminants is generally not trates a general weakness in all systems for as intensive (see chapter 111), Furthermore, setting priorities: chemicals on which there is the analytical methods for detecting sus- no information will automatically be given pected environmental contaminants may not low priority unless some room is left for large- be as prescribed as those for regulated con- ly intuitive judgments (1). The scientific cri- taminants. teria and methods used in determining what priority various toxic substances receive in Suspected environmental contaminants monitoring programs are discussed in more could be identified by surveying the universe detail in appendix F. of industrial chemicals and ranking them ac- cording to their potential for entering the food Although these exercises in setting prior- supply in toxic amounts (l). Such an ap- ities suffer from many limitations (including proach has been recommended to FDA by an lack of data, poor choices of criteria on which internal study group established by former to set ranks, deficiencies in the scoring and Commissioner Donald Kennedy (7), ranking systems, and deficiencies in scien- tific judgment), they still can serve a valuable This method has been employed by the function in guiding monitoring systems. Set- Food and Drug Administration (FDA) to devel- ting priorities is a prescreening exercise in op a list of chemical contaminants in food, which a compromise is struck between the ef- The criteria used in selecting the chemicals fort expended in preparing a priority list and included: occurrence in food, volume of pro- the effort that would be wasted in identifying duction, associated impurities or byproducts, and quantifying all the chemicals present in a predicted environmental stability, pattern of sample, Ch VII— Monitoring Strategies ● 83

Table 17. —Chemical Contaminants in Foodsa

up to 1 up to 1 1 up to 1 up to 1 ) 01-10 0.1-10 0.02-008 002-0.04 1.4

tr-6 7

03-2.0 t r-O. 1 0.2-0.8 unknown unknown 1

unknown unknown Niagara River. N Y unknown unknown }

0.03-162 0.03-162 0.03 -16.2 84 . Environmental Contaminants in Food

Table 17.—Chemical Contaminants in Foodsa—(cont.)

Chemical contaminant Range in ppm Locations

White Lake, Mich

Ohio River, Ohio

Pine River. Mich

Houston Channel Houston Channel

By applying appropriate criteria to the toxicity data on compounds, they may not be universe of industrial chemicals, it may be recognized as threats to human health. This possible to detect potential environmental class of substances is similar to suspected en- contaminants in food that have not been iden- vironmental contaminants in that there are tified as significant by other methods. Al- no stipulated analytical methods to detect though there is no truly independent way to them and no monitoring is mandated. Unchar- verify the reliability y of priority lists, such lists acterized environmental contaminants are could be generated for a pilot program de- different from suspected contaminants in signed to evaluate this approach vis-a-vis un- that none have been placed on lists of poten- characterized monitoring. tially harmful substances. Although validated analytical methods for Monitoring for Uncharacterized identifying uncharacterized environmental Environmental Contaminants contaminants may be lacking, data on their Uncharacterized environmental contami- presence or absence can be generated. Chem- nants are substances that may have entered ical analyses that are designed to show only the food supply but which have not been the presence or absence of a compound in a classified as regulated environmental con- sample are called qualitative analyses. In taminants or suspected environmental con- many cases an accepted analytical method taminants. Compounds may fall into this cate- for one class of compound will yield quantita- gory because they are not known or sus- tive results for those compounds and qualita- pected to occur in food. Because of a lack of tive results for others, Therefore, the pres- Ch. VII—— Monitoring Strategies ● 85 ence or absence of some suspected or unchar- or changes were observed, corrective regula- acterized environmental contaminants may tory actions could be taken. be determined even though the chemist is not specifically looking for them. Specimen Banking In setting up a system of monitoring for un- It is difficult to detect environmental con- characterized environmental contaminants, taminants unless one is specifically looking some preliminary judgments would be made for them, Monitoring methods for identifying about the chemical nature of the target sub- suspected and uncharacterized environmen- stances. Classes of compounds to be moni- tal contaminants promise to partially allevi- tored would be selected on the basis of their ate the problems. Yet, even as new analytical structural characteristics, their use, and instrumentation is developed and scientific their suspected toxicity. Trace metals, halo- knowledge expands, there will be new classes genated hydrocarbons, or radioactive sub- of compounds discovered in foods that have stances are examples of such classes. The been present for years but were undetectable class of compound determines the type of ex- with then-existing instrumentation. Informa- traction required to separate the substance tion on how long the compounds have been in from other constituents of food, as well as the food, what kinds of foods are affected, and instrumentation needed to detect the pres- from what areas the foods were derived ence of the substance. would be of great help to epidemiologists and public health officials who must decide The available analytical methods best whether or not the chemicals have had (or suited for the class or classes of compounds will have) an adverse impact on the public. under consideration would be selected. The question of whether analytical techniques One approach to this problem is the collec- are sufficiently advanced to support unchar- tion and storage of samples on a regular basis acterized monitoring is explored in chapter and in a manner that will protect their chem- VIII. ical integrity. In the future, when new instru- mentation is developed or a toxic compound is The establishment of an uncharacterized discovered in foods, samples can be with- monitoring program would require the devel- drawn from storage and analyzed. This in ef- opment of appropriate sampling and sample- fect would be retrospective analysis. Investi- handling guidelines. It would also be neces- gators could go back in time, reconstruct sary to modify currently used analytical pro- events leading to a current situation, and esti- cedures. For example, a typical program for mate human exposures from the consumed organic contaminants would involve prelimi- foods. nary screening to establish baseline levels of contamination in samples of food and water There are examples of this kind of retro- or selected indicator species over a given spective detective work. When high concen- period of time. This information would then trations of mercury were discovered in tuna be used to develop an appropriate sampling and swordfish, pollution was widely held plan to determine changes or trends over responsible, But analysis of museum speci- time. An increase in levels of an uncharac- mens that had been stored for decades in- terized substance indicates its entry into food dicated that the mercury levels were prob- and water. This finding would trigger addi- ably as high a century ago as now. Therefore, tional analytical efforts to characterize the the may be from natural new compound. Preliminary information on sources and may have always been so. This the substance’s structure would be trans- does not mean that the metal is not a potential mitted to toxicologists for evaluation and health threat but rather that the potential ex- comparison with available information on posure from eating the fish has not changed known toxic compounds. If alarming trends much over the years. 86 ● Environmental Contaminants in Food

There is a problem in utilizing most exist- EPA and the National Bureau of Standards ing collections for retrospective chemical are now working towards developing such a analysis. Since samples were not collected program by testing various methods of pre- for use in chemical testing; they were not serving samples for long periods of time with- stored in a manner to maintain their chemical out either adding unwanted chemicals or los- integrity. A 1975 survey of environmental ing ones that are already in the sample. A specimen collections in the United States by number of scientists are encouraging this Oak Ridge National Laboratories concluded program and similar efforts (3-5). If continued that few of the existing collections were suit- funding is made available for specimen bank- able for retrospective chemical analyses (2). ing of environmental samples (including Therefore, a need exists for a national pro- foods), future investigators will have an easi- gram to collect, store, and maintain environ- er job of assessing what impact environmen- mental samples (including food) to allow ret- tal contaminants in foods may have. rospective investigations,

SAMPLING

Sampling involves the systematic collection The number of samples to be taken de- of information from a portion of the environ- pends on how much risk of being wrong we ment. Sampling is done in such a way that the are willing to accept. In other words, to what collected samples represent the whole in degree of certainty do we want to know that terms of the information desired. In regula- our food is free from environmental contami- tory monitoring the samples must be food nants? One-hundred percent certainty would commodities because the intent of the pro- require analysis of every food item. Accept- gram is to determine the levels of regulated ance of a lesser degree of certainty allows the substances in the food supply. This informa- use of less costly sampling approaches. tion is the basis for enforcement actions. Before preliminary data are collected, Food samples may not be the best indi- there is no way to calculate the exact number cators if the monitoring is meant to serve as of samples needed to yield an answer of an early warning system—in other words, to specified certainty. The most difficult factor detect a substance soon after it enters the en- to estimate is the variation specific to each vironment and before it gets into foods. It may contaminant and how it changes over time be better to analyze nonfood samples such as and space. This kind of information would river sediments, water, or uneaten organs have to be collected in pilot programs for from food animals (the organs may concen- suspected and uncharacterized monitoring trate the substance to analytically detectable before a national sampling plan could be levels before it can be seen in the flesh). The developed. The number of samples taken will finding of an environmental contaminant in probably be constrained by the money, man- nonfood samples would trigger the examina- power, and available laboratory resources. tion of foods. The density and location of sampling sites The following discussion outlines the pri- depend on the socially acceptable level of mary considerations in selecting samples for uncertainty, whether the contaminant stems investigatory monitoring systems. Construct- from a point or nonpoint source, and how it is ing a sampling plan for such systems involves transported in the ecosystem. a number of decisions based on preliminary information about the nature and extent of If one is dealing with widely distributed environmental contamination. Such decisions nonpoint source environmental contaminants include the number, sites, frequency, and that are transported through water, the ideal types of samples. sampling locations would be rivermouths that Ch. VII—Monitoring Strategies ● 87

are discharge points of major watersheds. the production systems have been identified, one could very effectively monitor the in- the appropriate sampling frequency can be dustrial portion of Michigan by sampling determined (6), shellfish or fish from some two dozen major rivers just as they enter the Great Lakes. The selection of the types of samples to be Baseline levels in these foods could be deter- collected is also critical in identifying en- mined, but the origins of the contaminants vironmental contaminants as they enter the would be difficult to determine (6). food chain. Although biological samples offer many advantages in monitoring systems, non- The other extreme is to monitor food prod- living samples may be preferable in some in- ucts on a production, site-specific basis. If the stances. An example might be bottom sedi- aim were to inventory all industries (includ- ments from rivers, lakes, or estuaries. Bottom ing agriculture) that utilize and/or discharge sediments are derived, for the most part, a toxic substance, one could then routinely from erosion of land and often bring with monitor food products, fish, and game at each them to the aquatic environment substances identified site. Theoretically, the kepone and that are used on land such as or polybrominated biphenyls (PBB) situations pesticides. Moreover, once they are in the would have been detected much earlier with aqueous environment, they can “sorb” or such a system. The number of operations concentrate many substances found in indus- (large and small) that would need to be moni- trial discharges. The contaminated sediments tored is unknown, but the total appears to be then serve as a mechanism to expose the so large that the costs would preclude consid- plants and animals that live in the waters to a eration of this alternative (6). particular chemical. Contaminated sediments A reasonable compromise approach is to may also be used to pinpoint the source of a use information generated under the Toxic chemical once it has entered the river, lake, Substances Control Act on the types of chem- or estuary (6). icals manufactured at various locations to guide in the development of a sampling plan Other types of nonliving samples might in- for use in investigatory monitoring systems. clude air, river water, drinking water, or The sampling plan would focus on some food rain. All have certain advantages and disad- organisms and some nonfood items. The data vantages. For instance, the concentrations of derived from analyses of such samples would many environmental contaminants in air and yield the greatest information about environ- water are very low, causing problems for the mental contamination trends in the region analytical chemist. When a substance is from which the samples were drawn. found in air or water, it is sometimes difficult to determine where it entered the system. The frequency of sampling depends on the However, because we breathe the air and rates a t which the contaminant moves drink the water as well as eat the food from through, accumulates in, and decomposes out these environments, air and water cannot be of the food production system being moni- eliminated as potential samples (6). tored, Different food production systems have different genetic and environmental charac- The most appropriate biological samples in teristics that determine the rates of material an investigatory monitoring system should dynamics or transfer. For a beef feedlot, a reflect key elements in the human food chain. range of 50 to 180 days would include the Samples may include not only traditional period involving a single-batch process. For agriculture products but also fish, game, an apple crop, a single sample per year would shellfish, crustaceans, and wild fruits and suffice. The frequency of sampling may be nuts, Many of these wild foods also accumu- different for each type of production process. late both point and nonpoint source environ- Once the species and the characteristics of mental contaminants. Criteria for selection of

1– , 4— 88 ● Environmental Contaminants in Food

the exact organisms should include the fol- The supply of data must also be timely. If lowing characteristics: one wishes to regulate the level of an environ- mental contaminant in food when concentra- position in the food chain, tions vary weekly, a monitoring system that lifespan, reports data with a 6-month delay is not feeding behavior, workable. understanding of the organism’s physiol- ogy and biochemistry, Finally, to develop information on exposure body fat content, trends and determine the effectiveness of mobility, regulatory monitoring and enforcement, hu- availability for sampling, and man tissues, blood, and urine can be analyzed utility to humans (6). for the presence of environmental contami- nants. Human monitoring can be performed Once a sampling plan is developed and for either regulated, suspected, or uncharac- samples collected and analyzed, the data terized substances. A sampling plan to detect must be presented in a form useful to the trends in exposure among different popula- regulator. The data analysis should be rapid tion groups could be developed, but data gen- and provide information on trends as well as erated from human monitoring would be un- specific concentrations without significant able in most cases to identify the source of ex- distortion or deletion. Even with the best- posure. At the present time, EPA houses the designed computer retrieval system, the nec- principal human monitoring program. This essary data bank would become extremely program is primarily concerned with pesti- large, complex, and expensive. cide residues.

QUALITY ASSURANCE

To ensure that data generated by any moni- health might be affected. If the results are toring strategy are as accurate as possible, erroneously high, undue economic hardship schemes have been developed to pinpoint er- may be imposed on the food producer. rors. These schemes are called quality assur- ance programs. Such programs are manda- The Federal agencies that monitor foods tory in analytical laboratories because the for environmental contaminants are aware of possibility y of errors always exist. these problems. Thus they use standardized analytical methods that have been tested and Errors arise from a number of sources. For therefore offer some assurance that the re- instance, impure chemical reagents, dirty sults will be acceptable in court. When a vio- glassware, or sample containers can impart lative sample is found, the product or batch is contaminants to the sample that may inter- reanalyzed whenever possible to assure that fere with the analysis and result in false results are valid. readings. Instruments are not always stable and may give false readings. Some samples These two practices are part of a quality may contain substances that interfere with assurance program. There are others as well. analyses, or a substance may be bound in a When a new chemical extraction or analysis sample in such a way that normal extraction technique is tested, it is important to analyze met hods will not extract it. Another factor is a sample w i t h known composition t o check the potential for human error in the labora- the validity of the technique. Also, during tory. These factors, singly or in combination, routine determinations samples of known can lead to reported concentrations that are composition should be analyzed to check on in error. Since commodities that violate the other types of potential errors. The standards could be marketed and consumed samples of known composition are called if the results were erroneously low, human “reference material" and may come from Ch. VII—Monitoring Strategies ● 89

various sources including the National Collaborative studies that involve more Bureau of Standards and EPA, than one laboratory or group analyzing the same sample by the same or different meth- Available reference materials do not ods are part of a quality assurance program. always satisfy the needs of current chemical If all results are similar within acceptable monitoring programs, since such materials do limits there is some assurance that the meth- not contain many known environmental con- od(s) are precise and perhaps accurate, taminants. Moreover, the contaminants may not be stable under the storage method used. This is particularly true for synthetic organic Often, more than one method can be used to measure a given contaminant. Confidence chemicals. Another problem is that the type of reference material-i.e., beef liver—may in accuracy can be increased if the methods agree. This is one of the reasons that a moni- not be similar enough to the food samples to toring laboratory should have several meth- be analyzed-–i.e., fish—to be very helpful. ods available. This points up an important gap in our ability to analyze accurately for environmen- All of these aspects are important to tal contaminants in foods. The variety of assure that the proper answers are gener- reference materials and the variety of com- ated by a monitoring laboratory. All are time- pounds of known concentration in these mate- consuming and expensive. Therefore, any rials are insufficient to satisfy the needs of chemical monitoring program must allocate the analysts. More effort must be expended as much as 10 t o 20 percent of its time for this to correct this problem. increased workload.

CHAPTER VII REFERENCES Chapter VIII Monitoring Instrumentation

. Chapter VIII Monitoring Instrumentation

Various instruments have been designed to detect and quantify organic chemicals, trace metals, and radioactivity in foods. This chapter focuses on not only the technological state of this analytical art but also how such instrumenta- tion could be applied in an investigatory monitoring program.

ORGANIC ENVIRONMENTAL CONTAMINANTS

Analysis instruments and techniques. Among the most frequently used methods are mass spectrome- The vast number and wide diversity of nat- try, gas chromatography, and liquid chroma- ural and synthetic organic chemicals pose tography. Table 18 summarizes the tech- difficult analytical problems. Many methods niques available for qualitative and quantita- and instruments have been developed to de- tive organic analysis. More information can tect and quantify specific environmental con- be found in appendix G. taminants in foods. There are generally three steps required for each type of analysis: ex- The gas chromatography is an instrument traction, cleanup, and detection and quantifi- designed to separate, identify, and quantify cation. Figure 6 is a flow chart depicting how organic compounds. In simplified terms it these steps would be applied to the analysis consists of four components: an injection of the Environmental Protection Agency pri- port, a column, a detector, and a recorder. ority pollutants in food. There are many types of injection chambers, Extraction usually involves mixing the food columns, and detectors, but the principles of sample with a selected solvent. In this step, operation are similar. the chemical is removed from the food and Gas chromatography involves vaporization dissolved in the solvent. The time required for of the sample to be analyzed. The gaseous this step depends on the physical and chemi- sample then passes through a long tube or cal characteristics of the food sample and the column packed with a solid matrix which is substances. In some cases, the process may coated with an organic compound. This is take 24 hours or more. The initial extract con- called the “liquid phase. ” The rate at which tains not only the substance of interest but various organic compounds in the sample possibly other organic compounds of similar pass through this column is a function of volubility that must be removed before various chemical and physical properties, analysis. notably molecular weight. polarity, and geo- The second step is known as the cleanup or metric structure. The time of passage through isolation stage. The complexity and time re- the column is called the “retention time. ” quired for this procedure depends on the food Each compound has a characteristic reten- sample and the number of substances to be tion time in the column. removed. In many cases multiple cleanup To determine when and how much of each steps are necessary. The end products of substance leaves the column, a detector is cleanup procedures are fractions containing placed at the end of the column. The retention different classes of organic compounds. time can be used to identify each compound, The final step, detection and quantifica- while the strength of the electrical signal tion, requires the use of highly sophisticated from the detector indicates the quantity. The

93 94 Q Environmental Contaminants in Food

Figure 6.— A General Scheme for the Qualitative and Quantitative Analysis of the Organic EPA Priority Pollutants in Semisolid Foods

Extraction procedure Alternate * suitable for particular methods ● w I food sample type t I I I Initial cleanup - -0 ------J I I of extract b I

Initial interpretation I and quantification I

a Identification is based on the presence of characteristic ion fragments and associated chromatographic retention time data. Absolute identification would require a detailed interpretation of the complete mass spectrum of the organic com- pound of interest.

SOURCE John L Ldseter Approaches to Monitoring Environmental Contaminants [n Food OTA Working Paper 1978 identity of the compound is confirmed by com- gerprint” are compared to a standard “fin- paring its retention time to that of a series of gerprint” to determine how much of each standard solutions (of known composition) in- substance is present in the sample, If a peak jected into the gas chromatography under the in the sample fingerprint, for example, has same conditions. The areas under the the same retention time but twice the area as graphed peaks of a sample readout or “fin- one in the standard fingerprint, it is assumed Ch. VIII— Monitoring Instrumentation ● 95

Table 18.—Techniques Available for Qualitative and Quantitative Organic Analysis

Approximate Method detection limit, gm Specificity or common uses Gas chromatography Retention Indices. ., Detects most compounds Electron capture Halides, conjugated carbonyls, nit riles, di- and trisulfides Flame photometer. Phosphorus, sulfur Nitrogen/phosphorus Nitrogen, phosphorus Chemical methods Pyrolysis ...... Compound-type determinant ion Chemical reagents. ., Classical-functionality determination Electrolytic systems . Sulfur, nitrogen, halogens Instrumentation infrared-grating Compound-category-type identification Infrared-interferometer Compound-category-type identification Ultraviolet ...... Aromatics conjugated carbonyls Proton magnetic resonance Excellent for function, some molecular weight data. Mass spectrometer Batch inlet . Best for complete identification, GC-MS mode ...... molecuIar weight structure, and Multiple-ion detection . . . . function. Confirm any compound.

SOURCES Adapted f;om W McFadden Techniques of Comb!ned Gas Chromatography/Mass Spectrometry Wlleylntersc{ ence New York N Y 1973, p 4, and “Trace Orqanlc Analysls ‘ Enwronmental Science and Technology 12757 (1978) that the identity of the sample peak is the Also because less cleanup is required, their same as that of the known standard but that use reduces the possibility that an important twice as much is present. environmental contaminant may be removed in the process of preparing a sample for anal- Recent developments in analytical organic ysis. chemistry promise new, more sensitive tech- niques for monitoring of synthetic organic Although the capillary column offers a chemicals in foods. These developments in- number of advantages, it is more expensive clude high-resolution glass capillary columns and difficult to use than the standard packed for gas chromatography, and the linking of column. More highly trained personnel are gas chromatography to mass spectrometers. required to operate capillary columns and in- terpret results. However, many chemists feel The glass capillary column is much more that the advantages far outweigh any difficul- efficient in separating compounds than the ties. Glass capillary columns are not now older “packed” columns (figure 7). The gas widely used for monitoring synthetic organics chromatograph’s packed column separates in foods. Packed columns continue to play an the individual components so that they exit important role in the organic chemistry moni- the column and enter the detector one at a toring laboratory, but their future use may be time. Some of the early data on the pesticide restricted to more specific analyses. family DDT-DDE-DDD were incorrect be- cause polychlorinated biphenyls (PCBs) had Gas chromatography are currently the similar retention times in packed columns. As primary means by which laboratories moni- a result, reported concentrations of DDT, tor for organic chemicals. Some chemists, DDE, and DDD were often too high because though, prefer mass spectrometers because the peak areas were really reflecting a com- of their high sensitivity and flexibility (z). The bination of pesticides plus PCBs (l). Because coupling of a gas chromatography to a mass of their superior resolution power, glass cap- spectrometer introduces a new dimension illary columns avoid some of these problems. that allows a chemist to identify compounds 96 . Environmental Contaminants in Food

Figure 7. —Comparison of the Gas Chromatographic Therefore, instruments such as infrared Analysis of a Standard Mixture of Polychlorinated spectrophotometers and high-pressure liquid Biphenyls (Aroclor 1254) on Two Types of Gas chromatography are necessary. These instru- Chromatographic Columns ments add many thousands of dollars to the costs of setting up and equipping an up-to- A: Analysis Using a Conventional Packed Column date organic chemical-monitoring laboratory.

Application to Investigatory Monitoring Although the instrumentation described above has been available for several years, it has only recently been utilized for investiga- tory monitoring. Experimental projects incor- porating computer analyses are now under study in at least three research facilities: the r I I 1 I 1 I I 1 1 Bodega Marine Laboratory, the University of 18 3 0 6 12 24 New Orleans, and the Virginia Institute of Time (Minutes) Marine Science, These projects employ high- resolution glass capillary gas chromato- B: Analysis Using a High-Resolution Glass graphs coupled to mass spectrometers and Capillary Column sophisticated computer analyses, The result- ing “fingerprints” or readouts from a sample extract are highly complex and show the presence of not only known or suspected envi- ronmental contaminants but also many or- ganic compounds that must be classified as uncharacterized. The three research groups are now testing an approach to determine which compounds deserve further testing. The gas chromato- graphic fingerprints are computerized so that I 1 1 1 I I I 1 information on retention times and peak in- 0 10 20 30 40 tensities are stored. Subsequent samples are Time (Minutes) collected and analyzed and the data are com-

SOURCE John L Laset II Approaches to Monltorlng Environmental Contaml puterized in the same manner. By comparing nants Ir Foc)i OTA i’dork(ng Pdper 1978 results from the same location over time (or with samples taken from other locations), one can determine whether peaks are increasing classed as uncharacterized (3). Moreover, re- or decreasing, whether new peaks have ap- sults from a mass spectrometer can confirm peared or old ones have disappeared, or the identity of known chemicals whose pres- whether some peaks occur only in some loca- ence (as indicated by gas chromatography tions. alone) is in doubt because of sample contam- With appropriate computer programs or ination or other factors. Mass spectrometers ‘‘soft ware, ” the data bank could be asked if are expensive and in addition require compu- there is a peak that occurs in samples from terized data management systems. only one area or location. If so, the inves- Because of their physical and chemical tigator can assume that the compound is an- characteristics. many organic compounds thropogenic, or man-induced, The compound cannot be analyzed by gas chromatography. could then be identified and its source con- Ch. VIII— Monitoring Instrumentation ● 97

trolled before the concentration increased to that respond mainly to organic compounds the point that a major contamination episode containing halogens such as chlorine or bro- occurred. This type of monitoring could have mine. Because naturally occurring halogen- detected kepone in the James River, Va., long containing organic chemicals are rare and before thousands of pounds of the pesticide are far outnumbered by manmade ones, an entered the river. uncharacterized peak from a halogen-specif- ic detector may represent a manmade com- Another question which could he asked of pound, Historically, halogen-containing or- the data bank is whether there is a peak that ganics such as DDT, PCBs, polybrominated is increasing with time. In other words, does biphenyls (PBBs), and kepone have caused sampling over time indicate that a compound major pollution crises. Therefore, such infor- is becoming more concentrated? Such a find- mation may be sufficient to focus attention on ing would flag the compound as one which a given peak from a halogen-specific detec- merits attention, Had this type of monitoring tor. program been in effect and this question been asked earlier, widespread pollutants such as Uncharacterized substances may or may PCBs could have been detected long before not be dangerous to human health if con- they were, sumed. But no assessment of toxicity can be made until the compounds are identified. To Other information from the chemical labor- perform the chemical and physical tests nec- a tory car-r draw attention to an uncharacter- essary to identify even one uncharacterized ized substance deserving attention. For ex- peak may cost from $10,000 to $100,000, and ample, gas chromatographic detectors exist in some cases even more.

DETECTING AND QUANTIFYING TRACE METAL CONTAMINANTS

Trace metals pose many of the same prob- ples, thus eliminating the extraction step (4). lems that plague the analyst monitoring for But these instruments cannot detect all met- organic chemicals. However, the number of als at environmental concentrations. trace metals is much smaller. If only the total concentration of a metal is sought, more rig- Depending on what instrumentation is used orous extraction procedures can be em- in the final analysis, some cleanup of the ex- ployed. But some metals, such as mercury, tract may be required. The cleanup step is form organic complexes--for instance, meth- not nearly as frequent with trace metals, ylmercury--which give rise to a subset called however, as with organic chemicals except if metallo-organics. These metallo-organics are a metal lo-organic complex is sought. For not stable under harsh conditions and are metallo-organic compounds cleanup and sep- often changed during rigorous extraction pro- aration procedures may be similar to those cedures. for organic compounds. After extraction and possibly cleanup, the sample is ready to be Analysis analyzed by an instrument to determine which metals, and how much, are present. Analysis for trace metals usually requires There are a variety of instruments available extraction and cleanup before the quantita- for the determinations. tive analysis can be performed. Extraction often involves destruction of the sample’s or- Probably the most commonly used instru- ganic structure to release the metals from the ment is the atomic absorption spectrophotom- solid or liquid food. There are a few analyti- eter (AA). The cost of this instrument is nom- cal instruments such as X-ray emission spec- inal ($ 15,000 to $25,000), and i t can be oper- trographs that can accept solid or liquid sam- ated by a well-trained, motivated technician 98 . Environmental Contaminants in Food with a high school chemistry and physics This is analogous to uncharacterized organic background working under the supervision of monitoring where many unsought compounds a chemist. One drawback of this instrument can be found in the analysis of a food extract. is that only one metal can be analyzed at a Given the proper data systems, these “other” time. metals can be tracked, and perhaps pollution crises can be averted by detecting anomalous In contrast, more sophisticated instrumen- concentrations early enough. tation such as X-ray emission, proton-induced X-ray emission, or plasma emission spectrom- Speciation of Trace Metals eters can cost in excess of $100,000. With these instruments, 20 to 60 elements can be Often the active or functioning forms of ele- determined simultaneously. These more so- ments must be identified and measured. The phisticated instruments require more highly need for the development of analytical meth- trained personnel to operate and maintain. odology to identify and quantify the chemical Table 19 summarizes the techniques avail- form of metals found in the environment— able for qualitative and quantitative analysis i.e., chemical speciation—is pointed up by the of trace metals. More information can be following examples. found in appendix H. ● Although arsenic is toxic, the plus three Application to Investigatory oxidation state, As (III), is more toxic Monitoring than the plus five state, As (V). The com-

pound arsine, AsH3, is perhaps the most It is important to note that most AA units toxic chemical form of arsenic. can detect only the specific element (metal) that they are set up to analyze. By contrast, ● Alkyl (organic) mercury compounds pose the gas chromatography may sometimes detect greater propensities for bioaccumula- the presence of organics other than those tion and the associated health effects being sought. As a result, metals are more than do the more common inorganic likely than synthetic organics to escape detec- forms of mercury (but these also can tion simply because they are not programed vary greatly in toxicity). for analysis. Chemical forms or states are an important With the rapid development of lower cost determinant of a trace metal’s toxicity in bio- techniques for detecting more than one ele- logical systems. The selection of elemental ment at a time, it would be possible to obtain analysis techniques capable of specifically data on trace metals in foods that were pre- measuring those chemical forms most impor- viously not sought and therefore not obtained. tant from a biological-effects viewpoint

Table 19.—Techniques Available for Qualitative and Quantitative Trace Metal Analysis Ch. Vlll—Monitoring Instrumentation ● 99

should be a primary objective of any environ- of the important (threshold) concentration mental monitoring program. levels for various elements typically has been The status of the present analytical tech- rather crude. The most important current re- nology is inadequate for this purpose in most search need related to trace metal monitoring routine monitoring programs. As a result, the is the development of methods to measure the functioning forms of most elements cannot be species rather than the total amount of trace adequately studied. Thus the determination metals.

ANALYSIS OF FOODS

Analysis temperatures have not been excessive. If there is concern that insoluble particulate Measuring radioactivity in foods is a phys- may be present, it is necessary to use more ical process. It is most efficient when the drastic methods such as fusion to bring the radioactivity from a relatively large sample entire sample into solution. can be placed close to a detector. This means that direct measurements of radioactivity in In addition to bulk reduction, radiochemi- bulk samples are only useful when levels of cal separations are required to isolate the contamination are relatively high. Most meas- desired radionuclide both from the remaining urements are preceded by preparation and bulk constituents and from other radionu- possibly chemical separation to reduce the clides that would interfere in the measure- bulk of the material and to improve the effi- ment. It is also necessary to convert the final ciency of the measurement. product to a form suitable for presentation to the counter. This may involve electrodeposi- Foods generally have a high water content. tion, precipitation, or other processes. The primary method of reducing bulk is freeze-drying or drying at room or at elevated The actual mass of radionuclide that is temperatures. Most of the radionuclides of in- measured is almost always extremely small. terest are not volatile under these conditions. As a result, many of the normal chemical re- Only elements such as tritium or iodine may actions used in analytical chemistry to isolate experience losses. The dried materials can be an element are not appropriate. For instance, reduced further by ashing at elevated tem- precipitates may not form. Therefore, it is peratures, by cold-ashing with oxygen, or by common to add a few milligrams of a carrier wet-ashing with oxidizing acids. Dry-ashing is material, preferably the inert form of the the simplest process, but it is most likely to same element. When such an inert form does lead to loss of volatile elements. With care not exist, it is frequently possible to use simi- even cesium, polonium, or lead can be re- lar elements as a carrier (for example, substi- tained. The other processes should not result tuting barium for radium). The inert form in losses of elements of interest, except for then follows normal chemistry, carrying the iodine, tritium, and carbon. radionuclide with it. It is also worth noting that even when the separation technique does Another way to reduce bulk is to treat not depend on the mass of element present, a either the original sample or the dried or carrier may still be useful in preventing un- ashed material with acids or other solvents wanted coprecipitation or absorption on and thus remove the desired elements from glassware. the bulk of the sample. This requires consid- erable testing beforehand to be certain that In the chemical separations there is consid- the process operates in the desired manner. erable use of classical analytical chemistry The radionuclides of interest in ashed foods based on precipitation. In most cases, precipi- should be soluble in strong acids if ashing tation will be the final collection step in 100 “ Environmental Contaminants in Food

preparing the desired radionuclide for count- It is possible, however, to set a particular ing. In other cases the techniques of ion ex- total activity level as a screening level for a change, liquid extraction, distillation, and specific food, If the measured value is below electrolysis may be required to prepare the the screening level, no analyses for individual sample for counting. radionuclides are performed. In such a case, the measurements should be considered in- In certain cases, the total amount of a sam- ternal data and the numerical results should ple may be limited, and analysis for several not be published, Any report should merely radionuclides may be required. Procedures list the samples as having activities below the should be available for the sequential analy- stated screening level. Table 20 summarizes sis of single samples, even though separate the typical limits of detection for radionuclide samples are used for routine work. measurement. The measurement method used depends on Alpha Emitters.—The measurement of the type of radiation, the form of the sample, alpha activity is best carried out on a very and (to some extent) the amount of radioac- thin sample to avoid self-absorption of the tivity. The selected analytical procedure alpha particles. Most metals are electro-de- should be designed so that the sample is posited onto small smooth discs of stainless brought to a suitable form for the equipment steel, nickel, or platinum. Evaporation of pure and conditions that exist. solutions is used in some cases. The measure- The major emitted radiations are generally ment of the total alpha activity can be con- grouped as alpha, beta, and gamma. Alpha ducted either in thin-window counters or by radiation is characteristic of the natural and scintillation counting with zinc sulfide phos- artificial radionuclides of high atomic weight phor. Both techniques are highly efficient, but and consists of energetic particles with very the scintillation method can give a considera- low penetrating power. Its hazard is signifi- bly lower background with a consequently cant only within the body, where alpha-emit- lower limit of detection. ting nuclides can irradiate specific sensitive A somewhat less-sensitive technique is the tissues. Beta radiation appears in both natu- liquid scintillation spectrometer described in ral and manmade radionuclides, and consists the next section. of electrons possessing kinetic energy and modest penetrating power. Gamma radiation Alpha spectroscopy provides two other al- is pure electromagnetic radiation and is ex- ternatives: the Frisch grid ionization chamber tremely penetrating. Thus, it is hazardous ex- or the silicon diode solid-state detector. The ternally as well as when it is present in the Frisch grid unit can handle large area sam- body. Gamma radiation can be directly meas- ples but is slightly poorer in resolving closely ured in foodstuffs, while alpha and beta emit- separated energies. On the other hand, the ters generally must be separated from the silicon diodes available are all quite small bulk constituents of the sample before they and can only count samples of about 1-cm di- can be measured. ameter with high efficiency. It is possible to measure the total gamma, Beta Emitters.—Beta counting is a little total beta, or even the total alpha activity in a more flexible in the mass of material that can sample of food. Unfortunately, such data are be present at the time of counting. This is true valueless in estimating human exposure. The for higher energy beta emitters but carbon-14 accuracy of such estimates is very poor be- and tritium present a problem. Since each in- cause natural potassium usually interferes, dividual beta emitter gives off particles with and the chemical and radiation character- a range of energies from zero to a character- istics needed to evaluate possible hazards are istic maximum, beta spectrometry is not pos- not known. sible for food samples. Ch. VIII—Monitoring Instrumentation c 101

Table 20.—Typical Limits of Detection for Radionuclide Measurement

LLD (dpm) per sample for various counting times Counter Counter Chemical Nuclide efficiency % background cpm yield % 10 min 40 mln 400 mln 1,000 mln Amerlclum.241 25 0.001 60 03 015 0.05 003

Ceslum-144 22 04 75 6 3 09 06

Cesium-137 30 04 75 4 2 06 04

2 006 75 30 14 45 3

Cobalt-60, other activated products 1 005 — 100 50 15 10 Tritium 30 5 75 15 6 2 1.5

Iodine- 131 25 0.4 80 5 2 0.7 0.5

4 03 80 25 13 4 25

Phosphorus-32 45 03 60 3 1.5 0.5 0.3

Plutonium-239. -240 25 0005 85 06 0.3 0.09 0.06

Radium-226 56 02 90 1 0.6 0.2 0.1

40 0.001 85 1 0.5 0.15 0.1

Strontium-90 45 0.3 80 2 1 0,4 0.2

Thor! urn-230, -232 25 0001 70 0.3 0.1 0.04 0.03

U-Isotopic 25 0.005 75 0.7 0.3 0.1 0.07

SOURCE N H Harley Analysts of Foods for Radloactlvlty OTA Work Ing Paper 1979

The available counting equipment for used whenever the sample can be made mis- quantitative measurement includes geiger cible with the scintillating solution itself. This counters, proportional counters, and scintil- can even be done with solids by suspending lation counters. The geiger counter is rela- them in a scintillating gel. The advantage of tively inexpensive and requires only simple liquid scintillators is their high efficiency, electronics but is not popular and is generally even for the low-energy emitters carbon-14 not available as a counting system. The thin- and tritium. Scintillation systems for counting window proportional counter is used widely precipitates are not commercially available and has both reasonably high efficiency and at present. There are, however, many liquid low background. For low-level samples the scintillation systems on the market, most of background can be further reduced by anti- them with automatic sample changers. These coincidence techniques. These add to the use high levels of activity and short counting complexity and cost of the system but are times. The better systems also have a provi- sometimes necessary. sion for rather crude spectrometry. They can distinguish qualitatively and quantitatively among carbon-14, tritium, alpha emitters, Scintillation counting can be performed in and higher energy beta emitters. two ways. Solid scintillators can be used for counting chemical precipitates collected on Gamma Emitters.—Gamma rays are so filter papers, and liquid scintillators can be penetrating that the detector must have a 102 ● Environmental Contaminants in Food considerable mass to absorb enough energy balance requires a weighing of cost, man- to produce a response. Since energy absorp- power, and quality. tion is required for spectrometry, solid detec- tors are most useful, Sodium iodide is a popu- Applications to lnvestigatory lar detector because crystals can be fabri- Monitoring cated in large sizes and are transparent to the scintillations produced by radiation. So- The monitoring of foods for radioactivity dium iodide has high efficiency but poor en- should not be considered a primary defense ergy resolution and is now applied to samples against human exposure. The first indication where some separation has taken place. The of hazards should always come from informa- advantages are that samples of food can tion on releases or from measurements of ra- often be counted directly without chemical dioactivity in air or water. Once the existence preparation. Milk is a good natural example, of contamination has been established, foods as the metabolism of the cow removes most can be analyzed to evaluate the potential haz- gamma emitters other than isotopes of ce- ard to man. sium, iodine, and naturally radioactive Knowledge of the source of radioactive potassium. contamination gives a good indication of the More complex spectra can be resolved nuclides that can be expected in the sample. with the solid-state germanium diode detec- This information helps in planning the anal- tor. Interferences from radionuclide impur- ysis, since requesting a complete analysis for ities are greatly reduced compared to spectra all radionuclides or even for all types of from sodium iodide detectors. The efficiency radioactivity in a single sample would lead to of the diode is low and, for many analyses, a a lengthy and expensive operation. Indeed, spectrometer can only be used for one meas- monitoring for suspected contaminants is urement a day, Another disadvantage is that more applicable than the uncharacterized the detector must be kept at liquid nitrogen monitoring. temperature to maintain its detection capa- The general groups of nuclides that maybe bility. encountered include those that occur natural- Diode spectrometers may also be used to ly such as radioactive potassium and mem- measure the low-energy gamma-rays that ac- bers of the uranium and thorium series, arti- company alpha emission. This allows direct ficial fission products such as transuranic measurement in some environmental sam- elements, and other activation products that ples, but the levels in foods have not been result from nuclear weapon explosions and high enough for this technique. nuclear reactor operations. General Requirements.—The choice of a Fission products are a very complex mix- counting procedure depends on the precision ture when they are formed, but the short- required. The relative precision of a quantita- lived radionuclides die out rapidly and the tive counting measurement, in turn, is in- mixture becomes simpler within a few days. versely proportional to the square root of the The transuranics (plutonium, americium, etc. number of counts obtained. Thus any im- formed by activation of the basic fissionable provement in precision must be obtained by material) are of some interest because of increasing the number of counts. This can be their high toxicity when incorporated into the done by using larger samples, by counting for body. Present evidence, however, indicates longer times, or by using counters with higher that their uptake through the gut is relatively efficiency, A secondary improvement is possi- small and that dietary intake is not a signifi- ble for low-activity samples by decreasing the cant problem. The other activation products background, Each of these improvements has are frequently elements that make up steel or some drawback, and selection of the optimum other metal containers or structural ele- Ch. VIII—Monitoring Instrumentation ● 103 ments. Radioactive manganese, chromium, Almost all radionuclides of interest in cobalt, zinc, and iron are particularly com- cases of contaminating events now exist in mon and result from interactions of the mate- foods in small but measurable quantities. rials with neutrons released in the nuclear Short-lived nuclides are the exception and the reaction, Contamination of foodstuffs with transuranic elements are only present at single nuclides is extremely unlikely, and levels that require considerable effort in more than one member of any group will analysis. Since most of the radionuclides are probably be present in any sample. already present in foods, measurements In contrast to most other pollutants, the ef- made for background information should pro- fects of radiation are considered to have a duce a numerical answer, not merely an in- linear response regardless of the level, Thus dication that the amount is less than some there is no threshold and no absolutely safe pre-set value. This accumulation of back- limit. The analytical significance of this is ground data provides a valuable baseline for that the lower limits of detection for radioac- evaluating hazardous levels following a con- tive substances have been brought down to taminating event. The natural activity data very low levels. The simple yes-or-no testing are equally valuable, since the amount of in- for acceptability that satisfies regulators for formation on food concentrations is presently many other pollutants in foods cannot be insufficient for valid comparisons with man- used. made radioactivity.

ESTIMATED COST TO EQUIP A LABORATORY INVESTIGATORY MONITORING

Table 21 illustrates estimated costs, re- ber of samples to be collected and analyzed. quired space, and estimated downtime for a Information of this sort could be obtained laboratory designed to conduct investigatory through a pilot project to investigate the monitoring. In addition, the number of sam- feasibility of the two investigatory monitoring ples the system would be able to analyze in a approaches. This would determine the num- year are estimated. ber of laboratories, the number of people and These figures do not reflect the total costs their salaries, and costs of equipment mainte- for establishing a national system of inves- nance, supplies, and training. Once the sys- tigatory monitoring. Such a cost estimate tem is setup and running, some savings might would require information on the total num- be realized through automation.

54.5>5 0 - 79 - B 104 ● Environmental Contaminants in Food

Table 21 .—Estimated Costs to Equip a Laboratory to Conduct Uncharacterized Monitoring

Estimated Approximate down Analytical instrumentation capital cost Required space Samples per year time Synthetic organics 1. Small, high throughput, GC-MS- $100,000 5,000-6,000 ft. ’ 200-300 for 20- 50% data system with automated liquid laboratory supporting uncharacterized environmental Injection device. facilities and equipment contaminants as welI as known are required, environmental contaminants 2. Electron impact/chemical $200,000- 20- 50% Ionlzatlon-equipped high- $300.000 resolution mass spectrometer data system with automated injection device 3. Gas chromotographic flame $15,000 20-50 % ionization detector, electron capture detectors (additional chromotographic systems may be required). 4 Liquid chromotograph interfaced $20,000 20- 50% to mass spectrometer system 5. Central data management system $75,000- 20- 50% $1,000,000 6. Cold storage and processing $500! 000 facilities Synthetic organics subtotal $1,585,000- $1,935,000 Trace metals 1. Inductively coupled plasma $100,000 Approximately 2,500 ft.’ 3,000-5,000 3-15% of laboratory equipped Multielement atomic emission with supporting facilities system and equipment are required for items 1-4 2. Flame and furnace atomic $25.000 3- 15°/0 absorption spectrophotometer (single element mode) 3. Electrochemical instrumentation 3-15% 4. Central data management system. 3-15%

Trace metals subtotal.

Radionuclides 1. Four position alpha spectrometer $21,000 Approximately 1,500 ft. ’ 1,000-4,000 and detectors, multichannel laboratories equipped analyzer and output with supporting facilities and equipment are required 2. Germanium diode gamma $100,000 5 % spectrometer wit h detector, shield, electronics, PDP-11 computer and output 3. Four general purpose proportional $10,000 5 % counters, 4. Liquid scintillation spectrometer, $18.000 5% automatic sample. or alpha- counting capabilities Radionuclides subtotal $149,000 Total...... $1,934,000. $2,309,000

SOURCES Adapted from J L Laseter Approaches to Monltonng Environmental Contaminants In Food OTA Work!ng Paper 1978 R K Skogerboe Analytical Sys terns for the Determ!natlon of Metals {n Food an(j Water Supplles OTA Working Paper 1978 and N H Harley Analysls of Foods for Radloactlvlty OTA Work!ng Paper 1979 Ch. Vlll—Monitoring Instrumentation ● 105

1. Butler, P. A. Environmental Protection Agen- vironmental Contaminants in Food, OTA cy, Gulf Breeze Environrnen tal Research Lab- ~~[)rking paper, 1978. oratory, personal communication, 1978. 4, Sko,gerboe, R, K, “Analytical System for the 2. Anonlrnous. “Trace Organic Analysis,”” En- Determination of Lletals in Food and Water vironment~]f Science anci 7’ec~nology, 12(7): Supplies, ” OTA Working Paper, 1978. 757, 1978. 5, Harley, N. Ii. “Analysis of Foods for Radio- 3. Laseter, J. L. “Approaches to Monitoring In- activity,”’ OTA W’orking Paper. 1979. . .

Chapter IX Options

. Chapter IX 4 Congressional Options

The present system of controlling environmental contaminants in food con- sists of two parts: regulatory procedures to set and enforce limits for environmen- tal contaminants, and monitoring procedures to detect lots of food in violation of established limits. Each State has authority for regulating food grown and con- sumed within its boundaries. The Federal Government is responsible for regulat- ing food in interstate commerce. Congress can choose to maintain this system. But if it wishes to put greater emphasis on protecting consumers from contaminated food, one or more of the op- tions discussed below could be adopted. None of these options (except for the first) are mutually exclusive.

The current regulatory approach to con- people against exposure, In other words, ac- trolling environmental contaminants in food tion levels and tolerances permit a certain involves the setting of action levels (and occa- level of contaminant to be present in food. Un- sionally tolerances), coupled with regulatory less action levels or tolerances are reduced, monitoring for known [and a few suspected) little effort will be made to eliminate the con- contaminants. Food containing an amount of taminant, The threshold concept on which ac- a contaminant that exceeds the action level tion levels and tolerances are based—that or tolerance can be identified through such there are exposure levels to toxic substances monitoring. This food is then removed from below which there are no effects on health the marketplace. Public exposure is thus —is being increasingly challenged (especially theoretically limited to those foods containing when carcinogens are involved). quantities of contaminants that fall under prescribed action levels or tolerances. If high action levels or tolerances are es- tablished, exposure is not reduced. Lowered Pros: There are two principal advantages action levels and tolerances may reduce pub- to maintaining this system. No additional ap- lic exposure. But even low limits are set on propriations or legislation are required. No the basis (among other things) of nationwide changes in existing regulations are neces- per capita consumption of a particular food. sary. Contamination problems, however, may be lo- Cons: There are a number of disadvan- calized and further influenced by regional tages in retaining the current system. The food consumption patterns. Thus, a local pop- time needed to identify an environmental con- ulation may be highly exposed to a toxic sub- taminant in food and take corrective action stance although the tolerance, based on na- would not be shortened, The people would tional consumption, may be low. continue to be exposed until a contaminant Moreover, there is no requirement for re- was detected and identified, and an action view once an action level has been estab- level put into effect, lished. An agency is under no pressure to ac- Moreover, a c t ion levels and tolerances tively seek out new data that might alter a tend to institutionalize rather than protect prescribed level,

109 110 ● Environmental Contaminants in Food .

Tolerances and action levels have other ficult to prevent some shipments from reach- weaknesses. They are often not easily applied ing the marketplace. when the environmental contaminant in- volved is a suspected carcinogen (such as Finally, there are procedural problems in polychlorinated biphenyls (PCBs)). Further- the present system. States have no clearly de- more, the time required to perform a com- fined authority to which they can turn when plete chemical analysis for contaminants in they suspect environmental contamination of nonprocessed foods such as fish makes its dif- food.

OPTION 2-AMEND THE FOOD, DRUG, AND COSMETIC ACT

The Food, Drug, and Cosmetic Act has been ance proposal in the Federal Register, along amended several times since its passage in with (as required by recent court rulings) the 1938 to deal with new food regulatory prob- rationale and factual data underlying the lems. Environmental contamination of food is proposal. The public can then respond to the now a national problem which Congress has proposal with written comments. FDA may never directly addressed through legislation. also hold legislative-style public hearings to Thus, Congress could choose to give regula- allow presentation of oral arguments and tory agencies more guidance by clarifying its evidence. position on environmental contaminants in food. After considering all of the comments, FDA publishes a final rule (in this case, a toler- An amendment to the Food, Drug, and Cos- ance). This final rule includes explanation of metic Act could include one or all of the fol- any changes from the original proposal and lowing points. None are mutually exclusive. responses to factual points raised by the pub- lic. Most agencies now use this model for Option 2A-Simplify rulemaking. It is a process that can be car- Administrative Procedures ried out expeditiously with modest investment of an agency’s resources. Under current law and regulations, the Food and Drug Administration (FDA) sets an Pros: Adoption of this streamlined rule- action level for a contaminant soon after an making procedure would reduce the time and interstate shipment of contaminated food is expense now involved in setting a formal tol- discovered. FDA will then presumably launch erance. It may encourage FDA to move from the elaborate rulemaking proceedings that action levels to tolerances, thus bringing culminate in the establishment of a tolerance more public participation into the process. (under section 701(e) of the Food, Drug, and Cons: Because action levels are adminis- Cosmetic Act). In reality, the costs and delays trative guidelines, they can easily be changed involved in the complex rulemaking proce- dures now required for the adoption of toler- when new scientific information becomes ances have discouraged FDA from moving available. Even if FDA comes to use a simpli- from the first (action level) to the second (tol- fied procedure to set tolerances, it may still erance) state of the process. be slow to revise them in the light of new data. Congress could amend the Food, Drug, and Cosmetic Act to simplify the administrative Option 2B--Require the procedure through which tolerances are set. Establishment of Tolerances The changes could be modeled after section 553 of the Administrative Procedures Act. Congress could amend the Food, Drug, and This process involves publication of a toler- Cosmetic Act to require the establishment of Ch. IX—Congressional Options ● 111

a tolerance within a specific time after the require up to a year for generating the neces- setting of an action level. sary data. Thus, weighing the costs is best Pros: This change would encourage FDA to suited for setting tolerances. not action levels. gather additional information on a contami- nant’s toxicity and the public’s exposure. It The advantage in including costs in toler- would speed up a process. that now operates ance-setting decisions is that adverse eco- under no deadlines. And it would result in de- nomic impacts are likely to be reduced. The finitive tolerances that FDA could enforce disadvantage is that tolerance levels are like- with less concern about judicial questioning. ly to be higher than would be the case if costs were not considered. Cons: This option, however, would substan- tially increase FDA’s workload unless toler- Pros: By clearly defining to what extent ance-setting procedures were simplified. In- costs can enter into tolerance-setting deci- deed, it has been the costs and delays of the sions for environmental contaminants in food, current rulemaking process that have de- Congress would eliminate ambiguities of in- terred FDA from moving from action levels to terpretation and provide clear guidance to tolerances. Thus, tolerances should not be re- FDA. quired without also simplifying present rule- Cons: FDA has interpreted the Food, Drug, making procedures, and Cosmetic Act as allowing cost considera- tions, Legislation requiring economic assess- Option 2C--Clarify the Role of ment could limit FDA’s discretion to weigh Economic Criteria the costs of food lost if it judges the situation Congress could amend the Food, Drug, and warrants such a treatment. Cosmetic Act to clarify to what extent eco- nomic criteria can be used in setting toler- Option 2D--EstabIish Regional ances. The Act does not specify that costs Tolerances (the adverse economic effects) of a proposed tolerance be considered when setting a toler- This option would give FDA the flexibility ance. FDA, in practice, does weigh the cost of to set different action levels or tolerances for food lost when establishing a tolerance. different regions, based on expected levels of exposure, regional levels of contamination, The Food, Drug, and Cosmetic Act could be and eating patterns. amended to prohibit FDA from considering costs when setting a tolerance. Prohibiting Pros: Action levels and tolerances may not any economic assessment would ensure that be set low enough to protect those popula- public health would be the first priority in set- tions that are most highly exposed, previously ting a tolerance. exposed, or most vulnerable. States may not exercise their authority to set tolerances Conversely, the Act could be amended to which are more restrictive than the Federal require FDA to weigh the costs against the tolerance because of budget limitations, in- benefits of a proposed tolerance. Requiring adequate information, or political pressures. FDA to evaluate the economic consequences FDA can provide guidance to States and sug- of a tolerance would give FDA clear authority gest more restrictive tolerances or warnings to weigh such estimated effects together with to the public, but FDA has no authority to in- the potential health risks when establishing a tervene if the contaminated food does not tolerance. Congress could require FDA to enter into interstate commerce. gauge only the primary costs (as is now done with food lost) or all associated costs (food Cons: Regional tolerances would compli- lost, employment impacts, distributional and cate monitoring and enforcement programs. indirect effects) for a proposed tolerance. Regional tolerances might also be viewed as The techniques available for estimating costs Federal infringement on State authority. 112 ● Environmental Contaminants in Food

OPTION 3-ESTABLISH AN INVESTIGATORY MONITORING SYSTEM

Environmental contaminants could be de- gies, and reduce public exposures to environ- tected earlier in the food chain by improving mental contaminants as much as possible. present environmental monitoring capabil- ities —establishing an investigatory monitor- Option 3A-Establish a National ing system while maintaining current regu- Monitoring System for Suspected latory monitoring programs. Environmental Contaminants Congress could set up a national investi- Suspected environmental contaminants gatory monitoring system that monitors for are substances that are most likely to enter either suspected or uncharacterized environ- the food supply and pose potential health mental contaminants. A system combining hazards, Lists of such substances could be elements of both approaches could also be es- drawn up for organic chemicals, trace met- tablished. Since any of these monitoring ap- als, and radioactive substances in order that proaches would require some research and they be monitored in the food, Various cri- development before a fully operational sys- teria such as toxicity, volume of production, tem could be devised, Congress could choose occurrence in the environment, persistence, to create a pilot program. Such a program and biodegradability could be used in putting would spur research and development and together the lists. assess the feasibility and cost-effectiveness of the various approaches. Pros: Present food monitoring efforts are not designed to detect new environmental The investigatory monitoring systems dis- contaminants in food. The limited amount of cussed in this assessment would call for dif- monitoring for suspected contaminants that ferent sampling and quality control proce- does exist is primarily concerned with trace dures. The development of these procedures metals. But far more of this type of monitoring is as important as the development of moni- is needed to anticipate new contaminants in toring technology (some of it still in the ex- food. perimental stage). Indeed, there is no com- prehensive investigatory monitoring system Cons: To draw up such lists successfully, for toxic substances in food and the environ- considerable information is needed. Sub- ment at any level of government. stances for which there is little or no data Consequently, Congress might opt for a would automatically be given low priority. In pilot project to assess the capabilities and re- large part, the makeup of the lists would de- source requirements of various national mon- pend on scientific judgments. However, scien- itoring systems instead of mandating a par- tific judgments often vary (or even conflict). ticular monitoring approach. Such a pilot Thus, the reliability of the lists may be in project would focus on the monitoring of or- question, ganic chemicals, inorganic, and radioactive Priority lists of trace metals and radioac- substances. It would determine the technol- tive substances, which are limited in number ogy, sampling, and quality control needs for and already well-investigated, would be more monitoring these three toxic substance cate- reliable than a list of organic compounds. gories in food, air, water, and soil. The broad- There are thousands of such organic agents er purpose of the project would be to develop manufactured, And there is little toxicologi- a comprehensive monitoring program that cal or environmental information available on would meet the Government’s regulatory a great many of them. needs, provide data to make cost-effective- ness assessments of the alternative strate- If such lists were developed, the number of substances monitored would necessarily be Ch. IX—Congressional Options c 113

limited. Standards would be set up for deter- tion is developed, regulatory agencies could mining which substances would get priority. then take appropriate action. Of course, there would be no certainty that unlisted substances might not get into the This approach tries to create a mechanism food supply and threaten human health. The for quantitatively measuring uncharacter- cost of such a monitoring system would de- ized substances in food. Proper guidelines are necessary since the cost of quantitatively pend on how many substances were being identifying one substance can range from traced as well as the expenses of equipping and staffing laboratories. $10,000 to $100,000. Moreover, this monitor- ing approach requires sophisticated equip- ment for analyzing food samples. Also, the in- Option 3B--Establish a National formation generated by the analyses has to Monitoring System for be computerized. Computer technology makes it possible to correlate and interpret chemical Uncharacterized Environmental data to provide a continuous surveillance of Contaminants the levels in food. Uncharacterized monitoring would be de- Pros: The uncharacterized monitoring ap- signed to detect substances that are: lacking proach is in the research stage at several lab- toxicity data for potential human risk, not oratories in the country. If it is successfully known to be present in food, and not even developed, this type of monitoring would re- known to exist in the environment. This kind duce the time during which the public is ex- of monitoring is most needed for synthetic posed to high concentrations of uncharacter- organic chemicals. The purpose is to detect ized environmental contaminants in food. Ke- changes in the levels of various synthetic pone, for example, was polluting the James organics in environmental or food samples River, and people were eating kepone-con- over time, taminated fish for several years before the chemicals presence was discovered. With an Scientists would not necessarily know the uncharacterized monitoring system, kepone identity of individual substances they were may have been detected years earlier. monitoring. But if the concentration of a par- ticular compound substantially increased, Cons: This combination of sophisticated in- they could analyze it further to establish its struments, dependence on computers, and identity, The literature would be reviewed on highly trained personnel is expensive. And the substances toxicological properties. Or the “hardware” needed is generally not perhaps the substance would undergo toxico- found in Federal or State monitoring labora- logical testing. Depending on what informa- tories.

OPTION 4--IMPR0VE FEDERAL RESPONSE TO NEW CONTAMINATION INCIDENTS

All of the major food contamination inci- ance, Congress could choose to designate a dents have been marked by confusion. This lead agency or establish a center for the col- stems from the involvement of three Federal lection and analysis of data. agencies—FDA, the U.S. Department of Agri- culture (USDA), and the Environmental Pro- Option 4A-Designate a Lead Agency tection Agency (EPA)—in the monitoring and regulation of environmental contaminants in The problem of conflicting Federal assist- food. To cut down on confusion and to im- ance efforts was dramatized by several re- prove delivery of Federal technical assist- cent incidents including the most recent PCB 114 ● Environmental Contaminants in Food * contamination. This sort of situation is not human health impacts. It would have no regu- unique, Official reactions following the poly- latory function. brominated biphenyl (PBB) episode in Michi- The group would be able to mobilize within gan and the kepone incident in Virginia were 24 hours. It would be the lead Federal organi- similar. zation that affected States could initially con- The lead agency would serve as a clearing- tact. In the wake of an episode, the team house for all information coming from and go- could be responsible for followup scientific ing to States. FDA would be the most likely research that would lay the groundwork for candidate for lead agency when food contam- epidemiological studies of the exposed popu- ination is suspected. lation. It would also be able to conduct a Pros: With a clearly delineated lead agen- range of short-term toxicological tests to de- cy, States suspecting contamination of food termine the mutagenicity and potential car- would have one reliable source of technical cinogenicity of uncharacterized substances. assistance. Conflicts of opinion would be set- It would be able to examine an exposed popu- tled internally, and public statements and lation for any adverse health effects from in- gesting a particular contaminant in food. technical assistance provided with less am- biguity. The new technical center, in essence, Cons: Designation of a lead agency might would be similar to the Center for Disease Control (CDC). It would have the same capa- decrease the amount of technical expertise bilities in chemical epidemiology as CDC has made available to the States. It would require in infectious disease. The new center could the lead agency to develop new agreements with the other two agencies, During a food be given responsibility for investigating all contamination crisis the lead agency would toxic substance problems in the United need to coordinate responses from the other States, not just those limited to food. If it were placed under CDC, it could build on work now agencies. underway in that organization’s Environ- Option 4B--Establish a Center to mental Hazards Unit. Furthermore, the cen- ter would be able to capitalize on the long- Collect and Analyze Toxic standing relationships between CDC and the Substances Data States. Since the center’s sole mission would be the development of information on environ- Major delays in protecting the public from mental contaminants, it would not be linked environmental contaminants in food now re- to regulatory decisions. sult from the time-consuming process of gen- erating sufficient data on a substance’s toxic- As an alternative, the center might be lo- ity and dispersal in the food supply. Congress cated within FDA’s National Center for Toxi- could overcome this problem by setting up a cological Research or another Federal labor- new technical center. atory with existing analytical capabilities. It could then use the laboratory’s equipment Such a center would be able to rapidly as- and trained personnel for investigative ana- semble technical teams skilled in the identifi- lytical chemistry and toxicology. cation and analysis of organic, inorganic, and radioactive substances. A team would consist It would be better not to locate the center of a multidisciplinary group of experts, in- and the investigation teams in a regulatory cluding chemists, toxicologists, food and ani- agency with direct responsibilities for food. mal scientists, epidemiologists, biochemists, This would ensure that there is no possible biostatisticians, medical doctors, and others. conflict of interest between its job of factfind- Its mission would be to identify the cause of ing and the agency’s responsibility for regu- an actual or potential contamination incident, lating, By distancing itself from the regula- and assess the possible environmental and tory process, the new center team would Ch. IX—Congressional Options ● 115

strengthen its credibility with the media and Reaction time would be shortened and dup- the public. lication of effort reduced by a technical or toxic substances investigation team which Pros: Accurate scientific information on could react quickly in emergencies. the nature and extent of food contamination has to be generated quickly when an incident Cons: There is no assurance that a special occurs. Individual States often lack the scien- team could generate information more quick- tific expertise to develop such data. The Fed- ly than is now the case. If the center were not eral Government does have the necessary ca- part of a regulatory agency, it might not only pabilities, but the resources are scattered in grow out of touch with the needs but also several agencies with differing areas of au- could duplicate the investigatory work of the thority. Such a dispersal of expertise hinders agency. Furthermore, the potential for an the gathering of urgently needed information adversary relationship exists. following a contamination episode. Jurisdic- tional problems crop up, and States find they have to deal with more than one agency. The It could be argued that better coordination resulting delays slow the making of necessary among FDA, USDA, and EPA would accom- health, environmental, and regulatory deci- plish the same goals without the expense of sions. establishing a new research center. Appendixes Appendix A

Substances Whose Production or Environmental Release Are Likely to Increase in the Next 10 Years*

by Clement Associates, Inc.

OTA requested Clement Associates to develop a list identifying new chemical sub- stances likely to be manufactured in the near future and known chemicals likely to pose an increased burden to the environment because of increased production, new applica- tions, or new technological developments, including new energy technology. For these projections, several factors were considered, including market trends, Federal regula- tory activity, available substitutes for recently banned or restricted chemical substances, and consumer needs. During the development of the approach to this phase of the project, certain prob- lems and limitations became apparent. The nature of chemical substances under re- search and development but not yet introduced to the market is usually closely guarded proprietary information and therefore not available. In addition, there are no data sys- tems which bring together chemical information to facilitate the retrieval of necessary data. An approach was developed to obtain a maximum amount of information in a lim- ited amount of time. Sources, including personal contacts, were identified for information on new chemi- - - cals or chemicals whose production, use, and release to the environmentwere likely - to in- crease sharply because of future needs. These sources of information are listed in the reference list. Following are the list of chemicals and a brief indication of why these chemicals were selected.

CHEMICALS LIKELY TO HAVE SIGNIFICANTLY ENVIRONMENTAL RELEASE IN THE NEAR FUTURE

Chemicals Whose Production and Use derivatives are the most versatile of the soluble Are Likely to Increase Sharply polymers. Others are polyethyleneimine, polyvi- nyl sulfonic acid, polyacrylic acid with such che- Because of Future Needs lating groups as thiourea, 8-hydroxyquinoline, SoIuble Polymers iminodiacetic acid and hydroxyciniline, and poly- mers based on acrylic acid. (C&EN, Mar. 27, 1978, Soluble polymer “completing” agents are p. 24) being developed to remove toxic metals from waste waters or remove radioactive metals from nuclear waste fluids. Polyethylene glycol and its The principal organic flocculants are polyac- rylamides, polyamides, and polyepichlorohydrins. *Excerpt from OTA Working Paper entitled “Priority Setting of Toxic Substances for Guiding Monitoring Programs. ” A com- Stiffening waste treatment regulations to reduce plete copy of the paper can be obtained through the National sludge are making flocculants more attractive. Technical Information Service. (See app. 1.) The chemicals make up a $100 million per year

119 120 c Environmental Contaminants in Food market that may double within the next 5 years. phosphate out of the heavy-duty home laundry These polymers may also be used for enhanced oil market. Zeolite 4A is a cubic crystalline sodium recovery if the price of oil would rise to make aluminosilicate with the formula enhanced recovery economical. (C&EN, Jan. 23, 1978, p. 9) Textile Chemicals Multifunctional Acryiates The market for textile chemicals is expected to The use of radiation curing (ultraviolet or elec- top the billion-dollar market by 1980. Examples tron beam) is rapidly increasing. Radiation curing are: contributes significantly to critical energy sav- Chemical Use ings and pollution abatement. Demand for the fol- Biphenyl ...... Dye bath lowing multifunctional acrylates used for ultra- o-phenylphenol. . , . , ...... additives violet inks and coatings is expected to increase: 1,2,4-trichlorobenzene...... pentaerythritol triacrylate, 1,6-hexanedioldiacry- Methyl naphthalene...... Perchloroethylene ., , ...... late, trimethylolpropane triacrylate, and tetraeth- Alkyl aryl sulfonic acids ...... ylene glycol. (C&En, Jan. 23, 1978, p. 12, Celanese Ethoxylated alcohols , , , ., ., ., ...... Chemical Co. Advertisement) Quaternary ammonium compounds ...... Ethoxylated nonyl phenol, ...... Polyester Resin Alcohols of alkyl aryl sulfonic acid ...... Concern about energy will augment the growth Esters of phosphoric acid...... of lightweight polyester resin in the United States Acrylic latex , ...... Finishing agents through the 1980’s. The need for lighter weight Styrene-butadiene rubber latex ...... materials in the automotive market, as well as the Polyvinyl alcohol , ... , ...... Polyvinyl acetate , ., ., ...... demand for corrosion-resistant products in many ...... areas, will increase the demand for polyester Starches...... resin at an average of 9.2 percent a year through Polyvinylchloride . . . ., ... , ., ...... 1987. Polybutylene terephthalate (PBT) and poly- Fluorochemicals ...... Glyoxal (dimethylol, dihydroxyethylene urea) ethylene terephthalate (PET) will be among the Carbamate (isobutyl, 2-methoxyethyl)...... fastest growing polyester resins. PBT and PET are Tetrakis phosphonium sulfate ‘ammonia . . expected to grow from a combined demand of 38 Cyclic phosphonates ...... , . million lbs in 1977 to 322 million lbs by 1987—an Maleic anhydride...... Printing average annual growth rate of 24 percent. PBT is Polyacrylic acids ...... chemicals currently the most widely used thermoplastic pol- Acrylates ...... Butadiene acrylonitrile ...... yester resin. However, it is expected that PET pro- Hexamethylol melamine formaldehyde. . . duction will increase more rapidly, overtaking Dioctyl sulfosuccinates ...... PBT production. The total polyester resin market Ethoxylates ...... is expected to rise 140 percent in the next 10 Pentachlorophenon ...... Bacteriocides years. other uses and average annual projected o-phenylphenol. . . , ...... production increases from 1977 to 1987 include Low (L and aromatic distillates...... unsaturated reinforced polyester (9. 1 percent), Urea ...... thermoset surface coatings (5 percent), cultured Ethoxylated alkyl phenols ...... [dispersing Sodium lauryl sulfate...... agents/mulsifiers marble and other unsaturated thermoses (4.0 ((;&EN, May 29, 1978, p. 12) percent), strapping (mostly scrap, 28 percent), and fibers (4.0 percent). (C&EN Jan. 30, 1978, p. Pumice 11) Pumice has many similarities with asbestos. They are both mineral oxides, low in density, heat Zeolites resistant, nonflammable, chemically inert, and Zeolites (aluminosilicates) hold promise as de- low in cost, Rhodes (Division of Beatrice Foods tergent builders, A 25-million-lb market in 1977 Co., Des Plains, Ill.) claims pumice may be the safe has a prospect of a possible 400-million-lb market and economical alternative to asbestos (C&EN, in 1982. Zeolites containing detergents perform May 22, 1978, p. 9). The manufacturer suggests “equivalent or roughly equivalent” to the phos- that pumice be used in paints; chemicals (filtra- phate-silicate formulations that now command 75 tion media and chemical carrier); leather buffing: percent of the current heavy-duty powder market. compounders (powdered hand soaps and glass Environmental problems will continue to push cleaners); metal and plastic finishing; and ap- Append/x A—Substances Whose Production or Environmental Release Are Likely to Increase in the Next 10 Years ● 121 placations in the dental, rubber, glass, furniture, Chlorobromination electronics, and pottery industries. Bromine chloride has been shown to be more ef- Polyvinylacetates fective as a disinfectant than chlorine in field tri- als. It is a heavy red liquid and 12 times more sol- Acrylic and acetate resins compete in the mar- uble in water than chlorine. Its vapor pressure is ket as textile and binder emulsions for nonwoven one-third as great. Bromine chloride completely fabrics. Currently, acrylics dominate both of hydrolyzes almost immediately to hypobromous these markets by two-thirds of the total, Polyvinyl- acid and reacts with ammonia in sewage to form acetates make up only 14 percent of the market, bromoamines. It is believed that chlorobromina- with the remainder going to other resins. New tion is the best alternative to current chlorination technical developments may change the status practices. quo. By 1987, acetates could surpass acrylics in many quality paint and textile markets. (C&EN, Polyethyleneterephthalate Mar, 20, 1978, p. 11) DuPont is continuing a vigorous effort to devel- op a market for molded plastics made from con- Hydrazine and Its Derivatives ventional forms of thermoplastic polyethylene ter- Originally used in rocket fuels, today hydrazine ephthalate. (See also Polyester resin, ) This resin and its derivatives are commercially used in her- has a multibillion-pound demand as polyester bicides, pesticides, blowing agents for plastics, fibers and films and is finding a large market in and water treatments. Its consumption is growing plastic beverage bottles. DuPont’s trade name for by 15 percent per year according to Olin Chemi- PET is Rynite. Rynite’s prime targets will be cals’ advertisement in C&EN. metals replacement, especially in automobiles. Olefinic Thermoplastic Elastomers Chemicals Produced or Released Due Olefinic thermoplastic elastomers are also to Energy Development Technology called TPO rubbers and were introduced in 1973, having a market volume of 1.5 million lbs. It is an- Coal Liquefication and ticipated that the demand for TPO rubber will Gasification Program reach 44 million lbs by 1980 and that uses in auto- The conversion of coal to liquid to gaseous hy- mobiles will account for more than half of it. drocarbons for fuel technologies will result in the Mechanical goods and wire and cable uses ac- release of many chemicals to the environment. count for 12.7 and 18,2 percent of TPO demand, The contaminants may enter the environment via respectively. Olefinic thermoplastic elastomers two pathways: 1) emission into the atmosphere are also used by carmakers in electrostatically with the consequent potential for long-range paintable body filler panels, air deflectors, and transport and 2) direct discharge via runoff and stone shields. Recently they have been used as leaching into the aquatic and terrestrial domain sound deadening material on diesel-powered vehi- where impact might be expected to be more local- cles, jacketing and insulating material for wire ized. Contaminants from coal combustion can be and cable coating, and many custom-molded and classified into three groups: 1) organic chemicals, extruded mechanical goods. (C&EN, Oct. 23, 1978, 2) inorganic chemicals, and 3) radionuclides. p. 9) Organic Contaminants Silicones Organic contaminants from coal-derived proc- New markets for silicones as PCB replacements into several categories as in electrical transformers, in brake fluids, and as elastomers are developing. Methylchloride is used as an intermediate in the production of sili- cones, Other methylating agents could be used to replace methyl chloride in most applications, but the potential substitutes (e.g., dimethyl sulfate, methyl bromide, methyl iodide) are much more ex- pensive and have toxicity and handling problems which make them less desirable. 122 ● Environmental Contaminants in Food

Hetrocyclics ...... Pyridines Fine particulate , ...... Sulfur particulate Quinolines Respirable coal dusts Simple aromatic Tar hydrocarbons ., ., . . . . Benzene soots Toluene Gasses ...... Carbon monoxide Xylenes Sulfur dioxide Phenols...... Phenols Sulfur trioxide Cresols Sulfur tetraoxide Xylanols Nitrous oxide Polycyclic aromatic hydrocarbons ...... Benzo(a)pyrene Radionuciides Dibenzofluorene Dibenzoanthracene Emissions from a 100-MW electricity genera- Benzoanthracenes tion powerplant that burns coal at a rate of ap- Benzo(a)anthrone proximately 100 tons per hour are estimated to Dimethylbenzoanthracene contain 1 ppm of uranium and 2 ppm of thorium, A Chrysene plant of this size is expected to release 1 percent Methylchrysenes Benzocarbozoles of its fly ash to the atmosphere. Under these con- Idenopyrenes ditions thorium-228 and -232, radium-224, and Carbozoles lead-212 each contribute approximately 5 x 10-’ Pyrenes Ci per year; uranium-234 and -238, thorium-230 Biphenyl Acenapthalene and -234, radium-226, lead-210, polonium-210, Acenapthalyne and bismuth-2 10 each contribute approximately 8 Fluorene x 10-:~ Ci per year; uranium-235 and praseo- Alkylanthracenes dymium-231 both contribute approximately 3.5 x Alkylphenanthracenes 10-4 Ci per year; and radon-220 and -222 together Anthracene Perylene account for approximately 1.2 Ci per year. (ERDA Benzoperylene report 77-64, August 1976) Coronene Radioactive emissions of bituminous coals have Sulfur compounds ...... Thiophenes a high uranium content (75 ppm). Mercaptans Carbon disulfide Methyl thiphene Solar Heating and Cooling Organometallics. , ., ..., Nickel carbonyl Tetraethyl lead of Buildings Naphthyl ,. ..., Napthyl Ammonium thiocyanate Water contamination can occur in solar-heated domestic hot water systems at heat exchanger in- Trapped organic compounds and aromatic terfaces. Serious health consequences could be units in coals that were isolated, separated, and expected if the contaminated water is ingested. identified by gas chromatography and mass spec- Water contamination could result from the heat trometry are shown in table A-1. exchange fluids themselves, or in water-based Inorganic ChemicaIs systems from such additives as: Category Examples Corrosion inhibitors —Chromates, berates, ni- Acids ...... ,,, Sulfuric acid trates, nitrites, sulfates, sulfites, arsenates, ben- Nitric acid zoate salts, various triazoles, silicates, and phos- Hydrochloric acid phate compounds. Chromium salts ...... Chromium chloride Chromium sulfide Freeze protestants—Glycols. Sulfur compounds ...... Hydrogen sulfide Heat transfer fluids—Paraffins, aromatic and Carbon disulfide other synthetic hydrocarbons. Trace elements. . . ,, . . ! . Manganese Arsenic Mercury Bacteriacides— Chlorinated phenols. Barium Molybdenum Solar collectors used in heating and cooling Beryllium Nickel Bismuth Selenium systems utilize organic chemical compounds as in- Cadmium Silver sulators that can emit highly toxic substances Chromium Tellurium under overheat or fire conditions. Fumes usually Cobalt Thallium consist of simple starches and phenolic com- Copper Tin pounds: ammonia, hydrochloric acid, hydrofluoric Fluorine Uranium Gallium Vanadium acid, toluene diisocyanate (TOI), and hydrogen Lead Zinc cyanide. Appendix A—Substances Whose Production or Environmental Release Are Likely to Increase in the Next 10 Years ● 123

Table A-1 .—Trapped Organic Compounds and Aromatic Units in Coal

45 83 46 84 47 85 48 86 49 50 87 51 88 52.

53 89 54 90 91 55 56 92 57 58 93 59 60 94 61. 95 62 96 63 97 64 65 98 66 99. 67 68 69 100 101 70 71 102 72 103 73 104 74 105 75 106

76 107

77 108

78

79 80 81

82

B = branched T = identlflcatlon tentative ~ = identlflcatlon uncertain

Fuels From Biomass Water can be affected by the residuals produced from thermochemical conversion. Low-molecular Thermochemical biomass conversion can pro- weight oils, phenols, leachates from char and ash duce gases, tars, oils, and unconverted residue residues, and scrubber solution runoff may enter (char) and ash, depending on the particular con- water bodies by direct discharge or by percola- version process. Thermochemical reactions gen- tion to subsurface waters from evaporation erated sulfur-containing (H2S, COS, CS2, SOX) and points. nitrogen-containing (HCN, NOX, NH3) gases. 124 “ Environmental Contaminants in Food

REFERENCE LIST FOR SUBSTANCES WHOSE PRODUCTION OR ENVIRONMENTAL RELEASE ARE LIKELY TO INCREASE IN THE NEXT 10 YEARS

Chemical Marketing and Economics Abstracts. the Health and Safety Laboratory, Energy Re- The Division of Chemical Marketing and Econom- search and Development Administration, New ics of the American Chemical Society (ACS) pre- York, N.Y. 10014, reviewing the literature on one sents papers at ACS national meetings on sub- class of pollutants. Information on the environ- jects related to the responses of the chemical in- mental levels, ecological effects, and potential dustry to economic changes as well as responses toxicity to man of 35 elements that may be re- of the financial community to changes in the leased into the environment by coal combustion or chemical industry. Abstracts of these papers are gasification is presented. published by ACS. Fossil Energy Update. This monthly journal A Study of Industrial Data on Candidates for compiled by the Department of Energy lists ab- Testing. This document, published by the U.S. stracts of current scientific and technical reports, Environmental Protection Agency, Office of Toxic journal articles, conference proceedings, theses, Substances (EPA Contract No. 68-01-4109, No- and monographs on all aspects of fossil energy, in- vember 1976) contains market forecasts for 10 cluding factors involving the environment, health, major classes of chemicals (109 individual chemi- and safety. cals) with an annual production greater than or Chemical Engineering News. The Chemical En- equal to 1 million lbs. Chemicals that are used ex- gineering News is a weekly publication of the clusively as drugs or pesticides are not included. American Chemical Society that contains rele- The market forecasts include: 1) a discussion of vant information in such sections as “Chemical production and trade statistics, 2) consumption World, “ “This Week,” “ Business Concentrates, ” patterns, whenever possible, 3) growth trends, 4) and “Science/Technology,” and in profiles on a brief summary of current uses as well as poten- selected chemicals. tial new applications, and 5) growth trends in end- Chemical Marketing Reporter. This weekly, market consumption. A discussion of possible sub- published by Schnell Publishing Company, Inc., stitutes for some of the chemicals is also included. contains information on chemical market reports Environmental Development Plans. The Envi- and profiles on selected chemicals. ronmental Development Plans (EDPs) published by Trapped Organic Compounds and Aromatic the U.S. Department of Energy (March 1978) were Units in Coal. This article, published by Tyoichi conceived and prepared as basic documents to Hayatsu et al. in Fuel 57:541 (1978), contains a assist in planning and managing environmental detailed analysis of the organic constituents of programs of energy technology development. Ap- three coals: a lignite, a bituminous, and an an- proximately 30 EDPs covering major developing thracite. Organic compounds trapped in the coal energy technologies were prepared. matrix, residuals, and products of the original A Review of Current Information on Some Eco- coalification process were described. logical and Health-Related Aspects to the Release Personal Communications With Several Orga- of Trace Metals Into the Environment Associated nizations, Including the Chemical Development With the Combustion of Coal. This document by Association and Chemical Marketing Research Merrill Heit is a technical report (HASL-320) from Association. Appendix B

Friday June 29, 1979

Part XII

Department of Health, Education, and Welfare Food and Drug Administration

Polychlorinated Biphenyls (PCB’s); Reduction of Tolerances

125 126 ● Environmental Contaminants in Food

30330 Federal Register / Vol. 44, No. 127 / Friday, June 29, 1979 / Rules and Regulations

DEPARTMENT OF HEALTH, proceeding (see 42 FR 17489), one result can be done to remove the PCB’s from EDUCATION, AND WELFARE of PCB contamination of the the water or from the fish; their presence environment has been contamination of is, in that sense, unavoidable. Because Food and Drug Administration certain foods. In the Federal Register of the initial contamination of fish with July 6,1973 (38 FR 18098), FDA issued PCB’s cannot be avoided (nor the PCB’s 21 CFR Part 109 regulations to deal with the problem of processed out), the only way to avoid PCB contamination of food. Among PCB’s in fish is to remove fish from [Docket No. 77N-0080] those regulations was one establishing commerce if it contains PCB’s above a temporary tolerances for unavoidable given tolerance level. The degree of Polychlorinated Biphenyls (PCB’s); PCB residues in various categories of avoidance accomplished by this method Reduction of Tolerances food. Those original tolerances are now is, of course, a function of the level at codified in $109.30 (21 CFR 109,30). The which the tolerance is set, In this way, it order FDA is issuing in this document is theoretically possible to avoid PCB’s AGENCY: Food and Drug Administration. reduces certain of those tolerances. in fish absolutely by removing from ACTION: Final Rule. FDA’s authority to issue tolerances for commerce all fish that contain any amount of PCB’s. SUMMARY : The Food and Drug unavoidable food contaminants is Administration [FDA) is reducing the derived from sections 402(a)(2)(A) and Section 408 of the act authorizes FDA tolerances for unavoidable residues of 408 of the Federal Food, Drug, and to make a practical judgment in dealing the industrial chemicals polychlorinated Cosmetic Act (the act) (21 U.S.C. with such environmental contaminants: biphenyls (PCB’s) in several classes of 342(a)(2)(A) and 346). Section Based on an assessment of the degree to food. Specifically, the agency is reducing 402(a)(2)(A) deems food adulterated, which the contaminant poses a threat to the tolerances in milk and dairy and thus prohibited from interstate consumers, the agency can decide to products from 2.5 parts per million (ppm) commerce, if it contains “any added tolerate the contaminant’s presence in to 1.5 ppm (fat basis), in poultry from 5 poisonous or added deleterious food up to a level the agency considers ppm to 3 ppm (fat basis), in eggs from 0.5 substance” that is unsafe within the appropriate to protect the public health, ppm to 0.3 ppm, and in fish and shellfish meaning of section 4080 Section 408 taking into account, among other factors, from 5 ppm to 2 ppm. deems any added poisonous or the extent to which the presence of the deleterious substance to be unsafe contaminant is unavoidable, In making DATES: Effective August 28, 1979; objections on or before July 30, 1979. unless its presence in the food is this judgment, the agency’s paramount required in the production thereof or concern is protection of the public ADDRESS: Written objections to the cannot be avoided by good health: The tolerance cannot be set Hearing Clerk (HFA-305), Food and manufacturing practice. Section 408 also above the level. the agency finds Drug Administration, Rm. 4-85, 5000 authorizes the agency to promulgate necessary to protect the public health Fishers Lane, Rockville, MD 20857. regulations limiting the quantity of such adequately. But in determining what FOR FURTHER INFORMATION CONTACT: a required or unavoidable substance tolerance level provides an adequate Howard N. Pippin, Bureau of Foods that can be present legally in food. Such degree of public health protection, FDA (HFF-312), Food and Drug limits, called tolerances, are to be set by is required by section A08 to consider Administration, Department of Health, FDA at the level found necessary to the extent of unavoidability—in the case Education, and Welfare, 200 C St. SW., protect the public health, taking into of PCB contamination of fish, the Washington, DC 20204,202-24$3092. account the extent to which the . amount of PCB-contaminated fish that SUPPLEMENTARY INFORMATION: In the substance is required or unavoidable must be disposed of to reduce human Federal Register of April 1,1977 [42 FR and the other ways the consumer may exposure to PCB’s to a tolerable level, 17487), FDA proposed to reduce the be affected by the same or other As a practical matter, of course, a temporary tolerances for unavoidable poisonous or deleterious substances, tolerance, if it is to be enforceable, residues of PCB’s in several classes of Once a regulation establishing a cannot be set below the level at which food. The agency received over 100 tolerance has been promulgated for a the contaminant can be reliably comments on the proposal from particular poisonous or deleterious measured for enforcement purposes by interested individuals, consumer groups, substance, food containing that available analytical methods. businesses, trade associations, State substance in an amount exceeding the The toxicological data available on government agencies, and others. The tolerance is deemed adulterated under PCB's make it clear that, in an ideal agency has considered these comments section 402(a) (2)( A]. situation, it would be preferable not to and is now issuing a final order reducing One of the primary purposes of have PCB’s in food at any level. As the PCB tolerances as originally section 408 of the act is to enable FDA discussed more fully below, the data do proposed. Following a brief discussion to deal effectively with environmental not permit the identification of any level of the background, this document will contaminants such as PCB’s, These of PCB exposure that can be said to respond to the comments the agency substances often enter food as a result provide an absolute assurance of safety, received and explain the agency’s of events beyond the reasonable control It is equally clear, however, that the reasons for adopting the reduced of the food manufacturer or processor reduction of PCB exposure from food tolerance levels. and, once in the food, usually cannot be sources to zero, or to a level I. Background removed by good manufacturing approaching zero, would require practice. For example, in the case of elimination of large amounts of food, PCB’s are a class of toxic industrial PCB’s, some species of fish have become especially fish. Hence, in deciding the chemicals that have become persistent contaminated to varying degrees as a appropriate levels for PCB tolerances and ubiquitous environmental result of the dumping of PCB-containing under section 408, FDA has had to make contaminants as a result of past industrial waste into the nation’s waters some extraordinarily difficult judgments. widespread, uncontrolled industrial use, (see the Federal Register of March 18, It has had to decide, in effect, where the As explained in the preamble to FDA’s 1972 (37 FR 5705)]. Once the proper balance lies between providing proposal initiating this rulemaking contamination occurs, there is little that an adequate degree of public health Append/x B— FDA Derivation of PCB Tolerances— Federal Register ● 127

Federal Register / Vol. 44, No. 127 / Friday, June 29, 1979 / Rules and Regulations 38331

protection and avoiding excessive losses advantage in delaying this action levels of PCB contamination are now of food to American consumers. because it will take years to resolve routinely found. The comments received on the certain of the shortcomings in the data Based on the declining incidence of proposal reveal that by far the most on PCB’s, if they can be resolved at all. PCB contamination, which means that controversial aspect of this rulemaking For example, no chronic toxicity studies PCB’s are avoidable in food to a greater proceeding is the balancing judgment have been performed on the specific, degree now than they were earlier, as FDA made in proposing to reduce the chemically distinct composition of PCB’s well as the new toxicity data suggesting PCB tolerance in fish from 5 to 2 ppm. found in fish residues. Even if such chronic toxic effects, FDA decided the Some comments argued that the studies were begun immediately, it PCB tolerances should be reduced. proposed reduction would cause an would be 3 to 4 years before results In the preamble to the proposal, the excessive loss of food and significant could be available. That plainly is too agency analyzed the new toxicity and adverse economic impact without long to wait to take action necessary for exposure data as they bore specifically providing any significant increase in the protection of the public health. on the tolerance for PCB’s in fish(42 FR public health protection. Other Because of the emphasis the 17492-3). The agency concluded that comments argued the converse, i.e., that comments placed on the proposed reduction of the tolerance from 5 to 2 the proposed reduction to 2 ppm would reduction of the fish tolerance, this ppm was necessary to protect the public not adequately protect the public health document reviews the basis on which health adequately, even though that and that the tolerance should be the reduction was proposed and reduction would result in the estimated reduced to 1 ppm (the lowest level at explains why, after considering the loss of a minor percentage of marine fish which PCB residues in fish can be comments, the agency has decided to (approximately 0.2 percent) and up to 25 reliably measured fop enforcement promulgate the reduction as proposed. percent of freshwater fish shipped purposes), despite the additional losses After discussing the fish tolerance and interstate (the loss of marine and of food that a reduction to 1 ppm would the major points raised about it in the freshwater fish having a combined cause. In each case, the comments comments, this document responds to landed value of approximately $8 bolstered their arguments by contending the remaining comments received on million per year). The agency concluded that FDA has either overestimated or other aspects of the proposal. that the increment of public health underestimated the toxicity of PCB’s and protection afforded at least theoretically the impact of the proposed reduction of 11. The Tolerance for Fish and Shellfish by a further reduction of the tolerance to the tolerance in terms of food loss and In the preamble to the April 1, 1977 1 ppm did not justify such a reduction in adverse economic consequences. proposal, the agency discussed new light of the substantially greater loss of The comments criticizing the proposed toxicity data that had become available food that would result (a combined reduction of the fish tolerance highlight after the original PCB tolerances were landed value, marine and freshwater. of the difficulty of the judgment FDA must promulgated in 1973 [42 FR 17488-9). In approximately $18 million per year). sometimes make in establishing contrast to the data underlying the As noted, a large majority of the tolerances. Not only must FDA make a original tolerances, which consisted comments on the proposal dealt with qualitative judgment about the proper primarily of data from retrospective some aspect of the agency’s proposal to balance between adequate public health studies of humans in Japan who were reduce the fish tolerance to 2 ppm. Some protection and excessive loss of food, it exposed to high doses of PCB’s and of the comments agreed that the also must often make that judgment on showed acute toxic effects from the proposal struck a proper balance the basis of data that are incomplete, or exposure (42’FR 17487+), the new between the need to protect the public even in dispute, and that can easily lead toxicity data consist primarily of animal health and the need to avoid excessive reasonable people to differing studies showing an association between loss of food. Other comments argued conclusions. As the comments illustrate, PCB exposure and serious subchronic that the tolerance should be reduced to 1 it is nearly always possible to conduct and chronic toxicities, including adverse ppm in light of the new toxicity data on additional studies and investigations to reproductive effects, tumor production, PCB’s, despite any additional loss of refine further the knowledge of a and, possibly, carcinogenicity, as well food that might result. substance’s toxicological profile, the as effects on numerous biochemical Most of the comments on the fish incidence and-degree of human systems [42 FR 17488-9). Although the tolerance, however, were submitted by exposure to it, and the impact a given data do not fully resolve such important members of the fishing industry, by tolerance reduction will have on the questions as the carcinogenicity of trade associations, and by agencies of food supply. As an agency whose first PCB’s, they lead to the conclusion that State governments involved in responsibility is to protect the public neither “no effect” nor “allowable daily commercial fishing matters, who argued health, however, FDA must act on the intake” levels for PCB’s can be that reduction of the tolerance to 2 ppm basis of the information available to it, established with any confidence and is not justified. Some of these comments even when the information is that, from a toxicological point of view, contended that the health hazard incomplete. Neither the agency nor the human exposure to PCB’s should be presented by occasional consumption of public can afford to wait until every reduced. fish containing 5 ppm PCB’s is not uncertainty is resolved. See Ethyl Corp. The preamble to the proposal also significant and that any reduction in risk v. Environmental Protection Agency, 541 discussed data FDA had gathered on to consumers accomplished by reducing F. 2d 1, 24-29 (D.C, Cir,) (en hmc), cert. human exposure to PCB’s, especially the tolerance to 2 ppm would be minor. denied, 426 U.S. 941 (1976]. from dietary sources (42 FR 17489-90), These comments also argued that any In the case of PCB’s, even though These data show that the current such risk reduction would be there are obvious shortcomings in the incidence of PCB contamination of food outweighed by the resulting adverse available data, which are discussed has declined significantly in comparison economic consequences, which some below, FDA considers the data to to that on which the original PCB argued would be far in excess of those provide a more than adequate basis for tolerances were based (see 37 FR 5705), cited by the agency in its proposal. In the exercise of its judgment in reducing Indeed, the new data show that fish are support of the latter argument, some of the PCB tolerances. There would be no the only food group in which detectable these comments estimated the impact a 128 ● Environmental Contaminants in Food

38332 Federal Register / Vol. 44, No. 127 / Friday, June 29, 1979 / Rules and Regulations

2 ppm tolerance would have not only on reduce the fish tolerance. Nothing in the even that cannot be guaranteed. The the commercial fish catch, but also on comments and nothing discovered Work Group on Risk Assessment of the employment and income in the fishing during FDA’s reassessment of the Interagency Regulatory Liaison Group and related industries and on toxicity data alters the validity of that (IRLG) has recently prepared a report recreational fishing. Arguing that the fundamental judgment. The agency that discussers many of the principles States would curtail recreational fishing therefore concludes that it is important involved in risk assessment. in certain areas if the tolerance were as a matter of public health protection to As part of its review of the reduced to 2 ppm, the comments minimize human exposure to PCB's. toxicological justification for reducing projected large losses of sales among The real question raised by the the fish tolerance, the agency has those supplying boats, licenses, tackle, comments is whether the degree of risk performed a risk assessment aimed at and bait to sport fishers. reduction accomplished by lowering the comparing the estimated risks Due to the large volume of comments fish tolerance to 2 ppm is sufficient to associated with PCB exposure at the challenging the proposed reduction of justify the increased loss of food that the various levels of exposure that would the fish tolerance, the agency has lower tolerance will cause. This is an result from different tolerance levels carefully reassessed the justification for extremely difficult question because it is The written report on this risk lowering the tolerance from 5 to 2 ppm. not now possible for toxicologists to assessment has been made a part of the It has reviewed the toxicological data quantify precisely, on the basis of record of this proceeding as Reference and has attempted to estimate in toxicity data derived from animal 45. quantitative terms the degree to which studies, the risks posed to humans. As explained in that report, the risk lowering the tolerance would reduce Using classical toxicological methods, assessment involved the use of the most risk to consumers. In addition, it has re- the most that can be done reliably is to recent available data on the incidence of examined the question of how much make qualitative judgments about risks: PCB contamination of fish to calculate additional loss of fish would occur as a A statistically significant increased the level of exposure to PCB’s that could result of the proposed reduction. Based incidence of adverse effects in animals be expected to result from tolerance on its reassessment, the agency is good evidence of a risk to humans, levels of 5,2, and 1 ppm (Table 4, Ref. concludes that reduction of the and, generally, the greater the incidence 45). These calculations were based on tolerance for PCB’s in fish to 2 ppm of effects in animals, the greater the risk the assumption that under a given strikes the proper balance between the to humans (Ref. 43). Having identified tolerance level, no fish containing PCB’s need to protect the public health and the the risk of a chronic toxic effect from in an amount above that level would be need to avoid unnecessary loss of food. exposure to a substance, classical consumed, It is true, of course, that an Hence, the reduced tolerance is being toxicological principles lead to the FDA tolerance level directly affects only promulgated as proposed. conclusion that reduction in exposure fish shipped in interstate commerce, but will reduce the risk (Ref. 44). Again, States often adopt FDA’s tolerance A. Risk Reduction there is no evidence that these levels for application to intrastate and As noted earlier in this preamble, the principles do not apply to PCB’s. recreational fishing. Thus, even if the proposal to reduce the fish tolerance Scientists have recently developed exposure calculations used in the risk was based in part on new toxicity data methods, incorporating mathematical assessment (Table 4, Ref. 45) do showing a relation between PCB extrapolation models, for making somewhat overstate the absolute exposure and an increased incidence of quantitative estimates of risks to amounts of exposure reduction, they various subchronic and chronic toxic humans based on toxicity data from nevertheless demonstrate that a effects, including adverse reproductive animal studies. These risk assessment reduction of the PCB tolerance for fish effects, tumor production, and, possibly, methods do not purport to quantify would result in a significant reduction of carcinogenicity (42 FR 17487-9). The precisely the expected human risk, but PCB exposure [e.g., for heavy consumers proposal itself noted certain factors that rather attempt to estimate in of the affected species, reduction of the complicate the evaluation of PCB quantitative terms an upper limit on the fish tolerance from 5 to 2 ppm reduces toxicity (e.g., varying degrees of toxicity risk to humans that can be expected exposure from an estimated 20.1 among the several forms of PCB’s, the from a given level of exposure to a toxic micrograms (µg) per day to an estimated presence of toxic impurities such as substance, assuming humans are no 14.9 µg per day). Such significant chlorinated dibenzofurans in more susceptible to the effects of the reductions in PCB exposure from fish commercial preparations of PCB’s, the substance than are the most susceptible are especially important in terms of risk differences in chemical composition members of the animal species for reduction because fish are the only food between commercial PCB’s and PCB which toxicity data are available. These group in which detectable levels of residues in fish, and varying risk assessments can be useful as a PCB’s are still regularly found. susceptibilities of different animal means of comparing risks at various Based on the calculations of exposure species to the toxic effects of PCB’s); exposure levels and illustrating the at various tolerance levels and toxicity these complicating factors were also toxicological judgment that a reduction data from animal studies, the agency pointed out in some of the comments in exposure will reduce risk. Because of used a linear extrapolation method to received on the proposal, all the problems inherent in estimate the upper limits on certain risks Notwithstanding these factors, extrapolating from animal data to the posed by exposure to PCB’s, This however, there is little genuine dispute expected human experience, however, analysis resulted in estimates of over the fact that exposure to PCB’s the numbers produced by a risk significant potential risk to humans who must be considered to pose a risk of assessment must be interpreted consume PCB-contaminated fish on a serious, chronic toxic effects in humans. cautiously: They are estimates of upper continuing basis, especially fish The toxicological judgment that flows limits on risk and, though potentially contaminated at or above the 5 ppm from this fact—i.e., that a reduction in useful for comparative purposes, cannot level (Tables 6 and 7, Ref. 45). For human exposure to PCB’s will reduce be said to quantify actual human risk example, using the total malignancy this risk-was an important part of the precisely. These assessments attempt to data from the National Cancer Institute

agency’s rationale for proposing to avoid underestimating human risk, but (NCI] Bioassay (Ref. 19), it is estimated Appendix 6— FDA Derivation of PCB Tolerances—Federal Register . 129

Federal Register / Vol. 44, No. 127 / Friday, June 29, 1979 / Rules and Regulations 38333 that the upper limit on the lifetime risk from a 2 ppm tolerance for PCB’s in fish these more recent data, the value of the of cancer for heavy consumers of fish would be approximately $8 million in fish projected to be lost at tolerance most affected by the tolerances is 9.8 landed value, compared to levels of 5 ppm and 2 ppm is incidence of cancer per 100,000 of the approximately $1 million for the 5 ppm substantially less than was projected in population, assuming the tolerance is 5 tolerance and $18 million for a 1 ppm the proposal. The loss projected under a ppm; 7.2 per 100,000, assuming the tolerance, The estimated $8 million loss 1 ppm tolerance would remain about the tolerance is 2 ppm; and 4.4 per 100,000, resulting from a 2 ppm tolerance same (Table B, “Regulatory Analysis for assuming the tolerance is 1 ppm (Table encompassed a negligible percentage of Final Regulation for Reduction of 6, Ref. 45). Stated another way, it is the marine-fish catch (about 0.2 percent) Temporary Tolerances for estimated that the upper limit on the and about 25 percent of the freshwater Polychlorinated Biphenyls in Food,” Ref. number of new cancers per year among catch (42 FR 17492), 46). Specifically, the amount of heavy consumers of fish most affected The agency arrived at these figures by commercial fish now projected to be lost by the tolerances is 46,8, assuming a assuming that all fish containing PCB’s as a result of a 2 ppm tolerance is about tolerance of 5 ppm; 34.3, assuming a above the tolerance would be removed $5.7 million (expressed in 1974 dollars) tolerance of 2 ppm; and 21, assuming a from both interstate and intrastate compared to the previously estimated $8 tolerance of 1 ppm (Table 7, Ref. 45). commerce (“Economic Impact million; the current estimated loss of fish As explained in the report (Ref. 45), Assessment for Proposed Reduction of under a 5 ppm tolerance is about $0.6 the utility of this risk assessment for Temporary Tolerances for million (compared to the previously evaluating actual risk to humans from Polychlorinated Biphenyls in Food,” Ref. estimated $1.1 million). Under a 1 ppm exposure to PCB’s is extremely limited. 39). There are several difficulties tolerance, however, the projected fish This is due both to difficulties inherent inherent in this assumption. On the one loss, using the new sampling data on in making such extrapolations from hand, it may tend to overstate the loss PCB levels, is about $16 million animals to humans and, perhaps more because (a) some states may not apply (compared to the previously estimated importantly in this instance, to gaps and FDA’s reduced tolerance to intrastate $18 million). The percentage of the uncertainties in the data available for fish, (b) some violative fish will be part freshwater fish catch now estimated to this particular risk assessment. For of nonviolative lots, and (c) some be lost under a 2 ppm tolerance is 14 example, the toxicity studies on which violative lots may enter commerce percent (compared to the 25 percent that the risk assessment is based used undetected. On the other hand, it may had been estimated from the 1974 data); commercial preparations of PCB’s, tend to understate the loss because once under a 1 ppm tolerance, the currently which are chemically different from the the violative percentage of a given estimated loss of freshwater fish is 35 PCB residues found in fish and which species reaches a certain level, percent (compared to the previously contain small amounts of highly toxic commercial fishers may stop fishing that estimated 43 percent) (Ref. 46), impurities (e.g., dibenzofurans) not species altogether. Some of the As noted earlier in this preamble, known to be present in fish residues, comments cited these difficulties in many of the comments argued that the Also, in making the exposure estimates support of arguments that FDA had impact of the proposed tolerance required for the risk assessment, it was either overestimated or underestimated reduction must be measured not only by necessary to use existing data on the the amount of fish that would be lost as the amount of the resulting fish loss but numerical distribution of PCB levels in a result of a 2 ppm tolerance. Despite its also by other economic impacts, such as fish and rely on the assumption that the acknowledged limitations, adoption of potential unemployment and loss of effect of a given tolerance level is to the assumption is a necessary and income in the fishing industry and remove from commerce all fish reasonable method for dealing with the postulated disruption of the recreational containing PCB’s exceeding the uncertainties inherent in predicting the fishing industry (e.g., reductions in boat, tolerances. It is possible that neither the impact of a tolerance reduction. None of tackle, and bait sales]. The comments assumption nor the data precisely reflect the comments suggested an alternative provided figures ranging into the what actually occurs. method for estimating the amount of fish hundreds of millions of dollars on the For these reasons and others that would be removed from commerce total economic value of these industries discussed in the report (Ref. 45), the risk as a result of the proposed tolerance and, without offering any further assessment does not provide a basis for reduction. analysis, contended that the impact on precise quantification of the amount of Because of the comments it received them would be “severe” or “major.” The risk reduction accomplished by reducing questioning the justification for the predicted impact on recreational fishing the fish tolerance, Despite the proposed reduction in the fish tolerance, was premised on the possibility that limitations inherent in the risk the agency has re-examined its State governments would severely assessment, however, the agency projections of the food loss expected to curtail recreational fishing if the regards it as illustrative of the basic result from such a reduction. The tolerance were reduced to 2 ppm. validity of the toxicological rationale for projections made in the preamble to the In establishing a tolerance for PCB’s reducing the tolerance for PCB’s in fish: proposal were based on data obtained in fish, FDA must take into account the Reduction of the tolerance will result in in 1974 on the levels of PCB’s in amount of fish a given tolerance would a significant reduction in risk among commercial fish, primarily from the remove from commerce. Section 406 of those who consume PCB-contaminated Great Lakes (Ref. 39), In making those the act, however, neither requires nor fish. FDA considers this risk reduction original projections, the agency was authorizes FDA to weigh secondary to be of significant public health value, forced to rely on the assumption that economic impacts when it considers the even though it cannot be precisely PCB levels in freshwater fish nationwide level at which a tolerance should be set. quantified. were as high as those found in the Great Consideration of such impacts would be Lakes. FDA now has more recent and B. Loss Of Food inconsistent with the paramount more representative data on PCB levels concern of section 406, which is In the preamble to the proposal, the in commercial fish, which it obtained protection of the public health, and agency estimated that the loss of food through a nationwide sampling program would complicate the decisionmaking from commercial channels resulting conducted in 1978 and 1979, Based on process under section 406 in a way 130 “ Environmental Contaminants in Food

Federal Register / Vol. 44, No. 127 / Friday, June 29, 1979 / Rules and Regulations.

Congress did not intend. Obviously, levels well below 2 ppm. Over the long can fairly be termed “avoidable’’ -even consideration of the amount of food loss term, the adjustments will help to among the-more highly contaminated caused by a tolerance helps to ensure minimize the net economic impact of the commercial species most likely to be that the direct economic consequences tolerance reduction on both individual affected by a reduced tolerance, only a of the tolerance (in this case, decreased fishers and the overall commercial minor percentage (about 1.5 percent) sales and employment in the commercial freshwater fishing industry. contain PCB’s at levels as high as 5 ppm fishing industry) will not be None of the comments attempted to (Table A, Ref. 46). The agency’s disproportionate to the increased degree quantify in dollar terms the impact of judgment is that the balance between of public health protection accomplished the tolerance reduction on the public health protection and loss of food by the tolerance; but the agency recreational fishing industry, but several is properly struck by a 2 ppm tolerance. considers secondary economic postulated a “severe” or “major” impact As noted, as 2 ppm tolerance effects a consequences, such as potential impact premised on voluntary decisions by meaningful decrease in risk to on the recreational fishing industry, individuals not to fish and mandatory consumers while still excluding from totally beyond the scope of section 408. curtailments of recreational fishing by commerce only a relatively small None of this should suggest that the State authorities. FDA is in no better amount of food (about $5.7 million agency is unaware of, or unconcerned position than were those submitting the landed value in 1974 dollars). about, the economic consequences of its comments to make precise predictions Several comments argued that an actions. It is keenly aware that actions it about the future behavior of individuals adequate degree of public health takes to protect the public health can and State agencies. However, the protection can be provided only by have adverse economic consequences, agency considers the premises lowering the fish tolerance to 1 ppm, the both direct and indirect, and that these underlying the projections of “major” or lowest level at which PCB’s can be consequences can sometimes be felt “severe” impact to be somewhat reliably measured in fish for with particular severity in certain speculative and of questionable validity. enforcement purposes. Indeed, as one narrow segments of the economy. For AS noted to the extent that the behavior would expect, the risk assessment example, some of the comments on the of individual recreational fishers is performed by the agency, and discussed proposal argue that the impact of a 2 affected by the tolerance reduction at above, indicates that the estimated risks ppm PCB tolerance for fish will be all, they, as much as commercial fishers, that might be experienced by consumers especially severe for small-scale, can be expected to adjust to the of contaminated fish would be reduced freshwater fishers who specialize in tolerance by shifting their activities to even further by a reduction of the certain species that happen to be the less contaminated species and tolerance to 1 ppm (Tables 6 and 7, Ref. heavily contaminated. The agency locations. Also, even if State agencies 45). Based on the evidence now before acknowledged this possibility in the decide that some curtailment of it, however, the agency does not preamble to the proposal (42 FR 17492). recreational fishing is necessary in light consider a reduction to 1 ppm necessary in the present case, however, the of the reduced tolerance, it is reasonable or appropriate in light of the policy of agency has reason to believe that the to expect that their actions will be section 408 of the act. claims of adverse economic impact are tailored by species and location. In the The risk assessment the agency made exaggerated. Based on the 1078/1979 past, the most common response of incorporated several conservative data on PCB levels in freshwater fish, a State agencies to FDA’s PCB tolerance assumptions that were designed to 2 ppm tolerance will remove from for fish has not been the mandatory avoid understatement of the human risk. commerce about $5.7 million worth of curtailment of recreational fishing. Thus, it is expected that the actual risk commercial fish. Although it is possible Instead, they have issued warnings experienced by consumers of the 12 that fishing for certain heavily concerning particular species and more heavily contaminated species contaminated freshwater species may locations and made suggestions covered by the risk assessment is less cease entirely in locations where PCB regarding both limitations on than that estimated. Moreover, the contamination is concentrated, at least consumption of particular species and average consumer, who eats fish from a some affected fishers-both commercial methods of preparing and cooking fish variety of freshwater and marine and sport-can be expected to adjust to that minimize the amount of PCB’s sources, will actually experience a far the reduced tolerance by increasing actually consumed from contaminated lower level of PCB exposure and a their catch of other species or fish. Thus, there is little reason to correspondingly lower degree of risk transferring their activities to other, less believe that a 2 ppm tolerance will lead than those whose fish consumption is contaminated locations within their to widespread, mandatory curtailment concentrated among the more heavily current area of operation. of recreational fishing and the resulting contaminated (predominantly In evaluating claims of economic drastic economic impact the comments freshwater) species. For these reasons, impact, it is theoretically and postulate. notwithstanding the quantified risk pragmatically sound to take into account estimates produced by the risk the motives and opportunities for C. Conclusion assessment, the agency reaffirms the adaptive behavior by affected Based on the data now before it, the conclusion it expressed in the preamble individuals and fins. If the public agency concludes that a reduction of the to the proposal: The 2 ppm tolerance demand for commercially caught fish fish tolerance from 5 to 2 ppm will result provides an adequate degree of remains stable or increases, and if the in a meaningful decrease in the risk protection for all but those who attractions of sport fishing remain experienced by consumers from consume above-average amounts of strong, it can be expected that some exposure to PCB’s. Some reduction of freshwater fish taken from fishing activity will shift to species that the tolerance is clearly in order because contaminated waters (42 FR 17493). are not contaminated above the the toxic effects associated with In the agency’s judgment, the tolerance. Over time, the shifts will exposure to PCB’s are serious and additional increment of public health become easier as the levels of PCB irreversible; and, due to declining levels protection that might be provided by contamination decline because more of PCB contamination the current 5 ppm reducing the tolerance to 1 ppm does not and more species will have average PCB tolerance permits contamination that justify the additional loss of food that Append/x B— FDA Derivation of PCB Tolerances— Federal Register ● 131

Federal Register / Vol. 44, No. 127 / Friday, June 29, 1979 / Rules and Regulations 38335 would resuIt. First, as discussed above,PCB’s are forthcoming, the agency might age, especially pregnant and lactating the agency estimates that under a consider establishing a 1 ppm tolerance women, who may have consumed, or are tolerance of 1 ppm, approximately $16 despite the effect that would have on the consuming, higher than normal amounts million worth of the commercial fish availability of food. of PCB-contaminated fish. Data that catch would be vioIative and thus, For these reasons, the agency were discussed in the preamble to the presumably, removed from commerce.concludes that at this time a 1 ppm proposal (42 FR 17468-9) suggest an This is nearly triple the $5.7 milIion tolerance would not strike the proper association between PCB exposure and worth estimated to be violative under abalance between protection of the reproductive disjunction in rats and 2 ppm tolerance. It is far more likely public health and the need to avoid monkeys. They also show acute toxic under a 1 ppm tolerance than under a excessive2 loss of food. effects in the nursing offspring of ppm tolerance that the more heavily Though FDA considers 2 ppm to be maternal monkeys that had been contaminated species of freshwater fish the appropriate tolerance level for PCB’s exposed to toxic levels of PCB’s. Data would be violative in percentages high in fish under the criteria imposed by gathered by FDA since it issued the enough to put an end to their section 406 of the act, the agency is proposal in 1977, and discussed in the commercial exploitation and, possibly, concerned about the health of certain report on FDA’s risk assessment on force some segments of the freshwater groups that may not be adequately PCB’s (Ref. 45), establish more clearly fishing industry to cease operations protected by a 2 ppm, or even a 1 ppm, the link between PCB exposure and completely. Thus, the actual loss of food tolerance. As noted, sport fishers and adverse reproductive effects in the resulting from the 1 ppm tolerance could others who consume abnormally large rhesus monkey. They also confirm the greatly exceed even the $16 million amounts of the more highly earlier data showing acute toxic effects landed value (1974 dollars) estimated contaminated species may be at risk in the nursing offspring of PCB-exposed above. from PCB’s regardless of any tolerance maternal monkeys. As explained in the Second, for the average consumer, FDA establishes. (The agency’s risk risk assessment report (Ref. 45), it is not current exposure to PCB’s in fish is at a assessment, using data from a study of possible at this time to determine with tolerably low level, when considered in Lake Michigan sport fish eaters, confidence the significance of these data light of the criteria of section 406 of the estimated that the upper limit on the in terms of human risk. There have been act, without a 1 ppm tolerance, The lifetime risk of cancer for heavy eaters no reports of human reproductive average consumer eats a modest amount of sport fish from Lake Michigan is abnormalities or overt toxic effects in of fish from a variety of sources, both about 12 to 14 times greater than the nursing human infants that can be freshwater and marine, most of which corresponding risk for heavy eaters of attributed to PCB’s. That fact is of only yield fish with PCB levels below 1 ppm. those commercial fish most affected by limited significance, however, because Because their exposure is thus low to a PCB tolerance, even assuming the epidemiological studies adequate to begin with, they are adequately tolerance remained at 5 ppm (Table 6, detect such adverse effects in humans protected by a 2 ppm tolerance, which Ref. 45).] Those individuals, whose high have not been conducted. ensures that they will not be exposed to exposures to PCB’s tend to result from An additional reason for concern in the unusually high levels of PCB’s found localized conditions and fishing this area is that PCB’s ingested by in some species of fish. The slight practices beyond the control of FDA, human mothers are found, and to some additional protection these average should take steps to reduce their extent are concentrated, in human consumers might gain from a 1 ppm exposure to PCB’s. FDA urges State and breast milk [see the discussions in the tolerance does not justify the local health officials to evaluate the preamble to the proposal and in the risk significantly greater impact such a situation in their own localities and assessment report (Ref. 45)). Ina recent tolerance would have on the availability determine what steps, if any, they can nationwide survey, consisting of 1,038 of food. On the other hand, atypical take to address these special situations. samples of human breast milk collected heavy consumers (e.g., the Great Lakes In the past, some State and local in 44 States, the mean concentration of sport fisher who catches and consumes agencies have made FDA’s tolerance PCB’s was estimated to be in the range large quantities of the contaminated level for PCB’ s applicable to fish in of 1.00 to 1.10 ppm (on a fat basis) (Ref. species) would likely not be adequately intrastate commerce and have issued 45). Though the data are scanty, it is protected by even a 1 ppm tolerance advisories to sport fishers warning that reasonable to assume that among because of the amount of fish they eat consumption of certain species of fish women who consume above-average and because those fish are seldom should be minimized and suggesting amounts of PCB-contaminated fish, or affected by FDA tolerances (either other ways in which PCB exposure who are exposed to PCB's from other because they are sport fish or are from could be reduced. These agencies should sources, the levels of PCB’s in breast intrastate commercial channels and, in review their past actions in light of the milk are significantly higher, As noted it either case, are outside FDA’s current state of knowledge about PCB’s is not now possible to determine the jurisdiction). Protection of these and make the changes or take the significance of these facts in terms of consumers depends on actions by State additional steps that may now be increased risk to the nursing infant. authorities. appropriate, FDA will cooperate with In sum, although the agency concludes Finally, though the new toxicity data these agencies, as it has in the past, by that a 2 ppm tolerance for PCB’s on PCB's clearly support the need to providing technical advice and adequately protects most consumers, reduce exposure to this contaminant, the assistance. FDA is sending letters to the women of childbearing age, especially uncertainties in the data (discussed governors of States most affected by pregnant and lactating women, are above) cast some doubt on the degree to PCB’s in fish, discussing the agency's among those who should be careful to which consumers are at risk from concerns about aspects of the PCB avoid abnormally high exposure to extremely low levels of PCB exposure, problem that may require an up-to-date PCB’s in fish, They can avoid such and therefore weigh against lowering review in their States. exposure by minimizing consumption of the tolerance to 1 ppm. If, for example, The agency is advising that State both commercial and noncommercial more definitive and incriminating data health departments be particularly fish from waters known to be on the reproductive risks posed by concerned about women of childbearing contaminated with PCB's and avoiding 132 “ Environmental Contaminants in Food

38336 Federal Register / Vol. 44, No. 127 / Friday, June 29, 1979 / Rules and Regulations

entirely those species of sport fish argued, the tolerance should be applied exposure to PCB’s even further. known to contain high levels of PCB’s selectively to both the species and the Reduction of the tolerance for PCB’s in (e.g., coho and chinook salmon from the body of water. fish will bean especially effective step Great Lakes, and freshwater trout, The individualized approach to toward accomplishing that goal, because striped bass, and catfish from some establishing and enforcing tolerances for fish are the one remaining significant locations). State and local governments environmental contaminants suggested source of dietary exposure to PCB’s. have the important role of advising by this comment is not feasible because 4. One comment contended that fish consumers about conditions in the necessary species-by-species, products are being subjected to an particular localities. location-by-location data on PCB entirely different regulatory standard The agency is aware that its decision occurrence do not exist. Furthermore, than are poultry products, with no to set the fish tolerance for PCB’s at 2 many lots of fish, as currently packaged reasonable basis for the different ppm, rather than leaving it at 5 ppm or and shipped, do not bear the water-of- treatment. The comment stated that the reducing it further to 1 ppm, is inherently origin information required for the emphasis in establishing 2 ppm as the judgmental in character. Section 406 of recommended regulatory approach. tolerance for fish appears to have been the act provides no formula for These limitations make it necessary for safety to the consumer despite a balancing public health protection the agency to establish tolerances for considerable economic impact. Yet, it against loss of food, and, hence, there is fish on a generic basis. The result is a argued, the higher level of 3 ppm for no way for the agency’s decisions under uniform regulatory approach for all poultry is based on economic section 400 to be arrived at species, which provides clear and fair considerations relating to feed mechanically or quantitatively or to rules for all segments of the fishing contamination, apparently without appear clear-cut in every case. In this industry and is necessary to ensure that public health considerations. The case, for example, forceful arguments uncertainties and limitations in data will comment went on to state that the have been made in the comments in not result in increasing human exposure average per capita consumption of fish support of both a 5 ppm and a 1 ppm to PCB’s. To the extent that certain is 19 grams (g) per day compared to 63 g tolerance, but those arguments all reflect species only occasionally have PCB per day for poultry products. According the subjective” judgments of those who levels above 2 ppm, the economic to the comment, this means that under made them. In the end, the agency has impact of the reduced tolerance will be the proposed tolerances, and assuming been mandated by the Congress to make slight. maximum permissible levels in all foods, its own informed judgment about what 2. One comment stated that any FDA the average person will receive five is necessary to protect the public health. regulatory action regarding PCB’s in fish times as much PCB’s from poultry as It has done that herein setting the fish should apply to sport fish as well as from fish. tolerance at 2 ppm. commercial fish. The agency does not agree with the The statute provides an opportunity FDA’s regulatory authority extends comment’s contention that the for a public hearing on the agency’s only to foods shipped in interstate considerations involved in establishing order lowering the PCB tolerance for commerce and clearly does not extend the tolerances for poultry and fish result fish. Such a hearing would provide to fish caught and consumed by in different or conflicting regulatory persons adversely affected by the order individual sport fishers. FDA cooperates approaches for these products, First, the an opportunity to present any additional with the State agencies who have 3 ppm tolerance for poultry is based on evidence they may have bearing on the authority over sport fishing by sharing PCB residues in the fat of the bird, not in matters that influenced the agency’s data and views regarding toxicological, all the edible tissue as it is for fish, judgment. As always, the agency is analytical, and compliance matters, but Poultry generally averages about 10 prepared to reevaluate its position in FDA has no direct control over the percent fat; hence, the 3 ppm tolerance light of evidence adduced at a hearing. regulatory approaches adopted by the is comparable to a level of about 0.3 States. As noted, however, the agency ppm for the entire edible portion. Thus, D. Other Comments on the Fish and urges State and local health officials to even taking into account the higher Shellfish Tolerance look closely at the PCB problem in their average level of chicken consumption In addition to the points addressed areas and take whatever steps they find and assuming all foods contain above, the comments raised several necessary to address those aspects of maximum permissible amounts of PCB’s, other points relating to the tolerance for the PCB problem, such as the exposure poultry will actually be regulated at a fish and shellfish: of sport fishers, that are beyond FDA’s level that will result in a substantially 1. One comment recommended that authority. lower intake of PCB’s from poultry than FDA review its entire mechanism for 3. One comment requested from fish. Second, data show that handling recurrent problems of reconsideration of the proposal to detectable PCB residues occur so environmental contaminants in fish. The reduce the fish tolerance on the ground infrequently in poultry that exposure to comment stated that the PCB tolerance that overall ingestion of PCB’s is PCB’s from that source is already at an should remain at 5 ppm for marine fish reportedly declining. Because levels in insignificant level. Hence, further because the levels in those fish are low other foods have already decreased reduction of that tolerance would not enough that a reduction to 2 ppm would considerably, it was argued, there is less significantly reduce dietary exposure to have no increased protective effect, but need to lower the fish tolerance, PCB’s and would not enhance protection would result in economic problems that The agency is aware that PCB levels of the public health, Fish data, on the are unnecessary for species with only in foods other than fish have declined other hand, show frequent occurrence of occasional high PCB levels. The and that overall PCB intake is lower PCB residues at significant levels, so comment stated further that tolerances than it was in 1973, when the original that reduction of the tolerance will should be set for freshwater fish based temporary tolerances were established. result in increased protection for on their individual place in the market— However, as discussed in the preamble consumers of fish. their tonnage, distribution patterns, and to the proposal and in section II of this As explained in the preamble to the consumption patterns. When such preamble, toxicological considerations proposal (42 FR 17491-2), the agency factors combine to present a risk, it was now make it desirable to reduce dietary selected 3 ppm (fat basis) as the Append/x B— FDA Derivation of PCB Tolerances— Federal Register ● 133

Federal Register / Vol. 44, No. 127 / Friday, June 29, 1979 / Rules and Regulations 30337

tolerance for PCB’s in poultry to allow authority to regulate the pace at which that can significantly reduce dietary for the regular use of poultry feed particular species of fish are exploited exposure. contaminated up to, but not exceeding, commercially. 9. Two comments requested FDA to the 0.2 ppm tolerance for PCB’s in 7. Several comments stated that the hold a public hearing before finalizing poultry feed. (0.2 ppm is the lowest decision to reduce the tolerance for fish reduction of the fish tolerance. feasible tolerance for PCB’s in poultry should be reconsidered and the current The agency does not consider a public feed because of limitations on analytical 5 ppm level reaffirmed because PCB’s hearing on the fish tolerance to be cap ability.) The 3 ppm level took into are being steadily eliminated from the necessary or appropriate at this time. account the biomagnification of PCB’s in environment and may be expected to Tolerances are established under poultry that results from regular feeding disappear as a significant problem section 406 of the act under the formal with poultry feed contaminated up to, within the next decade. rulemaking procedures set forth in but not above, 0.2 ppm. The agency The agency does not agree that PCB’s section 701(e) of the act (21 U.S.C. reasoned that it would be inconsistent can be expected to be an insignificant a71(e]]. Under those procedures, any to set tolerances on two products at problem within 10 years. Although person adversely affected by this order levels such that the use of one product EPA’s continuing activities have may file objection within 30 days and that complies with the applicable resulted in a significant decrease in the request an evidentiary hearing on the tolerance causes the second product to amount of PCB’s being introduced into issues raised by those objections. The be illegal and, thus, that it would be the environment especially into water, opportunity for a hearing ensures that inappropriate to do so in the absence of the stability and persistence of these all genuine, material issues relating to other overriding considerations [e.g., chemicals and the likelihood that some the PCB tolerances will be fully aired. safety). For the reasons stated in the amount of additional contamination will Holding a hearing before issuing this preceding paragraph, the 3 ppm continue to occur from waste disposal order would only duplicate the tolerance for poultry adequately opportunity for a public hearing already sites ensures that PCB contamination protects the public health and is thus will remain a problem for the available in formal rulemaking and consistent as a matter of public health foreseeable future. Moreover, that PCB unnecessarily delay the proceedings. protection with FDA’s other tolerances levels are declining (i.e., that PCB’s are 10. One comment stated that the 5 for PCB’s. becoming more avoidable) is a reason to ppm tolerance for PCB’s in fish should 5. One comment stated that any consider lowering the tolerance. not a be retained but requested that FDA decision to lower the fish tolerance provide guidance to State agencies justification for leaving it unchanged. made in reliance on the regulation of regarding use or implementation of the 2 point source discharges and 8. One comment argued that the ppm tolerance if it is adopted. manufacture of PCB’s should not fail to decision to reduce the fish tolerance As noted, FDA provides data and consider the fact that PCB levels in should be reconsidered because by views to the States on a range of matters contaminated waters are not expected lowering the fish tolerance, thereby related to implementation of tolerances to decline for many years. preventing consumption of for PCB’s in food and will continue to do The agency is aware that, despite contaminated fish, some might be led to so. Environmental Protection Agency believe that the problem of exposure to 11. One comment asked whether (EPA) antipollution activities and the PCB’s had been solved. This procedures other than reducing the fish resulting gradual decline in PCB levels misconception could in turn reduce the tolerance have been evaluated as in at least some contaminated waters, pressure to attack the real problem— alternative means of reducing intake of there will continue to be a significant pollution, However, the comment PCB-contaminated fish. occurrence of PCB’s in fish for at least argued, if the environmental The agency has considered the use of the next several years because of the contamination itself is viewed as the general public warnings and/or labeling stability and persistence of the PCB’s “real” PCB problem of importance, as ways to limit consumption of now contaminating the environment. changing the fish tolerance is almost contaminated fish. Such approaches That fact was taken into account in irrelevant, given the small quantity of have been rejected, except as they apply deciding to reduce the tolerances. PCB’s affected. to certain heavy consumers of 6. One comment stated that, because PCB contamination of the contaminated sport fish (discussed pollution of water with PCB’s is environment is itself an important part above]. A public warning about fish expected to continue, PCB levels in fish of the PCB problem because, as generally, or even about particular will continue to rise, and susceptible discussed in the preamble to the species of fish, would not be effective in fish should be harvested now before the proposal (42 FR 17469-90), some human protecting the general public from increased contamination makes them all exposure to PCB’s comes from the air commercial fish because, assuming no inedible. and water, though the amount is changes are made in labeling, Although the levels of PCB’s in waters probably minimal. EPA is addressing consumers have no way to determine currently contaminated may not that part of the problem. However, FDA the species or waters of origin of most decrease substantially in the near disagrees with the view that exposure to commercially prepared fish products. In future, the agency does not expect those PCB’s from dietary sources is addition, general public warnings might levels to increase, nor does it expect the insignificant in comparison to the unduly discourage consumption of fish, levels of PCB’s in fish to increase. Better amount of exposure from the air and most of which is safe to eat and control of PCB levels should result from water. The agency has based the nutritious. Similarly, the requirement of efforts by the EPA and industry to proposed tolerance reductions on its warning labels on fish products in lieu of control discharge of additional PCB’s conclusion that dietary exposures to a tolerance, even on a species-specific into the environment. Hence, even if it PCB’s pose significant risks to basis, is not a sufficiently precise were possible to harvest whole species consumers, which can be reduced by regulatory approach because not all fish of fish now, that step would not have reducing exposure. That there is some from even the most heavily the effect of preventing increased future exposure to PCB’s from other sources is contaminated species contain levels of exposure to PCB’s. Finally, FDA has no not a good reason for withholding action PCB’s above the 2 ppm tolerance level. 134 “ Environmental Contaminants in Food

38338 Federal Register / Vol. 44, No. 127 / Friday, June 29, 1979 / Rules and Regulations

Thus, as with general public warnings, 14. Some comments included requests more sensitive to PCB’s than the dog or warning statements on labels are likely for compensation for commercial fishers rat. to discourage consumption of safe fish. and processors whose livelihoods are This comment is based on an 12. Some comments contended that destroyed by reduction of the fish apparent misunderstanding of the most Americans would probably prefer tolerance. One comment asked, in effect, toxicological rationale underlying the 2 to be warned of the potential danger that the effective date of the tolerance ppm fish tolerance. In evaluating the from PCB residues and retain the option reduction be delayed for 10 years so safety of substances in food, FDA of eating freshwater fish, rather than be fishers would have time to adjust ordinarily attempts to determine the deprived of any choice in the matter by economically. “no-effect” level for the substance, i.e., having the fish removed from the For reasons discussed earlier in this the highest level of exposure at which market. document, the agency considers it no adverse effect is observed in unlikely that the reduction of the fish appropriate animal studies, It then uses The agency acknowledges that some tolerance to 2 ppm will have the dire appropriate safety factors to extrapolate people would probably prefer to be left consequences on which these comments the results of the animal studies to the with the choice of whether to consume are premised. Moreover, FDA has human situation and determine safe fish contaminated with PCB’s above the neither the authority nor the resources levels of human exposure. In this case, 2 ppm level. As noted, however, the to provide compensation for economic however, the reduction of the fish consumer of commercially marketed fish losses that might be suffered as a result tolerance is not based on any “no- generally lacks the information on of regulatory actions it takes. The affect” level, It is based instead on a water-of-origin, size, and sometimes proposed 10-year postponement of the body of data that associate PCB even species that is needed to control effective date would be inconsistent exposure with several serious chronic his or her intake of PCB’s. Under these with the agency’s conclusion that a effects but that do not permit the circumstances, there is no genuine reduction of the tolerance is necessary to establishment of “no-effect” levels for opportunity to exercise informed choice. protect the public health. those effects, Thus, the comment’s Moreover, the agency believes that as a 15. A number of comments were argument that one species is more general matter it is obligated under concerned that if the 2 ppm tolerance for sensitive to PCB's than another and that section 408 of the act to exercise its fish is adopted, FDA will close certain the tolerance should be based on the scientific judgment and determine what waters to fishing or prohibit fishing of “no-effect” level observed in the most level of exposure, and thus what certain affected species in certain sensitive animal species is not relevant tolerance level, will provide an waters. They requested that more to the toxicological rationale the agency adequate degree of public health studies be carried out before relies on for reducing the PCB tolerance protection, determining whether such steps should to 2 ppm. 13. One comment referred to the be taken. 18. One comment disagreed with the agency’s decision not to reduce the The concern underlying these proposal to establish a 2 ppm tolerance temporary tolerances for infant and comments is misdirected, FDA does not for fish instead of a 1 ppm tolerance. junior foods and for animal feeds on the have authority either to close waters to One ppm is the lowest level of PCB ground that the current tolerances are fishing or to prohibit harvesting or residues in fish for which there is “at the lowest level at which PCB’s can possession of fish. Any actions to close analytical methodology suitable for be reliably determined for enforcement waters to fishing would have to be enforcement purposes. The comment purposes” and argued that lowering the instituted by State agencies. stated that toxicological information, tolerance for fish would probably create 16. One comment suggested that if the especially that suggesting the much greater economic hardship than 2 ppm tolerance is adopted, the counties carcinogenicity if PCB’s coupled with would developing and using more affected should be allowed to conduct the presence of PCB residues in human sensitive analytical methods so that more comprehensive testing of residue milk, requires the lowest possible other tolerances could be lowered, levels in the fish before any ban or tolerance. The agency acknowledges that the impoundment of fish in interstate With respect to the carcinogenic fish losses resulting from a 2 ppm fish commerce is imposed. potential of PCB’s, NCI has concluded tolerance would probably be greater In enforcing the fish tolerance, FDA that PCB’s (specifically, Aroclor 1254, than the costs of developing and using will sample and analyze individual lots the commercial PCB most similar the more sensitive enforcement analyses of fish in interstate commerce and take chemically to the PCB residues in fish) that would be necessary for a reduction regulatory action against lots, or the are not carcinogenic in Fischer 344 rats of the other tolerances. The occurrence shippers of lots, that exceed the under the conditions of the bioassay of PCB residues in infant and junior tolerance, There is nothing to prohibit (Ref. 47). After thoroughly reviewing foods and animal feeds is now so any interested party, including local and NCI’s report, the Data Evaluation/Risk infrequent, however, that those foods do State authorities, from conducting Assessment Subgroup of the not contribute significantly to dietary comprehensive testing of fish before Clearinghouse on Environmental exposure to PCB’s. Thus, spending the shipment in interstate commerce and Carcinogens accepted the report’s resources to develop more sensitive from withholding from commerce fish conclusion that PCB’s were not methodology and thereafter reducing the that exceed the tolerance. demonstrated to be carcinogenic in that tolerances for these foods would not 17, One comment suggested that the study, but suggested that PCB’s might significantly increase the protection of proposed 2 ppm tolerance for PCB’s in act as a tumor promoter. For the reasons consumers, and it still would be fish is inadequate for protection of discussed in the preamble to the necessary to reduce the fish tolerance, public health, The comment stated that proposal (42 FR 17489), FDA considers Because PCB’s do occur consistently at the tolerance levels must be based on the question of the carcinogenicity of the significant levels in some fish, the the “no-effect” level observed in-the PCB’s unresolved. For the purposes of its reduction of the fish tolerance can most sensitive animal species for which risk assessment on PCB’s (Ref. 44), provide increased protection for toxicological data are available, and it however, the agency treated the various consumers. suggested that the rhesus monkey is PCB’s as though they were carcinogenic, Appendix B— FDA Derivation of PCB Tolerances—Federal Register . 135

Federal Register / Vol. 44, No. 127 / Friday, June 29, 1979 / Rules and Regulations 38339

and it considers the carcinogenicity of data obtained in the two toxicity tests of 4. One comment opposed reduction of PCB’s to be a matter worthy of further PCB’s in rats and dogs. The comment the temporary tolerance for PCB’s in serious inquiry. suggested that FDA reconsider the eggs on the ground that there are no The agency has long been concerned proposed temporary tolerances on the substantial data that suggest that the with the exposure of nursing infants to basis of a reevaluation of the data from current temporary tolerance is not PCB’s in human breast milk. This, too, is the two long-term studies and, if judged sufficient to protect consumers of eggs. an area in which more must be learned necessary, propose new tolerances or The comment contends that lacking such before definitive statements can be reopen the matter for public comment. evidence, there is not justification for made about the incremental risks posed Though the data from the two long- reducing the tolerance. by this particular avenue of exposure. term toxicity studies of PCB’s in rats and The agency acknowledges that the For reasons discussed earlier in this dogs referred to in the comment were data indicate that eggs do not contribute document, however, the agency considered, the human toxicological measurably to dietary PCB exposure considers a 2 ppm tolerance adequate to data formed the primary basis for and that reduction of the egg tolerance protect all but those who consume developing the original temporary will not significantly affect PCB intakes. above-average amounts of the more tolerances for PCB’s. This fact was However, tolerances established under heavily contaminated species. stated in the July 6, 1973 document section 408 of the act are intended to establishing the tolerances and in the permit only those residues that are 111. Response to Comments on Other unavoidable. Because the available data Aspects of PCB’s preamble to the April 1,1977 proposal to reduce some of the tolerances. indicate that residues above the Following are the agency’s responses The agency is aware that doubt has analytical limits in eggs are avoidable to the comments that did not specifically been cast on the validity of the two and because no evidence was presented address the reduction of the fish long-term toxicity tests of PCB’s in rats to the contrary, it is appropriate to tolerance: and dogs referred to in the comment, as reduce the temporary tolerance for 1. Some comments recommended that well as on the validity of numerous PCB’s in eggs as proposed. the government regulate PCB’s only in unrelated toxicity tests performed by the 5. One comment requested that the the environment rather than in food laboratory facility in question (Ref. 48). temporary tolerances for PCB’s in products. Another comment suggested Therefore, the results from these two animal feed be reduced, but it did not that the limits on PCB’s in foods not be tests are no longer considered worthy of present a rationale to justify reduction. reduced until PCB levels have been reliance and, as explained earlier in this The presence of PCB’s in animal feed reduced in the environment, where the document, these studies played no part is of concern because PCB’s transfer and foods are produced. in the agency’s decision to lower the accumulate in human food products EPA has the authority to control PCB tolerances. derived from animals that consume environmental pollution and has already contaminated feed. The tolerance for taken important steps to prevent further 3. One comment asserted that the finished animal feed is currently set at pollution by PCB’s. Some environmental agency’s statement in the preamble to 0.2 ppm—the lowest level at which contamination with PCB’s already the proposal that it was unaware of any available analytical methodology can exists, however, and will undoubtedly consumers who had suffered deleterious measure PCB’s in animal feed for effects caused by PCB ingestion (42 FR persist for some years. FDA would be enforcement purposes. It would serve no failing in its duty to protect the public 17491) is misleading, in that the useful purpose to reduce the tolerance health if it withheld the actions statement actually reflects a lack of below this level in the absence of necessary to minimize human exposure knowledge rather than awareness of the analytical methodology for enforcing a to PCB’s from dietary sources until the results of properly designed reduced tolerance. Moreover, in light of long-term problem of environmental epidemiological studies. the rare occurrence of PCB’s in animal contamination has been solved. In making this statement the agency feeds, the agency considers the 0.2 ppm 2. One comment asserted that the relied on, the results of an level to provide an adequate degree of primary toxicological basis upon which epidemiological study carried out with public health protection, For these FDA established the temporary sport fishers in Michigan that failed to reasons, the agency declines to reduce tolerances for PCB’s in 1973 (38 FR establish a correlation in humans the tolerance for PCB’s in finished 18096) consisted of two long-terms between the ingestion of PCB’s and the animal feed. feeding studies in rats and dogs that occurrence of deleterious effects (Ref. were performed by the same testing 40). The study is discussed in the IV. “Temporary” Status of the laboratory and that demonstrated a “no- preamble to the proposal (42 FR 17492- Tolerances effect” level for PCB’s at 10 ppm. This 3). The only purpose of the statement As currently codified in S 109.30, the comment also suggested that these same was to cite an instance in which tolerances for PCB’s are designated as two studies serve as the primary basis relatively high exposure to PCB’s in fish “temporary,” The term “temporary” was for the current proposal to reduce those had not resulted in overt, acute toxic used to reflect the fact that the original temporary tolerances: The effects, such as occurred in the Yusho tolerances are subject to revision as comment stated that discrepancies and incident in Japan (42 FR 17488). The new data become available. In the inconsistencies have recently been Michigan example was intended to preamble to the proposal for reducing found in these two feeding studies, as illustrate the observation the agency the tolerances, the agency stated that it well as in other unrelated studies from made in the preamble to the proposal would retain the “temporary” the same testing laboratory, which that the amount of PCB’s in designation because of the possibility would indicate that toxic effects might environmental samples required to that further downward revisions of the actually have been produced in both cause Yusho-type effects is not known. tolerances might be necessary (42 FR rats and dogs, at dietary levels as low as This study has no direct bearing on the 17493). The agency has now 1 ppm. The comment requests that FDA agency’s conclusion that the chronic reconsidered this use of the term extend its audit of the testing laboratory effects of PCB’s require a reduction of “temporary” and has decided to in question to include a review of the the tolerances, abandon it. The term has never had any 136 . Environmental Contaminants in Food

30340 Federal Register / Vol. 44, No. 127 / Friday, June 29, 1979 / Rules and Regulations

legal significance as applied to Toxicology,” Academic Press, 1976, pp. made. Each numbered objection on tolerances established under section 406 124-31. which a hearing is requested shall 45. An Assessment of Risk Associated with specifically so state; failure to request a of the act, and its use is not provided for the Human Consumption of Some Species in FDA’s procedural regulations hearing for any particular objection of Fish Contaminated with Polychlorinated shall constitute a waiver of the right to a governing tolerance setting in Part 109 Biphenyls (PCB’S), 1979, FDA document. (21 CFR Part 109). When circumstances 46. Regulatory Analysis for Final Regulation hearing on that objection. Each are changing so rapidly that a particular for Reduction of Temporary Tolerances for numbered objection for which a hearing tolerance level is likely to be rendered Polychlorinated Biphenyls in Food, 1979, is requested shall include a detailed inappropriate in the near future, the FDA document, description and analysis of the specific agency establishes an action level rather 47. National Cancer Institute Carcinogenesis factual information intended to be than a tolerance (see $ 109.6(c) (21 CFR Technical Report Series No. 38, 1978. presented in support of the objection in 48. Letter from Donald Kennedy, the event that a hearing is held; failure 109.6(c))). In the case of PCB’s, however, Commissioner of Food and Drugs, to the agency has concluded that formal to include such a description and various clients of Industrial Bio-Test analysis for any particular objection tolerances are appropriate. The term Laboratories, Inc., June 1977. “temporary” is being abandoned to shall constitute a waiver of the right to a avoid the suggestion that the legal status Therefore, under the Federal Food, hearing on the objection. Four copies of of the PCB tolerances is something other Drug, and Cosmetic Act (sees. 306, all documents shall be submitted and than that of a formal section 406 402[a), 406, 701[a), 701[e), 52 Stat. 1045- shall be identified with the Hearing 1046 as amended, 1049 as amended, tolerance. Clerk docket number found in brackets 1055, 70 Stat. 919 as amended (21 U,S,C. in the heading of this regulation, Any FDA tolerance, just like any other 336, 342(a], 346, 371(a], 371[e))) and Received objections may be seen in the regulation, is “temporary” in the sense under authority delegated to the above office between the hours of 9 a.m. that it is subject to reevaluation and, if Commissioner of Food and Drugs (21 and 4 p.m., Monday through Friday. necessary, revision as new data become CFR 5,1), Part 109 is amended in $109,30 Effective date. This regulation will available. The agency will continue to by revising the section heading and become effective for foods initially monitor the PCB problem and, if paragraphs (a)(l), (2), (3), (4), and (7] and introduced into interstate commerce appropriate in light of changing (b) to read as follows: after August 28, 1979 except as to any circumstances or new data, will propose provisions that may be stayed by the revisions in the PCB tolerances. $109.30 Toterences for polychlorinated biphenyls (PCS’S). filing of proper objections. Notice of the V. Analytical Methodology filing of objections or lack thereof will (a) ● ● ● be announced in the Federal Register, Section 109.30(b) has been revised to (1) 1.5 parts per million in milk (fat refer to FDA’s updated compilation of basis). [Sees. 306, 402(a), 408, 701(a), 701(e), 52 Stall (2) 1.5 parts per million in 1045-1046 as amended, 1049 as amended, analytical methodology for PCB’s 1055, 70 Stat. 919 (21 U.S.C. 336, 342(a), 346, “Analytical Methodology for manufactured dairy products [fat basis). 371(a], 371(e)]) Polychlorinated Biphenyls, June 1979.” (3) 3 parts per million in poultry (fat basis). Dated: June 26, 1979. There have been improvements in the Donald Kennedy, analytical methodology for measuring (4) 0.3 part per million in eggs. PCB residues since 1973, and most of the ● **** Commissioner of Food and Drugs. revised procedures have now been (7) 2 parts per million in fish and [FS Dec. 79-z4wM Fded %2s-7s s.45 am] published in scientific journals. A copy shellfish (edible portion), The edible BILLINQ COI)E 411 O-O3-M of each procedure or a reference to the portion of fish excludes head, scales, appropriate journal is provided in the viscera; and inedible bones. updated compilation. As stated in ● 3 109.30(b), the compilation is available (b) For determining compliance with from the Hearing Clerk, Food and Drug the tolerances established in this Administration, Room 4-65, 5600 Fishers section, a compilation entitled Lane, Rockville, MD 20857. “Analytical Methodology for VI. References Polychlorinated Biphenyls, June 1979" is available from the Hearing Clerk, Food The preamble to the proposal cited 42 and Drug Administration, Department of references the agency relied on in Health, Education, and Welfare, Room developing the proposal and stated that those reference documents had been 20857, placed on file with the Hearing Clerk, ● **** FDA (42 FR 17493-4). The following Any person who will be adversely additional references, which are cited in affected by the foregoing regulation may the foregoing preamble, have also been at any time on or before July 30, 1979, placed on file with the Hearing Clerk, submit to the Hearing Clerk (HFA-305), FDA, Rm. 4-85, 5800 Fishers Lane, Food and Drug Administration, Rm. 4- Rockville, MD 20857, and may be seen 65, 5600 Fishers Lane, Rockville, MD between 9 a.m. and 4 p.m., Monday 20857, written objections thereto and through Friday. may make a written request for a public 43. DuBois, K. P. and E. M. K. Ceiling, hearing on the stated objections. Each “Textbook of Toxicology,” Oxford objection shall be separately numbered University Press, 1959, pp. 24-28. and each numbered objection shall 44. Ariens, E. J., A, M. Simonis, and J. specify with particularity the provision Offermeier, ‘*Introduction to General of the regulation to which objection is Appendix C

Methods for Toxicologic Testing *

by SRI International

METHODS FOR TOXICOLOGIC TESTING

137 138 ● Environmental Contaminants in Food scheme, experiments are designed and data are Selection of this other species " . . should con- collected based on expected results such as func- sider such factors as comparative metabolism of tional systemic changes, teratogenicity, or car- the chemical and species sensitivity to the toxic cinogenicity. By the use of an appropriate experi- effects of the test substance . . . . “(l) The route of mental design, several endpoints can be assessed administration chosen is that most nearly iden- in the same experimental period such as is done tical to the potential human exposure. Doses are in FDA’s three-generation studies (2). usually chosen to give results in the 20- to 80- For ease of presentation, this appendix has percent lethal or effective range and are usually been subdivided by endpoint into sections on Sys- separated by 0.5 log units (5). Many modifications temic Toxicity, Carcinogenicity, Mutagenicity, of this basic procedure are accepted. Teratology and Effects on Reproduction, Metabo- lism, and Structure-Activity Relationships. This Irritation appendix is not intended as an exhaustive survey The irritation potential of substances is tested of all testing methods used, but is meant to give an by observation of the reflex behavior of the ani- overview of those methods most commonly used mal and by direct observation of the site of con- today by toxicologists. tact with the agent, Attempts to quantitate reflex behavioral responses, i.e., eye rubbing, regurgita- LocaI and Systemic Toxicity tion, or shallow breathing, have met with little success. The simplest protocols for evaluating ir- Some of the fastest and simplest methods for ritation involve the skin and eyes. Semiquantita- determining the toxicities of substances involve tive systems for scoring skin and eye irritation the observation of changes in the structure and have been proposed by several authors (6-12) and function of organs and organ systems. These involve placing the suspected irritant in contact methods generally involve absolute and relative with the skin or eye of New Zealand White rab- weight changes, gross and microscopic structural bits. Protocols for skin irritation involve contact alterations, and primary and secondary tests for with both intact and abraded skin to differentiate organ, system, or whole animal function. With ad- the agent’s ability to penetrate the skin barrier, vances in the chemical, physiological, and behav- and occlusion of the contact site to maximize the ioral sciences, modifications for testing systemic response. The severity of erythema and edema is toxicity have been proposed that make these pro- scored as the endpoint. Eye irritation studies are cedures more sophisticated and relatively compli- carried out without washing and with washing at cated. Several good texts are available which re- various intervals to determine the effectiveness of view systemic toxicity [3,4). removal of the agent to reducing the adverse ef- Range Finding fect, Opacity, area affected, iris reaction to light, The classic determinations of toxicity involve hemorrhage, swelling, and discharge are scored. percent lethal or effective dose, concentration, or Results from these tests can vary greatly depend- time. These tests may employ any route of expo- ing on the method of application of the substance, whether dry or premoistened, etc. sure, the ones chosen usually being based on fac- tors such as chemical and physical properties of Other potential sites of irritation such as the the agent and potential routes of exposure from sensory nerves, respiratory system, urinary sys- the environment. The results obtained from these tem, and gastrointestinal tract are usually evalu- determinations are usually specific for the spe- ated secondarily or through necropsy. Secondary cies, sex, age, and condition of the organism, and effects include shallow breathing, regurgitation, for the route of exposure and environmental con- agitation on urination, and eye rubbing, which ditions before, during, and after exposure. The are broadly categorized as reflex behavior, and endpoints of these tests may be either structural blood in feces, urine, and sputum, or nasal dis- or functional changes, but they are usually lim- charge, which are more indicative of the primary ited to gross effects such as death or narcosis. irritant effect. These results are not quantifiable These tests are primarily used to determine rela- by present methodologies. Primary evaluation tive toxicities of various agents and for range through necropsy is also a qualitative procedure finding for maximum tolerated dosage prelimi- and has not undergone the extent of standardiza- nary to beginning a subacute study. They are not tion and validation that the skin and eye tests usually used to directly evaluate the hazard. have. Other methods for testing irritation such as In general, these tests will employ young adult resistance/compliance tests of the pulmonary sys- rats and another mammalian nonrodent species. tem, direct observation by scope of the esophagus Appendix C—Methods for Toxicologic Testing ● 739

and gastrointestinal tract, and roentgenographic the agent is unpalatable or that it is interfering examination with and without radio-opaque dyes, with the energy balance, the central nervous sys- have not received wide use as testing techniques tem (CNS) regulation of food or water consump- for Government regulatory purposes. tion, or the motivation of the animal. Although substances administered by other routes of expo- Sensitization sure may also interfere with palatability of food, Some substances, although not necessarily pri- through direct or indirect effects on the sensory mary irritants, elicit an irritant-type response nerves, this is less common. Increases in weight after repeated contact with the organism. Tests may be caused by a proliferating tumor mass. The for this sensitization potential involve exposing evaluations of structural changes can be obtained the animal to an agent at doses below those neces- from animals exposed during subacute experi- sary to produce signs of primary irritation, wait- ments. ing an appropriate interval, and then challenging In general, subchronic or subacute experi- the animal with the substance again at a different ments are designed to last approximately 10 per- site (11). If the response on retest is substantially cent of the animals’ lifespan (90 days for rats). Im- higher than the initial response, the agent can be mediately preceding and during the experimental classified as a sensitizer. Various test methods period, observations on animals should include and modifications have been proposed for testing rate of growth, food and water consumption, de- sensitizing potential (11, 13-21). Perhaps the most meanor, and reflex behavior; blood, urine, and common direct test for sensitization is the guinea feces should be collected. During the experimen- pig maximization test. In this test, the agent is tal period, tissue biopsies may be taken for obser- presented in Freund’s complete adjuvant which vation of structural changes. These techniques increases the response. Studies on the mechanism may be unreliable, however, if a structural of sensitization have shown that the agent or a change is localized and not included in the biopsy metabolize of it (antigen) may induce the lym- material, and such manipulation is often not phocytes of the body to form a complex molecule allowed by regulatory testing guidelines. If biop- (antibody) which reacts with the antigen to form sies are done, additional animals are required to an antigen-antibody complex. This reaction may maintain the statistical validity of the experiment. be with circulating free antibody or with lympho- At the end of the experimental period, the animals cyte-bound antibody, The formation of the are sacrificed and the organs are inspected for antigen-antibody complex induces the production gross changes, removed and weighed, and pre- and release of histamine and other compounds served for histologic treatment and microscopic which cause the erythema and edema at the site examination (5). The specific organs and tissues of antigen attack, or may cause anaphylaxis if the removed and examined will depend somewhat on antigen reaches the blood stream (22,23). The the expected action of the agent administered problem with testing methods based on this (usually perceived from preliminary testing) but mechanism, such as immunoelectrophoresis, should include at least the brain, liver, kidneys, radioimmunoassay, ring test, hemagglutination spleen, heart, testes (and epididymis) or ovaries tests, or microphage migration, is that they do not (and uterus), thyroid, and adrenals. measure the actual adverse effect [dermatitis or Changes in organ weights may signal a func- shock) but measure an indicator response. The tional change in this organ or in other organs; for methods are valuable, however, in demonstrating example, an increase in heart weight could be due the presence of antibody capable of producing to a decrease in oxygen diffusion from the lungs, these health effects. Several comparative tests an increase in adrenal weight could signal a have demonstrated that these in vitro techniques blockage of steroid synthesis within it, etc. The are often more sensitive indicators of the hazard weights of organs can be directly compared with than the classic in vivo ones (22,23). those from control animals; however, this often in- troduces an artifact since experimental and con- Structural Effects trol body weights are usually different. It is com- The basic determination of structural effects mon practice therefore to determine the relative on organs and systems begins with the determina- weights of the organs in relation to the total body tion of absolute and relative weight changes. weight of the animal. Recently it has been pro- Decreases in absolute body weight or rates of posed that the relative weights should be taken as weight gain for a test with a substance incorpo- a function of the animal’s brain weight, the postu- rated into the food or water may show either that lation being that the brain’s growth curve devi- 140 ● Environmental Contaminants in Food ates the least of any tissue in the body. While nor- compared with animal control values and are not malization based on this procedure would tend to necessarily comparable between species, Meth- emphasize changes more than other techniques ods for evaluation of pulmonary function (28), car- currently in use, it is not yet widely accepted. diovascular function (29), and brain and neural Tissue dry weight, after desiccation or ashing, activity (30) have been modified for human and has also been used as a tool for determining animal use. These methods include testing ven- mechanisms of growth and metabolic balances tilator flow, resistance, compliance, and gas dif- (24). This method has a major drawback, how- fusion capacity for the pulmonary system; elec- ever, because it removes the organ from further trical activity of the heart, and blood flow and studies such as microscopic examination. pressure for the cardiovascular system; electrical After gross observation and weighing, the activity of the brain (field and single unit) and organs and tissues are preserved for histological muscles; perception threshold, reflexes, and preparation and microscopic examination. The chronaxy for neural function. The significance of most common methods for the preparation and changes in brain activity as a determinant of tox- staining of individual tissues involve fixing with icity is under question at present, however. 10-percent buffered formalin solution, embedding Generally, in these evaluations, each animal in parafin, and staining with hemotoxylin/eosin, serves as its own control. Baseline data for each Many pathologists prefer other fixing and embed- procedure is determined prior to administration ding media, and certain tissues require different of the agent, and any changes in function are procedures. There are also special stains for noted during and after administration, since it is highlighting different cellular components. There important to determine whether the agent causes is no one best method for preparation and obser- reversible or irreversible changes in function. vation of the tissues. The most valuable procedure from the pathologists’ viewpoint is to prepare the Various other methods of testing for organ or tissues in a number of ways, which allows com- system function rely on both primary and second- parison of various aspects such as specific cellu- ary parameters. For example, liver function may lar components, nuclei, cell membranes, etc. be assessed by dye clearance studies (primary) or (25-27). by analysis of serum enzyme concentrations (sec- ondary). Most of the secondary procedures are Special consideration can also be given to tech- now automated and available through various niques in histochemistry and electron micro- clinical laboratories at a reasonable cost, Many scopy. These methods are not used routinely in toxicological evaluation and depend on a knowl- investigators, however, still prefer to perform the edge of the mode of action of the toxic agent. They tests manually, and standard procedures are can, however, indicate changes in cellular metab- well-defined and available in several texts olism or structure before those changes become (31-33). Various modifications of these tests for manifest by the conventional histological proce- specific animal systems have been developed and dures, and therefore they may be more suitable published, Tests usually considered appropriate for observing changes from agents whose toxici- include total and differential blood counts, serum ties are low or develop slowly. The equipment enzyme and ion analysis, urinalysis (especially for necessary for these techniques is generally more metabolizes of the agent), and liver and kidney expensive than that needed for the more conven- function tests (dye clearance). The value of these tional microtechnique methods. They are also tests is that abnormal results will often precede more time consuming and less standardized than obvious structural damage of the organ system in conventional methods. The histochemical meth- question and will be apparent at lower dose ods, although they might be more appropriately levels. classified as tests of organ function, are becoming The study of hematologic effects encompasses more widely accepted with investigators studying changes in the bone marrow as well as those in mechanisms of toxic: action. the cells of the circulating blood. Observations are made of the cells and of their absolute and Functional Effects relative numbers. Specific tests such as dye dilu- Frequently, changes in organ or system func- tion for blood volume, specific gravity, sedimenta- tion are observable before any change in struc- tion rates, osmotic fragility, hematocrits, or clot- ture becomes apparent. The test methods dis- ting time, are not routinely performed but may be cussed below have generally been adapted from indicated. Serial bone marrow biopsies may also human to animal use, and results are ordinarily be performed for hematologic effects; the results Appendix C—Methods for Toxicologic Test/rig ● 141

give both structural and functional information, indicator of hazard and can be used as an “early but these techniques are also not widely used. warning system. ” Several good reviews of behav- The functions of the liver may be tested for ioral toxicology are available (34-37). biliary obstruction (icterus index, alkaline phos- Some of these methods rely on newer methods phatase), liver damage (thymol turbidity, plasma of analysis such as contingent negative variation protein ratios, cholesterol ratio, glucose level, (CNV). Some are really an application of preexist- transaminase level, and cholinesterase level), ex- ing principles such as dorsal evoked potentials cretory function (bromsulphalein clearance, bili- and neuromuscular transmission time that have rubin tolerance), and metabolic function (glucose been widely used by experimenters in the field of tolerance, galactose clearance), The most com- neurophysiology. Most of these techniques have mon tests used in toxicology are the serum alka- only recently been turned to the evaluation of tox- line phosphatase and serum transaminases, and icity. in some cases a dye clearance (bromsulphalein) or glucose tolerance. At the present time, standardization and vali- The kidneys are responsible for excretion of dation of behavioral techniques has not been ac- certain substances, e.g., urea, and for concentra- complished. The question often raised by regula- tion and dilution of urine. Tests for excretion in- tory agencies is how do you relate an observed be- volve dyes like phenolsulphonphthalein and also havioral decrement to an adverse health effect, measure such substances as urea and creatinine, especially if there is no concurrent structural Concentration and dilution tests involve measure- change apparent in the nervous system. Because ment of urine specific gravity after fasting for of these factors, behavioral studies are often der- various periods. These tests are generally not ogated by these agencies when setting exposure used in toxicology screening studies unless there limits for toxicants. Current research is being is reason to believe the toxicant acts on the conducted, however, under Government con- kidneys. tracts to answer some of these questions. The evaluation of these tests may proceed with Comparison of Short- and Long-Term or without modification of the tissue metabolism. Methods for Systemic Toxicity That is, promotors and inhibitors of enzyme sys- tems, e.g., SKF-525A for mixed function oxidase, Most of the procedures noted in this section are may be used to enhance the susceptibility of a equally applicable to short- and long-term testing, particular organ or system to damage from a toxi- obvious exceptions being irritation and sensitiza- cant. This in effect maximizes the response so tion tests. The value of long-term testing for sys- that the toxic action can be more readily ob- temic toxicity lies in the ability to use low doses served. that do not produce detectable adverse effects in Many other specific tests are available for a short time period to see whether bioaccumula- evaluating various organs and systems such as tion and cumulative effects occur. Predictions of sperm motility, specific gravity of cerebrospinal the effect of bioaccumulation can be made know- fluid. calcium-phosphorus ratios for the skeletal ing the effects of short-term high doses, but final system along with tensile strength and compac- evaluation of the toxicity depends on the long- tion, epinephrine sensitivity of heart muscle, term effects observed, As will be pointed out in acetylcholine test of lungs, metabolism of excised the metabolism section, toxicants can be poten- tissue, work and strain measurements of the vari- tiated or inhibited by the metabolism and relative ous muscle systems, etc. accumulation of the toxic moiety. Without defi- nitely knowing the various metabolic reactions, Behavioral Effects rates, and probabilities, it is impossible to ac- A recent addition to the field of toxicology has curately predict toxic effects. High short-term been behavioral testing. Testing methods have doses may induce a toxic reaction, such as death been devised for everything from simple percep- from pulmonary edema, that might mask long- tion to complex tasks involving perception, learn- term, low-level exposure effects such as liver ing, judgment, motivation, and motor activity. The cancer, The differences in short-term and long- value of the behavioral methods lies in the ability term tests for systemic toxicity include the num- of the nervous system to respond to toxic agents ber of interim measurements allowed, the ability a t doses much lower than those necessary to pro- to ascertain the types of effects which might duce “classic” signs of toxicity in the organism. develop only over a long period and the progres- Therefore, these methods are a potential sensitive sion or time course of toxic manifestations, and 142 w Environmental Contaminants in Food the ability to evaluate mechanisms of bioaccumu- ducible. Finally, the test results should be quan- lation or adaptation in the organism. titatively applicable to mutagenesis in humans. Since no single test can fulfill these requirements Mutagenicity and none is reliable enough to stand alone as an indicator of mutagenic potential, mutagenicity Rapid identification of a food contaminant as a ‘testing should include a variety of systems se- possible mutagen is necessary to reduce the po- lected to show whether the test substance or its tential genetic risk to humans who might contact metabolize produce any of a range of genetic ef- the contaminant. Mutagenic effects on humans fects. The test battery approach includes systems often cannot be directly detected, and deleterious that will detect several types of gene mutations, effects on the human gene pool may not become chromosomal aberrations, and DNA repair; thus, apparent for many generations if, for instance, this approach offers the greatest reliability for the deleterious effect is due to a recessive gene. determining mutagenic potential, Tests that eval- Heritable genetic damage in humans may result uate effects in intact mammals are essential for from any of several types of effects on the genetic predicting mutagenicity in humans. material, The two major classes of effects are Since screening large numbers of chemicals by point mutations, which generally affect a single the test battery approach (39,40) may be prohibi- gene or part of a gene, and more extensive chro- tively costly, a hierarchical approach to mutage- mosomal effects such as gross changes in struc- nicity testing, known as tier testing (41), has been ture or changes in number. suggested. Tier 1 consists of relatively inexpen- Only a few tests are available that directly sive short-term prescreening tests. These use evaluate genetic effects of exposure of mammals micro-organisms or other in vitro systems to de- to chemicals: however, the potential of a chemical termine priorities for indepth testing. Substances to produce heritable genetic alterations in man that produce positive results in these tests, as can be evaluated indirectly from its effects on ge- well as those that are negative but are structur- netic material in various biologic test systems, in- ally similar to known mutagens or to which there cluding micro-organisms, mammalian cell cul- is a substantial risk of exposure for humans dur- tures, insects, and intact mammals. ing or preceding their reproductive years, should For substances that cannot feasibly be elimi- continue into Tier 2. nated from the human environment, it is not suffi- Tier 2 tests are usually designed to detect sub- cient to identify the existence of a genetic hazard; stances that are not mutagenic in vitro but are quantitative assessment of the risk involved is metabolized to an active form in the intact mam- necessary for appropriate regulatory activity, mal. Tests used at this level may include the domi- such as establishing action levels or tolerances nant lethal test, in vivo cytogenetic tests, the host- for food contaminants. mediated assay, and body-fluid analysis. Sub- Mutagenicity testing is also used to prescreen stances that are negative in Tiers 1 and 2 are gen- chemicals as an indicator of carcinogenic poten- erally considered safe for use and are given very tial and, less frequently, other toxic effects such low priority for further testing. as teratogenicity. This application is based on em- Only substances for which it is important to as- pirical demonstration or correlation between mu- sess risk are subjected to Tier 3 testing, designed tagenicity and carcinogenicity of chemicals (38) to permit quantitative evaluation of mutagenic po- and does not depend on the assumption that the tential. Tests used at this level include multigen- same mechanism is involved in both types of ef- eration mammalian studies, such as the heritable fect. translocation test, X chromosome loss test, and specific loci test in mice. Approaches to Testing Tier testing may represent an efficient use of As in all toxicological tests, mutagenicity tests resources in large-scale mutagenicity testing, but may produce false negatives (a negative result the use of prescreening tests carries serious dis- when the substance is actually mutagenic) and advantages in determining mutagenic potential, If false positives (a positive result when it is not test chemicals are prescreened by a single micro- mutagenic), and correlation between the results bial test, the proportion of false negatives may be from two test systems may be poor. Ideally, a mu- unacceptably high, and potentially hazardous or tagenicity test system should be sensitive enough useful substances may escape further testing. to detect any chemical that may cause heritable The use of two or three tests at this level, in- genetic damage and its results should be repro- cluding both micro-organisms and mammalian Appendix C—Methods for Toxicologic Testing . 143 . . . , ...... —- >-- cell cultures with and without activation by mam- means of metabolic activation is the addition of malian enzyme systems, may substantially in- microsomal mixed-function oxidase enzymes, typi- crease the reliability of prescreening (96). Never- cally from rodent liver homogenates, to metabo- theless, such cell systems may not approximate lize the test chemical in vitro, This activating sys- metabolic events in the intact mammal closely tem is added to the culture medium as part of the enough to reveal the mutagenic action of some Ames testing procedure with S. typhimurium, and substances that are potential human mutagens. it has provided evidence for the mutagenicity of Whatever the testing approach and test sys- many substances that have no direct mutagenic tems selected, mutagenicity tests should include a effect on these bacteria (51). Microsomal enzyme positive control as well as negative (untreated activation is also used with other microbial test and solvent) controls. The positive control sub- systems (46-48). The major drawbacks of this sys- stance, a known mutagen in animal systems that tem are that it would not detect chemicals metab- is selected for its structural similarity to the test olized to mutagenic intermediates by mechanisms chemical, serves to demonstrate the sensitivity of other than liver microsomal enzymes, e.g., sub- the test organism and the efficacy of the metabol- stances metabolized by the intestinal flora, and it ic activation system used. is possible that the in vitro metabolism of the substance does not adequately mimic its metabo- Current Test Systems lism in the intact organism because of competing Chromosomal effects of many substances have reactions. Another drawback is that a standard- been demonstrated in plants such as Vicia faba ized in vitro activation system has not been de- and Tradescantia (42), and the latter organism vised to date. has also been used in detection of somatic muta- Another widely used bacterial system is the tion (43). While the genetic events involved (alter- multipurpose strain of developed ations in DNA) are the same as those in mammali- by Mohn and coworkers (49). This strain can be an cells, their relevance to human mutagenesis used to measure reverse mutations restoring the has been questioned because of the major phylo- ability of the bacteria to synthesize the nutrients genetic and physiologic differences between arginine and niacin. Forward mutation rates in plants and animals; thus, a negative result in two genes controlling galactose metabolism can plants does not indicate that a substance is not a also be scored in this strain of E. coli. Use of this mutagen in mammalian systems. test organism permits the detection of several Of the many bacterial species that have been types of mutation in a single experiment. used to detect point mutations, the most exten- Eukaryotic micro-organisms that are used to sively employed are the Salmonella typhimurium detect the ability of chemicals to produce point mutants developed by Ames (44,45). The Ames mutations include haploid strains of the yeasts test uses a series of histidine-requiring mutant Saccharomyces cerevisiae and Schizosaccharo- strains that revert to histidine-independence by myces pombe (47,5o) and of the ascomycete Neu- specific mechanisms, either base-pair substitu- rospora crassa (51). A diploid strain of S. cerevi- tions or frameshift mutations, The original strains siae permits detection of chromosomal damage have undergone several further modifications expressed as mitotic recombination that produces that increase their sensitivity to mutagens by in- phenotypic color changes (52). terfering with DNA repair or modifying the cell Whole-animal activation mechanisms can cir- wall to enhance the penetration of chemicals into cumvent this problem but they are generally much the cell. Bacteria treated with the test chemical less sensitive than tests using in vitro activation. are plated on selective media or cultured in liquid In these systems, rodents are exposed to the test suspension to determine the number of revert- chemical by an appropriate route, and the effect ants. A reproducible mutation rate twice the of rodent metabolizes on microbial genetic mark- spontaneous (control) rate is usually considered ers is determined by either body-fluid analysis or evidence of mutagenic activity. host-mediated assay procedures, Because microbial cell systems do not possess In the body-fluid analysis (53-56), the micro- the metabolic capabilities of mammals, they will organisms are treated with the urine, blood, or not detect chemicals that exert a mutagenic effect homogenized tissues of the exposed animals. Cau- through metabolic intermediates, Several activat- tion is necessary in interpreting negative results ing systems have been developed for use with in of these studies, in the absence of supplementary vitro test systems to duplicate the effects of mam- pharmacologic data, since even if the chemical is malian metabolism. The most extensively used metabolized to a mutagen, other factors such as 144 ● Environmental Contaminants in Food tissue-specific activation and detoxification by autoradiographic methods, has also been used mechanisms and the half-life of the compound and to examine a variety of chemicals (65,66). Sister its metabolizes may affect test results. In the host- chromatid exchange in various cell systems has mediated assay (57,58), the micro-organisms are been demonstrated following exposure to known exposed to mammalian metabolic products of the mutagens, but additional work is needed to define test substance by being introduced into the perito- the extent to which results are correlated with neal cavity, circulatory system, or testes of the more traditional mutagenicity tests, host mammal. The host is treated with the test The fruit fly Drosophila melanogaster is used in substance, and after an appropriate incubation a comprehensive and extensively characterized period, the indicator organism is removed and ex- mutagenicity test system which can detect all amined for mutations. types of mutagenic activity at a fraction of the Genetic damage in micro-organisms can also be time and cost of in vivo mammalian testing (67). assessed indirectly through the use of DNA re- The large number of genetic markers and known pair-deficient strains of bacteria (44,59). These chromosomal aberrations make it possible to tests organisms and otherwise identical strains assay a chemical for many types of mutagenic ac- that have normal ability to repair DNA are tivity in a single test. The sex-linked recessive treated with the test substance. Toxic action of lethal test (68) in Drosophila is a very efficient the test substance produces zones in which bac- mutagenicity assay, since about 20 percent of the terial growth is inhibited, and the difference in insect’s genetic material is located in the X size between the inhibition zones in repair-defi- chromosome. Recessive lethal changes caused by cient and normal strains indicates the extent to point or chromosomal mutation can be mapped which this toxicity is due to damage to the DNA. and in most cases the nature of the change caus- A number of mammalian cell systems in culture ing the mutation can be determined. This test also have been developed for detecting point muta- permits a quantitative assessment of mutagenic tions, including cell lines derived from mouse lym- activity. Because of the large size of Drosophila phomas, Chinese hamster ovaries and embryos, chromosomes, this organism can also be used and human fibroblasts and lymphoblasts (60-63). readily to assess meiotic and mitotic recombina- In addition, gross chromosomal changes such as tion, dominant lethality, translocations, and dele- breaks, gaps, and rearrangements can be micro- tions. Some indirect mutagens that required meta- scopically observed in these cells. Stable re- bolic activation have been shown to be mutagenic arrangements, such as translocations and inver- in Drosophila, indicating that these insects have a sions, are considered evidence of heritable microsomal mixed-function oxidase system (69). changes. The induction of only gaps and breaks is However, to determine whether Drosophila test not regarded as evidence of mutagenicity, be- results are useful for risk assessment, more in- cause these aberrations often occur as a result of formation is needed on how their metabolism of general cytotoxicity and thus may be present only foreign chemicals compares to that in humans. A in moribund cells. Like microbial systems, in vitro few other species of insects have also proven use- mammalian cell tests can be used in conjunction ful in mutagenicity testing, including ‘several with activation by mammalian enzymes or with species of the parasitic wasp, Habrobracon (70). whole-animal activation to permit detection of Clearly the most accurate predictions of muta- mutagenic effects by metabolic products of a test genic potential in humans can be drawn from substance. tests that determine direct genotypic and pheno- Mammalian cells, including human white blood typic effects in mammals exposed to the test cells, are also used to detect chemical damage to chemical by routes relevant to human exposures, DNA by measuring unscheduled DNA synthesis Several direct mammalian tests exist, but these (63). This indirect indicator of genetic damage is have the disadvantage of detecting only a few of evaluated by measuring the uptake of radioactive the possible types of genetic damage. Chromo- thymidine for repair of damaged DNA during somal damage occurring in vivo can be detected those stages of cell growth when DNA synthesis in several different cell types, such as bone mar- does not normally occur. A similar test with row cells and circulating lymphocytes, and the mouse spermatocytes exposed in vitro or in vivo presence of micronuclei in red blood cells (7 I). demonstrates effects on DNA in germinal cells Cytogenetic tests using mammalian lymphocytes (64), and the micronucleus test offer the advantage of Sister chromatid exchange, a reciprocal ex- permitting direct comparison with effects on hu- change of segments at homologous loci. measured mans resulting from accidental exposures; how- Appendix C—Methods for Toxicologic Testing ● 145

ever, these tests demonstrate only effects on so- Carcinogenicity matic cells and do not provide direct evidence of heritability. In the event of massive or long-term environ- Cytogenetic changes in mammals can also be mental contamination of food destined for human evaluated in germinal tissue from the testes. In consumption, one of the decisions to be made is the direct spermatocyte test (72), male mice are whether the contaminant appears to pose a sig- exposed to the test substance. After sufficient nificant carcinogenic risk. With this in mind FDA time for the treated spermatogonia to reach the submits the candidate compound to the Chemical spermatocyte stage, they are examined for cyto- Selection Working Group at the National Cancer genetic abnormalities. This test allows for the ac- Institute for consideration under the carcinogen tual observation of induced cytologic changes in bioassay screening program (76,77). premeiotic male germ cells, but it does not permit detection of effects in postmeiotic cells or their Prerequisites for a Carcinogenicity transmission to the offspring. Study Effects on offspring can be evaluated in mice by Once the compound has been selected, it is the heritable translocation test (73) and the X screened using a chronic or lifetime exposure chromosome loss test (74). In the heritable trans- regimen (76,78,79). However, before the long-term location test, F, male offspring of treated mice are study is undertaken, specific toxicologic profiles mated to determine sterility, and indication o must be obtained. Young healthy adult animals of possible translocation heterozygosity. Chromo- each sex and strain to be used in the long-term somal effects are then confirmed by cytogenetic studies should be used in the preliminary studies. analysis of the germinal cells of the male of offspr- The animals should be of uniform age and weight ing. The X chromosome loss test permits the and should be tested using the same formulation detection of chromosome loss resulting from non- and route of exposure to be used in the long-term disjunction in the female, since, unlike somatic studies. The first is an acute study designed to chromosome aneuploids, animals of XO genotype gain additional information on the acute toxicity, are usually viable, Aneuploidy for the X chromo- assuming there is a paucity of data on this aspect some can be detected by genetic markers and con- of toxicity, and to determine the lethality of the firmed by cytologic observations. test compound. The duration of this test should The dominant lethal assay, usually performed not exceed 24 hours and should include at least in the rat or mouse, uses fetal loss as an indicator three dose levels determined by a geometric pro- of induced chromosomal mutations in male ger- gression. one of the dose levels selected should minal cells (75). The death of the zygote is assum- represent the highest dose to be used in subse- ed to result from chromosomal abnormalities in quent studies. Throughout the investigation, all the sperm of male mice exposed to the test chem- relevant clinical signs should be recorded. Ne- ical. This test is relatively easy to perform and its cropsies should be performed on a random selec- results have been positively correlated with muta- tion of animals of each sex and strain, and any ab- genicity in other animal systems. Preimplantation normal histopathologic changes should be noted. loss alone is not used as an indication of muta- genicity since it has been found to occur for rea- After the 24-hour study, a 14-day investigation sons other than chromosomal changes in the should be initiated in an effort to ascertain the sperm. Disadvantages of this test are its relative doses necessary for the subchronic study, the insensitivity and difficulty in clearly distinguish- next prerequisite investigation for the chronic ing weakly positive results. study. This toxicologic study requires five dose Only one test is available at present that can levels, with the highest one, estimated from the detect heritable gene mutations induced in mam- 24-hour acute study, producing no more than 10- malian germ cells. In the specific locus assay in percent lethality. The other dose levels should mice, forward mutations at seven loci, affecting represent geometric decrements of the highest characteristics such as coat and eye color are dose. Animals should be treated with the test mated with mice homozygous for recessive alleles substance for no more than 14 days, held another at these loci (75). Because such a small number of 24 hours, and then sacrificed for necropsy. loci are involved, this test required the scoring of Throughout the study, the animals should be ob- 20,000 to 30,000 offspring at each dose level to served for clinical signs of toxicity. Other toxicity produce reliable results and is therefore very data, such as those derived from organ function costly and time consuming. tests and metabolism studies, are also necessary. 146 ● Environmental Contaminants in Food

The next toxicity study involves the administra- The animals in any one test group should be tion of the test substance for 90 days and is used sacrificed at an adjusted or prearranged date. as a predictor of the maximum tolerated dose However, a group can be terminated earlier if (MTD). This can be defined as the highest dose there has been high cumulative mortality. Mori- given during a chronic study that can be pre- bund animals should be sacrificed immediately dicted to not alter the animals’ normal longevity upon discovery to lessen the likelihood of unob- from effects other than carcinogenicity, In prac- served deaths and subsequent autolysis or canni- tice, MTD is considered to be the highest dose that balism. Control groups should be sacrificed ac- causes no more than a 10-percent decrement in cording to the original or adjusted sacrifice date; weight compared to controls. Five dose levels are the later date is preferred. Because of the strong required in this study, with a minimum of 10 ani- dependency on histopathologic results and to mals of each sex and strain in each dose group. avoid possible criticisms of the study, necropsies The highest dose level used should be the lowest and histopathologic examinations should follow concentration that produced any detectable un- standard procedures required by regulatory toward toxic effects in the 14-day study. The re- guidelines. maining dose levels should be determined as in The major drawback to the carcinogenic bio- the 14-day study, If the selected dose levels do not assay procedure is the time and cost required to produce a discernible no-effect level, the study complete and analyze such a study. If the study is should be repeated with lower doses. done properly, however, the results should be conclusive and for the most part indisputable, al- Carcinogenic Bioassay though there still remains the question of extrap- The chronic study (76,78,79) represents the olation of results to the human population. essence of the carcinogenicity bioassay. It is used to determine the carcinogenicity of a compound in Short-Term Testing as a Prediction of males and females of two mammalian species, Carcinogenicity usually the rat and mouse. The species selected Evaluation of carcinogenicity has generally re- are tested throughout their entire lifespan. Each lied on the results of long-term animal studies, To test group should consist of a statistically repre- use this kind of testing approach for every sub- sentative number of animals. The highest selected stance suspected of being carcinogenic would be dose should represent MTD and the remaining cost-prohibitive and certainly impractical in dose levels should be adjusted accordingly. There terms of the overall time required to test all sus- should be at least one control group, in which the pected chemicals. Therefore, short-term tests are animals receive only the vehicle used for adminis- being developed to identify carcinogenic sub- tration of the test material, If no vehicle is used, stances. There has been much criticism concern- this control group should be untreated but iden- ing the comparison of short-term testing results tical in every way to the experimental groups. In from different laboratories because of the vary- addition to the concurrent control group, a colony ing conditions and refinements in techniques or historical group should be used for the com- practiced among testing facilities. The protocols parison of longevity, spontaneous diseases, and for these short-term tests, especially those involv- spontaneous tumor incidence. The historical con- ing mammalian enzyme activation systems, have trol may also be used for statistical comparisons. not been standardized or validated through inter- In some studies, a positive control group that has laboratory comparative testing procedures; thus, been treated with a compound structurally simi- comparisons of the data from one laboratory to lar to the test compound and known to be carcino- the next have often produced conflicting conclu- genic in the test species may be indicated. How- sions. In this light, a study conducted in one lab- ever, because of the added risk of handling a oratory that compares several short-term testing known carcinogen, a positive control group is sel- systems has added importance in clarifying the dom used. relative usefulness of the compared systems, Throughout the study, animals must be ob- In a recent study by Dr. Ian Purchase (80), 120 served for signs of toxicity. Every animal should organic chemicals (50 known carcinogens and 62 be examined carefully each week, Animals should noncarcinogens, based on published experimen- be weighed and food consumption measured. In tal data) were evaluated for activity in six short- some cases, it is desirable to evaluate tissue term test systems. These systems included: 1) mu- distribution and concentration of the substance tation of Salmonella typhimurium (45), 2) cell or its metabolizes. transformation (81), 3) degranulation of endoplas- Appendix C—Methods for Toxicologic Testing ● 147 .— mic reticulum (82), 4) sebaceous gland suppres- assays, the cells were used with and without sion (83), 5) tetrazolium reduction (84), and 6) le- metabolic activation with S-9 fraction as sion formation after subcutaneous implant (85). described previously. Without activation, Four additional tests used in a preliminary study very few carcinogens transformed hamster were found to be insufficiently accurate or sen- or human cells in the period of study. With sitive to justify a full evaluation. The tests re- activation, all cell lines detected carcinogens jected were transplacental blastomagenesis (86), with an accuracy of 88 percent or better. piperidine alkylation (87], iodine test (88), and the Furthermore, by the use of both the hamster acridine test (89). cells and either of the human cells, the Although there were considerable variations overall accuracy was improved to 94 percent between tests in their ability to predict car- (91 percent for carcinogens and 97 percent cinogenicity, two tests were quite accurate in for noncarcinogens). distinguishing between the known carcinogens ● Degranulation. The procedure used is that of and the noncarcinogens. These were the cell Williams and Rabin (82) and the test is com- transformation test and the bacterial mutation monly called the Rabin test. The test meas- test, which had accuracies of 94 and 93 percent, ures the loss of ribosomes (degranulation] respectively. The use of cell transformation and from isolated rat liver endoplasmic reticu- bacterial mutation together provided an advan- lum following incubation with the test com- tage over the use of either alone, predicting 99.19 pound. The overall predictive value was 71 percent of carcinogens. Not surprisingly, the in- percent for both carcinogens and noncar- clusion of the other four tests in a screening bat- cinogens. tery with these two resulted in an improved abili- ● Sebaceous gland test. The procedures used ty to detect carcinogens (99.97 percent), but were those of Bock and Mund (83) where test greatly decreased the accuracy and discrimina- chemicals were applied directly to the skin of tory value of the battery. It is important to note mice and a depression in the ratio of seba- that all tests generated both false positives and ceous gland to hair follicles indicates a posi- false negatives: the percentages for both can be tive response. The overall predictive value of readily calculated by subtracting the positive the test was 65 percent (67 percent for car- predictability values from 100 percent. cinogens and 64 percent for noncarcino- A description of each of the tests, with compar- gens). ative percent accuracies for predicting carcino- ● Tetrazolium reduction. The procedures used genicity as determined by Purchase et al. (80) is were based on those described by Iversen as follows: and Evensen (84). Test solutions were ap- ● Bacterial mutation. The procedures used plied directly to the skin of mice and the skin were those of Ames, in which four strains of samples were incubated in tetrazolium red S. typhimurium (TA 1535, TA 1538, TA 98, TA solution. An increase in the in situ biologic 100) were tested with each compound in an reduction of the colorless tetrazolium to a assay medium containing a metabolic activa- colored formazan compound indicates a posi- tion system composed of rat liver postmito- tive response. The overall predictive value of chondrial supernatant (S-9 fraction) and co- this test was 57 percent (40 percent for car- factors. The overall accuracy of the test in cinogens and 71 percent for noncarcino- predicting the carcinogenicity of the com- gens). pounds in this study (80) was 91 percent for ● Subcutaneous implant. The procedures used carcinogens and 94 percent for noncarcino- were essentially- those of- Longstaff and gens. These figures agree with the previously Westwood (85] and involved the subcutane- published value of 90 percent by McCann et ous implantation of a filter disc overlaid with al. (90) but are considerably higher than a gelatinous suspension of the test com- those published by Heddle and Bruce (91) pounds into mice. After 3 months, the sur- who found 65 and 81 percent, respectively. rounding tissues were scored for lesions. The ● Cell transformation. The procedures used overall predictive value for this test was 68 were those of Styles (81) and involved three percent (37 percent for carcinogens and 95 types of mammalian cells—human diploid percent for noncarcinogens). lung fibroblasts (WI-38), human liver- While the work of Purchase and associates (80) derived cells (Chang), and baby Syrian does present a very salient assessment of the hamster kidney cells (BHK 21/cl 13). In all more pertinent in vitro assays, it also points out 148 ● Environmental Contaminants in Food the shortcomings of this type of approach. Short- ty is also a consideration. For teratogenic studies, term tests do not use the induction of cancer as an the toxicant is usually administered daily on the endpoint, but each has a parameter, such as in- specific days of gestation representing the period duction of a point mutation, that varies with the of greatest sensitivity. Administration of the test carcinogenicity or noncarcinogenicity of the test substance should begin at or before implantation substance. Accordingly, the authors feel that and should continue throughout the period of ma- these test parameters should not be given a jor organogenesis. greater weight than that of any other arbitrary The selection of an animal species for evalua- response, regardless of how biologically signifi- tion of teratogenicity is an important considera- cant any of these tests might appear to be with tion. Test protocols currently in use recommend respect to the theories of the chemical induction at least two mammalian species, the first being a of cancer. Also, however much generalized data rodent (e. g., mouse. rat, hamster) and a nonrodent might be generated to support the predictive ac- mammalian species (e.g., rabbit). One species curacy of the given test, this accuracy should not should be the same as that used in the test for be assumed to apply uniformly to compounds of reproductive effects. every chemical class, In conducting the teratogenic investigation, at least three dose levels should be used, along with Teratology and Effects on concurrent control groups. These control groups Reproduction should consist of untreated animals, animals treated with the vehicle of administration only, An investigation of a teratogenic agent involves and, possibly, animals treated with an agent that the study of congenital malformations other than is known to cause the effect that is being in- those that are inherited. Teratogens themselves vestigated. The use of historical or colony con- act as triggers for malformation induction. Mal- trols may also be helpful in evaluating the data. formations may include gross, histological, mo- Both test and control animals must be young, lecular, and behavioral anomalies, mature, prima gravida females of uniform age, The sensitivity of an animal to a teratogen is de- size, and parity. The control groups should be termined by: 1 ) the period in which the insult is re- handled and maintained like the test groups. The ceived during the gestation period (this includes highest dose level to be considered should pro- before germ layer formation and during embry- duce signs of embryo or fetotoxicity as suggested ogenesis or organogenesis); 2) the dose and the by fetal growth retardation and more significant- route of administration of the compound; 3) pla- ly by maternal or fetal mortality: however, mater- cental transfer of the suspect teratogen, including nal mortality should not exceed about 10 percent. its lipid volubility, protein binding ability, and The other dose levels can be obtained in a de- creasing logarithmic fashion to a suspected no-ob- metabolism : and 4) uterine and dietary factors. A toxicologic profile of a suspect teratogen re- servable-adverse-effect level. In the classical quires an evaluation of potential hazards to re- teratology study, treatment may be by gavage so production and, particularly, to developmental that the administered dose is accurately known. It processes that respond to environmental insult has been shown, however, that use of this route of through mutation, chromosomal aberrations, mi- administration may induce anomalies in progeny totic interference, altered nucleic acid synthesis, that are a consequence not of the test compound enzyme inhibition, and altered membrane charac- but of the stress of dose administration (94). teristics. This is particularly important in the con- Prior to initiating the study, it is important to ceptus, embryo, and the neonate where the bio- determine whether sires and dams have success- chemical, morphologic, and physiologic proper- fully mated, This includes examinations for the ties change rapidly. Thus risk assessment must presence of plugs or evidence of sperm in vaginal not only address the teratogenicity of a contami- smears. Throughout the study, females should be nant but also its effect on reproduction. observed for behavioral changes, food and water consumption, body weight, vaginal bleeding in- Classical Approach dicating possible abortion, and spontaneous For an adequate teratologic assessment (79,92, deaths. Females showing signs of aborting or de-

93), exposure to the toxicant should parallel as livering prematurely should be sacrificed. closely as possible that expected in the human Fetuses should be obtained 24 hours before an- population. The pharmacologic activity of the ticipated parturition by cesarian section, Dams compound as well as its acute and chronic toxici- should be sacrificed and a complete necropsy per- Appendix C—Methods for Toxicologic Testing ● 149 test material for at least 120 days. At this point they are bred to produce the F, generation. The types of data to be collected include growth and time of delivery for each weanling as well as overt signs of toxicity, the general behavior and condition of the mothers, measure- ments of spermatogenesis in all Ff generation males used to produce the F: generation, litter size, number of stillborn/live births, and any phys- ical or behavioral anomalies. A statistically valid number of animals (males and females) obtained from the F, generation and used to produce the F~ generation should be sacri- ficed and examined at the appropriate time, with special emphasis on the histopathologic state of the reproductive system. In addition, an adequate number of weanlings of each sex from each dose level, including controls, should also be sacrificed and used for histopathologic analyses. Data derived from the above should be eval- uated for the existence of a relationship between exposure and the incidence and severity of ef- fects on reproduction and behavior, tumors, and mortality. The no-observable-adverse-effect level should also be determined. Compared to a teratology investigation, which requires about 3 months to complete the exposure and to analyze the results, a three-generation reproduction study represents a considerably longer expenditure of time, approaching 1.5 years between the initiation of the study and the com- pletion of the analysis of the gathered data. How- ever, an elementary profile on teratogenicity and reproductive performance can be obtained within a period of a year or less as a part of a continuing long-term toxicity study by adding the appropri- ate number of animals at the beginning of the long-term study (2).

Metabolism

Metabolic assessment studies are not directly used for the assignment of risks, tolerances, or ac- tion levels, These studies are used to determine parameters, such as absorption, distribution, storage, and excretion, that may affect the per- formance of materials in biologic systems, thus enabling the researcher to design testing proto- cols that measure the overall effect of exposure to the material rather than just measuring a part of the biologic response. These protocols can then take into account such factors as tissue concen- tration, length of time in contact with specific organs or tissues, ease of reactivity, and the pro- duction of significant (or nonsignificant) changes 150 ● Environmental Contaminants in Food in overall body concentrations that may alter the rally occurring substances in the body. Foreign outcome of a specific testing regimen. compounds that closely resemble normal body Metabolic study systems are usually centered constituents are frequently metabolized by the around the concept that the circulatory system is same specific enzyme systems that metabolize the major means of transportation of the material, their normally occurring analogues. Most foreign regardless of the route of exposure, Although spe- compounds, however, have no endogenous coun- cific materials may be readily metabolized at or terpart and must be metabolized by relatively near the site of entry into the body (e.g., the lungs, nonspecific enzyme systems. These nonspecific skin, and intestines), the majority of materials are enzymes catalyze many different types of reac- transported unchanged to the liver and then dis- tions leading to a diversity of metabolic products. tributed via the circulatory system to other tis- In general, the nonspecific enzyme reactions can sues and organs. At the same time, some of the be categorized into two types, The first includes material may be excreted unchanged in the urine, the conversion of one functional group into feces, and air or converted to various metabolizes another (oxidation of alcohol to aldehyde), the that are excreted or bound in the tissues. The pro- splitting of neutral compounds to fragments hav- portions of the metabolic products that are ex- ing polar groups (hydrolysis of esters and amides), creted or bound depends on the chemical nature or the introduction of polar groups into nonpolar of the original compound, the dose, route of ad- compounds (hydroxylation). The second type in- ministration, species, strain, sex, diet, and envi- cludes the conjugation of the created polar group ronmental factors. with glucuronate, sulfate, glutathione, or methyl The objective of the metabolic study is to math- groups to form a soluble, excretable product. The ematically evaluate the rates and relative im- product formed in the first reaction may be either portance of these processes in limiting the con- more or less toxic than the parent compound or centration of materials in the tissues of the body. may possess a different type of toxicity. Often, it For this purpose, the body is usually visualized as is this product that actually causes the toxic ef- a group of pharmacokinetic compartments, a sim- fects, including cancer, mutations, cellular necro- plistic view that surprisingly approximates fairly sis, hypersensitivity, fetotoxicity, and blood dys- accurately the complex, interdependent proc- crasias. A portion of the chemically reactive esses that actually occur in the body (95). These metabolize formed becomes bound to tissue mac- compartmental models allow the researcher to romolecules, such as cellular proteins, DNA, use measurements of blood concentration as in- RNA, glycogens, or lipids. In this way, it disrupts direct estimates of tissue concentrations and to the normal function of the macromolecule, caus- determine the length of time the material remains ing adverse biologic effects. In contrast, the ma- unaltered (i. e., the biologic half-life). Rates of ab- jority of the second type of reaction products are sorption from one compartment to another can usually either nontoxic or considerably less toxic also be readily measured, thus giving the re- than the parent compound, searcher information on the rate constants of dif- fusion into tissues and on the rates of distribution The metabolism of a foreign material is con- from the blood to various tissues. Clearance or trolled by enzymes, and any factor which affects elimination rates can also be determined for the these enzymes also affects the metabolism of the excretion of the material from the body via the compound and consequently its toxicity. The me- kidneys, gastrointestinal tract, lungs, saliva, and tabolism of a compound may be inhibited or stimu- perspiration. lated by the presence of competing substrates. The rate of metabolism of materials depends on Pretreatment with drugs, steroids, food additives, many factors, among the most important of which pesticides, polycyclic hydrocarbons, polycyclic are the physiochemical characteristics of the amines, and normal constituents of food can re- molecule itself. Polar compounds are usually ex- sult in an increase in the activity of enzymes that creted very rapidly from biologic systems largely metabolize foreign compounds. This increase in unchanged because of the chemical activity of enzyme activity differs according to the inducing these compounds, whereas nonpolar compounds chemical, but is mediated through an increased such as lipids usually remain in the body longer rate of synthesis of enzyme protein, a decreased because they must be metabolized to polar com- rate of turnover of enzyme, or an activation of en- pounds before excretion occurs. zyme, possibly by changes of structure or con- Another important factor in the metabolism of formation. Because of this phenomenon, chronic a compound is its structural resemblance to natu- administration of a foreign material may enhance Appendix C—Methods for Toxicologic Testing ● 151 the activity of the enzymes that catalyze its genie potential per se. The determination of meta- metabolism. bolic activation can permit the experienced re- Several types of biologic phenomena are readi- searcher to predict the kind of adverse effect that ly studied using a metabolic test system. These in- is likely to be elicited from the parent molecule, clude activation, antagonism, synergism, and po- thus enabling the researcher to better design ex- tentiation. Activation has been briefly touched on periments to observe these effects. above with the description of the process of tox- ification: it involves the rendering of an inactive From a properly designed and well-carried-out molecule into an active molecule, usually through metabolic study, the researcher can gain valuable the removal or substitution of a neutralizing fac- insight into the potential toxic actions and bio- tor attached to the molecule, but it may also in- logic effects to be expected from a foreign com- volve the direct addition of a constituent to the pound. If a compound is not absorbed or if the molecule. One of the most common examples of compound is destroyed or rapidly eliminated from activation, the conversion of a precarcinogen to a the body there is little likelihood of pronounced carcinogen, is the addition of oxygen molecules to toxic effects. If the principle metabolic products polycyclic aromatic hydrocarbons via the micro- are polar, conjugation (and thus elimination) somal NADPH-dependent cytochrome P-450 rapidly occurs and again there is little potential mixed function oxidase enzyme system. This con- for pronounced toxicity. On the other hand, if me- version to the oxide produces a carcinogenic tabolism produces an activation product, toxicity agent of considerable potency, whereas the par- is enhanced and biologic effects may be pro- ent material does not possess a direct carcino- nounced.

APPENDIX REFERENCES

1. 41 CI+’R 51206 (1976). 17. Llarzulli, F. N., and Ef. 1. klaibiich ( 1970). J. Soc. 2. Food and Drug Administration Advisory Commi[- Cosmet. Chem. 21 :695. tee on Protocols for Safety Evaluations: Panel on 18. h4aguire. H. C., Jr, ( 1973). ]. Soc. Cosmet. Chern. Reproduction Report on Reproduction Studies in 24: 151. the Safety Evaluation of F’ood Adciit ives and Pest i- 19. Levine, B. B. ( 1960). ~. Exp. A4efi. 112: 1131. icicle Residues ( 1970). Tox, App). Pharm. 16:264. 20. Salvin, S. B. (1965). Fed. Proc. 24:40.

3, A40dern Trends in Toxic~Jlogy ( 1968). E. Eloyiand 214 Turk, J. L. (1964). Int. Arch. Allergy 24:191. and R. Coulding, ecis,, But terworths, London. 22. Bio]ogy of the Immune Response ( 1970). P, Abro- 4. Methods in ‘1’oxicolog}’ ( 1970). G, E. Pagett, cd., noff and hf. LaVie, eds., hlcGraw-Hill, N.Y. Davis, Philadelphia. 23. Textbook of Immunop(lthologv ( 1976). H. G. hlLLl- 5. Frawlev, J. P. ( 1955). Fooff Drug Cosmetic L(IW J. ler-Eberhard, cd., Grune and St rat ton, N. Y.. 1 o: 703. 24. Medicul Physiology ( 1968). V. B. hlountcastle, cd., 6. Sm}th, 1{. I+’,, Jr,, an(i (;. P. Carpenter (1944). J. hfosby, St. Louis. Jnd. Hyg. (Jnd Toxicd. 26:269. 25. A Textbook of Histology ( 1964). W. Bloom and D. 7. Friedenwt~ld, J. S., W. l“, Hughes, Jr., and H. Her- W. Fa~cet t, eds., Saunders, Philadelphia. mann ( 1944). Arch. of Oph th. 31 :279. 26. Pfltholog}r (1971). L$’. A. D. Anderson, cd., Mosbv, 8. Carpenter, C. P., and H. F, Smyth, Jr, (1946). Am. St. Louis, J. Ophth. 29: 1363. 27. Nelson, A. A. [ 1955). Foo(j Drug Cosmetic L~Jw j. 9. Draize, J. H., G. Woodard, and H. O. Calvery 10:732, [ 1944),]. Phc~rm. find Expt. “rherc~peu. 82:377. 28. Comroe. J. H., Jr,, R. Il. F“orster II, A. B. Dubois, W. 10. I)raizc, J. 11., and E. A. Kelly ( 1952). Proc. of the A. Briscoe, and E. Garleson ( 1962). ‘rhe Lung: (lin- Sci. Sec. of Toilet G(MN]S Assoc. 17:1. ic{d Physiology (lntl Pulm~~n(] r-y Func ti(~n Tests, 11, Ilraize, J. H. ( 1955). Foo{j Drug Cosmetics Low ]. Yearbook hfedical Publ., (;hicago. 10:722. 29. Armstrong, hl. L. ( 1974). Ii~[~ctr[)cf)r-( ii(~gr{lms, 12. Draize, J. 11. ( 1965). Assoc. of Food and Drug Yearbook hledical Publ., Chicago. Of ficicds ~)f the U. S., Topeka. Kans., 46. 30. Current Practice of Clinical El[~ct~()(>rlc[?l~h~)l()g- 13. Lansteiner, K., and J. Jacobs (1935). ]. Exp. Med. r(lphy ( 1977). D, W’. Klass and D. D. Daly, eds., 61: 643. Raven. 14. Lansteiner, K,, and J. Jacobs (1936). ]. Exp. Meal, 31, Goodale, R. 11. ( 1965). Clinicul Interpretation o.f 64:625. Lahor~ltory Tests. Davis, Philadelphia. 15. Buehler, E. V. (1!365). Arch. Derm, 91 :171. 32. Henry, R. J. ( 1974). (linic(d (;hernistry: Principles 16. Baer, R. L., S. Rosenthal, and C, J. Sims [1956). J. an{j Techniques, Harper hledical. Invest. Derm. 27:249. 152 ● Environmental Contaminants in Food

33. Amuino, J. S, (1976). Clinical Chemistry: Prin- Detection, vol. 1, A. Hollaender, cd., Plenum ciples and Procedures, Little. Press, New York, 34. Xintaras, C., and B. L. Johnson (1976). Essays in 59. Slater, E. E., M. D. Anderson, and H. S. Rosen- Toxicology, W, J. Hayes, cd., 7:155, Academic kranz (19i’1). Cuncer Res. 31 :97o, Press, New York. 60. Chu, E. H. Y. ( 1971). In: Chemicul Mutagens: Prin- 35. Ekel, G. J., and W. H. Teichner (1976), An ciples and Methods for Their Detection, vol. 1, A. Analysis and Critique of Behavioral Toxicology in Hollaencier, ed., Plenum Press, New York. the U. S. S. R., 13 HEW (NIOSH) Publication No. 61. Kao, F-T, and T. T. Puck (1974). Methods Cell BioJ 77-60, Cincinnati. 8:23. 36. Behavioral Toxicology (1974). C. Xintaras, B. L. 62. Cline, D., and J-A F. S. Spector (1975). Mut. Res. Johnson, and 1, DeGroot, eds. DHEW (NIOSH) 31: 17. Publication No. 74-126, Cincinnati. 63. Regan, J. D., and R. B. Setlow ( 1973). In: Chemical 37. Weiss, B., J. Brozek, H. Hanson, R. C. Leaf, N. K. Mutugens: Principles and Methods for Their De- Mello, and J. M. Spyker ( 1975). Principles for Evu]- tection, vol. 2, A. Hollaender, cd., Plenum Press, uu ting (Jhemicais in the Environment t, 198 Na- New York. tional Academy of Sciences, Washington, D.C. 64. Sega, G, A., J. G. Owens, and R. B. Cumming 38. Alc(; arm, J., and B. h!. Ames ( 1976). Ann. fV. Y. (1!276). Mut. ~t?S. 36: 193. Ac{](I. Sci. 271 :5. 65. ‘1’aylor, J. H. ( 1958). Genetics 43:5 15. 39. Bridges, B. A. ( 1973). Environ. Heo]th Perspect 66. Allen, J. W., and S. A. Latt ( 1976). Nature 6:22 1, 260:449. 40, Bridges, B. A. ( 1974). Mu(. H[?s, 26:335. 67. Abrahamson, S., and E. B. Lewis (1971). In: Chem- 41. b’lamm, W. G. ( 1974). h’fut. Res. 26:329. ical Mutagens: Principles (]n(f Methods for Their 42. Ehrenberg, L. ( 1973). In: Chernicul Mu ta~ens: Detection. vol. 1, A. Hollaender, eci., Plenum Princi{j]es (In(i M[~lh(xls for Their Detection, vol. Press, New York. 2. A. }Iollaender, d., Plenum Press, New York. 68. Vogel. E., and El. Leigh ( 1975), Mut. Hes. 29:383. 43. Underbrink, A. E., L. A, Schaierer, and A. H. 69. Casida, J, E. ( 1969). In: Micr-osomes (Infl Drug Oxi- Sparrow (1973), In: Chemical Mutagens: Princi- du tion, J. R. Fouts and G. J. hlannering. eds., Aca- ples and Methods for Their Detection, vol. 2, A. demic Press, New York. Hollaender, cd., Plenum Press, New York. 70, Smith, R. 11., and R. ~;, von Borstel ( 1973). In: 44. Ames. B. N.. F. 1). Lee, and W. E. Durston (1973). (jhemic~d Mut(],gens: Principles und Methods for Pr(m. lV(It, Ac(Id. Sci., U.S.A. 70:782. Their Detection. vol. 2, A. Ilollaender, ed., plenum ’45. Ames, B, N., J. hlc(hnn, and E. Yamasaki ( 1975). press, New York. Mut. ~[?S. 31 :347. 71. !+:hmid, W. ( 1975). Mut. R[?s. 31 :9. 46. I.[]ishes, B. A., and 11. E’. S[ich ( 1973). Biochern. 72, [,eoll:lrd, A. ( 1 g73), In: Chemical Mutugens: Prin- Biophys, Res. [lmlmun. 52:827. ciples un[j Meth(~(js for Their Deter; tion, vol. 2, A. 47. Brusick, D. J., :]nd V. W“. Nfavcr ( 1973). Env. }Iollaender, cd., Plenum Press, New York. He({lth Persp[’c[. 6:83. 7:3. I,eonard, A. ( 1975). Mut. HCS. 3 I :291. 48. ong, ‘1”-hf, []nd 1{. V. ifal]in~ ( 1975). Mut. lies. 74. Russell, 1,. B. ( 1976). [n: (11[’mi~;~ll Mut{lgens: 31 : 195. Principles (ln(j Meth(xis ff~r Their fletection, vol. q:] . hlohn. {;.. J. Ell[?nlwrger, :]nd D. hlc~re~or ( 1974). 4, A. Ilollaencier, cd., Plenum Press, New York. MU (. 1{[]s. 25: 187. 75. Batem~~n A. J. :~rld S. S. ~PStein (1971). In: 50, Mortimer, R. K., and T. R. Manney (1971). In Chemic~lf Mut~lgens: Principles (In(i Metho(is for Chemicul Mutagens: Principles and Methods for Their Detection, vol. 1, A. Ilollaender, cd., Plenum Their Detec~tion, vol. 1, A. Hollaender, cd., Plenum Press, New York. Press, New York. 76. Sontag, J. N1., hf. P, Page, and U. Saffiotti ( 1976). 51. De%rnes. ~’. J., and If. V. hl:]lling ( 1971). In: In: Cui(ielines for Curcin(~gen Bimlss[]y in Sm(dl [;hcmi[;(il MU tflgens, Principles (]n(i Meth(xls for Ho(jen ts. U.S. Department of Health, Education Their Detecti{jn, vol. 1, A. 1 Iollaender, cd., Plenum and Welfare, Public Health Service, National In- Press, New York. sti tutes of [iealth. National Technical In forma- 52. Zimmerman l+’. K. ( 1975). Mut RCS. 3 1:71. t ion Service PB-264 061, Springfield, Va. 53. Gabri(ige, Ll, C., A. L)eNunzio, and M. S. Legator 77. Saffiotti, [J., :]nd 11. Autrup ( 1978). In: In Vitro ( 1969).N(ltur[?68:221, (1] rc;inogenesis; Guide to the Liter(lture. A(l- 54. Siebert, D. ( 1973). Niut. Res. 17::~07, v(lnces (ln~l Lahor(ltory Procedures, U.S. Depart- 55. Durston, M’. ~;., fln(~ Il. N. Ames (1!174). Proc. Nut. ment of I Ieal t h, Education and Welfare, Public Ac(I(I, Sci. USA 71 :737. }iealth Service, National Institutes of Health. Na- 56. Commoner, B.. A. J. frith:lv;ithil. and J. Henry tional Technical Information Service PB-281 737, ( 1974). N(lt urc 249:850. Springfield, Va. 57, Brew en, J. (;., P. Nettesheim, i~nd K. P. Jones 78. (~arcinogenicit y ( 1973). In: The Testing of Chem- ( 1970), Mu(. 11(:s, 10:645. ic(ils f~~r t~cll.[;inf~g(~nic;ity, Mu t(lgenic;ity, (Ind ‘I’eru - 58, I,eg[}tor, hf. S., (l nd I{. ~’. hi;} II ing [ 1971 ). In: Chem togenicity, Nlinistry of Ilealth and Welfare, i[;(li Jlfu t(]gens: Prin(; iples (ln(l Metho(ls for Their (;anada . Append/x C—Methods for Toxicologic Test/rig ● 153

79. 43 (;F’R 37336 ( 1978). 90, McCann, J., and El. N . Ames ( 1976). Proc. Nat. 80. Purchase, I. F. H., E. Longstaff, J. Ashby, J. A. Acud. Sci. U.S.A. 73:950. Stvies, D. Anderson, P. A. Lefevre, and F, R. 91. IIeddle, J. A., and W . R. Bruce ( 1977). In: Origins Westwo(d ( 1978). Br. /. (hncer (in press). of Humun Cflncer. 13(x)k C. Human Risk Assess- 81. St Vlcs, J. A. ( 1977), Br. ], Cancer 36:558. ment. H. H. Hiatt, J. D. VL’atson, and J. A, W“insten, 82. VL’illiams, I). J., and B. R. Ral)in ( 1971). Nclture eds. Cold Spring Harbor 14abor a torv. 232: 102. 92. ‘reratogenicity (1973). In: The ‘1’[?s ting of Chem- 83. Bock; h’. 11., and R. Mund ( 1958), (llncer Heseurch iculs for C(1 rcinogcn ici ( 1, M u t(lgf’n ici ty, (ln(i Terf I- 18:887. tr)genicity, Ministr>r of Ile:]lth an(i L’1’elfarc, 84. Iverscn, (), }1.. and A. Evensen ( 1962). In: F:XJX>ri- Canada. men t~~~ Skin (l] rein ogencsis in Mice, Norwegian 93. Courtney, K. D., and hi. ( ;h[?rnoff ( 1974). In: ‘~rf]in- IJniversit y Press. ing M(lnuol for 7’er[l t{jl~)g}’. office of Research 85. I,ongst{]ff. E,, and F. R. L1’estwood ( 1978). [Jnpub- and Development, U.S. Environmental Protection lishe(i. Agency, Research ‘1’ri:]n~le Park, N.C. 86. DiPaol:), J. A., R. L. Nelson, P. J. Donovan, and C, 94. Green, E. L. ( 1962). Gcnctif;s. N.Y. 47: 1085. 11, Ev:)ns ( 197:3). Archs. P(]thol. 95::380. 95. Compartment ts, Pools (~n(i Sj)fJces in Mwliccd 87. Epstein, J., R, W’. Rosenth:]l, and R. J. F;ss [ 1955). Physiol[~g~ ( 1967). P.-Ii. E. [3ergner ;ind f;. C. ],ush- Anfll. fllcrn. 27: 1435. baugh, eds. U.S. Atomic F;nergy (X)mmission, Divi- 88. Szcnt-GVorg yi, A., 1. Is(:nberg, and S. IJ. Baird sion of ‘1’echnicnl In forma t ion, 0:] k Rid8e, ‘1’enn. ( 1960). f%(~c. N(lt. Ac~Ici. Sci. USA. 16: 1444. 96. Purchase, I. ( 1976). N(It ure 264:624. 89. Szent-Gvorg}’i, A., :]n(i J. hlcLau8hlin (1!361 ). Prf)c. Nflt. Ac[I~l, Sf;i. 11. S.A. 47: 1397. Appendix D

Review and Evaluation Methods of Determining Risks From Chronic Low- Level Carcinogenic Insult *

by Kenny S. Crump and Marjory D. Masterman

BASIC PRINCIPLES AND CONSIDERATIONS

To aid in determining a proper regulatory ac- order to be 99-percent sure of detecting any car- tion regarding a carcinogen that is present in cinogenic response at all (i. e., for there to be a man’s environment, whether it be a feed additive, probability of 0.99 that at least one animal gets industrial pollutant, or otherwise, it is helpful to cancer). If background or spontaneous carcino- have some knowledge about the number of extra genesis is present even larger numbers of animals cancers that are likely to be caused by the pres- will be required, On the other hand, the extra hu- ence of the carcinogen in the environment. It is man risk that we may want to estimate resulting also helpful to have some knowledge of the likely from environmental exposure is usually (and change in number of extra cancers that would ac- hopefully) smaller than 1/100 for any given chem- company some projected increase or decrease in ical, perhaps on the order of 1/1,000,000. It is the level of human exposure occurring either as a clear that it would not be practical to conduct an result of regulatory action or inaction. This kind experiment with enough animals to measure di- of information is usually impossible to obtain di- rectly an increase in risk this small. rectly from human data. For this reason it is often For these reasons the procedure has been de- necessary to use data from animal feeding experi- veloped of conducting lifetime animal feeding ex- ments to estimate human risk. This procedure in- periments using, in addition to a control dose of volves two difficult steps: 1) relating the animal zero, several doses at which the projected extra risk at high doses to doses very near to zero and 2) cancer risk may be 1/10 or larger, These high- relating the animal risk to risk in humans. dose data are then used to estimate the extra risk Typically, animal experiments use on the order at a dose where the extra risk may be no larger of 100 animals at each experimental dose. If a than, say, 1/1,000,000. An equally important vari- particular experimental dose causes a lifetime in- ant to this problem is the calculation of the so- crease in cancer risk of 1/10, this increase can be called “safe” dose, that is, a dose for which there measured with a small degree of accuracy using is some measure of statistical assurance that 100 animals. But if the increased cancer risk is the extra risk at that dose is no more than, say, less than 1/100 this increase will often not even be 1/1,000,000 These problems are often referred to detectable by an animal feeding experiment. For collectively in the literature as the ‘‘low-dose ex- example, if the true risk is 1/100 it would require trapolation problem. ” that over 400 animals be tested at that dose in

Performing a low-dose extrapolation involves the choice of statistical procedures to apply to the the choice of a mathematical function to model mathematical function. The choice for this mathe- the dose-carcinogenic response relationship and matical function turns out to be extremely crucial to the outcome of low-dose risk estimation. If the *Excerpt from OTA Working Paper entitled “Assessment of assumed relationship between tumor occurrence Carcinogenic Risks From PCBS in Food.”’ A complete copy of the paper can be obtained from the National Technical Infor- and dose does not apply in the regions to which mation Service. (See app. J.) the extrapolation is being made, a serious over-

154 Appendix D—Review and Evaluation of Methods of Determining Risks From Chronic Low-Level Carcinogenic Insult ● 155 estimate of the “safe” dose may result (Mantel significantly better fit to data than several others and Bryan, 1961, p. 458). Chand and Heel (1974) this would still not furnish assurance that this compared five standard dose-response models function would necessarily correlate in any way and observed that they could differ by many with the true dose response at very low doses orders of magnitude at low dose levels for which where it is not feasible to measure the true extra extra risks are on the order of 1/100,000,000. risk directly. As a consequence of the great dis- It might be supposed that it should be possible parity of dose-response functions at low doses it is to discriminate among the various potential dose- imperative that the dose-response function be se- response functions on the basis of experimental lected, neither arbitrarily nor solely on the basis data but, unfortunately, two different dose-re- of how well it can be made to fit experimental sponse functions can often fit experimental data data, but, insofar as is possible, it should reflect equally well but still differ by several orders of known or at least plausible information regarding magnitude at very low doses. Moreover, even if a the biological mechanisms through which a chem- particular dose-response function were to give a ical induces or promotes cancer.

WHAT SHAPE SHOULD BE EXPECTED FOR THE DOSE-RESPONSE CURVE AT LOW DOSES?

Tumors of so many different types arise in such ly flat at zero dose. What this means is simply that a diversity of different tissues, their etiology is so when risk is plotted against dose response on or- little understood, and the agents that cause dinary linear scales, the tangent line to the dose- tumors affect a subject in such diverse ways, that response curve at zero dose should have a posi- it might seem that no general conclusions can be tive slope. When a dose-response function has drawn. However, for a certain broad class of “di- this property we will say it is linear at low dose. rectly acting” chemical carcinogens the range of This simple property can have far-reaching con- uncertainty associated with the shape of the dose- sequences on low-dose extrapolation. For exam- response curve at low doses can be greatly nar- ple, consider the two potential dose-response rowed. As used in this paper, the term “directly functions: acting carcinogen” encompasses (Guess, Crump, and Pete, 1977) carcinogenic agents for which either the agent itself or a metabolize acts directly Both of the curves give a risk of 1/10 at a dose of d at the cellular level and produces a heritable = 1 and are practically indistinguishable at change that eventually to the formation of a higher doses. However, at a dose of d = 1/1,000: tumor. Carcinogens that are carcinogenic by rea- 1) predicts a risk of 1/100,000 and 2) predicts a son of their mutagenicity should fall into the cate- risk of 1/10,000,000, a difference of two orders of gory of “directly acting carcinogens. ” According- magnitude. We note also that 1 ) has a tangent line ly, carcinogens that test positively using the Ames with a positive slope at d = O whereas 2) does not. mutagenicity screening test for carcinogenicity One explanation of why the dose-response are very likely to be directly acting (see McCann function should be linear at low dose when back- and Ames, 1976). In a recent study (McCann, ground is present may be found in Crump, et al. Choi, Yamasaki, and Ames, 1975) in which about (1976) and Peto (1977) and will be briefly outlined 300 carcinogens and noncarcinogens were tested here. When background carcinogenesis is pres- using the Ames test, 90 percent (157 out of 175) of ent, the cellular mechanism through which the the carcinogens were mutagenic including almost test agent produces cancer should already be op- all of the known human carcinogens. This indi- erative in producing background tumors. When cates that the class of directly acting carcinogens this is true the effect of the test agent is to add to may encompass most of the known carcinogens, an already ongoing process. The result of this A partial solution to the low-dose extrapolation additive effect is illustrated in figure D-1. The problem for the case of directly acting chemical dose-response curve is for all tumors produced carcinogens has been given in Peto (1977), Crump, through the mechanism through which the test Heel, Langley and Peto (1976), and Guess, et al. agent acts. Background carcinogenesis is allowed (1977). The key result is that, at least as long as for in the figure by an effective background dose background carcinogenesis is present, we should do. We see that, in this case, the added risk expect the dose-response curve not to be absolute- caused by a dose d of the test agent should be ex- 156 ● Environmental Contaminants in Food pected to increase approximately linearly near d a number of events are required to occur at the = O (i.e., the tangent line at d = O will have a posi- cellular level to initiate cancer. Although all mod- tive slope). Implicitly assumed by the way figure els in this class are linear at low dose provided D-1 is drawn is the fact that an added dose of a background carcinogenesis is present, a sizable carcinogen acting through this mechanism does subclass of them are linear at low dose in the not produce a smaller risk, If background carcino- absence of background carcinogenesis (Crump, et genesis is allowed for as in figure D-1 by positing al., 1976). an effective background dose do that is estimated Watson (1977) has recently proposed a more from the data, then the wide range of risks ob- specialized model for cancer induction and pro- tained using different models effectively disap- motion based on reversible epigenetic cellular pears (Pete, 1977). We note that the existence of a changes. Watson concludes that his model sup- tangent line with a positive slope at zero dose ports the low-dose-linearity hypothesis and states does not, in itself, imply any lower bound for extra “as suggested by a different argument of Crump, risk at low doses since the slope of the tangent et al. . . . it is reasonable to assess the risk due to line could possibly be very small. an additional carcinogen at low constant dosage The evidence given above for a positive slope to by a linear relation. ” the dose-response function at zero dose applies The evidence for low-dose linearity given above particularly to the case in which background car- applies mainly to directly acting carcinogens. An cinogenesis is operative. This does not imply that indirectly acting carcinogen might be one that we expect the dose-response curve not to be lin- causes some gross physiological change such as ear at low dose in the absence of background car- suppression of ovulation that could predispose cinogenesis. For example, the multistage models the subject to cancer. For such carcinogens the of cancer (Armitage and Doll, 1961) are a fairly shape of the dose-response curve at low dose is broad class of models in which it is assumed that highly speculative. There could possibly be a threshold dose below which the agent has no car- cinogenic effect at all on an individual, However even if a threshold mechanism is operative, there is likely to be considerable variation in individual Figure D-1 .—Illustration of Why the Dose” Response thresholds in a large population. Consequently the Curve Should Be Linear at Low Dose in the Presence dose-response curve for the entire population

of Background Carcinogenesis. dO Is the Effective could still exhibit a linear trend at risks as low as Background Dose and d Is the Dose of the 1/1,000,000 or lower. Carcinogen of Interest The effects of metabolic activation and detox- ification on carcinogenic dose response have been recently considered by Cornfield (1977) through a kinetic model that encompasses free toxic sub- stance, metabolize, deactivator, and the interac- tions of these substances. Only a steady-state sit- uation is studied in that variation over time of the concentrations of these agents is not considered. The model predicts a threshold dose below which there is no carcinogenic risk under the assump- tion that the deactivator is 100-percent efficient in deactivating the carcinogen. However, in a nat- urally occurring process it is likely that deactiva- tion would not be perfect and would be less than 100-percent effective in always combining with 100 percent of the carcinogen before an amount of the active metabolize reaches a cancer target at d =0 site, Any of a number of modifications to the model to allow for nonperfect deactivation would rule out a threshold and would lead directly to a model for which carcinogenic response varies linearly with dose at low doses, Cornfield’s own modification of perfect deactivation, that of al- Appendix D—Review and Evaluation of Methods of Determining Risks From Chronic Low-Level Carcinogenic Insult ● 157

lowing the deactivating reaction to be reversible, a mechanism that would lead to a more conserva- leads, as Cornfield points out, to a model which is tive dose-response relationship such as the risk linear at low dose. This occurs regardless of how varying approximately as the square root of dose slowly the reverse reaction takes place, as long as at low dose. In view of these uncertainties it the possibility is not eliminated entirely. Further- would seem reasonable to base estimates of more, even in the extremely unlikely case of added risk of cancer on a mathematical model perfect deactivation, an otherwise realistic model that encompasses low-dose linearity unless, of should still imply low-dose linearity since the course, the mechanism through which the car- theoretical time required for perfect deactivation cinogen operates is sufficiently understood so would not be zero and would likely be infinite. that low-dose linearity can be conclusively ruled For most, perhaps all, carcinogens, the mecha- out. Once the principle of low-dose linearity is ac- nisms through which cancer is produced are not cepted the problem of estimation of risks at low sufficiently understood so that the shape of the dose is nearly solved. This is because the dis- carcinogenic-response curve can be theoretically agreement between the upper statistical con- predicted with certainty. As pointed out earlier, fidence bounds on risk at low doses based on a neither can experiments of sufficient size be con- model that incorporates low-dose linearity and ducted that would permit direct experimental in- one that does not is typically several orders of vestigation of the dose-response curve at low magnitude whereas the corresponding disagree- dose. We have noted that there are plausible ar- ment between two reasonable models both of guments that the dose-response curve is linear at which incorporate low-dose linearity is usually low dose for many carcinogens. On the other much less than this. hand, this author knows of no serious proposal of

Mantel= Bryan that a set of DES data (Gass, Coats, and Graham, 1964, C3H females) exhibited a probit slope of The Mantel-Bryan procedure as originally pro- one-half, but the general use of a probit slope of b posed (Mantel and Bryan, 1961) and “improved” = 1 was still suggested. (Mantel, Bohidar, Brown, Ciminera, and Tukey, With the probit slope parameter fixed at b = 1 1975) is for the purpose of conservatively choos- the remaining parameters, a and C, are estimated ing a “safe” dose of a carcinogen, a “safe” dose from the data and then adjusted so as to produce being defined as one for which it can be expected a higher level of risk at a given dose that corre- that, with a given level of statistical assurance sponds to an upper 99-percent statistical limit on (e.g., 99 percent), the true dose producing a pre- the true risk at a given dose. The safe dose is then assigned “safe” level (e. g., 1/1,000,000) of risk determined to be the one producing a given low will lie above the “safe” dose. In the Mantel- risk (e. g., 1/1,000,000) based on the adjusted val- Bryan procedure the mathematical model used ues of a and C. for the dose-response model is the probit function: As pointed out in Mantel, et al. (1975), the Mantel-Bryan procedure rewards larger and bet- ter experiments in that the more evidence there is where d represents the dose of the carcinogen of safety, the higher the calculated safe dose will and P(d) represents the probability of a cancerous be. However this advantage should be shared by response in an animal subjected to a dose d. The any extrapolation method that uses reasonable parameters in the model are an intercept parame- statistical procedures. ter a, a probit slope parameter b, and C, which Some have considered the Mantel-Bryan proce- represents the probability of a response in un- dure to be too conservative (Federal Register, vol. treated animals. The parameter b is not estimated 42, 1977, p. 10419) in that it involves three con- from the data but rather is arbitrarily set equal to servative choices (99-percent statistical assur- 1. This choice is stated as being conservative ance, lifetime risk of 1/1,000,000, and probit slope (Mantel and Bryan, 1961), the argument for this set equal to 1) and that any one of these assump- being that typical dose data exhibit a probit slope tions alone could provide adequate protection to in the experimentally observable region above 1- the public. The first two of these choices are regu- percent incidence that is greater than one, In latory decisions that would have to be made with Mantel and Schneiderman (1975) it was observed any extrapolation procedure. However, the arbi- 158 ● Environmental Contaminants in Food trary selection for the slope parameter seems to . response relationship does not apply at the low be peculiar to the Mantel-Bryan procedure. To in- risk levels to which extrapolation- is ‘being made. vestigate its effect upon the extrapolation proce- Thus, before the degree of conservation can be dure a typical fit is presented in figure D-2 (from evaluated for any procedure, the properties of the Crump, 1977b) of the Mantel-Bryan probit curve dose-response curve at very low doses must be (equation 1) to experimental carcinogenicity data evaluated, when the probit slope parameter is fixed at b = 1. As described earlier, there are strong argu- As can be readily seen the probit curve typically ments that indicate the dose-response curve provides a very poor fit, curving downward even should be “linear at low dose” particularly for when the trend of the data is toward an increas- directly acting carcinogens in the presence of ingly upward curvature. This typically bad fit to background carcinogenesis. This has led Peto data of the probit curve raises serious questions (1974) to recommend extrapolation procedures regarding the validity of statistical procedures as- using only dose-response functions from a class sociated with the Mantel-Bryan method (see Sals- containing only dose-response functions that are burg, 1977; and Crump, 1977b). linear at low dose, At the very least, however, it On the other hand, the Mantel-Bryan proce- would seem prudent not to go to the opposite ex- dure utilizing the choice b = 1 was put forth as treme and use a dose-response function that rules conservative procedure and it gives that appear- out linearity at low dose by assumption, However, ance in figure D-1 since the probit curve appears the Mantel-Bryan procedure, through its use of to lie far above the trend of the data at the lowest the probit curve (equation 1) rules out linearity at doses. However as mentioned earlier and also low dose in favor of a “flatness property” (see pointed out by Mantel (Mantel and Bryan, 1961, p. Hartley and Sielken, 1977; Mantel, 1977; and 458) a procedure may, while appearing conserva- Crump, 1977b) at low dose which is anticonserva- tive at experimental dose levels, at the same time tive to the extreme. This property implies that seriously overestimate the “safe” dose (i. e., be mathematical derivatives of all orders of the seriously anticonservative) if the assumed dose- probit curve approach zero (through positive val- ues) as the dose approaches zero. This unusual Figure D-2.—Typical Fit of Mantel” Bryan Curve property is more often discussed within the con- to Experimental Data (From Crump, 1977b) text of mathematical oddities rather than in con- nection with a scientific investigation. It implies o that if the true dose-response curve comes from an extremely broad class of functions known as analytic functions and which pervade scientific applications of mathematics, then the probit curve will eventually underestimate the true risk Mantel Bryan at low doses. Furthermore, at low enough doses, maximum likelihood the degree to which the risk will be underesti- curve mated will be arbitrarily large (i.e., the ratio of \ the true risk to the probit estimate will grow ar- bitrarily large). — It was emphasized earlier that it is important “safe” dose curve when extrapolating to low doses for the assumed dose-response function to incorporate known or at least plausible facts about the mechanisms of carcinogenesis. In neither the original paper (Mantel and Bryan, 1961) nor in the paper outlin- ing the improved version is biological justification given for the selection of a curve having the above described “flatness property.” It should be men- tioned at this point that the incorporation of back- ground carcinogenesis into the Mantel-Bryan probit model (equation 1) using the parameter C o implies that the mechanism through which the 0 2 4 6 test carcinogen produces cancer is independent Dose of the mechanisms through which all of the back- Appendix D—Review and Evaluation of Methods of Determining Risks From Chronic Low-Level Carcinogenic Insult ● 159 ground cancers are produced (Crump, et al., while retaining the straight line relationship. 1976). In keeping with the discussion in the last When data at other experimental doses are avail- section it would seem more proper to incorporate able this method of linear extrapolation has the background into the probit model by positing an obvious shortcoming of not fully utilizing the effective background dose do which adds to the available data. dose d of the test carcinogen. If background is in- A linear dose-response curve is linear at low corporated in this way the probit curve no longer dose but the converse is not necessarily true. A has the “flatness property” and becomes linear curve can be linear at low dose and still have a at low dose (Guess, et al., 1977). In fact. with high degree of nonlinearity at higher doses. background incorporated in this way, the probit curve assumes a shape similar to the one-hit mod- Linear extrapolation is viewed by some as a el, sometimes referred to as the most conservative very conservative procedure. For example, com- of all procedures (e.g., Mantel, 1977). ments were made during the decision on which Even though the “flatness property” implies extrapolation procedure to incorporate into the the probit curve should at suitably low doses be SOM document to the effect that linear extrapola- anticonservative to the extreme, the “flatness” tion is the most conservative of all procedures. property holds only for doses approaching zero Crump, et al. (1976) examined the extent of the and the feature of arbitrarily fixing the probit conservatism of a linear dose-response function slope at 1 mitigates the anticonservativeness im- when compared with a multistage dose-response plied by the “flatness” property at any given low model (Armitage and Doll, 1961). The multistage model assumes that a cell must go through a num- dose (although the property itself will hold for all choices of the parameters a, b, C), The cancer ber of different stages before cancer is initiated risks that are typically extrapolated to are in the in that cell and the model can encompass a high risk ranges 1/10,000 to 1/100,000,000. We will ex- degree of nonlinearity. It was determined that the amine the outcome of Mantel-Bryan extrapola- maximum possible degree of conservatism of a tions to these risk levels in a later section when linear model relative to a multistage model de- we compare them with extrapolations based on pended rather heavily on the incidence at the ex- the multistage model. perimental dose relative to the background inci- dence. (This is consistent with the general rela- Linear Extrapolation tionship between background carcinogenesis and linearity at low dose as discussed earlier. ) For ex- The technique for linear extrapolation was ample, when the incidence at the experimental recommended by Heel, Gaylor, Kirschstein, Safi- dose is four times the incidence at zero dose the otti, and Schneiderman (1975) for use on an in- extra incidence at low dose derived from the lin- terim basis until better procedures could be de- ear dose response differs from the incidence de- veloped. The procedure is straightforward and is rived from the multistage model by, at most, less based on an assumed linear relationship between than a factor of 2.5 regardless of the number of dose and response at low dose. The procedure stages in the multistage process. Thus, when utilizes only the data for the group of control background carcinogenesis is present, the linear animals and a single other dose group, usually dose-response curve is not overly conservative either the highest dose that elicits no response or relative to the multistage dose-response curve. In else the lowest dose that elicits some response. In fact the linear dose-response curve is anticon- the case there are no cancers in the control ani- servative when compared to the one-stage or one- mals the “safe” dose, based on a maximum risk of hit models. 1/1,000,000 and 99-percent statistical assurance, Linear extrapolation has long been proposed is calculated as follows: An upper 99-percent con- for use in radiation carcinogenesis (see Brown, fidence bound is calculated for the cancer risk in 1976, for a review of the relevant reports). The the dose group of animals. From this risk and dose BIER (1972) report on radiation risks from the Na- one extrapolates back toward zero dose and zero tional Academy of Science recommended linear risk using a straight line relationship. The dose extrapolation as a “best estimate” approach as corresponding to a risk of 1/1,000,000 on this opposed to a conservative approach. Brown re- straight line is taken to be the “safe” dose. If viewed arguments both for and against linearity there are cancers in the control animals this pro- and concluded that “linear extrapolation of cedure is modified to allow for the statistical human data from high dose of low LET radiation treatment of the response in the control group cannot be said to overestimate the risk at low 160 ● Environmental Contaminants in Food

doses. In fact, there is some doubt as to whether The dose-response relation (equation 2) con- the risk is not underestimated. ” tains all of the Armitage and Doll (1961) multi- Certainly much remains to be learned about stage dose-response models as special cases but both radiation and chemical carcinogenesis. How- also contains curves that are much flatter at low ever, if both radiation and chemicals cause can- dose than any of the multistage curves. cer through similar mechanisms then it should be The two extrapolation procedures based on expected that there would also be similarities be- equation 2 have some features that are different. tween the respective carcinogenesis dose-re- When computing “most likely” estimates the pro- sponse functions. Direct damage to DNA by the cedure of Guess, Crump, and Deal uses an infinite carcinogenic agent has been implicated as one dimensional maximization procedure so that it is cancer-initiating mechanism for both radiation not necessary to specify a value of k, the degree of and chemicals (Brown, 1976; and McCann and the polynomial in equation 2. However the two Ames, 1976). Thus, the findings related to the methods differ chiefly in the way the statistical potential linearity of the dose-response function confidence intervals are computed. There have for radiation has implications for chemical car- not yet been sufficient comparative calculations cinogens as well, particularly for “directly act- made to determine how safe doses may differ ing” carcinogens. using the two approaches, Mantel (1977) has made a critical review of the statistical procedure Extrapolation Methods Based used by Hartley and Sielkin for calculating the on the Multistage Model “safe’ dose. Both the Hartley and Sielkin and the Guess, Two methods of low-dose extrapolation which Crump, and Deal (as extended by Daffer et al., are alternatives to the Mantel-Bryan or linear 1979) procedures can utilize times at which procedures have recently been proposed inde- cancer is detected in the experimental animals pendently by Guess, Crump, and Deal (Guess and rather than just the dichotomous information of Crump, 1976, 1978; and Crump, Guess, and Deal, whether or not an animal contracted cancer be- 1977) and Hartley and Sielkin (1977). Both of fore it died of some other cause or before the ter- these methods utilize a multistage dose response mination of the experiment, The utilization of time function of the form: data in low-dose extrapolation is important for at least two reasons: 1) the age at which cancers oc- where q~, ql, . . ., q~ are all non-negative param- cur should be important in assessing the magni- eters to be estimated from the data. This dose-re- tude of the harmful effect of a carcinogen on man sponse function is general enough to yield a con- (e.g., cancers that occur early in life should be siderably wide range of responses at low dose. On viewed as more serious than those that occur in the one hand, if q, >0 and q,= O for i~ 2 the dose- extreme old age); and 2) in many animal carcino- response function (equation 2) becomes the one- genicity experiments the response data at the stage model which yields risks at low doses com- highest doses lies below the trend of the lower parable to what would be obtained with linear ex- dose data, This sometimes appears to be due to trapolation, On the other hand, the model can pro- the fact that at the highest doses some of the ani- duce risks even as low as the probit curve (equa- mals are being poisoned by the chemical before tion 1) down to any fixed positive low dose. Thus they have a chance to develop cancer, When this this model is capable of fitting both highly linear occurs the high-dose data is often just deleted and highly nonlinear dose-response relations. from the analysis. However if the times at which Since the model has the property of “linear at low the animals die are properly used the high-dose dose” if q,> O and does not have this property if data might not appear anomalous. More research needs to be done on the proper utilization of ani- ql = O, use of this model does away with having to make the arbitrary but crucial decision of having mal time of death data to assess the harmful ef- to either assume linear at low dose as in linear ex- fects of chemicals to man. trapolation. Thus “safe” doses computed using this model should provide a more realistic meas- Comparisons Between Mantel-Bryan ure of the true uncertainty of low-dose extrapola- and Multistage Extrapolations tion than would “safe” doses based on either an assumed linear curve shape or an assumed highly To compare low-dose extrapolations using the nonlinear curve shape. Mantel-Bryan probit model (equation 1) with Appendix D—Review and Evaluation of Methods of Determining Risks From Chronic Low-Level Carcinogenic Insult ● 161

those using the multistage model (equation 2) we also Crump, 1977 for further discussion of this present figure D-3 based on the same data as point.) On the other hand, we have seen that there figure D-2. In this figure, the Mantel-Bryan “safe” are quite plausible arguments for the true dose- dose is plotted on a log-log scale as well as both response curve to have the same shape at low the multistage “most likely” curve and the “safe” dose (linear) as the estimated multistage curve. dose based on (equation 2) and computed as out- We note that both the multistage “safe” dose lined in Crump, et al. (1977). A 99-percent statisti- curve and “most likely” curve have a slope 1 in cal assurance was used for both “safe” dose figure D-3 as plotted on the log-log scales which is curves. We note that the Mantel-Bryan “safe” equivalent to the curves being linear at low dose. dose lies above the multistage safe dose curve at The fact that the “most likely” curve has slope 1 values of added risk below 5 x 10-4, The Mantel- is due to the fact that with this particular data set Bryan “safe” dose curve lies above the multistage the linear coefficient q, in equation 2 will always “safe” dose curve by a factor greater than 20 for be linear at low dose regardless of whether or not an added risk of 10~ and by a factor greater than the linear coefficient q, is estimated to be positive. 300 for an added risk of 10~. Because of the This property should be shared by any valid sta- “flatness” property of the Mantel-Bryan probit tistical procedure based on a dose-response func- function (equation 1) described earlier, the tion that does not rule out linearity at low dose by Mantel-Bryan “safe”’ dose curve will lie above the assumption as Mantel-Bryan does. Just as it is not multistage safe dose curve by arbitrarily large possible to prove statistically the existence of a factors at extreme low doses. Guess, et al. (1977) threshold, it is likewise not possible to rule out the have compared the Mantel-Bryan “safe” dose possibility that the true dose-response curve is curves to the multistage “safe” dose curves and linear at low dose on the basis of statistical found this to be a typical situation. Thus it is clear analysis. (See Guess, et al. (1977) for a thorough that if the true dose-response curve could be discussion of this important point.) The Mantel- similar to the multistage dose-response function Bryan obtains “safe’” dose estimates which are (equation 1) then the Mantel-Bryan procedure considerably higher than those obtained using the could not be justifiably called conservative (see multistage model because it assumes away linear- ity at low dose, an assumption that we have seen is probably unwarranted for that majority of car- cinogens which are classified as ‘‘directly acting” Figure D-3.—Comparisons of “Safe” Doses carcinogens. Computed From the Mantel” Bryan Procedure Since extrapolation based on a model such as and From a Procedure Based On the Multistage the multistage model (equation 2) must always be Model (From Crump, 1977b) linear at low dose, the question arises as to how different the result will be from simple linear ex- r trapolation. For some data the difference will be minimal. For example, for the data upon which figure D-2 is based, “safe doses” computed using the multistage model are almost identical with *’safe’” doses based on linear extrapolation. For some data sets, however, the difference could be considerable. For example, with the Gass, et al. (1964) DES using the C3H female mice, the “safe” dose based upon linear extrapolation is lower than the “safe” dose based on the multistage model by a factor of about five and there are other data sets where this difference is greater than an order of magnitude.

The Gamma Multihit Carcinogenesis Dose-Response Model Rai and Van Ryzin (1978) have proposed basing Added Risk risk estimation on the gamma multihit model: 162 ● Environmental Contaminants in Food

by orders of magnitude at low doses. The reason these large differences can occur is as follows. The multistage family of models contains mem- bers which are simultaneously linear at low dose and exhibit upward curvature at moderate doses. where P(d) is the lifetime probability of cancer in For example, the particular multistage model: a tissue when subjected to a constant dose rate, d, of the carcinogen. This model is obtained by as- is linear at low dose since q, >0 and still can ex- suming that cancer due to the carcinogen occurs hibit upward curvature at moderate doses since randomly according to a Poisson distribution. The qz >0. This means there are dose-response curves manner in which background carcinogenesis is in- in the multistage class which are both linear at corporated into the model is equivalent to assum- low dose and can adequately describe data of the ing that the event “cancer occurs due to the ac- type exemplified by figure D-4b. On the other tion of the carcinogen” is independent of the hand, this will generally not be true of gamma event “cancer occurs spontaneously, ” This as- multihit models. All dose-response curves in the sumption would not apply, for example, to proc- esses in which the effect of the carcinogen is to speed up the rate at which the “spontaneous” Figure D-4a.— Example of Data Exhibiting events occur which lead to the background can- Downward Curvature cers. At low dose rates, the response is approxi- mately given by:

Consequently, this dose response model is linear at low dose rates when and only when k = 1. Rai and Van Ryzin calculate confidence intervals for added risk at a given dose and for the dose pro- ducing a fixed added risk using asymptotic max- imum likelihood theory. Although in the theoreti- % cal development k must be an integer, in the appli- cations k is allowed to assume any positive value. x Lower statistical confidence limits on the dose producing a given low amount of extra risk (“vir- tually safe dose” or “VSD”) computed using this procedure can be compared with those computed Dose from the multistage model (equation 2) by consid- ering two general classes of data. If the data exhibit a general downward curva- ture as illustrated in figure D-4a, lower con- Figure D-4b.—Example of Data Exhibiting fidence limits on a VSD computed from the gamma Upward Curvature multihit model (equation 3) should generally be less than or equal to corresponding lower con- fidence limits computed from the multistage model (equation 2). In certain instances the gam- ma multihit lower limits could be much less than the corresponding multistage lower limits. This could occur when the data is consistent with k <1 in the gamma multihit model (presumably cor- responding to a fraction of a hit). On the other hand, if the data exhibit a general upward curvature as illustrated in figure D-4b the reverse situation will hold; gamma multihit lower confidence limits on VSDs will generally be greater than or equal to corresponding multistage limits, Gamma multihit limits will not in general share the low-dose linearity of multistage limits Appendix D—Review and Evaluation of Methods of Determining Risks From Chronic Low-Level Carcinogenic Insult ● 163

gamma multihit class that are linear at low dose multihit model may be either “super linear” (cor- must exhibit downward curvature and conse- responding to k < 1) or “sublinear” (correspond- quently would generally not be consistent with ing to k > 1). Superlinearity is achieved by making the data in figure D-4b. When this is true gamma the model too broad in that a fraction of a hit is multihit lower confidence limits on VSDs will be allowed which has no biological basis. On the sublinear (e.g., quadratic) at low dose and conse- other hand, sublinearity is achieved by making quently much larger than the multistage confi- the model possibly too restrictive in that models dence limits. which are reasonable from a biological viewpoint Confidence limits based on the gamma multihit are ruled out by assumption. model will be approximately correct whenever In addition to questions as to the appropriate- this model is the correct model, A similar state- ness of the gamma model, there is also a very sig- ment could be made for the probit model, multi- nificant difficulty with the statistical procedure stage model, or any other model to which reason- applied to the model in the Food Safety Council able statistical methods are applied. However, (1978) report. This difficulty is illustrated in table there may be considerable uncertainty as to what D-1. Here are listed lower confidence limits for the true model may be in a particular situation. VSDs for the gamma multihit model which were The multistage model not only reflects some rea- taken from table D-1 in Food Safety Council sonable assumptions regarding the carcinogenic (1978). For each of the 14 data sets a goodness-of- process which dovetail nicely with epidemiologi- fit test was conducted for compatibility of the cal data for many cancers (Pete, 1977 b), but it data to the subclass of gamma models for which also reflects some of the uncertainty with regard k =1. As indicated by table D-1, the subclass of to the true model by virtue of encompassing a models with k = 1 provided an adequate fit in 10 relatively large class of dose-response functions, out of the 14 data sets. For these 10 data sets are For example, as noted earlier, the multistage listed the VSD predicted by the best fitting gamma class contains dose-response functions which are model with k = 1. The significant point is that in 8 linear at low dose and also exhibit upward curva- of these 10 cases, the lower confidence limit on ture at moderate doses. On the other hand, the the VSD is greater than the VSD from the model gamma multihit class is more restrictive at this with k = 1, greater in some cases by enormous fac- point in that it does not permit such behavior. Is tors, e.g., a factor of 3,600 for B,, a fac- this extra restrictiveness of the gamma multihit tor of 1,000 for diethyl-nitrosamine, and a factor model justified? To help answer this question, of 18,000 for sodium saccharin. Put another way, consider the following modification to this model. in the case of the sodium saccharin data, there is Suppose that the hits (phenomenological events a member in the gamma multihit class which fits which are required to occur in a tissue in order the data quite adequately and for which the VSD that a cancer appear) can possibly occur sponta- is less by a factor of 18,000 than the lower con- neously in the absence of the carcinogen. Part of fidence limit for the VSD reported in Food Safety the effect of the carcinogen could then be to speed Council (1978). This obviously implies there is a up the rates at which the spontaneous hits are oc- serious problem with the computation of the curring. For example, if one of the “hits” is in in- lower confidence limits for the VSD. These lower correct base substitution in DNA during mitosis, confidence limits are much too large, even based the carcinogen could speed up the rate at which on assuming the gamma multihit model is the cor- these “hits” are occurring in individual cells by rect model, for these 8 data sets. This problem providing more aberrant bases for use in such an also exists for the 4 data sets which could not be incorrect substitution. With this modification to fit adequately by a model with k =1. This is well- the model, the gamma multihit lower confidence illustrated by the graph in Food Safety Council limits on VSDs will no longer be sublinear at low (1978) for the NTA data. In this graph, the lower dose and should be at least as small as corre- 97.5-percent confidence limit for the VSD corre- sponding limits calculated from the multistage sponding to a risk of 1O-G falls almost directly on model (equation 2). Thus, in order to obtain top of a data point for which the measured risk is sublinear lower confidence limits on VSDs with 1/91. Put another way, even though the measured the gamma multihit model a modification such as risk at this dose is 1/91, the statistical procedure the one described above must be ruled out, not on used indicates, with 97.5-percent assurance, that the basis of data, but by assumption. the risk is no greater than 1/1,000,000. By way of summary, confidence limits based on These difficulties can be overcome by using a the multistage model will always be linear at low different statistical procedure. However, when dose. Confidence limits based upon the gamma this is done, the lower confidence limits will be 164 ● Environmental Contaminants in Food

Table D-1.—Comparison of Lower 97.5-Percent Confidence Limits on Virtually Safe Doses From Gamma Multihit Model Given in to Virtually Safe Doses From Best Fitting Gamma Multihit Model With k = 1 —— —— —. — -— No. of experimental No.——— Substance dose levels—. 1 ‘“ ‘- ‘ -- ‘“ 2 3 4

.0013 5 .035 .000032 6 .020 7 .000029 .00020 8 .0120 p <.005(9 d.f .) significant lack of fit 9 .011 .00016 10 .76 .000042 11 25.3 significant lack of fit 12 13 14 many times smaller than those reported in Food with k = 1 by reason of strong upward curvature. Safety Council (1978). For data sets such as the 10 The four data sets in table D-1 which fall into this referred to in table D-1 which are compatible with category all involved a large number of animals k= 1, lower confidence limits for VSDs must and a number of different experimental dose essentially be calculated from a model with k <1 levels. These four data sets contained from 330 and consequently be smaller than those calcu- animals distributed over 11 dose levels to 720 lated from the one-hit model. This will lead, in animals distributed over 6 dose levels. Most car- many cases, to super-small confidence limits, as cinogenicity bioassays are smaller than these, For illustrated by the lower confidence limit for vinyl example, the standard bioassay for a given sex chloride, For those data sets, such as botulinum and strain in the National Cancer Institute pro- toxin-type A, which are incompatible with k = 1 gram seems to be 200 animals distributed over 4 because of their strong upward curvature, a dose levels. For experiments of this size it is more lower confidence limit for the VSD would still be difficult to rule out k= 1. Consequently it would larger than one computed from the one-hit model. seem that, for most of the data sets that are likely However, the theoretical objections raised earlier to result from a large screening program, appro- to the mathematical form of the gamma model priate statistical procedures applied to the gam- would still apply. ma multihit model would yield lower confidence It may not be generally realized that it is un- limits for VSDs which are even smaller than those common to find data sets which are incompatible calculated from the one-hit model. Append/x D—Review and Evaluation of Methods of Determining Risks From Chronic Low-Level Carcinogenic Insult “ 165

APPENDIX D REFERENCES

1. Armitage, P. and R, Doll. Stochastic models for 15, Hartley, H. O., and R. L. Sielken. Estimation of carcinogenesis. In Proceedings of the Fourth “safe doses” in carcinogenic experiments. Bio- Berkeley Symposium on Mathemuticul Statistics metrics 33:1-30, 1977. and Probability, vol. 4 (Berkeley and Los Angeles, 16. Heel, D. G., D. W. Gaylor, R. L. Kirschstein, U. Saf- Calif.: University of California Press, 1961), pp. fiotti, and M. A. Schneiderman. Estimation of risks 19-38. of irreversible, delayed toxicity. J. o.f Toxicology 2. BIER. The effects on populations of exposures to and Environmental Health 1:1 33-151, 1975. low levels of ionizing radiation, National Academy 17. Mantel, N. and W. R. Bryan. Safety testing of car- of Sciences. National Research Council (Washing- cinogenic agents. J. Nat]. Cancer Inst. 27:455-470, ton, D. C., 1972). 1961. 3. Brown, J. M. Linearity vs. nonlinearity of dose re- 18. Mantel, N., and M. Schneiderman. Estimating sponse for radiation carcinogenesis. Heulth “safe” levels, a hazardous undertaking. Cancer Physics 31: 231-245, 1976. Research 35:1379-1386, 1975. 4. Chand, N., and D. G. Heel. A comparison of models 19. Mantel, N., N, R. Bahidar, C. C. Brown, J. L. Cimi- for determining safe levels of environmental nera, and J. W. Tukey. An improved Mantel-Bryan agents. In Reliability and Biometry, F. Prochan and procedure for the “safety” testing of carcinogens. R. J. Serfling (eds. ) (Philadelphia, Pa.: SIAM, 1974), Cancer Research 35:865-872, 1975. pp. 382-401. 20. Mantel, N. Aspects of the Hartley-Sielken ap- 5. Cornfield, Jerome. Carcinogenic risk assessment. proach for setting “safe doses’” of carcinogens. In Science 198:693-699, 1977. origins of Human Cancer Book C, Human Risk As- sessment, H. H, Hiatt, J. D. Watson, and J. A. 6. Crump, K. S., D. G. Heel, C. H. Langley, and R. Pete. Fundamental carcinogenic processes and Winsten (eds. ) (Cold Springs Harbor, N. Y.: Cold their implication for low dose risk assessment. Springs Harbor Laboratory, 1977), pp. 1397-1401. Cancer I-lesearch 36:2973-2979, 1976. 21. McCann, J., E. Choi, E. Yamasaki, and B. N. Ames. Detection of carcinogens as mutagens in the Sal- 7. Crump, K, S., H. A. Guess, and K. L. Deal. Con- monella/microsome test. Assay of 300 chemicals. fidence intervals and tests of hypotheses concern- Proc. Nat. Acad. Sci. USA 72:5135-5139, 1975. ing dose response relations inferred from animal 22. McCann, J., and B. N. Ames. Detection of carcinogenicity data. Biometrics 33:437-451. 1977. carcinogens as mutagens in the Sa]monella/micro- 8. Crump, K. S, Response to open query: Theoretical some test. Assay of 300 chemicals: Discussion. problems in the modified Mantel-Bryan procedure. Proc. Nat. A cad. Sci. USA 73:950-954,1976. Biometrics 33:752-755, 1977. 23. Pete, R. Paper delivered at the NIEHS Wrightsville 9. Da ffer, P. Z., K. S. Crump, and M. D. Masterman. Beach conference on Low-Dose Risk Estimation, Asymptotic theory for analyzing dose response 1974. survival data with application to the low-dose ex- 24. Pete, R. The carcinogenic effects of chronic trapolation problem (submitted), 1979. exposure to very low levels of toxic substances. 10. Food Safety Council. Proposed system for food Presented at an NIEHS conference on cancer risk safety assessment. Food and Cosmetic Toxicology estimation, Pinehurst, N. C., Mar. 10-12, 1976. En- (December), 1978. vironment tal Health Perspectives, 1977. 11. Gass. G. H.. D. Coats, and N. Graham. Car- 25. Pete, R. Epidemiology, multistage models, and cinogenic dose response curve to oral ciiethylstil- short-term mutagenicity y tests. In Origins o.f Human bestrol. j. Nat]. Cuncer Inst. 33:971-977, 1964. Cancer, Book C, Human Risk Assessment, H. H. 12. Guess, H. A. and K. S. Crump. Low-dose rate ex- Hiatt, J. D. Watson, and J. A. Winsten (eds. ) (Cold trapolation of data from animal carcinogenicity Springs Harbor, N. Y.: Cold Springs Harbor Lab- experiments—analysis of a new statistical tech- oratory, 1977 b), pp. 1397-1401. nique. Ma thematicul Biosciences 30:15-36, 1976. 26. Rai, K. and J. Van Ryzin. Risk assessment of toxic 13. Guess, H. A. and K. S. Crump. Maximum likelihood substances based on a generalized multihit dose estimation of dose-response functions subject to response model. Proceedings of SIMS Conference, absolutely monotonic cons t ra ints. Annals of Sta tis - Alta, Utah, June 26-30, 1978. tics 6:101-1 11, 1978. 27. Salsburg, D. S. Open query: Theoretical problems 14. Guess, H. A., K. S. Crump, and R. Pete. Uncertainty in the modified Mantel-Bryan procedure. Bio- estimates for low-dose rate extrapolations of ani- metrics 33:4 19-421, 1977. mal carcinogenicity data. Cancer Research 28, Watson, G. S. Age incidence curves for cancer. 37:3475-3483, 1977. Proc. Nat. Acad. Sci. USA 74:1341-1342, 1977. Appendix

Measuring Benefits and Costs

by Donald S. Epp

tolerance level

substance over a period of 3 years or longer will result in a 50-percent loss of function of the arms The benefits to be evaluated are defined as the and hands is a medical and biological science con- reduction in hazards to human health. This defini- clusion. Until a determination of the health conse- tion follows directly from the concern of the Food, quences is made, very little can be done toward Drug, and Cosmetic Act and the activities of FDA. comparing benefits and costs of regulation, Any effects on the health of animals or plants, ex- Health hazards from exposure to food contam- cept as they become food and thus affect the inants may be stated in various ways depending health of humans, is not considered in this report. on the health effects of the substance and on the This is not to say that economic analysis is incapa- state of knowledge about those effects. The haz- ble of considering other values associated with ard from a substance that increases mortality in animal and plant life. Rather, it is to limit discus- the exposed population could be stated as the sion to those aspects most germane to the decision number of premature deaths per year. If illness under consideration. requiring time-off from regular activity results The measurement of human health and the de- from a certain exposure level, the hazard to the termination of a relationship between a particu- population might be stated as the number of per- lar contaminant in food and subsequent human son-days lost to illness per year. When it is not health impairment must be made in biological, certain that exposure will cause a particular con- medical, and physical science terms. For exam- sequence, but that the likelihood is increased, ple, the conclusion that daily consumption of food probabilities may be attached to the outcomes. containing x parts per million (ppm) of a given The hazard from exposure would then be stated that some percentage of the exposed population would die prematurely or would lose a specified *Excerpt from OTA WoPi ing Paper entitled “Priority Setting of Toxic Substances for Guiding Monitoring Programs. A com- number of days from normal activity due to ill- plete copy of the paper can be obtained from the National ness, This method of stating hazard is encoun- Technical Information Service. (See app. J.) tered with cancer risks due to smoking tobacco.

166 Appendix E—Measuring Benefits and Costs ● 167

In many instances the substance in question of the appropriate medical service could suffice. . . may not have been studied for health effects or it Where the contaminated food is consumed in par- may have been developed so recently that long- ticular regions of the country, e.g., catfish, the term effects of chronic exposure are uncertain. It health effects would likely be concentrated in may, therefore, be impossible to calculate proba- those regions also and the appropriate treatment bilities, let alone specify the numbers of people af- costs should relate to the same regions. fected, This does not mean the matter is of no con- While the cash cost approach is relatively sim- cern. Based on the chemical structure of the sub- ple to calculate, it ignores some of the true costs stance, scientists may believe that the long-run to society from an untimely death or an illness. consequences may be very serious. In such cases One of the most obvious omissions is the failure to a precise probability cannot be attached to a spe- account for an individual’s contribution to eco- cific event occurring, It cannot even be said with nomic output, had sickness or early death not in- certainty that a health hazard exists. What can tervened. This weakness is corrected with the for- be said is that a serious health risk may be pres- gone-earnings approach. ent and the consequences of that risk are so grave that society may wish to avoid whatever risk may Forgone-Earnings Approach result. While this type of statement of the health hazard is more difficult to use in a benefit-cost Generally, the forgone-earnings approach uses analysis, it is not impossible. the discounted value of the future stream of earn- No matter how the hazard is stated, the first ings as the appropriate estimate for the cost of an task in economic analysis is to convert the hazard untimely death, and thus, the appropriate esti- statement into units of measurement that are com- mate of the value of a life saved from a premature parable with the statements of costs to be devel- death. While it is clear that if an individual had oped later. Many (although not all) of the costs as- not died or become ill in a particular year he sociated with those decisions are measured in would have continued to be productive for a num- monetary units—dollars. Likewise, at least some ber of years, it is not immediately obvious of the most important benefits to be achieved from whether the correct measure to use is gross earn- restricting exposure to health risks are also meas- ings or net earnings. ured in dollars. The most fruitful way to proceed, Those who support the use of gross earnings then, is to assemble the appropriate monetary argue that what is of interest to society is lost pro- measures and make conversion into monetary duction, and an individual’s contribution to that terms wherever possible for those effects not so production is measured by his or her earnings measured. stream. Thus, the loss to society is the discounted present value of the stream of annual earnings Cash Cost Approach weighted by the probability that the individual will be alive and well enough to earn that year’s One of the simplest ways to measure the cost of income. This approach is best suited to the anal- illness and at least part of the cost of premature ysis of illness rather than death and has been death is to add up the expenditures made for used to study the loss to society from mental ill- treatment or burial and other out-of-pocket costs, ness, syphilis, and illness of all kinds collectively, These costs can be obtained in a straightforward Those who prefer to use net earnings argue manner and summed for the number of individu- that, in the case of death, society may lose the pro- als that will be affected at the tolerance level duction of the decedent, but it also no longer has being evaluated, to supply goods for the decedent’s consumption. Even though the procedure does not involve Death releases resources to the rest of society for complex calculation, this approach does involve their use. Thus, the appropriate measure of loss decisions and judgments about the appropriate to society is the discounted present value to the costs to us. Medical service costs vary greatly be- stream of differences between an individual’s ex- tween procedures and between locations, It pected income in each year and that individual’s would be advisable, therefore, to obtain medical consumption during the same period. Each annual costs for the specific types of procedures re- net earnings would be weighted by the probability quired and for the part of the country where af- of the individual surviving to that year. With this fected individuals would likely reside, In some method of calculating, the cost to society of a cases where the contaminated food would likely death could be negative (or a benefit) if the indi- be consumed nationwide, a weighted average cost vidual consumes more than he or she produces.

5U-515 o - 79 - 12 168 ● Environmental Contaminants in Food

Retired and unemployed individuals are likely to advantage of being observable to the market- have such negative life valuations. place. If an individual agrees to undertake a haz- The possibility of a negative value for a life has ardous occupation which increases his probabil- led to several criticisms of the approach. Some ity of death in exchange for a given sum of money, individuals produce products that are not mar- that sum can be used as an estimate of his willing- keted, but are none-the-less, real and valuable, ness to pay for safety, Thaler and Rosen used The largest single example is the output of house- data that measures the relative riskiness of jobs wives. Other objections have been raised to the to estimate the tradeoff between wages and risk. conclusion that the unemployed and retired are of Thirty-seven broadly defined job classifications no value. It must be admitted, however, that these shown to be actuarially riskier than the average latter objections appear to confuse the economic occupation were matched against a cross-section- value of an individual with an ethical value of a al earnings survey. The results showed that in- human life, It is clear that the forgone-earnings dividuals in the risky occupations received in- approach examines only one portion of an individ- creased compensation of about $200 per year ual’s contribution to society—the contribution to (1967 dollars) for jobs where the risk of death was those things measured in the net national product 1 in 1,000 greater than the average. (NNP). If the goal of society is the maximization of This estimate may be conservative when ap- NNP, then assessing the costs or benefits of a plied to the general population. The occupations health-impacting project according to the for- surveyed in the Thaler and Rosen study are ap- gone-earnings approach can be rationalized. If so- proximately five times riskier than the average ciety holds other goals as well, then the use of U.S. occupation. People who take these jobs have forgone earnings is merely an expedient approx- different reservation prices for risk than individ- imation. uals who pick less risky jobs. The derived esti- mate of compensation required to offset greater Willingness-to-Pay Approach risk is, therefore, an extremely conservative ap- proximation of aggregate willingness to be com- The most telling argument against the previ- pensated. A further rationale for believing this es- ously discussed evaluation techniques is that they timate to be conservative is that the riskiness of are conceptually incorrect for use in benefit-cost these jobs can in some measure be affected by the analysis. They are directed to find the answers to individual employee. He or she is to some degree questions like “What is my life worth to other peo- in control of the personal risk level within the con- ple like my heirs and society in general?” The con- text of the risk level of the occupation. If the in- ceptually correct question to ask is “How much dividual is not able to affect his or her personal would I pay to avoid a small probability of my risk exposure, it is quite likely that they would death or illness?” Approaches that answer this need to be compensated to a greater extent to question are consistent with other measures of so- undertake the unwanted risk. cietal welfare since they estimate the aggregate consumer surplus involved in the reduction in The other technique for assessing willingness mortality or morbidity rates. to pay is the questionnaire method. A representa- What has come to be known as the “willing- tive sample of individuals are asked how much ness-t o-pa y’” approach has received extensive they would be willing to pay to achieve a specified treatment in recent economic literature. A review change in a particular condition. For the problem of much of the theoretical literature and the dis- of environmental contamination of food, the ques- cussion of some technical points related to a per- tion would most likely be directed toward the will- son’s willingness to pay for small changes in mor- ingness to pay for a reduction in the probability of tality rate may be found in Epp, et al. (ch. 4), That suffering specific health effects, such as speech review will not be repeated here, but rather, the impairment and nerve sensations from exposure remainder of this section will be addressed to re- to methylmercury. Questionnaire design ranges viewing techniques for determining an individ- from direct questions about dollar amounts for a ual’s willingness to pay. Two approaches have specific risk change to a series of questions about theoretical validity: compensation for risk-taking items related to the specific risk change of inter- and questionnaires of willingness to pay. est from which the analyst can infer an answer to Compensation for risk-taking as a technique for the “what would you pay?” question, estimating the loss in consumer surplus due to in- While the willingness-to-pay approach is ask- creased risk of death, illness, or injury, has the ing the conceptually correct question, it has seri- Appendix E—Measuring Benefits and Costs ● 169

ous difficulties in application, because the items In spite of the difficulties associated with em- of interest are going to be consumed in equal pirical measurement of willingness to pay, sever- amounts by everyone and no one can be excluded al studies have been done and have reported some from consuming the item if it is provided for any- success with various methods of presenting the one. For example, a reduction in the risk of lung question to the respondents. Randall, et al. (1974) disorders due to reduced sulfur dioxide levels in used pictures of the powerplant at Fruitland, N. the air will be available to everyone in the area Mex., at various levels of smoke emission to elicit regardless of their preferences or payment, If it is responses from residents and tourists as to the provided for one person, it is impossible to ex- amount they would pay (in the form of a sales tax clude others. The communal consumption aspect increase) to move from a less preferred view to of this large class of similar goods and services the respondent’s preferred view. The technique provides an incentive for people to misrepresent was used in a similar fashion to examine ques- their true preferences when asked, For instance, tions about a powerplant location in southern if a person believes he will be required to pay Utah (Brookshire, et al. ) and about reclamation of some amount in proportion to his answer yet there stripmined land in Kentucky (Randall, et al., is a high probability the good will be provided 1978). A modification of the technique was used whether or not he contributes and if it is provided, by Mann, et al. to determine willingness to pay for he will be able to enjoy it, then it is rational for small changes in human mortality. Their study him to respond that he is unwilling to pay any- found that people are able to comprehend risk thing for it, even if he knows he really would pay changes of the order of IO-6. Although the risk pre- some amount, if necessary. On the other hand, if miums stated in the Mann, et al. study were sub- an individual knows that someone else will pay stantially different from those’ found by Thaler for the good if it is found to be valued highly will and Rosen, the study indicated that individuals find it rational to grossly overstate its value to were able to answer the type of question posed, him. Thus, the incentive to be a free rider makes it and that with improvements in the survey instru- difficult to accurately measure the value to soci- ment, the willingness-to-pay questionnaire may be ety of an improvement in food safety. a workable approach. COSTS

The costs associated with establishing a partic- procedures can tell what proportion of various ular tolerance level are the loss of social welfare food products from any specified area will not resulting from the reduction in supply because meet the tolerance level. The task of the economic some otherwise useful food products can no analysis is to evaluate the impact of such a reduc- longer be sold for consumption. The reductions in tion in the supply of that food product. consumer and producer surplus, which are dis- When analyzing the effects of a tolerance level cussed in detail in this section, account for the that removes some food products from the market, economic value of resources, such as labor and there are two major alternative methods. These production capital, which may no longer be em- two are what have been called the alternative ployed in producing the food product being con- cost method and the opportunity cost method. sidered, It is possible, however, that some re- Either of these methods may employ models of the sources may not have any alternative employ- affected industry, but as will be shown in the fol- ment, In such case the tolerance level may cause lowing sections, the opportunity cost method re- these resources to be unemployed. While this lack quires more sophisticated models and delivers a of alternative employment is not strictly an eco- more comprehensive description of likely effects. nomic cost, it is clearly a consequence that a deci- sionmaker may wish to consider. Therefore, other Alternative Cost Approach effects (some might wish to call them costs) of the decision are examined in the following section en- With the alternative cost method, the analyst titled “Distributional Effects. ” examines the additional cost necessary to achieve The starting point for assessing the costs of a given objective using the next best alternative moving to a particular tolerance level is the phys- method to the one under study. Applying this to ical product that will be affected. The results of the PCB contamination of lake trout in Lake Supe- the monitoring program and various special test rior, for example, would require that the analyst 170 ● Environmental Contaminants in Food

examine the next best method for producing change in tolerance level, while curve Sl repre- 164,000 lbs of lake trout that did not exceed 2 sents supply under the higher cost next best alter- ppm of PCB to replace that amount from Lake Su- native method of production which does not vio- perior which does exceed 2 ppm of PCB. This ap- late the new tolerance level. The alternative cost proach requires that there be some alternative method examines the increase in cost of obtaining method available which will achieve the same the initial equilibrium quantity, Q,. This increase level of output in order that the costs of the two in cost is represented by the area ABP,PO. The op- alternative ways of producing the output can be portunity cost method, on the other hand, recog- compared. nizes that a new equilibrium price and quantity If there are several alternative methods avail- will emerge after the shift in supply. Quantity will able for producing a “replacement” amount of shift to QI and price to Pz. Comparing the new production, the alternative cost approach re- equilibrium with the old, the analyst using the op- quires not only that the cost of total replacement portunity cost method observes that the consumer with each method be examined, but that combina- surplus has been reduced by an amount equal to tions of methods be considered also. To properly area ACPZPO in figure E-1. This reduction in well- evaluate the alternative cost figure, the analyst being is clearly less than the one calculated with must know the costs of various amounts of prod- the alternative cost method. Because the opportu- uct from each alternative method. To continue nity cost method recognizes changes in produc- with the PCB contamination of lake trout example, tion and consumption, it is the preferred method. the analyst must consider not only the cost of The alternative cost method tends to overstate the 164,000 additional lbs of lake trout with less than cost of more restrictive tolerance levels. 2 ppm of PCB from each of the other Great Lakes (and any other large sources), but the costs of Figure E-1 .—Analysis of a Shift in Supply various increments of production, such as 10,000 lbs. from each source. The least cost combination of increments would be found by starting with the least costly way of obtaining 10,000 additional lbs and adding to that the next least costly way of get- ting 10,000 more lbs until a total of 164,000 lbs are accounted for.

Opportunity COSt Approach

With the opportunity cost method the analyst examines the additional cost necessary to achieve a new market equilibrium amount of a food prod- uct which reflects the likely higher price and, therefore, lower consumption of consumers. Since restricting the supply of a food product from pre- vious sources means a shift to higher cost alter- I I native sources, we can use our knowledge of mar- o! i 1 kets and people’s preferences for the food prod- Q1 Qo uct to estimate their adjustment to the restriction, This gives a different estimate of costs due to re- The above diagrammatic analysis presents the striction than does the alternative cost method essential features of the budgeting approach to which assumes no change in the amount of the the opportunity cost method. This is discussed in food product consumed. more detail below. Another approach, modeling, A simplified partial analysis of the difference carries the above analysis further to show between the alternative cost method and the op- changes in the market for factors of production as portunity cost method is presented with the use of well as other product markets, While this method figure E-1. The decision to establish a tolerance cannot employ graphic analysis due to the com- level such that some portion of the product cur- plexity of relationships, the mathematical model- rently offered for sale may no longer be sold is ing techniques described below permit a much shown by the shift in the supply curve. Curve So more comprehensive analysis of likely effects of a represents the industry supply curve prior to the change in tolerance level. Appendix E—Measuring Benefits and Costs ● 171

As was noted above, the opportunity cost ap- Comparison of Approaches proach measures effects of a change in tolerance level as the change in consumer surplus in the The alternative cost and opportunity cost ap- economy. There is abundant literature discussing proaches are not equally adapted to handling all the problems of estimating consumer surplus; problems—each has several advantages and dis- some authors even question the value of the con- advantages. The alternative cost approach has cept in many empirical contexts. For this exer- the advantage of using data that is more easily ob- cise, however, it seems appropriate that consum- tained and not requiring extensive development er surplus be used. The ultimate consumer of the of mathematical models prior to the analysis. contaminated food products should provide the Since the alternative cost approach compares the basis for evaluation. This becomes particularly status quo with the next best alternative, subject crucial where ramifications of the production to the change in tolerance level, the only addi- shift may include a variety of products, not just tional data needed beyond present production the contaminated product. Consumer surplus be- techniques is an estimate of how the same amount comes the common denominator allowing compar- of product could be produced under the best al- ison among a variety of production effects. In ternative available with the new tolerance level. practice, the opportunity cost approach may em- This data can frequently be obtained from ex- ploy either of two methods: budgeting or model- perts in the production of the food under con- ing. sideration. The major disadvantage of the alternative cost Budgeting approach is that it is conceptually erroneous. Our With budgeting the analyst acknowledges that knowledge of economic adjustments to changing production shifts are likely for the contaminated production conditions recognizes that adjust- product. This technique uses data on inputs used ments will be made in the enterprise combina- in the production of a product to calculate the tions and in factor combinations for producing a costs of producing a particular amount of that particular food product. If a significant portion of product, assuming that other commodities will be the total production is removed from use or a spe- produced in the same amounts and at the same cific location is no longer capable of producing a prices as with the status quo. The budgeting ap- safe food product, the ensuing production adjust- proach is a more limited and restricted approach ments will not be limited to factors of production, than the modeling one, but for some circum- but may also include changing the foods produced stances may prove advantageous. or the location of production. The amount con- sumed of the food for which a new tolerance level Modeling is being established will probably adjust to The second and more complete method of ap- changes in the cost of obtaining the food in a safe plying the opportunity cost approach is through form (within tolerance levels). The alternative the use of production models. With this method, cost approach ignores all of this knowledge by as- mathematical models of the relevant portion of suming that the food in question, using the next the economy are employed to trace the shifts in best alternative, will be produced at the same supply curves and the changes in the amount and level as before. price of various commodities and factors which The conceptual error leads to a second error— result from the restriction on the use of a contami- the overstatement of the costs associated with the nated product. If these models are specified to in- tolerance level, Because this approach ignores clude geographic areas and the various alterna- the adjustments in the amount of a food product tive production activities which take place in produced, it leads the analyst to a cost figure that each of these areas, they are able to project is greater than the one that would actually result. changes in the location of production of particu- As production costs increase it is likely that the lar crops and changes in the use of various pro- quantity produced will decrease. Thus, the alter- duction factors in each region of the country. This native cost approach has the analyst multiply a more realistically describes the likely reactions to higher cost per unit by more units than would ac- the change in tolerance level and permits the cal- tually be produced. This leads to an erroneous culation of a more accurate estimate of the calculation of the change in consumer surplus change in consumer welfare resulting from the and to the overstatement of the cost of setting action level decision. tolerance at a particular level. 172 ● Environmental Contaminants in Food

The major advantage of the opportunity cost clude new developments in factor prices, product approach is conceptual: it allows the analyst to prices, and production technology. Thus, the mod- see the adjustments that the economy is likely to els are expensive to maintain. They are also ex- make in response to changing production costs as pensive to create in the first place. Most models a new tolerance level removes some of the prod- require a great deal of prior research on the tech- uct from the market. Where agricultural products nical relationships in production and specifica- are involved, the use of an econometric model of tion of the factors related to consumption. While the agricultural sector facilitates the analysis of much of this work has been done for agricultural changing comparative advantage and the result- commodities, it is recognized that the material is ing production pattern, This gives a more accu- frequently inexact and often the models are out of rate indication of the ultimate cost of making the date, It would be necessary, therefore, to under- tolerance level decision, take some rather expensive research in order to The use of models can also facilitate distribu- incorporate the opportunity cost approach for a tional analysis if those models have spatial (re- commodity that did not already have substantial gional) variables introduced into them. Through prior work. This is likely to be the case for sea- the use of regional supply and demand models, foods, freshwater fish, and certain agricultural the analyst is able to estimate not only the market products produced in a few local areas. equilibrium supply and demand adjustments, but Obviously, the detailed knowledge of produc- also the regional production adjustments in re- tion relationships and therefore the expense is re- sponse to the overall market changes. It is thus duced if one uses the budgeting method rather possible to note changes in the regional location than the modeling method in the opportunity cost of production and to estimate the impact of these approach, Budgets usually involve a partial anal- changes on various income and social groups that ysis of the adjustments and therefore do not re- are distributed differently in the various regions quire the development of production relationships of the country. Thus, the models that are used for for commodities not closely related to the com- the opportunity cost approach can facilitate a modity under consideration, Even so, the budget- more detailed and sophisticated analysis of the ef- ing approach requires more information than the fects of a change in the tolerance level. The rea- alternative cost approach because of the consid- sons for examining distributional effects as well eration of production changes. It is unlikely that as a brief description of some ways these may be the budgeting approach can be used for a region- done are included in the next section of this alization analysis that involves more than a very paper. few regions. Thus, the lesser cost produces less The major disadvantage of the opportunity cost information. approach is that it requires a very large amount Since the two approaches, alternative cost and of data. Production costs, outputs, factor require- opportunity cost, differ in the amount of data that ments, factor prices, and a variety of other bits of they require and in the cost of acquiring and proc- information are required for each alternative essing the data, it is necessary for an analyst to food production activity in each region including choose between these two methods. Several alternative ways of producing the food under con- points should be considered when making a selec- sideration. Even if the supply equations and activ- tion of the most appropriate analytical method. ities are limited to those most likely to enter into Demand Elasticity the solution, the requirements are formidable for most agricultural and fishery commodities. The If the product for which the tolerance level de- demand side also requires regional consideration cision is being made has a very inelastic demand, ’ with specification of the demand for each of the it may be appropriate to use the alternative cost commodities included in the supply side of the approach. With a very inelastic demand, it is like- model. Careful attention must be given to the in- ly that the assumption of producing the same clusion of complements and substitutes so that the quantity regardless of cost is a reasonably close - model will give a reasonable approximation of ac- approximation to reality, Thus, the disadvantages tual market adjustments. enumerated above for the alternative cost ap- A second disadvantage of the opportunity cost proach are of less significance for many agricul- approach is that the models developed for analyz- ing production and market shifts are usually short-run and static, This means that these models must be revised periodically in order to in- Appendix E—Measuring Benefits and Costs . 173 tural products, such as certain vegetables, sugar, are fairly well-developed for most of the impor- and cereals for human consumption, as well as tant alternative commodities as well as the com- for fish. modity under consideration, it may be possible to modify the existing work relatively inexpensively Information Availability to obtain the information needed for the tolerance Since the opportunity cost approach requires a level decision, For example, a great deal of work great deal of information about production rela- has been done with the feed grain-food grain sec- tionships, not only for the commodity under con- tors of American agriculture. Less work has been sideration but for alternative commodities, it is done on the livestock sectors, although there are helpful to have these relationships previously de- some models that incorporate feed grains and veloped. Much of this information is so time con- livestock and some models include the dairy sub- suming to develop that it would be virtually impos- sector. If the decision to be made involves food or sible to create a research program that could give food grains, serious consideration should be given results in time to be useful for any action level to using the opportunity cost approach and some decision if a great deal of groundwork had not of the models that have been developed, with been laid prior to that research effort. If no in- whatever modification seems appropriate. On the formation is available concerning production of other hand, there has been very little modeling the product in question or the alternative product that includes specialty crops, such as specific that might be produced or consumed, it is likely fruits and vegetables, into a general agricultural that the alternative cost approach would be model. It would be expensive and of dubious value adopted. This approach requires less data and to develop a programing model for a tolerance the material that it requires probably can be pro- level decision involving those crops. duced in time to be useful in making a tolerance level decision. . It is readily apparent that the key assumption of the alternative cost approach—no change in Complexity of Interrelationships the quantity of product produced after restricting If a product is produced in many areas with a the amount of a particular contaminant that may large variety of production alternatives, both with be present—is not realistic in most cases and can regard to factors of production and to alternative lead to substantial error in estimating the social products, one needs to use an opportunity cost ap- costs of a decision. It is the opinion of the author proach. In such a situation the interrelationships that the alternative cost approach is appropriate are so complex that it is difficult to judge from only in those cases where an alternative source of observation of the data the likely combination the product is available at virtually the same that will result from a tolerance level decision. effective price or where alternative production Under these conditions it is very desirable to use techniques are available at no increase in per the opportunity cost approach if at all possible. unit cost of production and which would involve not more than negligible shifts in the location of Availability y of Mathematical Models production. These are very restrictive conditions. The stage of development of the mathematical For most tolerance level decisions involving major programing models for a particular product or agricultural products these conditions do not hold sector of the economy is an important considera- and the opportunity cost approach is strongly ad- tion in choosing a method. If programing models vised.

DISTRIBUTION EFFECTS

The previous two sections have outlined consid- jor interest here is the analysis of how a decision erations and methods for determining the benefits impacts on various groups within society. Even and costs of a change in tolerance level for a par- though a decision may produce an improvement in ticular contaminant in a specific food product. welfare for the whole of society, there may be From an economic perspective and a society-as-a- particular groups for which welfare is reduced. whole viewpoint, these sections cover the econom- Such distribution effects may be due to the bene- ic analysis of welfare changes. There are, how- fits and costs being shared differently by dif- ever, related areas of concern which may be cru- ferent income groups or by different regions of cial to the social acceptability of a decision and the country, There may also be concern about for which economic analysis can be useful. Of ma- how the consequences of the decision affect the 174 ● Environmental Contaminants in Food

environment in various locations or its effect of The interregional differences in the impact of esthetic considerations. Each of these points is re- benefits and costs may also be of interest to the viewed briefly in this section. decisionmaker. Frequently the alternative source The analysis of benefits and costs may show of product to replace that which is contaminated greater addition to benefits from a particular tol- above the new tolerance level will be in a differ- erance level change than the addition to costs; ent area. Where a particular stream or valley thus, a desirable decision, The consequences of produces contaminated food, the production may that decision may not, however, be desirable for shift only a few miles to a nearby river or valley. all affected parties, It is, therefore, important in Then the impact on the economy of a particular some cases to examine the distribution of benefits community may be slight. In other cases, a wide- and costs among social and economic groups. spread area, such as a large estuary or part of a For example, the decision to set a very restric- State, may be contaminated and production will tive tolerance level on a particular contaminant shift to a completely different part of the country, in fish may mean that fishing in certain bodies of Interregional differences may arise from prod- water is no longer possible. That may have a very ucts which are produced in one area, but con- slight effect on consumers, since they can easily sumed in another, For example, most of the winter switch to fish from other areas. The fishermen head lettuce eaten in the United States is grown in who lose their employment may, however, have the Imperial Valley of California, yet only a minus- much lower incomes than the average consumer cule portion of the crop is consumed there. If a tol- and may have few or no alternative sources of em- erance level for some contaminant made it impos- ployment. In this case the benefits in health sible to sell Imperial Valley lettuce, almost all of hazard avoided by a large number of people may the cost would be borne by that small part of Cali- greatly exceed the costs of idled resources in fish- fornia while the health hazard reduction would be ing and thus, the decision is appropriate. It is well shared throughout the rest of the country. for a decisionmaker to realize, however, that the The data for constructing the regional impacts major portion of the costs will be borne by people of a tolerance level decision can be found most of much below average incomes with few alterna- easily in mathematical programing analyses un- tives to mitigate the losses. der the opportunity cost approach. Not all pro- Another decision may involve restriction of graming models have regional specifications, but contaminant levels in a food eaten primarily by a most agricultural models do, Careful review of specific ethnic or racial minority whose members budgeting results by experts on the production have low incomes, The producers of this food may have several alternative products they can pro- and marketing of the foods in question can also re- veal some of the regional impacts or can call at- duce and the consumers may also find substitutes tention to the likely importance of regional differ- readily available. In this case the costs are slight and a low-income minority receives substantial ences. Further regional analysis can be per- formed if desired. The alternative cost approach health benefits. does not provide much data on regional impacts. Most of the data needed to describe the distri- Here again, however, a careful review of the re- bution of effects among income groups is avail- sults by a knowledgeable analyst can suggest the able from the analysis of benefits and costs. In likelihood of regional shifts that warrant further some cases it is not clear which income or social analysis with finer detail on regional specifica- group is affected, but the analyst can usually de- tion. termine this at the time he collects the data on ef- fects. Thus, it is important that benefit-cost ana- The fact that a particular decision will severely lysts be alert for information about the incidence harm a particular region or social group does not of effects as they collect their data. What may be necessarily imply that the decision should be relatively easy to determine while the original avoided. If the society-wide analysis of benefits source is being contacted may be time-consuming and costs suggests a net gain from the decision, and difficult to find out at a later point in the there is good reason to go ahead. The analysis of analysis. The presentation of distributive impacts the distribution of effects may serve to point out would probably be in narrative form rather than the need for special programs to assist those a quantitative analysis. The important point is to harmed adjust to the decision. For example, re- include this additional information for the deci- gional impact analysis may point out the need for sionmaker so that the fullest possible knowledge a retraining program for unemployed workers in of impacts is available when the decision is made. a particular region far enough in advance that the Appendix E—Measuring Benefits and Costs ● 175

program can be ready to go when the tolerance production to shift to an area where environmen- level decision takes effect. tal damage would increase. For example, prohibiting the sale of milk from The analysis of impacts on various regions and one area may cause production to concentrate in social groups can provide the starting point for another, leading to increased wastes from dairy analysis of the sociological and psychological im- farm holding pens and milk-processing plants. pacts of the tolerance level decision. In some, per- What may have been an acceptable load on the haps many, cases the effect on a particular com- environment previously may become excessive. munity or group of people may have serious con- Analysts should be alert to the possible environ- sequences for social structure, organizational mental and esthetic impacts of decisions. In some stability, and even the mental health of the af- cases these effects can be measured in monetary fected people. Such a possibility was pointed out terms and compared with benefits and costs. In and briefly discussed by O’Mara and Reynolds other cases, the effects must remain in biological (PP. 61-66) in their analysis of restricting fishing terms, but described in such a way as to make in the James River of Virginia due to kepone con- them useful in a tolerance level decision. The use tamination. Since the present report is directed of environmental food chain models and impact toward economic analysis, the techniques of soci- matrices has been suggested by the author with ological analysis will not be discussed here. Some respect to pesticide decisions (Epp, et al., chs. 5 of the analyses likely to be used are briefly out- and 6) and the interested reader is referred to lined in Epp, et al. (pp. 134-138). that discussion for greater detail, Not all toler- Another area of potential concern is the impact ance level decisions will involve a significant en- of a tolerance level decision on the environment vironmental effect, but the procedure of analysis or on the esthetic properties of a region. It is pos- should include a notation to check for such effects sible that a tolerance level decision may cause and an analytical protocol for use when needed.

ALTERNATIVE TOLERANCE LEVELS

The sections on benefits, costs, and distribution that would produce even greater net benefits. The effects have described how an analysis might be proper use of economic methods requires that conducted of a particular decision to change the consideration be given to the likely changes in tolerance level from one point to another. For ex- benefits and costs if the tolerance level were ample, one decision might move from no restric- changed a little either way. If it seems possible tion to prohibiting the sale of a particular food that net benefits would increase significantly by with 5 ppm or more of a specific contaminant. If moving to a different tolerance level, then a de- benefits exceed costs and distributional as well tailed analysis as described above should be as other effects are not overwhelming, then the made of the change from the originally proposed move to a 5-ppm tolerance is efficient. That does tolerance level (5 ppm in the example) to the new not say that 5 ppm is the most efficient tolerance level (2 ppm). level. There may be some other level, say 2 ppm, 176 ● Environmental Contaminants in Food

LITERATURE CITED IN APPENDIX E

Brookshire, D. S,, B. C. Ives, and W. D, Schulze, 1976, McNally, J. M., 1978b, “Polybrominated Biphenyls: En- “The Valuation of Aesthetic Preferences,”’ }ournul vironmental Contamina t ion of Food, ” Congressional of Environment ta] Economics and A4anagement Research Service, Washington, D. C., OTA Working 3:325-346. Paper. Epp, D. J., et al., 1977, Identification una’ Specification 0’Mara, G. K,, and R. R. Reynolds, 1976, Evaluation of of Inputs for Benefit-Cost A40deiing of Pesticide Use, Economic and Social Conseffuences of Restricting EPA-600/5-77-01 2, U. S. Environmental Protection Fishing Due to Kepone Pol]ution in the ~arnes River, Agency. Washington, D,C. Virginiu, U. S. Environmental Protection Agency, George, P. S., and G. A. King, 1971, “Consumer Demand Washington, D,C. for Food Commodities in the United States with Pro- Randall, A. B., et al,, 1974, “Benefits of Abating Aes- jections for 1980, ” Giannini Foundation Monograph thetic Environmental Damage from the Four Cor- Number 26, California Agricultural Experiment ners Power Plant. ” New Nlexico State University Station, Berkeley. Agricultural Experiment Station Bulletin, Las I,eW,is, p F., 1977. “Economic Impact Assessment for Cruces, N. Mex. Proposed Reduction of Temporary Tolerances for Randall, A., et al., 1978, “Reclaiming Coal Surface Polychlorinated 13iphenyls in F’ood. ”” Unpublished Mines in Central Appalachia: A Case Study of Bene- manuscript, U.S. Food and Drug Administration, fits and Costs,”’ L(~nd Economics 54:472-489. Washington, D.C, Reichelderfer, K. H., et al., 1979, “’Pesticide Policy Eval- hlann. S. H., et al., 1976, “The Determination of the uation i?lo(iels, ” NRED Working Paper (unnum- Economic Effects of Pesticide Use on Human Health: bered draft), Natural Resource Economics Division, A Gaming Approach to Willingness-to-Pay, ” VVork- Economics, Statistics, and Cooperatives Service, ing Paper Series No. 32. ‘rhe Center for the Study of U.S. Department of Agriculture, VVashington, D.C. Environmental Policy. The Pennsylvania State Uni- Thaler, R.. and S, Rosen, 1973, The V(llue of Saving a versity, University Park, Pa. Life: Evidence from the I.(lhor Market. National Bu- LlcNally, J. M.. 1978a, “Poly(:h(llorirl[~ted” Biphenyls: rea u of Economic Rcsca rch, Con ference on Income Environmental Contamination of k’oad, ” Congres- and Weal t h. sion:i 1 Research Service, Washington, D. C., OTA ‘1’racer Jitco, Inc., 1978, “Risk Assessment of Polybro- Wrorking Pi]per. mina ted 13iphenyls, ‘“ 0’1’A Ll}orking Paper, Appendix F

Priority Setting of Toxic Substances for Guiding Monitoring Programs *

by Clement Associates, Inc.

INTRODUCTORY REMARKS ON THE ESTABLISHMENT OF PRIORITIES

There are many purposes for which the Federal problems in numerical order. Good systems for es- Government must set priorities among toxic chem- tablishing priorities will inevitably involve a sub- icals in the environment. The universe of environ- stantial element of sound scientific judgment in mental contaminants is very large, and resources weighing and interpreting incomplete and unsys- for research, testing, monitoring, and regulation tematic data. The second task in designing an effi- are limited. Some statutes explicitly require Gov- cient priority-setting system is thus to make a rea- ernment agencies to set priorities for testing or sonable compromise between the use of objective regulation, and in many other cases agencies may criteria and scientific judgment. be called on to explain their decisions to tackle This report is concerned specifically with the some problems before others. Accordingly it is establishment of priorities for monitoring toxic very desirable to develop rational systems for set- chemicals in food. There are two general methods ting priorities. by which this might be done: 1) directly, by sam- Establishment of priorities involves preparing a pling the food supply for chemical contaminants list of problems, ordered according to their per- and ranking them according to potential hazard; ceived *’importance” or urgency, before attempt- and 2) indirectly, by surveying the universe of in- ing to solve them. It is by definition a preliminary dustrial chemicals and ranking them according to step and should be designed to maximize the effi- their potential for entering the food supply in tox- ciency of the problem-solving process, Although it ic amounts. The first method has obvious advan- is desirable to use as much relevant information tages and in fact has been pursued by FDA. The as possible in setting priorities, it is also possible method is, however, limited by the available ana- to use too much information. If too much time is lytical methodology: specifically, most of the spent in overelaborate priority-setting exercises chemicals recognized to date as food contami- the gain in efficiency achieved by tackling the nants are those relatively easily detected by gas most urgent problems first is offset by a delay in chromatography or atomic absorption spectrom- taking any action at all. The first task in designing etry. The principal question addressed in this an efficient priority-setting system is thus to make report, therefore, is whether the second method a reasonable compromise between the effort ex- can complement the first by identifying potential- pended on this preliminary step and the gain in ef- ly significant contaminants that have not yet been ficiency achieved by doing it well. detected in foods. Although this report indicates Priority-setting exercises are usually limited that it can do so, it should be recognized that the primarily by the unavailability of data. (If exten- second method also is limited to some extent by sive data are available, then it is usually obvious available analytical methodology. The factors we which problems should be tackled first. ) For this can use to identify the potential of a chemical for reason there is usually little to be gained by devel- entering the food supply are selected on the basis oping elaborate multifactorial schemes to rank of a knowledge of the properties and environmen- tal behavior of other chemicals already known to do so. This in turn is based on our knowledge of *Excerpt from OTA Working Paper entitled “Priority Setting the extent of contamination and hence on our ana- of Toxic Substances for Guiding Monitoring Programs. ” A com- plete copy of the paper can be obtained from the National lytical capabilities. Thus, there is an inherent “rechnical Information Service. (See app. j.) bias towards identifying as potential food con-

177 178 ● Environmental Contaminants in Food

taminants those chemicals that are similar to in all systems for setting priorities: chemicals on chemicals already identified in food. This bias which there is no information at all will automati- can only be offset by the use of good scientific cally be given low priority, unless some room is judgment. This point illustrates a general problem left for largely intuitive judgments,

CRITERIA AND METHODS USED IN PREVIOUS EFFORTS TO ESTABLISH PRIORITIES AMONG TOXIC CHEMICALS IN THE ENVIRONMENT

Survey of Toxic Chemical recommendations of experts. The likelihood of Priority Lists errors of omission when this method is used will depend on the extent of knowledge and experi- Thirty-two priority lists of toxic chemicals com- ence of the experts. These methods can be com- piled during the last 5 years have been identified. bined by supplementing source lists with chemi- Most of these lists (28 out of 32) were prepared cals of concern to experts. for the Federal Government. For each list, a de- Reducing the size of the comprehensive list. It scription sheet indicating the purpose of the list is almost always necessary to reduce the size of and the criteria used in selecting chemicals was the comprehensive list before ranking. Since the prepared. In addition, the name of the list, the resources needed for compiling and evaluating Federal agency or other institution for which the the necessary data for setting priorities are usu- list was compiled, the date of completion, the con- ally limited, a “narrowing” or “truncating” step tract or document number, type of substances, is performed. Errors are very likely to occur at the number of chemical substances, the means by this step because, for practical reasons, decisions which the chemical was identified (for example, at this stage are usually based on incomplete in- CAS number or chemical name), whether the list formation, In a multistage priority selection is machine readable, and relevant comments scheme, such as the one by which the TSCA Inter- were included. agency Testing Committee selected chemicals to recommend to the EPA Administrator for testing, Methods additional information is obtained at each stage of reduction. In that effort, an initial listing of Chemical Selection 3,650 substances was reduced to a master file of Compiling a comprehensive list. The first step 1,700 substances, then to a preliminary list of 330 in priority-setting efforts is usually compilation of substances from which approximately 100 chemi- a comprehensive list defining the universe of cals were selected for dossier preparation. Twen- chemicals under consideration. Since chemical ty-one chemicals or groups of chemicals from that compounds and mixtures are often referred to by list have been designated for recommendation by several different names, each chemical compound the committee. In this type of process, the earliest on the comprehensive list should be uniquely iden- stages may be the most subjective and the most tified. This can be done by specifying for each likely to result in errors of omission. chemical on the list the IUPAC chemical name, Grouping. Individual chemicals are often com- synonyms, molecular structure, and/or the Chem- bined in groups or classes. This is usually done in ical Abstracts Service (CAS) registry number. The an attempt to reduce a list of chemicals to a more use of CAS numbers facilitates computerization. manageable set, There are, however, other rea- Most often, the source of chemicals for this sons for grouping. Some monitoring or analytical comprehensive list is previously existing lists. techniques do not distinguish between closely re- Errors of omission are likely with this method. A lated chemicals. Some classes of chemicals, such chemical not considered previously will not be as fluorocarbons, may be considered as a group considered by someone using existing lists as a for regulatory purposes. Such chemicals as PCBs source, The more comprehensive the source lists, are not manufactured or used as individual chem- the less likely chemicals are to be omitted from ical compounds and are, therefore, nearly always consideration. An alternative method for select- considered as a group. Groups can be based on ing chemicals for consideration is reliance on chemical and physical properties, on chemical Appendix F—Priority Setting of TOXIC Substances for Guiding Monitoring Programs ● 179

structure, on uses (e.g., flame retardants, fluores- Determining Factors To Be cent brighteners), or on some combination of Considered in Setting Priorities these factors. Sometimes a single compound or The purpose for which the list is to be used several compounds are selected as representative must be clarified. Once the problem the priority of the group. list is intended to solve is understood, factors rele- An advantage of grouping chemicals with simi- vant to that problem can be determined. Criteria lar structure or functional groups is that it some- based on those factors can then be formulated for times allows chemical, physical, and biological assigning chemicals a place on the list. General properties of individual compounds to be pre- criteria must often be defined in terms of specific dicted. Members of a chemical class may behave parameters for which information is available. in a qualitatively similar fashion. In addition, Those parameters for which quantitative data some properties increase or decrease systemati- are readily available are the easiest to use for cally along homologous series. Attempts to derive scoring and ranking. On the other hand, data are structure-activity correlations are much more frequently lacking on the factors of greatest likely to be successful for chemical and physical relevance to the specific problem. In such cases properties than for biological properties. Because factors of less relevance may have to be selected. of some unique properties of biological systems, The choice of factors to be used in setting prior- such as enzyme specificity, it is possible for close- ities often involves a compromise between the fac- ly related chemical compounds, even such stereo- tors that are most relevant and the factors on isomers as aldrin, dieldrin, and endrin, to have which sufficient information is available. very different biological properties. Some weaknesses associated with grouping Assembling Data on Each Chemical are: Sound decisions on assignment of priorities ● Members of the group may vary widely with require reliable data on each chemical to be respect to the factor being scored, making it ranked. The available information may, however, difficult or meaningless to assign a single be massive, contradictory, unverifiable, scanty, score to the group. or absent. It is usually difficult or impossible to ● No single compound can truly represent an obtain existing information that has been classi- entire class. fied as proprietary or confidential. ● It is often difficult to assign compounds with When information is needed on a large number several different functional groups to a of chemicals and only a limited amount of time or single chemical class. This difficulty can money is available for assembling data, second- sometimes be overcome by computerized ary sources are usually relied on. Certain types of substructure search systems that allow all information (e. g., annual production volume, LD~l~, chemical compounds with a specific func- octanol-water partition coefficients) are often tional group to be retrieved regardless of the tabulated in secondary sources. Other types of in- presence or absence of other functional formation (e.g., fate and transport in the environ- groups. ment, metabolic pathways) may be obtainable At one stage of the National Science Founda- only from primary reports of research studies. tion (NSF) Workshop Panel, multistage effort These individual studies require considerable grouping was done in an attempt to reduce the list time and effort to locate, obtain, and review. In a of chemicals to a more manageable set. Fourteen multistage screening process, primary sources chemical classes were defined, and a representa- may be used only in the last stage, when the num- tive chemical was identified for each. Weak- ber of chemicals has been reduced to a small por- nesses the panel found to be associated with this tion of the comprehensive initial list. In a very lim- procedure included: ited effort, primary sources may not be used at ● The classes were not completely valid either all. chemically or functionally. Assigning Priorities ● The physical and biological properties of compounds in a given class varied widely. Two general methods used for assigning prior- ● No single compound was truly representa- ities to toxic chemicals are the ranking of individ- tive of a class. ual chemicals [or classes) and the assigning of in- The class concept was abandoned in the next dividual chemicals (or classes) to priority catego- stage of the effort and selected members were ries. Only 6 of the 32 lists surveyed are actually considered as individual chemicals. ranked. In most of the other lists, chemicals were 180 ● Environmental Contaminants in Food

either assigned to a single “high priority” cate- quantitative information is available. For exam- gory or distributed among several priority level ple, the octanol-water partition coefficient can be categories. used as an index of bioaccumulation, but a more Some numerical index or score is nearly always sophisticated scoring system would consider such used for ranking and often used for assigning other factors as chemical reactivity, metabolic chemicals to priority level categories. When fate, and vapor pressure of the compound. chemicals are to be ranked solely by parameters The human health hazard posed by a chemical for which comparable quantitative information is substance is a function of exposure to the sub- available (e.g., oral LDfO in the rat or annual pro- stance and the inherent toxicity of the substance duction volume) these data can be used directly. itself. Scores for various exposure and toxicity When the criteria for ranking cannot be defined factors are usually combined for hazard-ranking in terms of parameters for which comparable purposes, No consensus has yet been developed, quantitative information is available for all chem- however, as to the optimum algorithm for hazard icals under consideration, chemicals are usually ranking in this way. assigned scores. Scoring is also usually used Linear weighting schemes are efficient and when quantitative and nonquantitative factors easy to use, and they have produced reasonable are jointly considered for ranking. results. These schemes rely on the assignment of The scoring process involves assigning numeri- scores to a number of factors. Weights are as- cal values to one or more parameters to produce signed to each factor, and the weighted factor an index or series of indices for each chemical. In- scores are summed to produce an overall score dices can then be weighted and combined to pro- for the chemical. The choice of weight is usually a duce a single score for each chemical, subjective judgment, involving consideration of Scoring can be totally subjective, totally objec- both the importance of each factor and the relia- tive, or somewhere in between, Subjective scoring bility of the information used in assigning the involves the use of experts to assign priorities. score. The experts may be provided with information on Multiplication of individual factor scores pro- each chemical or may simply be asked for their vides a wider spread of overall scores, Multiplica- opinion. The credibility of subjective scoring will tion is frequently used with actual unscored quan- depend on the perceived expertness of panel titative data. This is especially likely where some members. Delphi techniques can be used to im- of the data are in the form of fractions. For exam- prove the validity or rankings, Scoring systems ple, fraction of production lost would be multi- based on the Delphi method rely on the develop- plied by total production or fraction of dose ab- ment of consensus among experts acting inde- sorbed would be multiplied by total administered pendently or as part of a panel. In the NSF Work- dose. Multiplication cannot be used unless some shop Panel effort to select organic compounds correction is applied to allow for data gaps. In the hazardous to the environment, panel members development of one list, an algorithm for toxicity identified chemicals of concern on the basis of was applied which compensated for types of tox- their own experience and refined their judgments icities for which data were unavailable. The by considering available data for production, im- scores actually assigned to a chemical for various ports, uses, disposal, and toxicity. Through suc- categories of toxicity were summed [T(total)]. The cessive examinations by panel members, individu- maximum toxicity possible for each of these cate- ally and together, in progressively greater detail, gories for which data were available were also a priority list was developed. A weakness of such summed [T(possible)]. The ratio of these sums subjective scoring methods is their probable lack [T(total) ÷ T(possible)] was used as the toxicity of reproducibility, which exists because different factor in the hazard-ranking algorithm. The haz- experts would be likely to reach different conclu- ard-ranking algorithm used for ranking manufac- sions, (However, we know of no investigation of tured organic air pollutants was: the reproducibility of such results in which differ- ent groups of experts were asked to assign scores to the same chemical s.) Where scores can be assigned to specific val- ues of parameters for which quantitative informa- tion is available, objective scoring may be prefer- able. Often, however, modifying factors may make the use of some degree of subjective judgment de- sirable in the assignment of scores even where Append/x F— Priority Setting of TOXIC Substances for Guiding Monitoring Programs ● 181

In some cases the scores assigned to certain number of chemicals. Production data for chem- factors are logarithms of data or are selected sub- icals produced by less than three manufacturers jectively in such a way as to approximate loga- are considered proprietary. In the near future, rithms. In such cases, addition of the scores the EPA Toxic Inventory List, being prepared achieves the same result as multiplication of the under section 8 of the Toxic Substance Control actual data. Weighting of the factors before addi- Act, may be a convenient source of production in- tion then corresponds to changing the base of the formation, but at this time it is not clear how much logarithm. of this information will be released to the public. In addition to annual production volume, data Criteria on production capacity, transport volumes, im- ports, exports, net sales, or levels of consumption The two general criteria used for assessments may sometimes be available. A major problem of health hazard are exposure and toxicity. Previ- with using any of those figures as indices of en- ous efforts to set priorities among toxic chemicals vironmental exposure is that these figures are in the environment have been reviewed, and spe- based on commercial production and sales. The cific criteria used to assess the extent of exposure amounts of a chemical used as a captive inter- to these chemicals or the toxicity of these chemi- mediate are not included although such a chemi- cals have been considered. cal may be an occupational health hazard or a neighborhood pollutant in the vicinity of the plant. Exposure Very toxic and highly reactive chemicals are The three aspects of exposure that are of great- often manufactured at the same plant in which est importance in hazard ranking are the number they are used. Manufacturing byproducts may of people exposed, the frequency of exposure, and never enter commerce but because they are the quantities of the chemical to which they are waste products they are very likely to be released exposed. Where direct information is available on to the environment in plant emissions and in ef- these exposure factors, e.g., from monitoring of fluents. Toxic chemicals formed in the environ- the ambient environment, of food, or of human tis- ment (byproducts of microbial degradation or sue, it is obviously desirable to use this informa- photochemical reaction) also will not be reflected tion. In most cases, however, we do not have di- in production figures. rect information regarding the amount of a chemi- For naturally occurring chemical substances, cal to which members of a population are ex- such as heavy metals, the most relevant param- posed, so that we must rely on indirect informa- eter is not production, but transfer from one com- tion, Several factors have been used, alone or in partment of the environment to another or con- combination, as surrogates for direct measures of version from one chemical or physical form to exposure. another. Such transfers are measures of the in- Production.—The most commonly used index of crease in the amount of the substance converted exposure is the annual production volume. Since into a form in which it is potentially available for this information is quantitative, it is easy to use human exposure. for scoring or ranking. In most cases, production Occurrence in the environment. Actual meas- volume does provide a rough indication of the urements of chemicals occurring in the environ- amount of a chemical substance potentially avail- ment are available for only a small number of sub- able for release into the environment. However, stances. Federal agencies tend to monitor on a the fraction actually released may vary over sev- regular basis only when required to by law. Air eral orders of magnitude, from less than 0.001 for pollution measurements are available for carbon well-contained industrial intermediates to 1 for monoxide, total hydrocarbons, nitrogen oxides, chemicals which are used dispersively. sulfur oxides, total suspended particulate, and For organic compounds produced by three or photochemical oxidants. Water pollution monitor- more manufacturers and produced in volumes ing data are available primarily for pesticides, greater than 5,000 lbs, production information is PCBs, and heavy metals. Monitoring information compiled in the annual reports of the Interna- in Government data banks is often not retrievable tional Trade Commission and is easily available. in readily usable form. There may be a timelag of production information can sometimes also be several years before a Federal agency summa- found in other literature sources or obtained from rizes, analyzes, and reports monitoring data. For manufacturers or trade associations, but it is not qualitative or quantitative data on occurrence in easy to acquire and will be missing for a large the environment of chemicals not included in 182 ● Environmental Contaminants in Food monitoring programs, it is necessary to search the 1971 the sole U.S. producer of PCBs voluntarily literature for individual studies. Since these restricted sales to those for closed-system ap- studies will usually report data for a small num- plications. They assumed that this would mini- ber of samplings at a limited number of locations mize risk to human health and the environment. (often one), they are difficult to use as a basis for They overlooked the fact that most environmental the comparisons needed for ranking chemicals. release of PCBs occurs, not through use, but The number of reported incidents of environmen- rather through disposal. tal occurrence will often be a function of interest The quantity of a chemical released into the en- in the compound rather than of actual frequency vironment is usually unknown so that scores as- of occurrence. signed for this factor must be based on estimates. An advantage of using this type of information, The “release rate” is an interesting index re- when available, is that it provides direct evidence cently developed by the NSF Workshop Panel to that the chemical is an environmental contami- Select Organic Compounds Hazardous to the Envi- nant. ronment. The release rate (R) was defined as fol- Use. Patterns of use can be very useful indica- lows: tors of potential for human exposure. A greater exposure risk is usually associated with disper- sive uses than contained uses. Contained uses are where P = overall annual U.S. production I = annual quantity imported those in which the chemical is not systematically L = fraction of the production lost at plant site during released to the environment as a direct result of manufacturing, conversion, and formulation use (e.g., PCBs in transformers). In dispersive D = fraction of the material which goes to noninter- uses, chemicals are released to the environment mediate dispersive uses as a direct consequence of use (e.g., pesticides, Estimates of L and D are usually semisubjec- aerosols, and volatile solvents). The size of the tive, based on knowledge of the manufacturing population at risk is usually greater for consumer process and on the likely breakdown of uses. uses than for commercial or industrial uses. The Environmental transport. A universal or widely class of the use (e. g., food additive, pesticide, or dispersed contaminant will usually be given drug) will often determine which Federal agency higher priority than a contaminant found only at has regulatory responsibility for a chemical. The specific locations or a contaminant whose entry use pattern will be a major determinant of the into the environment results from a unique event type, frequency, and amount of human contact or sporadic events. with a chemical substance. For a persistent chemical, the distinction be- Unfortunately, accurate, complete, and up-to- tween a local, widely dispersed, or universal con- date use information is very difficult to acquire. taminant may be a function of time. Unexpectedly Many industrial uses are considered “proprie- wide transport has been observed in recent years tary. ” Uses can change rapidly so that informa- for chemicals such as DDE and PCBs, PBBs en- tion compiled in secondary sources is frequently tered the environment as a result of a single acci- out of date. One problem encountered in using use dental occurrence and contaminated a small num- information is the difficulty of obtaining quantita- ber of Michigan farms. Since these chemicals are tive data. Since some uses are much more likely to persistent and have entered the soil, water, and lead to human exposure than others, a quantita- food supply, they will continue to disperse. tive breakdown of uses would be particularly val- Substances that appear to be relatively innocu- uable. Even when use data are available, it is im- ous or inert at their point of release may be trans- portant to realize that not all human exposure to ported to segments of the environment where they chemicals occurs during use. Exposure can also have significant adverse effects. Useful models occur during production, formulation, and dis- for the reliable prediction of the movement of posal or after environmental transformation. chemicals in the environment would be extremely Release. The rate at which a chemical sub- valuable for setting priorities among toxic chem- stance is released into the environment can be a icals for purposes of monitoring, control, or regu- useful parameter in ranking toxic chemicals. lation. Such information would also be valuable in Chemicals that are found in effluents or emissions determining which segments of the environment are most likely to become environmental pollut- should be monitored or controlled. Unfortunately, ants. Information on routes of entry of chemicals although models are available for problems such into the environment can in certain cases be as plume dispersion for stack emissions, there are essential to determining population exposure. In no generally accepted, easy to apply, widely used Appendix F—Priority Setting of Toxic Substances for Guiding Monitoring Programs ● 183 models for environmental transport. Moreover, quired for removal of half of the molecules of a realistic models of environmental transport are given compound from the environment or from a usually extremely complex, and it is impractica- specific compartment of the environment. Half- ble to use them in priority-setting exercises. life in the atmosphere, for example, would be a Persistence. Chemicals can be transformed and function of such properties as photoreactivity, degraded in the environment through chemical, reactivity towards active forms of oxygen, water photochemical, or microbial processes. The rel- volubility, volatility, and adsorption to fine par- ative importance of these processes will be in- ticulate. fluenced by the compartment of the environment Bioaccumulation. Bioaccumulation can be de- with which we are primarily concerned. Reac- fined as the long-term presence of a xenobiotic tions leading to simpler compounds which are substance in living tissue at concentrations sig- part of a series leading ultimately to the free ele- nificantly higher than those in the surrounding en- ments or elemental oxides are considered to rep- vironment. Chemicals that bioaccumulate are resent “degradation,” Other reactions lead to often given priority over other chemicals in haz- ‘‘secondary pollutants. ard-assessment schemes. Concentrations of chem- Persistent chemicals are often given priority icals that bioaccumulate are likely to increase over easily degraded or transformed chemicals in along a food chain, a process referred to as “bio- hazard assessment systems, on the assumption magnification. ” For such chemicals, residue that persistent chemicals are more likely to levels tend to be higher in herbivorous animals spread through the environment and will continue than in plants, higher in carnivorous animals than to accumulate as long as they are released. When in herbivores, and highest in carnivores that eat persistence is used as a criterion for hazard other carnivores. Since humans may eat orga- assessment, it is important to be aware, however, nisms in any of these categories and since chemi- that there are cases in which degradation prod- cals which bioaccumulate will usually be stored ucts or secondary pollutants are more toxic than in human tissues, if such chemicals have any toxic the chemical substances initially released, properties they may pose a higher risk to human Biodegradability. Although all living things can health than would be predicted on the basis of en- play a role in transformation of chemicals in the vironmental concentrations alone. environment, the lead role is attributed to micro- Since bioaccumulation is often associated with organisms. The role of micro-organisms in trans- a high lipid volubility relative to water volubility, forming chemicals is extremely important in the the partition coefficient can be used as an indi- lithosphere and also important in the hydro- cator of the tendency to bioaccumulate. The parti- sphere. Micro-organisms are frequently efficient tion coefficient is a measure of the distribution of at metabolizing natural chemicals in their envi- a solute between two immiscible liquid phases in ronment. Industrial chemicals are inherently less which it is soluble. For a single molecular species, likely to be rapidly degraded by micro-organisms. the partition coefficient is a constant and does not Parameters often used as indices of biodegrad- depend on relative volumes of solutions used. The ability are the biochemical oxygen demand (BOD) most commonly used coefficient indicates relative or chemical oxygen demand (COD), Estimates of volubility in organic and aqueous phases, provid- biodegradability can sometimes be based on ing information on the hydrophobic or hydrophilic chemical structure. nature of the compound. Chemical reactivity. The assessment of the Although absorption and accumulation of toxic chemical reactivity of a compound is often based pollutants are related to their partition coeffi- on known physical and chemical properties such cients, partition coefficient alone does not deter- as: mine bioaccumulation. Other characteristics of ● physical state, the chemical (reactivity, vapor pressure) and of ● vapor pressure, the organism (metabolic pathways, nutritional ● solubilities, status, fat content of body tissue) will also in- ● auto-oxidation propensity, fluence the extent of bioaccumulation. Active ac- ● redox potential, cumulation of a substance that resembles another ● acid-base characteristics, chemical is frequently predictable from the prop- b rate of hydrolysis, and erties and structures of both, ● rate of photochemical degradation. For inorganic compounds relative volubility is One parameter which can-be used for persist- not the primary factor in determining extent of ence in the environment is half-life, the time re- storage in body tissues. Passive accumulation

r .- II - “ 1 - 1 ‘, 184 w Environmental Contaminants in Food through formation of complexes with organic precise measures of degrees of hazard is debat- ligands, especially those containing sulfhydryl able, where available they are useful in priority- and amine groups, can also occur. The synthetic setting exercises because they represent an at- compounds most likely to accumulate in this way tempt to weigh all toxicity data available at the are organometallic chemicals. The tendency of a time when they were established. chemical to accumulate in this way can be pre- Metabolic factors. dicted from knowledge of the likelihood of the 1. Uptake and absorption.—Even if the amounts chemical to form complexes with natural bases. to which people ‘may be exposed could be Population at risk. The size or type of popula- determined, we still would not know the tion exposed may be considered as a factor in set- amount that could be expected to enter the ting priorities for toxic chemicals. Types of popu- human body and cause biological damages. lations at risk could be: The term “penetrability” is sometimes used ● the general human population, to represent that portion of the chemical to Q population of a geographical region, which the person is exposed that is expected “ population of a specific neighborhood, to be absorbed into the body, This factor in- ● population in the immediate vicinity of an in- cludes consideration of both the routes of ex- dustrial plant, posure and the fraction of the chemical ab- ● people whose diets contain a high amount of sorbed after exposure by each route, particular foods, 2. Biotransformation. -The metabolic conver- ● occupational groups, and sion of a chemical within the human body ● highly susceptible groups. should be considered, as far as possible, in Population groups can be further classified ac- assessing potential health effects. A chemi- cording to whether their exposure is voluntary or cal can be detoxified, activated, or converted involuntary. Priority considerations should be to a compound with different biological prop- given to chemicals to which there is extensive in- erties. The metabolic pathway can be dose- voluntary public exposure or to which susceptible dependent. segments of the population are exposed. Some 3. Elimination. —The turnover rate in the ranking schemes consider certain segments of the human body can be an important factor in population more ‘‘valuable’ than others (e.g., the assessing biological effects. If, however, a young more “valuable" than the old). This ap- chemical has been scored for bioaccumula- proach is highly subjective and ethically ques- tion as an exposure factor, it may not be nec- tionable. essary for it to be scored for turnover in humans, as these factors are closely related, Toxicity Absorption, bioconversion, and elimination are The toxicity of a chemical is the second deter- complex factors and are difficult to handle for minant of hazard. The difficulties encountered in ranking purposes, Information is not readily estimating the biological activity of a chemical available and is rarely compiled in easy-to-use are as severe as those encountered in estimating secondary sources. For many chemicals, no meta- exposure. Here we find ourselves at ‘‘the fron- bolic information is available. When metabolic tiers of scientific knowledge” in attempting to studies are found, they are often reported in evaluate the significance of laboratory data for qualitative rather than quantitative form. Relat- human experience. ing metabolic information to adverse health ef- Most priority-setting efforts have scored chemi- fects often requires a high level of expertise in cals on a series of toxic effects, There are, how- toxicology. Despite all these problems, a toxico- ever, some general indices of toxicity, such as the logical ranking system which takes into account threshold limit values for atmospheric exposure these metabolic factors is more sophisticated in the workplace established by the U.S. Depart- than and usually superior to a system which does ment of Labor and the American Council of Gov- not, ernmental and Industrial Hygienists, tolerance Acute toxicity. The biological effect of a chemi- levels for pesticides in food set by EPA, or “action cal depends on the quantity of the chemical with levels’” for food contaminants set by FDA. These which the organism must deal, Chemicals having indices are based on reviews of a range of toxicity adverse effects at low dosage would be consid- data and represent attempts to identify threshold ered more toxic than chemicals having similar ef- levels for hazard or for lack of assurance of safe- fects only at much higher dose levels. It is difficult ty. Although the reliability of such indices as to perform this ranking at low effect levels. The Append/x F—Priority Setting of TOXIC Substances for Guiding Monitoring Programs ● 185 effects of low dosages are often subtle. Direct ex- moderately carcinogenic, weakly carcinogenic, perimental estimation of the level affecting 1 per- co-carcinogenic, or having no neoplastic effects. cent of the population may require several hun- There is usually not enough information avail- dred animals to obtain adequate statistical pre- able, however, to do this. Dose-response curves cision. For these reasons, lethal dosage levels are are rarely determined for carcinogens. Negative usually used in ranking chemicals for acute tox- results in carcinogenicity tests are usually not ac- ic i t y. cepted as proof of noncarcinogenicity. Compari- A common parameter of acute toxicity is the sons between carcinogenic “potency’ of various chemicals have been made on the basis of per- animals would die. To be comparable, results cent age of a test population developing malignant should be based on animals of the same species, tumors, the number or types of species in which strain, sex, and age. The same route of adminis- positive results have been reported, the types of t ration should be used. tumors observed, the lowest dosage causing ma- Other commonly used parameters of acute tox- lignancy, or the lag time between administration icity based on lethal doses are: of chemical and observation of tumors. LC-,(; = the concentration which is lethal to half Chemicals are sometimes classified with re- the test population. The duration of ex- spect to the degree of certainty of their carcinoge- posure should be specified. LC is usu- nicity (e. g., known carcinogen or suspected car- ally used for concentration in air, but cinogen). EPA has recently attempted to order the can also be used for concentration in NIOSH list of suspected carcinogens according to the ambient water to which aquatic or- the relative degree of concern that might be war- ganisms are exposed ranted regarding possible human carcinogenic LD, ,, = the lowest reported lethal dose potential. A four-digit code was used. The first LC,,, = the lowest reported lethal concentra- digit represented the species in which carcino- tion genic response was reported. The second digit designated the number of different species for Available data on these parameters for 16,500 which a carcinogenic response was reported. The different chemicals are summarized in the Regis- third digit was assigned on the basis of route of try of Toxic Effects of Chemical Substances pub- administration. The last digit was a count of the lished by NIOSH. The major objection to acute number of different species-route combinations. lethality as a basis for ranking environmental The most complete source of data and refer- contaminants is that the acute lethal doses are ences related to chemical carcinogenesis is the very much higher than those levels one would ex- Public Health Service’s Survey of Compounds pect to find in the environment. Furthermore, Which Have Been Tested for Carcinogenic Activ- values for acute lethality can vary widely be- ity. A master index of the series is maintained on tween species. Use of this parameter can be justi- tape at the National Cancer Institute headquar- fied, however, on the basis that the LD-),, is at least ters. a crude index of biological activity. It is reason- able to consider a chemical w i t h a very low LD,(, to Evaluations of carcinogenic risk are made by be a relatively toxic chemical. on the other hand, the International Agency for Research on Cancer a chemical with a high LD,,, is not necessarily under the auspices of the World Health Organiza- safe, since long-term exposure a t low concentra- tion. They do not use a formal ranking system. tions may have such adverse effects as carcino- Mutagenicity, A mutation is defined as any her- genicity or damage to the reproductive system. itable change in genetic material. The results of Data on acute toxic effects other than lethality mutations are usually undesirable and may in- are less frequently compiled, are more difficult to clude abortion, congenital anomaly, genetic dis- use for ranking purposes since they arc not usual- ease, lowered resistance to disease, decreased ly reported in numerical form, and may not be in- lifespan, infertility, mental retardation, senility, dicative of effects occurring at the low dosage or cancer. levels ordinarily found in the environment. Never- The reliability and accuracy of the prediction theless, nonlethal acute effects can provide useful of mutagenicity in man from the ability of a chem- information about target organs and should be ical to produce mutations in experimental tests considered as part of the overall picture when systems is controversial. Many types of tests are subjective judgments are made. currently available for use as indicators of muta- Carcinogenicity. In principle, it would be useful genic potential. Some (e. g., the Ames test, the to categorize chemicals as strongly carcinogenic, dominant lethal test) are actual tests for mutage- 186 ● Environmental Contaminants in Food

nicity. Others test for a variety of events relating tured mammalian embryos, and incubating chick to, or at least correlated with, mutagenesis. Some embryos. Of these systems, chicks and to a lesser such tests are for chromosome aberrations, DNA extent fish are the most widely used, damage or repair, increased rate of sister chro- matid exchange, or other phenomena considered It is difficult to rank chemicals on the basis of to be relevant to mutagenic potential. Because teratogenic potency. The type of dose-response even those tests that actually demonstrate muta- data needed for quantification of teratogenic risk genicity are highly specific for detecting a single is seldom available. Interspecies extrapolation is type of mutation, results of a battery of tests are poorly understood. There is an enormous range of desirable for an evaluation of mutagenic potential variability of teratological end points. Chemical of a chemical substance, structure has been found to be of little use in Structure-activity relationships have met with predicting teratogenicity. limited success in prediction of mutagenicity. Other toxic effects. Many chemical mutagens (but not all) act as elec- 1, Effects. . observed in humans, —Reports of tox- trophiles permitting a prediction of mutagenic po- ic effects in humans can be based on epi- tential of compounds that might form a reactive demiological or clinical observations. There electrophilic species. are severe problems in interpreting this type Many difficulties are encountered in extrapo- of information. Humans are rarely exposed lating to man results obtained in microbial, mam- to only a single chemical substance, It is malian cell culture, or test animal systems. Poten- usually difficult to find control groups with tial for mutagenic effects in man should be as- zero exposure to an environmental contami- sessed only after results of testing in various ex- nant. Results based on occupational expo- perimental systems are considered in conjunction sure may underestimate risk, because sus- with other relevant factors such as chemical ceptible segments of the population (infants, structure, biochemical activity, and metabolism children, the aged, the chronically ill) are ex- in man. cluded from the workplace. When workers Teratogenicity. Certain chemicals, when ad- become sick they often stop working and are ministered to a pregnant female, can interfere at no longer part of the occupational cohort. a critical stage of embryogenesis or organogene- Another problem in interpreting human data sis and can cause embryonic death or malforma- is that the doses are often unknown. tion, A chemical with such properties is a tera- togen. Testing for teratogenicity should ideally in- 2. Chronic toxicity, —Biological effects of a volve administration of the test substance only at chemical administered in low doses over an the period critical for organogenesis of the system extended period of time may be quite dif- sensitive to the teratogen. Sensitive periods of or- ferent from acute effects. Chronic exposure ganogenesis are characteristic of particular test protocols are more relevant than are chemical teratogens. acute protocols to the types of exposure due At the present time, only tests in pregnant to environmental contaminants. mammals are generally accepted by scientists Because of the wide range of end points and regulatory agencies as valid tests for terato- observed during chronic exposure tests, genicity, because the maternal-placental embry- comparisons are difficult to make. Compar- onic relationship cannot be duplicated in sub- ing one type of effect with another (for exam- mammalian test systems. Other test systems ple, comparing impaired fertility with liver sometimes used as indicators of teratogenic po- damage) on the basis of severity is highly tential include bacteria and other unicellular subjective. Ranking chemicals on the basis of organisms, somatic cells in culture, tissue culture, chronic toxicity therefore requires a high organ culture, intact invertebrate embryos, cul- level of scientific expertise. Appendix G

Approaches to Monitoring Organic Environmental Contaminants in Food*

by John L. Laseter, Ph.D.

INTRODUCTION

Trace quantities of potentially toxic organic selected, it is imperative that it provide une- compounds are frequently found in the environ- quivocal results. ment. These compounds sometimes possess prop- Most methods for monitoring and for the analy- erties that may have teratogenic, mutagenic, and sis for trace levels of organic pollutants in the en- even carcinogenic effects on humans and ani- vironment necessarily consist of several steps< mals. Some of the compounds, such as pesticides, These include sample collection and storage, sam- have been intentionally released into the environ- ple workup, and component identification and ment (1), while others have found their way into quantification. the environment by accident, through careless- The collection and storage of samples is an im- ness, or as byproducts of industrial processes. portant phase of an analytical method if meaning- Many of these compounds are subject to break- ful interpretation of the data obtained is to be down in the environment as a result of both physi- achieved. The sample selected for analysis must ochemical and biological processes, such as be representative of the whole system being ex- chemical weathering, photodecomposition, me- amined, and must be free of contamination due to tabolism, and biodegradation by micro-organisms improper collection and handling techniques. (1,2). In many cases, it is not uncommon for the Once collected, the samples must be stored under breakdown products to have greater health ef- conditions that will reduce or eliminate changes fects on humans and animals, than the parent in their composition. compounds. Another important phase in an analytical Because of the potentially toxic properties of so method is the sample workup. During this stage, many of these compounds and their breakdown the sample usually undergoes an extraction proc- products, several pieces of Federal legislation ess whereby the compounds of interest are re- have been enacted in recent years to monitor, moved from the sample matrix. Organic solvents, evaluate, and control the amounts of the pollut- and in some cases inert gases, are usually em- ants in order to protect our health and environ- ] ployed in the extraction process. Because the ex- ment. As a result, monitoring (screening) pro- traction process is seldom very selective, many grams have been established for many types of or- organic compounds in addition to those of interest ganic pollutants in the environment. All of these are also extracted. In order to reduce the programs rely heavily on the ability of analysts to amounts of these other compounds that may inter- correctly identify and quantify these compounds fere with the analysis, a sample cleanup proce- at the parts per million (ppm) and parts per billion dure usually follows the extraction process. (ppb) levels in a variety of sample matrices (3). In The last and most difficult phase of an analyti- order to successfully obtain useful information at cal method is the identification and quantification these concentration levels, it is essential that stage. The identification process is often accom- methods which are sensitive and selective be de- plished by comparison of the physical and chemi- veloped and used. Whatever method is ultimately cal properties of the unknown compound against the same properties of an authentic standard * ~~t)r purp{)st’s of this (i[)fwmont. fl)()[l r(?pr(w}nts ,ill st)ll(i, compound. For complete unknowns, the identifi- sf)m IS(JI i(i, ;I n(i IICIII i[i forms of f{)()[i pro(iu{ts. including I)l)t t I[v I w,1 t [’r, ( t )Ils u m[ x ] I]v m:] n. cation process can be very difficult. The quantifi- “I I1(! t[’rms ‘‘m{ Im t I ) r I n:” :] n(i ‘‘ S( rm’n i n~” ;I rc usmi in t (>r- cation process can only be accomplished after the ( il(i nq(’,~l)ii I n t il is (iIJ{ L]m(’n t. unknown compound has been identified and usu-

187 188 ● Environmental Contaminants in Food ally involves comparison of the detector response etry (12-14). This technique expands the area of for the compound of interest against the detector trace organic analysis to the identification and response for known quantities of an authentic quantification of compounds that are not suitable standard. for analysis by gas chromatographic techniques. In the analysis for trace levels2 of organic com- pounds in the environment, it is often very diffi- cult to obtain accurate and reliable qualitative as well as quantitative results. This is evident by the Table G-1 .—Techniques Available for Qualitative countless examples in the literature of errors in and Quantitative Organic Analysis both qualitative identifications and quantitative estimations of trace quantities of many organic detection Specificity or compounds in the environment (5). common uses Of the many methods currently available for the qualitative identification and quantification of Detects most organic compounds, few are sensitive and spe- compounds cific enough for meaningful trace analysis. Table Halides, conjugated carbonyls. nit riles. di- G-1 summarizes the techniques for organic analy- and trisulfides sis and some of their advantages and disadvan- Phosphorus, sulfur tages (4,6). The most common techniques in use for trace Nitrogen, phosphorus organic analysis are gas chromatography with the use of selective detectors (7) such as the elec- Compound type tron capture (EC). Hall electrolytic conductivity, determination and flame photometric detectors; high perform- Classical functionality ance liquid chromatography (HPLC) (8); and com- determination Sulfur, nitrogen, bined gas chromatography-mass spectrometry halogens (GC-MS). Selective detectors for gas chromatog- raphy are necessary because environmental samples are often very complex and the selectivi- ty of the detectors simplifies the analysis by allowing only certain compound classes to be de- tected at any one time. The most powerful of these techniques is the GC-MS technique (4,6,9-1 1), since it not only provides qualitative information of nanogram quantities of single compounds pres- ent in the sample, but also provides information Best for complete which can be used for quantification of individual identification, components in the sample. molecular weight structure, and A more recent technique involves the combina- function Confirm any tion of liquid chromatography and mass spectrom- compound E’Or purposes Of this II[)fument, ‘‘~ rare levels” is df?finfxi [ls a ronccn tr(] t ion helt)w lhf’ IOU p{] rts-pcr-m ill ion level. I)ctf?rtion at these rf)nrcnt r{) t ions is import;lnt he(tius[! m:inv org:]nirs arc biologir{i 1 Ii’ ~] (’ t ivc even :] t t hf: pa rts-pf!r-t rl I lion If?vel. Append/x G—Approaches to Monitoring Organic Environmental Contaminants in Food ● 189

ANALYSIS FOR EPA PRIORITY POLLUTANTS IN FOODS AND WATER

The recent EPA/NRDC consent decree estab- food samples. An estimate of the space, man- lished an analytical procedure for the analysis of power, and cost associated with the analysis of 129 priority pollutants (chemical indicators of or- the present 114 organic EPA priority pollutants in ganic pollution) in industrial waste water (1 5). Of food and water samples is given in table G-2. It these, 15 pollutants are metal, with the remainder should be pointed out that considerable research being individual organic compounds and com- and development effort must be expended to pound classes. The procedure for the analysis of adopt the methods proposed to foods in general. the organics is strictly a GC-MS computer method designed to provide qualitative as well as quan- Figure G-1 .—A Simplified Diagram for the titative information about the presence of the Qualitative and Quantitative Analysis of priority pollutants in waste waters. the Organic EPA Priority Pollutants Basically, the procedure requires that four separate analyses be performed on the sample. These are an analysis for the more volatile or-

involve liquid-liquid extractions, and subsequent NOTES 190 ● Environmental Contaminants in Food

Table G-2.—Estimated Space, Manpower, and Cost Associated With the Analysis of Priority Pollutants in Tissue and Water Samples According to EPA Analytical Protocola

Category Response and/or cost A. Instrumentation ...... GC-MS data system with automated Iiquid injection device. -$200,000 each B. Space...... 1,500 ft2 equipped with typicalb laboratory facilities and furniture including adequate air-conditioning. C. Downtime for instrumentation 30% (this figure can vary from 20 to 50% as a function of staff experience and logistics support). D. Minimum assignable ...... Ph. D or equivalent - 1 manpower M.S. or equivalent - 1 B.S. or equivalent - 2 Total 4C Estimated operational cost per routine sample -$2,000 to $2,500 (Scheme One as illustrated in figure G-l). Estimated operational cost per routine sample -$2,500 to $3,000 (Scheme Two as illustrated in figure G-2, as a function of cleanup difficulty).

a It , ~ a~~umed t-hat the laboratory ha~-~ther ~n~~ing actlvlt Ies and further that samples would be analyzed on Iy once and In succession The ultimate size and cost of such a monitoring faclllty IS based on processing only a few hundred samples per year No capital costs are calculated Into the operational cost per sample bTyplcal laboratory facllltles and furniture Include laboratory benches tables with sinks shelves for rea9ents and equl Pment. vented hoods, refrigerators, laboratory balances and scales, pH meters, hot plates, laboratory glassware, c hem Ical, etc This Includes special toxic chemical and carcinogen processing fac I I ltles Laboratory space of the type desc rlbed dbOW cost on the order of $lOO/ft Z CManpower requ ,rements are not Included for maintenance and 1091Stlc suPPort

SCREENING FOR UNSPECIFIED POTENTIALLY TOXIC COMPOUNDS AND CHEMICAL CLASSES IN FOOD AND WATER

There is a possibility that other compounds or steroids, phenols, amines, halogenated organics, classes of compounds not included in the EPA polynuclear aromatic hydrocarbons, and any priority pollutants list will find their way into the newly appearing organic that illustrates a environment. For this reason, it is essential that marked increase in concentration over a period of some form of monitoring system be established time. that will look for the appearance of particular The monitoring program design would involve compounds or classes of compounds in food and the establishment of appropriate sampling and water over a period of time. A recent National sample-handling guidelines, together with the Research Council report on environmental moni- modification of currently used analytical pro- toring (20) recommends the establishment of new cedures to include the new compounds or classes monitoring programs to anticipate pollution prob- of compounds. A typical monitoring program lems and to discover environmental pollutants in would involve the use of high-resolution gas their early stages of development so that ap- chromatographic techniques using the universal propriate corrective measures can be imple- flame ionization detector (FID) capable of detect- mented before the problem becomes unmanage- ing traces of known candidate compounds, and able, or worse, irreversible. A typical monitoring computer techniques that allow for rapid com- program is the EPA Mussel Watch Program (2 I), parisons of samples to establish trends. With the the EPA National Pesticide Monitoring Program, use of high-resolution gas chromatography (em- and the National Pesticide Monitoring Network ploying glass capillary columns”), one can readily for Birds operated by the U.S. Fish and Wildlife separate in an environmentally derived sample, Service. In establishing a monitoring program, the type of compounds to be monitored would be selected as candidate compounds on the basis of their chemical class, their use, and their suspected tox- icity, These candidate compounds might include Appendix G—Approaches to Monitoring Organ/c Environmental Contaminants in Food . 191 several hundred organic species, the vast majori- and use of pattern recognition techniques.5 The ty of which have not been characterized. monitoring would be set at a concentration level The analytical program would involve a prelim- below the actual legal accepted 1evel (for exam- inary screening to establish baseline levels of the ple, 0.01 ppm or one-tenth of the action level if it is candidate compounds in samples of foods and known), water, or selected indicator species over a given Should the screening result in an observation of period of time, followed by periodic screening of an increase in levels of either unknown or se- similar samples to determine changes or trends lected compounds or classes of compounds, sug- with time. If alarming trends are observed, cor- gesting the entry into the food or water of new rective actions and measures can be imple- materials, additional analytical efforts would be mented. employed to attempt characterization of the new Of the techniques available for analysis, the compound(s) observed. Preliminary information methods best suited for the class or classes of would be transmitted to associated toxicologists compounds under consideration would be se- for evaluation and comparison with information lected for routine monitoring. These would in- available on known toxic compounds. clude high-resolution gas chromatography or high-performance liquid chromatography and supporting computer methods which would in- clude the use of internal and external standards,

LABORATORY REQUIREMENTS

Because organic analysis at the submicrogram center would be equipped with data-archiving fa- (ppm and ppb) levels is a complicated and difficult cilities for both GC, LC, and GC-MS data and data- process requiring sophisticated instrumentation processing methods that would allow for rapid re- and expertise, the laboratory selected to perform trieval, comparison, and evaluation of these data, analyses for trace organics in foods and water Some of these data-handling techniques are now must be equipped with state-of-the art technology available and others are under development and experienced personnel. The best setting for (23-26] for such a central data management sys- such a laboratory is in a location where it can tem (see table G-3). Figure G-3 illustrates a pro- establish ties with the R&D community where ex- posed analytical scheme to monitor for the ap- perts in a variety of disciplines can be found who pearance of unknown organics in addition to the can serve and participate as consultants at vari- current EPA priority pollutants in food and water ous levels. It should be pointed out that analytical samples. techniques may require modifications to upgrade In support of the nationwide monitoring pro- the technology and a substantial expenditure of gram would be a quality control laboratory which funds to keep the program operating at maximum would coordinate intercenter calibrations, and efficiency and information output. would spot-check and confirm selected data de- The typical nationwide monitoring program veloped by the regional laboratories. Of necessi- would have several regional centers, each well- ty, the quality control laboratory would be better equipped with the appropriate instrumentation equipped than the regional centers, so that it and personnel to perform the analyses and con- could resolve the problems and issues that can oc- duct the program. In addition each center would cur during routine analyses. A minimum of 10 per- have a review board composed of senior scientific cent of all samples analyzed by the regional cen- personnel in such fields as toxicology, analytical ters would be confirmed by the quality control chemistry, environmental chemistry, etc., to as- laboratory, Additionally, a library of samples sess and interpret the data developed. To facili- used in the actual analytical studies would be tate the handling of all the data acquired, each stored for future reference. 192 “ Environmental Contaminants in Food

Table G-3.—Estimated Space, Manpower, and Cost Figure G-3.—General Analytical Scheme To Detect Associated With the Monitoring of Unknown and Monitor New Trace Organics and Priority Potentially Toxic Compounds in Addition to Pollutants in Food and Water Samples the EPA Priority Pollutants

Category Response and/or cost A Instrumentation a Small, high throughput GC-MS data-system with automated Iiquid injected device I -$50,000. b El/Cl-equipped high-resolution MS-data system with automated in- jection device -$200,000 to $300,000 c GC-FID-EC system $15,000 (addi- tional chromatographic systems may be required). d LC Interfaced into MS system -$20,000 II (? Central data management system

I r,

B Space

C Downtime for Instrumentation

I z i ~ratl ,n D Minimum I ~ assignable manpower NOTES

1 A vdrlel} of f r,+c I Ionat 1011 .in[j I>oldl ion pro( (>(1 (J ret t ion to I Ia u I(i I I(I u Id ex t rdc t Ion

E Estimated operational cost per routine sampIe Appendix G—Approaches to Monitoring Organic Environmental Contaminants in Food ● 193

DEVELOPMENTS THAT WILL IMPACT LOW CONCENTRATION ORGANIC MONITORING

Of all the developments in technology for chem- (FT-IR) detectors; and 3) developments in com- ical analysis, those which will have a greater im- puter system software capable of handling mas- pact on monitoring programs are the following: sive volumes of chromatographic as well as mass 1 ) developments in mass spectrometry instrumen- spectrometric data of the type obtained in a moni- tation, such as pulsed positive-negative chemical toring program. Developments in other methods ionization and detection techniques; 2) develop- such as electrochemical techniques and plasma ments in selective detectors for gas chromatogra- chromatography show some promise for trace or- phy and liquid chromatography, such as atomic ganic monitoring, but will only be useful if they absorption (AA) and atomic fluorescence (AF) can be coupled to GC or LC systems. spectrometry, and Fourier transform-infrared

ACKNOWLEDGEMENTS

The technical assistance and cooperation of 1. FDA and EPA and Mr. R. Finnigan, of Finnigan R. DeLeon and the help of D, Trernbley in prepar- Corpora t ion, for their helpful technical discus- ing this manuscript is gratefully acknowledged. sions. The author also mishes to thank the personnel of

SELECTED REFERENCES

15. U.S. Envi ronmen t{]) Pr[)tert i[)n AgencP’, ‘‘Sampling and A II;] lYIS is Procedu r [?s f( )r S r re[?ni n~ I ndus t r i:] I Efflut?nts for Prioritl’ Pollutants,”” ( ;in(innnti, ohio” (Revised, April 1977). 16. 1. R. Poli tzer, S. Gi t hens, B, J, L)(Iw tk, and J. 1,. 1,ase- tf;r, ]. (l]r~)m(lt[)~r. Sci., 13, 378 ( 1 975). 17. B, J. Dowtpr, 1., E, (;ret;n, :]nd J. L. L{]set[?r, AIIIII (;hem., 48, 947( 1976]. 18. I. R. Poli tzer, B. J. Ilowtl, :ind J. 1,. I,[]s[?t(!r, (linic[rl (;hf?m., 22, 1775( 1976). 19. I). L\’. Kuehl, An(l). f;hem., 50, 182 ( 1978). 20. “‘Environmental hioni torirl~, ~’(llllm(l IL’, N:) t ion:]l 6. A(:adem; of Scien(:c?s, [\’[lshing ton. ]]. (;., pp. 6-8 ( 1977). 21. E. I). (j[)ldhf:rg, ~’ ‘l’. 130wen, J. bf’. F’a rrington. (;, 7. Htlrvey, J. 11. hlartin, p l,. p:] rkcr, ~. P\’. Rise- t)rough, W“, Robertson, E. S(:hnci(i[?r, an(i E. (;aml)l(~. 8. Fl]iiron, CI)ns(>rv. 5, 1 ( 1978). 22. ‘1’. L. Iserlhour, H. R. Kow[llski, and P. (;, Jurs, (;li(; (Jritic(ll }ICJ in An~lJ (jh[~m, (July 1974). 9. 23. E. B. Ov[?rton, (1. F’, Steele, and J. 1,, Las(?ter, ‘ ‘Im- 1 (), proved Data Prwx?ssing Software fur (;1:{ss (;iipil- lar~ Separ~]ti(Jn of (;omplf?x Erliir(]rlnl[?rlt:l! S;lm- pies, ” Pittsl)ur~h (h)nf[?r~?nc[? F’t;h. 27-hli~r. :3, 1978, 11. (;level:~n(i, ohio.” 22, 4’2, 3 ( 1978). 24. 1. I{. Suffet :ln(i E. R. (jl;]s(?r, /, ( ;h r~)rn~l t{),gr. .${i,, 1 (;, 12 ( 1978), 25.1. t!’. h:]r[~sok.. ln~lut(ri(~l R ‘/1, ,20, 113 (April 1978). ~ [j. J-:. 13. ( )v(lr t on, (;, F. S I (?(?1 [?. i 111( i J. 1,. 1,:] s[; t [: r, J. H i v 1) H(~t{)l~l [](~n (;lll~)r]lflff)gz{]]~llk (sul)mitt[”(i 1 978]. 194 ● Environmental Contaminants in Food

GLOSSARY OF ABBREVIATIONS AND TERMS USED

Abbreviations sample results from the bombardment under high vacuum of the sample molecules by a beam of elec- AA—atomic absorption trons, usually at an energy of 70 electron volts (eV). AF—atomic fluorescence Flame Ionization Detector—a universal detector for CI—chemical ionization gas chromatography. This detector is fairly sensitive DS—data system and very linear, and responds well to most organic EC—electron capture compounds. EI—electron impact Fourier Transform-Infrared Spectrometry—this is the FID—flame ionization detector state-of-the art of infrared spectrometry. This tech- FT-IR-Fourier transform infrared nique employs minicomputers and Fourier transform GC—gas chromatography methods in the acquisition of infrared spectra. Ad- GC-MS--combined gas chromatography-mass vantages of this technique include the making of spectrometry measurements in a fraction of the time required for HPLC--high performance liquid chromatography the more conventional methods, and increased sen- LC—liquid chromatography sitivity . LC-MS--combined liquid chromatography-mass Gas Chromatography —one of the most widely used spectrometry analytical techniques for trace organic analysis. The MS—mass spectrometry technique is simple and very rapid to use, is extreme- TLC--thin-layer chromatography ly sensitive, allowing the use of minute amounts of UV—ultraviolet samples, and can be very useful for preliminary screening of environmental samples. Terms Gas Chromatography-Mass Spectrometry—a very Atomic Absorption Spectrometry—a form of spectro- powerful analytical technique that combines the chemical analysis usually applied to the determina- features of gas chromatography and mass spectrom- tion of the elements. The sample is heated to a rela- etry. tively high temperature to cause dissociation of the High-Performance Liquid Chromatography—high-res- chemical compounds into atoms. A source of radia- olution, high-speed, and high-sensitivity liquid chro- tion characteristic of the element to be determined is matography. passed through the sample. If the sample contains Liquid Chromatography—a separation technique that the element, absorption of the radiation by the sam- allows the partition of the sample between two ple atoms occurs and the amount of absorption can phases, a liquid and a solid, or two liquid phases. be measured for quantitative determinations. Liquid Chromatography-Mass Spectrometry—a re- Atomic Fluorescence Spectrometry—a form of spec- cently developed analytical technique that combines trochemical analysis usually applied to many organic the features of liquid chromatography and mass compounds and some inorganic compounds which spectrometry. emit radiation energy after they have first absorbed Mass Spectrometry—a very powerful tool for provid- radiant energy of a particular frequency. The sim- ing the structural identity of complex organic mole- plest form of AF is the fluorescence provided by a cules. Mass spectra furnish information about the monoatomic vapor, such as sodium. Chemical Ionization—a method for ionizing samples for mass spectrometric analysis. The ionization of the radical ion, which is usually produced by electron sample results from the reaction between the sample bombardment in the ion source of the mass spectrom- molecules and low-velocity reagent ions which re- etry. sults in the transfer of a charged species other than Thin-Layer Chromatography—a form of liquid-solid an electron. chromatography conducted on the surface of spe- Electron Capture Detector—a very sensitive and selec- cially prepared plates. This technique is generally tive detector for gas chromatography. This detector very quick and simple to use and is effective in per- responds to the presence of a variety of compounds forming separations of small amounts of sample. containing atoms with an affinity for electrons, such Ultraviolet Spectrometry—a form of absorption spec- as the halogens--chlorine, bromine, and fluorine— trometry generally suited for analysis of compounds and o t her atoms, such as oxygen, and sometimes. that are capable of absorbing ultraviolet radiation. even sulfur. These include aromatic compounds, conjugated ke- Electron Impact—a method for ionizing samples for tones, and other conjugated compounds, such as mass spectrometric analysis. The ionization of the polyolefins. Appendix H

Analytical Systems for the Determination of Metals in Food and Water Supplies

by R. K. Skogerboe, Ph. D.

INTRODUCTION

The analysis of food products or water supplies in sufficient quantities to cause deleterious for toxic or potentially toxic elements, e.g., heavy effects on the consumer population? metals, is the central problem addressed in this This question may be considered qualitative in report. An evaluation of this problem in terms of that it actually requires only a yes or no answer. the capabilities and limitations of applicable ana- Given the knowledge that each of the chemicals lytical techniques is the primary thrust of this being monitored must be present at or above some report. Thus, recommendations regarding the se- threshold’ concentration before they individually lection of the best analytical methods and tech- or collectively produce observable effects on the niques for the determination of toxic elements are recipient population, the answer is no if all are presented below. Before doing so, it is appropri- below their respective threshold effect levels and ate to delineate the criteria used in the develop- yes if one or more is above. As a result, it is quite ment of the recommendations and to discuss phil- common to use analysis approaches capable of osophical rationales on which these judgmental determining or detecting the chemicals only at guidelines are based. levels down to, but not below, their respective threshold levels. Although this practice can often The Analytical Process be justified on an economic basis, it is subject to challenge for scientific reasons. Any chemical analysis can be divided into three The astute recipient of the answer to the cen- sequential steps: 1) collection of the sample(s); tral question will immediately raise other ques- 2) chemical and/or physical preparation of the tions regardless of whether the answer is positive sample(s) for analysis: and 3) measurement of the or negative. If informed that all constituents of concentrations of the target constituents in the concern are below their individual or collective sample(s). Although these steps are interdepend- threshold levels, two questions are obvious: ent and should not be considered otherwise, the ● What degree of confidence can be assigned present discussion will focus on sample prepara- to the results? tion and measurement. Sampling should be dis- cussed in the broader context of the overall prob- lem of monitoring. The selection of an appropriate analytical technique must be based on the type of information desired and the purposes of collect- ing that information. In the present context, this may be delineated in fairly general terms. The Monitoring Question With rare exceptions, the central question as- sociated with a monitoring program is: ● Are there one or more chemical entities pres- ent in the target material (e.g., food or water)

195 196 c Environmental Contaminants in Food

● Are the concentrations of any chemicals in- There is one general criterion that may be de- creasing significantly over time for a partic- fined on the basis of the above discussion. All ular monitor and/or a particular collection analyses should rely on analytical methods that site? are capable of determining all target contami- When advised that one or more constituents nants at concentration levels below the threshold are present above threshold levels, these same effect levels. While methods capable of making two questions are raised and a third becomes per- these measurements at concentrations 100 to tinent, i.e., 1,000 times below these levels would clearly be ● What is the cause or source of the observed desirable, the use of methods providing measure- contamination? ment capabilities 10 to 100 times below the levels These questions are clearly quantitative in may prove more practical on an economic basis. nature. Decisions regarding possible impacts on This discussion provides a basis for delineating consumer populations, or the prevention thereof, a general protocol for the operation of the analyti- should not be based on less than quantitative and cal laboratory responsible for answering the defensible information. Adherence to this philoso- above questions. A brief discussion of this is pre- phy is complicated by the facts that reliable desig- sented here to provide a general basis for suc- nations of threshold effect levels are often lacking ceeding topics. and that two or more contaminants may act syner- gistically or antagonistically.

OPERATIONAL PROTOCOL FOR A MONITORING LABORATORY

A general flow diagram of the laboratory oper- that all required analyses can be completed and ation is presented in figure H-1. The analysis se- to provide comparative information relevant to quence given is depicted in the context of the mon- validation of results; and 3) a data storage-re- itoring questions, the checks required to validate trieval system consistent with the requirements of the results, and the regulatory actions likely to the monitoring questions, the quality assurance prevail. Examination of this protocol indicates the program, and regulatory actions. The general util- probable need for inclusion of: 1) a quality-assur- ity of this protocol diagram and the actions it por- ance program as a means of validating results; trays will be expanded upon below. 2) complementary analytical methods to ensure

CRITERIA FOR SELECTION OF ANALYSIS METHODS

The selection of an appropriate analysis ap- states of the element that may be present. Such proach must be based on the information require- measurements may, in fact, be relatively nonspe- ments. These may be formulated on the basis of cific indicators of potential or actual deleterious the answers to two questions: effects on biological systems. In numerous in- . What are the chemical entities that must be stances, the identification and measurement of determined? the active or functioning forms of the elements is c What are the concentration levels (threshold actually needed. The following examples testify to levels) of primary concern for these target the general importance of this statement. entities? 1. Although-arsenic is toxic, the plus three oxi- Although answering the first question should dation state (As(III)) is clearly more toxic not be particularly difficult in most instances, the than the plus five state (As(V)); the compound issue for toxic elements has often been obscured arsine (AsH,) is perhaps the most toxic chem- by the use of expedient analysis approaches ical form of arsenic. which may or may not provide the information re- 2. Although chromium is classified as a nutri- quired. Historically, most analysis methods used tionally essential element, Cr(III) is toxic for the determination of elements simply measure while Cr(VI) is relatively innocuous. the total amount of an element present without 3. A measurement of the level of vitamin Bl~ in differentiating between the various chemical animal tissue is ordinarily more useful than a Appendix H—Analytical Systems for the Determination of Metals in Food and Water Supplies ● 197

element (in all of its chemical forms) can be ra- tionalized. For example, if the analysis indicates that the total concentration is below the threshold effect level for any one or combination of the par- ticular chemical forms for an element, the need for measurement of the concentrations of the chemical forms is negated. The total concentra- tion measurement consequently serves as a screening device indicative of the possible need for chemical form measurements. Once the elements and the concentration ranges of interest for each have been defined, the criteria which must be invoked in the selection of

ty, selectivity, reliability, scope, sample prepara- tion requirements, and time-cost considerations. It would be folly to select an approach incapable of providing measurements a t or below the re- quired concentration (threshold effect) levels. Thus, selection of a sufficiently sensitive method is of paramount importance. Although it is com- +7 ’+7 mon to discuss sensitivity in the context of the term detection limit, it is also usually impossible to obtain a sufficiently accurate concentration measurement when said concentration is barely detectable. Thus, sensitivity should be considered in association with the term determination limit, i.e., the lowest concentration at which a suffi- ciently reliable concentration measurement can be carried out. The scope of an analytical approach is defined on several bases. The ability to determine reliably large numbers of individual chemical constituents each of which may be present in a variety of sam- ple types within a broad concentration range is the primary connotation, Thus, universal applica- bility (utility) is an alternate terminology implying the ideal case. It can also be argued that an ana- lytical method which requires simple sample preparation operations offers greater scope than one requiring more complex preparative steps: the ideal case involves direct analysis of the samples without prior chemical treatment. Analytical selectivity is usually interpreted synonymously with specificity. The analysis ap- proach used must provide an unequivocal means of identifying each chemical constituent of inter- est irrespective of the compositional characteris- tics of the sample material being analyzed. In ad- dition, the measurement of the concentration of a n y constituent of interest should provide accu- rate results independent of the variations in the concentrations of other constituents present in the sample materials. Analysis methods that do not satisfy these conditions lack selectivity (or 198 ● Environmental Contaminants in Food specificity): such methods are subject to what are selections are often economic errors when consid- often termed interference or matrix effects, In ered on a longer term basis, At the other extreme, essence, reliability (accuracy) is the criterion management may invest in expensive facilities which defines a sufficiently selective analysis; and expertise and then require that they be used confidence in the results and the decision(s) based to carry out analyses that can be accomplished on them is explicit. more economically via another approach. This, Minimization of the time-cost commitments too, can prove to be a false economy. must be an objective considered in the selection of an analysis approach. It is unfortunate that this Summary consideration often leads to one of two extreme stances. Management often opts for the adoption This background discussion has been pre- of analytical technology of limited utility on the sented as a preface to the evaluations presented basis of lower initial capital costs and/or the fact below; the intent has been to provide a common that such technologies frequently can be applied basis for comments, evaluations, and suggestions by analysts with lower levels of expertise. Such which follow.

PREMISES FOR SELECTION OF LABORATORY FACILITIES

In considering the operation of a laboratory samples in an “as received” form may be consid- dedicated to monitoring food and water supplies ered ideal. Sample preparation operations are for toxic elements, several premises may be es- also primary sources of contamination or loss of tablished. These are essential in defining the nec- the analytical constituents. The direct analysis, essary facilities and require discussion. The fol- minimal preparations, capability is desirable lowing list is not necessarily all inclusive nor is it from the accuracy standpoint as well. Finally, the set down in order of importance. accuracy requirement indicates the need for a 1. The majority of analysis requests will re- highly specific analysis approach which is not quire the determination of several elements subject to significant interference problems and per sample. the maintenance of a quality assurance program. 2. To be most effective, the laboratory must be No single analytical technique will necessarily able to comply with these requests in a rea- satisfy all of these requirements for the elements sonable (short ) time. and/or sample types of interest; a combination of 3. The analyses carried out should be accom- techniques will be required. Properly selected, plished at reasonable costs. the techniques used may be complementary in 4. The analytical results must be sufficiently terms of providing the range of elemental anal- accurate to avoid challenge of the integrity yses required and in terms of providing redun- of any decisions based on them. dant analyses for some elements. The latter will The combination of these premises clearly im- be useful for accuracy validation purposes (see plies that the ideal laboratory facility would be discussion below). one capable of accurately analyzing for all con- In effect, these premises lead to the defensible stituents requested at concentration levels down conclusions that: 1) the laboratory facilities to and below their respective threshold effect lev- should be primarily comprised of multicomponent els in a short period of time. The use of a system analysis systems; 2) more than one such system capable of simultaneously measuring all constitu- will likely be required: 3) the selection of systems ents of interest in each sample is definitely im- that will provide some redundant analytical in- plied. The time and cost premises further imply formation is desirable; and 4) the inclusion of the desirability of utilizing analysis techniques secondary analytical systems to be used in sup- which do not require extensive sample prepara- portive capacities may be essential. The following tion operations: the ability to directly analyze evaluation is predicated on these bases. Appendix H—Analytical Systems for the Determination of Metals in Food and Water Supplies ● 199

EVALUATION OF ANALYTICAL TECHNIQUES FOR THE DETERMINATION OF TOXIC ELEMENTS

The qualitative identification and quantitative Table H-1 .—Determination Limits for Neutron determination of toxic or potentially toxic ele- Activation Analysesa ments in food products and water can be based on — Concentration several different analytical techniques for which c Element value, ppb commercially available instrumentation exists. Ag ...... 0.005 The primary techniques to be considered include: As ...... 0.05 1. neutron activation analysis (NAA); Be ...... 2.5 2. molecular absorption and fluorescence spec- Bi 25 trophotometry: Cd : : : : : : : : : : : : 2,5 3. solids (spark source) mass spectrometry Co ...... 0.25 Cr. ., 50 (SMS); Cu 0.05 4. atomic absorption (AAS) and atomic emis- Fe ...... 2,500 sion spectrophotometry (AES); Hg . . 0.5 5. electrochemical techniques Mn ...... 0.003 Mo . . . . . 5 —anodic stripping voltammetry (ASV), Ni 2.5 —differential pulse polarography (DPP); Pb : : : : : : : : : : : : 500 6. plasma emission spectrometry (PES); Sb ...... 0.25 7. X-ray techniques Se ...... 250 —X-ray emission spectrometry (XES), Sn ... . 25 Te. . . 2.5 —proton-induced X-ray emission spectrom- TI ., . — etry (PIXE). V ...... 0.05 Each of these are considered below. Zn...... 5

aData taken from R- K Skogerboe and G H Morrison, Trace Analys!s E ssen Neutron Activation Analysis tlal Aspects, In Treatise on Analy~lcal Chernls(ry (1 M Kolthoff and P J El vlng, eds ), New York Wiley and Sons, 1971, pp 5842-5843 bNanograms of element that must be present In sample activated to Permit a This technique is perhaps the most sensitive of quantitative determlnatlon, detection Ilmlts are approximately a factor of 5 lower Interference free measurement condt!lons are assumed those available when all elements are considered. = cAss\, mlng a 10 gm samp Ie IS act Ivated, thus concen trat Ion values absolute Absolute determination limits, expressed as nano- values – 10 These values should be Increased by a factor of 100 to 1 000 !O compensate for the 10ss of neutron flux I f a neutron accelerator were used I n grams (1 x 10-(J gm) that must be present in the stea[j of a nuclear reactor sample for quantitative determination, are sum- marized in table H-1. Such determination limits To achieve the sensitivity required, the use of a should be converted to actual concentrations to nuclear reactor providing a high flux of thermal lend them perspective for the present evaluation. neutrons is required. The acquisition cost of the For the analysis of food products or water, a 10- reactor is several million dollars; the operational gm or 10-ml sample generally represents a rea- costs are also comparatively high. Considerable sonable upper limit on the sample size that can be reduction of these can be achieved by replace- activated for analysis. Thus, taking 0.5 ng as the ment of the reactor with a neutron accelerator. determination limit for arsenic as an example The neutron fluxes available with such accelera- (table H-1) this element can be determined at ap- tors are, however, about a factor of 100 to 1,000 proximately 0.5 ng/10 gm = 0,05 ng/gm (rig/ml for less than those typical of a reactor. The analytical water) or 0.05 parts per billion by NAA. Evalua- sensitivity available is reduced in proportion [see tion of the other elements of interest on this prox- footnote to table H-1). imate basis indicates that most could readily be If chemical separations are carried out after determined at required concentration levels via activation of the samples, proponents of NAA this technique. While this is an encouraging con- argue that analytical interferences can be virtu- clusion, there are other factors which detract ally eliminated. This argument is often specious from it. simply because the use of chemical separations is 200 ● Environmental Contaminants in Food often a primary source of errors. Since such sepa- for the formation of the absorptive and/or fluores- rations, even when free of errors, are time con- cent products, While such reactions are generally suming and add to the expense; activation ana- available for the elements of interest, the analysis lysts usually prefer what is called the purely in- of any one sample for several constituents would strumental approach. necessarily have to rely on carrying out several Two types of interferences are common to this individual reactions. Even then, there are only approach. Direct interferences occur when two limited instances for which a reaction will occur or more sample constituents emit radiation (gam- for only one constituent (a specific reaction). Most ma rays or beta particles) of nearly the same en- reactions involving a particular reagent set tend ergy. If the emitting species undergo radioactive to occur for each of several elements having simi- decay at significantly different rates, correction lar properties and, as a result, their absorption of for these interferences can be based on measure- fluorescence spectra tend to be quite similar. ment of the radiation at different times. Other- Such spectra are subject to some degree of wave- wise, the interfering species must be chemically length coincidence such that spectral interfer- separated prior to the measurement step. Indirect ences are not uncommon; the measurements are interferences due to contributions to the sample less specific than desirable, While other features spectrum from Bremsstrahlung and Compton in- of these techniques could be discussed, their teractions are also common, Correction for these chemical reaction requirements coupled with the must be obtained by subtraction techniques. The specificity problem are deleterious from the mul- net effect is, however, a reduction in the analyt- ticomponent analysis standpoint. The techniques ical sensitivity and/or an increase in the uncer- should not be classified as essential for the pres- tainties associated with the measurements, While ent elemental analysis purposes. these instrumental approaches are widely used in NAA to avoid the need for chemical separations, Solids Mass Spectrometry such avoidance still restricts the potential scope of the analyses. They also tend to lengthen the This analytical technique offers sensitivity time required to obtain analysis results. On these competitive with that characteristic of NAA (see bases, NAA does not appear to be the best choice table H-1) for a wide range of elements. To utilize for a laboratory facility: capital costs, operational it, nonconducting samples must be rendered elec- costs, and operator training requirements are pri- trically conducting. The constituents of an aque- mary weighting factors influencing this negative ous sample would ordinarily be analyzed after the judgment. water was evaporated away; the residue would be mixed with a conductor for analysis. Solid Molecular Absorption and materials such as foods are also typically non- Fluorescence Spectrophotometry conductive; by mixing them with a conductor such as graphite or silver powder they become conduc- These techniques have been used extensively tive. Food products must also be oxidized (by wet for elemental analysis. They generally rely on car- or dry oxidation techniques) to destroy the organ- rying out a reaction of the element of interest with ic constituents which cause serious spectral in- 2 a reagent (or series of reagents) to form a product terferences in the analysis step. These sample which has properties required for the absorption preparation steps are quite extensive and can ob- of light. Identification of the element incorporated viously be the sources of serious errors unless into the light-absorbing product is based on the carried out with caution. wavelength of light absorbed while measurement In practice, the analyses must be based on two of the concentration relies on the extent of ab- sample mixtures to obtain a complete analysis sorption; hence the term molecular absorption, subject to less interference problems. Elemental Some light-absorbing species regain a more stable subset A may be determined based, for example, energy configuration by release of the light en- on mixing the sample residue or ash with high- ergy absorbed as light (fluorescence or phosphor- purity graphite. However, the carbon polymers of escence). Measurement of the wavelength at which graphite is composed are observed in the which this occurs and the intensity of the light mass spectrum and preclude the possibility of emitted is used to identify the species responsible analyzing for those elements that would normally and the amount present, See the following” reference for [i more complete discussion The use of these techniques requires that spe- of the prohlems tind r;]p{~bilities: C. A. Evans and G. H. Mor- cific or semispecific chemical reactions be used rison, Anal. (;hem., W, 869 ( 1969). Append/x H—Analytical Systems for the Determination of Metals in Food and Water Supplies ● 201

be measured at these mass positions. The analysis Atomic Absorption and for elemental subset B would subsequently be Emission Spectrophotometry based on the use of high-purity silver, gold, cop- per, or aluminum powder as the conducting ma- Recent instrument sales figures indicate that trix. Again, the use of any of these choices to only gas chromatography is more widely used achieve sample conductivity results in spectral in- than atomic absorption spectrophotometry. When terferences which necessitate analyses based on a flame is used as the energy source required to two or more subsets of elements. The conductive produce the gaseous atomic populations from the materials used must be high purity, i.e., 6-9s or sample dissolved in aqueous solution, atomic ab- 99.9999-percent pure. This requirement limits the sorption offers favorable analysis capabilities for possibilities and affects their costs, Finally, it a reasonably impressive array of elements (see should be noted that solid mass spectrometric table H-2]. Atomic emission from the same atom analyses require unusually long times even populations in the flame is totally complementary though it is possible to obtain analyses for -40 to and supplementary to atomic absorption. Factors 60 elements on each sample. Although the sample which affect absorption also tend to affect emis- preparations required are time consuming, the sion; the ultimate sensitivity achieved with either analysis itself is also rather slow. Given samples measurement approach is limited by the ability to ready for analysis, a well-organized SMS labora- produce the atomic populations. As a result, it tory would be hard-pressed to analyze more than can be shown that atomic absorption is generally 5 to 10 samples per man-day. These factors, cou- most favorable (on the basis of sensitivity) for the pled with high acquisition costs (~$250,000) and determination of those elements requiring more high operational and instrument maintenance energy to produce atomic emission, i.e., an excita- costs, place solids mass spectrometry in a nega- tion potential above approximately 4.5 electron tive posit ion relative to other possibilities, volts (eV). Elements with lower excitation poten-

table H-2.—Determination Limits for Flame Atomic Absorption and Emission and Furnace Atomic Absorption’

b Flame methods Flame AA Furnace Furnace AA Atomic Atomic analysis of atomic analysis of Elements absorption emission tissue digestsc absorption tissue digestsc Ag 002 0.04 0.2 000004 0.0004 As 05 5 5 0.02 02 Be 001 0.5 0.1 00002 0002 Bi 02 2 2 0.001 0.01 Cd 0005 2 0.05 0.00002 0.0002 co 003 0.1 0.3 0.001 0.01 Cr 002 002 0.2 0.001 0.01 Cu 001 005 0.1 0.001 0.01 Fe 0.02 0.1 0.2 0.0006 0.006 Hg 10 2.0 10 0.02 0.2 Mn 001 0.02 0.1 0.0001 0.001 Mo 02 05 2 0,008 0.08 Ni 002 0.1 0.2 0.002 0.02 Pb 0.05 05 0.5 0.001 0.01 Sb 05 2 5 0.006 0.06 Se. 05 10 5 0.02 02 Sn 01 2 1 0.01 0.01 Te 05 200 5 005 05 TI 01 0.1 1 0.0006 0.06 V. 01 0.05 1 002 02 Zn 001 0.2 01 0.00002 0.0002

‘Data I rom J D VJI n~ fordner el al APPI SPec trosc Rev/e ws 712J 147 (1973) A II va I ~jes g lven In P 9 m I for aqueo~ls sol ~~ tlons tlDeflne{j for ~)e~t flafllf> con(jltlons for each element values In l,q ml ~ppm~ r Based on [1 I SSOIVI n q 10 q m o f h(r]locjlcal I I s$ue I wet welg ht I per 100 m I of dc Id va Iues I n Ig g m (ppm) wet wmg h I d~ef, nP(f for ~lDt, n),)” f”, “ac e ~ ~ncj, 1,Ons , n ea~ h case and “se of a 25,, I sample lnjec I Ion values In ,,g m I (ppm) 202 “ Environmental Contaminants in Food tials are best determined by atomic emission ance heating have been developed to high levels when flames are the energy media. These gener- of refinement. Liquid samples are delivered to a alizations require the use of an instrument that is furnace, which is resident in the optical path of well-designed for both absorption and emission an AA unit, and a programed heating cycle is ini- measurements (two of the three major U.S. atomic tiated, In this cycle, the liquid is first evaporated absorption manufacturers supply such instru- at ~120° C; the salt residue remaining is “ashed” ments). to convert it to a “common” chemical form at 0 In spite of the combined capabilities offered by -3000 to 500 C; and the ash is rapidly vaporized flame emission and absorption, they are frequent- and atomized at -2,0000 C for the absorption ly inadequate for the types of analyses in question measurement. The use of such systems provides unless preconcentration and/or separation proce- improvements in the analytical sensitivity for dures (e.g., solvent extraction) are used. These in- most elements (see table H-2). The extent of im- adequacies may be due to a lack of sufficient sen- provement is limited by the amount of sample that sitivity and/or to the occurrence of interference can be placed in the furnace ( ~25 to 50 µ1) but effects. These, coupled with the difficulties at- generally amounts to a factor of 10 to 100 when tendant to separation/preconcentration proce- compared with flame AA capabilities. dures, have led to the development of what is These capabilities combined with the general often called nonflame atomic absorption. The first simplicity of operation and lower instrument costs significant development in this area involved the have been largely responsible for the widespread reduction of mercury ion in aqueous solution to acceptance of atomic absorption. During the prin- atomic mercury (Hg) such that it could be carried cipal time of AA development, the primary com- directly to the gas phase into the optical path of peting techniques were flame, arc, and spark- an AA instrument for measurement. The meas- emission spectrometry. The last two were rather urement actually involves separation of the Hg quickly abandoned by the analytical spectrometry from the sample under conditions less subject to community because they were “notably subject to interferences. If carried out under appropriate interference problems associated with the vapori- conditions so that the Hg arrives at the measure- zation-atomization system. It is rather ironic that ment cell rapidly, an effective preconcentration is the same community, in less than 10 years, re- also realized. In effect, the measurement sensi- verted to the use of furnace AAS systems which tivity is determined by the concentration of atoms are subject to the same interference problems for delivered to the measurement cell per unit time. the same reasons. It is also ironic that an exten- The nonflame methods are all designed to take ad- sive fraction of the furnace AA research reported vantage of this thereby enhancing the ability to in the past 5 to 10 years has dealt with the study analyze at lower concentrations; methodological of interference effects and means for their elimi- developments for this purpose have taken two nation. A large percentage of these studies are general tacts, i.e., chemical generation and fur- reaching the same conclusions and developing the nace vaporization. same compensatory methods that resulted from Beyond the Hg method described above, the arc and spark spectrometry interference effect hydride generation method for the determination studies before the advent of atomic absorption. of arsenic, selenium, germanium, lead, tellurium, Nevertheless, atomic absorption analysis is well- tin, antimony, and bismuth has received wide at- established and here to stay. The furnace meth- tention. The hydrides of these elements are rapid- ods, in particular, clearly offer the required sen- ly formed by reaction in acid media with sodium sitivities for a wide range of analysis problems. borohydride. The metal hydrides, being gases at A primary historic limitation has been that ambient temperatures, are readily transported atomic absorption has been basically a single ele- via carrier gas to a flame or a heated ( - 800° C) ment analysis technique; analyses for several ele- quartz cell for atomization and measurement. The ments in a sample are carried out sequentially in success of this general approach has led to the time. The emergence of plasma emission spectro- commercial availability of hydride generator ac- metric systems which allow the simultaneous cessories for atomic absorption. It has also been analysis of several elements (to be discussed adopted in commercial autoanalyzer systems. De- below) has forced the atomic absorption commu- tection limits of less than 1 ng metal/ml (1 ppb) nity to the development of multicomponent anal- solution are readily obtained. ysis systems. Only one commercial instrument of- Furnaces, fabricated from graphite or tung- fering this capability for more than two elements sten, that can be temperature ramped by resist- is available at this date, This unit is, in fact, a se- Appendix H—Analytical Systems for the Determination of Metals in Food and Water Supplies ● 203 quential analysis system since the elements of in- improved. A variety of these differential pulse ap- terest are not measured simultaneously. Its use proaches have been developed to achieve analyti- saves analysis time but not as much as it would in cal measurements at concentrations 10 to 1,000 the simultaneous mode. At least one other manu- times lower than previously possible. The tech- facturer will introduce a truly simultaneous multi- niques which appear to offer the most significant element AA analysis system in the next 1 to 2 capabilities are differential pulse polarography years. Prototypes of such a unit have been devel- (DPP) and differential pulse anodic stripping volt- oped and tested at Colorado State University.3 The ammetry (DPASV). At least one commercial in- tests have shown that sets of 5 to 10 elements can strument offering both of these measurement cap- be determined simultaneously without sacrifice of abilities is available. analytical sensitivity. The availability of such in- The electrochemical literature indicates that strumentation will advance the state of the art for -16 to 20 elements can be determined by DPP, atomic absorption. DPASV, or cathodic stripping voltammetry (CSV). The inclusion of atomic absorption in an ele- This is true, but all such elements cannot be de- mental analysis laboratory may be considered termined under a single set of experimental con- worthwhile as a secondary facility at least. If said ditions. Some must be determined using a gold instrument offers the simultaneous measurement working electrode (for example) while others re- capability, the inclusion may be justified on a bet- quire the use of a mercury electrode. Some deter- ter economic basis. minations require the use of a specific supporting electrolyte solution while others do not. Finally, Electrochemical Techniques the redox potentials of some entities are suffi- ciently similar that they cannot be determined in These techniques are perhaps the most classi- the presence of each other without the use of cor- cal of those considered herein, And yet, electro- rection techniques. In some cases, prior separa- chemistry has been reborn in the past two dec- tions are required. ades largely through the development of what The electrochemical analysis literature indi- may be classified as pulse or differential tech- cates that Cd, Cu, Pb, Zn, Tl, and Fe can usually niques. The electrochemical techniques of pri- be determined simultaneously in a single support- mary interest in the present context rely on the ing electrolyte solution using mercury as the measurement of the Faradaic current (FC) pro- working electrode material, These elements can duced or used during oxidation or reduction reac- be determined at levels of 1 ng/ml or less in the tions involving the element to be determined; the electrolyte solution by DPP or DPASV. As a result, potentials at which FC changes occur depend on these techniques are quite widely used for the the elements (ions) involved in the reactions. Such analysis of the above elements. Although Cu and reactions are almost universally accompanied by Zn interfere with each other via formation of the non-Faradaic processes involving other sample Cu:Zn intermetallic in the mercury electrode and constituents which also result in the production the reduction of Fe(III) to Fe(II) interferes with the or utilization of current. In classical direct cur- Cu determination, these can be corrected via use rent polarography or voltammetry, the ability to of expedient instrumental procedures. Other ele- measure a low Faradaic current (low analyte con- ments frequently determined are As(III) by centration) is limited by the magnitude of the non- DPASV or DPP and Se(H) by CSV. Again, analyses Faradaic current changes occurring simultane- at or below the part per billion concentration ously. The revitalization of electrochemistry has level are common. been largely based on the fact that when the po- The electrochemical methods detect only the tential required to induce a redox reaction is re- electroactive species, e.g., the ions, This state- moved, the non-Faradaic current decays more ment must be qualified in terms of the time-scale rapidly than the Faradaic current. Thus, by of the measurement step. To illustrate, consider “pulsing” the potential up to the reaction level re- an electroactive metal ion (M+ 11) which may be quired, shifting it back to below the reaction level, present in the sample solution primarily as a com- and waiting an appropriate time period for the plexed or molecular species designated, for exam- + n non-Faradaic current to decay; the Faradaic to ple, by MXn. To measure the total M plus MXn non-Faradaic current ratio can be significantly concentration, the following reaction must occur either prior to or during the measurement period: -, M + n + nX - MXn ‘See for example: F, S. Chuang, D. F. S. Natusch, and K. R. If the duration of the measurement is short, the 0’Keefe, Anal, Chem. 50, 525 [1978). above reaction must go to a reproducible state of 204 ● Environmental Contaminants in Food

completion within that time period; in such cases ence effects due to matrix differences may be the complexed/molecular entity may be classed as eliminated or reduced. As a result, a single set of labile. Thus, there is a growing investigative ef- operational conditions may be used for the simul- fort in the electrochemical community involving taneous determination of 20 to 60 elements, the use of this conceptual approach as a means Quantitative determination limits for the ICP- for the identification of the chemical forms of and DCP-Multielement Atomic Emission Analysis metals in natural systems. The general thrust of Systems (MAES) are listed in table H-3. Examina- these efforts involves either the use of chemical or tion of these data indicates that these systems are electrochemical means for systematically shifting adequate for the simultaneous determination of a the above type of reactions toward the formation major fraction of the elements of interest at levels of an electroactive species for identification pur- commensurate with the anticipated threshold ef- poses. A typical electrochemically induced shift fect concentrations. This is one reason why the experiment, for example, might involve ASV. By ICP-MAES and DCP-MAES manufacturers have electrodepositing from the sample solution for enjoyed significant annual sales improvements successively longer periods and measuring the over the past 5 years. amount of M + I’ reduced during each period, the Some elements of high concern, because of time required to drive the above type of reaction their toxicities or propensities for bioaccumula- to completion might be deduced. This may be in- tion, cannot be determined at low enough concen- dicative of the “lability” of the complexed/molec- trations by direct solution analysis, e.g., Hg, As, ular form, i.e., its thermodynamic or kinetic stabil- and Se. However, the hydride generation ap- ity, Extensive research must be completed to es- proaches used to solve this problem when AAS is tablish the potential utility of such approaches. the analysis method, are equally applicable to However, these possibilities combined with the MAES. In fact, all hydride-forming elements can high sensitivities that can be achieved with mod- ern electrochemical techniques suggest that such facilities should prove to be valuable laboratory Table H-3.—Multielement Atomic Emission facilities. (See further discussion below.) Determination Limits for Two Common Plasma Excitation Systemsa Plasma Emission Spectrometry ICP-MAES DCP-MAES It has been suggested that atomic emission For tissue Tissue spectrometry offers what may approximate the In solution, digest in solution, digest ideal approach to multielement analyses. Indeed, ~lglml ~lg/gm (ppm)b ~~g/ml g/gin (ppm)b the use of the radio frequency inductively coupled 0.01 0.1 — — plasma (ICP) as the atomic excitation source in 0.3 3 0.2 2 0.02 0.2 0.02 0.2 combination with a direct-reading emission spec- 0.002 0.02 0.05 0.5 trometer permits the simultaneous determination 0.2 2 — — of numerous elements at low concentration lev- 0.01 0.1 0.05 0.5 els.4 Similarly, the use of a direct current plasma 0.01 0.1 0.08 0.8 (DCP) excitation source for direct-reading emis- 0.02 0.2 0.02 0.2 0.005 0.05 0.01 0.1 sion spectrometry shows comparable promise. 0,01 01 0.01 0.1 The evolution of such systems over the past dec- 0.1 1 0.01 0.1 ade has brought emission spectrometric analysis 001 01 0.05 0.5 back to the forefront of analytical capabilities. 0.03 0.3 0.05 0.5 0.05 0.5 0.05 0.5 Such plasmas are principally used to excite ana- 0.1 1 0.05 0.5 lytical constituents delivered to them via solution 1 10 — — nebulization (aerosol production) systems. The 0.2 2 0.2 2.0 conversion of solid samples to aqueous solutions 1 10 1 10 0.4 4 — — for analysis (e.g., by wet oxidation) results in 1 10 — — elimination of many of the major compositional 0,2 2 0.5 5 differences between samples such that interfer- 0,02 0.2 0.02 0.2 0.01 0.1 0.02 0,2

4 good introductory review has been presented by V. A. aData for the Icp trom R K Wlncje et al S~ectrochlm Acta, 326 327 ( 1977) A Data for the DCP from R K Skog;rboe H E Taylor and G W Johnson, Spec- Fassel and R. N. Kniseley, Anal. Chem, 46, 11 10A (1974). See trochlm Acta, In press also, %ience, 202, 183( 1978). bFor 10 ~m O( biological Ilssue (wet wt ) dissolved In too ml of Solutlon Appendix H—Analytical Systems for the Determination of Metals in Food and Water Supplies ● 205 be determined at required concentrations on a si- available. Such energy-dispersive detection sys- multaneous basis. (Tissue detection limits of 0.01 tems have rather rapidly replaced the convention- µg/gm (ppm wet weight) can be achieved. ) al wavelength-dispersion units. The production of Although the measurement of atomic emission X-ray fluorescence (emission) by X-ray bombard- offers a high degree of qualitative specificity, ment and measurement with an energy dispersive there are two common types of interferences detector offers reasonable analysis capabilities which affect the quantitative specificities of such for several elements of interest (see table H-4). systems. Both plasma types are highly efficient In 1970, the potential of heavy, charged parti- excitation media. As a result, extremely intense cles for X-ray excitation was recognized, and im- radiation from common elements such as Ca and proved capabilities have been demonstrated. The Mg is delivered to the spectrometer-measurement capability improvements of accelerator (particle) system, The result is the observation of stray light beams are due to: 1) the high particle fluxes avail- interference effects for some other elements. The able, Z) the relatively low background radiation magnitudes of these effects depend on: the design associated with the excitation process: 3) the fact characteristics of the spectrometer used, the con- that the excitation cross-sections of many ele- centrations of the elements from which the stray ments for particles are higher than for photons or light originates, the concentrations of the ele- electrons, and 4) a single charged particle can in- ments subject to the interferences, and the types duce emission of several X-ray photons as it pene- of approaches used to alleviate the effects. An ex- pedient means of correcting for these effects may be based on determination of the concentrations of the causative elements. To illustrate, let Cm represent the measured concentration of a partic- ular element (analyte) subject to interference due to a concomitant constituent present at concen- tration C(. The true (corrected) concentration (C1) of the analyte may be determined from: where a is a correction coefficient determined by simple experimental procedures. The other type of interference effect involves interelemental processes in which the presence of one constituent changes the extent of excitation of another in the plasma, While such interelement effects are less common, the above correction procedure can also be used for compensation. Commercial ICP- and DCP-MAES systems are rou- tinely equipped with minicomputer or microproc- essor systems for control, data acquisition, and data correction purposes. The use of the types of corrective procedures described above is, thus, easily automated. In effect, the general capabilities of plasma- MAES are such that it should be considered a pri- mary facility.

X-Ray Emission Techniques

The bombardment of samples with X-rays to produce X-ray fluorescence (XRF) has remained in wide usage. In fact, the development of the lithi- um drifted germanium, Ge(Li), or silicon, Si(Li), de- tectors has advanced the status of this approach by the reduction of spectral interference prob- lems and a general increase in the sensitivity 206 ● Environmental Contaminants in Food trates the sample. As a result, proton induced X- lytical substrate, Ions in solution may also be pre- ray emission analysis (PIXE) has rapidly emerged concentrated for analysis using filters impreg- as a sensitive analysis approach which may be nated with ion exchange resins. ’ Determination subject to fewer interference problems than X- capabilities for XRF and PIXE are listed in table ray induced emission. Thus, the X-ray source is H-4, replaced by a van de Graaff accelerator to pro- The general capabilities of these techniques duce proton beams in the 2.5 to 3.0 Mev energy are such that they can be used in a laboratory of range. the type considered. It should be emphasized, A potentially significant capability associated however, that quantitation of the measurements with PIXE analysis is that of direct analysis of bio- is subject to difficulties particularly for direct logical tissue sections or blood. Recent publica- measurements. Proponents of the techniques tions5 6 have shown that Cl, K, Fe, Cu, Zn, Pb, Se, argue that these problems can be readily over- Br, Sr, and S could be directly measured in a 30 come; others (cynics??) argue that this will re- µm thick section of human kidney; that the same quire extensive development efforts. The truth elements plus Mn, Ni, Hg, Rb, and Zr could be de- appears to be intermediate between these ex- termined in a thick section of carp muscle; and tremes. The ultimate decision to include the X-ray that several elements can be directly measured in capabilities in the laboratory facility should prob- tissue sections of liver, kidney, lung, and bone. ably be based on the essentiality (desirability) of The analysis of liquid- or wet-digested materials being able to analyze solid samples (e.g., tissue may be based on evaporation of the liquid phase thin sections) directly for several elements of po- to leave a residue deposit on an appropriate ana- tential interest. This capability may be consid- ‘lJ. L. Campbell et al. “Acivances in X-ray Analysis, ” vol. 17 ered by some to be advantageous simply from the (C. L. Grant et al,, eris.), Plenum Press, New York, 1974, pp. semiquantitative screening standpoint. 457-466. ‘P. S. ong et al., [bid., vol. 16 (L. S. Birks et al., ecis. ), 1973, S. L. Law and W. J, Campbell, “Advances in X-ray Anal- pp. 124-133. ysis, ” vol. 17, Plenum Press, New York, 1974, pp. 457-466.

PREPARATION OF (BIOLOGICAL TISSUE) FOR ANALYSIS

Many techniques require that solid samples be volatilization, convection, and retention losses is converted to solution form for analysis. The litera- higher, however, unless ashing conditions are ture on methodology for sample decomposition is carefully controlled, These risks have led to the immense, The procedures cited find both wet- fairly widespread adoption of low temperature (acid digestion) and dry-(ashing) oxidation meth- ashing (LTA) based on reaction of the sample of ods extensively used. Dry-ashing methods are usu- the oxygen free radical generated via a radio fre- ally implicated when problems with recovery or quency field. While it is true that materials can be losses of analytical constituents are reported. In converted to oxides under these conditions, it is comparing wet- and dry-ashing methods, the pau- also apparent that several elements may still be city of data specific to real-life samples makes it lost by volatilization at typical operational tem- inappropriate to state categorically that one peratures of 100° to 150° C. These are largely ele- method is superior to another. Some generaliza- ments with a tendency to form volatile chlorides, tions can, however, be made on the basis of proce- oxychlorides, or hydrides, Thus, the extent of dural differences. Wet oxidation has the advan- their loss may be particularly influenced by the tage of requiring a minimum of apparatus and is halogen (chloride) content of each sample, The less prone to volatilization and retention losses use of programed-temperature ashing furnaces than dry-ashing. Wet-ashing may suffer in that has been shown to be effective in preventing or relatively large amounts of reagents having signif- reducing volatilization losses. Raising the temper- icant levels of contamination may be required and ature at a rate permitting slow charring and oxi- contact with glassware may account for a higher dation of the sample is to be highly recommended. risk of contamination than dry-ashing. Dry-ashing Nitric acid is a universally used wet-oxidant, requires few, if any, reagents and handling of The azeotrope boils at 120° C, a factor which as- larger samples presents less problems. The risk of sists in its removal after oxidation but also limits Appendix H—Analytical Systems for the Determination of Metals in Food and Water Supplies ● 207 its effectiveness in completing the oxidation proc- tion rapidly and prevents bumping and frothing. ess. The most effective medium for wet oxidation Recovery studies run on a range of elements sug- is a mixture of nitric and perchloric acid; the aze- gest that loss problems are minimal. Bovine liver otrope boils at 180° C to force oxidation. Extra- is notorious for being difficult to wet digest. The ordinary care to avoid explosions and fume hoods report10 that 2 gm (wet weight) or 1 gm (dry is required. The precaution of keeping sufficient weight) of beef liver can be digested in 10 ml of ni- nitric acid present until all easily oxidized materi- tric/perchloric acid in 2 to 3 minutes indicates an al is gone (cessation of brown fumes) is particu- attractive capability that will surely guarantee larly important in wet digestion of tissues having extensive use of the microwave system. lipid (fat) contents. Because the acids used are For any of the analysis techniques which re- never absolutely free of metal contamination, the quire sample dissolution, the approximate upper objective of wet oxidation should always be to limit of “dissolved salts” that can be tolerated is 2 complete the oxidation with the smallest possible percent (wt/vol). Thus, for a wet biological tissue amount of acid. This minimizes the blank problem. which yields 10 percent ash, this upper limit The use of reflux digestion apparatus equipped would be equivalent to dissolving 20 gm of the with condensers is focused on this objective as tissue (wet wt) per 100 ml of analytical solution. well as that of minimizing losses. Recent systems Working with half that weight per 100 ml would which rely on microwave ovens for heating the have practical advantages. acid-sample mixture also show considerable promise. 8 9 The microwave system heats the solu- “A. Abu-Samr~, ]. S. Alorris, ~n[j S. R. Koirtyohann, Anal. Chem., 47, 1475 (1975]. “’Report to the U.S. FDA (Contract No. 223-75-2268] Der. 2, ‘IU. S. Patent No. 4,080.168 (Mar. 21, 1978). 1976.

ASSURANCE OF ANALYTICAL ACCURACY

Quality assurance in an analysis laboratory re- information but all should be included in the pro- fers specifically to the question: Are the analyt- gram. Moreover, any of the above tests should be ical results valid (accurate or reliable]? Obtaining carried out incognito. Otherwise, the analysts the answer is complicated by the fact that all may be tempted to devote inordinate attention to analyses are subject to random (indeterminate) the check analyses. errors and may also be subject to systematic (de- All of these methods add to the expense of lab terminate) errors. Errors are cumulative; those operation. It is necessary, however, for the lab which characterize the accuracy of any analysis personnel to be acutely aware of the ways in result may include the composite contributions of which they or their analytical methods can fail the random and systematic errors inherent in the and (ideally) know when they have failed. Other- analysis method(s) used and those characteristic wise, their products—the analysis results—will of each analyst, Laboratory, or set of equipment be subject to challenge. Experience with the legal involved in the analysis. To maintain quality as- process suggests that the agreement between re- surance, an operating laboratory must base its sults obtained by two or more independent analyt- program on some combination of the possible stra- ical methods is a primary indicator of success (ac- tegic approaches. These include: curate analyses). Thus, the planned inclusion of 1. Recycling of submitted samples to obtain analysis redundancy (5 above in particular) in the cross check analyses, laboratory operation is to be recommended; a typ- 2. Recycling of certified or secondary standard ical quality assurance program would involve reference materials to estimate analytical check analyses amounting to 5 to 10 percent of accuracy, the total load. 3. Spike-recovery studies to estimate accuracy, 4. Participation in collaborative test programs, and ‘ I An expanded discussion of quality assurance programs and 5. Comparison of results obtained by independ- the necessity thereof is given in the paper by R, K. Skogerboe ent analysis methods. and S. R. Koirtyohann, 4’Accuracy Assurance in the Analysis of Each of these provides specific types of relevant Environmental Samples.’” NBS Special Bulletin 422 ( 1976). 208 ● Environmental Contaminants in Food

LABORATORY INSTRUMENTATION AND FACILITIES REQUIREMENTS

Consideration of the above discussion indicates tralaboratory, interactive (time-shared) mode and that the majority of the analytical requirements tied to a larger computer system (external to the can be satisfied by plasma emission spectrometry laboratory) which would perform those data man- or by atomic absorption spectrometry with heavy agement functions that need not be carried out on reliance on furnace atomization systems. The in- a real-time (fast response) basis. Systems which clusion of more than one measurement technique use internal, instrumentation-coupled, computers in the laboratory can also be justified on various for control and data acquisition purposes and bases. The advantages, limitations, and tradeoffs transmit the data to central management comput- involved will be considered below. Regardless of ers are presently in operation at the USGS Water which instrumentation facilities are chosen as Resources Laboratory in Denver and the EPA Lab- primary; the space requirements, ancillary facil- oratories in Cincinnati. Although these operations ities requirements, and personnel requirements may not be the best model examples, the concep- are comparable. Thus, discussion of these prior to tual approach which they embody is recom- consideration of the analytical instrumentation mended for the present operation. Further com- facilities is appropriate, ments relating to this are inserted in appropriate sections which follow. Personnel, Space, and Ancillary Sample Receiving and Storage Room—The re- Facility Requirements ceiving operation will necessarily include facili- ties for logging in samples and the associated ana- The general design of the laboratory should in- lytical work requests. A computer terminal clude four physically separated types of space: should provide the most effective capability. The 1) office space for personnel, 2) a sample receiv- storage facilities should include: shelving or cabi- ing and storage room, 3) a sample preparation nets for those samples that can be stored under laboratory, and 4) the analysis laboratory. Work- ambient conditions, a cold room (4°C) for storage ing desk space should be included in the last three of water samples, and a freezer for storage of cer- types of space in addition to the facilities dis- tain biological samples. The size of the room and cussed below. All lab facilities will require tem- storage facilities required will depend on the an- perature control to plus or minus 5° F and humidi- ticipated sample load and variations in the sam- ty control (less than 50 percent). ple submission rate. Since such data have not It is recommended that computer capabilities been supplied, size estimates and costs are not in- should be a primary ancillary facility included in cluded in this report, the laboratory operation. This is based on several Sample Preparation Laboratory—The design of rationales including: this laboratory will be sample type and sample 1. several types of instrumentation likely to be load dependent. If the sample types to be ana- present are most effectively used under a lyzed include those which must be ground in the computer control-data acquisition mode of dry form such that atmospheric contamination operation: can occur, the grinding facilities should be physi- 2. personnel requirements, time commitments, cally isolated from the other sample preparation and human errors can be reduced in a com- operations, In addition, the sample preparation puter-oriented operation; and lab should basically be a clean room12 operation 3. a computer system may be essential as a or it should be equipped with laminar flow, filter data management, quality assurance evalua- hoods in which certain sample preparation opera- tion, warning assessment, and report prepa- tions can be carried out. Both wet- and dry-oxida- ration tool in a laboratory operation of the size likely for the present program. (;lean room or clean hood environments are classified on Although the computer configuration selected will the b[]sis of (x)ntroliing the concentrations of particles (dust) in be dependent on the laboratory purpose(s), the air which rnn cont:lminate the samples by f;~llout. A class-100 analytical facilities installed, the size of the oper- (’le;~n room, for t:xtlmple, must ht]ve no more than 100 particles in the 0.5- to 5- micrometer diameter size range per cubic foot ation, and other subsidiary factors, it is likely a of air. A cI:lss- 1,000 room permits 1,000 pa rticu]a tes per cubic computer acquired with one of the instruments foot in the ahove r:tn,gc. ‘1’o meet these specifications, air filtra- described below could be adapted for use in an in- tion is re(~uire(i. S(;e th[! 1963 Fe(teral Stand[~rd No. 209:1, Appendix H—Analytical Systems for the Determination of Metals in Food and Water Supplies ● 209 tion facilities should be included in the laboratory general area of instrumental trace analysis deal- plan to permit handling of various sample types in ing particularly with the types of instrumentation the most expeditious manner. A list of sample to be used. A working familiarity with basic elec- preparation facilities likely to be required is given tronics and minicomputers would be desirable. with their approximate costs in table H-5. Access The availability of major repair capabilities to a computer terminal to permit convenient entry through instrument maintenance contracts of sample preparation data is also recommended. should be insured. Analysis Laboratory—The instruments se- For the lab personnel, it may be assumed that lected will influence the facilities and size re- the average analysis production hours per man- quirements of this laboratory. It is assumed that day will range between 5 and 6. The remaining the prepared samples will be delivered to this lab time will be utilized for preventive maintenance in closed containers and will be handled in ways of equipment, recordkeeping, cleanup, etc. Al- that will minimize the probability of contamina- though there are nominally 260 working days per tion via atmospheric contact or fallout. Thus, al- annum, this reduces to about 200 days when vaca- though rigorous clean room operation has not tions, holidays, sick leave, and refresher training been recommended for this lab, it will be neces- time are taken into account. The number of per- sary that all analytical operations which can re- sonnel required and the sample throughput capa- sult in simple vaporization be carried out under bilities of the lab must be defined on this basis. ventilation conditions where the discharge is ex- ternal to the lab. Personnel Requirements and Performance— Table H-5.— List of Sample Preparation Stipulation of exact personnel requirements will Facilities Required again be dependent on the sample load and the a types of samples received. The sample logging Preparation facility Approximate cost, $ operation can be handled by an individual with no $2.500 8.000 automatic chemical training but with secretarial or key- 3,000 manual punch skills: this individual might also perform 4.000 some of the simpler sample preparation opera- tions, e.g., weighing and grinding. The prepara- 4.000 tion laboratory should be staffed by individuals 8.000 with B.S. degree training in chemistry with 1 to 3 6,000 years of wet chemical experience.13 One such indi- vidual can typically prepare 15 to 30 biological 8,000 tissue samples per day by wet- or dry-oxidation 4,500 techniques depending on the complexity of the 2,000 preparative steps involved. The analysis lab $5,000-10,000 should also be staffed by chemists with at least B.S. degree training plus 2 to 4 years of experi- ence.14 The experience should preferably be in the

ANALYTICAL INSTRUMENTATION

The previous discussion has presented the gen- H-6. Although the data presented in that summary eral analytical features, capabilities, and limita- have been based on personal experience and dis- tions of the various analytical techniques consid- cussions with other practicing analysts, they must ered applicable to the present problem. A summa- be considered as estimates only. This applies in ry of the cost, space, and throughput features of particular to the cost per analysis because it will these types of instrumentation is given in table be strongly influenced by the applicable salary 210 ● Environmental Contaminants in Food

Table H-6.—Summary of Cost-Productivity Estimates for Various Analytical Techniques

- ————— Approximate - Space Estimated Approximate Data Estimated acquisition requirements no. analyses cost per system down a b d e f Instrument facility cost, $ sq. ft. per man-dayc—————— analysis, $ requirement time, % ICP-MAES. ,-:- ...... ; .-: 100,000 200-400 1,500-2,000 0.1-0.3 Yes” 3-5% DCP-MAES ...... 60,000 200-400 1,200-1,500 0.1-0.3 Yes* 3-5% Flame & furnace AAS Single element modeg . . 25,000 200-400 100-200 (flame) 0.7- 1.0 Desirable 3-5% 50-100 furnace 1.0- 1.5 Desirable 3-5% Multielement modeh ...... 40,000 200-400 400-800 (flame) 0.3-0.5 Yes’ 3-5% 200-400 (furnace) 0.5-0.8 Electrochemical i ...... 25,000 200-400 50-100 1.0- 1.5 Yes* 3-50/0 N. A.A.J. ... . 40-50,000 k 200-400 200-700 ? Yes* ? X-ray fluorescence . . . 60-100,000 200-400 200-700 7 Yes* 3-50/0 PIXEJ ...... , 300,0001 300-500 400-800 7 Yes* ? — aFOr ~o.~~e-ment ~“~1~ SIs Capab;l llles lncludl~g ;strument mstallatlon bin addltlon 10 sample preparation and office space but allowlng for adjacent working sPace. Csamples Previously prepared for analysls, data management system available, 5 hours on the Instrument per8.hour day, one anal Ysls defined as determination of one element per sample, approximately 10 percent of total analytical loadI nvolvlng quail ty assurance assumed, all analyses to Involve simultaneousmultlcomponent determinations of 10 to 15 elements on each sample dcost per element per Sample, :jamples Previously prepared for analysis, approx 5-year instrument depredation assumed, see footnote C above el nstrument control, data acquls.ltlon and data management system considered essential for all facllltles, . Ind!cates Incluslon of a minicomputer or microprocessor of at least 8K memoryIn acquwtlon costs fAn Upper Ilmlt estimate Intenderj to Include preventive mmlenance 9AnalYSeS based on rjeternllnatlon of one element at a time Cost estimate Includes hollow cathode lamps for 30 elements and for autOmatlC sample lnjeCtlOn but nOt for a computerlmlcroprocessor system hAnalySes based on determination of 5 t. 7 elements at a tfme cost estimate Includes hollow cathode lamps for 30 elements, an auton’ratlc sample lnJE!CtOr, and a data acquislt!on and control system ‘Llmlted principally to analyses for As, Cd Cu, Fe, Pb, Se, Tl, and Zn, other metal analyses possible but not widely practiced Isame type of readout facllltles could be used for botf’r tecflnlques Radlatlon protection required particularly for NAA and PIXE systems ‘Does not Include reactor costs

IReduce by -2513000 If surplus van de Graaff accelerator IS available structure, to actual types and numbers of anal- been based on several factors including: 1) lack of yses to be performed, the efficiencies of the em- information regarding sample types and numbers, ployees involved, and the final laboratory design 2) the influences of the preparation procedures configuration, The instrument acquisition esti- actually selected on costs, and 3) the fact that the mates include initial installation and assume that sample prep costs will be essentially the same for the experience of the operating analyst will mini- each of the analytical techniques considered. mize startup time requirements. The estimated Comparisons of the estimates presented indi- number of analyses per 8-hour day assume sam- cate that the plasma emission techniques offer ples have been previously prepared for the anal- superior economic advantages which are comple- ysis; that the instrument operation will rely on mentary to the sensitivity advantages previously computer-controlled data acquisition and man- discussed. Although more analyses per day can agement; that the salary of the operation would be carried out with the plasma techniques, the approximate $20,000 to $24,000 per annum; and present estimates have been based on the expec- that the instrument would depreciate completely tation that it will be necessary to analyze each over a 5-year period. These estimates do not allow sample via two different sample introduction for laboratory refitting or remodeling costs or for methods to achieve the required sensitivity. Thus, overhead costs above the direct instrument oper- several elements may be determined by direct ational costs. Such factors have been excluded nebulization of the aqueous sample solutions. A because the costs involved are highly contingent second set of elements, e.g., As, Bi, Hg, Sb, Se, and on the extent of lab refitting required, the location Sn, may have to be determined using the hydride of the lab, and the wide variations in overhead generation method to achieve the required sensi- charges. It should also be noted that the overhead tivity (low-threshold effect levels). A similar situa- costs will be essentially the same regardless of tion is anticipated for atomic absorption spectro- the lab instrumentation facilities selected. The photometry; direct flame measurements will be footnotes to table H-6 provide further qualifying adequate for some elements while furnace meas- information. urements will be required for others. Analysis in The decision to exclude estimates of the costs the single-element mode for AAS is clearly more of preparing the samples for the analyses has costly. The inclusion of the simultaneous multiele- Appendix H—Analytical Systems for the Determination of Metals in Food and Water Supplies ● 211

ment AAS estimates is based on present experi- rational, The inclusion of an AAS system and an ence at Colorado State University. Conversion of electrochemical (EC) system as support (e.g., single-element AAS units to the multielement cap- cross-check) techniques should be recommended. ability can be anticipated in the near future. The The cost estimates given for the AAS and EU units long-term cost advantages are obvious. in table H-6 were based on the expectation that The restricted capabilities of the electrochemi- each might be the primary lab facility. When their cal system tend to remove it from competition ex- roles are reduced to a secondary (support) level, cept as a supplementary or specialized capability. less sophisticated (high versatility and produc- Primary arguments for the use of NAA or the X- tion) units may be acceptable. The cost of an ade- ray techniques are based on the ability to directly quate AAS facility could thus be reduced by analyze tissue (solid) samples without dissolution. -$8,000 to $10,000 and that for the electrochemi- Unless this capability is important, the cost differ- cal unit by -$5,000 to $10,000. It should also be entials evident in table H-6 argue against their noted that these lower cost units could be inter- use. faced to the ICP-MAES computer system prefer- In view of these data, the use of the ICP-MAES ably using inexpensive ($300 to $500) microproc- system as the primary instrument appears quite essors as buffers.

ESTIMATED ANNUAL SAMPLE THROUGHPUT LOAD

The above estimates of sample throughput cap- 3. The analytical facility could be used for addi- abilities may be used to approximate annual ana- tional types of analyses which do not require lytical productivities. These are based on a total sophisticated (time consuming) sample prep- of 260 working days per annum minus 60 days for aration operations, e.g., water samples. vacations, holidays, downtime, quality assurance Comments on the last two possibilities are ap- time, etc., leaving 200 effective 8-hour days. The propriate. projection below is based on the assumption that the primary load will involve the analysis of bio- Primary factors which affect the sample prepa- logical (organic) tissues; such analyses probably ration time/cost requirements include: the ele- represent the most rigorous time/cost require- ments to be determined and the types of samples. ments, The determination of toxic elements in natural Using the estimate of 15 to 30 samples per day and effluent waters is perhaps the simplest case. as the load that can be handled by a sample prep- The water samples must be filtered and appropri- aration technician, the annual preparative capa- ate added; these operations must bility ranges from 3,000 to 6,000 samples. Analy- be carried out as soon after sample collection as sis of these samples for 30 elements (as an upper possible and preferably in the field so essentially limit example) using the plasma emission capabili- no lab prep operations are required, Such anal- ty involves performing 90,000 to 180,000 anal- yses could occupy the additional time available on yses. Applying the lowest analytical throughput the instrumentation. The preparation of animal (1,500 analyses/day; table H-6) suggests that 60 to tissues (particularly liver) is at the other extreme. 120 days would be required for these analyses These contain varying amounts of fatty materials once the samples were prepared. In essence. this (lipids) which are difficult to decompose by wet emphasizes a fairly universal observation, i.e., oxidation techniques thereby requiring that rigor- preparation of samples for analysis is often the ous conditions be used. Although dry oxidation es- factor which limits laboratory productivity. Sev- sentially circumvents this problem, the fact that eral inferences may be drawn from the above several elements may be lost by volatilization sample preparation and analysis estimates: forces the use of slower oxidative methods, e.g., 1. The analytical capability proposed is suffi- temperature-programed furnaces, low-tempera- cient to keep 2 to 4 sample preparation per- ture ashing units, or sealed [high-pressure) bomb sonnel busy if biological tissues comprise the systems. The estimates given above have been principal workload of the lab. predicated on the use of wet oxidation (reflux) 2 The sample preparation methods used need techniques in common use; these appear to be less to be upgraded in terms of throughput per susceptible to problems than the dry techniques. unit time. It has also been assumed that complete destruc- 212 ● Environmental Contaminants in Food

tion of lipids present is required. An examination ise in alleviating the digestion time problem.15 of the literature, however, has not produced con- Spike-recovery studies carried out when food vincing evidence that this is essential. The point products were digested via this method are very raised is simply: Can the most significant fraction encouraging. The ability to reduce sample diges- (e.g., >90 percent) of the elements of interest be tion times by a factor of two or more appears to be “extracted” from the tissue matrices via wet di- a reasonable estimate. Factors of this nature gestion without complete destruction of the fat, must be considered in the final planning stage for lipids, or cellulose present? If so, the sample prep- the present program. aration times can be prominently reduced. Final- Ad[;] At)u-%mra, J. S. hlorris, and S. R. Koirtyohann, An:il. ly, it should be noted that the wet ashing of bio- Chem., 47. 1475 ( 1975). Set! a]so reports on L?SPH contrnct No. logical materials in microwave ovens shows prom- 223-75-2268 and U.S. P{~tent 4,080,1 t38.

ANALYTICAL ACCURACY COMPARISONS

As emphasized above, the validity (accuracy) of of analytical errors, the errors which may accrue the measurements strongly influences the integri- during the preparative steps will be essentially ty of any decision based on the results, Any meas- the same for all techniques. Second, the analysts urement is subject to errors which may be influ- involved can be responsible for the cause or pre- enced by several factors including the measure- vention of errors depending on the expertise and ment methods used and the analyst responsible. caution they exercise. The analyst that tends to This is the primary reason for stressing the impor- use poor technique or judgment when applying tance of a quality assurance program. The gen- analysis method A will, in all probability, do the eral degree of accuracy that can be achieved with same for method B; the errors for which he is re- analytical techniques discussed above can be an- sponsible will be comparable in both cases. Final- ticipated to be essentially the same for all tech- ly, the measurement accuracies of each technique niques. This first approximation expectation is discussed above are generally similar. As a re- based on the following. First, each of the tech- sult, there are no clear-cut, easily defensible rea- niques discussed essentially requires the same sons for suggesting preference for one approach sample preparation procedures. Since these are over another on the basis of improved accuracy. likely to prove to be significant (primary?) sources

RESEARCH NEEDS ASSOCIATED WITH THE PRESENT PROGRAM

In the section discussing the criteria involved in and 3) what are the operative threshold-effect the selection of an analytical method, the general concentration levels? cannot be considered trivial importance of being able to identify and measure problems. It is unfortunate that simply measuring the chemical forms or oxidation states of several the total amount of a particular element present elements was emphasized. This emphasis was in an ecological system may, in fact, be only a based on the fact that all segments of the ecologi- gross indicator of potential or actual deleterious cal system are affected to varying extents by the effects, What is often needed is the identification elements (e.g., metals) present. Indeed, the state- of the active or functioning forms of the elements ment that “the life processes of every living cell in question. Although numerous examples which are conditioned by the types and amounts of met- support this statement can be cited, the fact re- als present” would be accepted by a majority of mains that chemical form is often extremely im- scientists. Some metals are essential to the health portant. of living systems and, yet, they may also be insidi- ous pollutants because of their generally nonbio- For this reason, the development and refine- degradable nature. Only a few metals are com- ment of analytical methods and techniques for the pletely nontoxic at any concentration level; most identification of chemical form (chemical specia- cause deleterious effects at some exposure level. tion), at the trace to ultratrace concentration The ultimate definition of: 1) what constitutes a levels so often of interest, has received considera- deleterious effect?; 2) what actually causes it?; ble attention in the past decade, Although this is Appendix H—Analytical Systems for the Determination of Metals in Food and Water Supplies ● 213 / clearly an important area of research, progress those elements which may appear to be more im- has been slow for several reasons. These include: portant from the effects standpoint. 1. The analytical techniques most suitable for Technology presently exists for the differentia- the determination of chemical form are also tion between the organic (alkylated) and inorgan- generally those which lack sufficient sensi- ic forms of the toxic elements: arsenic, lead, mer- tivity for use at the low concentrations char- cury, and selenium. These rely on the fact that the acteristic of many elements in biological sys- inorganic forms can be easily converted to gase- tems. ous forms (mercury vapor and the hydrides) by 2. Biological systems are inherently very com- reaction with borohydride while the organic plex mixtures of a wide variety of chemical forms must first be photodecomposed by exposure constituents. It is a truism that the complexi- to ultraviolet (UV) radiation, Thus, analysis of the ty of the identification problem is dramat- gaseous reaction products before and after UV ically enhanced by the compositional com- irradiation provides measurements of the rele- plexity of the target system. vant organic and inorganic concentrations. This 3. While it can be argued that chemical iden- chemical differentiation approach has been inter- tification techniques are necessary for the faced with atomic absorption and emission spec- delineation of what chemical forms are bio- trometry as the measurement tools; inclusion of logically or toxicologically important, it ap- this capability in the laboratory is recommended. pears equally valid to argue that such knowl- Similarly, the ability to differentiate between edge on an a priori basis is highly beneficial the possible oxidation states for some metals is as a n aid i n focusing the development t work. important, e.g., As(III) versus AS(V) or Cr(III) ver- 4. The analytical chemical community is often sus Cr(VI), Such differentiation for As, Cr, and only partially aware of what types of chemi- some other metals can be accomplished by use of cal speciation measurements are considered selective oxidation-reduction (redox) reactions. most important by the medical, toxicology, The redox reactions may be controlled by proper and ecology communities. Similarly, the lat- selection of chemical reagents or by judicious use ter are often only peripherally aware of the of electrochemical (polarographic or voltammet- most promising emergent analytical technol- ric) principles. Electrochemical measurements ogy. can also be used to differentiate between electro- 5. The scientific communities involved in this active and nonelectroactive (chemically bound) chemical speciation question have often forms of some elements and between “labile” and been forced by the pressing circumstances ‘‘nonlabile”’ bound forms of some elements. In this so frequently associated with deleterious ef- context, labile and nonlabile refer to the thermo- fects to expend their efforts on the use of less dynamic and/or kinetic stabilities of the metal- than satisfactory approaches leaving less ligand systems in question. Although this type of time for the required types of development. differentiation may be only semiquantitative or Although other contributing factors could be semiempirical in nature, such measurements cited, the case in point is simply that the chemical have been shown in some instances to be perti- speciation capabilities so badly needed are avail- nent in experimentally defining bioavailability, able in only limited instances. The development of transfer mechanisms, transfer rates. etc. It is cer- adequate speciation technology should be a cen- tain that research must still be done to further de- tral thrust of research efforts involving collabora- lineate the diagnostic potentials of such types of tive efforts between the toxicology, biological ef- differential measurements, Present knowledge fects, and analytical chemical communities. Gov- seems sufficient, however, to support the inclu- ernmental and private funding agencies should sion of polarographic and voltammetric instru- strongly encourage these efforts, Because propo- mentation in the laboratory facilities for “labil- gation of chemical speciation developments will ity’” and/or oxidation state measurements. ultimately be essential to the success of the over- A limited number of other possibilities exist for all monitoring program, the laboratory facilities chemical speciation measurements. These are available at the outset should take advantage of largely very specialized in terms of applicability present capabilities. Moreover, the prime stance to certain types of samples and will not be dis- of the program should be that of providing feed- cussed herein. It is certain that the need for chem- back to the relevant scientific communities as one ical speciation exists and that its importance will means of focusing speciation research efforts on become even more apparent as more definitive 214 “ Environmental Contaminants in Food monitoring programs evolve. It is also certain that tion problems will involve the application of some no single analytical technique for speciation or combination of chemical and physical principles, any other purpose will be a universal panacea. The research programs most cognizant of this are The ultimate solutions to the majority of specia- most likely to be successful.

GLOSSARY OF ABBREVIATIONS

AAS—atomic absorption spectrophotometry Ng—nanograms—one billionth of a gram AES—atomic emission spectrometry PES—plasma emission spectrometry ASV—anodic stripping voltammetry PIXE—proton induced X-ray emission (spectrometry) cc—cubic centimeters ppb—parts per billion ( 1 x 10-’) gms/gm on a weight cm--square centimeters basis or 1 x 10-j µg/ml on a weight per unit volume DCP—direct current plasma basis) DPP—differential pulse polarography ppm—parts per million ( 1 x 10-6; gm/gm or 1 µg/ml as for FC—Faradaic or analytical current ppb) Ge(Li)—designates a germanium crystal which has Si(Li)--a silicon crystal which has lithium drifted into it lithium drifted into it to give it uniquely defined to provide unique physical properties properties SMS—solids mass spectrometry gin-gram µg—one millionth of a gram ICP—inductively coupled (radiofrequency) plasma µl —one millionth of a liter MAES—multielement atomic emission spectrometry µm—one millionth of a meter Mev--million electron volts XES—X-ray emission spectrometry ml—milliliter XRF—X-ray fluorescence (emission spectrometry) NAA—neutron activation analysis Appendix I

Analysis of Foods for Radioactivity*

by Naomi H. Harley

The analysis of foods for radioactivity should siderable effort in analysis. Since most of the ra- not be considered as a primary defense against dionuclides are already present, measurements human intake. The first indication should always made for background information should produce come from information on releases or from meas- a numerical answer, not merely an indication that urements of radioactivity in the airborne or wa- the amount is less than some pre-set value. The terborne releases. Once the existence of contami- accumulation of background data provides a valu- nation has been established then the foods can be able baseline for evaluating excursions following analyzed to evaluate potential hazard to man. a contamination event, The natural activity data In contrast to most other pollutants the effects are equally valuable since the amount of informa- of radiation are considered to have a linear re- tion on food concentrations is presently insuffi- sponse regardless of the level, thus, there is no cient for valid comparisons with manmade radio- threshold and no absolutely safe limit. Instead it activity. is necessary to set some lower level below which This report will describe the requirements and the radioactivity in foods is no longer of interest considerations for establishing a system to pro- as compared with other sources of radiation or duce acceptably accurate measurements of radio- other hazards of life. The analytical significance nuclides in foodstuffs. The basic concepts will be of this is that the lower limits of detection for ra- described, but to maintain the necessary brevity, dioactive substances have been brought down to the detailed procedures that might be used will be very low levels and the simple yes or no testing for given only by reference. acceptability that satisfies regulations for many During the preparation of this report, FDA has other pollutants in foods cannot be used. proposed certain recommendations for State and The radionuclides of interest in the case of con- local agencies on Accidental Radioactive Contam- taminating events are almost all present now in ination of Human Food and Animal Feeds. This foods in measurable quantities. Short-lived nu- material appeared in the Federal Register for De- clides are the exception and the transuranic ele- cember 15, 1978, page 58790, and is interesting ments are only present at levels that require con- background material for this topic.

RADIONUCLIDES

The most useful classifications of a radionu- The emitted radiations are generally grouped clide are those based on the characteristics of the as alpha (cY), beta (ß), and gamma (~). Alpha radia- radiation emitted and by the identity of the chemi- tion is characteristic of the natural and artificial cal element. The former is both a guide to the na- radionuclides of high atomic weight and consists ture of the hazard involved and to the measure- of energetic particles with very low penetrating ment required. The chemical species (e.g., ele- power. Its hazard is significant only within the ment, oxidation state) regulates the metabolic body, where alpha-emitting nuclides can irradiate pathways in the biosphere as well as the nature of specific sensitive tissues. Beta radiation appears any chemical separations required in the meas- in both heavy and light natural and manmade ra- urement procedure. dionuclides, and consists of electrons possessing kinetic energy and having modest penetrating * f{k( [:rpi from [III () I A 11 ~~rhllw I)(lp[’r l:rl II tlfxj “‘Arl;ili \is [II power. Gamma radiation is pure electromagnetic 1 II{)(Is for R:i[iro;i(ll\it} .’” A (l)mpl{’tc rIIpV of the p(iper (:ln 1)()

{ )t) t ii I rl~x 1 [rt )m t ht; Nii t I I JIII i 1 1 I;( h rl] ~ i~ I I n [( )rm; i t I( )rl S(I rv I( t’. ( SC(? radiation and is extremely penetrating. Thus, it i i[)~). J . ) can be a hazard externally as well as when it is

215 216 ● Environmental Contaminants in Food present in the body. For the present purpose, we tion. Many radionuclides are not potent health are only concerned that the penetrating nature of hazards, particularly those that are not metabo- gamma radiation allows its direct measurement lized by the body. The general groups of nuclides in foodstuffs, while alpha and beta emitters gen- to be expected include the natural activities, spe- erally must be separated from the bulk constitu- cifically radioactive potassium and members of ents of the sample before measurement is possi- the uranium and thorium series, and the artificial ble. fission products, transuranic elements, and other There are other processes in radioactive disin- activation products that result from nuclear tegration that produce emissions. Alpha emission weapon explosions and nuclear reactor opera- is usually accompanied by low-energy gamma tions. rays that may be used for measurement. X-rays Fission products are a very complex mixture at can be produced by electron capture and some the time of formation but the short-lived radionu- gamma emitters decay by internal conversion, a clides die out rapidly and the mixture becomes process where a fraction of the gamma rays are simpler within a few days, The transuranics (plu- converted to monoenergetic electrons. These tonium, americium, etc., formed by activation of processes do not really modify our measurement the basic fissionable material) are of some inter- concepts but the decay modes of the significant est because of their high toxicity when incorpo- radionuclides must be known for their accurate rated into the body but present evidence indicates measurement. that their uptake through the gut is relatively The physical half-life of the radionuclide tends small and that dietary intake is not a significant to control its persistence in the environment. For problem. This should be true particularly if the example iodine-l 31, with a half-life of about 8 relative hazard of other radioactive contaminants days, is important for only a few weeks, while probably present in the same sample is taken into cesium-137, with a half-life of about 30 years, may account, The other activation products are fre- be a problem for centuries. quently elements that make up steel or other A third classification that has some value is the metal containers or structural elements. Radioac- source that produced the radionuclides. Knowl- tive manganese, chromium, cobalt, zinc, and iron edge of the source of radioactive contamination are particularly common and result from inter- gives a good indication of the nuclides that are to actions of the materials with neutrons released in be expected in the sample. This is of considerable the nuclear reaction. It is worth pointing out that assistance in planning the analysis since request- contamination of foodstuffs with single nuclides is ing a complete analysis for all radionuclides or extremely unlikely, and that more than one mem- even for all types of radioactivity in a single sam- ber of any group will probably be present in any ple would lead to a lengthy and expensive opera- sample.

DISTRIBUTION OF RADIONUCLIDES

The source of the radionuclides involved will Nuclear reactors in normal operation release generally control their distribution in the environ- chiefly the radioactive noble gases that are not of ment and their consequent transfer to the food interest in considering foods. Reactors do contain chain. The sources considered here will include large inventories of fission products, transura- natural radioactivity, releases from operation of nics, and other activation products, however, and nuclear reactors and processing plants, and fall- accidental releases can contaminate vegetation out from nuclear weapons tests. by deposition or through the water pathway. Gas- Natural activity may be of concern when it is eous releases would most likely involve the vola- enhanced by man’s intervention, say by mining to tile elements such as iodine and tritium or those bring material to the surface and processing the with volatile precursors, such as strontium-90 ore to yield either products or wastes that may and cesium-137, Aqueous releases would follow concentrate the radionuclides. Good examples failure of the onsite ion exchange cleanup system are radium in uranium tailings, in phosphate rock and any water-soluble elements could be in- waste, or in slags from phosphorus production, volved. Radium may enter the food chain by dissolving in Processing plants could also have either gas- ground water and transferring through plant eous or aqueous releases, but only fuel reproc- roots. essing is likely to be a significant contributor. In Appendix I— Analysis of Foods for Radioactivity ● 217

this case, the fission products are aged before being more or less, depending on the size of the processing and iodine and the gaseous precursor weapon and the conditions of firing (high altitude, radionuclides are not released. Tritium and car- surface, underground). bon-14 are the major airborne products, while the In summary, the deposition of airborne materi- waterborne radionuclides are the same as for al on vegetation or on soil is the route by which reactors. foodstuffs become contaminated and the subse- Atmospheric nuclear weapons tests distribute quent behavior of the radionuclide is controlled their fission products, transuranics, and other ac- by its chemical nature, including solubility and tivation products globally, with local deposition plant or animal metabolism.

CONTAMINATION OF FOODSTUFFS

Contamination of foods can occur either 2. It is marketed rapidly, so that short-lived ra- through atmospheric deposition or by transfer dionuclides. such as iodine-l31, can be eval- with water. In the first case it is possible for the uated. radioactive material to be in the form of insoluble 3. It is a major diet component in the United particulates rather than in a more available form States, both directly and as an ingredient of where it will follow the chemistry of the elements prepared foods. involved. A knowledge of the pathway is not abso- 4. There is a lot of background information lutely necessary but it does assist in deciding on available on previous contaminating events. the proper preliminary treatment of the sample of It is worth noting that some of these “advan- evaluating exposure. For example, surface con- tages’ are the factors that contribute t o the role tamination may have a different significance than of milk as a source of human exposure. the same material present in a plant through root Milk is a poor indicator of many contaminants. uptake. The natural activities, the transuranics. and the Since pathway information is not always avail- activation products have relatively low concen- able i t is generally considered proper to measure trations in milk. The first two are low because of radioactivity in samples that have been prepared poor biological transfer and the last because their as if for eating, so as to approximate the true ex- pathways are almost entirely through the aquatic pected intake. This will usually result in stripping or marine food chain. Thus, the best approach is off or washing off of a considerable fraction of to know what is in the environment through other surface contamination. Cooking is generally not monitoring systems, and to design the food anal- part of the preparation, as the mode of cooking ysis program to fit the circumstances. and the use o f j u ices, cooking water, and the like Other monitoring data are also necessary in cannot be predicted, fixing the geographical extent of a contaminating Milk is often recommended as an indicator food event. As a general rule, nuclear tests are global for studying radioactive contamination. It has in radionuclide distribution, with enhanced levels many advantages: near the test site. Releases from other nuclear op- 1. It is available locally at most desired sam- erations tend to be more local in their effects and pling locations. the food-monitoring plan can be modified to suit.

SAMPLING

The mechanics of obtaining representative food as to its place of origin. For evaluating intake i t is samples will not be considered here since the pro- possible to collect total diet samples for a partic- cedures are common to all types of food analysis. ular population group and to measure the radio- There are certain points that must be considered activity in this composite diet. This approach has however. The first is whether the measurements been described by FDA(2) and by EPA(3) and is are being made to determine human intake or the normally applied in institutions where mass feed- source of the contamination. In the latter case the ing is carried out. A more elaborate procedure is early approach used by FDA(1) for radioactivity is to simulate a total diet by measuring a number of approp r i a t e. There, each sample was identified component food classes selected on the basis of 218 ● Environmental Contaminants in Food statistical information regarding consumption. and radioiodine. Most other foods give limited This approach was originally used by the Atomic geographical or seasonal coverage and are less Energy Commission(4) and was applied originally satisfactory. to three major cities. This approach does allow The preservation of samples in the field during identification of specific food types that are con- transport and in the laboratory awaiting analysis taminated but requires much greater effort and is only an esthetic matter in the case of radionu- cost in analysis. elides. Decomposition processes do not change The use of indicator foods as an intermediate the radioactivity, and sample contamination by type of monitoring is widespread. As mentioned radioactivity is unlikely. Thus, freezing, formalde- previously, most of the systems depend on the hyde addition, or any other method that will main- sampling of milk which is available in most parts tain the sample is adequate. Any additive should of the country either with specific information as be checked to assure it does not contain signifi- to place of origin or the general area of the milk- cant amounts of the radionuclide sought. shed. Part of the reason for using milk is the ease For the purpose of this report we will assume in sampling but it also has significance as a pri- that an adequate quantity of a representative mary food for the youngest and most susceptible sample is available for analysis and that another population group and it also does tend to pick up portion is available for storage, either permanent- several of the fission product nuclides of dietary ly or until the analytical results are accepted as significance such as radiostrontium, radiocesium, satisfactory.

SAMPLE PREPARATION

The measurement of radioactivity is a physical esses should not lead to losses of elements of in- process and it is most efficient when the radioac- terest with the exception of iodine and tritium, tivity from a relatively large sample can be placed mentioned above, and of carbon. close to the detector. This means that direct meas- Another approach that is essentially one of re- urements of bulk samples are only useful at rela- ducing bulk is to extract either the original sam- tively high levels of contamination and that most ple or the dried or ashed material with acids or measurements are preceded by preparation and other solvents and thus remove the desired ele- possibly chemical separation to reduce the bulk of ments from the bulk of the sample. This requires the material and to improve the efficiency of the considerable testing beforehand to be certain measurement. that the process operates in the desired manner. As mentioned previously, sample preparation All of these procedures reduce the bulk of the may include removal of inedible portions of the sample and in the case of extraction may also foods or those portions not generally eaten. For separate the desired constituents from some of example, citrus rinds, apple cores, outer leaves of the remaining inert material. They do not, how- leafy vegetables, and aboveground portions of ever, separate the radionuclides of interest com- root vegetables would normally be discarded. The pletely from the other radionuclides present in general goal is to prepare the foods as if for cook- the sample. Such separations will be covered in ing or consumption. the next section, but they may not be necessary if Foods generally have a high water content and the measurement technique can provide both a primary method of bulk reduction is drying at qualitative and quantitative information. In most room temperature, at elevated temperatures, or cases this limits the possibilities to gamma spec- freeze drying, Most of the radionuclides of inter- trometry on the prepared sample. est are not volatile under these conditions and If the samples are going to be subjected to losses must be considered only for elements such chemical analysis, the sample preparation must as tritium and iodine. The dried material can be include the dissolution of the dried or ashed mate- reduced further by ashing at elevated tempera- rial. This has already been done, of course, in tures, by cold ashing with activated oxygen, or by preparing the wet ashed or extracted solutions. wet ashing with oxidizing acids. The sample dry- The radionuclides of interest in ashed foods ashing process is most likely to lead to loss of should be soluble in strong acids if ashing tem- volatile elements but with care even cesium, polo- peratures have not been excessive, and this treat- nium, and lead can be retained, The other proc- ment is usually accepted. If there is concern that Appendix l—Analysis of Foods for Radioactivity “ 219

insoluble particulate may be present, it is neces- Since unforeseen questions often arise, it is rec- sary to use more drastic methods such as fusion to ommended that as many of these quantities be bring the sample into solution. This should not be measured as is possible. necessary, however, if human hazard is the prob- lem under consideration. The requirements for measurement of radionu- At the time of preparation, if not before, the clides in foods are such that sample preparation basis of measurement must be established. De- is best handled by the group responsible for the pending on the use of the data, wet weight, dry rest of the analysis. The chief difference from weight, ash weight, volume, or even numerical other types of analytical work is the initial sam- count (e. g., eggs) has to be determined. Frequent- ple. The usual range is from 1 to 20 kg, and the ly, this may have to be done in the field, but for- reduction of this amount of material is a special- tunately, relatively crude measures are adequate, ized problem.

RADIOCHEMICAL SEPARATIONS

Radiochemical separations are required to iso- ble to measure these losses in analysis and to late the desired radionuclide both from the re- make a correction at the end. One approach is to maining bulk constituents and from other radionu- measure the carrier at the end of the analysis clides which would interfere in the measurement. either by gravimetric or instrumental analysis. If In addition, it is necessary to convert the final there was none of the carrier substance present product to a form suitable for presentation to the in the initial sample, the fraction recovered will counter. This may involve elect redeposition, pre- be equal to the fraction of desired radionuclide cipitation, or other processes. There are a num- recovered. Where carriers are not available or ber of manuals giving the details of specific radio- where they are present in variable amounts in the chemical procedures and these details will not be original sample it is frequently possible to use repeated here. There are a few generalities how- radioactive tracers, that is radioactive isotopes of ever that may be of interest. the same element or a similar element. The frac- The actual mass of radionuclide that is meas- tion of the amount added that is left at the end of ured is almost always vanishingly small. This the analysis can be used to determine the recov- means that many of the normal chemical reac- ery of the radionuclide sought. tions used in analytical chemistry will not take In the chemical separations there is consider- place; for instance precipitates will not form. For able use of classical analytical chemistry based this reason it is common to add a few milligrams on precipitation and in most cases it will at least of carrier material which is preferably the inert appear as a final collection step to put the desired form of the same element. Where this does not ex- radionuclide in a condition suitable for counting. ist it is frequently possible to use similar elements In addition the general techniques of ion ex- as carrier, such as the substitution of barium for change, liquid extraction, distillation, and elec- radium. The inert form then follows normal chem- trolysis are used. istry, carrying the radionuclide with it. It is also Since many of the separations in radiochemical worth noting that even when the separation tech- analysis are carried out in small volumes the cen- nique does not depend on the mass of element trifuge is widely applied, It has become common present, a carrier may still be useful in prevent- practice to redissolve and reprecipitate rather ing unwanted coprecipitation or absorption on than to wash precipitates carefully. This may be glassware. repeated several times very rapidly and will gen- Because of the high degree of purification re- erally give good decontamination (separation quired in radiochemistry, it has been customary from other radionuclides). As noted above, where to make a number of repeated separations either the element sought and the contaminant have identical or different to insure purity. To carry similar properties it is preferable to use two dif- these out in a reasonable time it is better to lose a ferent precipitations, a precipitation followed by small amount of the nuclide sought while remov- an extraction, or any two widely different steps to ing a large fraction of the undesired material achieve good decontamination. rather than retaining all of the nuclide and much One process that is frequently of value is called of the undesired material. Fortunately, it is possi- scavenging. This term is usually applied to a pre- 220 ● Environmental Contaminants in Food cipitation carried out to remove contaminating ra- sample preparation are discussed: for example, dionuclides after the bulk matrix of the sample the sample area and mass should be reproducible has been removed. The process involves the addi- and the sample should be mounted on a sample tion of a group carrier, precipitation, and discard- holder identical with that used for the counter ing of the precipitate. A typical example would be standard. Since it is frequently necessary to de- an addition of iron carrier followed by hydroxide termine the weight of precipitate for recovery de- precipitation to remove rare earths and other termination, this factor must also be considered. heavy metals in a determination of an alkali or The selection of a mounting technique is usual- alkaline earth. Frequently the scavenging proce- ly a compromise between convenience and the dure is repeated to improve decontamination, but counting requirements. Besides consideration of it is only rarely that the scavenge precipitate is the type and energy of emission, practical matters redissolved and reprecipitated to recover any of such as the counter size and sample mounts avail- the desired constituent that may have been ab- able must be weighed, sorbed. In certain cases, the total amount of sample After suitable radiochemical separations have available may be limited and analysis for several been made, it is necessary to collect and mount radionuclides may be required. Procedures the sample for counting. This is most often done should be on hand for the sequential analysis of by precipitation or by elect redeposition. In the single samples, even though separate samples are section on measurement, the requirements for used for routine work.

MEASUREMENT

The method of measurement to be selected de- on the separated material for radionuclide identi- pends on the type of radiation, the form of the fication. The latter step may be omitted if other sample, and to some extent on the amount of ra- considerations limit the possibility to a single ra- dioactivity. It is necessary that the complete ana- dionuclide. lytical procedure be designed so that the sample The equipment available for quantitative meas- is brought to a suitable form for the equipment urement of radioactivity is sufficiently sensitive and conditions that exist. for all foreseeable cases. Instruments for detec- It is possible to measure the total gamma, total tion of the three major emissions are described beta, or even the total alpha activity on a sample here, and their applications are shown in table 20 of food, Unfortunately, such data are valueless in (chapter VIII) in terms of the detection limits for estimating human exposure. The accuracy of the various radionuclides. determination is very poor, natural potassium It is worth pointing out that a number of the in- usually interferes, and the chemical and radia- struments described are not commercially avail- tion characteristics needed to evaluate possible able. They have been in the past, but the low de- hazard are not known. It is possible, however, to mand has removed them from the market, The set a particular total activity level as a screening necessary electronic components are available level for a specific food. If the measured value is but the mechanical assemblies for the detectors below the screening level, no analyses for individ- are not, and the larger laboratories tend to build ual radionuclides are performed. In such a case, their own systems. the measurements should be considered as inter- Alpha Emitters. The measurement of alpha ac- nal data only and the numerical results should not tivity is best carried out on a very thin sample to be published. Any report should merely list the avoid self-absorption of the alpha particles. This samples as having activities below the stated is even more true for spectrometry since degrada- screening level. tion of the original alpha energy will give a spec- Qualitative identification of radionuclides on trum with poor resolution. The measurement of original samples of foods is limited to gamma emit- the total alpha activity can be carried out either ters at relatively high levels. The sensitivity can in thin-window counters or by scintillation count- be increased if some bulk reduction, as described ing with zinc sulfide phosphor. Both techniques under “Sample Preparation, ” is carried out. The have high efficiency but the scintillation method identification of alpha and beta emitters depends can give a considerably lower background with a on radiochemical separation for element identifi- consequently lower limit of detection. Unfortu- cation and the measurement of energy or half-life nately, neither system is readily available as a Appendix I— Analysis of Foods for Radioactivity . 221

complete unit ami must be assembled from com­ incidence techniques. These add to the complex­ mef(~ially available components. A somewhat less ity and cost of the system but are sometimes sensitive technique is to use the liquid scintilla­ necessary. tion spectrometer rlescribed in the next section. Scintillation counting can be performed in two Alpha spectroscopy also has two possibilities. ways. Solid scintilla tors can be used for counting the Frisch grid ionization chamher or the silicon chemical precipi ta tes collected on fil ter papers diode solid-sta te detector. The Frisch grid unit anci liquid scintilla tors can be used whenever the can handle large area samples but is slightly sample cnn be made miscible with the scintillat­ poorer in resolving closely sepa ra ted energies. On ing solution itself. This can even be cione \vith the other hancL th(~ silicon diodes available are all solids by suspending them in a scintilla ting gel. quite small and can only count samples of about 1- The advantage of liquid scintilla tors is their high em diameter with high efficiency. This is not a efficiency, even for the low-energy emi tters car­ serious limitation, since the Frisch grid systems bon-14 and tritium. Scintillation systems for are not pres(mtiv Hvailable in this country. This counting precipitates are not commercially avail­ m(~,lT1S that the form of the sample presented to able at present. There are. hovvever, many liquid the counter is limited. i\lost metclis are electrode­ scintillation systems on the market. most of them posited onto small smooth discs of stainless steel, with automatic sample changers. They enjoy a nickel, or pl,lIinum, vvith evaporation of pure solu­ good demand in medical and biological studies tions being used in some cases. with tritium and radiocarbon. These use high A specializ!~d t(~chniqlle is available for meas­ levels of activity and short counting time. The bet­ uring raclium-226. by means of its gaseous daugh­ ter systems also have a provision for rn ther crude ter product radon-222. If the sample can be put spectrometry. They can distinguish qualitatively into solution and stored for about 3 weeks the aiHi quantitatively among carbon-14. tritium, r;!(jon-222 \vill be close to equilibrium (equal ac­ alpha emi tters. and higher energy beta emi t ters. tivity) vvith the raclium-226. The gaseous radon Gamma Emitters. Gamma rays are so penetrat­ can then be transferred to a scintillation counter ing that the detector must have a considerable or an ionization chamber for measurement of its mass to absorb enough to produce a response. For "I .. h" ""i;";i,, 'rh;" ;" "nih"," n""";"l;",,,rl n~r~ ih", oqJII<1 <1LllVlly. 1111" 1<> I rtlllt"1 "PC\,IClLIE,Gll. rtllll lilt" spe(~tr()m8try, complete energy RIJs{)rpti{)n is re ... equipment is not listed in this report since radi­ qui red, so solid detectors are most useful. Sodium um-226 can be determined by other procedures. iodide is a popular detector, since crystals can be For exampl(~. sepa ra tion and c()un ting as the sul­ fabricated in large sizes and the crystal is trans­ fate or chromate using barium as carrier is quite parent to the scintillations produced hy radiation. adequate. Sodium iodide is used for most of the gamma Beta Emitters. Beta counting is a little more counting today, and is still useful in spectrometry. flexible in the mass of material that can be pres­ It has high efficiency but poor energy resolution ent at the time of counting. This is true for higher and is now applied to samples where some sepa­ energy beta emitters but carbon-14 and tritium ra tion has taken place. i\,1ilk is a good na turnl ex­ present a problem. Since each individual beta ample, as the metabolism of the cow removes most emitter gives off particles with a range of ener­ gamma emitters other than isotopes of cesium and gies from zero to a characteristic maximum. beta iodine. spectrometry is not possible for food samples. More complex spectra can be resolved with the Some quaiitative information may be obtained solid-state germanium diode detector. A compari­ from absorption measurements. which allow an son of a sodium iodide with a germanium diode estimate of the characteristic maximum energy of detector spectrum is shown in figure 1-1. It is single emitters or simple mixtures (5). possible to resolve the sodium iodide spectrum of The available counting equipment for quantita­ a sample containing several radionuclides but tive meaSllf!:ment includes geiger counters. pro­ this requires a sizable computer and very careful portional counters. and scintillation counters. work to allow for drift in instrument character­ The geiger counter is relatively inexpensive and istics and for the possibility of other radionu­ requires only simple electronics but is not popular elides than those sought being present. The diode and is generally not available as a counting sys­ detector gives much better resolution and present tem. The thin-window proportional counter is spectrometer systems frequently include suffi­ used widely Hnd has both reasonably high effi­ cient computer capacity to resolve the resulting ciency and low background. For low-level samples spectra and produce quantitative data. At the the background can be further reduced by antico- same time, interferences from radionuelide im- 222 “ Environmental Contaminants in Food

Figure I-1 .—Comparison of Gamma Spectra Taken With a Sodium Iodide Detector (Upper Curve) and a Germanium Diode Detector (Lower Curve). Note that the main single peak in the sodium iodide spectrum is resolved into 9 peaks for 6 radionuclides with the germanium diode spectrum “

o

purities are greatly reduced compared to spectra foods have not been high enough for this tech- from sodium iodide detectors. The efficiency of nique. the diode is low and, for many analyses, a spec- General Requirements. The choice of a count- trometer can only be used for one measurement a ing procedure depends on the precision required. day. Another disadvantage is that the detector In turn, the relative precision of a quantitative must be kept at liquid nitrogen temperature to counting measurement is inversely proportional maintain its detection capability. Newer diodes to the square root of the number of counts ob- have been developed that do not require storage tained. Thus, any improvement in precision must at liquid nitrogen temperatures; however, they be obtained by increasing the number of counts. must be cooled during measurement. Since the This can be done by using larger samples, by system is in use most of the time, this may not be a counting for longer times, or by using counters significant advantage. with higher efficiency. A secondary improvement Diode spectrometers may also be used to meas- is possible for low-activity samples by decreasing ure the low-energy gamma-rays that accompany the background. Each of these improvements has alpha emission. This allows direct measurement some drawback, and selection of the optimum bal- in some environmental samples, but the levels in ance requires a degree of experience to weigh Appendix l—Analysis of Foods for Radioactivity “ 223

cost, manpower, and quality. For example, han- In addition, maintenance of control charts(5) will dling larger samples increases the effort in sam- signal when instrument problems arise. ple preparation and radiochemistry while longer Experience with modern nuclear instrumenta- counting times require more counters, and in- tion has been good, and downtime of the order of 5 creased counter efficiency or lower background percent is common. This requires that service be is expensive. available immediately or the next day following a Counting-room operation requires the mainte- breakdown. The increasing complexity of measur- nance of detailed records on standardization, ing systems tends to preclude in-house servicing background, and sample measurement. This in- in most cases, but some diagnostic capability and formation can frequently allow the recovery of competence in minor repairs is very valuable. bad data or the correction of calculational errors.

CONTROLS

An analytical laboratory carrying out measure- This requires that standard samples be available ments of radioactivity in foodstuffs will require a for testing the full procedure. Additional controls program to document the validity of the measure- would include running of blind duplicate samples ments, This is necessary in legal cases and is also to test reproducibility of analyses, and blank highly desirable when data are being presented samples to check on the possibility of laboratory for use in decisionmaking. A suitable program of or reagent contamination. These three types of quality control should be carried on in addition to samples should make up at least 10 percent of the the necessary calibrations and standardizations. laboratory output if the quality of measurement is Calibration is the determination of the relation- to be followed closely. It is also necessary that the ship between a desired quantity and the response data be published with the ordinary laboratory of a particular instrument. In the case of radioac- results and that any deviations from the expected tive materials this may mean that the instrument results should be used to make suitable correc- must be calibrated with each radionuclide to be tions in laboratory operations. measured unless there is evidence that the instru- There is always some danger that a laboratory ment response is independent of the energy or may drift out of control in spite of an adequate in- other characteristics of the radiation. Alterna- ternal quality control program, This is possible tively a complete response v. energy calibration since an internal program may emphasize consist- may be substituted. These calibrations, to have a ency while the absolute values drift. Therefore, legal standing, should probably be traceable to some part of the quality control effort must be the U.S. National Bureau of Standards or compar- devoted to intercomparisons with established able authority. This turns out to be a requirement groups, such as IAEA or EPA. This should be done that is far from trivial in the effort required. on an annual basis, at least. Fortunately, a complete calibration is not re- It must be noted that standard samples are fre- quired at frequent intervals and, depending on quently not available, and recourse must be had experience, may not be needed oftener than every to “spiking,” which is the addition of a known few months. In the meantime of course it is neces- amount of a radionuclide to a sample that is rela- sary to have assurance of the proper operation of tively free of that nuclide. Recovery of the spike is measuring equipment but this can be done with not necessarily a good test of a chemical proce- simple standards or even with samples that are dure, since a natural sample may be more diffi- reproducible over a period of time. For example, a cult to analyze. simple counter standard may be run every morn- A final component of a good quality control sys- ing before starting operations just to be sure that tem is a careful, responsible review of all the data the counter is working properly. Similar stand- produced. This means checking the arithmetic, ards should also be available for checking spec- knowing the characteristics of the equipment and, trometer energy response. hopefully, having a sixth sense that recognizes An ideal quality control program should in- that certain results do not look right. This is clude the checking of the complete procedure especially true when relying on automatic count- from sample preparation through measurement. ers or on computer processing of the data. 224 . Environmental Contaminants in Food

STAFF AND FACILITIES

A minimum facility could be designed around a pability already exists in the overall organization, staff of six, including a B.S. or M.S. senior chemist it may be fruitful to copy advanced noncommer- with experience, a B.S. junior chemist and three cial instruments, but staffing specifically for this technicians for chemistry, sample preparation, function is probably not economical. Thus, most and counting-maintenance, A- secretary-adminis- small laboratories must operate with the less- trative assistant could handle reports, local pur- than-optimum commercial equipment, chasing, and similar duties. With this small staff, If significant uses are to be made of gamma considerable versatility and flexibility would be spectrometry there will be a need for at least required. modest computing facilities. These can be part of The measurement of radioactivity in foods re- the purchased spectrometer system but this is an quires fairly extensive facilities to handle the expensive method if suitable computer time is varied analyses that might be called for, In addi- available otherwise. The former approach is used tion to a modest office space, separate rooms here, and it might be noted that the spectrometer would be necessary for sample preparation, wet computer has capacity for other work. chemistry, measuring equipment, and for mainte- Additional space is required not only for stock- nance support and storage. The first two lab- ing necessary reagents and materials but also for oratory rooms would require hoods, chemical storage of incoming samples and for storage of benches, and laboratory safety devices such as residual material from samples that have been showers and eye fountains. Since the samples to run. As a general rule it is desirable to take a be measured are normal foods there are no spe- larger sample than would be required and to set cial requirements for radiation safety, If the sam- aside a portion of this for possible contingencies ples were radioactive enough to be a personnel or if the data are later brought into question. hazard in the laboratory they certainly would not The output of a group this size should run be- require this type of precise measurement. tween 1,000 and 4,000 samples] a year, depending If the laboratory is to be in continuous opera- on the difficulty and the activity level, One limit- tion it is most likely that the principal chemical ing factor is that, at background levels, each and measurement systems will have to be avail- counter can only turn out one or two samples per able at least in duplicate. This and similar consid- day. Following a contaminating event, it should be erations lead to the conclusion that a certain possible to process smaller samples and to count minimum size and sample throughput are not nec- for shorter times, both of which would allow high- essary if the radioactivity laboratory is part of a er output. It might be possible to add 3 more lab- larger operation which can furnish support. One oratory staff plus 50 percent more space and continuing problem is the need for electronic equipment dollars and essentially double the out- maintenance of radiation measuring equipment. put. Further increases would probably run into This is rather specialized and must be considered other bottlenecks due to the need for added sup- either when staffing the laboratory or in locating port. it where such services are readily available. Table I-1 details the costs for space modifica- It is not always possible to purchase the ideal tion and furnishings ($140,000), for equipment counting equipment for a particular purpose. The ($200,000), and for supplies ($20,000). These costs total demand for many systems is small and com- are based on a building in place, lights and parti- mercial instrument makers are not interested. tions in place, and utility stubs in each room, It is Large laboratories can make some of their own also assumed that building services are provided. equipment but this is not possible for the group described here. The construction of alpha and beta radiation detectors requires services of a first-class machinist plus sufficient electronic IA sample is defined as a ~x]mp]ete gamm{i spectral an[]lysis know-how to transfer the detector signal to the or a single raciionuclicie an[]lysis that requires preparation plus available commercial equipment. Where such ca- r:idi(x:hemist rv. Appendix I— Analysis of Foods for Radioactivity ● 225

Table I-1. —Costs for Space, Furnishings, Equipment, and Supplies

Cost cost Space modification and furnishings Chemical laboratory—20 x 30 ft. Base cabinets $ 8,500 Wall cabinets 2,500 Storage cabinets 1,800 Sinks (2) 1,200 Hoods (3) 7,000 Services and Installation . 40,000 Preparation area— 10x 30 ft. Base Cabinets 3.800 Wall cabinets 1,800 Storage cabinets 1.500 Sink 600 Hood 2,500 Benches 600 Services and Installation 25,000 Counting area— 10x 20 ft. (air-conditioned) Benches 1,200 Rack 1,000 Base cabinets 1,700 Wall cabinets 600 Desk, etc. 500 Services and installation 12,000 Storage area— 10x 20 ft. Acid storage 800 Solvent storage ,. 600 Cabinets 3,000 Services and Installation 1,500 Office area —library 10x 20 ft. Desks, etc (4) 2,000 Cabinets, bookcases 3,000 Services and Installation 1,000

Equipment Chemistry area Hot plates (2) 600 Stirring plates (6) 900 pH meter 350 Demineralizer 500 Centrifuges (3) 1,800 Platinum ware(12 items) 12,000

APPENDIX I REFERENCES

1. E. P. I.aug, et al. ‘radioactivity in F’oods, ”’ journal Data un(~ Heports 15, 647-56 ( 1974). (rI’his is a later A(~A(; 44, 52.3-30 ( 1961), :+Iso 4’Strontium-90 An~~l- report thti t summa rizes much of the EPA program. ) yses of I [uman an(i Animal Foods, H(l(liologic(]] 4. J. H. Harley and Joseph Rivera, Summ[lr} of Av(lil- HefJ](h fljt{] 1, 36-40 (hfay 1960). able h] ta on the Stron fium-{~[~ Con tent of F(mds (]n(l 2. ‘1’een-Ager Diet Survey, Food and Drug Administra- Total Diets in the United St{ltcs, LJ, S. AEC Report tion, ll{~(lif)]f)gicfl) Fff?fI)th ll{!(fl 4, 18-22 ( 1963) HASL-90 (August 1960). 3. R. E. Simpson, E. J. Barat ta, and (;. h’. Jelinek, “Radio- 5. J. 11. Ilarley IEciitor). EML Procmiures M(lnu(ll, U.S. nucli[ies in F’(Nxis: Nloni (orin~ Program. ” Rcl(fio (i{)n I)OE Report F{ ASL-300 ( 1972, revised annually). 226 ● Environmental Contaminants in Food

GENERAL REFERENCES

Radiochemical Methods International Atomic Energy Agency, Metrology and Radionuclides (Symposium), IAEA, Vienna ( 1960). National Academy of Sciences, Nuclear Science Series Crouthamel (Editor), Revised by Adams and Dams, Ap- (NAS-NS 3001 ff), Office of ‘1’echnical Services, plied Gamma-Ray Spectrometry, Pergamon, New Washington (1960 and later), A Series of Mono- York (1970). graphs on Individual Elements and Techniques. WHO-FAO-IAEA, Methods of Radiochemical Analysis, World Health Organization, Geneva ( 1966). Data Compilations Johns (Editor), National Environmental Research Cen- ter, Las Vegas, Handbook of Radiochemical Analyti- Lederer and Shirley, Table of Isotopes—7th Edition, cal Methods, EPA-680/4-75-001, February 1975. John Wiley, New York (1978). Harley (Editor), Environmental Measurements Labora- Health, Scintillation Spectrometry, Gamma-Ray Spec- tory, DOE, EML Procedures Manual, HASL-300, trum Catolog (3rd Edition), USAEC Report 1972 (Updated Annually). ANCR-lOOO ( 1974). International Atomic Energy Agency, Reference Meth- Chanda and Deal, Catolog of Semiconductor Alpha-Par- ods for Marine Radioactivity Studies (Sr, Cs, Ce, Co, ticle Spectra, USAEC Report IN-1261, March 1970, Zn), Technical Reports Series No. 118 (1970). Niday and Gunnink, Computerized Quantitative Anal- Ibid. (Transuranics, Ru, Zr, Ag, I) ysis of Gamma-Ray Spectrometry, USAEC Report Technical Reports Series No. 169 (1975) UCRL-51061, 4 Volumes (Includes Decay Scheme Data Library). Measurement of Radioactivity

National Council on Radiation Protection and Meas- Quality Assurance urements, A Manual of Radioactivity Measure- ments Procedures, Report 58 ( 1977). Kanipe, Handbook for Analytical Quality Control in Ra- National Council on Radiation Protection and Meas- dioanalytical Laboratories, EPA Report 600/7- urements, Environmental Radiation Measurements, 77-088 (April 1977), Report 50 (1976). Ziegler and Hunt, Quality Control for Environmental International Commission on Radiation Units and Meas- Measurements using Gamma-Ray Spectrometry, urements, Measurement of Low-Level Radioactiv- EPA Report 600/7-77-144 (I)ecember 1977). ity, Report 22 ( 1972). See also: International Atomic Energy Agency, Measurement of Metrology Needs in the Measurement of Environmental Short-Range Radiations, Technical Reports Series Radioactivity Environment International 1, 7ff No. 150 ( 1973). (1978),

GLOSSARY

Activation Products—The radionuclides formed in Fission Products—The radionuclides formed when fis- either fissionable material or in surrounding mate- sionable material splits into two or more atoms. rial during a nuclear reaction, usually by capture of Geiger Counter— A detector based on gas ionization neutrons, which converts any ionization within the detector Carrier—A stable element added in radiochemical into a large electrical pulse. analysis to provide sufficient mass that a radionu- Germanium Diode Detector—A gamma-ray detector clide can follow various separation steps. composed of very pure germanium, usually acti- Detector—Any device that transforms the radiation vated with lithium. emitted by a radionuclide into a signal that can be Indicator Food—A food that is measured to give an handled electronically. The most usual signal is an estimate of the total dietary intake of man. electrical pulse which can be recorded. Ionization Chamber—A radiation detector based on Disintegration-The spontaneous process by which a gas ionization which converts any ionization within radionuclide gives off energy in the form of nuclear the chamber into an equivalent electrical pulse. radiation. Pathway—The route by which a radionuclide is trans- Fissionable Material—Any of the heavy radionuclides ferred from the source through the environment to which can undergo fission, or splitting into two or man. more lighter atoms. The fission process is accompa- Picocurie—See units of radioactivity. Equal to 2.2 dis- nied by a large release of energy. integrations per minute. Appendix I—Analysis of Foods for Radioactivity . 227

Proportional Counter—A detector based on gas ioniza- Tracer—A radionuclide added in radiochemical anal- tion which converts any ionization within the detec- ysis to follow the distribution of the desired constit- tor into an amplified electrical pulse proportional to uent in various separation steps. the amount of ionization. Radionuclide—Any atomic species which is unstable Transuranic Element—An artificial element having a and gives off nuclear radiation to attain stability. higher atomic number than uranium, formed by ac- Scintillation Detector—A detector which converts nu- tivation, usually with neutrons. clear radiation to a pulse of light. This, in turn, can be converted to an electrical pulse with a photo- tube. Units of Radioactivity—There are various ways of ex- Sodium Iodide Detector—A scintillation gamma-ray de- pressing the disintegration rate of a radionuclide. tector made up of a single crystal of sodium iodide In this report the disintegrations per minute (dpm) activated with silver. unit is used. In other reports, the curie and its sub- Spectrometry—The measurement of the energy of radi- multiple (millicurie, microcurie, picocurie) are ation emitted during decay of a radionuclide or mix- used and the new International Standard nomen- ture of radionuclides. Quantitative as well as qual- clature is the bequerel, equal to one disintegration itative information may usually be derived. per second. Appendix

Commissioned Papers

1, Clement Associates, Inc., Washington, D. C., “Assessment of Carcinogenic Risks From “Priority Setting of Toxic Substances for PCBS in Food. ” Guiding Monitoring Programs. ” 5 Dr. Donald Epp, Department of Agricultural 2. Congressional Research Service, Science Pol- Economics and Rural Sociology, Pennsylvania icv Research Division, State University, “Determining the Economic ● Christopher H. Dodge, “Contamination of Impact of Alternative Action Levels. ” Food by Radioactive Nuclides. ’ 6. Dr. Naomi H, Harlev, Institute of Environmen- ● Sarah L. Hartman, “Nitrate Food Contami- tal Sciences, New “York University Medical nation: A Case Study. ” Center, “Analysis of Foods for Radioactivity. ” ● Jo-Ann M. McNally, “Mercury: Environmen- 7. Dr. John L. Laseter, Environmental Affairs, tal Contamination of Food, ” “Polybromi- Inc., New Orleans, La. “Approaches to Moni- nated Biphenyls: Environmental Contamina- toring Organic Environmental Contaminants tion of Food, ‘“Polychlorinated Biphenyls: in Food’ (appendix G). Environmental Contamination of Food. ” 8, Dr. Frank Lu, School of Medicine, University 3. Dr. William E. Cooper and R. V, Farace, De- of Miami, “Environmental Contaminants in partment of Zoology, Michigan State Universi- Food: International Activities. ” ty, “Toxic Substances in Food Information 90 Dr. Rodney K. Skogerboe, Department of System: Design and Management.”’ Chemistry, Colorado State University, “Ana- 4. Dr. Kenny S. Crump and Marjory D. Master- lytical Systems for the Determination of Met- man, Science Research Systems. Ruston, La., als in Food and Water Supplies” (appendix H), 10. SRI International, Menlo Park, Calif., “Assess- ment of Methods for Regulating ‘Unavoidable’ Contaminants in the Food Supply.”’ 11, Tracer Jitco, Inc., Rockville, Md., “Case Studies on Consumer Risk from Environmental Contaminants in Food. ”

,

228 Appendix K

Executive Agency Observers at Advisory Panel Meetings

U.S. Environmental Protection Agency Fred Arnold* Marty Halper* Rufus Morrison Janice Ryan

U.S. Department of Health, Education, and Welfare Food and Drug Administration John Wessel* Charles Jelinek Pasquale Lombardo Bart Puma Sidney Williams

U.S. Department of Agriculture Food Safety and Quality Service Grace Clark* Richard Ellis William Leese John Spaulding

*Official agency liaison to the assessment,

229