Click for next page ( 34


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 33
2 SAMPLE SELECTION THE UNIVERSE AND THE SAMPLING FRAME According to an estimate based on the Chemical Abstracts Service (CAS) Registry, the universe of known chemicals consists of over 5 million entities. me Committee on Sampling Strategies and the Committee on Toxicity Data Elements used a subset of the universe, a select universe, that contains 6S,725 listings of substances to which humans may be exposed. The committees considered many options in choosing the most appropriate select universe for accomplishing the task effectively. First, one might use the CAS list and select substances to which extensive human exposure is known or likely. This approach was not practical. Although the CAS list is computerized, it is not based on a potential for human exposure, and manual evaluation of such a large number of substances was beyond the physical resources of the study. Second, one could search for existing lists of substances preselected for human exposure potential and computerized for reasonably easy access. The committees' decision was based on a desire to include to the fullest feasible extent the substances of possible interest to NTP; to avoid the unnecessary inclusion of substances not of interest to NTP; to choose a select universe that, through analysis, most completely reflects true toxicity-testing needs; and to take advantage of chemical-name lists that were computerized, are associated with CAS Registry numbers, and permit retrieval for sampling purposes. In consideration of these factors, the committees established the select universe by identifying five classes of chemicals: (1) pesticides and inert ingredients of pesticide formulations; (2) cosmetic ingredients; (3) drugs and excipients used in drug formulations, (4) food additives, including direct and indirect additives, colors, flavors, chemicals generally regarded as safe (GRAS), and chemicals listed in the chemical-name category of the Food and Drug Administration (FDA) Bureau of Foods dictionary of chemical names; and chemicals in commerce, as listed in the Toxic Substances Control Act (TSCA) Inventory (U.S. Environmental Protection Agency, 1979, 1980~. Chemicals in commerce were separated according to 1977 production reported by manufacturers: (5) production of 1 million pounds (454 metric tons) or more, (6) production of less than 1 million pounds, and (7) production unknown or inaccessible because of manufacturers' claims of confidentiality. The Committee on Toxicity Data Elements and the Committee on Sampling Strategies regarded the contents of the select universe as closely approximating the universe of interest to NTP. The committees recognized that this select universe did not systematically include substances that were environmental decomposition products, manufacturing contaminants, or natural substances (e.g., plant pollens and foods). However, the committees elected not to include a miscellaneous category of such substances, because a suitable list could not be identified. Therefore, the sum of the above, 65,725 entries, was taken as the select universe of substances for purposes of 33

OCR for page 33
this study. Statistical adjustment for duplications indicated that the select universe contained about 53,500 distinct entities. The samples selected by the Committee on Sampling Strategies were for the use of the committees in fulfilling the study's objectives. They were chosen only as representatives of the select universe, not for any intrinsic interest of their own. Each of the seven categories can be divided into subcategories based on a variety of criteria. For example, drugs could be divided according to route of administration (e.g., oral, parenteral, or intravenous), strength of pharmacologic activity, intended-use volumes, production volumes, or structure-activity relationships. Some of the subcategories might cross the boundaries of categories of intended use. The Committee on Sampling Strategies recognized that the select universe of chemicals could theoretically have been divided into classes based on criteria other than intended use--e.g., chemical structure, degree of toxicity, extent of exposure, or biologic availability--and that each such class could be further divided into categories that are different from those listed above. For the purposes of this study, however, discussions are based on the seven categories and their possible subcategories. The sources and scope of representative lists of chemicals contained in the seven categories are presented in Table 1. A major role of the Committee on Sampling Strategies was to develop a sampling procedure that would yield statistically valid estimates of the select universe. The procedure is described later in this chapter. The statistical validity of the estimates was especially critical, because the success of the study required, in part, that the sample be used to estimate the status of other chemicals in the select universe. The seven categories of the select universe embody a large collection of chemicals that formed the sampling frame from which the sample was drawn. An overview of the sampling process is presented in Figure 1. The sampling procedure began with the selection and preparation of the lists that most accurately presented the seven categories of the select universe (see Table 1~. The characteristics of each depended on the purposes of the organization for which it was constructed. The sizes, contents, and formats of the lists varied. The lists of pesticides and inert ingredients in pesticide formulations, cosmetic ingredients, drugs and excipients used in drug formulations, and food additives had characteristics, such as use functions, that were more consistent internally than the characteristics of the TSCA Inventory of chemicals in commerce. They were also substantially smaller. The characteristics of each category in the select universe are described below. 34

OCR for page 33
us :~ o In In i - as c) o In i - In JJ u ~ ~ 3 S U :~ ~ A' al EN en In .,1 S O Cat U} ret . - Ld o U] U] to ,. C) ~ Cat te 1 1 1 ~lo us Q 1 1 1 ~ In ~1 1 1 ~o U] , ~t_ US QO id ~D L ~fir o As Cat ret ~co #) 54 Lit U] ~ ~ ~1 ~ O :3 ~ U .~4 _I ~ ~S ~ Jet {15 ~ ~.- tl ~~ > ~-~ > ~_ ~dJ N J ~ ~l I ~ ~1 .- = - S i4 ~Q. Ld ~ ~ ~S ~ O ~ a: ~ =4 54 D~: - . U Q4 o U U] U 3 U] s 4 - - ~ :D :~ U cn 1 ~ ' - CR ~00 ' ~ :D 00 ~q 1 :>,U] .,- ~. - ~ ~ ~ ~ U . -O ~ U] ~ C ~ ~ O U] ~U! ~ ~ ~ ~ O O ~U] _t ~ ~ ~ - ~ ~ ~ ~, - ,- ~ ~ y Q~ ~ ~ ~Q ~ O - Q C (IJ 1:~ C~ Qd OO U} _1 U] - O 04 te ~ tl ~ U] - ~1 - ~ tQ J~ ~_~ ~ - - Ul 41 :~ ~ ~ ~ ~ ~ ~ ~ ~ ' ~ ~ ~ ~ ~ ~ ~ - ~ ~ ~ ~ ~ ~ ~ ~ ~D4O ~ cm ~s ~ ~ ~ ~ S ~ ~ ~ ~ - O ~ 1 ~ ~ ~- ~ ~ ~ ~ U v ~ 0 (Q u s ~v ~ 0 "a ~ ~ 1 - .,, eq ~l ~ ~. - ~ - ~ ~ ~ - ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ m a~ v ~: ~ ~ ~ ~ ~ ~ ~ - ~ ~ ~ c ~ v S ~ ~ ~ O - ~S ~ ~ ~ ~ ~C ~ O ~ O C X ~ ~ ~ C-) ~ ~ ~ 0 ~ ~ ~ ~ - V- ~ - ~ ~ ~ ~ ~ ~ O P. C) ~Q C ~ ~ 1 - ~S ~ V C ~' - :, O O rr, Q C 61 0 ~ C O ~ ' ~ ^ ,4 ~, ~ ~ ~ ~ ~ ~O C ~ - C V O ~ ~ ~ ~ _ :> ~n ~0 ~q .,~, .,~ ~ ~ ~, ~ .,. - Ll ~ ~V ~ ~ ~ ~ ~ O ,. ~ ~ ~ - ~ ~ ~ - ~ 0 ~ ~ O v ou e 0 0 ~ s ~ 0 0 ~ 0 r ~ c a: ~O E ~ O~OE~<- 0 m.~= {Q a ~ 35

OCR for page 33
A o ~ By GO - 3c .,. 8 - m o U] In . - U] U) In ,' 4 ~ of o me CO a' `: 0 ~ JJ .~4 N ~ .,1 Q S ~QJ O o U U) ~4 3 U] U) .,, s ~ rid O U) ~5 ~ at I: CO ~ a, U] U. O ~ ~ O _1 Q 3 ' - a) 0 A' A' to U) 3 3 a: U] .,. ~1 o tl] 1 O . - I: O 3 3 m ~ c o .,, 0 0 0 0 0 Us ~U. o ~ ~ ~D ~ U~ co ~ ~ ~ ~ '~ r~ ;n ~ ~ Q ~ e3 - :~, Ll U' 3 ~1 U] a; U] . - ,l o o rl cn ~ . - cn h ~ O ~ ._~,- C) O . - ' ~ U] O o W U] ~4 CQ o Q 3 U] V . ~ U) O ~ ' ... S ~ - 5: a' _1 Q o o o V 4 ~ O O O :D O O . - O ~ U] . O O _d ~ O O tt5 .s ~^lv ~ :>~ 4~ Q Q ~ G. Q O O - . - U) U] 3 3 0 C: O P. 3 '- 36 Ln ~D ~ 0 Q _I .,1 ..H Q ~ - U "l 00 1 ~q C U) ~ 1 O ~ Ll ~n C Ll ~ C . ~a ~ :^ ~q C o o ~ - 1 O C ,. 1 ,~ , tO 3 ~n o O J~ 4~ C O ~ O T5 ~ ~n ~n ~ C JJ n ~ tn O U] ~ U U) U] C~ Q ~1 U] ,' U U] o C U] .,. S o C . - ~n ~n ~4 3 U] g :3 C r1 cn C

OCR for page 33
Select Universe Assemble lists of pesticides and inert ingredients of pesticide formulations, cosmetic ingredients, drugs and excipients in drug formulations, food additives, and chemicals in commerce as inventoried under TSCA Select a sample from each category using systematic sampling with random start (see Table 1) Randomize the substances in each category of the sample ' 1 ' Apply a screen for minimal toxicity information (see Table 3) to the randomized chemicals in each category of the sample, in sequential order of randomization , On the basis of the screen, select a subsample of predetermined size from each category consisting of substances with minimal toxicity information that appear first in the randomized sample (see Table 1) FIGURE 1 Process used to draw sample and subsample from select universe 37

OCR for page 33
PESTICIDES AND INERT INGREDIENTS OF PESTICIDE FORMULATIONS The NIH/EPA Chemical Information System (CIS) lists of registered active pesticides and inert formulation ingredients were used. CIS is a collection of data bases with computer programs to search them. The list of registered active pesticides contains 2,483 entries, including substances that at the time of inclusion were for experimental use or that were analogies, salts, or acids of other substances in the list. Of the 2,483 entries, 265 were duplicate substances on the inert-formulation- ingredients list, leaving 2,218 unique substances. (EPA has versions of the list that were shortened by clustering chemicals that have similar structural backbones, that are salts or acids of a given chemical, or that are closely related analogies. For the purposes of the sampling methods in this study, it was important to maintain the integrity of each substance by means of its own identity. Therefore, the expanded list of 2,483 entries was used.) The list of registered inert ingredients contains 1,132 ingredients that are present in pesticide formulations, but have no claim of pesticidal action (not necessarily implying that they might not have adverse ef feats on human health) ; because of the potential for human exposure to pesticide-formulation ingredients, the inert ingredients were included in the select universe. Of the 1,132 entries, 265, as indicated above, were duplicate substances on the active- pesticides list. Consideration of duplicate substances is described later in this chapter. COSMETIC INGREDIENTS The list of the Cosmetic, Toiletry and Fragrance Association was used. It contains the names of 3,410 ingredients used in cosmetic formulations. Entries were made in three groups at different times. Each group is arranged alphabetically, so the list contains entries in three alphabetical cycles. DRUGS AND EXCIPIENTS I N DRUG FORMULATIONS The FDA Bureau of Drugs provided entries from its chemical-ingredient dictionary as it existed in October 1982. This included all nonproprietary prescription drugs (879) and nonprescription drugs (218) on the U.S. market in October 1982, as well as excipient substances used in prescription-drug formulations (717) and excipient substances used in nonprescription-drug formulations (284) at that time. This total list of 2,098 substances was purged of duplicates--e.g., nonproprietary prescription drugs that are also nonprescription drugs--for a final list of 1,815 substances. It did not contain substances under investigative new drug (IND) status at that time, substances not approved for use in the United States, drugs for veterinary use only, chlorofluorocarbon aerosol propellants used with drugs, or color additives used in drug formulations. Color additives are subject to certification under 38

OCR for page 33
separate provisions of the Food, Drug, and Cosmetic Act and were largely accounted for in the food-additives list (see below). Because of duplicate removal and specific exclusions, the list used in this study may be smaller than that reported in other sources. Furthermore, other lists of international scope may have different sizes and contents, because different criteria were used in placing substances on them. FOOD ADDITIVES The FDA Bureau of Foods chemical dictionary was used. This dictionary contains 19 chemical sorting codes. Six were used to make up the list of food additives from which the sample was drawn (see Table 2~. Later, it was found that a code containing 90 animal-drug additives from the food-additives category of the select universe had been excluded. The drugs in this code have veterinary applications for animals consumed by humans. Metabolites of these drugs are contained in the chemical-name code. Because they constituted a small fraction of the number of entries in the food-additives list, the probability of their selection for the sample would have been small, so this omission was not corrected. Cosmetic ingredients and drugs in the Bureau of Foods dictionary were specifically excluded from the list used to draw the sample, because they were contained in the lists of the cosmetic-ingredient and drug categories. The six components of the dictionary used provided a total of 8,627 entries, from which the sample was drawn. An undetermined number of these entries were altered forms of food additives that may appear in foods, even though their presence has not been confirmed by FDA. These substances, termed "theoreticals" by FDA, are possible products of known chemical pathways. They are not identified by any special designation and could not be removed mechanically from the list before sampling. The list was alphabetical, with an added minor portion of substances whose names began with numerical prefixes; this portion was organized according to ascending value of the numerical prefix. The alphabetical listing precluded separation of ingredients into code categories. Four of the six categories (direct, indirect, color, and flavor additives) implied categorization by use, but the fifth (GRAS) implied a decision by FDA under a "grandfather" clause that all substances in this category were safe for human use. It was an express desire in this study not to presume the degree of toxicity of any substance (such as those in the GRAS category), so that sampling could be based strictly on statistical premises and tenets. Therefore, the alphabetical integrity of the list was maintained, and entries in all six categories were selected by the prescribed random selection. CHEMIGALS I N COMMERCE The TSCA Inventory of chemicals~in commerce gave rise to special problems, because its construction was not restricted by specific use or 39

OCR for page 33
TABLE 2 Six Cedes of Chemical Classification in FDA Bureau of Foods Dictionary Used to Form Food-Additives Category from Which Food-Additive Sample Was Drawn Coded Fraction of Dictionaryb Direct food additives Indirect food additives Flavors Colors GRASC substances Chemical name sd Total Oe 015 0~042 O. 063 0~003 0~029 0~174 O .3;25 a Substances in 13 codes of Bureau of Foods chemical dictionary were excluded from list. 13 codes were animal-drug additives, food additives, biologics, cosmetic-label ingredients, cosmetic substances, indirect food additives (temporary file), drugs for human use, industrial chemicals, pesticide chemicals, and trade names for food additives, human drugs, pesticides, and veterinary drugs. As of February 4, 1981, when Bureau of Foods dictionary contained 25,401 preferred terms. Figures are fractions of substances in corresponding code files. c Generally regarded as safe. d This category contained substances that were included in food-additive petitions by manufacturers. Included were agents that (1) were awaiting assignment to more specific category, such as one of first five of table, or (2) were not assigned to a category, because they were intermediate products, impurities, or related compounds of safety interest only. 40

OCR for page 33
class, such as drugs, but rather was based on the amount of each chemical produced during 1977, as reported by manufacturers and processors. The Inventory contained (1) substances classified into 10 production ranges; (2) a group on which production data were absent; (3) a group on which production data were inaccessible to the general public, because of manufacturers' claims of confidentiality; (4) a group that was not produced during 1977, the year for which production data were amassed to assemble the Inventory; (5) a group used in processing of other substances (as opposed to their own manufacture), on which production data were not obtained; and {6) a group manufactured by industries that were not required to report production data under the terms of TSCA. Availability of production data in the Inventory was thus not uniform. This problem was exacerbated by other circumstances surrounding the Inventory's construction: Production volumes were reported in ranges too wide to permit accurate summation of volumes of all reporting manufacturers. Furthermore, manufacturers were not required to report a substance's production at plant sites where production began after the time of reporting. As a result, the indicated 1977 production of a given substance may have a large error. Under the terms of TSCA, an unknown number of chemical manufacturers, such as small businesses, were not required to report that they were producing a given substance. This introduced errors of unknown size in the production data in the Inventory on an unknown number of chemicals of unknown identity. - Processors and users were not required to report. EPA has estimated that, if this had been required, an additional 750,000 report submissions would have resulted without substantially increasing the number of substances. me total production of petroleum products and related substances is not accurately reflected in the Inventory. Although, on the basis of reported production volumes, gasoline is the leading substance and most of the next 10 high-volume substances are also petroleum products, some agents that are major fractions of mixtures are not reported as individual chemicals, but rather as parts of mixtures (e.g., benzene in gasoline). O Over 85,000 submissions of volume data from manufacturers were not verified by EPA. A given substance may be listed more than once with different CAS Registry numbers in the Inventory. About 75% of the known production data are on WCBs (unknown, variable composition, complex, or biologic products), such as petroleum products. 41

OCR for page 33
There are no production or use data on natural substances, such as asbestos, although these substances are listed in the Inventory. The TSCA Inventory was categorized by the Committee on Sampling Strategies into substances of which 1,000,000 lb or more was produced during 1977, substances of which less than 1,000,000 lb was produced during 1977, and substances on which production data were inaccessible or absent. Because of the large errors in production volumes tabulated in the TSCA Inventory, the sizes of the categories should be thought of as rough approximations. At the time the sample was drawn, the TSCA Inventory contained 48,523 usable entries distributed among the three categories, as shown in Table 1. CONSIDERATIONS IN DEVELOPING THE SAMPLING STRATEGY The development of a sampling plan for a complex study requires balancing of various competing objectives. For example: Will the most important inferences apply to the whole population, to independent segments of it, or to comparisons among segments? Can the sampling frame be defined in a way that is simultaneously precise and focused on the real objects of inquiry? What part of the total effort should be devoted to preparation of the lists from which the sample will be drawn? Does the difficulty of data collection vary substantially from one population member to another, and, if so, should the variation be used to reduce total costs (or expand sample size within the constraints of available resources)? In the present case, these and similar questions were particularly acute, because of the great cost and effort required for each substance to be subjected to full investigation and assessment. The Committee on Sampling Strategies believes that the sampling plan it adopted satisfies the competing demands and constraints efficiently. However, two issues required additional discussion: o The partition of the sample among the various lists and sublists. m e handling of interlist duplication. The committee determined that the sample should provide estimates for each of the seven categories. The sample sizes were then assigned to the categories with the aim of yielding estimates with roughly the same precision for each category. 42

OCR for page 33
In light of similar considerations, the Committee on Sampling Strategies regarded the 15:15:15:15:40 distribution (pesticides and inert ingredients of pesticide formulations:cosmetic ingredients:drugs and excipients in drug formulations:food additives:chemicals in commerce) in the subsample of 100 as illustrated in Table 1 to be a reasonable allocation of effort. The Committee further decided that a 10:10:20 split of the 40 substances from the TSCA Inventory would allow at least minimal inferences regarding specific subcategories, roughly in proportion to the expected need for information on their toxicity. These samples are probably at the lower limit of sample sizes that are usable for the present purposes; making some groups larger at the expense of other groups would have eliminated the latter from separate consideration, although they would still have contributed to inferences regarding the whole select universe. Some substances appear on more than one of the lists used to define the select universe. Such duplicates could have been identified in the whole universe or only for substances in at least one of the subsamples; and once identified, they could have been left in place or removed. It would have been possible to remove duplication among lists, for example, by establishing a priority order among the five lists and assigning each duplicated substance to the highest priority list of which it was a member. However, this would not have produced unbiased estimates for separate lists without additional steps to flag each substance on each list according to which of the other lists it was on. Careful matching of the entire lists would have involved little statistical advantage, but much effort. The committee also considered a modification of this approach in which only compounds in the sample would be matched against all lists for the purpose of describing the select universe. The advantage of having precise, rather than estimated, numbers of duplicates was judged not to offset the extra effort of matching the entire lists. Therefore, duplicates remain as they appear in the lists, and statistical methods were developed to reflect and adjust for this decision. The Committee on Sampling Strategies chose an approach to gain an understanding of the probability of the presence of information of adequate quality to enable assessment of the health hazards of a select universe of substances. Selections were made on a statistical basis without specific or deliberate reference to substances to which humans are primarily or extensively exposed. SAMPLING STRATEGY SAMPLE SIZE Both the sampling plan and the sample size were substantially affected by the large resource investment needed to study even a relatively small number of substances. Investigation of substances was a two-phase process: screening of the sample (rapid and inexpensive) to determine whether toxicity studies had been performed and reported, and a 43

OCR for page 33
detailed search (resource-intensive) for information on the subsample. It appeared that no more than 100 substances could be assessed in the subsample, and preliminary studies suggested that a sample size of approximately 700 would produce at least 100 substances in the subsample for the later detailed study. A sample of 700 was therefore selected to estimate the proportion of substances on whose toxicity there is sufficient published material to warrant further study. In the course of the study, two adjustments were made in the size of the sample, and that resulted in a subsample of 675. In the first adjustment, the 50 substances selected for each of the cosmetic-ingredients and food-additives categories were insufficient to produce the required 15 substances that would pass the minimal- toxicity-information screen for the categories of the subsample. (Minimal toxicity information is described later.) Therefore, 50 more substances were selected for each of the two categories in a random fashion identical with that used to select the first 50. In the second adjustment, the 250 substances selected for the category of chemicals in commerce with unknown or inaccessible production levels far exceeded the number needed to produce the 20 for the subsample. Therefore, this category was reduced by a random procedure to 125 substances. All 675 entities in the sample were screened to identify substances on which there was minimal toxicity information. The 675 substances were distributed among the seven categories of the select universe in a proportion of 50:100:50:100:125:125:125, representing pesticides {active ingredients and registered inert ingredients of pesticide formulations), cosmetic ingredients, drugs and excipients in drug formulations, food additives, and the three production categories of chemicals in commerce (see Appendix A). The available resources were enough to examine only about 100 substances on which there was at least minimal toxicity information so that their testing quality could be assessed. The Committee on Sampling Strategies faced the task of allocating the 100 substances among the seven categories of the select universe. Two approaches were considered in dividing the 100 among the seven categories: Make each of the seven subsample categories large enough to provide usable information on its corresponding category of origin. O Allocate category sizes with some appreciation of the relative size and importance of each list to NTP. The Committee on Sampling Strategies recommended the first approach, and the subsample of 100 with at least minimal toxicity information was drawn according to the proportion of 15:15:15:15:40. Furthermore, on the basis of the division of the chemicals in commerce into three production categories (125 substances produced at 1,000,000 lb/yr or more, 125 produced at less than 1,000,000 lb/yr, and 125 with unknown or inaccessible production), the 40 chemicals in commerce in the subsample were selected in a proportion of 10:10:20, respectively. The 44

OCR for page 33
sizes of the categories in the select universe, the sample, and the subsample are presented in Table 1. The contents of the sample and the subsample are presented in Appendix A. SAMPLING PLAN A stratified sampling plan was used to control the composition of the sample with respect to the seven categories in the select universe. This was critical in attaining the required sample sizes for all categories. For example, although the list representing chemicals in commerce, the TSCA Inventory, is larger than the other lists, it has a disproportionately small number of substances with toxicity information, compared with the other categories. Regulations mandate toxicity information associated with the use of substances in the other categories before registration and marketing. Variations in the toxicity data bases on substances in the seven categories resulting from differences in degree of exposure to them were considered by the Committee on Sampling Strategies to be a second reason for choosing a stratified sampling procedure. By stratifying according to these seven categories, it was possible to specify an adequate number of substances from all lists. The sample was drawn by using the same procedure seven times, once for each category. A random start was used to select the first sampled substance from a list. The rest of the substances in the category being sampled were drawn from the remainder of the category's list at equal intervals calculated to provide precisely the sample size needed. Thus, the sample was drawn from all parts of each list. This procedure, called "systematic sampling with a random start," is standard in such applications. me interval varied according to the sizes of the lists, because unequal sampling rates were necessary to achieve the required sample size. Each substance was selected according to its position on the list from which it was sampled. Because probability sampling was applied to each list, the sample was still statistically valid. Although the sample was an unbiased probability sample, it did not conform to the kind of sample design that lends itself to simple existing methods of variance estimation (for example, the committee considered whether periodicities in the lists might invalidate the sample). A systematic sample was selected, rather than a simple random sample, because it is easier and cheaper to select, and it will almost always have lower sampling errors. However, the Committee on Sampling Strategies paid special attention to this aspect of the sampling method and concluded that the effect of the systematic sampling was negligible and that, for variance purposes, the sample was obtained by a simple random process within seven strata. SCREENING THE SABLE FOR MINIMAL TOXICITY INFORMATION After the selection of substances for the sample, each entry in each category was assigned a random number. These numbers, each with its 45

OCR for page 33
assigned substance, were numerically ordered within each category, to provide seven randomly ordered samples totaling 675. The sampling plan called for a screening of the substances in the sample for minimal toxicity information to obtain the subsample of 100 substances. This , served two functions: In the assessment of the adequacy of toxicity-testing protocols by the Committee on Toxicity Data Elements, screening precluded the appearance in the subsample of a substance on which there was either no toxicity information or too little for use by that committee. - In the ensuing exhaustive search of the literature for all available toxicity information, screening obviated exhaustive searching for literature when it was nearly certain that there was none. me criteria for minimal toxicity information presented in Table 3 were delineated by the Committee on Toxicity Data Elements. That committee recommended that a combination of toxicity assays as described in the table--acute, chronic, subchronic, genetic, and reproductive or developmental, each including human and animal studies--be considered to constitute minimal toxicity information. During this critical screening process, two conflicting criteria needed to be satisfied: A sufficient body of toxicity-testing data on each substance in the subsample of 100 was desires, so that the committee members would be able to perform an analysis. The number of substances that could be screened as candidates for the subsample was constrained by available resources. Committee members used their collective experience to establish requirements for the amount and type of toxicity testing that would provide minimal toxicity information for each category of substances. These requirements, presented in Table 3, were varied to meet what should be existing toxicity information for categories with differing intended uses. By the terms outlined for each category in Table 3, the substances were listed in random order and, in that order, subjected to the screen for minimal toxicity information. Each category of the subsample contained the required number of substances that, according to the random order, were the first to pass the screen. The number of substances that were screened to find the number needed for the subsample varied among the seven categories (see Appendix A). Substances that the literature search revealed to have less information than required by the standard for minimal toxicity information were not allowed into the subsample. Frequencies by category for the select universe, the sample, and the subsample are presented in Table 4. Note that within each 46

OCR for page 33
~4 o u c A 0 3 C) a' 0 .. - .,. Ma o o . - L' o c .,4 ..~ x ED e ..~ c A: o U) c D A ID o Q. U. qua O C 4 U) C ,. - - 0 ,. 4~ CO . - ~ ~4 O: 0 D :~ m ~ ~ 0 En ~ 0 C0 a)c -0 ~- .~. X In pa ~a e :1 ~ C C 0 a ~- ~ 0 4~ c .- . - 4J ~ Q. 0 0 C 0 - ~ ~ C 0 P4 c o qua o D 0 of . - :' E o 0 C ~ U . ~ ~ 4J ~ 0 C ~ cn 41 ~* iC ~c >, ~ 1 45 0 D C ~ I - ~ C O - C ~: - :~ [q O - _I "l C C ~ :h, 1 . - X ~ O ~ ~' - ' - ~ I ~s0 e ~ m ~" ~ ~ ~ ~ ~ ~ C ~ U ~ ~ ~r~ 1 ~ ~ ~X X J~ ~ ~U ~ O C N . - ~ ~O O X ~ U]- ~ ~1 ~4~ p~ J~ O O .- t:C ' ~ V O 0~ - C ~O ~ ' ~ ' - S~ ~ ~ ~ 0 ~ :~ ~ ~S ~ 1 ~ C O ~ X 4J C ~ ~ O ~ C ~ ~ O O ~ ~ ~ ~ ~D -~ ~ 0 ~= ~4 C t): ~C 01 U) 47 c .,. c ., E 0 1:5 ~ 0 S - O a' c C~ O S V C C - ~ C n :, C ~1 a, ~ 3 0 U] C e ~ :3 s s - 0 3 ~q C ,C e ~ C ~ 0 X C 0 C c ~- 0 0~4 o J: e 0 a~u~ c ~4 ~4 :1 a 0s o Ll e ~a .~. :' U' a, 0 ~ ~ 4 e 0 c . ~ s w o ~n c .,. o ~ s .~ 0 c s o D E e ~ a, O D C 0 D 0 0 S ~ 4 S V ' 0 cn a4 :>' ~ 4~ 4~ :~ :o ~ 0 JJ 0 ~ 4 4 4 a' ~ S ,C C C O O J~ C C 0 0 4 4 D4 o Ll D e a, e e u Q C D :3 U] 0 U C 0 >' ~ ~ 0 4J U] 4 ,. Q. e U) n 0 4~ o 4 De e e D S 0 a, 4 :>' O ~ 0 S JJ W ,C .-1 S o :- :> C C C C O O 4~ ~ C C U) U) ~ ~4 0 U) 0 ~ U] ~o 0 0 ~ ~ O ~ ~ ~ ~ .,, ~ ~ ~ ~ 3 C ~ C C C C O D O O O O 0 ~ ~ - - ~ - ~ ~ ~ 4 a' ~ e ~ ~ ~ ~ 0 E E E E E 54 Ll ~ O O O O O ~4 ~ ~ ~ ~ ~4 s: C ~ C C C C E U ~d D

OCR for page 33
U1 In ~ ] ~ Q U) U] . U] n :D a In o U) N ~1 CO a) UO or EN cn U] a, Q In U] A' O S ~ Al 1~ lo O ;) .~ ~ O V - O ~ ~ H C) o ED ,4 U] ,' O ~ z In ski o ;) U] ,4 Lo ~O O O ~1 ~1 ~1 -~1 Lr, ~d' ~ ~ L`1 ~ O O O ~ ~ O UP O ~ ` ~Cal ,~1 ~1 ~ ~1 ~1 ~'A ~a' O ~1 11 of .n O ~O ~1 ~ ~1 ~ ~1 In 0O a o .. Y U] U] V] ~ O U] ~ ~ `S Q Q O - to '0 ~ ~X ~ O O O ~ '~ ~ ~ - U] ~ ~ O ~O S: ~ i ~ ~ ~ ~ ~ ,4 ,1 ~ tO ~IlS ~ ~ ~ ~ - O -4 .~ a) : ~a.) u~ 5, - ~ ~ ~ ~ Q. ~ ~ , ~ O ~ a., ~ I v a, ~ ~ 0 0 ~ ~0 S - O ~ ~;) 48 l L~ l U' E~ ,1 o o w ~q u] o ~ 1 ~ o ~ o a) 0 s 3 o a U] Q ,1 ,' o o .D O O S~ a' 3 o o U] 0 a X al o O S 1 O 0 i4 ~, .,~ ,1 U] .,, . a) 0 - :3 U] O .,~ - -

OCR for page 33
category the number of substances selected in the sample differs from the total number of substances used for later analyses. The differences are due to the overlap among lists constituting the select universe. For example, 106 pesticides and inert ingredients of pesticide formulations were used in the analysis of the sample. They included 50 substances that were selected for that category from the lists of pesticides and inert ingredients of pesticide formulations and an additional 56 substances that were selected from lists for the other categories but found to appear also in the lists of pesticides and inert ingredients of pesticide formulations. All 56 substances and some of the 50 had higher probabilities of selection than substances that appeared only on the list of pesticides and inert ingredients of pesticide formulations. The weights used in constructing the estimates that are based on the whole sample took these various probabilities into account to produce unbiased results. The statistical procedure used to prepare estimates is described in Chapter 4. Three chemical indexes were used in the screen for minimal toxicity information: CIS/Structure and Nomenclature Search System (SANSS), the National Library of Medicine's Chemline, and the Chemical Abstracts Service's Chemname. These indexes collectively contain over 5 million unique chemical substances by CAS Registry numbers, with synonyms and trade names for each chemical. The Registry of Toxic Effects of Chemical Substances (RTECS) and the Toxicology Data Base (TDB) offer toxicity data extracted from published research findings. TOXLINE houses 11 subfiles, including those generated for chemical-biologic activities, air pollution and industrial hygiene, toxicity bibliography abstracts on health effects of environmental pollutants, pharmaceutical abstracts, pesticide abstracts, Environmental Mutagen Information Center and Environmental Teratology Information Center files, and the toxicology section of Chemical Abstracts. These contain literature from 1965 to the present. For literature published from 1950 to 1965, the NRC Toxicology Information Center (TIC) card catalog was searched manually. The limited-search strategy used in screening had three steps: CIS/SANSS was searched to find alternative names of each substance in question and to point to other data bases that might contain information on that substance. If CIS did not provide this information, the National Library of Medicine's Chemline and the Chemical Abstracts Service's Chemname were searched. Once the location of available information was ascertained, the following sequence was implemented in an attempt to acquire minimal toxicity information on a substance: -- RTECS and TDB were searched for basic toxicity information (skin irritation, eye irritation, LD50, LCso, TDLo, etc.~. 49

OCR for page 33
-- If the minimal information requirement was not met by searching RTECS and TDB, the National Library of Medicine's TOXLINE was searched for information on acute, chronic, subchronic, genetic, and reproductive or developmental toxicity. -- If the requirement still was not met, the chemical name was searched in the card catalog of TIC. -- If the requirement still was not met after all preceding parts of the search strategy, the books and other resources of TIC were used--at least two appropriate toxicology reference books in each case. In this manner, 100 substances constituting a subsample with minimal toxicity information as defined by the Committee on Toxicity Data Elements were selected from a larger, randomly ordered, stratified sample of 675, which was itself a product of a large select universe of 65,725 listings. After selection of the 100 substances for the subsample, the screen for minimal toxicity information was applied to the remainder of the 675 substances in the sample. Thus, all statistical analyses of the sample included the data on all 675 substances. Under the direction of the Committee on Toxicity Data Elements, the subsample of 100 substances was then subjected to a comprehensive literature search for the assessment of the quality of toxicity testing by that committee. l 50