Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
THE MEDICAL FOLLOW-UP STUDY Men who are selected to serve in the Army are, in general, in better physical and mental health than their peers. Because their later health would also be expected to be better than average, it is inappropriate to compare their health and life experiences with those of the general U.S. male population. One option for this report would be a purely descriptive presentation of the findings derived from hos- pital records and a questionnaire. But the men exposed to the various chemicals differed in several ways, such as in age and date of testing, so a simple presentation of outcomes could be misleading. To permit useful conclusions to be drawn about the experience of men in the different chemical-group tests, comparative statistics were developed on the basis of observed and expected life experiences of comparison groups. ANALYTIC PROBLEMS _ The testing of chemical agents in human subjects at Edgewood began in 1955 and continued for some 20 years. The main objective of these tests was to determine effects of various chemical agents on the ability of test subjects to function effectively in a military situa- tion. It was not anticipated that any late effects would occur. In fact, two characteristics of the testing program make the demonstration of late effects extremely difficult: · Selection Bias: The volunteers received careful physical and mental screening examinations for contraindications to the planned tests. The health of a volunteer helped to determine the type of test in which he participated. The more healthy men were exposed to the active chemicals, and the less healthy were used as controls and in some cases tested equipment without being exposed to chemicals. Such selection bias means that the men not exposed to chemicals would be expected to have more illness; therefore, the likelihood of discovering effects in them (whether early or late) due to the treatments would be smaller. o Multiplicitv of Chemical Exposures: For the sake of efficiency, many volunteers were used in two or more tests. If a test substance produced detectable long-term adverse effects in a man who was also ex- posed to another substance, it could be difficult to ascribe the effect to the first substance alone, especially if many men were tested with both substances. 3
The Committee had many discussions on how best to evaluate the effects of the test chemicals, given the lack of suitable control populations and the multiple exposures of some of the men. It was decided that two comparison groups would be used. The first group included the subjects who received no test chemicals--the "no-chemical-test" (NCT) group. Although these men met the requirements for military service, they did not meet the rigorous standards demanded for exposure to chemicals. A second comparison group consisted of subjects who were tested with chemicals other than those being evaluated in a particular statistical comparison--the "other-chemical-test" (OCT) group. Suggestions to use additional comparison groups from other populations as surrogate controls were quickly turned aside, because the factors involved in selecting sub- jects for the Edgewood tests were not well understood and would involve variables that were not available for analysis, such as race, religion, socioeconomic status and characteristics related to subjects' desire to volunteer for the tests. Several other problems that make simple comparisons between exposed and unexposed groups difficult are discussed below. DIFFERENCES IN AGE AT FOLLOW-UP Posttest experiences of groups of volunteers cannot be compared solely by comparing numbers of events experienced; one must also take age differences among the test groups into account. Volunteers in the early test years were about the same age at the time of testing as men tested later (Table 1~. By 1984, the participants in the early tests had a much longer posttest period than the most recent participants; men tested more recently are younger now than-men tested earlier. Thus, those tested earlier are more likely to have had experiences that are more frequent with advancing age, even in the absence of exposure. The testing program spanned a period of some 20 years, and the chemical agents tested changed with time (Table 2~. For example, the testing of LSD derivatives was concentrated in the early years; more than half the doses had been administered by the end of 1959. At the other end of the testing period were tests of FDA-approved drugs, innocuous chemicals, and control substances. One can therefore expect health problems more common to the later years of life to have occurred in a much higher proportion of men in the LSD tests than of men who received approved drugs, innocuous chemicals, or control substances, even if LSD and its derivatives do not promote such health problems. Because of the age differences, comparisons between groups were adjusted for current age. 4
MAGNITUDE OF DETECTABLE EFFECTS OF CHEMICAL EXPOSURES How small a real effect can one reasonably expect to detect? The answer to this question depends on several factors: · The number of subjects exposed to each chemical of interest. This number is fixed for each chemical tested. · The number of subjects available for comparison purposes. · The proportion of members of each chemical-test group and comparison groups on whom follow-up information is obtained. · The level of significance at which the null hypothesis of "no effect" will be rejected. The 5% level is proposed for screening purposes. For this kind of examination, a one-sided test of signifi- cance, directed solely at identifying adverse exposure effects, is appropriate. (In making such a test of significance, it is assumed that the chemical-test participants have not benefited from their exposure.) · The "power" of a study is the probability that a true effect of a given size, or larger, will be detected in that study and labeled "statistically significant." "Power" depends on how big the true difference is (or is expected to be) and how many persons there are in the exposed group and the comparison group. The larger the true difference, the more likely it is to be detected, hence the greater the power. Similarly, the larger the number of persons observed, the more likely that a given difference will be found to be statistically significant. In general, bigger true difference and larger study both mean greater power. Power calculations are shown in Tables 3 and 4 for comparison with two base line groups of participants: the NCT group, consisting of subjects who did not participate in any of the chemical tests (those exposed only to FDA-approved drugs or placebo substances and those who were exposed only to equipment); and the OCT group, consisting of subjects who were not exposed to a chemical of interest, but were exposed to other test chemicals. Like use of the NCT group, use of the ACT group can decrease the probability of detecting effects of exposure to specific chemicals. In both comparison groups, health status, including fertility, might be lower than in a true control population. That could be the case in the NCT group because the selection process included volunteers who were less healthy. It could be the case in the OCT group if exposure to chemicals other than those of interest in a particular test had independent health effects on those volunteers. Table 3 summarizes power calculations based on 1,058 men in the NCT group who responded to the questionnaire (discussed later), and
Table 4 summarizes power calculations based on a typical number of respondents in the OCT group. As an example of how to use these tables, assume that 1,000 exposed respondents are being compared with the 1,058 NOT respondents (Table 3~. Assume further that the background, or base line, risk among the comparison subjects is 10% (0.100) and that we wish to discover the true effect of exposure that results in an additional 1% (0.010) above the background risk (the resulting risk in exposed subjects is therefore 11%~. For 1,000 exposed (the NOT comparison group), the intersection of the column representing the 10% base line risk and the row representing a 1% risk increase due to exposure con- tains the value 0.183, which signifies that there is an 18.3% chance of reporting a significant increase among exposed respondents. Table 4 is computed in the same manner as Table 3; however, the size of the OCT comparison group is not constant, but increases as the exposed group decreases, and by the same amount. Tables 3 and 4 show that the probability of observing a signifi- cant difference decreases with reduced numbers of exposed respondents and with an increased base line risk. As might be expected, the larger the true increase in risk, the larger the probability that the increase will be detected and labeled statistically significant. MULTIPLE COMPARISONS Tests of statistical significance lead to statements concerning the probability that some observed difference could have been due to chance alone. Probabilities are computed as though only one signifi- cance test, or comparison, had been conducted. However, one is seldom faced with a single-comparison situation. As the number of compari- sons increases, so does the probability that chance alone will lead to finding "significant" differences. In this report, the five chemical classes were analyzed separately as seven groups; the psychochemical class was analyzed separately for those exposed to LSD, Sernyl, and cannabinoids. Furthermore, the data on the men who were exposed to chemicals of only one class were analyzed separately from the data on those exposed to chemicals of that class and later to chemicals of another class. Therefore, in the following analyses, 14 chemical-test groups (seven with single chemical-class exposures and seven with multiple chemical-class exposures) are compared with two baseline groups on at least 27 end points (outcomes), for a minimum of 756 comparisons (14 x 2 x 27 = 756~. In the absence of any real adverse chemical-test effect, it is to be expected that several differences will be declared "significant" at the O.01 level, and possibly even one at the O.001 level, assuming that the end points are independent. Collateral data must therefore be brought to bear before even findings that are statistically highly 6
significant (i.e., ~ < 0.001) can be said to demonstrate that a true effect exists. Some of these collateral data are based on determina- tion of whether a dose-response relationship exists, whether the association is biologically reasonable, and whether the effect can be explained by peculiarities in or differences between the comparison populations. METHODS OF STUDY This report is based on a follow-up of the 6,720 men who partici- pated in experiments conducted by the Army at the Aberdeen Proving Ground, Edgewood, Maryland, in 1955-1975. Follow-up information was obtained from four sources: · A tape file, provided by the Army, of admissions of partici- pants to Army hospitals during calendar years 1958-1983. This file made possible the examination of the hospital experiences of volun- teers between the time of test participation and separation from the service. · A tape file, provided by the Veterans' Administration (Vi), of admissions of participants to VA hospitals from 1963 to 1981. This file was used to examine the use of VA hospital facilities after sepa- ration from the service. · Responses to a health questionnaire mailed to men who were still living in 1984. · Information obtained from men who initially did not respond to the questionnaire, but provided interviews when asked to do so in follow-up telephone calls. OBTAINING ADDRESSES The mail survey of health status was undertaken to obtain infor- mation related to health, social adjustment, and lifestyle during the posttest period. The survey was directed at 6,395 men for whom no claim for a burial allowance had been received by VA (Table 5~. Several sources of address information were used, including the VA compensation and pension files, records at the St. Louis Military Personnel Records Center, the National Institute for Occupational Safety and Health (which can obtain current address information from the Internal Revenue Service on persons with occupational exposure to possible health risks), and a commercial tracing agency. Addresses of 5,620 volunteers, or 88% of those not known to be dead, were ob- tained. Of these addresses, 624 (11%) were reported by the U.S. Postal Service to be incorrect, leaving 4,996 persons who are believed to have received the questionnaire; of these, 4,085 responded. The information on the total group of volunteers participating in the Edgewood studies follows: 7
Total participants Died before survey Could not be located Located, but failed to respond Located and responded No. TO 6,720 100.0 325 4.8 1,399 20.8 911 13.6 4,085 60.8 Table 5 shows the follow-up information on the persons in each chemical-test group. Presumably correct addresses were obtained for 4,996 men, or 78% of all living volunteers; this varied from 70% of men used in tests of LSD to 84% of those used in tests of the anticholinergics and cholinesterase Deactivators. There was sub- stantial uniformity from test group to test group in the proportion of persons responding to the questionnaire, once the participants were located. The response rate, 82% of all participants, was slightly higher among the participants in the chemical tests (81-84%) than among the NCT group (78%~. The likelihood of obtaining an address for a volunteer was greatest for those who participated in the most recent tests and decreased as the time since test participation increased (Table 6~. Valid current addresses were obtained for 83% of living men involved in tests conducted in 1970 or later, compared with 81% of men who participated in 1960-1964. For the men involved during the earliest years, before the replacement of the Army Service Number with the Social Security Number, the address yield fell to 57%. Knowledge of a person's Social Security Number is very important for successful tracing. The Internal Revenue Service failed to locate in its files only 171 (3%) of the 5,680 men whose Social Security Numbers were sent for tracing (Table 7~. The earliest test groups, Sernyl and LSD, had the highest proportions of "no record of SSN" responses; however, IRS still found in its files more than 95XO of the persons whose Social Security Numbers were known. The proportionate yield of good addresses of the living among the various chemical-test groups varied from 70% for LSD-tested subjects (early tests) to 84% of men exposed to anticholin- ergic chemicals (recent tests), as shown in Table 5. Addresses were obtained for 7670 of living subjects in the several test categories that constitute the NCT comparison group. CONTACTING BY MAIL Up to three mailings, 6 weeks apart, were made to a man at a given address. The first mailing contained the questionnaire and an explan- atory letter (Appendix E, Exhibit A). The second and third mailings included supplemental letters (Appendix E, Exhibits B and C) emphasi- zing the importance of a reply. An addressed, postage-paid envelope was enclosed with each mailing. The third mailing to a given address was sent by certified mail with return receipt requested as proof of delivery, and the enclosed letter invited the man to telephone the study director, reversing the charges, if he wished to inquire further concerning the survey. A mailgram (Appendix E, Exhibit D) was sent at 8
the time of the third mailing, advising the subject that the third request was on its way, in the hope of convincing him that the letter should not be treated as discardable mail. The mailing sequence ended on receipt of a completed questionnaire or a refusal. If the Postal Service returned a letter marked "address incorrect," another mailing sequence, to the next most likely address on file, was started. The priority for the use of addresses was (1) that obtained from the IRS, (2) that obtained from the VA compensation and pension file, (3) that obtained from service records, and (4) commercial tracing. The cutoff date for receipt of questionnaires to be included in the analysis was February 1, 1985. Questionnaires were received from 4,085 men, 64% of survivors (Table 8~. The response rate for men who were correctly located was 82~; 2Yo of located men declared their unwillingness to participate, and 16% of those located failed to respond. Those who failed to respond were telephoned, and when necessary visited, to determine the reasons for their not responding.* A total of 657 (74%) of the 891 nonrespondents were interviewed. The results of the interviews are discussed later. COMPARISON GROUPS The chemical tests were conducted by the Army for the sole purpose of studying immediate effects. The tests in which a volunteer parti- cipated were therefore determined by the tests being conducted at a given time. Volunteers participated in an average of 3.3 tests. Some were tested more than once with the same material, and some were tested with several materials. Only 43% of respondents in any chemical-test group were exposed to only a single class of chemicals. That is, most of the men were exposed to multiple chemicals from different chemical groups. Thus, comparisons of importance to the search for late effects are those between men who participated in tests involving a single class of chemicals and the 1,058 volunteers in the NOT comparison group. The NCT group includes: · 907 respondents who were not exposed to drugs (used mostly in tests of equipment). · 93 respondents who were exposed to FDA-approved drugs (58 dif- ferent). · 17 respondents who were exposed to relatively innocuous sub- stances (alcohol, caffeine, etc.~. · 39 respondents who were exposed to control substances (water, saline, bicarbonate of soda). · 2 respondents who were exposed to two of the above. * This survey of nonrespondents was conducted by the Research Triangle Institute, Research Triangle Park, N.C. 27709. _9 _
Evidence that the NOT group was composed of the less healthy volunteers dictated that other comparison groups be defined--groups less likely to have a built-in health bias. These groups, the OCT groups, consisted of all volunteers who participated in chemical tests other than those under immediate evaluation. Members of the OCT groups had received one or more of the chemicals. By definition, the OCT groups were likely to yield underestimates of possible effects of exposure to specific chemicals, because its members were exposed to other chemicals. Although the possibility of adverse effects due to other exposures must be considered, no such effects were noted when the OCT groups were compared. Table 9 is a sample of the computer tables used in the preliminary screening of the data. Multiple test-group comparisons (described below) were considered in deciding whether a given category of chem- ical appeared to have had an adverse effect on exposed volunteers. The categories shown in Table 9 are those used in the study of men exposed to anticholinesterase chemicals. Similar groups were defined for the other categories of chemicals. The column definitions are as follows: · A ALN (Alone): Persons who participated in tests of the category of chemical named in the table heading and no others (in this example, the anticholinesterases)O · A AND: Persons who participated in tests of the category named and in tests of at least one other category. · TTL A: The sum of A ALN and A AND (i.e., all persons exposed to the category named). · CNTRL: The no-chemical-test (NCT) comparison group defined earlier. · OTHER: All subjects in chemical tests except those involving the category named, i.e., the other-chemical-test (OCT) comparison group. · NOT A: All subjects (including the NOT group) minus volunteers who participated in tests involving the named category (i.e., the sum of CNTRL and OTHER). · TOTAL: All respondents (i.e., the sum of TTL A and NOT A). Each computer table consists of several panels representing cur- rent age categories in 1984 (the survey year) and a "total" panel representing the sum of all age groups. The responses to the various questions of the questionnaire are defined by the labels in the left margin. Table 9 deals with whether the volunteers requested informa- tion about the results of the study. The percentages are based on column totals for a given panel. -10-
Whether a class of chemicals might have had an adverse effect on exposed subjects was based first on inspection of the percentages of subjects with similar responses in each of the first five columns named above. This inspection was performed separately for each age group. If it appeared that the responses of those who were exposed to the chemical group of interest alone or in combination differed from the responses of the NCT group (control) or the OCT comparison group (other), the pooled data were compared with a more rigorous analysis that included adjustment for age differences and testing for statis- tical significance. Age was adjusted for by calculating the expected number of the given response for each age group. Expected numbers were based on the proportion of NCT subjects who responded in similar fashion. This proportion was then multiplied by the number of sub- jects in the exposed group to find the expected number of the given response for that age group. The expected numbers were then summed for all age groups to get a total expected number, which was compared with the total observed number of responses. A similar adjustment was made with the OCT comparison group. For assessment of fertility and other relevant variables, current marital status was also adjusted for. In these cases, the data were adjusted for age and marital status concurrently. That is, the popu- lation was divided into age groups, which in turn were divided into marital-status groups for purposes of determining expected values. Significance testing was performed with observed data for both the exposed and comparison groups. Each group was subdivided into age- specific groups, and the Mantel-Haenszel chi-square test procedure was used. 6 This procedure tests for differences between exposed and comparison groups by combining tests performed for each age group. For this purpose, the responses were reclassified in some cases into only two categories, i.e., dichotomized. For example, the total num- ber of children of each man was reclassified into no children and at least one child. In some instances, particularly with respect to occurrence of specific disorders, the total numbers of events were very small. For these comparisons, a Fisher exact test was used to generate the prob- ability of observing the number of events actually reported by the respondents, given the numbers observed in each of the two comparison groups. For example, assume that 10 of the respondents reported a particular disability, including 4 exposed subjects and 6 who were not exposed. The exact test provides the probability that the 10 reported events would include 4 or more exposed subjects (an excess as great as or greater than that observed). The response of interest in Table 9 is whether the volunteers asked to be informed about the results of the study. It is apparent that more men who were involved in tests of anticholinesterases asked to be informed of the results of the studies than NCT subjects. Of all the groups studied, the NCT group showed the least interest in the results. 11
RESULTS A total of 27 outcome variables (Table 10) related to the health, social adjustment, and reproductive experiences of the Edgewood participants were examined for each of the seven classes of chemicals tested. That led to 189 possible decisions concerning the existence of a chemical effect--27 decisions for each of the seven chemical groups. Only a few of these decisions were judged to have possibly a significant association between exposure and the outcome variable. The evidence leading to those judgments is presented in this chapter. In addition, detailed comparisons were made where there was some biologic reason to anticipate a specific response, even though preliminary examination of the data did not suggest an effect. ADMISSIONS TO HOSPITALS Admissions to Army Hospitals The Army provided a tape file of admissions of volunteers to Army hospitals in calendar years 1958-1983. This file does not cover the early test years, but the loss of information is limited to the early posttest years and to men who participated during the first 3 years of Army testing. The numbers of volunteers under observation during each posttest year were counted for each of the chemical groups of volun- teers; separation from the service terminated observation. 3 Follow- up of volunteers began at the time of testing, or in 1958 for those tested earlier, and numbers of person-years of observation were summed over successive posttest intervals. Rates of hospital admissions were estimated per l,OOO person-years of observation. Percentages of men hospitalized were based on the mean numbers of men being followed. Army hospital admission rates after test participation are shown separately for men exposed to a single class of chemicals (Table 11) and for men exposed to more than one class of chemicals at Edgewood (Table 12~; the two groups are mutually exclusive. These data, and those on the VA hospital admissions that follow, are reported for all volunteers, not only those who returned the questionnaire. For each class of volunteers, the need for Army hospital admission was greatest during the first 5 years of follow-up and decreased over succeeding intervals. However, neither table shows that the men in any of the chemical-exposure groups appear to have had more hospital care than did the NOT group. Hospital admission rates were more sensitive to the frequency of hospital care than were the percentages of men hospitalized. Numbers of men admitted to Army hospitals were compared for men exposed to only one chemical group (Table 13) and for men exposed to chemicals in more than one group, including the group of interest (Table 14~. The percentages of hospitalized men in each chemical-test group seldom exceeded that of NOT subjects more than trivially. No pattered of -12-
excess hospital use was shown for any chemical group when the one- class and multiple-class exposure groups were compared. The medical conditions responsible for admissions were examined in broad categories based on the Eighth Revision International Classifi- cation of Diseases (1967-1969), 26 referred to as ICD 8. Among the more frequently encountered diagnostic categories for all volunteers are respiratory disease (359 admissions, 16.1% of total), digestive disorders (260, 11.7%), infectious and parasitic diseases (202, 9.1%), musculoskeletal and connective tissue problems (178, 8.0%), and symptoms, ICD 8, codes 780-796 (162, 7.3YO). These five categories account for over half the 2,231 recorded Army hospital admissions of these volunteers since testing. Furthermore, these medical conditions are common in Army life. Conditions that might have resulted from the various chemical exposures resulted in few admissions, with rates similar to those for the NCT subjects. The medical conditions that might have resulted from exposures include malignant neoplasms; benign and unspecified neoplasms; -endocrine, nutritional, and metabolic diseases; diseases of the blood and blood-forming organs; mental disorders; diseases of the nervous system and sense organs; and diseases of the circulatory system. Several of these groups of disorders are examined separately below. Admissions to VA Hospitals Data on admissions to VA hospitals have been obtained from the VA for the years 1963-1981. Because only 15% of veterans' hospital-care needs are provided by VA,1 the rates of postseparation admission of men exposed to a single chemical group (Table 15) and to more than one group (Table 16) grossly underestimate hospital use by these par- ticipants. However, there is no a priori reason to suspect that those exposed to one chemical group are more likely to depend on VA for their hospital needs than another group. Thus, the rates shown for the sev- eral chemical groups are assumed to be comparable. Admission rates for various postseparation intervals show consid- erable variation between chemical groups. Although there are excep- tions, the NCT subjects generally experienced the highest admission rates. This relationship is seen clearly when the postseparation period is examined as one unit. The procedures used by the Army to select NCT subjects apparently had the effect of identifying, on the average, men more likely to use or require hospital care after dis- charge. In Table 16, the exposure group with the highest admission rate is the Sernyl group, but comparison with Table 15 shows that 38 of the 39 admissions were of men exposed to other chemicals, as well as Sernyl. The admission rate among those exposed to several chemicals including Sernyl was 57.0 per 1,000 person-years. However, because the total number of men, and consequently of person-years of observation, is so -13-
small for this group, the increase could easily be due to chance, rather than to exposure to Sernyl. Most of the increase occurred 5-14 years after exposure to the chemicals. Admissions for Selected Disorders Hospital admission rates constitute a good index of the need of the test participants for hospital care, but they can be distorted by the repeated admission of a few persons for a given condition. This problem can be eliminated by examining the numbers of men admitted to hospitals for selected conditions. Three categories of illness of particular interest for possible relationship with the chemical agents tested are malignant neoplasms (ICD-8, codes 140-239), mental dis- orders (ICD-8, codes 290-315), and diseases of the nervous system and sense organs (ICD-8, codes 320-389~. The numbers of men admitted to Army hospitals and to VA hospitals for these conditions for each chemical-test group are shown in Tables 17, 18, and 19. Two counts of "expected" admissions are also shown: those based on the experience of participants exposed to chemicals other than the group being considered for a specific comparison, the OCT group; and those based on the NCT group experience. Seven volunteers were admitted to Army hospitals for malignant neoplasms during their posttest period of service, of whom three were NCT subjects and four were participants in chemical tests (Table 17~. Chemical-test participants as a group experienced fewer admissions than were expected on the basis of the NCT experience. When each chemical-test group was compared with all other groups combined, the differences between observed and expected admissions for malignant neoplasm were in general no greater than might be expected owing to chance. All six cases of malignant neoplasms treated in VA hospitals, however, occurred among participants in chemical tests. Three chemical-test groups had more cases than expected: anticholines- terases, anticholinergics, and cholinesterase Deactivators. There is duplication among the cases associated with these three groups, because some of the subjects had been exposed to more than one chemical. Hence, the sum, nine admissions, is greater than the admissions to VA hospitals. The number of neoplasms among the anticholinesterase-exposed men was not significantly increased over the OCT expected number. When an exact test was performed with the four cases in the anticholinesterase group and no cases in the NCT group, the results were barely significant. An examination of the admission diagnoses revealed no consistent pattern of site involve- ment. There were two cases of cancer of the trachea or lung; one each of cancer of the lip, skin, and lymphoid tissues; and one myeloma. Mental disorders (Table 18) led to admission of 56 participants to Army hospitals and 109 participants to VA hospitals after separation Both the Army and the VA experiences clearly indicate that NCT subjects were more likely to have mental disorders requiring hospitalization -14-
than were the chemical-test participants and indicates that this group was not as fit as those who were given experimental chemicals. Because of the increased frequency of mental disorders among the NCT group, a more useful comparison for the chemical groups is obtained by using men tested with other chemicals to compute expected values. For no specific chemical group were more men admitted to Army or VA hos- pitals than expected on the basis of the NCT comparison group. In comparison with expectation that was computed from the OCT comparison group, there were slight excesses in four of 14 categories, but in no case was the excess statistically significant. Diseases of the nervous system and sense organs resulted in the admission of 43 participants to Army hospitals and 16 to VA hospitals (Table lay. Consistently more chemical-test participants were hos- pitalized for diseases of the nervous system than would be expected from the NCT experience, both while in the Army and after separation. The numbers of admissions for these disorders were significantly higher than three of the four expected values for the men exposed to LSD. Among the LSD-exposed volunteers, three were hospitalized for otitis media, two for diseases of the retina (not detachments), one for a cataract, and one for deafness. No Sernyl participants were admitted for this group of illnesses; Sernyl-exposed men constituted the only group with fewer participants admitted than expected on the basis of the NCT experience. Half the men admitted to Army hospitals for diseases of the nervous system and sense organs were first admitted within 5 years of testing, whether they were in chemical-test groups or the NCT group. When the distributions of specific diagnoses are compared among the test groups, no consistent pattern is seen that suggests that any of them can be attributed to the testing program. MAIL SURVEY Response Bias As noted earlier (Table 5), responses were received from 4,085 participants--64% of the 6,395 men who were believed to be living when the mail survey was conducted and 82% of the 4,996 men who were located. Because responses to the questionnaire survey were not obtained from all living men, it is necessary to compare the health of respondents with the health of those who were located but did not respond and of those who could not be located. The current health status of the three groups could not be evaluated directly, but it has been possible to examine the percentages of men still in the Army at successive intervals since testing and the percentages that were hospitalized during each interval (Table 20~. The upper panel of Table 20 shows the mean annual numbers of men who were still in the service during successive 5-year posttest intervals. The center panel shows the numbers of these men who were hospitalized by the Army, and the lower panel shows the percentages of hospitalized men among those still in service. -15-
The overall use of hospitals was highest during the 5-year period immediately after test participation. The proportion of participants hospitalized beyond the fifth posttest year was greatest among men who responded to the questionnaire, somewhat lower among men who failed to respond, and lowest among men who could not be located. These percen- tages suggest that men who could be the most likely to have health problems were most likely to respond. If a response bias exists, ~ therefore, it is in the direction of overestimation of current health problems of the living volunteers. Survey of Nonrespondents An attempt was made to interview the current nonrespondents (in person or by telephone) to answer four basic questions: ~ Did the man participate in the Edgewood testing program' · If sordid he receive any of the mailings addressed to him? . If so, did he respond to the questionnaire? · If not, why not? An arrangement was made with Research Triangle Institute (RTI), Research Triangle Park, North Carolina, to attempt to interview the 891 men who had not responded (20 of the nonrespondents were not identified in time to be included in the survey). The survey instrument used is shown in Exhibit E of Appendix E. Interviews were obtained with 657 nonrespondents, 74% of the total (Table 21~. The proportion of men interviewed from among the chemical- test groups varied from 71% for the irritant or vesicant group to 91% for the Sernyl group. Thirty-six nonrespondents (4%) refused to be interviewed--from none of the Sernyl group to 7% of the LSD group. RTI was unable to contact the remaining 198 (22%) of the nonrespon- dents (from 9% of the Sernyl group to 24% of the anticholinesterase group and irritant and vesicant group). These 198 included 74 participants who could not be located. The identification of participants among the nonrespondents was reasonably good; 643 (98%) of persons interviewed confirmed their having been on special assignment to the Aberdeen Proving Ground. Of the 643, 102 (16%) could not recall having received any of the three mailings (Table 22~. It is reasonably certain that the third letter was delivered, inasmuch as certified-mail receipts of delivery were returned for the nonrespondents. The letters probably were simply discarded after receipt. Of the 541 participants who remembered receiving the questionnaire, 62 (11%) claimed to have completed and returned it (Table 23~. Only 13 -16-
questionnaires were eventually received from these 62 men, including seven received after the cutoff (February 1, 1985~. The three most frequently reported reasons for not responding were lack of health problems to report (337 participants, 62%), concern about confiden- tiality (135, 25%), and judgment that the questionnaire was too personal (114, 21%~. There is no evidence that the men exposed to different chemical groups differed in their reasons for not responding, particularly where current health was concerned. Mail Survey Findings Age of Respondents. Although the ages of participants at the time of testing were similar among the exposed to the various chemical groups (Table 2), differences in the periods during which the different classes of chemicals were being tested are reflected in the ages of questionnaire respondents in 1984 (Table 24~. The mean age of the respondents in 1984 was 42.4 years, with a 9-year spread between the anticholinergic test group (mean age, 40.4 years) and the Sernyl test group (mean age, 49.1 years). Education of Respondents. The level of education of respondents varied little among those exposed to the various chemical groups (Table 25~. Median reported years of education, 13.3 for all respondents, ranged from 12.7 years for the Sernyl group to 13.7 years for the anticholinergic group; 2% of respondents failed to report their education. Family Relationships. Social status, living habits, and family composition of the various test groups were similar to those of both comparison groups. With one exception, the small observed differences disappeared when the results were adjusted for current age. The exception involves the reproductive experience of men who participated in tests of anticholinergics. Volunteers exposed only to anticholinergics and volunteers exposed to anticholinergics and other chemicals both reported having had significantly fewer liveborn children (stillbirths were not reported) after test exposure. In comparisons of numbers of children, men who did not respond to the question on children were assumed to have had no children. Most men who failed to respond to this item had never been married. Table 26 shows the numbers of children reported by men tested only with anticholinergics and with anticholinergics and other chemicals. The deficits in numbers of children of those exposed only to anticholiner- gics and those exposed to anticholinergics and other chemicals were the same: 1.81 children per respondent compared with an expectation of 2.01 children per respondent, based on the age-specific NCT-group reproduction rates. The expected number of men with given family sizes was also estimated on the basis of the age-specific reproduction 1 - 1
histories of the participants in tests of all other chemical groups--a mean of 1.95 offspring were expected per respondent. The deficiency of offspring among the anticholinergic test subjects increased when allowance was made for differences in current marital status (Table 27~. However, data were not available to control for duration of marriage after exposure. Allowance for age and education differences reduced the expectation to 1.93 children. An examination of the ratios of observed to expected numbers of men indicates that the anticholinergic- test subjects were less likely to report larger families (three or more children) than were the OCT subjects. The above analyses were performed with the total family sizes of the men, i.e., including children born before and born after the chemical tests. Approximately 16% of the children were born before testing or within a year after testing. Later analyses therefore used only children born later than a year after exposure. To estimate expected values, the rates were corrected for the ages of the men at the time of testing. That is, a man aged 20 at the time of testing could be expected to have more children born after testing than a man who was 35 at the time of testing. This would be true even if their current ages were the same. The anticholinergics (median year of testing, 1968) were among the chemicals more recently tested (see Table 2~. Because the mean ages of the men at testing were approximately the same for each chemical group, the current mean age of the anticholinergic-test group was the lowest (see Table 24~. There has been a general trend in the U.S. population toward lower birth rates and higher ages at conception.~5 Therefore, the apparent decrease in fertility could reflect this trend. Adjusting for the age of the respondent at time of testing does not adjust for such trends. The expected values presented in Table 28 are adjusted for age of respondents at time of testing and year during which the last testing was conducted. With these adjustments, the differences between observed and expected fertilities were sharply reduced when the OCT group was used as the comparison group. However, a deficit in the number of male children born after testing did not disappear after the corrections. When the NOT group was used as the comparison group, the findings were similar, i.e., no significant overall fertility difference and the deficit in number of male births remained significant (see Appendix D). Other information that might be related to the fertility of these men--duration of marriage after exposure, use of oral contraceptives, race, and socioeconomic status, etc.--was not available. A telephone survey of 50 fathers who reported having had three to five children was made to determine whether reporting had been accu- rate or whether men with large families underreported family size. Two additional liveborn children were discovered by the survey; a boy "pronounced dead at birth" and a girl born before test participation and therefore intentionally omitted from the questionnaire by the respondent. Six of the 163 reported children, two boys and four -18-
girls, lived only a few hours. Thus, the telephone survey strongly supported the conclusion that the reporting of liveborn children was accurate. Sex of First Child. The sexes of all the children born to the volun- teers exposed to anticholinergics and the children born to the OCT - group are shown in Table 28. There appears to be a decrease in male children born to the anticholinergic group. However, to evaluate the impact of an exposure on the sex of liveborn children, it is believed more appropriate to consider only the first child conceived by the - volunteer after exposure, hereafter referred to as the first postex- posure child. These are the first children born at least 1 year after exposure. If an effect exists, it should be strongest in the first postexposure child. Restricting these comparisons to the first post- exposure child born within a given period after the test--say, 1 or 2 years--would have yielded too few children for useful comparisons. A further reason for considering the first postexposure child, rather than all postexposure children, is that some epidemiologists feel that cultural and societal expectations might influence the proportion of children who are male in a completed family. (However, a mathematical "modeling of this phenomenon shows that it would have only a minor effect.) Among the first postexposure children born to men exposed to anticholinergics, 445 (48.7% ~ 1.7%) were males and 469 females. Of the first postexposure children of the OCT group, 714 (49.8% ~ 13 were males and 720 females. These values are not significantly different. Among the first postexposure children of the NCT group, 445 (54.9% 1~7Yo) were males, 370 females. This value is significantly differ- ent from both the anticholinergic-exposed group and the OCT comparison group. A national survey found that 51.2% of offspring were males.13 i5 For outcome characteristics, such as health, that are strongly affected by the selection of healthy volunteers at a given time, it is clear that comparing these volunteers withy U.S. male population is inappropriate. However, there is no reason why selection for general health at one time should affect the proportion of later children who are male. Therefore, it is appropriate to compare the NCT comparison group, which received no exposure, with this general population. The NCT has a significantly increased proportion of male children (p < 0.005), when compared with the national population. But the OCT group does not differ significantly in this regard. Because there is no a priori reason for the NCT group to differ from the national population, sampling errors might explain this extremely high propor- tion of males. Therefore, it was concluded that there was no clear indication that exposure to anticholinergics would affect the sex of later offspring in a significant way and that there were no animal or laboratory data to support such a finding. -19-
Respondents' Health. Several questionnaire items were aimed at deter- mining the respondents' health status. These ranged from a simple self-assessment of health on a four-point scale (excellent, good, fair, and poor) to the recording of illnesses requiring medical care over the preceding 5 years, confinements to bed for illness or injury during the preceding month, and impairments limiting ability to work and care for oneself. Responses to these items were strongly related to the respondents' ages, so adjustment for current age was particu- larly important. Of the 4,085 respondents, 25% reported their health to be excel- lent, 54% good, and 17Xo fair. Only 3% of respondents considered their health poor, and 1% failed to respond. None of the chemical-test groups, either alone or in combination with other groups, differed from either comparison group. When asked whether they had ever been told by health professionals that health problems they were experiencing were due to exposure to a toxic substance, 316 (7.7%) of the 4,085 respondents answered "yes." Only the men exposed to irritants and vesicants had a response pattern that showed a significant increase over either comparison group (Mantel-Haenszel test of "yes" versus "no" responses, ~ = 0.05~. The 538 men exposed to irritants and vesicants alone responded as follows: Reported health problem Observed due to toxic exposure No. % Expected No * % . No 464 86.2 485.3 90.2 Yes 47 8.7 37.4 7.0 Unknown 27 5.0 15.3 2.8 Total 538 538.0 * Based on the age-specific no-chemical-test experience. Much of the difference is due to an increase in "unknown" responses. But a review of the nature of the exposures of the volunteers exposed to irritants and vesicants and of a sample of NOT subjects who responded "yes" to this question clearly indicates that acute effects, mainly erythema and blistering during the immediate posttest period, were responsible for the difference. The irritants-vesicants class of chemicals produced no demonstrable excess of late effects. (Long-term health effects of men who experienced acute erythema and blistering are discussed later.) -20-
Half the respondents reported having sought medical care for a disease or illness during the preceding 5 years. There is no evidence that any of the chemical-test groups experienced an increased use of medical care. The irritants-vesicants group, the only one that differed significantly (n < 0.05) from either comparison group, reported a reduction in later use of medical care. Some 29% of respondents reported having been hospitalized during the preceding 5 years. The test volunteers as a group reported fewer hospital admissions than either comparison group. Hospital admissions of the anticholinergics group were significantly (p < 0.05) less than those of NCT subjects. Bed confinements of 1 day or more during the preceding month were reported by 20% of respondents. The experience reported by chemical- test participants tended to be more favorable than that of the NCT comparison group. None of the test-group differences from either comparison group was statistically significant. Health problems or impairments prevented 337 (8%) of the respon- dents from working and limited the type of work engagement for an additional 492 men (12%~. One test group, subjects exposed to anti- cholinergics alone (consisting of 353 men), reported significantly (p < 0.05) more men with work limitations, as shown below: Observed Ability to work No. % Expected (NCT group) No. % No limitations 259 73.4 279.7 79.2 Limits work 50 14.2 38.2 10.8 Prevents work 33 9.3 27.1 7.7 Unknown 11 3.1 8.0 2.3 Total 353 353 ~ 0 A review of the responses of anticholinergic-test and NCT men reporting work limitations indicates that no particular type of impairment was involved. Furthermore, there was no difference between the men exposed to anticholinergics and the OCT comparison group. No relationship to specific chemical type, route of administration, or dose was seen. The excess of reported work impairments among men exposed to anticholinergics most likely can be attributed to chance. Problems in performing household chores were reported by 376 (9%~: 159 (4%) reported that performance was "prevented," and 217 (5%) that it was "limited." None of the test groups differed significantly from either comparison group, nor were the directions of differences consistent. -21-
Problems in personal care were reported by eight respondents, and an additional 27 required help with their daily routines. No differ- ences between test groups and either comparison group were noted. Substance Use and Abuse. Four questions were related to the use and abuse of tobacco, alcoholic beverages, and drugs. Nearly one-third of the respondents never used tobacco regularly. The vast majority of those who smoked on a regular basis were currently smoking cigarettes; 1,493 respondents (37%) smoked at least one pack per day, and 167 cigar smokers and 130 pipe smokers smoked at least one cigar or one pipeful per day. Of the 2~,779 respondents who reported ever having smoked regularly (more than occasionally), 1,438 (52%) reported that they no longer smoked. None of the chemical-test groups differed from either comparison group in smoking experience, in terms of the number with a history of regular smoking or current smoking patterns. The drinking of alcoholic beverages daily was reported by 3,147 (77%) of the respondents; 68% drank beer, 44% wine, and 54% whiskey or other hard liquor. These figures are not different from that reported by adult males in a national survey, i.e., 69%.15 A total of 231 men (5.7%) reported having had drinking problems that required treat- ment. Similar drinking experiences were reported by the chemical-test groups and by the NCT group. Table 29 shows the reported use of selected substances of potential abuse. Most frequently reported was marijuana (37% of respondents), followed by "other narcotics, opiates" (34%), tran- quilizers (29%), amphetamines and other stimulants (18%), LSD (12%), barbiturates and other depressants (lOYo) ~ cocaine (10%), phencyclidine (3%), and heroin (2%~. With one exception, the use of the various substances shows no relationship to chemical-test group. The one exception is the LSD- test group (Table 30~. LSD-test subjects reported much more frequent use of the substance than the NCT group or the appropriate OCT group. Most of the excess use of LSD was among those who used it one to nine times, so much of the reported excess might have been due to the reporting of test exposures. However, nine subjects (2.8% of the total) reported having used LSD 10 or more times, compared with 1.2 subjects expected on the basis of the NCT group experience and 2.2 subjects expected on the basis of the OCT group experience. A few of the LSD-test subjects, therefore, appear to have continued to use LSD, inasmuch as no participant received more than four test exposures (or six test exposures, depending on the source of information). Perhaps continued substance abuse by the LSD-test group of soldiers explains in part their somewhat higher rate of questionnaire refusal to respond to the questionnaire for reasons of concern about confidentiality (Table 23). -22-
Groups of Special Interest. The responses of all 38 soldiers who experienced erythema or blisters as a result of exposure to mustard gas were examined in detail, because of the carcinogenic and mutagenic potential of this chemical (see Appendix A). None of the 38 mentioned cancer in response to the question of whether any health professional attributed a health problem to exposure (question 9) or the question of whether any disease or hospitalization had occurred in the preceding 5 years (question 13~. One person mentioned cellulitis as a recent health problem, but his physician had not attributed this to the expo- sure. The mustard-damaged skin site (shoulder) was in a different location from the cellulitis (lower leg). ; A second group of special interest consisted of subjects exposed to Sernyl, a purified form of phencyclidine. The primary focus for this group was whether any general health problems or mental disorders had developed as a result of exposure to this chemical. The follow-up responses of 48 men were received; this sample represented 60% of the 80 exposed men who were known to be alive at the time of the survey. In response to the question regarding their "overall health" (question 4), 35 men (73%) reported good to excellent health. The remaining 13 men reported health problems that bore no relation to their test experience in the Army. For example, five men reported "low back pain," three reported "hernias," and three others reported "prostate problems." No questions were explicitly related to the psychologic status of these men. However, there were only two reports of treatment for alcoholism, infrequent reports of drug use, and no unusual reports regarding employment, marital, and family status. There is no basis to infer above-average frequency or intensity of psychologic problems among these men. -23-
