1 15 9 https://www.nal.usda.gov/exhibits/speccoll/files/original/afc503475f8dac9b1e8778f4ed94b544.pdf b83094460edb1cb45fadfba3dadf3216 PDF Text Text Item ID Number °1786 Author Breslin, Patricia Corporate Author Report/Article Title Proportionate Mortality Study of US Army and US Marine Corps Veterans of the Vietnam War JOlirnal/BOOk TltlB Journal of Occupational Medicine Year 1988 Month/Day Ma Color a Number of Images v 8 Descriptor Notes Monday, June 11, 2001 Page 1787 of 1793 �Proportionate Mortality Study of US Army and US Marine Corps Veterans of the Vietnam War Patricia Breslin, ScD; Han K. Kong, DrPH; Yvonne Lee, MSc; Vicki Burt, ScM; and Barclay M. Shepard, MD The patterns of mortality among 84,835 US Army and Marine Corps Vietnam veterans were compared with that of 86,685 non-Vietnam veterans using standardized proportional mortality ratios. The veterans were a random sample of deceased Vietnam-era veterans identified in a Veterans Administration computerized benefit file. Military service information was obtained from military personnel records, and cause of death information from death certificates. Statistically significant excess deaths were observed among Army Vietnam veterans for motor vehicle accidents, non-motor vehicle accidents, and accidental poisonings. Similar findings have been reported in other studies of Vietnam veterans. Suicides were not elevated among Vietnam veterans. The Marine Corps Vietnam veterans appeared to have an increased mortality from lung cancer and non-Hodgkin's lymphoma. Although exposure to several environmental factors may be speculated, this study did not investigate possible etiologic factors for these elevated malignancies. here concern in the United States that postTserviceismortality among Vietnam are disproportionveterans is unusually high and certain causes of death ately elevated. Traumatic deaths such as motor vehicle accidents, suicides, and homicides are often cited as possible health outcomes associated with military service in Vietnam.1'6 Concern also persists that, as a result of exposure to Agent Orange and other chemicals in Vietnam, Vietnam veterans may be at increased risk for soft tissue sarcomas and other cancers.7"9 ApproxiFrom the Office of Environmental Epidemiology, Veterans Administration, Washington DC 20006-3868. Address correspondence to VA Office of Environmental Epidemiology (10B/AO8) Biddell Bldg, Rm 401, 1730 K St NW, Washington DC 80006-3868. 0096-1736/88/3006-4ia$oa.OO/0 Copyright © by American Occupational Medical Association 412 mately 2 million US military personnel served a oneyear tour in Vietnam during the Vietnam war. Findings of mortality studies of Vietnam veterans reported to date are not consistent with each other.1"6 Whether the variations among the studies are the result of the relatively small number of deaths analyzed, therefore reflecting lack of adequate statistical power, or whether they suggest an underlying difference in the mortality experience among the different Vietnam veteran study populations is not obvious. The number of deaths analyzed in these studies ranged from 246 to 923. In view of the public concern about the potential adverse health effects of military service in Vietnam and inconsistent findings in the scientific literature, a proportional mortality study of Vietnam veterans was undertaken. Approximately one third of all deaths which have occurred among the Vietnam veterans who served in tho US Army or Marine Corps was analyzed in the study. Materials and Methods Selection of Study Subjects Study subjects were restricted to ground troops, men who served in the US Army or Marine Corps at anytime from July 4, 1965 through March 1, 1973. Data published by the US Department of Defense indicate that over 80% of those who served in Vietnam were ground troops.10 Those having served in the Air Force, Navy, or Coast Guard were excluded because it is difficult to determine whether personnel who were considered to have served in the Vietnam theatre of operation were Mortality Study of Vietnam War Veterans/Breslin et al �ever actually "in country" Vietnam. Female veterans were also excluded from the study. It was determined that at least 50,000 eligible cases would be needed for the study in order to obtain adequate statistical power. The sample size of 50,000 deceased Vietnam era veterans would provide statistical power of over 90% for detecting a twofold increased relative risk of non-Hodgkin's lymphoma or lung cancer. The study would have excellent power to detect small increases in certain common causes of death. Potential study subjects who were reported to be deceased as of July 1,1982 were randomly selected from the Veterans Administration Beneficiary Identification and Record Locator Subsystem (BIBLS). The VA maintains the automated information retrieval system to identify and locate records of veterans who have received any of a wide variety of veterans' benefits including death benefits to their families. A study by the National Academy of Sciences indicates that the names of at least 94% of all deceased Vietnam-era veterans identified through independent means are in BIBLS.11 A subfile of 186,000 deceased Vietnam-era veterans who served in the Army or Marine Corps and whose service dates included the period 1964-1975 was assembled from BIBLS. If the service data (branch, service dates) were missing in BIBLS, veterans whose birth dates were between 1935 and 1957 (inclusive) were selected because of the high likelihood that they may have served during the Vietnam era. To achieve the desired sample size of approximately 50,000 eligible veterans, a random sample of 75,617 names was selected from the target population. Extra names were selected to allow for the exclusion of ineligible cases. The military personnel records for all 75,617 potential study subjects were requested from the National Personnel Becord Center in St. Louis, MO. Demographic data and information on military service such as branch of service, length of service, rank at discharge, and military occupational specialty were abstracted. In addition, for those who served in Southeast Asia, dates of service, principal duty, and unit addresses while in the theatre of combat were obtained. Of the 75,617 Vietnam-era veterans selected, 22,332 (29.5%) veterans were found to be ineligible upon reviewing their military personnel records. The ineligible cases included duplicate names; men who did not serve in the military from July 4, 1965 through March 1, 1973; men who served in the Navy, Coast Guard, or Air Force; men who were killed in action or were reported missing in action and subsequently declared dead; men who died in service before 1974; men who died of warrelated injuries; and all women. Eligibility for the study could not be determined for 1,032 veterans (1.4%) and they were excluded. The final sample consisted of 52,253 men who died between July 4, 1965 and July 1, 1982 and who served in the US Army or Marine Corps during the period July 4, 1965 through March 1, 1973. Death certificates were available from the VA files for about 70% of the sample; for the remaining 16,000 cases, the veteran's death certificate was requested from the state of his last known residence. The place of the veteran's death was identified by checking files of the VA, Social Security Administration, Internal Bevenue Service, and National Center for health Statistics National Death Index. Although death certificates were the preferred source of information, casualty reports issued by the Department of Defense were also used for active duty personnel or reservists who died outside the country and for whom no death certificate could be obtained. Most of these deaths were accidents and probably little additional information would have been obtained from the death certificate if it were available. Death certificates were the source of cause of death information for 96.9% of all cases. This was equally true for both those who served in Vietnam and those who did not. The underlying causes of death were coded by experienced nosologists at the National Center for Health Statistics using the International Classification of Diseases, 8th Bevision (ICDA8).ia The nosologists had no knowledge of the military service status of the veteran. Cause of death was ascertained for 51,421 veterans or 98.4% of the men determined to be eligible for the study. The cause of death for the remaining 1.6% was not obtained for one of the following reasons: the veteran died overseas and no certificate or cause of death information was available or the veteran's place of death had not been identified, and therefore the death certificate could not be located. Of the 51,421 men for whom military service data and cause of death information were available, 26,685 had not served in Southeast Asia; 24,235 had served in Vietnam. The remaining 501 were either known to have served elsewhere in Southeast Asia or their place of service in Southeast Asia was unknown. Analyses of mortality data were based on 24,235 Vietnam veterans and 26,685 non-Vietnam veterans. These procedures used to select the subjects are outlined in the Figure. Statistical Analyses The deaths observed among the Vietnam veterans were compared with expected numbers computed by applying the age- and race-specific proportions of deaths for each cause among the non-Vietnam veterans to the total number of deaths in the study group. Differences between observed and expected number of deaths for each cause were summarized in the form of the proportional mortality ratio (SPMB) which is the ratio of the number of deaths observed to that expected.13 The statistical significance of each ratio was tested by a xa with 1 df.14 The 95% confidence intervals for the SPMBs were also computed.10 Proportional mortality ratios standardized for age and race (SPMBs) were calculated separately for Army and Marine Corps Vietnam veterans for all major causes of death and for selected causes of death. The PMB analysis by branch of service was performed because these groups might have had different types of environmental exposure in Vietnam either by virtue of the Journal of Occupational Medicine/Volume 30 No. 5/May 1988 413 �Veterans Administration Beneficiary Identification and Record Locator Subsystem TABLE 1 Racial Characteristics of the 50,920 Deceased Vietnam-era Veterans by Branch and Vietnam Service Army Race Service in Vietnam Yes (N = 19,708), No (N = 22,904), Yes (N = 4,527), No (N = 3,781), 78.1 19.2 2.7 79.5 17.7 83.5 13.7 82.5 14.9 2.8 2.8 2.6 White Black Other Unknown Totals Random Sample (75,617) Marines Service in Vietnam Deceased Vietnam-era Veterans Army Marine Corps or Branch Unknown (186,000) * 100 * 100 100 100 •Less than 0.1%. Military Records Not Found (1,032) Qualified for Study (52,253) TABLE 2 Military Rank of the 50,920 Deceased Vietnam-era Veterans by Branch and Vietnam Service Army Cause of Death Unknown (832) Served in Thailand or Elsewhere in Southeast Asia (501), Excluded from Study Cause of Death Known (51,421) Served in Vietnam (24,235) Rank Served Places Other than Southeast Asia (26,685) Figure. Selection process of study subjects. location of their units or the types of duties they performed. Unlike the Army units, the Marine Corps units were primarily located within the I Corps area of South Vietnam. South Vietnam was divided into four tactical combat zones, I Corps being in the northernmost part of South Vietnam. Results The demographic characteristics of the sample are given in Tables 1 and 3. More than 50% of the veterans died at ages 25 through 34. Some died at ages less than 85 (5.5% of Vietnam and 11.7% of non-Vietnam veterans) and some died at ages older than 65 (0.74% of Vietnam and 8.6% of non-Vietnam veterans). There seemed to be no remarkable differences in the major cause of death categories between the men who served in Vietnam and their counterparts who did not serve in Vietnam with a few exceptions (Table 3). Deaths from external causes (ICDA codes E800-E989) were relatively more frequent among veterans who served in Vietnam than among those who did not. However, this excess is statistically significant only for Army veterans (PMB, 1.03; P< .01). More than half of all the deaths in the study population were due to accidents, accidental poisonings, or violence. Within this broad category, approximately 35% of the deaths were due to motor vehicle accidents 414 Enlisted Warrant officer Officer Unknown Totals Marines Service in Vietnam Service in Vietnam Yes No Yes No (N = 19,708), (N = 22,904), (N = 4,527), (N = 3,781), 92.8 93.4 93.6 95.8 2.1 5.1 1.0 5.5 0.9 5.5 0.3 3.9 * 100 * 100 * 100 100 •Less than 0.1%. (Table 4). Although the magnitude of the relative excess of motor vehicle accidents was about the same in both branches, only the SPMR for Army veterans was statistically significant (PMR, 1.05; P< .085). "Other transport accidents" were seen to be in excess primarily among Army personnel (PMR, 1.36; P < .01); 51% of these were aircraft accidents. Of the men who died in aircraft accidents, 88% had been helicopter pilots or crewmen and 84% of these had served in Vietnam. Many of these died while working as helicoptor pilots or crewmen in civilian life; others died in aircraft accidents while still in the military after the war. The category of "accidental poisonings" was elevated among both Army and Marine Corps veterans who served in Vietnam. In reviewing a sample of 100 of these deaths, it was found that 98% of these deaths were due to narcotic overdose, mostly heroin. Among enlisted Vietnam veterans, veterans with combat-related military occupations died from homicide significantly more frequently than veterans with non-combat-related military occupations: 9% excess for Army veterans (P < .05), 84% excess for Marine Corps veterans (P< .01). It also appeared that the excess deaths from motor vehicle accidents and accidental poisonings were greater during the first ten-year period of observations than the later years of observation for both Army and Marine Corps Vietnam veterans (Table 5). Deaths coded as suicide were relatively less frequent among those who served in Vietnam than among those who did not serve in Vietnam for both Army and Marine Corps veterans. Mortality Study of Vietnam War Veterans/Breslin et al �TABLE 3 Number of Deaths and Proportional Mortality Ratios (PMRs) Among Vietnam Veterans by Major Causes and Branch Army* Cause (ICDA No.) All other causes (210-228, 290315,740-759,780-796) Infective and parasitic diseases (000-136) Malignancies (140-209, 230-239) Endocrine, nutritional, and metabolic (240-279) Blood and blood-forming organs (280-289) Nervous systems and sense organs (320-389) Circulatory diseases (390-458) Respiratory diseases (460-519) Digestive diseases (520-577) Genitourinary diseases (580-629) Skin and subcutaneous tissues (680-709) Musculoskeletal and connective tissues (71 0-738) Accidents, poisonings, and violence (E800-989) Observed Marinest 95% Confidence Interval PMR Observed PMR 95% Confidence Interval 150 0.94 0.63-1.01 19 1.02 0.89-1.17 0.97 0.85 0.93-1.02 0.67-1.08 521 22 1.20 0.66 1 .0-1 .45 0.22-2.01 32 0.68 0.45-1 .03 8 3.22 0.51-20.5 167 0.95 0.77-1.18 27 0.86 0.17-4.44 0.98 0.93 0.99 0.77t 0.76 0.95-1 .01 0.69-1.25 0.94-1 .04 0.60-0.99 0.05-11.19 647 62 169 13 2 0.98 0.95 0.87 0.67 0.50 0.86-1.12 0.75-1 .21 0.70-1.08 0.33-1.35 0.06-4.18 7 0.87 0.22-3.41 2,880 1.00 709 0.91 127 0.80 2,452 135 3,578 406 1,001 80 8 29 10,984 1.55 0.8-3.0 1.03§ 1.02-1.04 * Army: deaths observed = 19,708. Expected numbers are based on 22,904 deaths in Army non-Vietnam veteran comparison group, t Marines: deaths observed = 4,527. Expected numbers are based on 3,781 deaths in Marine non-Vietnam veteran comparison group. :(:P<.05forx 2 with1cff. § P < .01 for x2 with 1 df. TABLE 4 Number of Deaths from Accidents, Accidental Poisonings, and Violence and Proportional Mortality Ratios (PMRs) Among Vietnam Veterans, by Branch Army* Cause (ICDA No.) Motor vehicle accidents (E810E827) Other transportation accidents (E800-E807, E830-E845) Accidental poisonings (E850E877) All other accidents/injury (E880E949, E970-E989) Suicide (E950-E959) Homicide (E960-E969) Marinest Observed PMR 95% Confidence Interval Observed PMR 95% Confidence Interval 3,884 1.05$ 1.01-1.09 1,011 1.07 0.97-1.18 493 1.36§ 1.19-1.56 117 0.75 0.56-1.01 461 1.15$ 1.02-1.30 120 1.10 0.93-1.30 2,323 1.05 0.99-1.11 593 1.01 0.98-1 .04 2,003 0.93§ 1.01 0.88-0.98 0.73-1.40 542 497 0.93 0.98 0.86-1.01 0.89-1 .08 1,816 * Army: deaths observed = 19,708. Expected numbers are based on 22,904 deaths in Army non-Vietnam veteran comparison group, t Marines: deaths observed = 4,527. Expected numbers are based on 3,781 deaths in Marine non-Vietnam veteran comparison group, t P<.025 for x 2 with 1 df. §P<.01for x 2 with1df. When all malignancies were grouped together, Vietnam veterans did not exhibit an excess of cancer when compared to their counterparts who did not serve in Vietnam. Differences between the services, however, were seen for specific cancer sites among those who served in Vietnam relative to men who did not (Table 6). The most interesting differences were the statistically significant elevation for lung cancer (PMR, 1.58; P< .025) and non-Hodgkin's lymphoma (PMR, S.10; P < .025) seen in the Marines who served in Vietnam relative to Marines who served elsewhere. The risk for soft tissue sarcoma was not elevated among Vietnam veterans as a whole or in any subgroup of these veterans. Discussion For most major causes, the distribution of deaths for veterans who served in Vietnam is not markedly different from those who did not serve in Vietnam except for selected malignancies and "accidents, accidental poisonings, and violence." Four states have conducted mortality studies of Vietnam era veterans: Wisconsin,1 West Virginia,8 New York,8 and Massachusetts.4 The Centers for Disease Control also reported the postservice mortality of US Army Vietnam veterans." The results of these studies were similar to what was seen here. They Journal of Occupational Medicine/Volume 30 No. 5/May 1988 415 �TABLE 5 Deaths from Selected Causes Among Enlisted Vietnam Veterans Who Died Between 1965 and 1982, and Who Had Only One Tour of Duty* Army Marines 1965-1975 1976-1982 1965-1975 Cause (ICDA No.) Observed Observed 717 34 1.11f 0.65 2,053 201 1.06 1.10 167 11 77 342 212 232 89 Motor vehicle accidents Other transport accidents Accidental poisonings All other accidents/injury Suicide Homicide Cancers (140-209, 230-239) PMR PMR 1.14 1.00 1.00 0.92 0.74t 255 1,186 1,089 1,008 655 1.09 1.05 0.94 1.03 0.87f 25 82 57 68 25 1976-1982 PMR Observed PMR 1.18 1.04 611 48 1.01 0.82 2.26f 1.01 0.65 1.27 0.61 1 Observed 71 364 368 314 176 1.04 0.93 1.05 0.91 1.31 * PMR, proportional mortality ratio of observed to expected numbers of deaths. Expected number was generated based on deaths from nonVietnam veterans with similar characteristics. tP<.05forx 2 with1df. TABLE 6 Number of Deaths from Malignancies Among Vietnam Veterans by Branch of Service Army* Cause (ICDA No.) Observed All other causes (000-136, 210E989) All malignancies Buccal (140-1 49) Esophagus (150) Stomach (151) Intestines and other gastrointestinal (152-1 54, 158, 159) Liver, bile ducts (155-1 56) Pancreas (157) Upper respiratory (160-161) Lung (162) Bone (170) Soft tissue (171) Melanoma of the skin (1 72) Prostate (185) Testis(186) Bladder (188) Kidney (189) Brain (191) Other nervous system (192) Thyroid and endocrine (193-194) Non-Hodgkin's lymphoma (200, 202) Hodgkin's disease (201) Multiple myeloma (203) Leukemia (204-207) Other cancers (163, 173-4, 187, PMR 17,256 0.97 0.92 1.24 1.12 0.96 34 82 29 632 27 30 145 30 90 9 55 116 43 15 108 1.04 0.87 1.14 1.03 0.82 0.99 1.02 0.92 1.12 0.56 0.87 0.97 0.558 0.59 0.81 92 18 202 281 1.16 0.77 0.88 1.03 95% Confidence Interval 1.00 2,452 71 46 88 209 Marinesf Observed PMR 95% Confidence Interval 4,006 0.98 0.93-1 .01 0.47-1.82 0.78-1.98 0.85-1.47 0.70-1.32 521 13 5 17 33 1.20 1.95 0.39 0.82 1.26 1.0-1.45 0.54-7.04 0.11-1.41 0.41-1.64 0.71-2.24 0.77-1.41 0.64-1.18 0.63-2.07 0.39-1.71 0.80-1.23 0.91-1.14 0.55-1.23 0.84-1.5 0.27-1.18 0.50-1.52 0.29-3.20 0.38-0.79 0.30-1.17 0.63-1.04 6 18 1 130 11 8 36 5 26 4 13 25 11 4 35 1.21 1.63 0.18 1.58* 1.38 0.71 0.94 1.29 1.29 2.41 0.89 1.07 0.93 0.57 2.10* 0.52-2.83 0.46-5.75 0.03-1.32 1 .09-2.29 0.09-21 .48 0.38-1.32 0.59-1.50 0.16-10.3 0.47-3.57 0.09-66.35 0.54-1 .46 0.16-7.14 0.49-1 .78 0.10-3.37 1.17-3.79 0.73-1.85 0.23-2.53 0.73-1 .06 0.93-1.14 22 2 42 54 1.33 0.45 1.14 1.07 0.67-2.63 0.01-17.13 0.18-7.14 0.60-1.91 190, 195-9, 208-9, 230-9) * Army: deaths observed = 19,708. Expected numbers based on 22,904 deaths in Army non-Vietnam veteran comparison group, t Marines: deaths observed = 4,527. Expected numbers based on 3,781 deaths in Marine non-Vietnam veteran comparison group, t P<. 025 for x2 with 1 off. §P<.01 for x 2 with 1 df. also reported that Vietnam veterans were more likely to die from "accidents, accidental poisonings, and violence" or from a few selected malignancies than their counterparts who did not go to Vietnam. However, within these two broad categories, the findings reported by these studies were not consistent with each other or with what was found here. It should be noted that the data in the state reports were not strictly comparable to the data given in this report. They differed from this 416 study in that numbers of deaths studied were much smaller—less than 1,000-and they included Vietnamera veterans from all branches of service. They also differed somewhat in the analytical methods and used different types of veteran comparison populations. Furthermore, military personnel records of the state study subjects were not reviewed to verify Vietnam service status. The New York,3 Massachusetts,4 and CDC8 studies Mortality Study of Vietnam War Veterans/Breslin et al �reported nonstatistically significant elevations of risk for suicide when veterans with service in Vietnam were compared with other Vietnam-era veterans. In this study no excess of suicides was seen among the veterans who served in Vietnam when compared with other Vietnam-era veterans. The ratio of observed to expected deaths among the Army and Marine Corps veterans who served in Vietnam relative to their counterparts who did not serve in Vietnam was less than one; for veterans who had served in the Army, the deficit was statistically significant at P< .01. Simon16 reported an association between attempted suicide and combat experience in World War II veterans. In our study there was no data element that indicated whether a man had been in combat. As a surrogate measure, enlisted men whose military occupational specialties would be likely to involve combat, ie, rifleman, artilleryman etc, were compared to the other enlisted men who had served in Vietnam. Among the enlisted men who served in Vietnam, those with "combat-related" occupational specialties in both branches of service had relatively fewer suicides than "non-combat" Vietnam veterans in the same branch of service (Army: N = 590, PMR = 0.96; Marines: N = 228, PMR = 0.83). For the Marines, the deficit was statistically significant atP<.05. It is known that suicides are under-reported on death certificates but there was no reason to believe that they were more underreported among veterans who served in Vietnam than among those who did not. Nor was there any reason to believe that suicide was more apt to be underreported among those likely to have been in combat in Vietnam than among other veterans. Several researchers suggested that 1.6% to 5% of motor vehicle accidents may be suicides.17'18 In our study, motor vehicle accidents were relatively less frequent among Vietnam veterans with combat-related occupational specialties than among those with noncombat occupations (Army: N = 1,205, PMR = 0.99; Marines: N = 466, PMR = 0.94). Even if one assumes that some of the motor vehicle accidents are "hidden" suicides, it is unlikely that these could account for the overall deficit in suicides among Vietnam veterans since the possible "hidden" suicides among motor vehicle accidents are reportedly relatively small. The apparent excess of drug-related deaths attributable to heroin use among Vietnam veterans is of concern. This observation was consistent with that of Rohrbaugh et al,19 who found that, whereas Vietnam veterans were no more likely than other Vietnam-era veterans to use drugs in general, they were more likely to use opiates than other illicit drugs. The CDC also reported that accidental drug poisonings were substantially elevated among Vietnam veterans. One of the major concerns of Vietnam veterans has been the possibility of developing cancer as a result of exposure to Agent Orange, a mixture of two phenoxy herbicides. Some studies have shown an association between soft tissue sarcomas and exposure to phenoxy herbicides.7'20'21 Although data from Wisconsin,1 West Virginia,2 and Massachusetts4 indicated that veterans who served in Vietnam may have an increased risk of soft tissue sarcoma, no excess of soft tissue sarcomas was seen among the Vietnam veterans in our study. No association between soft tissue sarcoma and military service in Vietnam was found by Greenwald et alsa in a case control study of 881 men with soft tissue sarcoma from New York State, nor was any association found between military service in Vietnam and 234 cases of soft tissue sarcoma occurring among Vietnam-era veterans admitted to Veterans Administration hospitals,23 or 817 soft tissue sarcoma cases referred to the Armed Forces Institute of Pathology.84 The veterans who served in the Marine Corps in Vietnam were seen to have a statistically significant (P < .085) excess of non-Hodgkin's lymphoma when compared with Marines who did not serve in Vietnam. West Virginia veterans with service in Vietnam had a statistically significant excess of Hodgkin's disease when compared with other Vietnam-era veterans.2 None of the other state studies indicated any excess of lymphomas among veterans with service in Vietnam.1'3'* Non-Hodgkin's lymphoma has been associated with exposure to phenoxy herbicides,8'9 arsenicals,86 dapsone,26 and certain viruses.27 The men who served in Vietnam had the potential for exposure to all of these agents. Agent Blue, a herbicide used in Vietnam, was an organic arsenical compound and dapsone, a sulfone, was used as an antimalarial drug by some of the troops in Vietnam. Dapsone26 has been shown to cause lymphomas in laboratory animals. Dapsone was given mainly to troops stationed in I Corps and the central highland areas of Vietnam where falciparum malaria was prevalent. Most of the Marines in Vietnam served in I Corps. It will be interesting to see whether the Army troops who were stationed in I Corps also exhibit an excess of lymphomas. The data necessary for this analysis are now being collected. Lung cancer was significantly elevated (PMR, 1.58; P < .085) among Marines who served in Vietnam relative to Marines who did not serve in Vietnam. The veterans from New York3 with service in Vietnam also had relatively more lung cancer than other Vietnamera veterans but the excess was not statistically significant. Although tobacco is the etiologic agent most commonly associated with lung cancer, this disease has also been associated with exposure to other substances such as arsenic28 and phenoxy herbicides.80'29 A survey of more than 89,000 Vietnam-era veterans in Wisconsin indicated that they were nearly twice as likely to be cigarette smokers as were men in the general population.1 There are no smoking histories available for the Marines in this study. If the lung cancer deaths in this study are associated with an increased use of tobacco by the men who served in Vietnam, lung cancer deaths should also be increased among Army troops in Vietnam. They were not. The present study has certain inherent limitations that make it difficult to draw firm conclusions. First, Journal of Occupational Medicine/Volume 30 No. 5/May 1988 417 �risk estimates obtained from PMR analyses can approximate the results from studies of cause-specific mortality rates or the standardized mortality ratio (SMR).30 However, PMRs may be inflated for certain causes when the overall mortality rate of the study group is lower than that of the comparison population. This would have been the case if the US general population had been chosen for the comparison population in this study. It was shown that the selection process for military service exerted a profound effect on the mortality of veterans after separation from service. The number of deaths among the World War II male Army veterans was only 83.5% of the expected number at concurrent death rates for US white men.31 A recent study published by the CDC showed that the mortality among the Vietnam veteran study population was 17% higher than the rate among the non-Vietnam veteran comparison populations.5 These suggested that SPMRs for lung cancer and nonHodgkin's lymphoma reported in this study could have been biased toward underestimating the risks. Second, it is possible that, with so many comparisons being made, the few significant elevations observed could be interpreted as chance findings. Findings from this study need to be replicated by other Vietnam veteran studies. Third, no exposure data on individual veterans were available so as to evaluate the possible etiologic factors of the malignancies which appeared to be elevated among Marine Vietnam veterans. Additional work needs to be done to find characteristics that may point to possible etiologic factors. Fourth, the observation period in this study, a maximum of 17 years, may have been still insufficient to observe the risk of dying from diseases with a long latency period. A periodic monitoring of Vietnam veteran mortality patterns is warranted. Despite the limitations described above, the present study is the largest mortality study of Vietnam veterans reported to date encompassing approximately one third of all deaths which have occurred among the US Army and Marine veterans who served in Vietnam. Having an equally large number of non-Vietnam veterans whose characteristics are well-defined and are similar to the study population except for service in Vietnam should be considered a major strength. Furthermore, unlike other PMR studies of Vietnam veterans, in this study military personnel records for almost all (98.6%) potential study subjects were retrieved and reviewed to determine eligibility of the veteran. Therefore, the chance of misclassification of the most important study variable, namely service in Vietnam, is minimal. In summary, the study shows no significant differences in the major cause of death between Vietnam veterans and non-Vietnam veterans with a few exceptions. Accidental and drug-related deaths were relatively more frequent among Army Vietnam veterans. Suicides were less frequent among Vietnam veterans. Vietnam veterans who served in the Marine Corps were seen to have statistically significant excess of lung cancer and non-Hodgkin's lymphoma. 418 Acknowledgments The Veterans Administration wishes to acknowledge the assistance and support received from many individuals and agencies without which the VA mortality study could not have been successfully completed. We are grateful to Gilbert Beebe, PhD (NIH), Chin Long Chiang, PhD (UC Berkeley), Joseph Fleiss, PhD (Columbia University), the late Bernard Greenberg, PhD (University of North Carolina), the late Abraham Lillenfeld, MD (Johns Hopkins University), and Richard Monson, MD (Harvard University) for their reviews of the study protocol and the many suggestions and recommendations made on the conduct of the study. We also would like to acknowledge the contributions of David Peterson and Carolyn Brooks of the National Archives Records Administration; Paul Gray, National Personnel Records Center; Richard Christian, US Army and Joint Services Environmental Support Group; Robert Bilgrad, National Center for Health Statistics, National Death Index; and the Social Security Administration; the National Institute for Occupational Safety and Health; and the Internal Revenue Service. John Ward of Westat and Elaine Kokiko of Moshman Associates assisted us in the collection of military service data and death certificates, respectively. The guidance provided by William Page, PhD (National Academy of Sciences) and Alvin Young, PhD (White House Office of Science and Technology Policy) in planning for the study is greatly appreciated. References 1. Anderson HA, Hanrahan LP, Jensen M, et al: Wisconsin Vietnam Veteran Mortality Study, Final Report. State of Wisconsin Dept of Health and Social Services, Division of Health, 1986. 8. Bailey C, Baron BC, Basanao E, et al: West Virginia Vietnam Era Veterans Mortality Study. West Virginia Residents 1968-1988, Preliminary Report. West Virginia Health Dept, 1986. 3. Lawrence CE, Reilly AA, Quickenton P, et al: Mortality patterns of New York State Vietnam veterans. Am J Public Health 1985:75:377-879. 4. Kogan MD, Clapp RW: Mortality Among Vietnam Veterans in Massachusetts, 1978-83. Massachusetts Office of Commissioner of Veterans Services, Agent Orange Program, Massachusetts Dept of Public Health, Division of Health Statistics, 1985. 5. The Centers for Disease Control: Postservice mortality among Vietnam veterans. JAMA 1987;857:790-795. 6. Hearst N, Newman TB, Hully SB: Delayed effects of the military draft on mortality: A randomized natural experiment. N Eng-1 J Mod 1986;314:680-684. 7. Hardell L, Sandstrom A: Case control study: Soft tissue sarcoma and exposure to phenoxyacetic acids of chlorophenols. Br J Cancer 1979;39:711-717. 8. Hardell L, Ericksson M, Lenner P, et al: Malignant lymphoma and exposure to chemicals especially organic solvents, chlorophenols and phenoxy acids: A case-control study. Br J Cancer 1981;43:169176. 9. Hoar SK, Blair A, Holmes FF, et al: Agricultural herbicide use and risk of lymphoma and soft-tissue sarcoma. JAMA 1986;856:11411146. 10. Dept of Defense, Directorate for Information, Operations, and Reports: Selected Manpower Statistics, Fiscal Year 1981. 11. National Academy of Sciences Commission on the Life Sciences Medical Follow-up Agency: Ascertainment of Mortality in the U.S. Vietnam Veteran Population. Report of Contract V101 (93) P-937, Washington, DC, 1985. 18. Eighth Revision International Classification of Disease, Adapted for Use in the United States. US Public Health Service, Washington, DC. 13. Monson RR: Analysis of relative survival and proportional mortality. Comp Blamed Res 1974;7:385-338. 14. Mantel N, Haenszel W: Statistical aspects of the analysis of data from retrospective studies of disease. JNCI 1959;88:719-748. 15. Spiegelman D, Wang JD, Wegman D: Epidemiologlc programs for computers and calculators. Am J Epidemiol 1983;118:599-607. 16. Simon W: Attempted suicide among veterans. J Nerv Ment Dls 1950;lll:451-468. Mortality Study of Vietnam War Veterans/Breslin et al �17. Huffine GL: Equivocal single-auto traffic fatalities. Life Threatening Behav 1971;l:83-95. 18. Schmidt QW, Shaffer JW, Zlotowitz HI, et al: Suicide by vehicular crash. Am J Psychiatry 1977;184:176-177. 19. Rohrbaugh M, Bads O, Press S, et al: Effects of Vietnam Experience on Subsequent drug use among servicemen. Int J Addict 1974;9:S6-40. SO. Lynge E: A followup study of cancer incidence among workers in manufacture of phenoxy herbicides in Denmark. Br J Cancer 1985:58:859-870. 21. Erickson M, Hardell L, Berg NO, et al: Soft tissue sarcomas and exposure to chemical substances: A case referrant study. Br JInd Med 1981:38:87-33. 88. Greenwald P, Kovasznay B, Collins DN, et al: Sarcoma of soft tissues after Vietnam service. JNCI 1984;73:1107-1109. 83. Rang H, Weatherbee L, Breslin P, et al: Soft tissue sarcoma and military service in Vietnam: A case comparison group analysis of hospital patients. J Occup Med 1986:88:1815-1818. 34. Rang HK, Enzinger FW, Breslin P, et al: Soft tissue sarcoma and military service in Vietnam: A case control study. JNCI 1987;79:693-«99. 85. Axelson O, Dahlgren E, Jansson CD, et al: Arsenic exposure and mortality: A case referrent study for a Swedish copper smelter. BrJInd Med 1978;35:8-15. 86. National Cancer Institute Carcinogenesis Technical Report; Series No. 80: Bioassay of dapsone for possible carcinogenicity. DHEW publication No. NIH 77-880. Washington, DC 1977. 87. Schottenfeld D, Fraumeni <TF: Cancer Epidemiology and Prevention. Philadelphia, W. B. Saunders Co, 1988, pp 770-771. 88. Ott MG, Holder BB, Goldon HL, et al: Respiratory cancer and occupational exposure to arsenicals. Arch Environ Health 1974:89:850-855. 89. Zack JA, Oaffey WR: A mortality study of workers employed at the Monsanto Company plant in Nitro, West Virginia. Environ Sci Res 1983:86:575-591. 30. Decoufle P, Thomas TL, Pickle LW: Comparison of the proportionate mortality ratio and standardized mortality ratio risks measures. Am JEpidemiol 1980;lll:863-869. 31. Seltzer CC, Jablon S: Effects of selection on mortality. Am J Epidemlol 1974:100:367-378. Journal of Occupational Medicine/Volume 30 No. 5/May 1988 419 � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Alvin L. Young Collection on Agent Orange Description An account of the resource <p style="margin-top: -1em; line-height: 1.2em;">The Alvin L. Young Collection on Agent Orange comprises 120 linear feet and spans the late 1800s to 2005; however, the bulk of the coverage is from the 1960s to the 1980s and there are many undated items. The collection was donated to Special Collections of the National Agricultural Library in 1985 by Dr. Alvin L. Young (1942- ). Dr. Young developed the collection as he conducted extensive research on the military defoliant Agent Orange. The collection is in good condition and includes letters, memoranda, books, reports, press releases, journal and newspaper clippings, field logs and notebooks, newsletters, maps, booklets and pamphlets, photographs, memorabilia, and audiotapes of an interview with Dr. Young.</p> <p>For more about this collection, <a href="/exhibits/speccoll/exhibits/show/alvin-l--young-collection-on-a">view the Agent Orange Exhibit.</a></p> Text A resource consisting primarily of words for reading. Examples include books, letters, dissertations, poems, newspapers, articles, archives of mailing lists. Note that facsimiles or images of texts are still of the genre Text. Box The box containing the original item. 066 Folder The folder containing the original item. 1786 Series The series number of the original item. Series III Subseries III Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Creator An entity primarily responsible for making the resource Breslin, Patricia Han K. Kang Yvonne Lee Vicki Burt Barclay M. Shepard Source A related resource from which the described resource is derived Journal of Occupational Medicine Date A point or period of time associated with an event in the lifecycle of the resource 1988-05-01 Title A name given to the resource Proportionate Mortality Study of US Army and US Marine Corps Veterans of the Vietnam War Subject The topic of the resource VA Mortality Study mortality trends cancer risk assessment ao_seriesIII Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Alvin L. Young Collection on Agent Orange Description An account of the resource <p style="margin-top: -1em; line-height: 1.2em;">The Alvin L. Young Collection on Agent Orange comprises 120 linear feet and spans the late 1800s to 2005; however, the bulk of the coverage is from the 1960s to the 1980s and there are many undated items. The collection was donated to Special Collections of the National Agricultural Library in 1985 by Dr. Alvin L. Young (1942- ). Dr. Young developed the collection as he conducted extensive research on the military defoliant Agent Orange. The collection is in good condition and includes letters, memoranda, books, reports, press releases, journal and newspaper clippings, field logs and notebooks, newsletters, maps, booklets and pamphlets, photographs, memorabilia, and audiotapes of an interview with Dr. Young.</p> <p>For more about this collection, <a href="/exhibits/speccoll/exhibits/show/alvin-l--young-collection-on-a">view the Agent Orange Exhibit.</a></p> Text A resource consisting primarily of words for reading. Examples include books, letters, dissertations, poems, newspapers, articles, archives of mailing lists. Note that facsimiles or images of texts are still of the genre Text. Box The box containing the original item. 037 Folder The folder containing the original item. 0799 Series The series number of the original item. Series III Subseries I Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Creator An entity primarily responsible for making the resource Kang, Han K. Barclay M. Shepard Alvin L. Young Source A related resource from which the described resource is derived Occupational Medicine - Current Concepts Date A point or period of time associated with an event in the lifecycle of the resource 1984 Title A name given to the resource Health Effects of Agent Orange Exposure Subject The topic of the resource dioxin herbicide toxicology veteran health and hygiene ao_seriesIII https://www.nal.usda.gov/exhibits/speccoll/files/original/f9a43b5489fcaeec623ab493d8807bbd.pdf 41584cca05f4093c521e8da0cedfb4ff PDF Text Text Item ID Number 0136? Author Wilkins, John R. Corporate Author RepOrt/ArtlClO Title Epidemiologic Approaches to Chemical Hazard Assessment JOUmal/BOOk Title Hazard Assessment of Chemicals: Current Development Year 1983 Month/Day Color a Number of Images 29 Descrlpton Notes Thursday, May 03, 2001 Page 1367 of 1403 �Epidemiologic Approaches to Chemical Hazard Assessment John R. Wilkins III Department of Preventive Medicine The Ohio State University Columbus, Ohio Nancy A. Reiches Comprehensive Cancer Center The Ohio State University Columbus, Ohio I. Introduction II. Developing Clues to Chemically Related Disease: Descriptive Approaches A. Fundamental Epidemiologic Tools B. Variations of Disease Occurrence in Time C. Variations in Disease Occurrence from Place to Place . . III. Testing Etiologic Hypotheses: An Overview of Analytic Approaches A. Case-Control Studies B. Cohort Studies IV. Crucial Aspects of Environmental Study Design A. Measurement of Dose B. Measurement of Response V. Relating Measures of Dose to Measures of Response VI. Conclusion References I. 133 135 135 141 145 153 154 158 161 161 173 178 180 182 INTRODUCTION During this century, populations of industrialized nations have experienced dramatic changes in the pattern and relative importance of various life-threatening illnesses. At one time, diseases such as tuberculosis and 133 HAZARD ASSESSMENT OF CHEMICALS: Current Developments, Vol. 2 6dt- Copyright © 1983 by Academic Press, Inc. All rights of reproduction in any form reserved. ISBN 0-12-312402-6 �x-r-r junii iv. yniKins iu ana Nancy A. Ketches smallpox claimed large numbers of lives, especially in younger age groups. Now, these and many other conditions of infectious origin are rarely encountered. The resulting increase in average life expectancy, however, has resulted in a new set of public health problems. One of these problems, chemical contamination of the human environment and the assessment of health risks that may result, is the focus of this article. Perhaps one of the most difficult questions in contemporary epidemiology and public health stems from evidence suggesting that most malignant disease is of environmental, perhaps chemical, origin. This supposition is particularly significant in the context of public health since cancer is the second leading cause of death, accounting for approximately one in every five deaths. Furthermore, mortality from cancer has increased in recent years, leading to the hypothesis that the spread and diversity of chemical contamination of the environment may largely account for this trend. To the extent that carcinogenic agents are exogenous in nature, there is potential for intervening in the environmental network and thus preventing the occurrence of disease. It is this concept that forms the cornerstone of epidemiologic research. The primary purpose of epidemiologic inquiry is to estimate quantitatively the effects of those factors that determine whether disease does or does not occur in human populations. Classic epidemiologic models, in combination with evidence from laboratory research, are powerful tools for both generating and testing hypotheses about the etiologic significance of environmental contaminants. Epidemiologic investigations often result in the institution of preventive or control strategies—i.e., interventions in the process of disease causation— even in the absence of knowledge regarding the underlying biologic mechanisms. Consequently, epidemiologic research assumes a central role in the protection of the public health. In this article, the major features of the epidemiologic approach to chemical hazard assessment are discussed. At the outset the fundamental sources of epidemiologic data and the process of generating testable hypotheses are described. Next, the methods by which such hypotheses are tested are considered. In particular, the focus is on the ways in which measures of dose and measures of response are derived and evaluated in epidemiologic research. Finally, questions regarding the interpretation of dose and response measures are addressed. The epidemiologic approach to disease may be described as proceeding in two major phases. The first phase, discussed in the following section, involves the conduct of "descriptive" epidemiologic studies. It is important to emphasize that the primary purpose of these studies is to generate hypotheses of cause and effect, a goal achieved primarily by examining Epidemiologic Approaches to Chemical Hazard Assessment 135 patterns of disease occurrence with respect to time and place. The second phase, discussed in Section III, involves conducting more rigorous "analytic" studies, studies whose purpose is to test hypotheses previously set forth. II. A. DEVELOPING CLUES TO CHEMICALLY RELATED DISEASE: DESCRIPTIVE APPROACHES Fundamental Epidemiologic Tools 1. Measures of Disease Frequency In this subsection, measurements of disease occurrence that are fundamental to any epidemiologic investigation are discussed. Subsequently, the role of these measurements in typical epidemiologic models is considered. There are several measurements that reflect various aspects of the frequency of disease in a population. In general, these can be divided into two major categories. The first relates to measures of morbidity, or illness; the second concerns mortality. One of the most basic concepts with respect to the measurement of disease occurrence is that of a rate. Most simply, a rate may be defined as the frequency of a condition in a defined population over a specific period of time. Clearly, an absolute count of cases, without reference to a population of known size, precludes direct comparisons between groups. For example, if it is known only that /i, cases of Disease X occurred in Community A, and n2 cases in Community B, there is insufficient information to determine in which community the occurrence of the disease is greater. If, however, the size of the population in which the cases were detected is also known, the rate at which disease occurs can be computed, thereby yielding figures that are comparable. With respect to morbidity, the rate that is usually of most interest is the incidence rate. This is a direct measure of the probability or risk of illness. Ideally, incidence rates would be based on prospective surveillance of a well-defined population in which only those individuals who are at risk of developing the disease under consideration are included in the denominator. Although denominators are rarely this precise, they should not include persons who already have the disease or are not susceptible. Mortality rates, on the other hand, represent the probability or risk of death. The number of deaths in a defined group constitutes the numerator of a mortality rate, while the denominator represents the total number �136 Epidemiologic Approaches to Chemical Hazard Assessment John R. Wilkins III and Nancy A. Reiches of persons in the defined group, i.e., the number of persons at risk of dying. Morbidity and mortality rates may be expressed as either crude, specific, or adjusted rates. Crude rates can be constructed from a minimal amount of information; they require knowledge only of the total number of events (e.g., births, deaths) and the size of the population in which the events occurred. However, to the extent that the risk of illness or death is not uniform (equally probable) across all members of a population, specific rates can provide more useful information. Typically, specific rates are defined with respect to age, race, sex, or some combination of these demographic characteristics. The basic formula for a specific rate includes in the numerator the number of events of interest in a homogeneous population subgroup, and in the denominator, the number of persons at risk in that same subgroup. Therefore, an age-specific death rate could be computed for persons 50-54 years of age; the rate could further be confined to white male members of the population in this age group. The general advantage of specific rates, whether for morbidity or mortality, is that they do not obscure potentially significant differences among population subgroups. Since these rates provide a high degree of detail, they are very useful for both analytic and health planning activities. This is particularly true if the condition under consideration exhibits great variation in occurrence among different age groups. Most chronic diseases, such as cancer and cardiovascular disease, manifest such a pattern. When morbidity and mortality rates are, however, used to compare disease experience in two or more populations, specific rates can present problems. Age-specific rates are often computed for either 5- or 10-year intervals, thus yielding as many as 18 rates per population. For many types of comparisons the task becomes cumbersome and the results difficult to interpret. Therefore, it is often preferable to compute some type of summary figure, which usually implies an adjusted rate. Adjustment is a procedure by which differences in the composition of groups are removed, so that the difference does not bias the comparisons of interest. The need for adjustment is illustrated by the difficulties that can be introduced when only crude rates are used. For example, if interest centers on a disease such as cancer, an illness occurring more frequently in older persons, the rationale for adjusting for age is clear. If Population A has a higher proportion of older individuals than does Population B, but the risk of dying for persons in any specific age group is the same, the crude rate will be greater for Population A. This would lead to a misinterpretation of the risk of dying in each population. What Hs required is a method of comparing the two groups as if the age distributions were identical. This can be accomplished by the "direct method" 137 of adjustment. By this method, age-specific rates are weighted not by the proportion of the population under study in the given age group, but rather by the proportion of an external standard population in the same age group. This procedure answers the question: "What would the rate in the study population be if its age distribution were equivalent to that of the standard population?" By .adjusting several population rates to the same standard, direct, unbiased comparisons between groups can be made. Although adjustment for age is probably the most common form of adjustment (or standardization), the same technique can be applied to remove differences with respect to other characteristics, such as sex or race. While the procedure is both well accepted and useful in a variety of analytic settings, it is not without disadvantages. First, it must be remembered that an adjusted rate is, in one sense, "fictional." That is, its magnitude depends not only on the "real" death rate in the study population, but also on the choice of the standard population. Second, because the adjusted rate is a summary figure,, it may obscure different trends among subgroups. For example, if only temporal changes in ageadjusted rates are examined, differences across selected age groups with respect to the rate of change, or even the direction of change, may go unnoticed. 2. Sources of Morbidity and Mortality Data This subsection considers some of the major sources of morbidity and mortality data commonly employed in epidemiologic studies of environmental phenomena. Some of the advantages and disadvantages of these sources are also discussed. First, we consider mortality data, a source of information that appears to be quite straightforward, but in truth is rather complex. The fundamental resource for mortality data is the death certificate. A standard form for death certification has been developed by the National Center for Health Statistics, the agency responsible for the detailed tabulation of all vital records. The death certificate contains demographic information, such as the decedent's age, race, sex, place of birth, place of death, place of usual residence, marital status, and occupation. The medical portion of the certificate includes data on the immediate and underlying causes of death and on other significant conditions that may have contributed to the death. There is also an item indicating whether or not an autopsy was performed. There has been a great effort to ensure uniformity in the death certification process and in the subsequent reporting of death records. In this regard, the Physician's Handbook on Medical Certification specifies rules for �138 John R. Wilkins III and Nancy A. Reiches recording the cause of death, rules that distinguish between the immediate cause and the underlying cause (760). The immediate cause of death is the disease or injury that directly preceded death; the underlying cause, which is the most important item for epidemiologic studies, is "the condition that started the sequence of events between normal health and the immediate cause of death." Additionally, there are internationally accepted rules for coding the medical information contained in the death certificate. Numeric codes, as specified in the International Classification of Diseases (ICD) manuals, are assigned to each cause on the certificate. The most crucial part of the coding process is the choice of underlying cause of death, since this item becomes the officially reported cause. Rules for this section are outlined in a National Center for Health Statistics publication entitled "Instructions for Classifying the Underlying Cause of Death, 1979" (767). However, owing to differences in the way in which physicians complete the death certificate form, the process of choosing the underlying cause does involve judgment and is therefore subject to both random and systematic error. In addition to outright errors in death certification and coding, there are some questions inherent in the way mortality data are reported that bear on their epidemiologic usefulness. One central question involves the forced choice of a single, underlying cause of death. Published statistics might lead one to believe that each death is the consequence of just one disease. However, the majority of death certificates, particularly for older persons, contain two or more diagnoses. This raises the medical question of how exact cause of death can be determined in an individual with multiple, potentially life-threatening infirmities (93). Furthermore, there is the data-management and statistical problem of tabulating multiple causes of death, which in practice is seldom done. Although computerization of death records is relatively recent, the coding of cause-ofdeath data into machine-readable form opens up the possibility of routine tabulations of multiple-cause data (60, 91). The validity—i.e., the accuracy—of mortality data has been studied extensively. In this context, questions of validity relate to difficulties in ascertaining the exact cause of death. This issue is extremely important, since mortality data are perhaps the primary source of information on the consequences of illness experiences. For example, in several studies reported cause of death has been compared with autopsy findings (64, 149) and hospital diagnosis (2). Results of these studies indicate that complete concordance between clinical diagnosis and stated cause of death does not exist. Sources of discrepancies have been identified (84), 1 and their effect on subsequently computed mortality statistics estimated Epidemiologic Approaches to Chemical Hazard Assessment 139 (773). While significant problems with the reliability and validity of death certificate data have been identified, the utility of this source of information remains high, assuming that the proper interpretive safeguards are heeded. Some of these cautions were pointed out in a study that directly compared the epidemiologic inferences that could be drawn using mortality versus morbidity data. In this study it was demonstrated that both sources of data lead to essentially the same conclusions (182). For examining trends in the occurrence of disease over time or among selected populations, mortality data are the only resource that satisfies the criteria of continuity and coverage. As is discussed below, there is no single source of information on morbidity that is available for an entire country. In studies of cancer, for example, mortality data have been employed extensively because such data are assumed to be adequate surrogates for incidence data. This will be true for any condition for which the interval between onset of disease and death is reasonably short and for which the case fatality rate is high. Since mortality data are comprehensive, it is a reasonably straightforward task to compute death rates, assuming that appropriate population figures are available. These rates can then be applied in a variety of analytic models. Although it is clear that mortality data are not error-free, morbidity data introduce additional complexities. With the exception of certain infectious and communicable diseases, incident cases of disease such as cancer are not reportable to any public health agency. Therefore, only through specialized programs are incidence data collected. With regard specifically to cancer cases, there are several data systems that compile information about individuals with malignant disease. The first such category of data systems is not population based, i.e., not all cases in a defined or definable population are included. Under these circumstances it is not possible to compute incidence rates. Despite this limitation, these systems have provided a substantial portion of our knowledge about the determinants of cancer. Most of these systems are hospital-based tumor registries, the purpose of which is to collect a standardized set of demographic and medical information for each cancer patient treated at the particular institution (727). However, since patients are not admitted or referred to hospitals on a random basis, the sample of patients in a tumor registry is rarely representative of all cancer cases in the community. Although population-based analyses cannot be done with hospital registry data, these programs serve several useful research purposes. First, the health or vital status of patients is monitored over time, thus yielding data for the computation of survival rates. These are, of course, the most common endpoints for evaluating the effectiveness of cancer-directed therapy. Second, registry data allow for studies of specific variables �140 Epidemiologic Approaches to Chemical Hazard Assessment John R. Wilkins III and Nancy A. Reiches related to prognosis. Third, registries are extremely valuable for locating subjects for studies of factors related to the occurrence of disease. An ambitious multiinstitutional tumor registry program has recently been undertaken in this country. The Centralized Cancer Patient Data System (CCPDS) is a network of tumor registries at institutions designated as Comprehensive Cancer Centers by the National Cancer Institute. Data for all patients treated at these major centers are entered into a common data base, and follow-up information is obtained annually. Between July 1, 1977 and December 31, 1980, data for 142,079 tumors were registered. Although these data are not population-based (therefore reflecting the selection factors intrinsic to referral centers), the value of the CCPDS lies in the high reliability and validity of the data that are abstracted. The system has rigid coding categories, a well-defined program of quality control, frequent training sessions for nonphysician abstractors, and computerized editing of each data item. The data base provides unique opportunities for the study of cancer etiology, particularly with regard to rare tumors, only a few cases of which may be treated at a single institution (77). In addition to tumor registry data, there are some sources of populationbased cancer incidence data in the United States. The largest of these is the Statistics, Epidemiology, and End Results (SEER) program (234, 235). The purposes of the SEER program include estimating cancer incidence in the U.S., monitoring trends in survival, and identifying etiologic factors. Each of these can be studied with respect to a variety of demographic and social characteristics of the population. There are 11 geographically defined areas that participate in the program, representing about 10% of the U.S. population. Although fairly representative of the entire U.S. population with respect to age, blacks and rural residents are somewhat underrepresented. This program is an outgrowth of the End Results program and the National Cancer Surveys (49, 50). Data are collected for all new malignancies occurring in the study communities. Cases are identified from hospital charts, pathology reports, radiation therapy records, death certificates, autopsy reports, tumor registries, and private laboratories. Complete information about the patient's demographic characteristics is collected, along with data describing the anatomic site and histology of the tumor, extent of disease, and first course of therapy. Active follow-up is maintained for all cases. Although it has been established that cancer mortality rates are an acceptable surrogate for incidence rates, programs such as the Third National Cancer Survey (TNCS) and SEER provide certain information that cannot be obtained from death records. Most notable in this regard 141 are data on histologic type of the tumor and on stage or extent of disease at the time of diagnosis. The former is significant because cell type might be related to etiologic variables. The latter is important because of its relationship to survival. B. Variations of Disease Occurrence in Time The relation of disease occurrence to time is an important aspect of any epidemiologic evaluation of chemical hazards. While the occurrence of disease may be measured in terms of morbidity or mortality, time itself may be measured in terms of any appropriate dimension (hours, days, weeks, months, years, etc.). The relation between time and disease may therefore be viewed from a number of perspectives. In general, this involves examining fluctuations in disease occurrence that take place either over relatively short periods of time (e.g., over a number of hours, days, or weeks) or over relatively longer periods of time (e.g., over a .number of years or decades). Although the temporal focus of the two approaches differs, the intent of each view is to gain insight into the reason or reasons why disease occurrence has fluctuated during the time period. 7. Short-Term Fluctuations When a limited, well-defined, and homogeneous population is subjected to a single and intense chemical exposure, the effects of such exposure are likely to be manifested in a matter of minutes, hours, days, or weeks. Although the average amount of time that elapses before disease appears will depend on, at least, the mode and intensity of the exposure and on the toxicity of the agent involved, the pattern of disease occurrence in time over the short term may be expected to parallel (at least qualitatively) patterns demonstrated by common-vehicle, point-source epidemics of microbial origin. Following this type of exposure, such acute outbreaks may be described by the shape and location in time of their epidemic curves. Characteristically, onset of disease is explosive in nature; the epidemic curve is positively skewed (i.e., the distribution of onset times is log-normal); and the time interval between exposure to the agent and clincial manifestation of the illness is short. In this regard, Sartwell's method for estimating median incubation periods for infectious diseases is a traditional epidemiologic tool (796). Although originally developed to study temporal patterns of infectious diseases with relatively short incubation periods, this technique has been successfully applied to various neoplastic diseases resulting from certain chemical exposures, including �142 John R. Wilkins III and Nancy A. Reiches Epidemiologic Approaches to Chemical Hazard Assessment angiosarcoma of the liver following exposure to vinyl chloride (205, 275) and tumors of the urinary bladder following exposure to dyestuff intermediates (4). Thus, conventional epidemiologic methods pertaining to the investigation ,of common-vehicle, point-source epidemics of infectious diseases may be appropriately applied to the study of acute outbreaks of chemically related illness. When the conditions of exposure previously set forth prevail (i.e., single and intense exposure of a limited, well-defined, and ^homogeneous population), the epidemiologic interpretation is usually Straightforward, because cause and effect are close in time (192). However, when the introduction of the toxicant into the population is gradual or intermittent, and thus occurs over a long period of time, the incubation, or latent, period is likely to be measured in years or decades rather than in hours, days, or months. In this case, the epidemiologic interpretation, i.e., the linking of cause and effect, is much more difficult. human origin (i.e., an increase or a decrease in a death rate over time may be artifactual), or that changes in mortality over time may indeed reflect a true change in the incidence of the disease. Real changes in death rates, of course, may result when the genetic composition of a population changes (possibly the result of population migration or the dilution of genetic isolates), or when the environmental milieu changes. Real changes in mortality may also result when the age distribution of a population shifts or when the case fatality rate for a given disease changes. Before it can be concluded that changes in mortality rates are real, however, alternative explanations should be ruled out. In this context, at least two sources of error, both of which relate to the numerator of a death rate, must be considered. First, medical advances are likely to result in more accurate methods of diagnosis, which in turn might yield more precise determination of underlying cause of death. Consequently, a decline in the degree of misclassification of the underlying cause of death for a specific cause may result in an apparent, but spurious, decline in the cause-specific mortality rate. For example, improvement over time in the ability to identify correctly the primary site of malignant tumors in all likelihood explains the steady decline in mortality from primary cancer of the liver over the past 40 years. Misleading trends in death rates may also result from revisions in the ICD, which occur approximately every 10 years. These revisions may involve either changes in the way various disease entities are defined or changes in the actual numerical code, or both. For example, Percy et al. (772) demonstrated that the 10% increase in the number of lung cancer deaths reported in 1968 over the number reported in 1967 was the result of a procedural change in the classification of malignant neoplasms that emphasized coding to a specific site rather than, as had been practiced before 1968, coding to unspecified or unknown categories. In general, changes in mortality that result from ICD revisions are likely to be quite striking and readily identified as such. With respect to changes in mortality rates that result from improved diagnostic methods, however, the evaluation is much more difficult. Specific techniques have been suggested in this regard (128). The denominator of a death rate, i.e., the estimated population at risk, is also subject to error. This error is usually one of underestimation, the net effect of which is an artificial inflation of the rate. If the degree of error in population enumeration varies from census to census, a misleading trend in mortality may be observed. To complicate the picture further, inaccuracies in the census may not be of consistent magnitude across age, race, and sex groups (203). 2. Long-Term Variations The epidemiologic view of time and disease also involves the examination of changes in disease frequency over long periods of time. These "secular .trends" are usually investigated in terms of mortality rates because adequate morbidity data are rarely available. For example, examination of the temporal trends in sex- and site-specific cancer mortality rates for the years 1930-1978 reveals several patterns worthy of further investigation. Particularly notable are the decline in gastric cancer mortality for both males and females and the disproportionate increase in lung cancer mortality among females as compared to males (204). By contrast, other neoplasms such as pancreas, bladder, and esophagus show little change over the time period. The observed increases in cancer mortality have raised questions about the role of chemicals in the human environment. One such question focuses on toxic chemicals, the chemical industry, ' and their relation to recent temporal trends in cancer rates (57, 779, 206, 277). A related question focuses on quantifying the proportion of all cancers attributable to "environmental" factors (29, 98, 99, 101, 232). Although quite a debate has ensued and several articles addressing these issues have been published, there seems to be, at present, little chance of reaching definitive conclusions in the near future, given the overwhelming lack of relevant scientific knowledge. Part of the debate alluded to above involves the interpretation of observed time trends in mortality rates. Although specific guidelines and techniques have been proposed (123, 128), the importance of systematically -..evaluating such trends is not always appreciated. It must be recognized that apparent changes in mortality over time may result from errors of 143 �144 Finally, we introduce a note of caution with regard to the interpretation of time trends in mortality rates. If it is observed that an increase in some disease is paralleled by a concomitant increase in some measure of a putative risk factor (and if it can be shown that the increase in disease is not artifactual), then can such an observation be used to support an argument of causality for the hypothesized factor? The answer is no, although one can legitimately say that such a result is not inconsistent with the stated hypothesis. If, on the other hand, there is no coincident rise (or fall) in the disease and the "risk factor," can it then be legitimately concluded that no causal link exists? The answer to this particular question is an emphatic no. Even if artifact is considered and judged negligible or somehow appropriately adjusted for, the general approach is sufficiently insensitive to support the hypothesis of no effect. Furthermore, it must be kept in mind that the results of time-trend analyses are only a small part of the overall process of making judgments about the causal role of some putative risk factor. In essence, time-trend analyses are most appropriate when the purpose is to generate hypotheses of cause and effect, not to test them. 3. Epidemiologic Approaches to Chemical Hazard Assessment John R. Wilkins III and Nancy A. Reiches Other Perspectives of Disease and Time Although acute outbreaks of disease and secular trends in mortality dominate epidemiologic interest regarding disease versus time considerations, other perspectives may be taken. For example, many diseases (including infectious and noninfectious ones) show some sort of cyclic or repetitious pattern of occurrence in time. While the focus of study in this context has been primarily on infectious conditions and their seasonal periodicity in relation to insect vectors and certain human activities, studies of various noninfectious diseases of early life [e.g., congenital anomalies (62)] have revealed variations in risk by season of birth, suggesting the possible influence of environmental factors operating in utero or in the early postnatal period (112, 128, 140). Another view of disease and time involves the investigation of temporal "clustering," i.e., the detection of epidemics (transient excesses in the incidence or prevalence of a disease or condition). In this regard, simple epidemic curves may be constructed, or more sophisticated methods like the scan statistic (762, 222, 226) may be applied. A related and somewhat broader view involves the examination of clusters of disease in time and space. So-called space-time clusters have been of interest with respect to several diseases, including leukemia (113, 122, 216) and Hodgkin's disease (220). Several statistical methods are available to test the significance of space-time clusters (134, 167, 176), although an in-depth discussion of these is beyond the scope of this article. C. 145 Variations in Disease Occurrence from Place to Place The examination of geographic patterns of disease occurrence also plays an important role in the epidemiologic evaluation of chemical hazards. A number of strategies may be employed, each generally involving differing definitions of "place." In this regard, in order to determine if differences in disease occurrence exist among different geographic areas, basically two types of comparisons can be made: groups of countries may be compared with respect to available morbidity and/or mortality figures; or, comparisons of disease rates may be made on an intracountry basis. 1. Intercountry Comparisons Differences between countries with respect to the occurrence of many diseases can be quite striking. For example, when cancer incidence rates (for both sexes and all sites combined) are compared worldwide, a threefold difference is obtained when the highest risk countries are contrasted with the lowest risk countries (59, 759, 224). When comparisons of high-risk and low-risk nations are of a sex- and site-specific nature, extremes in cancer incidence may vary by as much as a factor of more than 500 (232). Significant differences between countries with respect to cancer mortality have also been documented (799, 252). While racial or genetic differences among the populations compared, plus other endogenous factors, account for some of the observed variation between countries in cancer incidence and mortality, the magnitude of many of the differences suggests the influence of environmental factors (57, 232). Disease rates in the lowest risk countries may be considered "baseline" (i.e., spontaneous or genetically determined) levels of cancer. Thus, the amount of cancer (or other disease) above such "natural" levels may be the result of the action of environmental forces (759, 232). This particular inference, as some have argued (29,30,69), implicates exposures of human populations to chemical carcinogens. Others involved in the debate over the proportion of all cancers due to "environmental" factors have, however, used the word environmental in a much broader context—as a synonym for any extrinsic or exogenous exposure (98-100, 232). Thus, references to environmental factors, it must be emphasized, relate not only to chemical pollutants but also to physical carcinogenic agents such as ionizing radiation, biological carcinogenic agents such as tumor viruses, and life-style influences such as diet and behavior (7). Although the results of international comparisons of disease rates have relatively limited epidemiologic utility, the exercise can serve the very useful purpose of generating hypotheses of cause and effect; it may even suggest preventive strategies (759). Moreover, once it can be established �146 John R. Wilkins III and Nancy A. Reiches that observed differences among countries are not spurious, studies of migrant populations may then be initiated to attempt a separation of the influence of genetic factors from that of environmental factors (95, 111, 116). 2. Intracountty Comparisons Many diseases, for example, most forms of cancer (25, 104, 138), cardiovascular disease (70), and multiple sclerosis (3), to name a few, show a marked geographic variation in frequency when comparisons are made on an intracduntry basis. Unlike intercountry comparisons, however, where the size of the geographic unit of analysis is fixed by national boundaries, intracountry comparisons may be performed at many levels; theoretically, any geographic unit, from the largest of subnational units (regions or states) to the smallest of subnational units (census tracts, block groups, or the like), may be used. In practical terms, though, it is usually necessary to select for study a geographic unit that best satisfies the need to compare populations that are as homogeneous as possible and, at the same time, large enough to yield stable disease rates. As Hoover el al. (103) and Blot and Fraumeni (27) indicate, the optimal geographic unit of study seems to be the county, at least when epidemiologic interest centers on environmental causes of cancer. The initial work of Mason and McKay (737, 738) illustrates the approach. a. Mapping Cancer Death Rates. For the period 1950-1969, age-, race-, sex-, and site-specific cancer mortality rates were computed from National Center for Health Statistics death certificate data and U.S. Census figures for each of the 3056 contiguous U.S. counties (737). To allow valid comparisons among the counties, the death rates were agestandardized to the 1960 U.S. population, yielding average annual race-, sex-, site-, and county-specific rates for the 20-year period. From these summary data, an atlas of cancer mortality, color-coded to five levels, was created by comparing statistically each subgroup-specific county rate to the appropriate national rate and by, at the same time, classifying the rates into deciles (738). For the rarer malignancies, state economic areas (defined as groups of similar, contiguous counties) were used as the geographic unit of analysis. Virtually all of the resultant maps demonstrated that cancer death rates vary in magnitude across U.S. counties (or across state economic areas) in a nonrandom fashion. Furthermore, the number of identifiable "clusters" of high (or low) rate counties, the size of a given cluster, and the location of clustering depends on both the site of disease and on the sex-race subgroup examined. Striking geographic Epidemiologic Approaches to Chemical Hazard Assessment 147 patterns emerged for several malignancies, including, notably, cancers of the bladder, esophagus, lung, stomach, and oral cavity. This mapping approach is quite useful for a number of reasons. First, from a technical point of view, a large amount of data can be analyzed by computer relatively quickly and inexpensively. Second, because the data are presented visually, high-risk populations can be identified rapidly, which in turn, provides a firm basis to form causal hypotheses that can then be pursued by other means. Furthermore, mapping studies, sometimes referred to as the geographic pathology approach (87), serve as an important first step in a logical sequence of epidemiologic studies based on countylevel cancer mortality data (24). b. Correlation Studies. Although the benefits of mapping disease rates are clear, the identification of high-disease areas by this technique creates somewhat of an interpretational dilemma. Can the disease clustering be explained by the demographic and/or genetic characteristics of the people that inhabit the area, or are the chemical, physical, and biological (i.e., environmental) characteristics of the place responsible for the elevated disease rates in the resident population, or is the explanation some combination of these two factors? These questions give rise to a class of investigations often referred to loosely as "correlation" studies. The purpose of this type of study, simply stated, is to identify those demographic and environmental characteristics of the populations in question that covary with the disease rates, thereby providing etiologic clues. Such studies rely primarily on routinely collected data, such as U.S. Census figures and death certificate data. Quantitatively, although numerous statistical methods may be employed, correlation studies fall basically into one of two general categories: studies that employ standard univariate methods of analysis, i.e., studies that use the bivariate correlation coefficient to measure association between a single factor and a disease; and studies that employ multivariate methods of analysis. This typically involves the investigation of multiple risk factors for disease with standard multiple regression techniques. Schroeder's paper on various chemical and physical properties of finished drinking water and cardiovascular disease mortality exemplifies the univariate approach (198). In this study, average annual age-adjusted mortality rates from cardiovascular disease for the period 1949-1951 for white males aged 45-64 years were plotted as a function of the average drinking water hardness in the 48 contiguous United States plus the District of Columbia. Bivariate correlation coefficients were computed for four categories of cardiovascular disease and for all causes of death combined; �148 John R. Wilkins III and Nancy A. Reiches four of the five correlation coefficients were negative in sign and were statistically significant at the p < 0.01 level. A similar analysis of coronary heart disease and 21 constituents of water in the 163 largest U.S. cities yielded highly significant (negative) correlation coefficients for magnesium, calcium, bicarbonate, sulfate, fluoride, dissolved solids, specific conductance, and pH, prompting Schroeder to write "the data offer a clue to an environmental influence associated with the nature of public water supplies which affects adversely the course of degenerative cardiovascular diseases in the United States." This particular approach (i.e., the use of bivariate correlation coefficients to measure association between some factor and some disease) has some notable limitations. First, the magnitude of a correlation coefficient can be affected significantly by the size of the geographic unit used for analysis. Although the phenomenon is not always appreciated, the simple aggregation of geographic areas into larger units will usually result in an increase in the size of the correlation coefficient, an increase which can, in reality, be quite large. As Blalock (18) explains, a shift from smaller to larger geographic units will tend to reduce the effect of so-called nuisance variables, variables that are causally related to Y (the dependent variable of interest, i.e., disease) but that are unidentifiable and/or unmeasurable. Thus, as geographic areas are aggregrated they become more homogeneous with respect to the nuisance variables, which in turn allows the single independent variable of interest (X) to account for, or "explain," a greater proportion of the variation in Y. It is not surprising, therefore, that a fairly high (negative) correlation between hardness of drinking water and cardiovascular disease can be obtained when comparisons are made on a state-by-state basis, while the association, in general, disappears as the geographic unit of study gets smaller and smaller (43). Another important consideration is that the correlation coefficient itself does not measure the strength of an association, but merely reflects the degree of dispersion of the data points about a straight line. Since it is the regression, not the correlation, coefficient that measures the effect of changes in X on Y, its use is preferred as a measure of association between factor and disease. Further, the magnitude of a regression coefficient (i.e., the slope of a line in a bivariate analysis) is theoretically not influenced by shifts in the size of the geographic unit of study. The univariate approach discussed above, whether correlation or regression coefficients are used, cannot deal with the complex of factors related to the occurrence of human disease. Consequently, in order to help identify the cause (or causes) of environmentally related illness, a multivariate approach must be employed. In general, this involves the application of standard multiple regression techniques, a strategy based Epidemiologic Approaches to Chemical Hazard Assessment 149 on the assumption that disease rates (e.g., county-level cancer mortality rates) can be expressed as a linear or nonlinear function of a set of demographic, socioeconomic, and environmental characteristics of the population(s) in question. To illustrate, once maps of county-level cancer mortality data reveal geographic clusters of elevated death rates for specific neoplasms, multiple regression studies may be performed to identify those demographic, socioeconomic, and environmental characteristics of the apparent high-risk populations that are statistically related to the cancer death rates (2723, 76). The county-level cancer mortality data [their origin described earlier (137)] are treated as dependent variables in regression equations, and relevant county-level demographic, socioeconomic, and environmental data are entered as independent or predictor variables. Statistical significance of regression coefficients indicates which variables are significantly associated with the cancer rates, while the sign of each coefficient indicates the direction of the association. The county-level demographic, socioeconomic, and environmental data are obtained from such sources as U.S. Census publications and computer tapes. Quantities such as percentage of the population that is nonwhite, percentage that is urban, percentage residing on farms, population density, median family income, median number of school years completed by the adult population, percentage foreign stock, and various industrial indexes [derived from the U.S. Census of Manufactures (279)] are typically included in regression equations. For example, after controlling for the effects of demographic and socioeconomic differences, Blot and Fraumeni (21) found lung cancer mortality rates in white men to be significantly high in those U.S. counties where the paper, chemical, petroleum, and transportation industries tended to concentrate. Interestingly, death rates from lung cancer among white females were found not to be significantly associated with any of the industrial indices examined, a result consistent with the hypothesis that occupational exposures account for a measurable proportion of all lung cancer deaths. A methodologically similar study by Blot and Fraumeni (23) reported a statistically significant (positive) association between cancer of the urinary bladder among white males and the geographic concentration of the chemical and printing industries. As with their lung cancer study, industrial indices were found not to be related to bladder cancer mortality in white females. Comparable studies, one of which describes their method in detail (22), have investigated demographic, socioeconomic, and industrial correlates of oral (22) and esophageal (76) cancers. The initial focus of the multivariate approach described above is on a specific disease (or possibly on a group of diseases). In other words. �150 John R. Wilkins III and Nancy A. Reiches given a specific health outcome, a study is made to identify demographic, socioeconomic, and environmental factors of potential etiologic significance. It is possible, however, to reverse this logic by asking the following question: Given an a priori interest in a specific risk factor, what disease or group of diseases could be related to the factor in question? Thus, the initial focus of a correlation study might be on some environmental variable, perhaps a certain chemical exposure, rather than on a specific health effect. This approach, which generally utilizes the same multiple regression techniques discussed above, has been applied primarily to the investigation of various industries and/or occupations suspected of being cancer hazards (87,211), including the chemical (102), petroleum (79), metal electroplating (17), and furniture (32) industries. Studies of this kind have also addressed the possible cancer risk associated with the contamination of water by asbestos (739), fluoride (105), and organic chemicals (228)', of air by arsenical compounds (20); and of soil by uranium mill tailings (136). c. Nature of the Ecologic Study. The correlation studies referred to above, whether analyzed in univariate or multivariate fashion, share a very important characteristic: the data employed, and this relates to both the independent and the dependent variables, are in aggregate form, i.e., they (the data) are organized at a group level, thus providing information about human populations in a collective sense. Quantities derived from U.S. Census data, such as the proportion of a county population employed in a given industry, illustrate the point. Thus, investigations that rely on group- or aggregate-level data are commonly referred to as "ecologic studies," a descriptor having origins in the social sciences (61, 89, 150, 184). Ecologic studies, by virtue of their use of aggregate-level data, possess various limitations. A major concern in this regard pertains to the interpretation of the results of an ecologic analysis. First and foremost, since study subjects cannot be classified on an individual basis with respect to the study variables, any results suggesting an association between some factor and some disease must be regarded as indirect and therefore not conclusive. It should be appreciated that the interpretation of ecologic data is subject to an "ecologic fallacy," in this case, the error of ascribing to individuals associations or characteristics based on an analysis of aggregate-level data. This particular fallacy, the "aggregative fallacy," pertains to improper inferences made from the aggregate to the disaggregate (212). Improper inferences may also be made in the other direction, i.e., from the disaggregate to the aggregate; this type of ecologic fallacy is referred to as the "atomistic fallacy" (2/2). Ecologic studies, therefore, cannot be used to test formally some hypothesis of cause and effect. Epidemidogic Approaches to Chemical Hazard Assessment 151 They can, however, be used quite successfully to generate clues to the etiology of disease, clues that are pursued by more rigorous methods of study. Ecologic analyses are fraught with other methodologic problems. These include: 1. Inability to incorporate the concept of a latent period. It will usually not be possible with routine data to construct a proper temporal relationship between measures of the hypothesized cause and measures of the hypothesized effect, i.e., it may not be possible to take into account a latent period—the time between the biologic onset of disease and its clinical manifestation (792). For example, in studies (720, 170) of cancer and organic chemical contamination of public drinking water supplies, data on the hypothesized cause (drinking water) pertained to 1963 and data on the hypothesized effect (cancer mortality) pertained to the 20year period, 1950-1969. In this case, for the measure of cause to precede in time in an appropriate fashion the measure of effect, the drinking water data should have pertained to a time period well before the 19501969 vicennium. The actual magnitude of a latent period to incorporate into such an analysis depends, of course, on the specific chemical agent and disease in question, as well as on the nature of the exposure. Moreover, if the study period overlaps with the latency period, regardless of when population exposures occurred, the full effect of the exposure cannot be determined because not all cases of disease attributable to the exposure will have had time to become manifest. 2. Artificial nature of the boundaries of geographic units of study. Geographic areas demarcated by natural boundaries such as mountain ranges or major rivers, as contrasted to areas defined by political or administrative boundaries, are more likely to be homogeneous with respect to demographic and environmental characteristics of etiologic significance, and thus would be more likely to define areas of high (or low) disease occurrence. The boundaries of most (though not all) states, counties, cities, etc. do not coincide with natural boundaries. Therefore, the artificial and arbitrary nature of politically established geographic areas creates problems in ecologic analyses because such boundaries may either subdivide homogeneous regions or combine heterogeneous ones (140). Further, it may not even be possible to obtain data on all variables of interest for a given type of areal unit. This creates a comparability problem, which may be compounded if the political/administrative boundaries shift over time. Such changes may then preclude any investigation of the temporal relationship between an exposure and a disease (209). 3. Difficulties in measuring human exposures to toxic chemicals in ecologic studies. With the kind of data typically available for use in ecologic studies it will usually only be possible to employ fairly crude �152 Epidemiologic Approaches to Chemical Hazard Assessment John R. Wilkins ffl and Nancy A. Reiches measures of human exposure to toxic chemicals. In general, such measures will be indirect (rather than direct) and qualitative (rather than quantitative). Since ecologic measures of exposure are in aggregate form, like the other variables in an ecologic study, they cannot be truly representative of the exposure experience of individuals comprising the study population. This aspect of conducting epidemiologic studies of chemical hazards is discussed in more detail in Section IV. 4. Effects of population migration. Migration of people between geographic areas will affect adversely the sensitivity of an ecologic study. As Polissar (178) points out, a geographic area assumed to'be inhabited entirely by a so-called exposed population is actually inhabited by some who are and by some who are not exposed to the agent or factor in question because of in-migration of unexposed people. Polissar demonstrates how one measure of disease risk, the Standardized Mortality (or Morbidity) Ratio (SMR), can be affected by differing degrees of migration. He shows, with some simplifying assumptions, that the SMR is a function of (i) the proportion of the exposed population that is truly exposed because they have not migrated, (ii) the size of the exposed population, (iii) the rate of death or disease in the control (unexposed) population, and (iv) the ratio of death or disease in the exposed population to that in the control (unexposed) population. Polissar also shows that the magnitude of excess risk observed in ecologic studies in the presence of migration varies with the age of the population, the particular disease in question, the duration of the latency period, and the type of geographic unit used for study. 5. Some technical issues. Most of the technical (statistical) problems in ecologic analyses relate to assumptions associated with any multiple regression problem. In many cases, underlying assumptions of the general linear (or nonlinear) model cannot be met strictly by the data. These include assumptions of normality with respect to predictor and dependent variables, homoscedasticity of variances, and independence of observations. Fortunately, the regression model is quite robust with respect to the first two assumptions, i.e., they can be violated substantially before the validity of results is threatened. The third assumption, however, can pose serious problems. In many instances, the distributions of two or more predictor variables are not independent, which means that they are correlated. This situation adversely affects the estimates of the regression coefficients and their subsequent interpretation. In many cases, however, this problem, called multicollinearity, can be solved by the use of two-stage or even three-stage least squares, rather than ordinary least squares, regression. Perhaps the most serious problem in ecologic analyses relates to the question of specification of the model to be evaluated. If a predictor 153 variable that is significantly related to the dependent variable is omitted (owing to lack of data or lack of knowledge that the variable is important), specification errors will result. If the omitted variable is correlated with a variable that is specified in the equation, the included variable will appear to be more strongly related to the dependent measure than is actually the case. If this occurs, a variable might erroneously be associated with the disease under consideration. Errors of this type in hypothesisgenerating studies can be dangerous, since they will mislead the investigator and probably result in an untenable etiologic hypothesis. Unfortunately, there is rarely enough knowledge available at the time of a preliminary study to determine whether a specification error has indeed occurred. However, the possibility emphasizes the caution noted earlier that causal inferences cannot be drawn from correlational results, but that such findings must be regarded as tentative. Once descriptive epidemiologic tools have generated hypotheses regarding the potential adverse effects of chemical exposures on human health, studies will then be conducted to'test formally such hypotheses. Studies of this type, "analytic" studies, require individual-level data on the traits and characteristics of study subjects. With such disaggregate data it will be possible, as it is not in ecologic studies, to classify each study subject with respect to the study variables. III. TESTING ETIOLOGIC HYPOTHESES: AN OVERVIEW OF ANALYTIC APPROACHES In this section the major analytic approaches typically employed to estimate potential associations between an exposure and a defined health outcome are discussed. The two primary methods can be distinguished on the basis of how the study samples are selected. In the case-control approach, individuals with a specific disease are compared with persons believed to be free of the condition under study. The cohort approach evaluates the occurrence of disease within a group defined in terms of characteristics prior to the diagnosis of disease. Both cohort and casecontrol methods can be defined further on the basis of whether the study is conducted retrospectively or prospectively. Many case-control studies are conducted retrospectively, i.e., the data collected are historical in nature. However, it is also possible to conduct prospective case-control analyses, in which the sample is accumulated over time as new cases of the disease occur. Cohort studies can also be conducted forward or backward in time. In either case, the distinguishing characteristic of such analyses is the long-term observation of a group of people, which is accomplished by prospective monitoring of the study subjects or by �154 John R. Wilkins IH and Nancy A. Reiches historically tracing the experience of the sample over a defined time interval. Prospective cohort studies are sometimes referred to as concurrent studies, while historical cohort analyses may be called nonconcurrent (123). The major features of each analytic model and some of their relative advantages and disadvantages are now described. A. Case-Control Studies Case-control studies are an excellent method for evaluating the relationship of an exposure or hypothesized etiologic factor to the occurrence of disease. In its most fundamental form, the purpose of a case-control study is to determine whether individuals with a disease are more (or less) likely to possess some characteristic (or exposure) than persons without the disease. These studies are designed to assess whether exposure to some factor of interest places individuals at higher risk of disease than those individuals not exposed. Statistical techniques applied to data from case-control studies can evaluate risk associated with two or more levels of exposure. Perhaps most fundamental to the conduct of a case-control study are those issues that bear on the selection of the study subjects. In selecting cases it is essential to confirm that the potential subject is indeed a case, i.e., that he or she strictly meets the diagnostic criteria of the condition under study. With respect to studies of cancer, for example, such confirmation could be in the form of microscopic pathologic analysis of a tissue sample, rather than merely a clinical or radiologic diagnosis. Even under conditions of laboratory confirmation, misclassification can occur, so it is important to minimize this bias, where feasible (63). Furthermore, it has been suggested that cases have a reasonable probability that their disease might have been caused by the hypothesized agent, and not by another identifiable factor (109). Cases can be identified through several sources. These include all persons (or more usually a probability sample of persons) diagnosed during a specified period of time in a given community or in a single hospital or group of hospitals. Often it is more practical to identify cases through the records of one or just a few medical care facilities. However, this can introduce a bias into the sample, since there are systematic selection factors that guide certain individuals to a particular facility. If such a bias is present, study cases will not be representative of the entire population of persons with the disease, possibly leading to erroneous inferences about the etiologic factor of concern. In principle, sampling cases from a general population has great theoretical appeal, but can be laborious and expensive. To the extent possible, the assumption of rep- Epidemiologic Approaches to Chemical Hazard Assessment 155 resentativeness should be met or the possible extent of bias estimated. Even with a condition such as cancer, certain nonrandom cases do escape medical attention. The question of selecting appropriate controls poses even more difficult questions. Simply defined, a control is an individual with no clinical evidence of the disease under study. Ideally, the control group will be a representative sample of disease-free persons. Furthermore, it is desirable that the controls be members of the same general population from which the cases derive. Control groups can be selected from several sources. These include hospital patients, residents of the same geographic area as the cases, and relatives or other associates of the cases. Selecting cases from the same geographic area is appropriate if the cases are representative of that population. Hospital controls are often used since they are a relatively easy source to obtain. The major disadvantage of this source is simply that hospitalized persons are ill and may therefore be unrepresentative of the general population with respect to a complex of illness-related factors. This may introduce a particular type of selection bias, especially if many of the controls are of a similar diagnostic group. This effect may be minimized by choosing controls from several diagnostic categories (41, 135). The extent of selection bias has long been known (94) and has been comprehensively discussed (10). Although selection biases do not necessarily invalidate study findings, one must carefully interpret whether an observed association is likely to be real or spurious (34, 58). Another significant issue with regard to the selection of cases and controls involves the question of matching. Since the purpose of a casecontrol study is to measure the effect of a defined exposure, it is desirable to eliminate by design those factors that might potentially confound the results. Matching is a process of selecting controls known to be similar to the cases with regard to specific characteristics such as age, race, sex, or socioeconomic status. Effects of variables known to be associated with both the disease and the study factors can be controlled by matching (752). The primary disadvantage of matching is that the etiologic role of the matching variable cannot be evaluated, since, by definition, cases and controls are alike with regard to that characteristic. Also, matching can increase the complexity of a study, with respect to both design and analysis (13, 14,143). Finally, there is a risk of overmatching or unnecessary matching. In general, inappropriate matching can reduce the statistical efficiency of the case-control comparison (757, 200). Collecting accurate and valid information on exposure for both cases and controls is a crucial aspect of case-control studies, since resulting estimates of risk are directly related to these measures. The precise �156 Epidemiologic Approaches to Chemical Hazard Assessment John R. Wilkins III and Nancy A. Reiches meaning of exposure (to be expanded on in Section IV) must be defined, with regard to both intensity and duration. Most importantly, the exposure information must be comparable, with respect to reliability and validity, for cases and controls. If the data are incomplete, spurious associations will result. Medical or vital statistics records and interviews with study subjects provide the major sources of exposure data; both sources entail potential biases. For example, as discussed earlier, there are possible sources of error with public records, such as death certificates. Clinical records may exhibit some of the same problems, or they may simply be incomplete with regard to data concerning the exposure of interest. Since the purpose of the record is to chart the clinical course of disease, information required to study other phenomena may not be available (72). Additionally, questions of validity of data recorded in the medical record have been raised (7/5). Finally, reliable (reproducible) abstracting of clinical data from existing records cannot always be achieved (28). Despite these potential limitations, the medical record remains a key source of data regarding exposure; its intrinsic value in the study of disease etiology is unchallenged (779). Data collected by interviewing cases and controls also presents some possible biases, but this method also assumes great importance in epidemiologic analyses. Information obtained through personal interviews (or self-administered questionnaires) is subject to the pitfalls of faulty respondent memory, unintentional errors in reporting, or outright prevarication. Bias might occur, for example, if the occurrence of disease has prompted the respondent to recall certain related information that might otherwise have been forgotten. If corresponding information does not emerge for the control, a bias is introduced (194). Further, the passage of time since the relevant event might affect the validity of the reported information (270). In other instances respondents might have been unaware of exposure and consequently are unable to report it. A number of studies have been conducted to evaluate the reliability and validity of self-reported data, many of which offer encouraging results (67, 114, 187). Errors in the collection of exposure data or noncomparabilities of data between cases and controls can result in serious misclassification problems, i.e., erroneously determining an individual's exposure status. The result of misclassification is an inaccurate estimate of risk. It has long been appreciated that even random and independent errors can reduce the measured association between exposure and outcome (33, 164). This topic has been reviewed extensively, and methods to adjust for misclassification under specified conditions have been proposed (45,85,110). Careful attention to study design can preclude or diminish many errors 157 of misclassification. Standardized provisions for data collection may at least ensure a high degree of reliability, a major prerequisite for validity. The analysis of data from a case-control study is primarily a comparison between cases and controls regarding the presence of hypothesized etiologic factors in each group. Results of these analyses indicate whether there is an association between the factor and disease. In principle, one desires to estimate the relative risk associated with exposure (i.e., the incidence rate among those with the factor divided by the incidence rate for persons without the factor). However, the method by which cases and controls are assembled does not include all exposed and all unexposed individuals. Consequently, the incidence rates of interest cannot be calculated. If, however, assumptions about the representativeness of cases and controls can be met, a measure known as the odds ratio can be computed as an estimate of relative risk (46, 47). In the simplest case, data from a casecontrol study can be presented in the form of a 2 x 2 table, with columns representing the classification of cases and controls, and rows representing the presence or absence of the exposure factor (Fig. 1). The odds ratio, given by the expression (ad)/(bc), summarizes the probabilities of having or not having disease. The statistical properties of the odds ratio have been analyzed extensively. Numerous methods have been proposed for tests of significance (83, 230) and for approximating confidence intervals (46, 96, 214). Furthermore, there are techniques to adjust the odds ratio for the effects of confounding variables through stratification (75, 90, 135). Appropriate statistical management of the odds ratio also depends on the degree of matching in the study design (148). In addition to stratification, control can be introduced by logistic Status of study subject Case Control ' Present a b Absent c d a + c b + d Exposure • a+b+c+d Fig. 1. Cross-classification of subjects in a case-control study. �158 John R. Wilkins in and Nancy A. Reiches Epidemiologic Approaches to Chemical Hazard Assessment models, which permit adjustment for variables that were not matched in the study design (31, 180). The epidemiologic literature is replete with examples of investigations employing the full range of study designs and analytic techniques described above. For example, death certificates were used as a primary source of data in an analysis of sinonasal cancer among males (193), for which several chemical agents are hypothesized etiologic factors. Exposure to a variety of agents was inferred from occupational data on the death certificate; complete occupational histories and quantitative exposure data were not available. However, probability of exposure to nickel, cutting oils, and wood dust was estimated from occupational titles and industry of work. The odds ratio computed for nickel exposure was not statistically significant, but was with respect to cutting oils and wood dusts. An example of the evaluation of a multiplicity of factors can be found in an ambitious study of bladder cancer (706). The effects of several exposures, including tobacco, coffee, various nutrients and nitrates in food, and occupation were estimated. A logistic model was used to deal with the multivariate design, thereby affording an opportunity to measure the independent effects of exposures, their interaction, and the effects of confounding variables. The applicability of case-control studies to provide preliminary information that might account for an unusual cluster of cases is demonstrated by an analysis of mortality from pancreatic carcinoma (775). Although the findings are somewhat constrained by the limited residential and occupational information available on the death certificate, a significant odds ratio was obtained for persons who worked in oil refining or paper manufacturing. A small effect was detected among persons living near refineries. A study of this type is useful for defining requirements for a more extensive interview study and is particularly interesting because of the implication of occupational and ambient environmental risk. Finally, the conduct of an interview study is shown in an analysis of exposure to artificial sweeteners (157), and the use of more than one control series is demonstrated in another investigation of pancreatic cancer (729). B. Cohort Studies The second major approach in analytic epidemiology is the cohort study, of which the primary design features are discussed here. Although cohort studies differ from case-control studies with respect to the way in which study subjects are selected, the majority of issues that pertain to the validity and analysis of data obtained in cohort studies are equivalent 159 to the issues considered with respect to the case-control method. Specifically, similar and equal attention should be given to the representativeness of the sample, the effects of both disease status and exposure misclassification, other selection biases, sources of exposure data, and problems of confounding variables. Indeed, many of the statistical approaches to the resulting data are the same, or entail the same assumptions, and therefore are not considered in detail here. The basic concept of a cohort study is relatively straightforward. A sample of a population is selected, and it is determined which members of the cohort possess the study characteristic or are exposed to the hypothesized etiologic agent. The cohort is then followed over time, and the incidence rate of the disease under consideration is calculated for the exposed group and for the unexposed group. If the rate of disease is higher among those exposed, an association between the risk factor and disease is inferred. In the retrospective, or nonconcurrent, cohort approach, the period of observation is historical, a method often used to study specially exposed groups such as industrial populations. Several examples of this method will be discussed in the context of measuring response by computing standardized and proportionate mortality ratios (Section IV). One difficulty in retrospectively assembling a cohort is in assuring that all members can be identified. Sometimes, comprehensive data are not available (37, 221), thus limiting the generalizability of the findings. In addition to special exposure groups, cohorts can be defined because they can be followed over time and because methods for identifying outcomes of interest are available. Examples of groups that have been studied include persons enrolled in prepaid medical care plans, groups of insured persons, obstetric populations, and volunteer groups. Additionally, a cohort may be defined on the basis of geography, such as all members of a specifc community. One of the most crucial aspects of a cohort study relates to followup, i.e., the task of determining outcome, usually the appearance of morbidity or mortality. It is important that determination of outcome be equally complete for exposed and unexposed cohort members, or for cohort members in each level of exposure in the nondichotomous case. Otherwise, measures of association between exposure and disease will be biased. Therefore, it is important to establish a follow-up (or tracing) mechanism that applies equally to all study subjects, whether the surveillance entails review of routine records or special data collection efforts. The length of follow-up is also a significant determinant in the results of a cohort study. If follow-up is not sufficiently long, cases of the study �160 John R. Wilkins III and Nancy A. Reiches disease will not have yet become clinically apparent (especially if the latent period is long), and the rates of disease will therefore be underestimated. Comparisons of findings under two different follow-up periods have been reported. One such example involves two analyses of a cohort of workers exposed to beta-naphthylamine and benzidine (750, 757). The time-of-measurement effect has also been discussed in a theoretical framework (257). The consequences of losing some proportion of persons during followup can be considerable, particularly if the losses are not random. If losses are biased with respect to outcome, the absolute rates of the study condition will be influenced, but their relative relationship to each exposure category will remain the same. Substantial losses, however, can distort the measurement of risk. A more serious situation will result if followup losses are biased with respect to exposure category, since this will affect the relative rates of disease between exposure groups. In some circumstances it is possible to estimate the effect of follow-up losses, particularly if the date on which an individual leaves the cohort is known. There are many examples of cohort studies designed to evaluate the effects of environmental exposures. Despite limitations in data availability, results of these investigations can be very revealing. For example, analysis of a cohort of persons engaged in the manufacture of mustard gas was limited in that only 84% of the cohort could be traced (755). To compensate, additional calculations were made under the extreme conservative assumption that all persons untraced were alive at the termination of the follow-up interval. Even under these conditions, a positive association was detected. The relative advantages and disadvantages of case-control versus cohort studies can be summarized briefly. Case-control studies are reasonably efficient and inexpensive to conduct. Comparatively few subjects are required, since the study begins with the identification of cases. This efficiency is particularly apparent in etiologic investigations of rare diseases, although the same advantage can accrue to retrospective cohort studies. By contrast, it is impractical, if not impossible, to assemble a large enough cohort to study in prospective fashion the occurrence of a rare condition, since the probability of any cohort member exhibiting the disease is extremely small. Case-control studies also offer an advantage with respect to time, since it is not necessary to wait for the development of new disease. Analytically, the major disadvantage of a case-control study is that relative risk cannot be measured directly, and there is controversy regarding the most appropriate estimation and testing of the odds ratio. Further difficulties, such as selection of the most appropriate control group, have been alluded to above and are discussed extensively in the literature (41, 108, 109). Epidemiologic Approaches to Chemical Hazard Assessment 161 By comparison, cohort studies have the advantage of classifying persons with respect to exposure prior to the development of disease. This can minimize (although not necessarily eliminate) problems of bias and misclassification. Most significantly, cohort analyses can yield actual incidence rates, thereby providing a direct measure of risk. The primary disadvantages of the cohort method relate to obstacles encountered in the follow-up of a large number of study subjects. Prospective cohort studies are expensive to execute and generally represent very large-scale undertakings. Consequently, they are not efficient for exploring new hypotheses; their strength lies in the provision of additional evidence after a specific hypothesis has been posited. IV. CRUCIAL ASPECTS OF ENVIRONMENTAL STUDY DESIGN To estimate accurately health risks resulting from chemical exposures, the relation between the amount or concentration of the agent in the critical organ or tissue (i.e., the dose) and the proportion of the population at risk manifesting a specified biological effect (i.e., the response) must be determined. In the following two subsections, epidemiologic approaches (and attendant problems) with respect to making measurements of dose and response in human populations are examined. A. Measurement of Dose Perhaps the most problematic aspect of designing epidemiologic studies of chemical hazards in human populations involves the measurement of dose. Although attention in this regard centers (properly) on considerations of the intensity, duration, and mode of external exposure to environmental chemicals, the problem of dose estimation is actually much more complex. The difficulty is easily appreciated by considering the many factors, conditions, and forces that affect the actual degree of external exposure, and thus, ultimately, the amount of toxicant reaching the critical organ or tissue. Therefore, if at first just those factors that determine the transport and fate of chemicals in the ambient environment are considered, many relevant questions may be posed. For example, what is (are) the source (sources) of the chemical agent in question? Is it a point or a nonpoint source? In what amounts and at what frequency is the substance discharged into the environment? Is the substance primarily of natural or anthropogenic origin, or a combination of the two? What are the patterns of use and of disposal of the material? Once the substance enters the human environment, how does it behave in air, in water, in soil, in �162 John R. Wilkins III and Nancy A. Reiches biota? To what extent is the substance transported between the various environmental compartments? To what extent will the chemical undergo a transformation in either air, water, soil, or biota? If a transformation does occur, will the product be more hazardous or less hazardous than the parent material? Is it possible to identify populations that are most likely to be exposed and/or most likely to be exposed to the highest levels of the substance? Are there subgroups of the population that are particularly sensitive to the agent in question? And once human exposure does occur, what information is available on the absorption of the agent into the body, on the distribution of the agent in plasma and tissue, and on the pathways and rates of excretion? Before proceeding further, the difference between exposure and dose must be clarified. Unfortunately, the distinction is not always made. Exposure to a given chemical substance refers simply to the extent of contact between the toxicant and those surfaces of the human body where absorption may occur. Thus, measures of exposure, i.e., measures of external exposure, are expressed in terms of the concentration of the agent in environmental media (air, water, food) that interface with relevant body membranes (11). Dose, on the other hand, refers to the amount or concentration of the toxicant in a critical organ or tissue [the critical organ or tissue being that which exhibits the first or the most serious effect (77, 765)]. It is the dose that must be obtained in order to quantify health risks resulting from chemical exposures. In general, however, the amount or concentration of a toxicant in a critical organ or tissue cannot be measured directly. Thus, dose must be measured in some indirect fashion and in such a way that the index used will result in an observed dose-response curve that accurately depicts, in both qualitative and quantitative terms, the true or actual dose-response relation for the substance in question. Surrogate measures of dose, then, will generally be derived from data either generated by some form of biological monitoring [the "systematic collection of human or other biological specimens for which analysis of pollutant concentrations, metabolites, and biotransformation products is of immediate application" (77)] or by some form of environmental monitoring [the "systematic collection, analysis, and evaluation of environmental samples, such as air, water, or food for pollutants" (77)]. 7. Biological Monitoring Given the usual inability to measure directly dose of the toxicant at the effector site, it would seem appropriate to assume that levels of the toxicant (or of its metabolites) in blood or other accessible tissue(s) would correlate with levels in the critical organ or tissue and thus could serve Epidemiologic Approaches to Chemical Hazard Assessment 163 as a reliable and valid index of dose. If so, data derived from measurements made on human tissues or excreta could be used to clarify dose-response relationships. For example, levels of lead and other heavy metals (such as cadmium and mercury) in blood (755, 233), urine (233), hair (97), or nails (107) have been used in the past as dose indices. Other tissues or excreta, including breast-milk (756), adipose tissue (275), expired air (44), and others (77), are at present amenable to biological monitoring. Unfortunately, there are relatively few applications in epidemiologic studies of chemical hazards (42, 88). However, there is a growing opportunity and need for a greater integration of toxicologic and epidemiologic data for purposes of health hazard assessment, as evidenced by recent papers (55, 275, 229). For a more detailed discussion of biological monitoring per se, see references 77 and 236. 2. Environmental Monitoring Most surrogate measures of dose 'used in epidemiologic studies are actually measures of external exposure to some agent, since they are usually derived from some sort of environmental data. Further, indices of dose based on such data may be divided into two categories: those that can be characterized as simple classifications and those that are based on Haber's law. a. Simple Classification Schemes. Measures of external exposure to toxic chemicals should, ideally, reflect the intensity, duration, and mode of the particular exposure. Further, such measures should be quantitative in nature, should accurately summarize the exposure experience over time of any individual study subject, and should reflect the amount or concentration of the toxicant in the critical organ or tissue. In order to develop such measures, detailed, extensive, and individualized environmental data are required. Such data, however, are not, for a number of reasons, usually available. For one thing, the substance or group of substances in question may have only recently come to be thought of as hazardous, in which case ambient and/or occupational environments will probably not have been routinely monitored in past time periods. Further, there may exist no analytical techniques capable of measuring the substance or substances in air, water, soil, and/or biota; or, if analytical methods are available, they may be grossly inadequate. Also, there may be no mechanism, practical or otherwise, to link levels of exposure to the environmental contaminant(s) in question with specific individuals. Consequently, when numerical estimates of external exposure cannot be calculated for individual study subjects, for whatever reason, �164 John R. Wilkins III and Nancy A. Reiches what may be described as simple classification schemes are developed instead. Although these simple schemes may be devised from either qualitative or quantitative environmental data, they generally take the form of ordinal-level measurements, measurements that do provide a rank ordering of exposure categories but do not provide an indication of the "distance" between categories. Further, exposure classification schemes of this sort are usually (but not always) employed in aggregate population studies, i.e., in studies comparing morbidity and/or mortality rates among large geographic units such as states, counties, communities, etc. When the classification scheme is based on qualitative information only, the exposure variable itself will likely be constructed as a simple high/medium/low, or a simpler high/low dichotomy. For example, in the prevalence study of chronic obstructive respiratory disease by Detels et al. (55), several lung function parameters were compared between the populations comprising two California communities. The "high" exposure community was described as "chronically exposed to relatively high levels of photochemical/oxidant-type pollutants," and the other, the "low" exposure community, was "subjected to low levels of chemical ambient air pollutants." In addition to high/low schemes and variations thereof such as exposed/ unexposed, aggregate populations have also often been categorized in terms of their degree of urbanization. Since U.S. Census data make this a relatively straightforward procedure, many studies report comparisons of morbidity and/or mortality rates among "urban" and "rural" populations, as well as among populations classified in similar ways (such as, for example, SMSA1 county with central city/SMSA county without central city/non-SMSA county). A number of studies employing this general approach, including notably several studies of air pollution, have focused on the differences in sex- and site-specific cancer death rates between urban and rural populations. In this regard, attention has centered on the cancer death rates among populations characterized by differing levels of urbanization. After a review of several such studies the general conclusion, as Carnow and Meier (55) point out, is that mortality from respiratory cancer is roughly twice as high in urban areas as in comparable rural areas, results consistent with the hypothesis that the higher levels of airborne carcinogens generally found in urban areas are etiologically involved in pulmonary neoplasms. Similarly, recent studies of organic chemical contamination of drinking water supplies report the classification of numerous aggregate populations with respect to various raw water source and treatment characteristics, 'Standard Metropolitan Statistical Area. Epidemiologic Approaches to Chemical Hazard Assessment 165 from which several types of comparisons have been made (225). For example, sex- and site-specific cancer mortality (and in some cases cancer incidence) rates in populations served surface water have been compared to cancer rates in populations served groundwater. Populations served chlorinated water have been compared to populations served unchlorinated water, and so forth. Although the interpretations of these studies differ, the results of all of them to date seem to suggest a slightly elevated risk of certain gastrointestinal and urinary tract cancers in populations consuming drinking water containing the highest levels of trace organic chemical contaminants. Because the schemes previously discussed are generally based on qualitative information only, they may be improved somewhat by using quantitative environmental measurements to assist in the construction of exposure classes. Studies of air and water pollution, again, illustrate the approach. Morris et al. (156) compared mortality between two small Pennsylvania communities, one of which was in close proximity to a coal-fired electric power plant (and therefore presumably had higher levels of air pollution than the other). Unlike the study by Detels et al. (55), which assigned exposure categories (high/low) without, apparently, using quantitative environmental measurements, Morris and co-workers used specific air quality indices (dust fall, sulfation rate, suspended particulates, and sulfur dioxide levels) to verify that a significant difference in air quality existed between the two study populations. Similar air quality measurements have been used to assign communities some air pollution exposure ranking in a number of other studies (38, 77). Studies of the organic chemical content of public drinking water supplies by the U.S. Environmental Protection Agency (USEPA) have provided quantitative measurements on the levels of various waterborne organic chemical contaminants, data that have been used in several epidemiologic studies (225). To summarize, the simple classification schemes previously discussed (1) are based on either qualitative or quantitative environmental data, (2) take the form of ordinal-level measurements, (3) are usually used in aggregate population studies, and (4) are perhaps the crudest of techniques available for measuring human exposures to chemical hazards. It is, however, possible under certain circumstances to refine these simple schemes. For example, in epidemiologic studies of occupational hazards, in which "exposure" data of some kind are usually available for individual study subjects, qualitative information such as the occupation and/or industry of each worker can form the basis of the classification scheme. This "occupational title" (OT) approach, as described by Gamble and Spirtas (52), assigns workers to categories of jobs that are functionally similar (i.e., jobs that involve the same equipment or process) and/or �Epidemiologic Approaches to Chemical Hazard Assessment 166 167 John R. Wilkins III and Nancy A. Reiches that are materially similar (i.e., jobs that involve similar products). Appropriate morbidity and/or mortality measures can then be compared among the various groups of workers, since the groups can be thought of as fairly homogeneous with respect to occupational exposures. For example, in a study of mortality among workers in a rubber tire manufacturing plant, McMichael et al. (146) needed to identify 60 separate OTs in order to characterize the work histories of approximately 1500 men. In their analyses, the 60 OTs were grouped into 16 major work areas and the frequency of employment in each OT group (i.e., the "rate of exposure" to each work area or OT group) was compared among the case and a control series for 7 cancer and 2 noncancer causes of death. The strongest associations were observed between several neoplasms and those work areas most likely to have involved the greatest exposures to organic and inorganic chemicals. Notably, exposure to solvents at several stages of tire building was associated with lymphatic leukemia. Other studies by McMichael and co-workers (144, 147) of the rubber industry, including one that focuses on leukemia and exposure to solvents (147), illustrate the OT approach. The OT approach has also been employed in a study of steel workers (124), and in studies of occupational exposures to asbestos (752) and chloromethyl methyl ether (54). The OT approach has several advantages. First, even in the absence of quantitative environmental sampling data it is possible to characterize systematically chemically complex environments, such as those encountered in rubber tire manufacturing plants. Second, what is usually a very large number of specific jobs can be reduced to a manageable number (at least from a statistical point of view) of fairly uniformly exposed OT groups. Also, the results of such an analysis can be quite useful in either generating or refining hypotheses of cause-and-effect relationships, because it is not necessary to state a priori an interest in some specific health effect, nor is it necessary to have a clear understanding of the induction and latency periods involved. Furthermore, if the results of a study employing the OT approach identify a particularly hazardous work area, intervention strategies may be implemented without knowledge of the specific chemical substances responsible. Additional studies could focus in detail on' the process and/or product related to the apparent high-risk work area in attempts to identify the specific causal agent or agents. Before attention is turned to other ways of measuring external exposure to chemicals in epidemiologic studies it is important to realize that the classification of aggregate populations into ordinal-level exposure categories involves, at least implicitly, the following assumptions: • The degree of exposure among the individuals comprising each class is uniform, or nearly so. • The designated categories make a clear distinction between the groups with respect to exposure levels. In order to gain some insight into the implications of these assumptions, imagine that two well-defined communities (A and B) are selected for study. Imagine further that Community A is in very close proximity to a point source of pollution, say a lead smelter, and that the mere existence of this smelter serves as the basis for labeling Community A the "highexposure" population. Community B, without smelter, is therefore considered the "low-exposure" population. If it is then assumed that the only difference between the two populations is the existence of the smelter, the situation may be conceptualized with the help of a simple sketch (see Fig. 2). Inspection of the figure suggests that exposure to lead in both communities is not precisely uniform, but rather the definitions, high/low, reflect an average amount of population exposure. Certainly, exposure levels in each community vary about a mean, and these means are significantly different from each other. (For simplicity, the distributions have been given a "normal" shape, although the actual distributional form is more likely to be log-normal.) Since, in this case, the exposure classification scheme reflects a sizable difference between the mean population exposure levels—i.e., the exposure definition employed discriminates between the two populations—a comparison of lead-related health measures will be valid. Consider, however, two reasons why this hypothetical situation is not realistic. First, even if the actual (true) underlying distributions of exposure to lead for the study populations are significantly different, the "looseness" or imprecision of simple classification schemes such as with smelter/without smelter, high/low, etc. can, in the general case, create categories that are not homogeneous (with respect to exposure) like those portrayed in Fig. 2. In other words, a certain amount of misclassification will occur, weakening the "purity" of comparing health effect measures between the two groups. Second, it is quite unlikely that Exposure classification Low Kgh (-jt.immunity B j. A Fig. 2. >^ Community A "" s > >. s / * k-rff k Degree of exposure Hypothetical example of lead exposure in two populations. �168 John R. Wilkins III and Nancy A. Reiches we will be able, realistically, to find two or more well-defined populations so vastly different with respect to the degree of exposure to the agent in question, thus compounding the effects of misclassification. This is certainly true for chemical contaminants like lead and organic pesticides that are widely distributed in nature. Further, while the hypothetical communities A and B are rank-ordered in terms of exposure to lead, the "distance" between categories, i.e., the actual degree of difference between A and B with respect to lead exposure, cannot, with such a scheme, be estimated without more information. And, in the particular case of community-wide exposure to lead, exposed/unexposed categories would be unjustified since there are numerous pathways of exposure including food, water, and air. What, then, are the epidemiologic implications of employing simple exposure classification schemes? For one, there are statistical implications. Artifact (resulting from the way in which exposure classification schemes are constructed) and/or the "natural" or environmental characteristics of the agent in question will tend to result in the mixing of individuals with different exposure experiences, thus creating heterogeneous rather than homogeneous exposure categories. The greater the degree of "mixing," i.e., the more "impure" the comparison, the more alike the comparison groups will be in terms of exposure, which in turn means they will be more alike with respect to any health effect truly related to the particular exposure. Consequently, the difference between the comparison groups (with respect to the measure of effect) will diminish, compromising the sensitivity of the study. Health effects, particularly modest ones, may therefore go undetected, possibly giving the illusion of "safety." b. Variations on Haber's Law (Unweighted Models of Cumulative Exposure). A common way to model human exposures to environmental chemical involves the application of Haber's law, an elementary concept in toxicology. This approach can be taken when quantitative environmental data are available and when it is possible to relate such data to individual study subjects, as is often the case in epidemiologic studies of workplace chemical hazards. Mathematically, Haber's law states that the magnitude of toxic effect £ is a function of the product of the intensity of exposure (in concentration units Q and the duration of exposure (in time units /), so that £ = C x t (75). With this concept it is possible to compute, for each person in a study population, an index of total cumulative exposure (TCE) by simply summing the product C x t for each period of exposure to the substance in question (if exposure levels change over time) over the entire study period. The (artificial) data in Table I illustrate the calculations for two Epidemiologic Approaches to Chemical Hazard Assessment 169 TABLE I Calculation of Individual Total Cumulative Exposures Worker i 1 Job./ Intensity of exposure, C (arbitrary units) Duration of exposure, / (arbitrary units) 1 2 3 1 2 3 1 3 9 Total cumulative exposure for worker 1:2 = 2 5 15 Total cumulative exposure for worker 2: C x i 1 6 27 34 4 5 90 99 (of n) hypothetical workers, each having worked for varying amounts of time in three different jobs entailing differing degrees of exposure to a single chemical substance. Since C and t are given equal weight in the computations, this particular method is often referred to as a simple or unweighted model of cumulative exposure. Once the total cumulative exposures are computed for each study subject, categories of TCE are then created, and appropriate morbidity and/or mortality measures compared across the classes. For example, in a study of coke oven workers (142) an index of total cumulative exposure to coal tar pitch volatiles (CTPV) was computed for each worker as follows: TCEworker,-= 2 (mean level of exposure, job j) all jobs x (duration of exposure, job j) Mortality rates for all causes of death combined, for all cancers combined, and for lung cancer were then compared across four (increasing) categories of total cumulative exposure, for white and for nonwhite workers. Notably, mortality from all cancers combined and from lung cancer increased sharply with increasing total cumulative exposure to CTPV among the nonwhite coke oven workers, results suggesting a dose-response relation. �170 Epidemiologic Approaches to Chemical Hazard Assessment John R. Wilkins HI and Nancy A. Reiches The applicability and validity of the simple, unweighted model of cumulative exposure rest on several assumptions. First, it is assumed that quantitative environmental data are available for all relevant time periods and that such data are accurate. Unfortunately, the quantitative characterization of local environments (ambient or occupational) over long periods of time is often not possible. In order to investigate properly the cause or causes of chemically related illness, measures of which (like death rates) are usually contemporary, information must be obtained on exposures occurring prior to the development of the disease. For diseases with substantial induction and/or latent periods, exposure levels dating back years and perhaps decades are required. Even when data on historical conditions are available, the accuracy and representativeness of such data can be questionable, reducing individual exposure histories to crude "guesstimates." Second, when satisfactory environmental monitoring data are available, it is assumed that it will be possible to select the most appropriate way to summarize exposure levels. Since, by nature, the concentration of most toxics in air, water, soil, etc. will fluctuate over time, so will human exposures. Depending on the substance in question, levels of the agent in environmental media may vary by the hour, by the day, by the week, by the month, and by the year, which raises several questions. Will simple arithmetic means appropriately summarize exposure levels? Would time-weighted averages be better? Should sharp peaks over the short term be given more, less, or equal weight, compared to steady, consistent, long-term trends? Although the answers are not usually clear, they are important questions, questions that relate to yet another assumption of the model: since simple cumulative models of exposure give equal weight to C and t, the rate of exposure can be ignored. The essence of this assumption is that the risk of disease would be the same for a given TCE achieved as a result of high exposure over the short term or as a result of low exposure over the long term. Exposure-time units, which are analogous to the familiar and widely used concept of person-time units (201), may be accumulated in virtually an infinite variety of ways: 100 exposure-time units = 1 exposure unit x 100 time units = 100 exposure units x 1 time unit. The problem here arises because, at least for certain types of chemical exposures, the risk of disease for a given TCE is not independent of the mode of exposure. For example, in a study of asbestos workers (66, 68) the risk of respiratory cancer was, on average, about twice as high for men who had had intermittent (and relatively high) asbestos exposures compared to men who had had steady (and relatively low) asbestos exposures. Since the mean total cumulative exposures in both groups were about the same (230.7 versus 236.0 exposure- 171 time units), this finding suggests that the rate at which exposures are accumulated must be taken into account. Another aspect of this "transient dose states" problem is something that has been referred to as the wasted dose phenomenon (197). Because disease in this particular model is viewed as dependent on the maximum TCE, i.e., dependent on the highest level or category of cumulative exposure achieved, the possibility of a "lower effective dose" is not considered, as Schneiderman et al. (197) point out. Exposures occurring after the biologic onset of disease (disease presumably resulting from the TCE up to that point) continue to be added to a person's cumulative exposure. Since the TCE responsible for disease would be overestimated by including exposures occurring during the latent period, the risk of disease would be underestimated at any TCE above the causative or "effective" TCE. Clearly, the longer the latency, the greater the discrepancy between the effective TCE and the TCE employed in the analysis, and thus the greater the underestimation of risk (as long as exposuretime units are accumulated beyond the time of biologic onset of the disease). Underestimation of risk may also occur when the exposure period and the follow-up period overlap. As Enterline (66) views it, a "dose-response fallacy" can occur because entry into the highest TCE categories can only occur for study subjects who survive long enough, i.e., to the end of the follow-up period. As he suggests, "a high dose and death tend to be incompatible states." Although one possible remedy here, at least for occupational studies, is to limit the investigation to retired persons only, this type of study entails other kinds of problems. On the other hand, risks may be overestimated when the disease of interest has a latent period and when the amount of time elapsed from the onset of exposure is not taken into account. In this case, persons falling into the lowest cumulative exposure classes will tend to be those with the least amount of exposure time, thereby artificially reducing the risk in the lower TCE classes and, accordingly, artificially inflating the risk in the higher TCE classes. As Pasternack and Shore suggest (171), a solution would be to simultaneously assign persons to their rightful category of TCE and to the appropriate category of time since exposure began, thus controlling for differences across TCE classes with respect to length of time exposed. Finally, the simple, unweighted model of cumulative exposure cannot be used to study chemically complex environments, i.e., when exposures to more than one chemical agent occur simultaneously. This is not surprising since, in general, the investigation of health risks resulting from multiple chemical exposures is quite difficult. One reason for this is that the �172 John R. Wilkins III and Nancy A. Reicbes Epidemiologic Approaches to Chemical Hazard Assessment necessary environmental data are rarely available (225). Another reason, as Saracci points out (795), is that it may be impossible to study enough subjects to isolate the effects of exposure to one agent in the presence of one or more other agents, particularly when the statistical analysis involves the cross-classification of the sample into contingency tables. And third, even if a large enough sample could be obtained, choice of the most appropriate statistical approach would be a matter of judgment. In point of fact, studies of interaction may be based on either additive models of disease risk (189, 190) or on multiplicative models (100). See also references 777, 797, 223 for a more thorough discussion of synergy and antagonism. Recent attempts have been made to refine the epidemiologic study of the health effects resulting from multiple chemical exposures (205), an area of research in which the knowledge base is rudimentary at present. areas used as the weights. This method is also discussed by Land and McGregor (727). Other attempts have been made to take latency into account by either partially weighting late exposures or by ignoring them altogether (i.e., giving them zero weight). This is the so-called lagged-exposure model, which assumes that exposures occurring a certain number of years prior to disease or death may be discounted. Mazumdar and Redmond (747) discuss this technique as applied to their study of lung cancer in men exposed to coal tar pitch volatiles. Pasternack and Shore (777) discuss the application of the lagged-exposures approach to actually estimating the average latent period in a set of data. c. Other Variations on Haber's Law (Weighted Models of Cumulative Exposure). A major deficiency of the unweighted model is that it does not take into account the concept of a latent period, i.e., the notion of an "effective" cumulative exposure is not addressed. In seeking to measure an effective cumulative exposure it has been argued that some portion of exposure occurring during the exposure period may legitimately be discounted, i.e., differentially weighted, because the portion of exposure in question presumably plays little or no causal role. A strong case can probably be made that contemporary risks are independent of very recent exposures, particularly for diseases with substantial latent periods. The central issue, of course, is the lack of knowledge about when the biologic onset of disease occurs, which severely limits the estimation of latency. One approach to this problem involves, first, making certain assumptions about the temporal pattern of disease occurrence following a single exposure, i.e., assumptions are made about the latent period. After making assumptions about the shape, the standard deviation, and the central tendency of the distribution of latent periods, the distribution itself is used to derive a series of weights, which in turn are applied to the cumulative exposures in appropriate time periods. For example, in the Lundin et al. (126) study of lung cancer mortality in underground uranium miners, weights were derived by, first, assuming that the time between the first exposure to underground uranium mining and death from lung cancer was log-normally distributed with a standard deviation of 0.1761 and a median latency of 10 years. Data on the miners were used to estimate median latency, while estimates of the shape and the spread of the distribution were based on those observed for leukemia following a single high exposure to atomic radiation (16, 727). The log-normal density function was then integrated over the relevant time intervals and these 173 B. Measurement of Response As described earlier, the morbidity or mortality rate is a useful measure of the risk of disease or death in a population. In practice, however, there are often situations in which a rate, in its usual form, cannot be computed as a measure of response in a population to a given exposure. In other situations the computed rate is not reliable and therefore should not be employed. For example, the problem of unreliable rates is common in the study of cancer. Although cancer is a leading cause of death, the actual probability of an individual dying of cancer in a given year is quite small. If the population under study is not sufficiently large the resulting rate, particularly the age-specific rate, will be quite unstable, and the confidence interval encompassing that rate will be very large. A measure called the standardized mortality ratio (SMR) is commonly used in such situations. The major advantage of the SMR is the introduction of information from a large, stable population. Using this method it is possible to compare the mortality experience of a defined subgroup with the total population. The SMR is computed as a ratio of the observed number of deaths in the study population to the number of deaths that are expected to occur in that group. It is with respect to the denominator of this ratio that the concept of a standard population is required. To compute the SMR it is not necessary to know the number of deaths that occur in each age group of the study population; one only needs to know the number of persons at risk in each age group. An expected number of deaths is generated by multiplying this figure by the age-specific mortality rate of the standard population. Both observed and expected numbers of deaths are then summed over all ages, and the ratio computed. The SMR is interpreted with respect to its deviation from unity. To the extent that it exceeds unity, the risk of death is said to be greater in the study population. Statistical properties of the SMR are known and �174 Epidemiologic Approaches to Chemical Hazard Assessment 175 John R. Wilkins III and Nancy A. Reiches therefore significance testing is possible. In this regard, tables containing critical values of SMRs are published2 (6). The choice of a standard population is not a trivial matter, nor is the most appropriate selection always obvious. For example, a comparison of death rates for different socioeconomic groups in a population might use as a standard the highest socioeconomic group (207). Resulting SMRs for lower groups would then reflect the assumption that excess deaths occur because living conditions, medical care, etc., are inadequate. Interpretively, this raises different issues than might surface if the entire population had served as the standard. In selecting a standard population, attention should be given to the purpose of the comparison and to its potential limitations. SMRs are frequently utilized in the study of defined occupational groups. One such investigation evaluated the mortality experience of iron foundry workers (53). Death rates in this cohort were compared to rates for the general U.S. male population. Potentially confounding variables, such as length of employment in the industry and race, were considered. Although silicosis has historically been a health problem in the foundry industry, unusually high mortality from chronic respiratory disease was not observed in this analysis. However, it cannot be discerned from the data whether this was a result of exposure to low levels of silica-containing dusts, or of insufficient numbers of workers with long exposure histories, or of too short a follow-up interval to allow for the clinical expression of disease and subsequent death of cohort members. Overall, this investigation revealed lower total mortality in the worker population. This finding is not unusual in occupational studies, a phenomenon discussed later (8, 155). A similarly designed study of workers in a chemical company was intended to determine whether socioeconomic status or job classification was related to overall or cause-specific mortality (769). This type of study is important because it recognizes the variability of individual characteristics within a broadly defined worker cohort. SMRs were computed using the U.S. white male population as the standard. Overall mortality was lower than expected, but certain malignancies (such as urinary organ neoplasms) yielded high SMRs. Stratification of the data by socioeconomic level showed statistical differences: low SMRs in the high socioeconomic group. Differences with respect to job category were also detected. For example, plant mechanics and machinists had more malignancies than expected, while inorganic chemical production workers 2 Although this discussion pertains to mortality, morbidity data can be treated in the same way. showed a decreased rate of cancer. Additionally, the analysis addressed the question of age at entry to the study and age at death. This approach is significant because it presents the investigator with a new set of paths to follow to explain notable trends in the health of industrial populations. Careful stratification of the cohort is particularly important, since, as the study of the chemical workers demonstrates, there is measurable heterogeneity within the cohort. For example, the low rate of cancer mortality in the high socioeconomic group might reflect a lower prevalence of smoking. Once the specific mortality pattern of a well-defined group is understood, there is opportunity for careful testing of such hypotheses. Other investigations have also used the approach of employing data on job classifications and have reported increased rates of malignant disease for specific categories (124, 125, 168). This has been noted for mechanics and machinists in more than one industry. In addition to SMRs, a measure called the proportional mortality ratio (PMR) has been used in a large number of analyses, particularly in studies of occupational groups. The PMR differs from the SMR in that the demographic composition of the population at risk is not known. Rather, the PMR represents the proportion of total deaths attributable to a specific cause in a study population. Consequently, the PMR is not a rate. It is simply a measure of the relative importance of a given cause of death, not a measure of the risk of death (154). Despite this inherent limitation, PMRs do have a role in epidemiologic analyses. First, the data necessary to compute a PMR are relatively easy to obtain—they are essentially only the data that would normally constitute the numerator of a mortality rate. Hence, PMR studies are sometimes referred to as "numerator studies." Second, although the absolute risk of death cannot be determined, knowledge of the relative importance of a cause of death can lead to testable hypotheses about potential etiologic factors. That is, if a cause of death is proportionately greater in one group than in another, exposures unique to the former might explain the observed differential. The validity of a PMR study depends on the extent to which certain assumptions are met by the data. The difficulty is that the assumptions cannot be tested empirically, since they require data that are not available, namely population data. The basic assumption is that the relationship between the PMRs of two groups being compared is equivalent to the relationship between the actual mortality rates in the populations. If this latter information were known, however, there would be no need to compute PMRs; rather the rates or SMRs could be compared directly. There is a danger of erroneous interpretation of PMRs if this assumption is not tenable or if the PMR is inappropriately interpreted as a measure of risk. For example, consider a hypothetical case of two study populations, �176 John R. Wilkins UI and Nancy A. Reiches in each of which 1000 total deaths are observed. Also assume that 200 deaths in each group are due to cancer. The PMR for each group would thus be 0.20. Clearly, cancer assumes the same relative importance in each group—20% of all deaths. But if the population of the second group is larger, the actual death rate from cancer would be smaller than in the first group. That is, the risk or probability of dying of cancer can vary even if its proportionate contribution to total mortality is the same. Consequently, in interpreting PMRs one must not be tempted by the false impression that the comparative risk of death is being analyzed. There are many examples of PMR studies in the literature. One such analysis investigated mortality patterns among employees exposed to poly vinyl chloride (PVC) (40). Since the population at risk could not be determined, the proportional mortality in various subgroups of the worker population was compared to similarly defined PMRs for the U.S. population. A comparison of these figures is an indication of whether or not relative excess mortality has occurred in the study population. In this particular investigation, there appeared to be an excess number of cancer deaths among both white males and white females. For the reasons already given, however, this finding must be interpreted with caution. The suggestion, however, of excess cancer mortality does provide a lead for more definitive investigation, thereby demonstrating the value of PMR analyses. The importance of such analyses is similarly demonstrated in a study of mortality among workers in a newspaper printing factory (92). This example is noteworthy because it demonstrates that PMR analysis can be an efficient method for very preliminary investigation of a new hypothesis. This study was undertaken following anecdotal reports of a high incidence of bladder cancer among the printing workers. In the particular group studied there was no evidence that bladder cancer assumed unusual importance as a cause of death, although the PMR for all neoplasms combined was very high. However, this appeared to be the result of a large number of deaths from lung cancer, implicating smoking rather than an industrial hazard. The comparability (and differences) of the PMR and SMR methods is demonstrated in an analysis of workers exposed to low levels of methylene chloride for up to 30 years (75). Specifically, the investigators wanted to determine whether this cohort exhibited high rates of mortality from ischemic heart disease, since exposure to chlorohydrocarbons may result in increased cardiac sensitivity (183). Since population data were not available for this group prior to 1964, a PMR approach was adopted. The post-1964 cohort was analyzed by an SMR approach. Proportional mortality ratios did not reveal any unusual mortality trends for any of Epidemiologic Approaches to Chemical Hazard Assessment 177 the 17 major diagnostic categories that were analyzed. Further breakdown of the data for specific malignancies also failed to show any statistically significant differences. In the second part of the analysis, two different standard populations were selected. The first was the group of all other males working in the same plant; the second was a general population standard. The results obtained exemplify the point noted earlier regarding the effect of a particular control population on study findings. With respect to the industrial standard, the methylene chloride-exposed group did not have significantly different SMRs for any major cause of death studied. However, when compared to the general population, significantly fewer deaths than expected were observed for malignant neoplasms and circulatory diseases. Specifically, ischemic heart disease mortality was reduced. While it has been emphasized that PMRs are not direct risk-assessment measures, their usefulness for preliminary screening of data is generally accepted. The methylene chloride study, however, demonstrates a case in which potentially erroneous conclusions might have been drawn if only the PMR analysis were available. In this study the PMRs and SMRs are not directly comparable, since the data for each were derived from different time periods. However, one might argue that the differences are small and that the conflicting results reflect the method of analysis. The findings of this investigation do not negate the relative value of PMR analysis, nor do they wholly validate the SMR approach. Rather, they point out the need for cautious interpretation. There are in the literature several rigorous comparisons of the two approaches described here (52, 118, 181). Although these issues will not be discussed in detail, it is important to recognize, at least in concept, some of the primary constraints. Some, such as the choice of the standard population and the failure of PMRs to measure risk, have been previously alluded to. Other problems of the SMR have also been identified (80, 81, 145). For example, the SMR does not reflect the effect of a hypothesized hazard on life expectancy; it counts only the number of deaths, not the ages at which they occur (87). It has been demonstrated that populations with different life expectancies can yield the same SMR. Additionally, the SMR is dependent on the age distribution of the study population. If younger workers have a lower mortality rate than the standard population, the SMR will not correctly estimate the probability of death, since this probability is not precisely (mathematically) equivalent to the mortality rate (74). These two figures are related, however, and consequently one can compute the degree of age dependence in the SMR (39). Finally, the SMR is not independent of the length of follow-up of the study cohort. That is, if calculated periodically during follow-up, the SMR is not expected �John R. Wilkins III and Nancy A. Reiches Epidemiologic Approaches to Chemical Hazard Assessment to remain constant. If the risk of death in the study group is high, SMRs might exceed 1.00 early in the study, but decline as follow-up continues. A final point about the measurements of outcome that have been discussed in this subsection relates to a question of sample selection bias known as the "healthy worker effect." A variety of data indicate that the fact that persons are healthy enough to be employed intrinsically predicts that their mortality experience will be more favorable than that of the general population. This effect was first identified nearly 100 years ago and has been widely recognized in contemporary epidemiology (86, 166). Furthermore, it has been demonstrated that the magnitude of the bias is related to the age distribution of the industrially employed cohort and to the specific causes of death being considered (65). In addition to the fitness of workers at the time of employment, the issue is further complicated by the fact that the composition of the cohort is influenced by the dynamics of individuals leaving the industry for health-related reasons. The empirical effects of these questions on SMRs has been reported. One of the more comprehensive analyses involved a study of all PVC workers in Great Britain (75). The findings supported an association between exposure to the vinyl chloride monomer and angiosarcoma of the liver; furthermore, it was demonstrated that the observed rates of mortality were indeed related to the selection of workers into the industry, their continued employment, and the length of time the cohort was followed. The cause-specific nature of these biases has been shown in a study of workers in five chemical plants, using an approach designed to minimize selection effects on the resultant SMRs (202). sometimes provide sufficient information for designing intervention strategies, thus interrupting the causal chain of events even if they are not fully known. Even with this inherent strength, epidemiologic assessment of health risks resulting from chemical exposures can be enhanced by incorporating knowledge derived from theoretical studies of the pathogenesis of cancer. In this' section we do not provide a comprehensive discussion of the molecular theories of carcinogenesis; rather, we highlight a few general principles that bear on the design of epidemiologic investigations and the interpretation of their results. Of particular importance in this context are the concepts of initiation and promotion. These terms were coined in the 1940s to define operationally the extended period between the initial exposure to a carcinogen and the expression of a malignancy (755,188). Early empirical demonstrations of this process involved the direct application of a confirmed chemical carcinogen (usually a polycyclic hydrocarbon) to the skin of a mouse— the initiation phase. Tumor promotion was accomplished by the application of another chemical agent, which was by itself incapable of inducing neoplasia (9). Although this general procedure has been refined in recent years, it still provides one of the fundamental models for studying chemically induced cancer. Subsequent experiments have confirmed that certain compounds, such as benzo[o]pyrene and methylcholanthrene, possess both initiation and promotion activity, and therefore are complete carcinogens (26, 27). The various stages of carcinogenesis have been demonstrated in organs other than the skin. For example, a breast cancer model in rats and mice has indicated that application of a carcinogen without the appropriate hormones does not result in a malignant tumor (79). Additionally, both initiating and promoting agents have been identified for tumors of the dog and rat bladder, the mouse lung and forestomach, and the rat colon, bone marrow, liver, and thyroid (777). In each case the initiator is an agent whose metabolites can react with DNA. The corresponding promoters range from natural products to normal circulating hormones. There are several general characteristics of the multistage carcinogenic process that have implications for the ultimate control of malignant disease in human populations. One important finding in this regard is that the process of initiation is not reversible, while the process of promotion is. This bears directly on the potential for prevention of neoplastic disease. That is, if the exposure is discontinued before cells in the target tissue develop the ability to multiply in the absence of the promoter, then formation of a tumor may be avoided. Reduction of risk of lung cancer following cessation of cigarette smoking may be an example of this type of intervention (174). The declining risk suggests that promoters are the cancer-causing elements in cigarette smoke. 178 V. RELATING MEASURES OF DOSE TO MEASURES OF RESPONSE There are several important aspects of the process by which the functional relationship between measures of dose and measures of response is determined. For example, in the case of cancer this process is influenced by the investigator's assumptions regarding the underlying biologic mechanisms of carcinogenesis. Although cancer has been recognized as a distinct disease for thousands of years, only recently has there developed some understanding of the mechanisms responsible for the transformation of a normal cell into a malignant one. Clearly, an elucidation of the biologic mechanisms of carcinogenesis will substantially increase the potential for prevention and control of neoplastic disease. The lack of this type of evidence, however, does not thoroughly preclude the ability to intervene; associations discovered in epidemiologic investigations can 179 �180 John R. Wilkins III and Nancy A. Reiches The concepts of initiation and promotion are inextricably bound to the phenomenon of latency. The importance of accounting for the latent period is well understood in epidemiologic research. Investigators attempt to introduce appropriate temporal relationships between measures of dose and measures of response. The latent period is a general feature of the natural history of neoplasia, whether the relevant exposure is chemical, radiologic, or viral. In the practice of epidemiologic research, however, the concept of a latent period and the temporal relationships between initiation and promotion phases pose various difficulties. In general, the duration of the latency period is unknown. Furthermore, the relationship between dose level and duration of latency is uncertain. Experiments with laboratory animals have indicated that an increased dose does shorten latency, and attempts have been made to quantify this relationship (7). However, the same mathematical model does not appear to hold for human populations (121). For example, in a study of bladder cancer among persons occupationally exposed to dyestuff intermediates, no relation between dose and latency could be detected (36). These findings have led to speculation that the duration of the latent period is affected by variables other than the dose of the initiator. Some of these factors are probably endogenous characteristics of the host, such as levels of pituitary hormones and the genetic makeup of the host (12). Other modifiers are thought to be exogenous and may include dietary constituents (153, 163). To the extent that these factors are unknown, the assessment of risk in human populations becomes more complicated, since the variation in dose rate over time, the reversibility of initiation, and the distinction between initiation and promotion must be accounted for if causal inferences are to result. Although a number of diverse quantitative approaches to modeling carcinogenesis in human populations have been proposed, none is entirely consistent with available empirical evidence (48, 227). Many of these models have incorporated information regarding the age distribution of cancer cases and have measured the effective duration of exposure before onset of disease over a wide range of ages (5, 56). By this method the comparative risk of exposure to the same agent at different ages could be analyzed in relation to dose, duration of latency, and the effect of altering various promoters. VI. CONCLUSION The purpose of this article has been to discuss some of the concepts fundamental to the epidemiologic evaluation of potential health risks stemming from chemical contamination of the human environment. These Epidemiologic Approaches to Chemical Hazard Assessment 181 methods assume a central role in any comprehensive attempt to understand the effects of chemical exposures on human health. Combined with evidence derived from the fields of chemistry and toxicology, the quantification of human risk should ultimately result in substantially improved methods for intervening in the process of disease causation. The epidemiologic approach is characterized by its systematic examination of patterns of exposure and response in human populations. Since the occurrence of disease is not a random phenomenon, epidemiologic investigation is uniquely suited to the generation and testing of etiologic ./hypotheses. The development of clues to explain chemically related illness I generally begins with descriptive methods, whose major purpose is to i detect variations in disease occurrence with respect to time and/or place. <f Observed secular trends may reflect alterations in exposure to environmental hazards. Notable geographic differences in disease occurrence may result from the presence of a risk factor in some populations and its absence in others. Once observed temporal or geographic patterns are established as real (as opposed'to artifactual), epidemiologic investigation may proceed to a variety of aggregate population studies, usually entailing correlation or regression techniques. In this phase of the process, attention focuses on the identification of demographic, socioeconomic, and environmental factors that may have etiologic implications. Although the methodologic problems associated with ecologic analyses are well recognized, the method has substantial utility for generating hypotheses that may subsequently be tested by more rigorous methods. If descriptive epidemiologic studies suggest a potentially adverse effect from a chemical exposure, investigations can then be designed to test formally the possible association between the agent and the disease. The analytic methods employed in this phase of epidemiologic inquiry incorporate data for individual study subjects, as opposed to aggregate or summary data for a population. Although the two primary methodologic approaches, case-control and cohort studies, differ with regard to the assemblage of subjects, they share a number of characteristics. In both cases, the desired endpoint is some quantitative (statistical) measure of risk associated with the exposure in question. Concern for proper classificatinn_nf_stiidy suhjectsj\jth_respect to exposunTand disease, selection of appropriate^ comparison groups, the requirement of reliable and valid ""exposure data", jmd the needto control_gonfounding factors are common" elements in both approaches.^ATtEough there are a variety of advantages and disadvantages intrinsic to both methods, the choice for a particular study depends on the hypothesis to be tested, the availability of necessary data, the rarity of the disease under consideration, and the prevalence and intensity of the exposure factor. �182 John R. Wilkins III and Nancy A. Reiches To a great extent, the applicability of results from case-control or cohort studies is dependent on the method by which exposure (and by implication, dose) is measured. The exposure variable may range from a simple qualitative classification to a more complex quantitative estimate of total cumulative exposure. What is important to recognize is that the measurement of exposure must be consistent with an underlying biologic theory of disease causation. For example, studies of malignant neoplasms must account for periods of latency and possible differential effects between initiating and promoting agents. Finally, methodologic attention must focus on appropriate measures of response to a chemical contaminant. Quantitative measures, such as standardized and proportional mortality ratios, need to be carefully constructed and statistically analyzed. Each aspect of the epidemiologic approach we have described is itself an area that continues to be subjected to intense critical scrutiny. For example, there is a rich literature regarding the statistical properties of the odds ratio, the choice of controls for case-control studies, the appropriate length of follow-up for cohort analyses, etc. That controversy exists in each area does not invalidate the overall approach; rather, it enhances the investigator's ability to reach critical decisions about all phases of study design, execution, data analysis, and interpretation. Perhaps more than for anything else, the epidemiologic method can be recommended for its vigilance regarding the possibility of alternative explanations to account for any observed finding. In the ideal case, the interpretation of epidemiologies data guards against the chance that a hazardous exposure is judged to '- 'safe." Epidemiologic analyses thus contribute to the control of disease by quantifying the probability that a chemical exposure may pose risk to human health, and by specifying the co-occurring conditions under which such risk might exist. If a hazard is confirmed, appropriate intervention strategies may be devised to interrupt the causal chain, thereby reducing morbidity and mortality. In this context the epidemiologic approach is fundamental to the assessment of chemical exposures, their effects on human health, and the benefits to society that might result from reduced environmental contamination. References 1. Albert, R. E., and Altshuler, B. (1973). In "Radionuclide Carcinogenesis" (C. L. Sanders, R. H. Busch, and J. E. Ballow, eds.), pp. 223-253. U.S. At. Energy Comm., Oak Ridge, Tennessee. 2. Alderson, M. R., and Meade, T. W. (1967). Br. J. Prev. Soc. Med. 21, 22-29. 3. Alter, M., Loewenson, R., and Harshe, M. (1973). J. Chronic Dis. 26, 755-767. Epidemiologic Approaches to Chemical Hazard Assessment 183 4. Armenian, H. K., and Lilienfeld, A. M. (1974). Am. J. Epidemiol. 99, 92-100. 5. Armitage, P., and Doll, R. (1961). In "Proceedings of Fourth Berkeley Symposium on Mathematical Statistics and Probability," pp. 19-38. Univ. of California Press, Berkeley. 6. Bailer, J. C., and Ederer, F. (1964). Biometrics 201, 639-643. 7. Barna-LIoyd, G. (1978). Science 202, 469. 8. Beaumont, J. J., and Weiss, N. S. (1980). Am. J. Epidemiol. 112, 775-786. 9. Berenblum, I., and Shubrik, P. (1947). Br. J. Cancer 1, 379-380. 10. Berkson, J. (1946). Biometrics 2, 47-53. 11. Berlin, A., Wolff, A. H., and Hasegawa, Y., eds. (1977). "International Workshop on the Use of Biological Specimens for the Assessment of Human Exposure to Environmental Pollutants." Martinus Nijhoff Publishers, The Hague, Netherlands. 12. Bielschowsky, F. (1955). Br. J. Cancer 9, 80-83. 13. Billewicz, W. Z. (1964). Br. J. Prev. Soc. Med. 18, 167-173. 14. Billewicz, W. Z. (1965). Biometrics 21, 623-644. 15. Birch, M. W. (1964). J. K. Stat. Soc., Ser. B 26, 313-324. 16. Bizzozero, O. H., Jr., Johnson, K. G., and Ciocco, A. (1966). N. Engl. J. Med. 274, 1059-1102. 17. Blair, J., and Mason, T. J. (1980). Arch. Environ. Health 35, 92-94. 18. Blalock, H. M., Jr. (1964). "Causal Inferences in Nonexperimental Research." Univ. of North Carolina Press, Chapel Hill. 19. Blot, W. J., Brinton, L. A., Fraumeni, J. F., Jr., and Stone, B. J. (1977). Science 198, 51-53. 20. Blot, W. J., and Fraumeni, J. F., Jr. (1975). Lancet 1, 142-144. 21. Blot, W. J., and Fraumeni, J. F., Jr. (1976). Am. J. Epidemiol. 103, 539-550. 22. Blot, W. J., and Fraumeni, J. F., Jr. (1977). J. Chronic Dis. 30, 745-757. 23. Blot, W. J., and Fraumeni, J. F., Jr. (1978). JNCI, J. Nat. Cancer Inst. 61, 10171023. 24. Blot, W. J., Fraumeni, J. F., Jr., Mason, T. J., and Hoover, R. N. (1979). Environ. Health Perspect. 32, 53-58. 25. Blot, W. J., Mason, T. J., Hoover, R., and Fraumeni, J. F., Jr. (1977). In "Origins of Human Cancer. Book A> Incidence of Cancer in Humans" (H. H. Hiatt, J. D. Watson, and J. A. Winsten, eds.), pp. 21-32. Cold Spring Harbor Laboratory, Cold Spring Harbor, New York. 26. Boutwell, R. K. (1964). Prog. Exp. Tumor Res. 4, 207-250. 27. Boutwell, R. K. (1974). CRC Crit. Rev. Toxicol. 2, 419-423. 28. Boyd, N. F., Patu, J. L., Ginsburg, A. D., and Myers, R. E. (1979). J. Chronic Dis. 32, 327-332. 29. Boyland, E. (1967). Proc. R. Soc. Med. 60, 93-99. 30. Boyland, E. (1967). Prog. Exp. Tumor Res. 11, 222-234. 31. Breslow, N. E., Day, N. E., Halvorsen, K. T., Prentice, R. L., and Sabai, C. (1978). Am. J. Epidemiol. 108, 299-307. 32. Brinton, L., Stone, B. J., Blot, W. J., and Fraumeni, J. F., Jr. (1976). Lancet 2, 628. 33. Bross, I. D. J. (1954). Biometrics 10, 478-486. 34. Carlson, H. A., and Bell, E. T. (1929). J. Cancer Res. 13, 126-135. 35. Carnow, B. W., and Meier, P. (1973). Arch. Environ. Health 27, 207-218. 36. Case, R. A. M., Hosker, M. E., McDonald, D. B.. and Pearson, J. T. (1954). Br. J. Ind. Med. 11, 75-104. 37. Case, R. A. M., and Lea, A. J. (1955). Br. J. Prev. Soc. Med. 9, 62-72. �185 John R. Wilkins HI and Nancy A. Refches Epidemiologic Approaches to Chemical Hazard Assessment 38. Chapman, R. S., Shy, C. M., and Finklea, J. F. (1973). Arch. Environ. Health 27, 138-142. 39. Chiang, C. L. (1961). Biometrics 17, 57-78. 40. Chiazze, L., Nichols, W. E., and Wong, O. (1977). JOM, J. Occup. Med. 19, 623628. 41. Cole, P. (1979). J. Chronic Dis. 32, 15-34. 42. Colucci, A. V., Hammer, D. I., Williams, M. E., Hinners, T. A., Pinkerton, C., Kent, J. L., and Love, G. J. (1973). Arch. Environ. Health, 27, 151-154. 43. Comstock, G. W. (1979). Am. J. Epidemiol. 110, 375-400. 44. Conkle, J. P., Camp, B. J., and Welch, B. E. (1975). Arch. Environ. Health 30, 290295. 45. Copeland, K. T., Checkoway, H., McMichael, A. J., and Holbrook, R. H. (1977). Am. J. Epidemiol. 105, 488-495. 46. Cornfield, J. (1951). JNCl, J. Nat. Cancer Inst. 11, 1269-1275. 47. Cornfield, J., and Haenszel, W. (1960). J. Chronic Dis. 11, 523-534. 48. Crump, K. S. (1975). Biometrics 35, 157-167. 49. Cutler, S. J., and Ederer, F., eds. (1961). "End Results and Mortality Trends in Cancer: Part I. End Results in Cancer," Nat. Cancer Inst. Monogr. 6. U.S. Gov. Printing Office, Washington, D.C. 50. Cutler, S. J., Scotto, J., Devesa, S. S., and Connelly, R. R. (1974). JNCI, J. Nat. Cancer Inst. S3, 1565-1575. 51. Davis, D. L., and Magee, B. H. (1979). Science 206, 1356-1357. 52. Decoufle, P., Thomas, T. L., and Pickle, L. W. (1980). Am. J. Epidemiol. Ill, 263269. 53. Decoufle, P., and Wood, D. J. (1979). Am. J. Epidemiol. 109, 667-675. 54. DeFonso, L. R., and Kelton, S. C. (1976). Arch. Environ. Health 31, 125-130. 55. Detels, R., Rokaw, S. N., and Coulson, A. H. (1979). Am. J. Epidemiol. 109, 3358. 56. Doll, R. (1971). J. R. Stat. Soc. Part 2, 134, 133-166. 57. Doll, R. (1977). Nature (London) 265, 589-596. 58. Doll, R., and Hill, A. B. (1952). Br. Med. J. 2, 1271-1286. 59. Doll, R., Muir, C., and Waterhouse, J., eds. (1970). "Cancer Incidence in Five Continents." Springer-Verlag, Berlin and New York. 60. Dorn, H. F., and Moriyama, I. M. (1964). Am. J. Public Health 54, 400-406. 61. Duncan, O. D., and Davis, B. (1953). Am. Sociol. Rev. 18, 665-666. 62. 'Edwards, J. H. (1961). Ann. Hum. Genet. 25, 89-93. 63. Ehrlich, D., Li-Sik, M., and Modan, B. (1975). J. Chronic Dis. 28, 359-364. 64. Engel, L. W., Strauchen, J. A., Chiazze, L.. and Heid. M. (1980). Am. J. Epidemiol 111, 99-112. 65. Enterline, P. E. (1975). JOM, J. Occup. Med. 17, 127-128. 66. Enterline, P. E. (1976). JOM, J. Occup. Med. 18, 150-156. 67. Enterline, P. E., and Capt, K. G. (1959). Am. J. Public Health 49, 205-212. 68. Enterline, P. E., Decoufle, P., and Henderson, V. (1972). JOM, J. Occup. Med. 14, 897-903. 69. Epstein, S. S. (1974). Cancer Res. 34, 2425-2435. 70. Fabsitz, R., and Feinleib, M. (1980). Am. J. Epidemiol. Ill, 315-328. 71. Feigl, P., Breslow, N. E., Laszlo, J., Priore, R. L., and Taylor, W. F. (1981). JNCI. J. Nat. Cancer Inst. 67, 1017-1024. 72. Feinstein, A. R., Pritchett, J. A., and Schimpff, C. R. (1969). Arch. Intern. Med. 123, 448-461. 73. Filov, V. A., Golubev, A. A., Liublina, E. I., and Tolokontsev, N. A. (1979). "Quantitative Toxicology." Wiley, New York. 74. Fleiss, J. R. (1973). "Statistical Methods for Rates and Proportions." Wiley, New York. 75. Fox, A. J., and Collier, P. F. (1976). Br. J. Prev. Soc. Med. 30, 225-230. 76. Fraumeni, J. F., Jr., and Blot, W. J. (1977). J. Chronic Dis. 30, 759-767. 77. French, J. G., Lowrimore, G., and Nelson, W. C. (1973). Arch. Environ. Health 27, 129-133. 78. Friedlander, B. R., Hearne, T., and Hall, S. (1978). JOM, J. Occup. Med. 20, 657666. 79. Furth, J. (1975). In "Cancer: A Comprehensive Treatise" (F. F. Becker, ed.), pp. 75-120. Plenum, New York. 80. Gaffey, W. R. (1975). JOM, J. Occup. Med. 17, 128. 81. Gaffey, W. R. (1976). JOM, J. Occup. Med. 18, 157-160. 82. Gamble, J., and Spirtas, R. (1976). JOM, J. Occup. Med. 18, 399-404. 83. Gart, J. J. (1971). Rev. Int. Stat. Inst. 39, 148-169. 84. Gittelsohn, A., and Senning, J. (1979). Am. J. Public Health 69, 680-689. 85. Goldberg, J. D. (1975). JASA, J. Am. Stat. Assoc. 70, 561-567. 86. Goldsmith, J. R. (1975). JOM, J. Occup. Med. 17, 126-127. 87. Goldsmith, J. R. (1977). JOM, J. Occup. Med. 19, 533-539. 88. Goldstein, I. F. (1981). Environ. Res. 25, 236-240. 89. Goodman, L. A. (1953). Am. Sociol. Rev. 18, 663-664. 90. Goodman, L. A. (1969). J. R. Stat. Soc., Ser. B 31, 485-498. 91. Goodman, R. A., Maton, K. A., Nolan, T. F., Bregman, D. J. and Hinman, A. R. (1982). JAMA, J. Am. Med. Assoc. 247, 793-7%. 92. Greenberg, M. (1972). Br. J. Ind. Med. 29, 15-20. 93. Guralnick, L. (1966). J. Chronic Dis. 19, 979-990. 94. Guy, W. A. (1856). J. R. Stat. Soc. 19, 12-27. 95. Haenszel, W. (1975). In "Persons at High Risk of Cancer: An Approach to Cancer Etiology and Control" (J. F. Fraumeni, Jr., ed.), pp. 361-371. Academic Press, New York. 96. Haldane, J. B. S. (1955). Ann. Hum. Genet. 20, 309-311. 97. Hammer, D. I., Finklea, J. F., Hendricks, R. H., Shy, C. M., and Horton, R. J. M. (1971). Am. J. Epidemiol. 93, 84-92. 98. Higginson, J. (1979). Science 205, 1363. 99. Higginson, J., and Muir, C. S. (1979). JNCI, J. Nat. Cancer Inst. 63, 1291-1298. 100. Hogan, M. D., Kupper, L. L., Most, B. M., and Baseman, J. K. (1978). Am. J. Epidemiol. 108, 60-67. 101. Hoover, R. (1979). Ann. N.Y. Acad. Sci. 329, 50-60. 102. Hoover, R. and Fraumeni, J. F., Jr. (1975). Environ. Res. 9, 196-207. 103. Hoover, R., Mason, T. J., McKay, F. W., and Fraumeni, J. F., Jr. (1975). Science 189, 1005-1007. 104. Hoover, R., Mason, T. J., McKay, F. W., and Fraumeni, J. F., Jr. (1975). In "Persons at High Risk of Cancer: An Approach to Cancer Etiology and Control" (J. F. Fraumeni, Jr., ed.), pp. 343-360. Academic Press, New York. 105. Hoover, R., McKay, F. W., and Fraumeni, J. F., Jr. (1976). JNCI, J. Nat. Cancer Inst. 57, 757-768. 106. Howe, G. R., Burch, J. D., Miller, A. B., Cook, A. M., Esteve, J., Morrison, B., Gordon, P., Chambers, L. W., Fodor, G., and Winsor, G. M. (1980). JNCI, J. Nat. Cancer Inst. 64, 701-713. 184 �186 Epidemiologic Approaches to Chemical Hazard Assessment John R. Wilkins III and Nancy A. Reiches 107. Jenkins, D. (1977). In "The Use of Biological Specimens for the Assessment of Human Exposure to Environmental Pollutants" (A. Berlin, A. H. Wolff, and Y. Hasegawa, eds.), pp. 215-220. Martinus Nijhoff Publishers, The Hague, Netherlands. 108. Jick, H., and Miettinen, O. S. (1973). N. Engl. J. Med. 289, 63-67. 109. Jick, H., and Vessey, M. P. (1978). Am. J. Epidemiol. 107, 1-7. 110. Keys, A., and Kihlberg, J. K. (1963). Am. J. Public Health S3, 1656-1665. 111. Kmet, J. (1970). J. Chronic Dis. 23, 305-324. 112. Knobloch, H., and Pasamanick, B. (1958). Am. J. Public Health 48, 1201-1208. 113. Knox, E. G. (1971). Br. Med. Bull. 27, 43. 114. Kolonel, L. N., Hirohata, T., and Nomina, A. M. Y. (1977). Am. J. Epidemiol. 106, 476-484. 115. Koran, L. M. (1975). N. Engl. J. Med. 293, 642-646. 116. Kotin, P. (1970). J. Chronic Dis. 23, 293-304. 117. Kupper, L. L., and Hogan, M. D. (1978). Am. J. Epidemiol. 108, 447-453. 118. Kupper, L. L., McMichael, A. J., Symons, M. J., and Most, B. M. (1978). J. Chronic Dis. 31, 15-22. 119. Kurland, L. T., and Molgaard, C. A. (1981). Sci. Am. 245, 54-63. 120. Kuzma, R. J., Kuzma, C. M., and Buncher, C. R. (1977). Am. J. Public Health 67, 725-729. 121. Land, C. E., and McGregor, D. H. (1977). In "Prevention and Detection of Cancer" (H. E. Nieburgs, ed.), Vol. 1, Part 1, pp. 831-843. Dekker, New York. 122. Lewis, M. S. (1980). J. Chronic Dis. 33, 703-712. 123. Lilienfeld, A. M., and Lilienfeld, D. E. (1980). "Foundations of Epidemiology," 2nd ed. Oxford Univ. Press, London and New York. 124. Lloyd, J. W., and Ciocco, A. (1969). JOM, J. Occup. Med. 11, 299-310. 125. Lloyd, J. W., Lundin, F. E., Jr., Redmond, C. K., and Geiser, P. B. (1970). JOM, J. Occup. Med. 12, 151-157. 126. Lundin, F. E., Jr., Archer, V. E., and Wagoner, J. F. (1979). In "Energy and Health" (N. E. Breslow and A. S. Whittemore, eds.), pp. 243-264. Soc. Ind. App. Math., Philadephia, Pennsylvania. 127. Maclennan, R., Muir, C., Steinitz, R., and Winkler, A. (1978). "Cancer Registration and its Techniques," IARC Sci. Publ. No. 21. Int. Agency Res. Cancer, Lyon. 128. MacMahon, B., and Pugh, T. F. (1970). "Epidemiology: Principles and Methods." Little, Brown, Boston, Massachusetts. 129. MacMahon, B., Yen, S., Trichopoulos, D., Warren, K., and Nardi, A. (1981). N. Engl. J. Med. 304, 630-633. 130. Mancuso, T. F., and Coulter, E. J. (1959). Am. J. Public Health 49, 1525-1536. 131. Mancuso, T. F., and El-Attar, A. A. (1967). J. Occup. Med. 9, 277-285. 132. Mancuso, T. F., and El-Attar, A. A. (1967). J. Occup. Med. 9, 147-162. 133. Manning, K. P., Skegg, D. C. A., Stell, P. M., and Doll, R. (1981). Clin. Otolaryngol. 6, 165-170. 134. Mantel, N. (1967). Cancer Res. 27, 209-220. 135. Mantel, N., and Haenszel, W. (1959). JNCI, J. Nat. Cancer Inst. 22, 719-748. 136. Mason, T. J., Fraumeni, J. F., Jr., and McKay, F. W. (1972). JNCI, J. Nat. Cancer Inst. 49, 661-664. 137. Mason, T. J., and McKay, F. W. (1974). "U.S. Cancer Mortality by County: 19501969," DHEW Publ. No. (NIH) 74-615. U.S. Govt. Printing Office, Washington, D.C. 138. Mason, T. J., McKay, F. W., Hoover, R., Blot, W. J., and Fraumeni, J. F., Jr. (1975). "Atlas of Cancer Mortality for U.S. Counties: 1950-1969," DHEW Publ. No. (NIH) 75-780. U.S. Govt. Printing Office, Washington, D.C. 139. 140. 141. 142. 143. 144. 145. 146. 147. 148. 149. 150. 151. 152. 153. 154. 155. 156. 157. 158. 159. 160. 161. 162. 163. 164. 165. 166. 167. 187 Mason, T. J., McKay, F. W., and Miller, R. W. (1974). JAMA, J. Am. Med. Assoc. 228, 1019-1020. Mausner, J. S., and Bahn, A. K. (1974). "Epidemiology: An Introductory Text." Saunders, Philadelphia, Pennsylvania. Mazumdar, S., and Redmond, C. K. (1979). In "Energy and Health" (N. E. Breslow and A. S. Whittemore, eds.), pp. 265-282. Soc. Ind. Appl. Math., Philadelphia, Pennsylvania. Mazumdar, S., Redmond, C. K., Sollecito, W., and Sussman, N. (1975). /. Air Pollut. Control Assoc. 25, 382-389. McKinlay, S. M. (1974). Appl. Stat. 23, 372-383. McMichael, A. J., Andjelkovic, D. A., and Tyroler, H. A. (1976). Ann. N.Y. Acad. Sci. 271, 125-137. McMichael, A. J., Haynes, S. A., and Tyroler, H. A. (1975). JOM, J. Occup. Med. 17, 128-131. McMichael, A. J., Spirtas, R., Gamble, J. F., and Tousey, P. M. (1976). JOM, J. Occup. Med. 18, 178-185. McMichael, A. J., Spirtas, R., Kupper, L. L., and Gamble, J. F. (1975). JOM, J. Occup. Med. 17, 234-239. McNemar, Q. (1947). Psychometrika 12, 153-157. Medical Services Study Group of the Royal ,ColIege of Physicians of London (1978). Br. Med. J. 2, 1063-1065. Menzel, H. (1950). Am. Social. Rev. 15, 674. Miettinen, O. S. (1970). Am. J. Epidemiol. 91, 111-118. Miettinen, O. S. (1974). Am. J. Epidemiol. 100, 350-353. Miller, J. A., Kline, B. E., Rusch, H. P., and Baumann, C. A. (1944). Cancer Res. 4, 153-158. Monson, R. R., Peters, J. M., and Johnson, M. N. (1974). Lancet 2, 397-398. Morgan, R. W., Kaplan, S. D., and Gaffey, W. R. (1981). JOM, J. Occup. Med. 23, 13-21. Morris, S. C., Shapiro, M. A., and Waller, J. H. (1976). Arch. Environ. Health 31, 248-254. Morrison, A. S., and Buring, J. E. (1980). N. Engl. J. Med. 302, 537-541. Mottran, J. C. (1944). J. Pathol. Bacterial. 56, 181-185. Muir, C. S. (1975). In "Persons at High Risk of Cancer: An Approach to Cancer Etiology and Control" (J. F. Fraumeni, Jr., ed.), pp. 293-305. Academic Press, New York. National Center for Health Statistics (1978). "Physician's Handbook on Medical Certification: Death, Birth, Fetal Death," DHEW Publ. No. (PHS) 78-1108. U.S. Govt. Printing Office, Washington, D.C. National Center for Health Statistics (1978). "Vital Statistics: Instructions for Classifying the Underlying Cause of Death, 1979," Part 2a. U.S. Govt. Printing Office, Washington, D.C. Naus, J. I. (1965). J. Am. Stat. Assoc. 60, 532-538. Newberne, P. M., Weigert, J., and Kula, N. (1979). Cancer Res. 39, 3986-3991. Newell, D. J. (1963). Biometrics 19, 187-188. Nordberg, G. F. (1976). "Effects and Dose-Response Relationships of Toxic Metals." Elsevier, Amsterdam. Ogle, W. (1885). "45th Annual Report of the Registrar-General of Births, Deaths, and Marriages in England." Registrar-General, London. Ohno, Y., Aoki, K., and Aoki, N. (1979). Int. J. Epidemiol. 8, 273-280. �188 Epidemiologic Approaches to Chemical Hazard Assessment John R. Wilkins III and Nancy A. Reiches 168. Ott, M. A., Holder, B. B., and Langner, R. R. (1975). Arch. Environ. Health 30, 333-339. 169. Ott, M. A., Holder, B. B., and Langner, R. R., (1976), JOM, J. Occup. Med. 18, 171-177. 170. Page, T., Harris, R. H., and Epstein, S. S. (1976). Science 193, 55-57. 171. Pastemack, B. S., and Shore, R. E. (1977). In "Environmental Health: Quantitative Methods" (A. Whittemore, ed.), pp. 49-71. Soc. Ind. Appl. Math., Philadelphia, Pennsylvania. 172. Percy, C., Garfinkel, L., Krueger, D. E., and Dolman, A. B. (1974). Public Health Rep. 89, 418-428. 173. Percy, C., Stanek, E., and Gloeckler, L. (1981). Am. J. Public Health 71, 242-250. 174. Peto, R. (1977). In "Origins of Human Cancer. Book A. Incidence of Cancer in Humans" (H. H. Hiatt, J. D. Watson, J. A. Winsten, eds.), pp. 1403-1428. Cold Spring Harbor Laboratory, Cold Spring Harbor, New York. 175. Pickle, L. W., and Gottlieb, M. S. (1980). Am. J. Public Health 70, 256-259. 176. Pike, M. C., and Smith, P. G. (1980). Biometrics 24, 541-556. 177. Pilot, H. C., and Sirica, A. E. (1980). Biochim. Biophys. Ada 60S, 191-196. 178. Polissar. L. (1980). Am. J. Epidemiol. Ill, 175-182. 179. Pollack, E. S., and Horm, J. W. (1980). JNCI, J. Nat. Cancer Inst. 64, 1091-1103. 180. Prentice, R. (1976). Biometrics 32, 599-606. 181. Redmond, C. K., and Breslin, P. P. (1975). JOM, J. Occup. Med. 17, 313-317. 182. Reiches, N. A., and Savitz, D. A. (1980). Am. J. Epidemiol. 112, 425. 183. Reinhardt, C. F., Azar, A., Maxfield, M. E., Smith, P. E., and Mullin, L. S. (1971). Arch. Environ. Health 22, 265-279. 184. Robinson, W. S. (1950). Am. Social. Rev. 15, 351-357. 185. Roels, H. A., Buchet, J. P., Lauwerys, R. R., Bruax, P., Claeys-Thoreau, F., Lafontaine, A., and Verduyn, G. (1980). Environ. Res. 22, 81-94. 186. Rogan, W. J., Bagniewska, A., and Damstra, T. (1980). N. Engl. J. Med. 302, 14501453. 187. Rogot, E., and Reid, D. D. (1975). Int. J. Epidemiol. 4, 51-54. 188. Rons, P., and Kidd, J. G. (1941). J. Exp. Med. 73, 365-371. 189. Rothman, K. J. (1974). Am. J. Epidemiol. 99, 385-388. 190. Rothman, K. J. (1976). Am. J. Epidemiol. 103, 506-511. 191. Rothman, K. J. (1978). Am. J. Epidemiol. 108, 347-349. 192. Rothman, K. J. (1981). Am. J. Epidemiol. 114, 253-259. 193. Rousch, G. C. (1980). Am. J. Epidemiol. Ill, 183-193. 194. Sackett, D. L. (1979). J. Chronic Dis. 32, 51-68. 195. Saracci, R. (1978). Int. J. Epidemiol. 7, 101-111. 196. Sartwell, P. E. (1950). Am. J. Hyg. 51, 310-318. 197. Schneiderman, M. A., Decoufle", P., and Brown, C. C. (1979). Ann. N.Y. Acad. Sci. 329, 92-130. 198. Schroeder, H. A. (1960). JAMA, J. Am. Med. Assoc. 172, 1902-1908. 199. Segi, M., and Kurihara, M. (1972). "Cancer Mortality in 24 Countries." Japan Cancer Society, Nagoya. 200. Seigel, D. G., and Greenhouse, S. W. (1973). J. Chronic Dis. 26, 219-225. 201. Sheps, M. C. (1966). Milbank Meml. Fund Q. 44, 69-91. 202. Shindell, S., Weisberg, R. F., and Giefer, E. E. (1978). JOM, J. Occup. Med. 20, 807-811. 203. Siegel, J. S. (1974). Demography 11, 1-23. 204. Silverberg, E. (1982). Co—Cancer J. Cttn. 30, 15-31. 205. 206. 207. 208. 209. 210. 211. 212. 213. 214. 215. 216. 217. 218. 219. 220. 221. 222. 223. 224. 225. 226. 227. 228. 229. 230. 231. 232. 233. 234. 235. 236. 189 Smith, A. H., Waxweiler, R. J., and Tyroler, H. A. (1980). Am. J. Epidemiol. 112, 787-797. Smith, R. J. (1980). Science 209, 998-1002. Spiegelman, M., and Marks, H. H. (1966). Hum. Biol. 38, 280-292. Spirtas, R. and Kaminski, R. (1978). JOM, J. Occup. Med. 20, 427-429. Stavraky, K. M. (1976). J. Chronic Dis. 29, 435-444. Stolley, P. D., Tonascia, J. 'A., Sartwell, P. E., Tockman, M. S., Tonascia, S., Rutledge, A., and Schinnar. R. (1978). Am. J. Epidemiol. 107, 226-235. Stone, B. J., Blot, W. J., and Fraumeni, J. F., Jr. (1978). JOM, J. Occup. Med. 20, 472-477. Susser, M. (1973). "Causal Thinking in the Health Sciences." Oxford Univ. Press, London and New York. Tassignon, J. P. (1979). JOM, J. Occup. Med. 21, 10. Thomas, D. G. (1975). Comput. Biomed. Res. 8, 423-446. Thomas, J. A. (1979). Fed. Proc., Fed. Am. Soc. Exp. Biol. 38, 1873-1874. Till, M. M., Hardisty, R. M., Pike, M. C., and Doll, R. (1967). Br. Med. J. 3, 755758. Toxic Substances Strategy Committee (1980). "Toxic Chemicals and Public Protection." U.S. Govt. Printing Office, Washington, D.C. Unger, M., and Olsen, J. (1980). Environ. Res. 23, 257-263. United States Bureau of Commerce (1966). "1963 Census of Manufactures." U.S. Govt. Printing Office, Washington, D.C. Vianna, N. J., and Greenwald, P. (1972). Ann. Intern. Med. 77, 169-180. Wada, S., Miyanishi, M., Nishimoto, Y., Kambe, S., and Miller, S. W. (1968). Lancet 1, 1161-1163. Wallenstein, S. (1980). Am. J. Epidemiol. Ill, 367-372. Walter, S. D., and Holford, T. R. (1978). Am. J. Epidemiol. 108, 341-346. Waterhouse, J., Muir, C. S., and Correa, P., eds. (1976). "Cancer Incidence in Five Continents." IARC Sci. Publ. 15. Int. Agency Res. Cancer, Lyon. Waxweiler, R. J. (1981). Envrion. Health Perspect. 42, 51-56. Weinstock, M. A. (1981). Int. J. Epidemiol. 10, 289-293. Whittemore, A. S. (1978). Adv. Cancer Res. 27, 55-58. Wilkins, J. R., Ill, Reiches, N. A., and Kruse\ C. W. (1979). Am. J. Epidemiol. 110, 420-448. Woods, J. S. (1979) Fed. Proc., Fed. Am. Soc. Exp. Biol. 38, 1891-18%. Woolf, B. (1955). Ann. Hum. Genet. 19, 251-253. Wu, M. (1980). J. Chronic Dis. 33, 207-213. Wynder, E. L., and Gori, G. B. (1977). JNCI, J. Nat. Cancer Inst. 58, 825-832. Wysowski, D. K., Landrigan, P. J., Ferguson, S. W., Fontaine, R. E., Tsongas, T. A., and Porter, B. (1978). Am. J. Epidemiol. 107, 27-35. Young, J. L., Asire, A. J., and Pollack, E. (1978). "SEER Program: Cancer Incidence and Mortality in the United States, 1973-1976," Nat. Cancer Inst. DHEW Publ. No. (NI-H) 78-1837. U.S. Govt. Printing Office, Washington, D.C. Young, J. L., Percy, C. L., and Asire, A. J., eds. (1981). "Surveillance, Epidemiology, and End Results: Incidence and Mortality Data, 1973-1977," Nat. Cancer Inst. Monogr. 57, DHHS Publ. No. (NIH) 81-2330. U.S. Govt. Printing Office, Washington, D.C. Zielhuis, R. L. (1979). In "The Use of Biological Specimens for the Assessment of Human Exposure to Environmental Pollutants" (A. Berlin, A. H. Wolff, and Y. Hasegawa, eds.), pp. 341-355. Martinus Nijhoff Publishers, The Hague, Netherlands. � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Alvin L. Young Collection on Agent Orange Description An account of the resource <p style="margin-top: -1em; line-height: 1.2em;">The Alvin L. Young Collection on Agent Orange comprises 120 linear feet and spans the late 1800s to 2005; however, the bulk of the coverage is from the 1960s to the 1980s and there are many undated items. The collection was donated to Special Collections of the National Agricultural Library in 1985 by Dr. Alvin L. Young (1942- ). Dr. Young developed the collection as he conducted extensive research on the military defoliant Agent Orange. The collection is in good condition and includes letters, memoranda, books, reports, press releases, journal and newspaper clippings, field logs and notebooks, newsletters, maps, booklets and pamphlets, photographs, memorabilia, and audiotapes of an interview with Dr. Young.</p> <p>For more about this collection, <a href="/exhibits/speccoll/exhibits/show/alvin-l--young-collection-on-a">view the Agent Orange Exhibit.</a></p> Text A resource consisting primarily of words for reading. Examples include books, letters, dissertations, poems, newspapers, articles, archives of mailing lists. Note that facsimiles or images of texts are still of the genre Text. Box The box containing the original item. 052 Folder The folder containing the original item. 1387 Series The series number of the original item. Series III Subseries II Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Creator An entity primarily responsible for making the resource Kang, Han K. Alvin L. Young Barclay M. Shepard Date A point or period of time associated with an event in the lifecycle of the resource 1984-05-01 Title A name given to the resource Typescript: Health Effects of Agent Orange Exposure Subject The topic of the resource soft tissue sarcoma AOWG veteran health and hygiene ao_seriesIII https://www.nal.usda.gov/exhibits/speccoll/files/original/b95bc46d3bd42a85abaaa8a1f7e34f5f.pdf 5e2ebbef6d46e98d88e8635a9b2320b2 PDF Text Text Item ID Number 01759 Author Kang, Han K. COPporatB Author Agent Orange Projects Office, Department of Medicine a RODOPt/ArtlClO TitlO Typescript: Research Protocol, a Matched Case-Control Study of Soft Tissue Sarcoma, March 1983 Journal/Book Title Year 000 ° Month/Day Color G Number of Images 21 Descriptor! Notes Monday, June 11, 2001 Page 1760 of 1793 �Research Protocol A Matched Case-Control Study of Soft Tissue Sarcoma Agent Orange Projects Office Department of Medicine and Surgery Veterans Administration and Department of Soft Tissue Pathology Armed Forces Institute of Pathology March 1983 Investigators: Han K. Kang, Dr.P.H., VA Franz W. Enzinger, M. D., AFIP Lawrence B. Hobson, M. D., Ph.D., VA Barclay M. Shepard, M. D., VA �I. Introduction There is much concern in the United States that many veterans report health problems that possibly stem from their military service in Vietnam. Their complaints include a wide variety of medical problems such as psychological, dermatological and physiological illnesses, reproductive disorders and even cancer. Agent Orange was the herbicide most commonly applied in Vietnam by the United States Air Force between 1965 and 1971. It was a mixture of the two commercial herbicides, 2,4-D (2,4-dichlorophenoxyacetic acid) and 2,4,5-T (2,4,5-trichlorophenoxyacetic acid). The 2,4,5-T contained minute amounts of an extremely toxic chemical, dioxin (2,3,7,8-tetrachlorodibenzo-p-dioxin or TCDD), which contaminated the herbicide during the manufacturing process. TCDD is teratogenic and carcinogenic in experimental animals (Poland and Knutson, 1982; Kociba et al., 1978; NCI, 1980). The possibility that exposure to the herbicide may induce rare forms of cancer in humans such as soft tissue sarcoma (STS) has been suggested from recent studies in Sweden (Hardell and Sandstrom, 1979; Hardell, 1981). The Swedish studies have shown that persons reporting exposure to phenoxy herbicides have a 5 to 6 fold higher risk of developing STS compared to persons without such exposure. A similar risk was reported by one of the Swedish investigators for malignant lymphoma (Hardell et al., 1981). These significant observations have not yet been replicated by other research teams and studies by Finnish (Riihimaki, 1982) and New Zealand investigators (Smith et al., 1982) failed to show an association of STS with exposure to phenoxy herbicides. However, several confirmed cases of STS have been reported among workers involved in the manufacturing or use of phenoxy herbicides (Cook, 1981; Honchar and Halperin, 1981; Moses and Selikoff, 1981). These industrial workers, in contrast to the herbicide applicators, are believed to be exposed to relatively high levels of the TCDD contaminant. 2. �Soft tissue sarcomas are a complex and diverse group of malignant neoplasms that originate in nonepithelial extraskeletal supporting structures of the body, excluding the hanatopoietic-lynphatic system, the glia and supporting tissues of specific organs and tissues (Enzinger, et al.f 1969). Soft tissue sarcomas account for about 1% of all malignant neoplasms and for about 2% of all cancer deaths. The average annual age-adjusted incidence rate is 3.89 per 100,000 and it is estimated that about 8,000 patients are diagnosed with STS each year in the United States (Cutler and Young, 1975). The most common histologic types are malignant fibrous histiocytoma, leiomyosarcoma, sarcoma not otherwise specified, liposarcorna and fibrosarcoma in that order. Little is known about the etiology of STS. The epidemiologic study of STS has been especially difficult because of uncertainties in the morphologic classification of this diverse group of neoplasms. In addition, the International Classification of Disease (ICD), being site-oriented, does not distinguish between the heterogeneous types of sarcoma. A small proportion of cases are probably related to Mendelian syndromes and the familial multiple-cancer syndrome (Tucker and Fraumeni, 1981; Blattner et al., 1979). An excess of STS has also been reported in patients receiving threapeutic immunosuppression for renal transplantation and other conditions (Hoover and Fraumeni, 1973; Kinlen et al., 1979). Some cases are associated with genetically determined immunodeficiency syndromes (Spector et al., 1978). Patients with chronic lymphocytic leukemia are also prone to STS (Greene et al., 1978). There is very limited information on environmental risk factors for STS. A small fraction of STS is induced by heavy external radiation therapy for various benign disorders and malignant tumors. Nearly all cell types of STS have been described following radiation, the most common being fibrosarcoma (Kim et al., 1978; Czesnin and Wronkowski, 1978). Some radioactive materials used for diagnostic or therapeutic purposes may induce sarcomas at or near sites of deposition (Falk et al., 1979a; McKillop et al., 1978). The best known examples of associations between specific chemicals and STS of specific cell types is angiosarcoma of the liver and exposure to vinyl chloride or inorganic arsenical compounds. Among 168 deaths from hepatic angiosarcoma during 1964-1974 in the United States, 37 deaths were associated with vinyl chloride, thorotrast, or 3. �inorganic arsenic (Falk et al., 1979b). The increased risk of developing STS among Swedish workers exposed to phenoxy herbicides or chlorophenols was described earlier. In view of the concern raised by many veterans that their contact with Agent Orange during Vietnam service may increase the risk of developing STS and conflicting research findings in the scientific literature regarding association between exposure to phenoxy herbicides and STS, we have decided to conduct an independent epidemiologic study. �II. Research Questions A. Phase I: Study based on the existing records 1. Does service in Vietnam during 1965-1971 increase or decrease the risk of developing STS among veterans? 2. Is there a trend in the odds of developing STS with increasing probability of exposure to Agent Orange? 3. Does the histopathology and anatomic site of STS among Vietnam veterans differ from those of non-Vietnan veterans and non-veterans (i.e., individuals who never served in the military)? B. Phase II: Study based on the existing records and information obtained from interviews 4. What are the other host or environmental risk factors for the development of STS? Factors to be considered are: (a) Occupational and non-occupational exposure to phenoxyacetic acids herbicides and other chlorophenols; (b) Exposue to phenoxyacetic acid containing drugs such as clofibrate; (c) Other factors such as genetic syndromes, immunologic deficiency, lymphedema, trauma and exposure to ionizing radiation, asbestos, arsenic, vinyl chloride and steroids. �III. Study Design A matched case-control study design will be used, in which individuals with STS (cases) are compared with individuals without STS (controls) with respect to Vietnam service, probable Agent Orange exposure and other possible risk factors. The case-control method is chosen primarily because it is well suited to the study of rare diseases (annual incidence of STS is 3.9/100,000), it is relatively quick and inexpensive, and it allows the study of multiple potential causes of the disease. 1. Cases Cases will be drawn from accession lists of the Armed Forces Institute of Pathology (AFIP). The AFIP offers a unique resource to contribute to this study. The AFIP routinely provides consultation services for pathologists throughout the United States, especially for conditions such as STS which present special diagnostic problems. One quarter to one third of the STS's occurring in the United States are sent to AFIP for review. Thus, the AFIP is one of the largest single registries in the world for this group of tumors. The uniformity and high quality of diagnoses at AFIP give it an added advantage as a resource for epidemiologic studies. Selection will be restricted to males, who were diagnosed as STS patients sometime between January 1, 1975 and December 31, 1980 and were aged 20 to 40 at the time of diagnosis. These eligibility criteria are established 1) to restrict the study to persons who were potentially at risk of exposure to Agent Orange; that is, persons would have been aged 18 to 23 sometime during 1965 and 1971, the period when Agent Orange was most heavily used in Vietnam, 2) to allow a minimum of 4 years of latency period, and 3) to reduce selection bias by restricting the cases to those referred to AFIP before the recent publicity on Vietnam service (or Agent Orange exposure) and the risk of developing STS. 6. �2. Controls Controls will be selected from the patient logs of referring pathologists or their pathology department. This is to duplicate the selective factors (e.g., socioeconomic status, area of residency, etc.) which bring people to these hospitals or clinics. Excluded from consideration as controls will be diagnoses of STS, non-Hodgkins lymphoma and Hodgkins disease. The latter two conditions have been associated with exposure to phenoxyacetic acid herbicides, chlorophenols, or their contaminants (Hardell, 1981). A contact person in each referring pathology unit, usually a medical assistant or nurse, will be asked to select the two sequential patients who matched the case by race, sex and age (+_ 5 years): one with a malignant neoplasm, and one with a non-malignant disease in the log book following the STS case. A pilot study is needed to test the feasibility of this control selection method. Ihere are several reasons for choosing two controls per case: one control from other cancer patients and the other control frcra non-cancer patients. First, 2:1 matching will increase the statistical power of the study. This will be discussed in the following section. Second, possible recall bias and interviewer bias can be minimized by selecting other cancer patients as controls. The cancer patient may try harder to remember his exposure to well publicized chemical carcinogens and radiation. Recall bias may, therefore, occur when these patients are compared with individuals with no cancer. In addition, the interviewer may tend to probe the cancer patients or their families more intensively for histories of exposure than they might for the control subjects or their families. Third, on the other hand, it is also possible that exposure to phenoxy herbicides and chlorophenols may cause some of the cancers in the control group and this would mask an association between an exposure to these chemicals and STS. Having non-cancer patient controls would eliminate this possibility. Recall bias or interviewer bias will not be a problem for determining military service status because this will be verified by records kept by the Veterans Administration* and the National Personnel Records Center**. �* VA BIRLS (Beneficiary Identification and Records Locator Subsystem). The Veterans Administration maintains a file of nearly 40 million computerized records known as the BIRLS file. It contains the veteran's name, date of birth and social security number and/or service numbers. Although Vietnam service is one of the information categories represented, this information is not provided on most of the records. Prior to 1972 only veterans who filed a claim for VA services were placed in the BIRLS file. Since January 1973 the names of all discharged veterans have been listed. ** National Personnel Records Center (NPRC). Ihe NPRC is the major repository for records of veterans who have been discharged from the service. Ihis is believed to be the best records center for determining the Vietnam service status of veterans. Absence of the study subject's record in the center would indicate that he did not serve in the military or that he was still on active duty. �IV. Statistical Considerations 1. Sample size determination The number of people to be selected for the study depends on the specifications of four values: (1) the relative frequency of risk factor among controls in the target population, Po, (2) a hypothesized relative risk associated with the risk factor that would have sufficient public health importance to warrant its detection, R, (3) the desired level of significance, alpha; that is, the probability of making an error of claiming that the risk factor under investigation is associated with disease when in fact it is not; (4) the desired study power, 1-beta; that is, the probability of claiming that the risk factor is associated with disease when in fact it is. Sample size for the study was determined under the following conditions: (1) alpha = 0.05 (2) beta = 0.2 or 1-beta = 0.8 (3) R - 2 (4) P0 = 0.05 (5) two sided test (6) two matched controls per case Under these conditions we will need a total of 500 cases and 1,000 controls. Please see attachment 1 for detailed calculation. Assuming about an 85% success rate for obtaining appropriate records for other condidtions, we will start with 600 cases of STS from the AFIP file. Preliminary data provided by the AFIP indicate obtaining 600 cases will not be a problem. (Please see attachment 2) As of the end of 1980 about 1,100 cases had already met the criteria for cases. 9. �2. Analyses of Data The data will be analyzed using conventional epidemiologic and biostatistical methods including the following: a. When a single univariate binary risk factor is considered, the following matched analysis with two controls per case is used: X2 =f fi2 PI -i* = f P2 L s.e.(?2 ~ P1) j roB [(m-D B-mA]2 ~ •(-I P 1 P = 2= ' 'ttle proportion of controls having risk factor T-h, 3 ." A ' >Ine proportion of cases having the risk factor N where, A = the total number of controls with the risk factor B = the total number of either cases or controls having the risk factor N = the total number of matched triples m = 3 (1 case + 2 controls) ni = number of either case or controls having the risk factor within a given triple xi = number of controls with the risk factor within a given triple The odds ratio is calculated as follows: *• 0 = (m-1) (B-A) -L*;(nj. -5-* A x ~ 10. �b. An attempt to explore the individual and joint effects of a number of variables will be made using multivariate statistical analysis based on the linear logistic model. This technique enables one to investigate the effect of several variables simultaneously in the analysis while allowing for the matched design (Holford et al., 1973; Breslow et al., 1978). �V. Proposed Study Strategies A. Phase I: A study based on the existing records 1. Case selection a. Tabulate all male STS cases referred to AFIP during January 1, 1975 and December 31, 1980 by specific diagnosis and by age. b. Identify from the APIP records the males aged 20-40 at the time of diagnosis. c. Randomly select a total of 600 cases among all eligible cases. d. Obtain necessary information from AFIP records (name, age, name of pathologist and his location, etc.) for each case. 2. Control selection a. Secure the consent of the pathologist whose patients will be approached to participate in the study. b. Select controls from pathology log books with the cooperation of referring pathologists or their assistants. Controls will be be matched to case by sex, race, age (+5 years). The first two eligible patients (one with cancer excluding STS, non-Hodgkin's lymphoma, and Hodgkins disease and one with non-malignant disease) filed immediately after the case will be selected for controls. 3. Determination of military service status a. Provide the National Personnel Records Center (NPRC) in St. Louis a listing or computer tape containing full names of both cases and controls, social security number and other identifying information obtained from the AFIP, referring pathologists and primary care physicians. �b. The NPRC will search and pull military personnel records for on-site review by VA contractor employees. c. The contractor will review and extract necessary information from the file. d. Cases or controls from non-military hospitals whose records are not kept in the NPRC could be either non-veterans (never served in the military), or still on active duty. However, if one assumes that active duty servicemen use military hospitals especially for the treatment of illness that requires referral to pathologists and since these cases and controls are from non-military hospitals it would be almost certain that they did not serve in the military or that they are on reserve duty. e. The cases and controls from military hospitals will be referred to the military personnel records centers of each branch of service for ascertaining active duty status and obtaining appropriate military records. f. The names and social security number (SSN) of cases and controls will be cross checked with the VA BIRLS file. The BIRLS file contains a record for each VA beneficiary and as of January 1973, BIRLS began including records for all veterans at separation from military service. g. Develop an Agent Orange exposure ranking scheme based on MOS data and other information extracted from military records. 4. Initial analysis of data obtained from the available records. The first three research questions listed on page 5 can be addressed with information obtained from the records. �B. Phase II 1. Locate cases and controls with help from pathologists, the surgeon's office and/or primary physicians. Since cases and controls medical records go back only a maximum of 7 years, it may not be insurmountable to locate them. However, this effort will be complemented by the following tracing mechanisms: a. IRS-NIOSH-SSN b. c. d. e. Telephone directory Post Office State motor vehicle department Credit bureau 2. Develop a questionnaire 3. Prepare introductory and informed consent letters and obtain consent of cases and controls, or their next-of-kin prior to conducting interviews, in accordance with existing regulations. 4. Develop an interview schedule. Conduct a pretest and make necessary revisions. 5. Conduct telephone interviews of the cases and controls, or their next-of-kin. 6. Review, edit and code all completed interviews. 7. Analyze data. 8. Final Report. 14. �VI. Confidentiality Confidentiality of all records pertaining to individuals in the study will be carefully protected. Names of individuals will be used solely to locate persons for the purpose of determining their military service status and of interviewing. Personal identifiers will not be retained on any data record used for analysis, nor will they be included in any publication or other presentation of study results. Records with personal identifiers will be under the control of VA and AFIP investigators or their agents and will not be accessible to other individuals or groups. �Acknowledgements We are indebted to Drs. Kenneth Cantor and Shelia Hoar of the National Cancer Institute and Dr. Carolyn Lingeraan of the National Institute of Environmental Health Sciences for giving us the impetus to initiate this study. �References Blattner, W. A., McGuire, D. B., Mulvihill, J. J., et al. (1979) Genealogy of cancer in a family. JAMA 241:259-261 Breslow, N. E., Day, N. E., Halvorsen, K. T., et al. (1978) Estimation of multiple relative risk functions in matched case-control studies. American J. Epidemiol. 108:299-307 Cook, R. R. (1981) Dioxin, Chloracne and Soft Tissue Sarcoma. Lancet 1:618-619 Culter, S. J. and Young, J .L. (eds): Third National Cancer Survey: Incidence Data. NCI Mpnogi 41:1-454, (1975) Czesnin, K. and Wronkowski, Z. (1978) Second malignancies of the irradiated area in patients treated for uterine cervix cancer. Gynec Oncol. 6:309-315 Enzinger, F. M., Lattes, R. and Torloni, H. (1979) Histological typing of soft tissue tumors. World Health Organization, Geneva Falk, H., Telles, N.C., Ishak, K. G. et al. (1979a) Epidemiology of thorotrast-induced hepatic angiosarcoma in the United States. Environ. Res._ 18:65-73 Falk, H., Thomas, L. B., Popper, H. et al. (1979b) Hepatic angiosarcoma associated with adiogenic-anabolic leukemia steroids. Lancet 2:1120-1123 Greene, M. H., Hoover, R. N. and Fraumeni, J. F., Jr. (1978) Subsequent cancer in patients with chronic lymphocytic leukemia - a possible immunologic mechanism. JNCI 61:337-340 Honchar, P. A. and Halperin, W. E. (1981) 2,4,5-T, trichlorphenol and soft tissue sarcoma. Lancet 1:268-269 Hardell, L. and Sandstrom, A. (1979) Case-control study: Soft tissue sarcomas and exposure to phenoxyacetic acides or chlorophenolsa. Br. J. Cancer 39:711-717 Hardell, L., Eriksson, M., Lenner, P. and Lundgren, E. (1981) Malignant lymphomo and exposure to chemicals, especially organic solvents, chlorophenols and phenoxy acids: A case-control study. Br.J. Cancer 43:169-176 Hardell, L. (1981) Relation of soft-tissue sarcoma, malignant lymphoma and colon cancer to phenoxy acids, chlorophenols and other agents. Scand. J. Work Environ. Health 7:119-130 Hardell, L. and Eriksson, M. (1981) Soft-tissue sarcoma, phenoxy herbicides and chlorinated phenols. Lancet 2:250 �Attachment 1 a. Ihe relative frequency of the risk factor among controls in the target population, P0 No combat (0.32) low combat (0.43) rvice (0.36) Vietnam (+) (0.3) High combat (0.26) Males Non-service (0.64) Vietnam (-) (0.7) (1) risk factor = Vietnam service/high combat duty Assuming the servicemen in the Vietnam/High combat category were most likely exposed to Agent Orange, the P0 was calculated as follows: P0 - 0.36 x 0.3 x 0.26 = 0.029 (2) risk factor = Vietnam service P0 = 0.36 x 0.3 = 0.11 (3) risk factor = Vietnam service/combat (high + low) P0 = 0.36 x 0.3 x (0.43 + 0.26) » 0.07 (4) risk factor = Occupational and non-occupational herbicide exposure NCI assumes Po = 0.1 �References Holford, T. R. , White, C. and Kelsey, J. (1978) Multivariate analysis for matched case-control studies. American J. Epidemiol . 107:245-256 Hoover, R. and Fraumeni, J. F. , Jr. (1973) Risk of cancer in renal transplant recipients. Lancet 2:55-57 Johnson, F. E., Kugler, M. A. and Brown, S. M. (1981) Soft tissue sarcoma and chlorinated phenols. Lancet 2:40 Kim, J. H., Chu, F. C. , Woodward, H. 0., et al. (1978) Radiation-induced soft tissue and bone sarcoma. Radiol. 129:501-508 Kinlen, L. J., Sheid, A. G. R. , Peto, J. et al. (1979) Collaborative United Kingdom - Australian study of cancer in patients treated with immuno-suppressive dru< s Br> 3Med* J. 2:1461-1466 Kociba, R. J., Keyes, D. G. , Beyer, J. E. et al. (1978) Results of a two year chronic toxicity and oncogenicity study of 2,3,7,8-tetrachlorodibenzo-p-dioxin in rats. Ibxicol. Appl. Pharmacol. 46:279-303 McKillop, J. H., Doig, J. A., Kennedy, J. S., et al. (1978) Laryngeal malignancy following iodine - 125 therapy for thyrotoxicosis. Lancet 2:1177-1179 Moses, M. and Selikoff, I. J. (1981) Soft tissue sarcomas, phenoxy herbicides and chlorinated phenols. Lancet 1:1370 National Cancer Institute (1980) DHHS Publication Number NIH 80-1765 Poland, A. abd Knutson, J. (1982) 2,3,7,8-tetrachlorodibenzo-p-dioxin and related halogenated hydrocarbons. Examination of the mechanism of toxicity. American Rev. Pharmacol. ._ Ibxicol . 22:517-554 Riihimaki, V. (1982) Mortality of 2,4-dichlorophenoxyacetic acid and 2,4,5trichlorophenoxy acetic acid herbicide applicators in Finland. Scand. J. Work Environ. Health 8:37-42 Smith, A. H., Fischer, D. 0 , Pearce, N. and Teague, C. A. (1982) Do . agricultural chemicals cause soft tissue sarcoma? Initial findings of a case-control study in New Zealand. Community Health Studies 6:114-119 Spector, B. D. , Perry, G. S. and Kersey, J. H. (1978) Genetically determined immunodeficiency disease (GDID) and malignancy: Jteport from the immunodeficiency - cancer registry. Clin. Immunol. Immunopathol . 11:1 2-29 Tucker, M. A. and Fraumeni, J. F. , Jr.: Soft tissue in Cancer Epdimeiology and Prevention. D. Schottenfeld and J. F. Fraumeni, Jr. (eds). Philadelphia W. B. Saunders Co. 1981. pp 827-836 �Attachment 1 - Continued We have chosen a conservative number Po = 0.05 for the study. b. Sample size with two controls per case. (1) P0 = 0.05, alpha = 0.05, beta = 0.10, R=2 r _ n = [Z«,\,'(1 + 1/c) F-q> _ + r>2. Z^^P i q i + P0qo/c J/ (Pi - P 0 )2 where, P^ = po R/[1 + po (R-1)] = (P! + CPO) / d + o F 31 c = 1 - Plf and q- = 1 - p= number of controls per case n = [1.96 V '(1 + 1/2) 0.065 x 0.935 + 1.28\/0.095 x 0.905 + (0.05 x 0.95)/2] divided (0.095 - 0.052)2 = 509 (2) c. P0 = 0.05, alpha = 0.05, beta = 0.20, R=2 n = 372 Sample size with two matched controls per case (1) P0 = 0.05, alpha = 0.05, beta = 0.10, m = [ Z «'/2 + Zp \/P(1-p)]2 / (p-i/2)2 = 90 P M = Vd+R) = 2/3 = 0.667 ^mAP^ + PT qo) = 90/0.135 = 666 (2) P0 = 0.05, alpha = 0.05, beta = 0.20, R=2 m =67.7 M = 501.8 �SOFT TISSUE SARCOMAS AFIP 1975 -- 1980 Ages 20 to 90 years (Excludes military and dependents) 1 Number I % M 20- F 50- 40- Ages 50- 60- 70- 80- Malignant Fibrous Histiocytoma 1921 24 1114 805 80 119 208 588 515 416 197 Leiomyosarcoma 1285 15 592 692 56 124 206 295 527 216 63 Sarcoma NOS 987 12 520 466 238 158 158 154 140 118 41 Liposarcoma 800 10 485 514 55 104 150 185 195 105 2i Fibrosarcoma 510 6 288 221 86 77 71 102 91 58 25 Malignant Schwannoma 465 6 250 211 95 75 75 69 77 58 1 c Atypical Fibroxanthoma 411 5 290 119 8 15 21 65 89 124 9 : Hemangiopericytoma 556 4 144 212 59 61 70 78 45 29 1- Dermatofibrosarcoma Protruberans 541 4 170 169 75 96 66 59 28 14 •: Poorly Differentiated Sarcomas 540 4 192 147 96 47 45 57 54 55 1C Angiosarcomas 254 5 159 115 50 24 57 45 Ji 42 2] Synovial Sarcomas 247 5 154 115 ; 82 65 57 55 24 5 - Rhabdomyosarcomas 158 2 88 50 ! 72 14 15 17 11 5 4 Myxofibrosarcomas 2" 15 12 > 1 5 5 5 6 6 - Neurogenic Sarcomas _5 . 2 5 1 1 1 2 0 0 c, 4425 5647 8085 C J>vfc,< 1 - v/t> tf. » i >t> l l � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Alvin L. Young Collection on Agent Orange Description An account of the resource <p style="margin-top: -1em; line-height: 1.2em;">The Alvin L. Young Collection on Agent Orange comprises 120 linear feet and spans the late 1800s to 2005; however, the bulk of the coverage is from the 1960s to the 1980s and there are many undated items. The collection was donated to Special Collections of the National Agricultural Library in 1985 by Dr. Alvin L. Young (1942- ). Dr. Young developed the collection as he conducted extensive research on the military defoliant Agent Orange. The collection is in good condition and includes letters, memoranda, books, reports, press releases, journal and newspaper clippings, field logs and notebooks, newsletters, maps, booklets and pamphlets, photographs, memorabilia, and audiotapes of an interview with Dr. Young.</p> <p>For more about this collection, <a href="/exhibits/speccoll/exhibits/show/alvin-l--young-collection-on-a">view the Agent Orange Exhibit.</a></p> Text A resource consisting primarily of words for reading. Examples include books, letters, dissertations, poems, newspapers, articles, archives of mailing lists. Note that facsimiles or images of texts are still of the genre Text. Box The box containing the original item. 065 Folder The folder containing the original item. 1759 Series The series number of the original item. Series III Subseries III Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Creator An entity primarily responsible for making the resource Kang, Han K. Franz W. Enziger Lawrence B. Hobson Barclay M. Shepard Description An account of the resource <strong>Corporate Author: </strong>Agent Orange Projects Office, Department of Medicine and Surgery, VA and Department of Soft Tissue Pathology, AFIP Title A name given to the resource Typescript: Research Protocol, a Matched Case-Control Study of Soft Tissue Sarcoma, March 1983 Subject The topic of the resource VA/AFIP STSS study protocol ao_seriesIII https://www.nal.usda.gov/exhibits/speccoll/files/original/e9f2ffdda5301f1a4edffa2a2bc52518.pdf d4cd75ec9d13150a353331fd639e5cff PDF Text Text Hem ID Number °1762 Author Kan 9>Han K Corporate Author Agent Orange Projects Office, Department of Medicine a RODOrt/ArtlOlO Title Typescript: Research Protocol, a Matched Case-Control Study of Soft Tissue Sarcoma, March 1984 Journal/Book Title Year Month/Day Color D Number of unagos 46 Doscrlpton Notes Monday, June 11, 2001 Page 1763 of 1793 �Research Protocol A Matched Case-Control Study of Soft Tissue Sarcoma Agent Orange Projects Office Department of Medicine and Surgery Veterans Administration and Department of Soft Tissue Pathology Armed Forces Institute of Pathology March 1984 Investigators: Han K. Rang, Dr.P.H., VA Franz M. Enzinger, M. D., AFIP Lawrence B. Hobson, M. D., Ph.D., VA Barclay M. Shepard, M. D., VA Patricia P. Breslin, Ph.D., VA �I . Introduction is much concern in the United States that many veterans report health problems that possibly stem from their military service in Vietnam. Their complaints include a wide variety of medical problems such as psychological, dermatological and physiological illnesses, reproductive disorders and even cancer. Agent Orange was the herbicide most commonly applied in Vietnam by the United States Air Force between 1965 and 1971. It was a mixture of the two commercial herbicides, 2,4-D (2,4-dichlorophenoxyacetic acid) and 2,4,5-T (2,4,5-trichlorophenoxyaoetic acid). The 2,4,5-T contained minute amounts of an extremely toxic chemical, dioxin (2,3,7,8-tetrachlorodibenzo-p-dioxin or TCDD), which contaminated the herbicide during the manufacturing process. TCDD is teratogenic and carcinogenic in experimental animals (Poland and Khutson, 1982; Kociba et al., 1978; NCI, 1980). The possibility that exposure to the herbicide may induce rare forms of cancer in humans such as soft tissue sarcoma (STS) has been suggested from recent studies in Sweden (Hardell and Sandstrom, 1979; Hardell, 1981). The Swedish studies have shown that persons reporting exposure to phenoxy herbicides have a 5 to 6 fold higher risk of developing STS compared to persons without such exposure. A similar risk was reported by one of the Swedish investigators for malignant lymphoma (Hardell et al., 1981). These significant observations have not yet been replicated by other research teams and studies by Finnish (Riihimaki, 1982) and New Zealand investigators (Smith et al., 1982) failed to show an association of STS with exposure to phenoxy herbicides. However, several confirmed cases of STS have been reported among workers involved in the manufacturing or use of phenoxy herbicides (Cook, 1981; Honchar and Halperin, 1981; Moses and Selikoff, 1981). These industrial workers, in contrast to the herbicide applicators, are believed to be exposed to relatively high levels of the TCDD contaminant. 2. �Soft tissue sarcomas are a complex and diverse group of malignant neoplasms that originate in nonepithelial extraskeletal supporting structures of the body, excluding the hematopoietic-lymphatic system, the glia and supporting tissues of specific organs and tissues (Enzinger, et al., 1969). Soft tissue sarcomas account for about 1% of all malignant neoplasms and for about 2% of all cancer deaths. The average annual age-adjusted incidence rate is 3.89 per 100,000 and it is estimated that about 8,000 patients are diagnosed with STS each year in the United States (Cutler and Young, 1975). The most common histologic types are malignant fibrous histiocytoma, leiomyosarcoma, sarcoma not otherwise specified, liposarcoma and fibrosarcoma in that order. Little is known about the etiology of STS. The epidemiologic study of STS has been especially difficult because of uncertainties in the morphologic classification of this diverse group of neoplasms. In addition, the International Classification of Disease (ICO), being site-oriented, does not distinguish between the heterogeneous types of sarcoma. i A small proportion of cases are probably related to Mendelian syndromes and the familial multiple-cancer syndrome (Tucker ard Fraumeni, 1981; Blattner et al., 1979). An excess of STS has also been reported in patients receiving threapeutic immunosoppression for renal transplantation and other conditions (Hoover and Fraumeni, 1973; Kinlen et al., 1979). Some cases are associated with genetically determined immunodeficiency syndromes (Spector et al., 1978). Patients with chronic lymphocytic leukemia are also prone to STS (Greene et al., 1978). There is very limited information on environmental risk factors for STS. A small fraction of STS is induced by heavy external radiation therapy for various benign disorders and malignant tumors. Nearly all cell types of STS have been described following radiation, the most common being fibrosarcoma (Kim et al., 1978; Czesnin and Wronkowski, 1978). Some radioactive materials used for diagnostic or therapeutic purposes may induce sarcomas at or near sites of deposition (Falk et al., 1979a; McKillop et al., 1978). The best known examples of associations between specific chemicals and STS of specific cell types is angiosarcoma of the liver and exposure to vinyl chloride or inorganic arsenical compounds. Among 168 deaths from hepatic angiosarcoma during 1964-1974 in the United States, 37 deaths were associated with vinyl chloride, thorotrast, or 3. �inorganic arsenic (Falk et al., 1979b). The increased risk of developing STS among Swedish workers exposed to phenoxy herbicides or chlorophenols was described earlier. In view of the concern raised by many veterans that their contact with Agent Orange during Vietnam service may increase the risk of developing STS and conflicting research findings in the scientific literature regarding association between exposure to phenoxy herbicides and STS, we have decided to conduct an independent epidemiologic study. �II. Research Questions A. Phase I: Study based on the existing records 1. Does military service in Vietnam increase or decrease the risk of developing STS among veterans? 2. Is there a trend in the odds of developing STS with increasing probability of exposure to Agent Orange? 3. Does the histopathology and anatomic site of STS among Vietnam veterans differ from those of non-Vietnam veterans and non-veterans (i.e., individuals who never served in the military)? B. Phase II: Study based on the existing records and information obtained from interviews 4. What are the other host or environmental risk factors for the development of STS? Factors to be considered are: (a) Occupational and non-occupational exposure to phenoxyacetic acids herbicides and other chlorophenols; (b) Exposue to phenoxyacetic acid containing drugs such as clofibrate; (c) Other factors such as genetic syndromes, immunologic deficiency, lymphedema, trauma and exposure to ionizing radiation, asbestos, arsenic, vinyl chloride and steroids. 5. �III. Study Design A matched case-control study design will be used, in which individuals with STS (cases) are compared with individuals without STS (controls) with respect to Vietnam service, probable Agent Orange exposure and other possible risk factors. The case-control method is chosen primarily because it is well suited to the study of rare diseases (annual incidence of STS is 3.9/100,000), it is relatively quick and inexpensive, and it allows the study of multiple potential causes of the disease. 1. Cases Cases will be drawn from accession lists of the Armed Forces Institute of Pathology (AFIP). The AFIP offers a unique resource to contribute to this study. The AFIP routinely provides consultation services for pathologists throughout the United States, especially for conditions such as STS which present special diagnostic problems. One quarter to one third of the STS's occurring in the United States are sent to AFIP for review. Thus, the AFIP is one of the largest single registries in the world for this group of tumors. The uniformity and high quality of diagnoses at AFIP give it an added advantage as a resource for epideraiologic studies. Selection will be restricted to males, who were diagnosed as STS patients sometime between January 1, 1975 and December 31, 1980 and were aged 20 to 40 at the time of diagnosis. These eligibility criteria are established 1) to restrict the study to persons who were potentially at risk of exposure to Agent Orange; that is, persons would have been aged 18 to 23 sometime during 1965 and 1971, the period when Agent Orange was most heavily used in Vietnam, 2) to allow a minimum of 4 years of latency period, and 3) to reduce selection bias by restricting the cases to those referred to AFIP before the recent publicity on Vietnam service (or Agent Orange exposure) and the risk of developing STS. The subject of possible selection bias by restricting cases to those from the AFIP registry has been considered at length . It was concluded that if a decision to refer the cases to the AFIP was made without information on study factors (Vietnam service, Agent Orange exposure and other phenoxy herbicide exposure), the use of the AFIP registry for selecting cases is still valid. 6. �In other words, unless there are differential referral patterns with respect to the presence or absence of study factors, one can make use of this unique resource for a valid epidemiologic study. It will be almost impossible to prove that there was no selection bias. However, a limited review of APIP registry data has shown that the proportion of "military age" (25-40 years old) cases among the total male soft tissue sarcoma cases and the total number of male cases referred to in the AFIP stay remarkably the same throughout the study period (1975-1980), which may indicate no large influx of military age cases to the AFIP. (see Attachment 1) It is well demonstrated that misclassification of soft tissue sarcoma may present a greater problem than possible selection bias. In a recent paper Percy et al. (1981) of the NCI reported that only about 56% of the soft tissue sarcoma deaths coded on the death certificates were confirmed by hospital records. Furthermore, the Nationa Institute for Occupational Safety and Health (Fingerhut et al., 1983) reported in a scientific meeting that two of the seven cases of soft tissue sarcoma cases previously reported in industrial workers which had generated much attention were found to be carcinomas. This determination was made by Dr. Franz Enzinger (AFIP) and his associate after reviewing all seven slides. 2. Controls Controls will be selected from the patient logs of referring pathologists or their pathology department. This is to duplicate the selective factors (e.g., socioeconomic status, area of residency, etc.) which bring people to these hospitals or clinics. Excluded from consideration as controls will be diagnoses of ST3, non-Hodgkin's lymphoma and Hodgkin's disease. The latter two conditions have been associated with exposure to phenoxyacetic acid herbicides, chlorophenols, or their contaminants (Hardell, 1981). A contact person in each referring pathology unit, usually a medical assistant or nurse, will be asked to select the two sequential patients who matched the case by race, sex and age: one with a malignant neoplasm, and one with a non-malignant disease in the log book following the STS case. A pilot study is needed to test the feasibility of this control selection method. �There are several reasons for choosing two controls per case: one control from other cancer patients and the other control from non-cancer patients. First, 2:1 matching will increase the statistical power of the study. This will be discussed in the following section. Second, possible recall bias and interviewer bias can be minimized by selecting other cancer patients as controls. The cancer patient may try harder to remember his exposure to well publicized chemical carcinogens and radiation. Recall bias may, therefore, occur when these patients are compared with individuals with no cancer. In addition, the interviewer may tend to probe the cancer patients or their families more intensively for histories of exposure than they might for the control subjects or their families. Third, on the other hand, it is also possible that exposure to phenoxy herbicides and chlorophenols may cause some of the cancers in the control group and this would mask an association between an exposure to these chemicals and STS. Having non-cancer patient controls would eliminate this possibility. Recall bias or interviewer bias will not be a problem for Phase I, determining military service status, because this will be verified by records kept by the Veterans Administration* and the National Personnel Records Center**. 8. �* VA BIRLS (Beneficiary Identification and Records Locator Subsystem). Ihe Veterans Administration maintains a file of nearly 40 million computerized records known as the BIRLS file. It contains the veteran's name, date of birth and social security number and/or service numbers. Although Vietnam service is one of the information categories represented, this information is not provided on most of the records. Prior to 1972 only veterans who filed a claim for VA services were placed in the BIRLS file. Since January 1973 the names of all discharged veterans have been listed. ** National Personnel Records Center (NPKC). Hie NPRC is the major repository for records of veterans who have been discharged from the service. This is believed to be the best records center for determining the Vietnam service status of veterans. Absence of the study subject's record in the center would indicate that he did not serve in the military or that he was still on active duty. 9. �IV. Statistical Considerations 1. Sample size determination The number of people to be selected for the study depends on the specifications of four values: (1) the relative frequency of risk factor among controls in the target population/ Po, (2) a hypothesized relative risk associated with the risk factor that would have sufficient public health importance to warrant its detection/ R, (3) the desired level of significance, alpha; that is, the probability of making an error of claiming that the risk factor under investigation is associated with disease when in fact it is not; (4) the desired study power, 1-beta; that is, the probability of claiming that the risk factor is associated with disease when in fact it is. Sample size for the study was determined under the following conditions: (1) alpha • 0.05 (2) beta » 0.2 or 1-beta =0.8 (3) R = 2 (4) P0 - 0.05 (5) two sided test (6) two matched controls per case Under these conditions we will need a total of 400 cases and 800 controls. Please see attachment 2 for detailed calculation. Assuming about an 80% success rate for obtaining appropriate records and information for other condidtions, we will start with 500 cases of STS from the AFIP file. 10. �2. .Analyses of Data The data will be analyzed using conventional epidemiologic and biostatistical methods including the following: a. When a single univariate binary risk factor is considered, the following matched analysis with two controls per case is used: X2 = r PO - Pi -i* - s.e.(P2 - PI)J 2* B-A mB - %_ ni2 proportion of controls having risk factor ___ N(ra-1) P [(m-1) B-mA]2 , The proportion of cases having the risk factor N where, A « the total number of controls with the risk factor B » the total number of either cases or controls having the risk factor N « the total number of matched triples m =» 3 ( 1 case + 2 controls) nj[ » number of either case or controls having the risk factor within a given triple x^ * number of controls with the risk factor within a given triple The odds ratio is calculated as follows: 0 - (m-1) (B-A) A - £ 11. �b. An attempt to explore the individual and joint effects of a number of variables will be made using multivariate statistical analysis based on the linear logistic model. This techinque enables one to investigate the effect of several variables simultaneously in the analysis while allowing for the matched design (Hoiford et al., 1978; Breslow et al., 1978). 12. �V. Proposed Study Strategies A. Phase I: A study based on the existing records 1. Case selection a. Tabulate all male STS cases referred to AFIP during January 1, 1975 and December 31, 1980 by specific diagnosis and by age. b. Identify from the AFIP records the males aged 20-40 at the time of diagnosis. c. Randomly select a total of 600 cases among all eligible cases. d. Obtain necessary information from AFIP records (name, age, name of pathologist and his location, etc.) for each case. 2. Control selection a. Secure the consent of the pathologist whose patients will be approached to participate in the study. b. Select controls from pathology log books with the cooperation of referring pathologists or their assistants. Controls will be be matched to case by sex, race (white, black, other), age*. The first two eligible patients (one with cancer excluding STS, non-Hodgkin's lymphoma, and Hodgkin's disease and one with non-malignant disease) filed immediately after the case will be selected for controls. 3. Determination of military service status a. Provide the National Personnel Records Center (NPRC) in St. Louis a listing or computer tape containing full names of both cases and controls, social security number and other identifying information obtained from the AFIP, referring pathologists and VA BIRLS. * Age Criteria for Controls AFIP Accession Age Range for Year of STS cases Controls 1975 22 - 33 1976 23 - 34 1977 24 - 35 1978 1979 1980 25 - 36 26 - 37 27 - 38 11. �b. The NPRC will search and pull military personnel records for on-site review by VA contractor employees. c. The contractor will review and extract necessary information from the file. d. Cases or controls frcra non-military hospitals whose records are not kept in the NPRC could be either non-veterans (never served in the military), or still on active duty. However, if one assumes that active duty servicemen use military hospitals especially for the treatment of illness that requires referral to pathologists and since these cases and controls are from non-military hospitals it would be almost certain that they did not serve in the military or that they are on reserve duty. e. The cases and controls from military hospitals will be referred to the military personnel records centers of each branch of service for ascertaining active duty status and obtaining appropriate military records. f. The names and social security number (SSN) of cases and controls will be cross checked with the VA BIRLS file. The BIRLS file contains a record for each VA beneficiary and as of January 1973, BIRLS began including records for all veterans at separation frcra military service. g. In collaboration with the Army Agent Orange Task Force, the VA will develop an Agent Orange exposure ranking scheme based on military unit and other information extracted from military records. 4. Initial analysis of data obtained from the available records. The first three research questions listed on page 5 can be addressed with information obtained from the records. 14. �B. Phase II 1. locate cases and controls with help from pathologists, the surgeon's office and/or primary physicians. Since cases and controls medical records go back only a maximum of 9 years, it may not be insurmountable to locate them. However, this effort will be complemented by the following tracing mechanisms: a. IRS-NIOSB-SSN b. c. d. e. Telephone directory Post Office State motor vehicle department Credit bureau 2. Develop a questionnaire and obtain OMB clearance. 3. Prepare introductory and informed consent letters and obtain consent of cases and controls, or their next-of-kin prior to conducting interviews, in accordance with existing regulations. 4. Develop an interview schedule. Conduct a pretest and make necessary revisions. 5. Conduct telephone interviews of the cases and controls, or their next-of-kin. 6. Review, edit and code all completed interviews. 7. Analyze data. 6. Final Report. 15. �VI. Confidentiality Confidentiality of all records pertaining to individuals in the study will be carefully protected. Names of individuals will be used solely to locate persons for the purpose of determining their military service status and of interviewing. Personal identifiers will not be retained on any data record used for analysis, now will they be included in any publication or other presentation of study results. Records with personal identifiers will be under the control of VA and AFIP investigators of their agents and will not be accessible to other individuals or groups. 16. �Acknowledgements We are indebted to Drs. Kenneth Cantor and Shelia Hoar of the National Cancer Institute and Dr. Carolyn Lingeman of the National Institute of Environmental Health Sciences for giving us the impetus to initiate this study. 17. �References Blattner, W. A., MeGuire, D. B., Mulvihill, J. J., et al. (1979) Genealogy of cancer in a family. JAMA 241:259-261 Breslow, N. E.r Day, N. E., Halvorsen, K. T., et al. (1978) Estimation of multiple relative risk functions in matched case-control studies. American J. Epidemiol. 108:299-307 Cook, R. R. (1981) Dioxin, Chloracne and Soft Tissue Sarcoma. Lancet 1:618-619 Culter, S. J. and Young, J .L. (eds): Third National Cancer Survey: Incidence Data. NCI Monogi 41:1-454, (1975) Czesnin, K. and Wronkowski, Z. (1978) Second malignancies of the irradiated area in patients treated for uterine cervix cancer. GynecOncol. 6:309-315 Enzinger, F. M., Lattes, R. and Torloni, H. (1979) Histological typing of soft tissue tumors. World Health Organization, Geneva Falk, H., Ttelles, N.C., Ishak, K. G. et al. (1979a) Epidemiology of thorotrast-induced hepatic angiosarcoma in the United States. Environ. Res. 18:65-73 Falk, H., Thomas, L. B., Popper, H. et al. (1979b) Hepatic angiosarcoma associated with adiogenic-anabolic leukemia steroids. Lancet 2:1120-1123 Fingerhut, M.A., Halperin, W.E., Honchar, P.A., et al. (1983) Review of exposure and pathology data for seven cases reported as soft tissue sarcoma among persons occupationally exposed to dioxin-contaminated herbicides. Presented at Symposium on Public Health Risks of the Dioxin (The Rockefeller University, New York City, October 19-20, 1983). Greene, M. H., Hoover, R. N. and Fraumeni, J. P., Jr. (1978) Subsequent cancer in patients with chronic lymphocytic leukemia - a possible immunologic mechanism. JNCI 61:337-340 Honchar, P. A. and Halperin, W. E. (1981) 2,4,5-T, trichlorphenol and soft tissue sarcoma. Lancet 1:268-269 Hardell, L. and Sandstrom, A. (1979) Case-control study: Soft tissue sarcomas and exposure to phenoxyacetic acides or chlorophenolsa. Br. J. Cancer 39:711-717 Hardell, L., Eriksson, M., Lenner, P. and Lundgren, E. (1981) Malignant lymphomo and exposure to chemicals, especially organic solvents, chlorophenols and phenoxy acids: A case-control study. Br. J. Cancer 43:169-176 Hardell, L. (1981) Relation of soft-tissue sarcoma, malignant lymphoma and colon cancer to phenoxy acids, chlorophenols and other agents. Scand. J. Work Environ. Health 7:119-130 Hardell, L. and Eriksson, M. (1981) Soft-tissue sarcoma, phenoxy herbicides and chlorinated phenols. Lancet 2:250 18. �References Holford, T. R., White, C. and Kelsey, J. (1978) Multivariate analysis foe matched case-control studies. American J. Epidemiol. 107:245-256 Hoover, R. and Fraumeni, J. P., Jr. (1973) Risk of cancer in renal transplant recipients. Lancet 2:55-57 Johnson/ F. E., Kugler, M. A. and Brown, S. M. (1981) Soft tissue sarcoma and chlorinated phenols. Lancet 2:40 Kim, J. H., Chu, F. C., Wbodward, H. 0 , et al. (1978) Radiation-induced soft . tissue and bone sarcoma. Radio!. 129:501-508 Kinlen, L. J., Sheid, A. G. R., Peto, J. et al. (1979) Collaborative United Kingdom - Australian study of cancer in patients treated with ijnmuno-suppressive drugs. Br. Med. J. 2:1461-1466 Kociba, R. J., Keyes, D. G., Beyer, J. E. et al. (1978) Results of a two year chronic toxicity and oncogenicity study of 2,3,7,8-tetrachlorodibenzo-p-dioxin in rats. Ttoxicpl. Appl. Pharmacol. 46:279-303 McKillop, J. H., Doig, J. A., Kennedy, J. S., et al. (1978) Laryngeal malignancy following iodine - 125 therapy for thyrotoxicosis. Lancet 2:1177-1179 Moses, M. and Selikoff, I. J. (1981) Soft tissue sarcomas, phenoxy herbicides and chlorinated phenols. Lancet 1:1370 National Cancer Institute (1980) DHHS Publication Number NIH 80-1765 Percy, C., Stanek, E., and Gloeckler, L. (1981) Accuracy of cancer death certificates and its effect on cancer mortality statistics. Am. J. Public Health. 71:242-250. Poland, A. abd Knutson, J. (1982) 2,3,7,8-tetrachlorodibenzo-p-dioxin and related halogenated hydrocarbons. Examination of the mechanism of toxicity. American Rev. Pharmacol. Tbxicol. 22:517-554 Riihimaki, V. (1982) Mortality of 2,4-dichlorophenoxyacetic acid and 2,4,5trichlorophenoxyacetic acid herbicide applicators in Finland. Scand. J. Work Environ. Health 8:37-42 Smith, A. H., Fischer, D. O., Pearce, N. and Teague, C. A. (1982) Do agricultural chemicals cause soft tissue sarcoma? Initial findings of a case-control study in New Zealand. Community Health Studies 6:114-119 Spector, B. D., Perry, G. S. and Kersey, J. H. (1978) Genetically determined immunodeficiency disease (GDIO) and malignancy: Report from the immunodeficiency - cancer registry, din. Immunol. Lnmunopathpl. 11:12-29 Tucker, M. A. and Fraumeni, J. F., Jr.: Soft tissue in Cancer Epidemiology and Prevention. D. Schottenfeld and J. F. Fraumeni, Jr. (eds). Philadelphia W. B. Saunders Co. 1981. pp 827-836 19. �Attachment 1A DISTRIBUTION OF POTENTIAL CASES BY CALENDAR YEAR Calendar Year Age Potential Cases AFEP Total 1975 22-33 81 (10%) 806 1976 23-34 70 (.% 89) 789 1977 24-35 75 (.% 92) 819 1978 25-36 69 (.% 82) 843 1979 26-37 74 (.% 82) 905 1980 27-38 71 (.% 83) 853 440 �Attachment IB HALS SOFT TISSUE SARCOMAS DIAGNOSED BY THE AFIP BY 5-YEAR AGE CATEGORY AND CALENDAR YEAR Age Category Calendar Year 26-30 31-35 36-40 26-40 1975 66 37 54 157 (19.5%) 806 1976 47 37 42 126 (16%) 789 1977 68 37 55 160 (19.5%) 819 1978 42 34 32 108 (12.8%) 843 1979 51 45 39 135 (14.9%) 905 1980 54 36 56 146 (17.1%) 853 328 226 278 832 5,015 Total (-5) 07+ �Attachment 1C DISTRIBUTION OF POTENTIAL CASEs TYPE OF MEDICAL INSTITUTE Hospital Nurber BY Percentage Civilian Hospital 354 80.5 Military Hospital 50 11.3 Army Navy Air Force Veterans Hospital Public Health Service (2 2) (11) (7 1) 35 80 . 1 0.2 440 100.0 �Attachment 2A a. The relative frequency of the risk factor among controls in the target population, Po No combat (0.32) Low combat (0.43) Servi (0.36) Vietnam (+) (0.3) High combat (0.26) Males' Non-service (.4 06) ietnam (-) (0.7) (1) risk factor = Vietnam service/high combat duty Assuming the servicemen in the Vietnam/High combat category were most likely exposed to Agent Orange, the Po was calculated as follows: P0 = 0.36 x 0.3 x 0.26 = 0.029 (2) risk factor = Vietnam service P0 = 0.36 x 0.3 = 0.11 (3) risk factor = Vietnam service/combat (high + low) P0 - 0.36 x 0.3 x (0.43 + 0.26) - 0.07 (4) risk factor » Occupational and non-occupational herbicide exposure NCI assumes Po = 0.1 �Attachment 2B Sample Size and Power with Multiple Controls per Case (Ref: Case Control Studies, James J. Schlesselman, Oxford University Press, 1982, P 150-151) 1/c) P'q' + Z PiQi + P0 - P0)2 where p = (P-j + cPo)/( 1 + c) c = 1 or 2 P! = P0 V[1 +P0 (R- 1)1 P0 = Relative frequency of risk factor among controls in the target population. R = A hypothesized relative risk Ql - 1 - Pi qJ - 1 - p"' �STUDY POWER WITH TWO OCNTRODS PER CASE 200 Triplets R Po 100 Triplets R 2 3 1.5 1.5 0.05 14 34 72 0.10 20 52 0.15 26 63 300 Triplets R 1.5 2 3 400 Triplets R 1.5 2 3 98 38 82 99 92 99 59 97 99 97 99 72 99 99 2 3 27 56 94 30 72 91 35 79 99 48 96 45 89 99 60 alpha » 0 0 .5 R « relative risk P0 » relative frequency of risk factor among controls in the target population H �STODY PCWER WTEH ONE CXKTRQL PER QVSB 100 Pairs R 2 3 200 Pairs R 2 3 300 Pairs R T.5 2 3 400 Pairs R 1.5 2 3 P0 1.5 00 .5 10 23 55 17 42 85 22 56 96 28 69 99 0.10 15 39 80 27 65 98 37 82 99 46 92 99 0.15 20 50 90 31 79 99 48 92 99 59 97 99 alpha =. 0 0 .5 R = relative risk 1.5 �ID No. VA/ftFIP Soft Tissue Sarcoma Study Hello, my name is (interviewer's name]. I am calling on behalf of the Veterans Administration. As you know from the letter you recently received from us, you have been selected to participate in a study of environment and health, sponsored by the VA and the Armed Forces Institute of Pathology. I would like to ask you some questions about your [or (name of study subject)*s] jobs, smoking habits, medical history and so forth. Your participation in this study is voluntary. All of the information collected will be kept completely confidential and neither nanes nor any other identifying information will appear in any report of the study. The interview takes about a half hour to complete. Your participation in this study is most appreciated. Please make yourself comfortable and let us begin. Interviewer Initials �Date Time Started _ am/pn SECTION A - a\CKGRQUND INFORMATION If the respondent is not study subject, list the relationship of respondent to study subject: In the first section of the interview, I will ask you sane questions about your (your ) education, religion, background. | IF THE RESPQLNDgaff IS THE STUDY SUBJECT, GO TO A3 . | A1. How many years did you know your ? * ot years A2. Approximately how many days per month did you talk or visit with your during most of his adult life? Days/Month A3. What is your ( 's) date of birth? Month/Day/Year A4. Vhat city, county and state or foreign country were you (was your ) born? ,City/State/County or Foreign Country A5. How many years of schooling did you (your (Do not read categories to respondent.) Less than 8 years 8 through 11 years 12 years or completed high school Post high school training other than college (e.g., vocational or technical training) Some college College graduate Postgraduate Other (Specify) ) complete? 01 02 03 04 05 06 07 08 �SECTION A - 3ACKGKOUNO A6. In what religion were you (was your (Do not read categories to respondent.) None Catholic Jewish Latter Day Saints (Mormon) Protestant Other (Specify) ) raised? 01 02 03 04 05 06 Most people in the United States have ancestors who came frcra other part of the world. Some have mixed ethnic backgrounds. A7. What is your racial background? Are you White, Black, Hispanic, Asian or Pacific Islander or American Indian or Alaskan native (Circle all that apply). White Black Hispanic Asian or Pacific Islander American Indian or Alaskan native Other 1 2 3 4 5 6 A8. What is your ( 's) father's ethnic background? A9. What is your ( 's) mother's ethnic background? RECORD BELOW. IF MDRE THAN OJE ETHNICITY IS GIVEN FOR THE FATHER OR MOTHER, RECORD ALL THAT APPL I Father's English, Scotch, Welsh French German Greek Irish Italian 01 02 03 04 05 06 Spanish, Portugese Other European 07 08 Czechoslovakian Russian Other Eastern European (Polish, 09 10 Lithuanian, etc.) 11 �SECTION A - BACKGROUND INFORMATION Father's Scandinavian (Norwegian, Danish, Finnish, Swedish) 12 American Indian Central American Mexican Puerto Rico South American West Indian 13 14 15 16 17 18 Chinese Indian, Pakistani Japanese Other Asian Countries or Pacific Islanders 19 20 21 African Middle Eastern 23 24 Other (Specify) 25 Unknown 26 22 Mother's English, Scotch, Welsh French German Greek Irish Italian Spanish, Portugese Other European 01 02 03 04 05 06 07 08 Czechoslovakian 09 Russian 10 Other Eastern European (Polish, Lithuanian, etc.) 11 Scandinavian (Norwegian, Danish, Finnish Swedish) 12 American Indian Central American Mexican Puerto Rico South American West Indian 13 14 15 16 17 18 �SECTION A - BACKGROUND INFORMATION Mother's Chinese Indian, Pakistani Japanese Other Asian Countries or Pacific Islanders 19 20 21 African Middle Eastern Afro-American 23 24 25 Other (Specify) 26 Unknown 27 22 A10. How tall are you (was your Feet/Inches All. Before 19 , what was your ( •s) usual adult weight? Lbs. A12. In 19 , were you (was your divorced, separated or never married? ) married, widowed, Married Widowed Divorced Separated Never married 1 2 3 4 5 Not sure 6 �SECTION B - MILITARY HISTORY Now I am interested in whether you (your ) ever served in the U.S. military. B1. Did you (your Navy, or Air Force? ) serve in the U.S. military like the Army, Yes No 1 2 Not Sure 3 | IF NO OR NOT SURE 00 TO B10.| 82. In what years did you (your ) serve in the military? From 83. hhich branch did you (your (Read) Army Navy Air Force B4. Did you (your To ) serve in? 1 2 3 Marines Coast Guard National Guard 4 5 6 Reserves 7 Not Sure 8 ) serve in Vietnam? Yes No Not Sure 1 2 3 IF YES, CONTINUE, IF NO OR NOT SURE SKIP TO B . 8| B5. Could you tell me the names of places or areas in Vietnam where you (he) served? Not Sure B6. Do you think you were (he was) exposed to Agent Orange? Yes No Not Sure B7. How were you (was he) exposed? Describe Not Sure 1 2 3 �SECTION B - MILimRY HISTORY 88. When you (he) first entered the military were you (was he) drafted or did you volunteer? Volunteer Draft Not Sure 1 2 3 89. There is no requirement that you provide us your social security number or any other information for that matter. But we could get more information about your (his) troop movements from the military if we had your (his) serial number or social security number. Wbuld you mind giving us them? SSN Serial Number Not Sure BIO. Were you in Vietnam for some reason other than military service? Yes 1 No Not Sure 2 3 | IF YES, CONTINUE; IF NO OR NOT SURE SKIP TO SECTION C.| Bl 1. could you tell me the names of places or areas in Vietnam where you (he) worked? Not Sure B12. Do you think you were (he was) exposed to Agent Orange? Yes No Not Sure 313. How were you (was he) exposed? Describe Not Sure 1 2 3 �SECTION C - OCCUPATIONAL HISTORY How, I am interested in your (his) occupational history. C1. First, what was your ( 's) usual occupation during roost of your (his) adult life, that is, the job you (he) held the longest? Usual Occupation IF STUDY SUBJECT NEVER WORKED, CHECK HERE AND GO TOC24J C2. In what year did you (your occupation)? ) start working as a (usual Year C3. In what year did you (your occupation)? ) stop working as a (usual Year C4. What were your (his) activities or duties? Activities/Duties C5. For what kind of company did you (he) work, that is, what did they make or do? Type of Company C6. Vvhat is the name and location of this company? Name and location �SECTION C - OCCUPATIONAL HISTOHY Q. C7-C20. IF RESPONDENT ANSWERS YES TO BART A, ASK &-D, I.E., GO ACROSS HOWS BEFORE GOING DOWN THE COLUMNS a. Did you (you ) ever work. .. b. What is the name and location of the company or enployer you (your worked for? c. In what year did you (your first work... d. In what year did you (your ) last work... C7 . . .mixing or Yes 1 formulating pes No 2 ticides Year Year C8... treat ing Yes 1 seeds with fung No 2 icides Year Year Year Year C9...on a high- Yes 1 way, railroad, No 2 utility, or right-of-way maintenance crew? CIO.. as a gard- Yes 1 ner, a landNo 2 scaper, florist, or some other horticultural occupation? C1 1 . .at a non- Yes 1 farm job apply- No 2 ing pesticides, insecticides, herbicide or fungicides? C12..as a veterinarian? Yes 1 No 2 Year Year Year Year Year Year �SECTION C - OCOJlSVnCNAL HISTORY a. Did you (youi ) ever work. . . C13..in the Yes 1 chemical inNo 2 dustry, for example, manufacturing drugs, 01 chemicals (othei than pesticides] C14..in a sawrmill? Yes 1 No 2 b. What is the name and location of the company or employer you (your worked for? c. In what year did you (your first work... d. In what year did you (your ) last work... Year Year Year Year Year Year CIS.. in a wood- Yes 1 working occupa- No 2 tion, for example, furniture or cabinet making? C16..in the con- Yes 1 struction inNo 2 dustry, for example, as a builder, paintet or carpenter? Year Year C 17. .machining Yes 1 metal or refin- No 2 ing metal? Year Year C18..in a job with exposure to radiation Yes 1 No 2 Year Year C19..at an incinerator? Yes 1 No 2 Year Year Year Year Year Year C20..in manufac- Yes 1 turing or reNo 2 pairing electrical transformers and capacitors? C21..in lumber- Yes 1 ing, logging, or No 2 forestry? �SECTION C - OCCUPATIONAL HISTORY C22. While working at the various jobs did you (your contact with any of the following substances: Yes No ) cone in When Asbestos Arsenic compounds Defoliants or herbicides Insecticides or pesticides Degreasing chemicals Vinyl chloride X-ray or nuclear radiation C23. Have you ever been exposed for a month or more to a job or work area which you think may have been harmful to your health (excluding accidents)? Yes No Not sure 1 2 3 If yes, Job/Occupation When and for how many months? Industry What do you think the harmful substance was? C24. Did you (your ) ever work or live on farmland? Yes No Not sure | IF NO OR NOT SURE GO TO C30.1 1 2 3 �SBCTIGW C - OCCUPATIONAL HISTORY C25. When did you (he) work or live on farmland? From C26. Tto Vfere herbicides, that is, weed killers or defoliants, ever used? Yes No Not Sure 1 2 3 | IF MO OR NOT SURE, SKIP TO C 0 | 3. C27. What are the names of the herbicides that were used? C28. About how many days per year were you (was your ) usually exposed to herbicides and how many years were you (was your _ ) exposed to herbicides. Days/Year Total Year C29. Did you (your ) use any protective equipment when mixing or applying the herbicides, such as rubber gloves, masks, etc.? Yes No Not Sure 1 2 3 C30. Before 19 , did you (your ) ever use herbicides, that is, weed killers or defoliants, at home, in yard work, for gardening, or for other purposes not previously discussed? Yes No Not Sure I IF NO OR NOT SURE, GO TO SECTION D.| 1 2 3 �SECTION C - OCCUPATIONAL HISTORY C31. What were the names of the herbicides or weed killers you (your ) used? Herbicides Not Sure C32. For how many years did you (your ) apply herbicides? Years C33. Have you (has he) been engaged in hobbies that involve the use of chemicals? Yes No Not Sure C34. What was (were) the hobby (hobbies)? C35. What chemicals did you (he) use? Chemical Name C36. When did you first engage in that hobby? Month/Year C37. For how many years did you have that hobby? Years 1 2 3 �SECTION D - MEDICAL HIS'LOKY Now I would like to ask you some questions about your ( medical history. 's) D1. I am interested in medications and other medical treatments that you (your ) may have taken. Before~T9 , did you (your ) ever: Yes a. Have your (his) tonsils removed? b. Receive radiation (for example, cobalt treatment, radioisotopes) as part of a medical treatment? (Emphasize "Before 19 ".) i. To what part of the body did you (your ) receive radiation treatment? No 1 Not Sure 2 Year 3 1 2 3 1 2 3 Part(s) of Body c. Take any cholesterol-lowering drugs, for example, clofibrate? d. Take any medications for seizures or epilepsy? Was it: Dilantin Phenobartital Mesantoin Hydantoin 1 2 3 1 2 3 1 1 1 1 2 2 2 2 3 3 3 3 e. Take the drug chloramphenicol? 1 f. Receive blood transfusion? 1 g. Receive iron dextran, shots for anemia 1 i. How many times? ii. In which part of the body were the shots usually given? 2 2 2 3 3 3 h. Receive immunosuppresive therapy? i. Receive a drug to prevent from getting malaria? Was it: Dapsone Chlorcquine Other (Specify color) 1 2 3 1 2 3 1 1 2 2 3 3 1 2 3 j. Apply a tar ointment to your (his) skin. D2. Before 19 , did a doctor ever tell you (your had (Disease)? a. Chicken Pox b. Diabetes or sugar in your urine 1 1 ) that you (he) 2 2 3 3 �SECTION D - MEDICAL HISTORY Yes c. Allergies d. Infectious nononucleosis ("mono") e. Eczema f. Chloracne (a skin eruption resulting from a chemical exposure, not teenage acne) g. Heart disease h. Hypertension, high blood pressure i. Cancer (Specify Type/Site) j. Hepatitis, Jaundice, cirrhosis, or other liver disease k. Kidney stones or other urinary problems 1. Systemic. lupus erythematosus, SLE m. Celiac disease, nontropical sprue n. Neurofibromatosis, "Elephant man" disease o. Gardner's syndrome, familial polyps in colon p. Hemophilia q. Rheumatoid arthritis r. Other (1) (2) <3> Z Z Z Z D3. Before 19 , did you (your injuries from accidents? No 1 1 1 2 2 2 3 3 3 " 1 2 2 3 3 " 1 1 2 2 3 3 1 1 1 1 2 2 2 2 3 3 3 3 1 2 3 1 1 1 2 2 2 3 3 3 1 Not Sure Year ) ever have any serious 1 2 3 IF NO OR NOT SURE GO TO D7 . | D4*. What part of your ( 's) body was injured? D5. What type of injury D6. In what year were did you (your ) (was your ) have? injured? a• b. c. a• b. c. a• b. c. *SPECIEY RIGHT OR LEFT, IF APPLICABLE D7. Has anyone in your ( *s) immsdiate family, that is, your (his) mother, father, brothers, sisters, or children, ever had cancer of a n y kind? 1 2 3 | IF NO OR NOT SURE GO TO SECTION E.| ' �SECTION D - MEDICAL HISTORY D8. Please tell me who your ( 's) relative was wno had cancer. D9. What was the kind of cancer? D10. At what age was (disease) first diagnosed? Relative Initial location Age Relative Relative Initial location Initial location Age Age Relative Relative Relative Initial location Initial location Initial location Age Age Age �SECTION E - SMOKINS AND BEVERAGE HISTORY Now, I would like to ask you sorne questions about your (his) smoking history. El. Before 19 , were your (was he) a cigarette smoker? Yes 1 No 2 Not Sure 3 IF NO OR NOT SURE GO TO SECTION B6J E2. Did you (he) usually smoke filter or non-filter cigarettes? Filter Non-filter Not Sure 1 2 3 E3. How old were you (was he) when you (he) first regularly smoked cigarette? Age E4. For how many years did you (he) smoke cigarettes? Years E5. Before 19 , about how many cigarettes did you (he) usually smoke per day? Cigarettes/day E6. Before 19 , did you (your pipe for six months or longer? ) ever smoke cigars or a Yes No Not Sure E7. Before 19 , did you (your for six months or longer? 1 2 3 ) ever chew tabocco or use snuff Yes No Not Sure 1 2 3 Now I would like to ask you some questions about beverages you (your may have drunk. ) �SECTION E - SMOKING AND BEVERAGE HISTORY EC. Before 19 , did you (your ) regularly drink coffee? Yes No Not Sure 1 2 3 | IF NO OR NOT SURE GO TO E13.| E9. How old were you (was your regularly drank coffee? ) when you (he) first Age E10. How many years did you (your ) drink coffee? Years Ell. Did you (your ) usually drink decaffeinated coffee (i.e., Sanka, Brim, etc.) or regular coffee? Decaffeinated Regular Not Sure 1 2 3 E12. Before 19 , how many cups of coffee did you (he) usually drink per day? Cups/Day El3. Did you (he) drink alcoholic beverages sometime? Yes No Not Sure 1 2 3 | IF NO OR NOT SORB GO TO SECTION F | . E14. Before 19 , how many 4 ounce glasses of wine did you (your_ usually drink in a week? Glasses/Week E15. Before 19 , how many 1 1/2 ounce glasses of whiskey or hard liquor did you (your ' ) usually drink in a week? Glasses/Week E16. Before 19 , how many 12 ounces glasses or cans of beer, ale, or similar drinks did you (your ) usually drink in a week? Gl asses/week �SECTION F - HEALTH HABITS/SOCIAL FACTORS PI. Before 19 , did you (your _) have a regular physician? Yes , No Not sure 1 2 3 F2. Before 19 , how many hours of sleep did you (he) usually get at night? 6 hours or less 7 hours or more Not sure 1 2 3 F3. Before 19 , did you (he) take any of these vitamins? For how long? Vitamins No Yes, every- Yes, some- For how day times many years Amount Not Sure Multiple vitamins (One-a-day type) Vitamin A Vitamin C Vitamin E B complex Cod liver oil Nutritional yeast Other (Specify) F4. Before 19 , how many close friends did you (he) have? (People that you feel at ease with, can talk to about private matters, and can ask for help.) None 1 or 2 3 or more Not Sure 1 2 3 4 �SECTION F - HEALTH HABITS/SOCIAL FACTORS P5. Before 19 , how many relatives did you (he) have that you (he) feel close to? None 1 or 2 3 or more Not sure 1 2 3 4 SECTION G -CONCLUSION That concludes the interview. Thank you very much for your participation in this study. Time ended am/pea INTERVIEWER REMARKS G1. Respondent's cooperation was: Very good 1 Good 2 Fair Poor 3 4 G2. The respondent's: Did not know enough information regarding the topic Did not want to be more specific Did not understand or speak English well Was bored or uninterested Was upset or depressed Was physically ill Had poor hearing or speech Was confused by frequent interruptions Was emotionally unstable Other (Specify) 01 02 03 04 05 06 07 08 09 10 � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Alvin L. Young Collection on Agent Orange Description An account of the resource <p style="margin-top: -1em; line-height: 1.2em;">The Alvin L. Young Collection on Agent Orange comprises 120 linear feet and spans the late 1800s to 2005; however, the bulk of the coverage is from the 1960s to the 1980s and there are many undated items. The collection was donated to Special Collections of the National Agricultural Library in 1985 by Dr. Alvin L. Young (1942- ). Dr. Young developed the collection as he conducted extensive research on the military defoliant Agent Orange. The collection is in good condition and includes letters, memoranda, books, reports, press releases, journal and newspaper clippings, field logs and notebooks, newsletters, maps, booklets and pamphlets, photographs, memorabilia, and audiotapes of an interview with Dr. Young.</p> <p>For more about this collection, <a href="/exhibits/speccoll/exhibits/show/alvin-l--young-collection-on-a">view the Agent Orange Exhibit.</a></p> Text A resource consisting primarily of words for reading. Examples include books, letters, dissertations, poems, newspapers, articles, archives of mailing lists. Note that facsimiles or images of texts are still of the genre Text. Box The box containing the original item. 066 Folder The folder containing the original item. 1762 Series The series number of the original item. Series III Subseries III Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Creator An entity primarily responsible for making the resource Kang, Han K. Franz M. Enziger Lawrence B. Hobson Barclay M. Shepard Patricia P. Breslin Description An account of the resource <strong>Corporate Author: </strong>Agent Orange Projects Office, Department of Medicine and Surgery, VA, and Department of Soft Tissue Pathology, AFIP Title A name given to the resource Typescript: Research Protocol, a Matched Case-Control Study of Soft Tissue Sarcoma, March 1984 Subject The topic of the resource VA/AFIP STSS study protocol ao_seriesIII https://www.nal.usda.gov/exhibits/speccoll/files/original/9c9f23c26722d230cc07a7c9a8547588.pdf 4624d4d4dae9a614631f333ac701ae0f PDF Text Text Item ID Number 01755 Author Kan 9-Han K- Corporate Author RopOrt/ArtlOlO TltlO Typescript: Soft Tissue Sarcoma and Military Service in Vietnam: a Case Comparison Group Analysis of Hospital Patients, February 1985 Journal/Book Title Year 000 ° Month/Day Color n Number of Images 14 DoSCrlptOH NotOS Submitted for publication in the American Journal of Industrial Medicine. Monday, June 11, 2001 Page 1766 of 1793 �Soft Tissue Sarcoma and Military Service in Vietnam: A Case Comparison Group Analysis of Hospital Patients Han K. Kang, Dr.P.H., Lee Weatherbee, M.D.*, Yvonne Lee, M.S., Patricia P. Breslin, Ph.D., Barclay M. Shepard, M.D. Department of Medicine and Surgery Veterans Administration Washington, D.C. 20420 (202) 389-5534 * Laboratory Service, Veterans Administration Medical Center 2215 Fuller Road, Ann Arbor, Michigan 48105 Submitted for publication in the American Journal of Industrial Medicine. February 1985. �Abstract - Soft tissue4 sarcoma and military service in Vietnam: A case comparison group analysis of hospital patients. The possibility that exposure to Agent Orange or phenoxy herbicides may have increased the risk of soft tissue sarcoma (STS) has been of genuine concern to Vietnam veterans and their families. A hospital-based case comparison group study was undertaken to examine, through a comprehensive review of medical records and military personnel records, the association between previous military service in Vietnam and soft tissue sarcoma. The case group comprised 234 Vietnam-era veteran patients who served in the U.S. military between 1964 and 1975 and were treated in one of the 172 VA hospitals between 1969 and 1983 with a histologically confirmed diagnosis of soft tissue sarcoma. The comparison group consisted of 13,496 patients who were systematically sampled from the same Vietnam-era veteran patient population from which the cases were drawn. Military service information, in particular Vietnam service status, for each case and control patient was obtained from a review of the patient's military personnel records archived at the National Personnel Records Center in St. Louis, Missouri. No significant association of soft tissue sarcoma and previous military service in Vietnam was observed: odds ratio was 0.83 with a 95% confidence interval of 0.63-1.09. Key word: Agent Orange, dioxin, phenoxy herbicides, TCDD, Vietnam veterans, and Vietnam era veterans. �Introduction Two Swedish case-control studies have suggested that persons reporting exposure to phenoxy herbicides have a 5 to 6 fold higher risk of developing soft tissue sarcoma compared to persons without such exposure (Hardell et al., 1979; Erikson et al., 1981). In addition, several cases of soft tissue sarcoma have been reported in the U.S. among workers involved in the manufacturing or use of phenoxy herbicides (Cook, 1981; Honchar and Halperin, 1981; Moses and Selikoff, 1981). These studies have generated much concern in the United States for Vietnam veterans that, as a result of their exposure to Agent Orange in Vietnam, they may be at increased risk for soft tissue sarcoma (STS) in addition to several other medical and psychological problems. Agent Orange, a mixture of two commercial phenoxy herbicides 2,4-dichlorophenoxyacetic acid (2,4-D) and 2,4,5-trichlorophenoxyacetic acid (2,4,5-T), was the herbicide most commonly used by the U.S. military in Vietnam. The principle concern over exposure to Agent Orange stems from the fact that during the manufacture of 2,4,5-T trace amounts of a highly toxic dioxin, 2,3,7,8-tetrachlorodibenzo-para-dioxin (TCDD), appeared as a contaminant. During the five-year period from 1965 to 1970, the U.S. Air Force sprayed more than 11 million gallons of Agent Orange in South Vietnam. Approximately two million U.S. military personnel served one-year tours during the same period. Studies published subsequent to the Swedish studies have not yet demonstrated the association between STS and either exposure to phenoxy herbicides or military service in Vietnam (Riihimaki, 1982; Greenwald et �al., 1984; Smith, et al., 1982, 1984; U.S. Air Force, 1983; University of Sidney, 1984). Two of the 7 industrial workers previously reported to be cases of SI'S were also found to have not sarcoma but carcinomas (Fingerhut et al., 1984). In view of the public concern about potential health risk among Vietnam veterans and conflicting research findings in the scientific literature, a case comparison group analysis of hospital patients for soft tissue sarcoma was undertaken to determine the association between previous military service in Vietnam and soft tissue sarcoma. Materials and Methods 'Ihe Veterans Administration Patient Treatment File (PTF) was used to identify all Vietnam era veterans who were diagnosed as having soft tissue sarcoma from 1969 through 1983. Ihe PTF is a computerized hospital data base of in-patient records including patients' demographic data, surgical and procedural transactions, and patient movement and diagnoses. A record is created for each in-patient discharged from one of the 172 VA medical centers. The Vietnam era veterans are defined as veterans who served in the U.S. military sometime during August 5, 1964 and May 7, 1975. A total of 418 cases with International Classification of Disease (ICD) 171 diagnosis, i.e., malignant neoplasm of connective and other soft tissue, were identified by computer search of the FIF for Vietnam era . veterans who were hospitalized between 1969 and 1983. A pathology report for each ICD 171 case was requested from each treating VA medical center. A review of 394 pathology reports received for these cases was made by a pathologist (L.W.) who has particular interest and experience in this �group of malignancies. During the review he had no knowledge of Vietnam service status of any of the cases. On the basis of the review of the pathology reports, 151 ICD 171 cases were excluded as not likely being soft tissue sarcoma because of miscoding or misclassification and 9 ICD 171 cases were put in a doutbful STS category, leaving 234 histologically confirmed diagnoses of STS. All diagnoses were classified according to the WHO classification system for soft tissue sarcomas (Enzinger et al., 1969). The comparison group consisted of 14,931 patients who were systematically sampled from the same Vietnam era veterans patient population from which the STS cases were identified. Vietnam era veteran patients who have predetermined numbers in the last two digits of their social security numbers were selected among all Vietnam era veteran patients. Military service information, in particular Vietnam service status, for STS cases and control patient was obtained from a comprehensive review of the patient's military personnel records archived at the National Personnel Records Center (NPRC) in St. Louis, Missouri. The General Services Administration (GSA) under an agreement with the Department of Defense maintains the military personnel records of veterans including those from the Vietnam era. Military personnel records were located and abstracted for all of the 234 soft tissue sarcoma cases and 13,496 of the 14,931 (90%) control patients. �Results and Discussion Eighty-six of the 234 histoloqically confirmed soft tissue sarcoma cases, or 36.8%, had served in Vietnam. As Table 1 indicates there was no one predominant type of soft tissue sarcoma. Distribution of tumor type of the 234 STS cases was similar to the results from the recently published NY state study of 281 cases of soft tissue sarcoma and Vietnam service (Greenwald et al., 1984). Greenwald et al., reported that percentage distribution of malignant tumor of muscle tissue, fibrous tissue, adipose tissue, and other soft tissue were 23.8, 17.8, 16.4 and 42.0 respectively among the men with soft tissue sarcoma diagnosed from 1962 through 1980, who were between the ages of 18 and 29 years anytime between 1962 and 1971 and in the New York State Cancer Registry. Age distribution of STS cases was similar to the control group. No unusual influx of STS cases was observed at any interval as indicated by percent distribution of STS cases and control groups by hospitalization year (Table 2). Of the sample of 13,469 PTF Vietnam era patients, 5,544 or 41% had served in Vietnam (Table 3). No significant association of soft tissue sarcoma and previous military service in Vietnam was observed among the Vietnam era veterans who come to the VA hospital for inpatient medical care. The odds ratio was 0.83 with a 95% confidence interval of 0.63-1.09. This suggests that the chance of having diagnosis of STS among Vietnam veteran patients was not greater than that among veteran patients who did not serve in Vietnam. A differential ascertainment of military service status between the STS cases (100%) and the control patients (90%) should be noted. We believe, however, that the difference is primarily a reflection of levels of efforts and man-hours allocated for the personnel record search rather than any difference in availability of the military records between STS �cases and control patients, or Vietnam veterans and non-Vietnam veterans. For example, when the same levels of record search efforts as for the control patients were employed for the STS cases, the military record searchers at the NPRC were able to locate 214 of the 234 STS cases (91%); the yield for the control patients was 13,496 of the 14,931 (90%). Additional time-consuming manual tracking efforts were made for the 20 STS cases whose Vietnam service status was not determined because their personnel folder were misplaced, missing or on loan to other agencies. Of the 20 STS cases, 12 did not serve in Vietnam and 8 did serve in Vietnam: a ratio of 5:4. Prior to this exhaustive manual search the ratio among the 214 STS cases was 5:3. Even if one makes an extreme assumption, that is, that all of the remaining 10% of the control patients (1,435), whose military personnel records were not located, did not serve in Vietnam, the conclusion of the study would not be altered. This assumption results in the odds ratio of 0.98. The other extreme assumption, that is, that all of the 1,435 patients had served in Vietnam, results in the odds ratio of 0.66. There seems to be no propensity of ground troops (Army or Marines) among the STS cases as compared to the comparison group (Table 4). It has been suggested that ground troops, by nature of their military operation through defoliated zones and by practice of base perimeter spraying, might have a higher probability of direct or indirect contact with Agent Orange than Air Force or Navy personnel. The findings of this study are consistent with a case control study recently published by Greenwald et al. (1984). Greenwald et al. (1984) reported no significant association between STS among Vietnam era veteran age males and military service in Vietnam. �Other studies of Vietnam era veterans published to date also have failed to find an excess of STS among Vietnam veterans. A study of PANCH HAND personnel, a group of approximately 1,260 men who conducted the fixed wing aerial herbicide spraying missions in Vietnam from 1962 through 1971, did not reveal a single death from STS (USAF, 1983). A proportionate mortality analysis of deaths among New York State Vietnam era veterans between 1965 and 1980, exclusive of 1968 and 1969, also failed to show excess STS deaths among Vietnam veterans. Two of the 555 deaths reported among Vietnam veterans were due to cancer of connective and soft tissue (ICD 171), whereas 3 of 941 deaths among non-Vietnam veterans resulted from the same type of cancer. The mortality odds ratio (MOR) was 1.09 with a 95% confidence interval of 0.08-15.09 (Lawrence et al., 1985). A mortality study of Australian Vietnam era veterans reported 260 deaths among 19,205 Vietnam veterans and 263 deaths among 24,677 non-Vietnam veterans when followed from the end of their military service to January 1, 1982. There was no statistically significant difference in the death rates from STS (University of Sidney, 1984). However, in all three mortality studies, it should be recognized that the design of the study is such that only very high risks for STS were likely to be detected: the number of person years followed or number of deaths available for analysis was too small to detect moderately elevated relative risks of STS from Vietnam service. The absence of positive association between STS and Vietnam service might be a result of insufficient observation time since Agent Orange exposure in Vietnam. In general, it takes more than a decade for cancer to manifest itself if it is induced by a chemical carcinogen. Over 80% of STS cases in the study were observed less than 10 years after the last troops were exposed to Agent Orange in Vietnam. Another possibility is that although Agent Orange or dioxin can induce STS, Vietnam veterans as a 8 �group, were exposed to such small amounts that the conventional epidemiologic study cannot detect the excess risk resulting from Agent Orange exposure in Vietnam. Or, of course, there is the possibility that Agent Orange does not induce STS in humans after all. In conclusion, a study of STS cases and a comparison patients group in VA hospitals did not reveal a statistically significant positive association between STS and previous military service in Vietnam. \ �References 1. Cook, R. R. (1981). Dioxin, chloracne, and soft tissue sarcoma (letter). Lancet 1:618-619. 2. Eriksson, M., Hardell, L., Berg, N. 0., Moller, T., Axelson, 0. (1981). Soft tissue sarcomas and exposure to chemical substances: A case-reference study. Br. J. Ind. jted^ 38:27-33. 3. Fingerhut, M. A., Halperin, W. E., Honchar, P. A., Smith, A. b., Groth, D. H., Russell, W. 0. (1984). Iteview of exposure and pathology data for seven cases reported as soft tissue sarcoma among persons occupationally exposed to dioxin-contaminated herbicides. In Lowrance WW (ed): "Public Health Risks of the Dioxin." The Rockefeller University, pp 187-216. 4. Greenwald, P., Kovasznay, B., Collins, D. N., Iherriault, G. (1984). Sarcomas of soft tissue after Vietnam service. JNCI 73:1107-1109. 5. Hardell, L., Sandstrom, A. (1979). Case-control study: Soft tissue sarcoma and exposure to phenoxyacetic acids or chlorophenols. Br. J. Cancer 39:711-717. 6. Honchar, P. A., Halperin, W. E. (1981). 2,4,5-T, trichlorophenol, and soft tissue sarcoma. Lancet 1:268-269. 7. Lawrence, C., Iteilly, A. A., Quickenton, P., Greenwald, P., Page, W., Kuntz, A. (1985). Mortality patterns of New York State Vietnam veterans. Am. J. Public Health (in print). 8. Moses, M., Selikoff, I. J. (1981). Soft tissue sarcomas, phenoxy herbicides and chlorinated phenols. Lancet 1:1370. 9. Riihimaki, V., Sisko, A., Hernberg, S. (1982). Mortality of 2,4-dichlorophenoxyacetic acid and 2,4,5-trichloroacetic acid herbicide applicators in Finland. Scand. J. Work Environ. Health 8:37-42. 10. Smith, A. H., Fisher, D. 0., Pearce, N., Teague, C. A. (1982). Do agriculture chemicals cause soft tissue sarcoma? Initial findings of a case-control study in New Zealand. Community Health Stud. 6:114-119. 11. Smith, A. H., Pearce, N. E., Fisher, D. 0., Giles, H. J., Teague, C. A., Howard, J. K. (1984). Soft tissue sarcoma and exposure to phenoxy herbicides and chlorophenols in New Zealand. JNCI 73:1111-1117. 12. The Commonwealth Institute of Health, University of Sydney (1984). A retrospective cohort study of mortality among Australian national servicemen of the Vietnam conflict era. Australian Government Printing Service, Canberra. 13. USAF School of Aerospace Medicine, An epidemiologic investigation of health effects in Air Force personnel following exposure to herbicide. Baseline mortality study results. Epidemiology Division, USAF School of Aerospace Medicine, Brooks AFB, Texas. June 1983. 10 �Table 1 Soft Tissue Sarcoma Type By Military Service Status NDn-Vietnam Veteran Vietnam Veteran Phabdomyosarcomas 18 8 26 Le iomyosar comas J8 V2 2£ 26 20 46 (19.7) Type Histology Tumors of muscle tissue Total (%) Tumors of fibrous tissue Fibrosarcoma 26 13 39 (16.7) Tumors of synovia! tissue Synovial sarcoma 21 9 30 (12.8) Tumors of adipose tissue Liposarcoma 19 9 28 (12.0) Tumors of vascular origin Angiosarcoma 3 1 4 _1£ 2 11 13 3 16 (6.8) 43 32 75 (32.0) Malignant hemangiopericytomas Others Total (%) 148 (63.2) 86 (36.8) 234 (100) �Table 2 Distribution by Age and Hospital Discharge Year for STS Case and Comparison Group Category STS Cases Percentage Comparison Group Age group, years 20 - 29 9 6 30 - 34 18 29 35 - 39 42 37 40 - 44 11 11 45 - 49 4 4 50 - 59 10 8 6 5 6 7 1971 - 75 35 36 1976 - 80 42 41 1981 - 83 17 16 60+ Hospital ization, year Before 1970 �Table 3 Distribution of STS Cases and a Comparison Group of Patients by Vietnam Service Status Vietnam service STS Cases (%) Yes 86 (37) 5,544 (41) 5,630 No 148 (63) 7,952 (59) 8,100 234 (100) 13,496 (100) 13,730 Ibtal Comparison Group (%) Odds Ratio: 0.83 (95% confidence interval 0.63-1.09) X2: 1.78 (P>0.1) Ttotal �Table 4 Distribution of STS Cases and a Comparison Group of Patients by Branch of Service in Vietnam Branch STS Cases (%) Comparison Group (%) 45 (52) 3,528 (64) 6 (7) 367 (7) Marines 14 (16) 921 (16) Navy 21 (24) 721 (13) — 86 (100) 7 (*) Army Air Force Coast Guard Total *Less than 1% 5,544 (100) � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Alvin L. Young Collection on Agent Orange Description An account of the resource <p style="margin-top: -1em; line-height: 1.2em;">The Alvin L. Young Collection on Agent Orange comprises 120 linear feet and spans the late 1800s to 2005; however, the bulk of the coverage is from the 1960s to the 1980s and there are many undated items. The collection was donated to Special Collections of the National Agricultural Library in 1985 by Dr. Alvin L. Young (1942- ). Dr. Young developed the collection as he conducted extensive research on the military defoliant Agent Orange. The collection is in good condition and includes letters, memoranda, books, reports, press releases, journal and newspaper clippings, field logs and notebooks, newsletters, maps, booklets and pamphlets, photographs, memorabilia, and audiotapes of an interview with Dr. Young.</p> <p>For more about this collection, <a href="/exhibits/speccoll/exhibits/show/alvin-l--young-collection-on-a">view the Agent Orange Exhibit.</a></p> Text A resource consisting primarily of words for reading. Examples include books, letters, dissertations, poems, newspapers, articles, archives of mailing lists. Note that facsimiles or images of texts are still of the genre Text. Box The box containing the original item. 066 Folder The folder containing the original item. 1765 Series The series number of the original item. Series III Subseries III Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Creator An entity primarily responsible for making the resource Kang, Han K. Lee Weatherbee Yvonne Lee Patricia P. Breslin Barclay M. Shepard Title A name given to the resource Typescript: Soft Tissue Sarcoma and Military Service in Vietnam: a Case Comparison Group Analysis of Hospital Patients, February 1985 Subject The topic of the resource cancer risk assessment dioxin ao_seriesIII https://www.nal.usda.gov/exhibits/speccoll/files/original/873afa4efa90b2c9b94828ca63a23588.pdf dc8a294f981627c6a59dae605d36adfa PDF Text Text Item ID Number 01768 AllthOT Kang, Han K. Corporate Author Roport/Artldo TltlO Soft Tissue Sarcomas and Military Service in Vietnam: a Case Comparison Group Analysis of Hospital Patients JOUrnal/BOOk TltlU Journal of Occupational Medicine Yoar 1986 MOUth/Day December Color a Number of Images 4 Descrlpton Notes Monday, June 11, 2001 Page 1769 of 1793 �Soft Tissue Sarcomas and Military Service in Vietnam: A Case Comparison Group Analysis of Hospital Patients Hem K. Kong, DrPH; Lee Weatherbee, MD; Patricia P. Breslin, PhD; Yvonne Lee, MS; and Barclay M. Shepard, MD The possibility that exposure to Agent Orange or phenoxy herbicides may have increased the risk of soft tissue sarcomas has been of genuine concern to Vietnam veterans and their families. A hospital-based case comparison group study was undertaken to examine, through a comprehensive review of medical records and military personnel records, the association between previous military service in Vietnam and soft tissue sarcomas. The case group comprised 834 Vietnam-era veteran patients who served in the US military between 1964 and 1975 and were treated in one of the 172 VA hospitals between 1969 and 1983 with a diagnosis of soft tissue sarcomas. The comparison group consisted of 13,496 patients who were systematically sampled from the same Vietnam-era veteran patient population from which the cases were drawn. Military service information, in particular Vietnam service status, for each case and control patient was obtained from a review of the patient's military personnel records archived at the National Personnel Records Center in St Louis, Missouri. No significant association of soft tissue sarcomas and previous military service in Vietnam was observed: odds ratio was 0.83 with a 95% confidence interval of 0.63 to 1.09. suggested persons reporting exposure to phenoxy herbicides Twoa Swedishcase-control studies havedevelopingthat have five- to sixfold higher risk of soft tissue sarcomas (STS) compared with persons without such exposure.1'2 In addition, several cases of soft tissue sarcomas have been reported in the US among workers involved in the manufacturing or use of phenoxy herbicides.3-5 Prom the Department of Medicine and Surgery, Veterans Administration, Washington, DC. Address correspondence to: VA Office of Environmental Epidemiology (10X8B), AFIP, Washington, DC 20306-6000 (Dr Rang, Director). 0096-1736/86/S813-1215$03.00/0 Copyright Cc) by American Occupational Modical Association These studies have generated much concern in the United States for Vietnam veterans—concern that, as a result of their exposure to Agent Orange in Vietnam, they may be at increased risk for soft tissue sarcomas in addition to several other medical and psychological problems. Agent Orange, a mixture of two commercial phenoxy herbicides, 3,4-dichlorophenoxyacetic acid (3,4-D) and 2,4,5-trichlorophenoxyacetic acid (2,4,5-T), was the herbicide most comonly used by the US military in Vietnam. The principal concern over exposure to Agent Orange stems from the fact that during the manufacture of 2,4,5-T trace amounts of a highly toxic dioxin, 2,3,7,8-tetrachlorodibenzo-para-dioxin (TCDD), appeared as a contaminant. Agent Orange was sprayed in Vietnam for defoliation and crop destruction from 1965 to 1970 in a military operation named Ranch Hand. The aerial application of Agent Orange reached its peak in 1967, leveled off slightly in 1968 and 1969, and declined rapidly in 1970. During the five-year period the US Air Force sprayed more than 11 million gallons of Agent Orange in South Vietnam. Approximately 2 million US military personnel served 1-year tours during the same period. Studies published subsequent to the Swedish studies have not yet demonstrated the association between soft tissue sarcomas and either exposure to phenoxy herbicides or military service in Vietnam.6-11 Two of the seven industrial workers previously reported to be cases of STS were also found to have not sarcomas but carcinomas.18 In view of the public concern about potential health risk among Vietnam veterans and conflicting research findings in the scientific literature, a case comparison group analysis of hospital patients for soft tissue sarcomas was undertaken to determine the association between previous military service in Vietnam and soft tissue sarcomas. Journal of Occupational Medicine/Volume 28 No. 12/December 1986 1215 �Materials and Methods The Veterans Administration Patient Treatment File (PTF) was used to identify all Vietnam era veterans whose conditions were diagnosed as soft tissue sarcomas from 1969 through 1983. The PTF is a computerized hospital data base of inpatient records, including patients' demographic data, surgical and procedural transactions, and patient movement and diagnoses. A record is created for each inpatient discharged from one of the 172 VA medical centers. The Vietnam-era veterans are defined as veterans who served in the US military sometime during Aug 5,1964, and May 7, 1975. A total of 418 cases with International Classification of Diseases (ICD) 171 diagnosis, ie, malignant neoplasm of connective and other soft tissue, were identified by computer search of the PTF for Vietnam-era veterans who were hospitalized between 1969 and 1983. A pathology report for each ICD 171 case was requested from each treating VA medical center. A review of 394 pathology reports received for these cases was made by a pathologist (L.W.) who has particular interest and experience in this group of malignancies. During the review he had no knowledge of Vietnam service status of any of the patients. On the basis of the review of the pathology reports, 151 ICD 171 cases were excluded as not likely being soft tissue sarcomas because of miscoding or misclassification and nine ICD 171 cases were put in a doubtful STS category, leaving 234 diagnoses of STS. All diagnoses were classified according to the World Health Organization classification system for soft tissue sarcomas.13 The comparison group consisted of 14,931 patients who were systematically sampled from the same Vietnam-era veterans patient population from which the STS case subjects were identified. Vietnam-era veteran patients who have predetermined numbers in the last two digits of their social security numbers were selected among all Vietnam-era veteran patients. Military service information, in particular Vietnam service status, for STS case subjects and control patients was obtained from a comprehensive review of the patient's military personnel records archived at the National Personnel Records Center (NPRC) in St Louis, Missouri. The General Services Administration (GSA), under an agreement with the Department of Defense, maintains the military personnel records of veterans, including those from the Vietnam era. Military personnel records were located and abstracted for all of the 234 STS case subjects and 13,496 of the 14,931 (90%) control patients. Results and Discussion Eighty-six of the 234 STS cases, or 36.8%, had served in Vietnam. As Table 1 indicates there was no one predominant type of STS. Distribution of tumor type of the 234 STS cases was similar to the results from the recently published New York state study of 281 cases of soft tissue sarcoma and Vietnam service.7 Greenwald et al reported that percentage distribution of malignant tumor of muscle tissue, fibrous tissue, adipose tissue, and other soft tissue was 23.8, 17.8, 16.4, and 42.0, respectively, among the men with soft tissue sarcomas diagnosed from 1962 through 1980, who were between the ages of 18 and 29 years any time between 1962 and 1971 and in the New York State Cancer Registry. Age distribution of STS case subjects was similar to the control group. No unusual influx of STS case subjects was observed at any interval as indicated by percent distribution of STS case subjects and control groups by hospitalization year (Table 2). Of the sample of 13,469 PTF Vietnam-era patients, 5,544 or 41% had served in Vietnam (Table 3). No significant association of soft tissue sarcomas and previous military service in Vietnam was observed among the Vietnam-era veterans who had come to the VA hospital for inpatient medical care. The odds ratio was 0.83 with a 95% confidence interval of 0.63 to 1.09. This suggests that the chance of having a diagnosis of STS among Vietnam veteran patients was not greater than that among veteran patients who did not serve in Vietnam. A differential ascertainment of military service status between the STS case subjects (100%) and the control patients (90%) should be noted. However, the difference is primarily a reflection of levels of efforts and manhours allocated for the personnel record search rather than any difference in availability of the military records between STS case subjects and control patients, or Vietnam veterans and non-Vietnam veterans. For ex- TABLE 1 Soft Tissue Sarcoma Type By Military Service Status Type Histology Tumors of muscle tissue Rhabdomyosarcoma Leiomyosarcoma Fibrosarcoma Synovial sarcoma Liposarcoma Angiosarcoma Malignant hemangiopericytoma Tumors of fibrous tissue Tumors of synovial tissue Tumors of adipose tissue Tumors of vascular origin Others Total (%) 1216 NonVietnam Veteran Vietnam Veteran 18 8 26 21 19 3 10 43 8 12 13 9 9 1 2 32 148 (63.2) 86 (36.8) Total (%) 26 20 39(16.7) 30(12.8) 28(12.0) 4 12 75 (32.0) 234 (100) Soft Tissue Sarcomas and Agent Orange/Kang et al �TABLE 2 Distribution by Age and Hospital Discharge Year for Soft Tissue Sarcoma Case Subjects and Comparison Group TABLE 4 Distribution of Soft Tissue Sarcoma Case Subjects and Comparison Group of Patients by Branch of Service in Vietnam Percentage Category STS Case Subjects Comparison Group Age group (yr) 20-29 30-34 35-39 40-44 45-49 50-59 60+ 9 18 42 11 4 10 6 6 29 37 11 4 8 5 Hospitalization (yr discharged) Before 1970 1971-1975 1976-1980 1981-1983 6 35 42 17 7 36 41 16 TABLE 3 Distribution of Soft Tissue Sarcoma Case Subjects and a Comparison Group of Patients by Vietnam Service Status* STS Case Subjects (%) Comparison Group (%) Total Yes No 86 (37) 148(63) 5,544(41) 7,952 (59) 5,630 8,100 Total 234 (100) 13,496 (100) 13,730 Vietnam Service * Odds ratio: 0.83 (95% confidence interval 0.63 to 1 .09); x2: 1 .78 ample, when the same levels of record search efforts made for the control patients were employed for the STS case subjects, the military record searchers at the NPRC were able to locate 214 of the 234 STS cases (91%); the yield for the control patients was 13,496 of the 14,931 (90%). Additional time-consuming manual tracking efforts were made for the 20 STS case subjects whose Vietnam service status was not determined because their personnel folders were misplaced, missing, or on loan to other agencies. Of the 20 STS case subjects, 12 did not serve in Vietnam and eight did serve in Vietnam: a ratio of 5:3.3. Prior to this exhaustive manual search the ratio among the 214 STS case subjects was 5:2.9. Even if one makes an extreme assumption, that is, that all of the remaining 10% of the control patients (1,435) whose military personnel records were not located did not serve in Vietnam, the conclusion of the study would not be altered. This assumption results in the odds ratio of 0.98. The other extreme assumption, that is, that all of the 1,435 control patients had served in Vietnam, results in the odds ratio of 0.66. There seems to be no propensity of ground troops (Army or Marines) among the STS case subjects compared with the comparison group (Table 4). It has been suggested that ground troops in Vietnam, by nature of their military operation through defoliated zones and by practice of base perimeter spraying, might have a higher probability of direct or indirect contact with Agent Orange than Air Force or Navy personnel. Branch STS Case Subjects Comparison Group (%) Army Air Force Marines Navy Coast Guard 45 (52) 6(7) 14(16) 21 (24) 3,528 (64) 367 (7) 921 (16) 721 (13) Total 86 (100) 5,544 (100) The findings of this study are consistent with a case control study recently published by Greenwald et al.7 Greenwald et al reported no significant association between STS among Vietnam-era veteran-age males and military service in Vietnam. Other studies of Vietnam-era veterans published to date also have failed to find an excess of STS among Vietnam veterans. A study of Ranch Hand personnel, a group of approximately 1,260 men who conducted the fixed-wing aerial herbicide spraying missions in Vietnam from 1962 through 1971, did not reveal a single death from STS.10 A proportionate mortality analysis of deaths among New York State Vietnam-era veterans between 1965 and 1980, exclusive of 1968 and 1969, also failed to show excess STS deaths among Vietnam veterans.14 Two of the 555 deaths reported among Vietnam veterans were due to cancer of connective and soft tissue (ICD 171), whereas three of 941 deaths among non-Vietnam veterans resulted from the same type of cancer. The mortality odds ratio (MOR) was 1.09 with a 95% confidence interval of 0.18 to 6.70. A mortality study of Australian Vietnam-era veterans reported 260 deaths among 19,205 Vietnam veterans and 263 deaths among 25,677 non-Vietnam veterans when followed from the end of their military service to Jan 1, 1982. There was no statistically significant difference in the death rates from STS.11 However, in all three mortality studies, it should be recognized that the design of the study is such that only very high risks for STS were likely to be detected: the number of person-years followed or number of deaths available for analysis was too small to detect moderately elevated relative risks of STS from Vietnam service. The absence of positive association between STS and Vietnam service might be a result of insufficient observation time since Agent Orange exposure in Vietnam. In general, it takes more than a decade for cancer to manifest itself if it is induced by a chemical carcinogen. More than 80% of STS case subjects in the study were observed less than 10 years after the last troops were exposed to Agent Orange in Vietnam. Another possibility is that although Agent Orange or dioxin can induce STS, Vietnam veterans as a group were exposed to such small amounts that the conventional epidemiologic study cannot detect the excess risk resulting from Agent Orange exposure in Vietnam. Or there is the possibility that Agent Orange does not induce STS in humans after all. Journal of Occupational Medicine/Volume 28 No. 12/December 1986 1217 �1. Hardell L, Sandstrom A: Case-control study: Soft tissue sarcoma and exposure to phenoxyacetic acids or chlorophenols. Br J Cancer 1979:89:711-717. 2. Eriksson M, Hardoll L, Berg NO, et al: Soft tissue sarcomas and exposure to chemical substances: A case-reference study. BrJInd Med 1981;38:37-33. 3. Cook RR: Dioxin, chloracne, and soft tissue sarcoma, letter. Lancet 1981:1:618-619. 4. Honchar PA, Halperin WE: 2,4,5-T, trichlorophenol, and soft tissue sarcoma. Lancet 1981:1:268-869. 5. Moses M, Selikoff IJ: Soft tissue sarcomas, phenoxy herbicides and chlorinated phenols. Lancet 1981;1:1370. 6. Riihimaki V, Sisko A, Hernberg S: Mortality of 2,4-dichlorophenoxyacetic acid and 2,4,5-trichloroaeetic acid herbicide applicators in Finland. Sound J Work Environ Health 1982;8:37-42. 7. Grecnwald P, Kovasznay B, Collins UN, ot al: Sarcomas of soft tissue after Vietnam service. JNCI 1984:73:1107-1109. 8. Smith AH, Fisher DO, Pearco N, et al: Do agriculture chemicals cause soft tissue sarcoma? Initial findings of a case-control study in New Zealand. Community Health Stud 1988:6:114 119. 9. Smith AH, Perace NE, Fisher DO, et al: Soft tissue sarcoma and exposure to phenoxy herbicides and chlorophenols in New Zealand. JNCI 1984:73:1111-1117. 10. An epidemiologic investigation of health effects in Air Force personnel following exposure to herbicide: Mortality update. Epidemiology Division, USAF School of Aerospace Medicine, Brooks AFB, Texas. December 1984. 11. A retrospective cohort study of mortality among Australian national servicemen of the Vietnam conflict era. The Commonwealth Institute of Health, University of Sydney. Canberra, Australian Government Printing Service, 1984. 12. Fingerhut MA, Halporin WE, Honchar PA, et al: An evaluation of reports of dioxin exposure and soft tissue sarcoma pathology among chemical workers in the United States. Scand J Work Environ Health 1984:10:299-303. 13. Enzinger FM, Lattes R, Torloni H: Histological typing of soft tissue tumors, in International Histological Classification of Tumors, No. 3. Geneva, World Health Organization, 1969. 14. Lawrence C, Reilly AA, Quickenton P, ot al: Mortality patterns of Now York State Vietnam veterans. Am J Public Health 1986:75:277879. 1218 Soft Tissue Sarcomas and Agent Orange/Kang et al In conclusion, a study of STS case subjects and a comparison patients group in VA hospitals did not reveal a statistically significant positive association between STS and previous military service in Vietnam. References � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Alvin L. Young Collection on Agent Orange Description An account of the resource <p style="margin-top: -1em; line-height: 1.2em;">The Alvin L. Young Collection on Agent Orange comprises 120 linear feet and spans the late 1800s to 2005; however, the bulk of the coverage is from the 1960s to the 1980s and there are many undated items. The collection was donated to Special Collections of the National Agricultural Library in 1985 by Dr. Alvin L. Young (1942- ). Dr. Young developed the collection as he conducted extensive research on the military defoliant Agent Orange. The collection is in good condition and includes letters, memoranda, books, reports, press releases, journal and newspaper clippings, field logs and notebooks, newsletters, maps, booklets and pamphlets, photographs, memorabilia, and audiotapes of an interview with Dr. Young.</p> <p>For more about this collection, <a href="/exhibits/speccoll/exhibits/show/alvin-l--young-collection-on-a">view the Agent Orange Exhibit.</a></p> Text A resource consisting primarily of words for reading. Examples include books, letters, dissertations, poems, newspapers, articles, archives of mailing lists. Note that facsimiles or images of texts are still of the genre Text. Box The box containing the original item. 066 Folder The folder containing the original item. 1768 Series The series number of the original item. Series III Subseries III Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Creator An entity primarily responsible for making the resource Kang, Han K. Lee Weatherbee Patricia P. Breslin Yvonne Lee Barclay M. Shepard Source A related resource from which the described resource is derived Journal of Occupational Medicine Date A point or period of time associated with an event in the lifecycle of the resource 1986-12-01 Title A name given to the resource Soft Tissue Sarcomas and Military Service in Vietnam: a Case Comparison Group Analysis of Hospital Patients Subject The topic of the resource cancer risk assessment dioxin ao_seriesIII https://www.nal.usda.gov/exhibits/speccoll/files/original/1bdac9903088241f58735eea077fdfa9.pdf fd78e6b9c1761e3abc6221fe6b7dbadc PDF Text Text Item D Number °4974 Author Kan D N0t Scanned 9. Han K. Corporate Author Report/Article TitlO Health Status of Self-Selected Group of 86,000 Vietnam Veterans Journal/Book Title Year 1983 Month/Day Color Number of images D ° Descriptor! Notes Friday, February 22, 2002 Page 4974 of 5115 �Health Status of Self-Selected Group of 86,000 Vietnam Veterans Han K. Rang, Joshua Barwick, Yvonne Lee, Patricia Breslin, Barclay M. Shepard Agent Orange Projects Office, Veterans Administration Washington, D.C. 20420 Since 1978 the Veterans Administration has provided at VA hospitals a complete physical examination and a group of baseline laboratory tests to each Vietnam veteran who was concerned about the adverse health effects of Agent Orange. Agent Orange was the mixture of phenoxy herbicides most commonly applied in Vietnam by the U.S. Air Force during the Vietnam war. Self-reported health problems as well as veterans' recollections of exposure to Agent Orange were also recorded during the examination. Therefore, these data sets were limited by self-selection, subjective reporting of exposure to Agent Orange, and possible selective recall of symptoms and exposure. Notwithstanding the limitations we have analyzed 85,903 Agent Orange examinations that have been computerized as of May, 1983 because of the great concern placed upon the current health status of Vietnam veterans. Ihe distribution by branch of military service of veterans coming to VA hospitals for the examination closely paralleled the distribution of military personnel in Vietnam. A total of 23,030 veterans or 26.8% of all examinees had no symptoms. Among those veterans reporting symptoms 38.9% reported skin rash, 17.5% reported nervousness, 14.0% reported headache, 12.1% reported abdominal problems and 10.2% reported personality disorders. In 36,538 veterans, or 42.5% of all examinees the examining physician did not establish a diagnosis. Among the veterans diagnosed by physicians, skin disease was the most frequent (26.3%) followed by mental disorders (10.4%). A remarkably similar distribution of frequencies of symptoms and diagnoses was observed among each of the branches of military service. Distribution of malignant neoplasm cases in the Agent Orange Registry was similar to that of the reference population. The strengths and limitations of the data base, which is the largest of this kind, will be discussed as well as the implications of the registry analyses. � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Alvin L. Young Collection on Agent Orange Description An account of the resource <p style="margin-top: -1em; line-height: 1.2em;">The Alvin L. Young Collection on Agent Orange comprises 120 linear feet and spans the late 1800s to 2005; however, the bulk of the coverage is from the 1960s to the 1980s and there are many undated items. The collection was donated to Special Collections of the National Agricultural Library in 1985 by Dr. Alvin L. Young (1942- ). Dr. Young developed the collection as he conducted extensive research on the military defoliant Agent Orange. The collection is in good condition and includes letters, memoranda, books, reports, press releases, journal and newspaper clippings, field logs and notebooks, newsletters, maps, booklets and pamphlets, photographs, memorabilia, and audiotapes of an interview with Dr. Young.</p> <p>For more about this collection, <a href="/exhibits/speccoll/exhibits/show/alvin-l--young-collection-on-a">view the Agent Orange Exhibit.</a></p> Text A resource consisting primarily of words for reading. Examples include books, letters, dissertations, poems, newspapers, articles, archives of mailing lists. Note that facsimiles or images of texts are still of the genre Text. Box The box containing the original item. 167 Folder The folder containing the original item. 4974 Series The series number of the original item. Series VII Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Creator An entity primarily responsible for making the resource Kang, Han K. Joshua Barwick Yvonne Lee Patricia Breslin Barclay M. Shepard Date A point or period of time associated with an event in the lifecycle of the resource 1983 Title A name given to the resource Health Status of Self-Selected Group of 86,000 Vietnam Veterans Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Alvin L. Young Collection on Agent Orange Description An account of the resource <p style="margin-top: -1em; line-height: 1.2em;">The Alvin L. Young Collection on Agent Orange comprises 120 linear feet and spans the late 1800s to 2005; however, the bulk of the coverage is from the 1960s to the 1980s and there are many undated items. The collection was donated to Special Collections of the National Agricultural Library in 1985 by Dr. Alvin L. Young (1942- ). Dr. Young developed the collection as he conducted extensive research on the military defoliant Agent Orange. The collection is in good condition and includes letters, memoranda, books, reports, press releases, journal and newspaper clippings, field logs and notebooks, newsletters, maps, booklets and pamphlets, photographs, memorabilia, and audiotapes of an interview with Dr. Young.</p> <p>For more about this collection, <a href="/exhibits/speccoll/exhibits/show/alvin-l--young-collection-on-a">view the Agent Orange Exhibit.</a></p> Text A resource consisting primarily of words for reading. Examples include books, letters, dissertations, poems, newspapers, articles, archives of mailing lists. Note that facsimiles or images of texts are still of the genre Text. Box The box containing the original item. 047 Folder The folder containing the original item. 1170 Series The series number of the original item. Series III Subseries I Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Creator An entity primarily responsible for making the resource Young, Alvin L. Han K. Kang Barclay M. Shepard Source A related resource from which the described resource is derived Environmental Science and Technology Date A point or period of time associated with an event in the lifecycle of the resource 1983 Title A name given to the resource Chlorinated Dioxins as Herbicide Contaminants Subject The topic of the resource Vietnam soft tissue sarcoma congenital birth defects veteran health and hygiene ao_seriesIII