~ .!J .3D/. d< .' fl..3~1 r---- 1, I <' :~:' Air Force Health Study; I ~~~ e,.' An Epidenziologic Investigation of I Health Effects in Air Force Personnel ' I Following Exposure to Herbicides j ... ~; 1I 11 SAJCTeam ~~~ George D. Lathrop, M.D., M.P.H., Stella G. Machado, Ph.D. Theodore G. Karrison, Ph.D. '-William D. Grubbs, Ph.D. Wanda F. Thomas, M.S. Project Manager: W.F. Thomas Air Force Team Ph.D. COL William H. Wolfe, M.D., M.P.H. Joel E. Michalek, Ph.D. LTC Judson C. Miner, D.V.M., M.P.H. LTC Michael R. Peterson, D.V.M., M.P.H., Dr.P.H. Program Manager: R.W. Ogershok EPIDEMIOLOGY DIVISION USAF School of Aerospace Medicine Human Systems Division (A,FSC) Brooks Air Force Base, Texas 78235 October 1987 SUMMARY First Followup Examination Results January 191!5 to September 191!7 ContractNumbcr: F41689-85-D-OOIO SAIC Project Number 2-816-XX-1951254-XX (Distribution Unlimited) :;i_ ~--:. _--! :· ,j:~t'~-. :~. ' ..JiiMi-. . r" ',_...., '-r/ UNCLASSIFIED SECURITY CLASSIFICATION OF THIS PAGE REPORT DOCUMENTATION PAGE 1e. REPORT SECURITY CLASSIFICATION lb. RESTRICTIVE MARKINGS UNCLASSIFIED 2e. SECURITY CLASSIFICATION AUTHORITY 3. OISTAIBUTION/AVAILABILITY OF REPORT 2b. DECLASSIFICATION/DOWNGRADING SCHEDULE Approved for public release; distribution unlimited •. PERFORMING ORGANIZATION REPORT NUMBEA(Sl 5. MONITORING ORGANIZATION REPORT NUMBER(Sl USAFSAM-TR-87-36 6e. NAME OF PERFORMING ORGANIZATION Sb. OFFICE SYMBOL 7e. NAME OF MONITORING ORGANIZATION Science A~plications Inter-(/( applicabl~} (HSD) national or~~. L1fe Science Human Systems Division anrl Sv<:tPmc;. n;:~rtmPnt 6c. ADDRESS (City, Stat~ and ZIP COO.lO). ****Group-by-covariate Interaction. --8 Analysis not done. NS*: Borderline significant (O.OSC RH>C RH>C 11 TABLE 4. (Continued} Overall Suaaary Results of Unadjusted and Adjusted Analyses of Neurological Variables Direction of Variable Unadjusted Adjusted Results** Central Nervous Slstem Coordination Tremor NS* NS* RH>C Coordination NS Romberg Sign Gait CNS Summary Indexd NS NS 0.036 NS 0.042 RH>C **RH>C: More abnormalities in Ranch Hand group than in Comparison group. •oisease categories include: inflammatory diseases, heriditary and degenerative diseases, peripheral disorders, disorders of the eye, disorders of the ear, and other disorders. NS:Not significant (p>O.lO). bNo inflammatory diseases noted; borderline significant (p=0.069, RH>C) for other disorders; not significant for remaining categories. --Analysis not performed because of sparse number of abnormalities. cNo abnormalities present. NS*Borderline significant (O.OSRH NS NS NS NS NS*b NS*b RH>C NS NS 0.017 0.020 C>RH NS **** NS NS NS **** NS*b **** RH>C NS **** <0.001 **** RB>C NS NS 0.003 0.040 RB>C NS NS •aH>C -more abnormalities in Ranch Hands; C>RH -more abnormalities in Comparisons. blllnesses include psychosis, alcohol dependence, anxiety, and other neuroses. --Analysis not performed. NS: Not significant. NS*: Borderline significant (O.OSC Laboratory Testing SGOT SGPT GGTP Alkaline Phosphate Total Bilirubin Direct Bilirubin LOB Cholesterol Triglycerides Uroporphyrin Coproporphyrin NS NS* NS 0.009 NS NS NS NS NS 0.048 NS* NS NS NS NS NS NS NS NS NS NS 0.048 NS 0.008 NS NS **** NS **** **** NS* **** 0.029 NS **** NS NS NS NS NS NS NS NS NS* NS **** NS NS **** C>RH RH>C C>RH RH>C Questionnaire-Laboratory Corr elation Skin Bruises, Patches,-and Sensitivity 0.001 RH>C *C: Continuous D: Discrete **RH>C: more abnormalities, or higher mean value, in Ranch Hands. C>RH: more abnormalities, or higher mean value, in Comparisons. •Adjusted for blood type. NS: Not significant (p>O.lO). NS*: Borderline significant (O.OSO.lO). --Analyses not performed. NS*: Borderline significant (0.05O.lO). Statistical/ Clinical Analysis Discrete Continuous Manual Doppler Manual Doppler Manual Doppler Manual Doppler Manual Doppler Manual Doppler Manual Doppler Manual Doppler Unadjusted Adjusted NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS **** NS NS NS **** NS NS NS **** NS NS NS **** NS NS NS **** NS NS NS **** NS NS NS*:Borderline significant (0.05C (Est. RR: 1.25; 95% C.I.: (1.02, 1.54), p-0.054). In the analysis of peripheral vascular function, no significant group differences were observed for abnormalities involving radial, femoral, pop liteal, posterior tibial, dorsalis pedis, or three anatomic aggregates of these pulses, either by manual palpation or Doppler techniques. This overall finding was in distinct contrast to the 1982 Baseline examination, which by the manual palpation method, showed significant peripheral pulse deficits in the Ranch Hands. This favorable pulse reversal over the two examinations is primarily attributed to the rigid 4-hour tobacco abstinence applied prior to Doppler testing, although other factors may be related. The lack of group differences for pulse abnormalities was noted even though the manual and Doppler techniques differed significantly (p<0.05, p<0.001 for most) in the detection of abnormal! ties for all but one of the pulses or pulse combina tions. For manually-determined pulse abnormalities, there was a significant group-by-race interaction for the popliteal pulses, a significant group by-percent body fat interaction for the leg pulses, and significant group by-occupation interactions for the posterior tibial, dorsalis pedis, and the three pulse aggregates (leg, peripheral, and all pulses). No interactions were encountered in the adjusted analyses of the Doppler results, and none showed significant group differences. Statistical analyses i nvol vi ng the Original Ccxnpari sons also showed no significant differences in the cardiovascular measurements between groups, although slightly different interactions were detected in sane of the adjusted analyses. For the exposure analyses, the only statistically significant effects were those pointing to less bradycardia and less reported and verified heart disease in the medium exposure level category, as contrasted to the low exposure category, among the enlisted groundcrew. In many cases there were too few abnormalities within the occupational categories to permit formal statis tical tests. Overall, the exposure analyses were deemed as unsupporti ve of any meaningful dose-response relationships. The longitudinal analysis of the p'-!lse index confirmed the significantdifference in the change in the pattern of results from the Baseline examination to the followup examination, largely due to a relatively greater increase of pulse abnormalities in the Comparison group than in the Ranch Hand group.There was no significant change in pattern between the two groups in overall ECG findings between examinations. There was a similar distribution of the covariates between groups, exceptfor a slightly higher level of current Ranch Hand smoking (also observed at Baseline), and a corresponding slightly lower mean percent body fat. The general covariate effects were strong and showed expected, classical associations with the cardiovascular measurements. However, unexpected effects were consistently noted for personality score, with higher proportions of various cardiovascular abnormalities associated with scores in the Type B direction, a finding possibly attributable to the method of personality determination. Although smoking was positively associated with many of the cardiovascular measurements, negative associations were seen between current smoking and reported and verified essential hypertension and'between pack-years of smokingand verified hypertension. 25 In conclusion, of 27 cardiovascular variables, only one, verified heart disease, showed a significant excess in the Ranch Hands, but this finding was largely unsupported by other cardiac measurements. Both manual palpation and Doppler recordings of five peripheral pulses were similar in both groups, in marked contrast to the 1982 Baseline examination which found significant pulse deficits in the Ranch Hand group. This change at the followup examination was most likely due to required tobacco abstinence prior to the pulse measure~ ments. Exposure index analyses did not support a consistent dose~response relationship for any variable. Overall, there was remarkable similarity in the cardiovascular health between the Ranch Hand and Comparison groups. HEMATOLOGY The functional integrity of the hematopoietic system was assessed by the measurement of eight peripheral blood variables: red blood cell count (RBC), white blood cell count (WBC), hemoglobin (HGB), hematocrit (HCT), mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (l"iCHC), and platelet count (PLT). These variables were analyzed in the discrete form to detect differences in the percentages of values outside the designated laboratory range, as well as in the continuous form to detect shifts in mean values between the two groups. A summary of all of these analyses, unadjusted and adjusted for the covariates of age, race, occupation, and smoking, is presented in Table 10. The unadjusted discrete analysis of the percent abnormal values, both low and high, showed no statistically significant differences between the Ranch Hand and Comparison groups for any of the hematological variables. Similarly, the adjusted categorical analysis disclosed that none of the adjusted relative risks was significant for either group, and that no significant group-bycovariate interactions were present. The unadjusted continuous analysis did not detect any significant differences in group means for any of the eight variables. The adjusted continuous analysis found no significant group differences for·HGB, HCT, MCV, MCH, and MCHC, but encountered significant three-factor interactions for WBC (groupby-race-by-age, group-by-age-by-smoking history, and group-by-race-byoccupation), for PLT (group-by-race-by-smoking history and group-by-raceby-current level of smoking), and a borderline interaction for RBC (groupby-occupation-by-smoking history). Ranch Hand versus Original Comparison analyses revealed further significant interactions for HGB, HCT, MCV, and MCH. As no group strata demonstrated consistent patterns of hematologic impairment, biologic relevance was not assigned to the interactions. The covariate effects of age, race, occupation, and smoking history were highly significant for many of the hematological variables. The effect of race was particularly profound for all variables except PLT. There was fair consistency in the covariate effects upon the RBC-related variables. Generally, decreasing hematologic values were associated with increasing age and the Black race, and increasing hematologic values were associated with increasing smoking. The detection of these classical covariate effects lends credence to the overall finding of nonsignificant group differences for all of the hematological variables. Significant group differences found for MCV and MCH at the Baseline examination were not significant at the first followup. Other differences (e.g., covariate effects, interactions) 26 TABLE 10. Overall Su.aary Results of Unadjusted and Adjusted Analyses of Beaatological Variables Unadjusted Adjusted Mean Categorical Mean Categorical RBC NS NS NS* NS VBC NS NS **** NS HGB NS NS NS NS HCT NS NS NS NS MCV NS NS NS NS MCB NS NS NS NS MCBC NS NS PLT NS NS **** NS NS: Not sgnificant (p>O.lO). NS*: Borderline significant group-by-covariate interaction (0.05~<0.10). --Analysis not performed due to sparse data. ****Group-by-covariate interaction. Note: Significant group-by-covariate interaction, Ranch Hands versus Original Comparisons only, for HGB, HCT, HCV, and HCH. 27 between the Baseline and followup examinations may be due to small numeric shifts in the cohorts under study (see Chapter 2) and the selection of alternate statistical models, or due to chance. Unadjusted continuous exposure analyses in the Ranch Hand group revealed only one significant effect (RBC in enlisted groundcrew) and one borderline effect (HCT in enlisted groundcrew) but neither was consistent with a plausible dose-response relationship. The adjusted continuous exposure analyses found only one significant contrast (HCT, medium exposure versus low exposure, enlisted groundcrew). However, seven exposure level-by-covariate interactions were noted for four of the hematological variables. Discrete outcome analyses of the exposure level index revealed a significant result only for WBC in the enlisted flyers. The longitudinal analyses of MCV, MCH, and PLT found significant differences only for PLT values between the Baseline and followup examinations, with the Baseline group difference in mean values closing to near equivalence at the followup examination. In conclusion, none of the eight hematological variables were found to differ significantly between the Ranch Hand and Comparison groups. In fact, group equivalence was more apparent at the followup examination than at the Baseline examination. The classical effects of age, race, and smoking were demonstrated with most of the hematological variables. The longitudlnal analyses also suggested that nelther group manifested an impairment of the hematopoietic system. Exposure index analyses dld not support a plauslble dose-response relationship for any of the hematological variables. RENAL A summary of all renal variables, including unadjusted and adjusted analyses, is displayed in Table 11. A historical assessment of kidney disease/kidney stones by a reviewof-systems questionnaire showed no significant differences between the Ranch Hand and Comparison groups (see Table 12). An adjusted analysis did not alter this conclusion as an adjusted relative risk of 0.95 (95% C.I.: [0.71,1 .25], p=0.693) was demonstrated. These statistics appeared to be in marked contrast to the Baseline historical findings. Differences vis-a-vis the Baseline were most likely due to a difference in questionnaire techniques. Current renal function was evaluated by five laboratory variables: urine protein, occult blood, urine, white blood cell counts (WBC's), blood urea nitrogen (BUN), and urine speciflc gravity. Invasive procedures were not used. The unadjusted analysis of proteinuria showed no group differences (Est. RR: 1.18, 95% C.I.: [0.75,1 .86], p=0.485), but the adjusted analysis showed an interaction of group and diabetic class; appropriate stratified analyses revealed that the prevalence of proteinuria was lower in the Ranch Hands than in the Comparisons in the diabetic and impaired strata, but higher in the normal strata for the Ranch Hands. These results were in contrast to the 28 TABLE 11, Overall Summary Results of Unadjusted and Adjusted Analyses for Renal Variables Variable Unadjusted Adjusted Reported Kidney Disease NS NS Urinary Protein NS **** Urinary Occult Blood NS **** Urinary Leukocytosis NS **** BUN NS **** Urine Specific Gravity NS* **** NS: Not significant (p>O.lO). NS*: Borderline significant (O.OSO.lO). --b Analysis not performed. NS*: Borderline significant (0.05O.lO). ****Significant group-by-covariate interaction. Cell surface marker studies were conducted for total T cells (T11 ), helper T cells (T~), suppressor T cells (T8), B cells, monocytes, and HLA-DR cells; the ratio of T~IT8 cells was included in the analysis. Because of inherent significant day~to~day and batch-to-batch variation, all results (including functional stimulation studies) were adjusted for blood-draw day variation. Statistical testing of the seven phenotypic cell markers did not reveal any significant group differences (interactions excepted), either unadjusted or adjusted for the covariates of age, race, occupation, current smoking, lifetime smoking history (pack-years), current alcohol use, or lifetime alcohol use (drink-years). Similarly, none of the unadjusted or adjusted analyses of the functional stimulation studies (for phytohemagglutinin, pokeweed mitogen, or mixed lymphocyte culture) showed any statistically significant group differences. However, the adjusted analyses for total T cells, B cells, monocytes, HLA-DR cells, pokeweed mitogen, and net mixed lymphocyte culture stimulation showed some significant group-by-covariate interactions, precluding direct adjusted group contrasts. Overall, no discernible pattern was identified to suggest a detriment in any subgroup of either the Ranch Hands or Comparisons. Results were similar between the analyses of the total Comparison group and the analyses of the Original Comparisons. The covariate effects of age, race, smoking, and alcohol use were gener ally profound on most variables in the phenotypic and stimulation studies. Consistently decreasing values of all cell markers and stimulated cells were associated with increasing age, whereas increased levels of smoking were usually associated with increases in the values of those variables. Blacks 3~ had consistently higher stimulated cell counts than nonblacks, but this effect was not observed for counts of T cells, B cells, or HLA-DR cells. Enlisted personnel generally had higher cell surface marker counts than officers. Exposure index analyses of cell surface markers revealed no pattern consistent with a dose-response relationship. For enlisted groundcrew, the mean total T cell and suppressor T cell counts for the medium exposure level were significantly lower than those of the low exposure level, but were slightly lower than those of the high exposure level. The exposure index analyses of the functional stimulation tests revealed no consistent significant dose-response patterns for net PHA counts or net MLC counts. For net pokeweed counts, enlisted flyers in the high exposure level had a significantly lower adjusted count than enlisted flyers in the low exposure level, and a decreasing trend was apparent. The delayed hypersensitivity response was assessed by the skin test antigens of mumps, Candida albicans, Trichophyton, and staph-phage-lysate. The 48~hour measurements of skin induration and erythema for the four tests showed marked inter-reader variation. Analyses showed that one of the three skin test readers too often measured induration larger than erythema (a clinically unacceptable finding), in an average of 30 percent of the readings, and did not yield measurements that detected a case of possible or overt anergy, whereas the other two readers found this condition in 5.6 percent of the participants. Remaining data from Readers 1 and 3, however, were found to vary significantly in clinical interpretation over duration of the examination. Consequently, all skin test data were declared invalid, and were not used in the assessment of group differences. The skin test reading problems led to the use of additional clinical quality control procedures for the AFHS followup examination begun in May 1987. In conclusion, no significant group differences were judged present for the comprehensive cell surface marker or functional stimulation studies. The profound effects of age, smoking, and alcohol use were observed in these immunologic tests. The assessment of delayed hypersensitivity skin responses was precluded by poor data quality and excluded from further analysis. Overall, there was no indication of impaired immunologic competence in either group. RESPIRATORY SYSTEM A summary of the results on the analyses of reported history of respiratory illness and of radiological and clinical findings is given in Table 16. Based on the 31 December 1986 mortality data, there were seven deaths from respiratory conditions in the Comparison group and none in the Ranch Hand group. 35 TABLE 16. Overall s~lesults of Unadjusted and Adjusted Aaal.yses of Put.onary Disease Pul.onary Disease Unadjusted Adjusted Reported History of Respiratory Illness Asthaa NS NS Bronchitis NS NS Pleurisy NS **** Pneuaonia NS NS Tuberculosis NS **** Radiological and Clinical Findings Thorax and Lungs NS NS Asy.aetrical Expiration NS NS Hyperresonance NS NS Dullness NS NS Vheezes NS NS Rales NS **** X Ray NS NS NS: Not significant (p>O.lO) ****Group-by-covariate interaction. There were no group differences found for reported history of asthma, bronchitis, pleurisy, or tuberculosis based on the unadjusted analyses. Adjustments for age and lifetime smoking did not alter the findings of group similarity, although there was a significant group-by-pack-year interaction for pleurisy and for tuberculosis. Similarly, there were no significant group differences in the unadjusted analyses for the radiological and clinical respiratory findings of thorax and lungs, asymmetrical expiration, hyperresonance, dullness, wheezes, rales and x-ray interpretations. These findings were supported by the adjusted analyses, although there was a group-by~age interaction for rales. The exposure index analyses revealed no consistent dose~response pattern. Analyses of past history of respiratory illness and the clinical and radiological examination of the chest and lungs did not reveal any statistically significant differences between the Ranch Hand and Comparison groups suggestive of herbicide related disease. Several group-by-covariate interactions did exhibit statistical significance, but these findings did not indicate any consistent patterns suggesting different disease experience in the two groups. 36 CONCLUSIONS This chapter summarizes the conclusions drawn from the statistical analyses that have been conducted on the Air Force Health Study data base. The followup study, which began in 1985, was the logical extension of the 1982 Baseline, building upon the strengths of the Baseline study and utilizing the data collected at both the Baseline and the followup. The high level of Government support and outstanding participation of the study subjects that characterized the Baseline study were maintained through this first followup. STUDY PERFORMANCE ASPECTS Of the living Baseline study participants, 99.2 percent were located and asked to participate in the followup. Participation in the followup physical examination and questionnaire was very high. Of the fully compliant Baseline participants, 971 of the 1,045 Ranch Hands (92.9%) and 1,139 of the 1,224 Comparisons (93.1%) participated in the followup. Thus, there was no group difference in compliance of the Baseline partic~pants at the followup. Overall, the 2,309 participants in the followup (1 ,016 Ranch Hands and 1,293 Comparisons) represented a loss of 159 individuals and a gain of 199 since Baseline. One percent of the fully compliant Baseline population died between 1982 and the 1985 followup examination. The bias/compliance analyses suggested that there had been no change between Baseline and the followup in the way replacements volunteered for entry into the study, and that no additional bias had been introduced at the followup due to scheduling differences. Although replacements were not health-matched at Baseline as they were at the followup, they were similar to refusals with respect to reported health, medication use, and income level. This result supported the conclusion that there has been little, if any, selection bias due to differential participation in the Comparison group and supported the use of the total Comparison group in the main analyses presented in this report. POPULATION CHARACTERISTICS Overall, the Ranch Hands and Comparisons reported similar social and behavioral characteristics. No significant differences were found in age, educational background, religious preference, current military status, and income level. Significantly more Ranch Hands smoked cigarettes at the time of the followup examination than did Comparisons, but there was no significant difference between groups on past cigarette, cigar, and pipe use and on recent and past use of marijuana. A much higher percentage of participants reported past marijuana use at the followup than at Baseline. This difference was most likely due to a greater level of confidentiality afforded by the questionnaire technique. Risk taking behavior, assessed by questions on potentially dangerous recreational activities, revealed borderline significance. Slightly more Comparisons were scuba divers and more Ranch Hands raced motor vehicles. The difference in scuba diving was also significant at Baseline. 37 Patterns of Results Both the chapter conclusions and the final conclusions of this report have been predicated upon concepts of consistency, specificity, coherence, strength, and plausibility as they apply to the interpretation of group differences. In particular, careful consideration has been given to a variety of data and patterns of results that have emerged from the clinical evaluations. Specifically, there were few differences in the proportions of abnormalities between groups; the positive associations have not aggregated in the clinical areas of prime dioxin concern, nor have they been of serious clinical importance; the unadjusted results have been remarkably concordant with the adjusted results, both in terms of relative risk and p value; the analyses using the Original Comparison set have largely mirrored the results found with the total Comparison group; many of the group differences noted at Baseline have disappeared at the followup examination, and only a few new associations have emerged; almost all of the covariates have acted as expected in the adjusted analyses; and the exposure index analyses and the group-by-covariate interactions have not demonstrated biological patterns of concern and appeared to be more likely due to chance than not. Due to the acknowledged limitations of the exposure index used in this report (and considering the potential use of dioxin body burden levels at the next followup), dose-response relationships have not been emphasized in reaching final eonclusions. The overall pattern of these findings indicates that this followup study cannot be viewed as alarming from the traditional perspectives of clinical medicine or epidemiology. This study, in fact, demonstrates similarity in current health status between the Ranch Hand and Comparison groups. CLINICAL ASPECTS General Health The nonspecific assessment of general health showed relatively close similarity between the two groups. Ranch Hands rated their health as fair or poor more frequently, but this difference was found only in the enlisted groundcrew and not in the officers nor enlisted flyers. The perception of health in both groups had improved since Baseline. Physician-rated appearance of relative age was not found to be significantly different at the followup in contrast to the Baseline finding that a higher percent of Ranch Hands than Comparisons looked younger than their stated age. The categorical analysis of sedimentation rate showed that the Ranch Hands had more abnormalities than the Comparisons. These results were not supported by the continuous analysis of mean sedimentation rates and were opposite to the Baseline results, which showed that younger Comparisons had elevated sedimentation rates. The categorical analysis of percent body fat showed no significant differences between the two groups, which was consistent with Baseline. However, the continuous analysis founu that the Ranch Hands had a significantly lower mean percent body fat using age, race, and occupation as covariates. The detailed exposure analyses revealed no consistent exposure effects, and this result was consistent with the Baseline analysis. No longitudinal difference was found on perception of health. A significant group difference was found over time for the longitudinal analysis of sedimentation rate due to the change in the findings between the two examinations, possibly related to a change in laboratory methodology. 38 Malignancy Skin and systemic cancers, both suspected and verified by medical records, showed no significant group differences for the Baseline~followup interval (1982-1985). However, for all neoplasms combined (malignant, benign, and uncertain), a borderline significant excess in the Ranch Hand group was noted in an unadjusted analysis. The analyses of interval cancers revealed group interactions for verified and verified plus suspected basal cell carcinoma and verified plus suspected systemic cancers. Nonsignificant findings were observed for verified and verified plus suspected sun exposure-related cancers. Verified systemic cancers did not differ significantly between groups. The analyses of lifetime cancer found significant results for verified basal cell carcinoma and verified sun exposure-related skin cancers. Group interactions were noted for systemic cancer categories and for verified plus suspected basal cell carcinoma. The higher rate of basal cell carcinoma in the Ranch Hands versus the Comparisons found at Baseline was nonsignificant for the followup interval, but due to the effect of the larger number of Baseline cases and the significant confounding of average residential latitude, the adjusted analysis of lifetime basal cell carcinoma emerged as statistically significant. There were several disparities in the distribution of testicular, colon, and smoking-related tumors in the groups. Further, one case of soft tissue sarcoma and one possible lymphoma (both in Ranch Hands) were diagnosed in the interval, balancing the two similar cases found in the Comparison group at Baseline. Considering that the systemic cancer curves are in their early stages for both ~roups, with perhaps insufficient latency, the cancer results of the followup examination should not be viewed as disturbing, but as cause for continued monitoring. Neurological Assessment None of the 27 neurological variables demonstrated a significant group difference, although several variables had relative risks which were greater than one. Tnere was no group difference in reported neurological illnesses for the interval or for a lifetime history. Of the cranial nerve variables, speech and tongue position were marginally significant, with the Ranch Hands at a slight detriment. The analyses of peripheral nerve function showed no significant differences between the Ranch Hands and the Comparisons. In the analysis of central nervous system function, hand tremor was found to be of borderline significance, with the Ranch Hands faring slightly worse than the Comparisons. A borderline significant group interaction (Ranch Hand hand tremor by insecticide exposure) may have had biological and operational significance. Overall, substantially fewer neurological abnormalities were detected at the followup examination than at the Baseline examination. The exposure analyses showed only occasional statistically significant results, although no consistent pattern with increasing exposure was evident. In the longitudinal analysis of the Babinski reflex, a significant change over time was observed. This was due to a nonsignificant finding in the Ranch Hands at the followup, which differed from the significant adverse finding at Baseline. The covariates of age, alcohol history, and diabetes showed classical effects with many of the neurological measurements. Overall, the followup examination results were quite similar to the Baseline findings. 39 Psychological Assessment The reported and verified data on lifetime psychological illnesses showed no significant differences between groups. Distributional tests of the 14 Minnesota Multiphasic Personality Inventory (MMPI) scales, stratified by occupation, revealed that only 2 of the 42 results approached significance. For the total Cornell Medical Index (CMI), separate distributional tests were conducted with stratification by age, race, occupation, education, and current drinking status; a significant difference was found for one statum of each of the covariates. In all cases, the mean of the Ranch Hand distribution was greater than the mean of the Comparisons. The analysis of the 14 MMPI scales showed that there was a significant difference between the two groups for denial and masculinity/femininity, with more abnormalities in the Comparisons than the Ranch Hands. The results of the analyses for hysteria were of borderline significance, with more abnormalities in the Ranch Hands. There were more abnormalities in the Ranch Hands than the Comparisons for social introversion, which was of borderline significance. Differences in the total CMI and A~H area subscore were found to be significant, with more abnormalities in the Ranch Hands. There was no significant difference between the two groups on the Halstead-Reitan Battery impairment index, a measure of the functional integrity of the CNS. The exposure index analyses did not reveal any pattern consistent with a dose-response relationship. As expected, the effects of age, educational level, and alcoholic history showed profound effects on many of the psychological measurements. Gastrointestinal Assessment Although the followup gastrointestinal assessment disclosed more statistically significant findings than the Baseline examination, the abnormalities were distributed equally between the two groups, and there was no clinical, statistical, or exposure pattern consistent with an herbicide-related effect on health. No historical or biochemical evidence was found to suggest an increased likelihood of porphyria cutanea tarda (PCT) in the Ranch Hand group. Only sparse and nonsignificant liver disorders were reported for the interval between Baseline and followup. Also, for the lifetime history of livei disorders, there were no significant differences between groups. Further, there were no significant group differences in reported lifetime peptic ,ulcer disease. A review of digestive system mortality showed a relative excess in the Ranch Hands but a relative lack of malignant neoplasms. The results of the physical examination showed a borderline increase of hepatomegaly in the Ranch Hand group. There was a significantly lower mean 'Serum glutamic-pyruvic transminase (SGPT) level, a greater mean alkaline phosphatase level, and a lower mean uroporphyrin level in the Ranch Hand group. The analysis of coproporphyrin was of borderline significance, with the mean of the Ranch Hands in excess of the mean of the Comparisons. No group d~fferences were found for serum glutamic-oxaloacetic transminase (SGOT), gamma-glutamyl transpeptidase (GGTP), total and direct bilirubin, lactic dehydrogenase (LDH), cholesterol, or triglycerides. The numerous group-by-covariate interactions did not disclose any consistent subgroup patterns detrimental to the Ranch Hands. These findings were generally consistent witH the results of the 1982 assessment. The longitudinal analyses fo~ SGOT, SGPT, and GGTP showed no significant differences between results by group over time. 40 Dermatological Evaluation No significant group differences were identified in the dermatological evaluation. None of the questionnaire data showed an increased likelihood of past chloracne, as determined by anatomic patterns of acne, and no cases were diagnosed in the physical examination. Analyses were conducted on six dermatologic disorders (comedones, acneiform lesions, acneiform scars, inclusion cysts, depigmentation, and hyperpigmentation) and on a composite variable of 16 other minor conditions (the latter not generally associated with chloracne). Exposure index analyses did not reveal consistent patterns suggestive of a dose-response relationship. The longitudinal analysis, based on a composite dermatology index, showed no significant differences between the results over time. Substantially more dermatologic abnormalities were detected at the followup examination than at the Baseline examination. In general, however, the followup results were consistent with the findings at Baseline. Cardiovascular Evaluation Overall there was remarkable similarity in the cardiovascular health of the Ranch Hands and the Comparisons. Of the 27 cardiovascular variables, there was a significant difference for only one, verified heart disease, with an excess in the Ranch Hand group. This finding was largely unsupported byother cardiac measurements. The cardiovascular assessment was based on reported and verified heart disease; the measurement of central cardiac function by systolic blood pressure, abnormal heart sounds, and ECG findings; and the evaluation of peripheral vascular function by diastolic blood pressure, funduscopic examination, presence of carotid bruits, and detailed manual and Doppler measurements of five peripheral pulses. Doppler recordingsof five peripheral pulses were similar in both groups, a finding which was in marked contrast to the Baseline examination that found significant pulse deficits in the Ranch Hand group. This change was most likely due to a required 4-hour abstinence from tobacco prior to the pulse measurements. Overall, the exposure analyses were unsupportive of any meaningful dose~response relationship. The longitudinal analyses confirmed the changein pulse abnormalities in the Ranch Hand group over time, but showed no significant group change in overall ECG findings between the examinations. Hematological Evaluation The hematological evaluation found that neither group manifested an impairment of the hematopoietic system, consistent with similar findings at the Baseline. The evaluation was based on eight peripheral blood variables: red blood cells (RBC), white blood cells (WBC), hemoglobin (HGB), hematocrit concentration (HCT), corpuscular volume (MCV), corpuscular hemoglobin (MCH),corpuscular hemoglobin concentration (MCHC), and platelet count (PLT). Both the discrete and categorical analyses revealed no significant groupdifferences. The covariate effects of age, race, occupation, and smokinghistory were highly significant for many of the variables. Two group-by-covariate interactions in the analyses of mean differences did not appear to have a meaningful interpretation. The exposure index analyses did not support any plausible dose-response relationship. The longitudinal 41 analyses of MCV, MCH, and PLT found significant differences only for PLT between the Baseline and the followup, with the Ranch Hands exhibiting a slight decline in mean level from Baseline and the Comparisons showing an opposite change. Renal Assessment None of the six renal variables of reported kidney disease, urine protein, occult blood, urine white blood cell count, blood urea nitrogen, and urine specific gravity showed a significant difference between the two groups based on the unadjusted analyses. In the adjusted analyses of the laboratory variables, however, there were significant group-by-covariate interactions that did not yield a consistent pattern to suggest a renal detriment to either group. The finding of group equivalence for past kidney disease was in contrast to the Baseline examination, which found significantly more reported disease in the Ranch Hand group. The difference in findings is more likely due to a change in questionnaire wording than to a true change in renal health. Like the Baseline findings, the exposure index analyses showed very little evidence of a dose-response relationship. In the longitudinal analyses of blood urea nitrogen, there were no significant changes between the examinations by group. Endocrine Assessment In general, the endocrine health status of the Ranch Hands and the Comparisons was reasonably comparable. The examination found no significant differences between the two groups for past thyroid disease, or thyroid and testicular abnormalities determined by palpation. In the analyses of the seven laboratory values (T %Uptake; thyroid stimulating hormone (TSH); 3 testosterone; initial, second, and differential cortisol; and postprandial glucose), significant differences were found for TSH and testosterone, with higher mean levels in the Ranch Hands. These analyses were not supported by the categorical analyses. The thyroid test results were conflicting with respect to an assertion of hypothyroidism in the Ranch Hands (a possible dioxin effect). Mean levels of testosterone were significantly elevated in the Ranch Hand group as contrasted with the Comparisons in the 10-25 percent body fat category. The effects of personality score and percent body fat on the differential cortisol levels were not fully expected. Although tests of 2-hour postprandial mean values showed no significant group differences, comparable categorical tests revealed that significantly fewer Ranch Hands had impaired glucose levels, but conversely, had more (nonsignificant) diabetic levels of glucose. Analyses of the composite diabetes indicator (history plus 2~hour postprandial results) did not disclose significant group differences. The exposure index analyses suggested that the enlisted flyers in the medium exposure level were significantly different from those in the low exposure level for differential cortisol, postprandial glucose, and testosterone. The corresponding high to low contrasts were not significant. The longitudinal analyses were based on %Uptake, TSH, and testosterone, and revealed only T1 symmetrical and nonsignificant changes in the Ranch Hand and Comparison groups over the time interval. 42 Immunological Evaluation Overall, there were no significant group differences or any indication of impaired immunological competence in either group based on comprehensive cell surface marker and functional stimulation studies. Six cell surface markers (total T cells, helper T cells, suppressor T cells, B cells, monocytes, HLA-DR cells, and a constructed helper/suppressor ratio variable) and three functional stimulation studies (PHA, pokeweed, and mixed lymphocyte culture) were conducted on 47 percent of the study population. No significant differences were revealed for five of these variables. In the analyses of the other five variables, there were significant group-by•covariate interactions, but no discernible pattern was identified to suggest a detriment in any subgroup of either group. Skin test assessments of delayed hypersensitivity were characterized by inter-reader variation and shifting diagnostic criteria for anergy. The skin test data were judged invalid and were not subjected to statistical testing for group differences. No consistent pattern of immunological deficits could be associated with increasing levels of herbicide exposure in the Ranch Hand group. Pulmonary Disease The pulmonary assessment did not reveal any statistically significant differences between the Ranch Hand and Comparison groups that were suggestive of an herbicide-related disease. The analyses consisted of group assessments of respiratory disease incidence, physical examination abnormalities, and the current prevalence of x-ray abnormalities. There were no significant differences between the Ranch Hands and Comparisons for history of asthma, bronchitis, pneumonia, or for six of seven clinical variables (excluding rales) determined by x-ray or auscultation. Analyses of history of pleurisy, history of tuberculosis, and rales showed Significant but inconsistent group-by-covariate interactions. These findings did not indicate any patterns suggesting a different disease experience in the two groups. The exposure index analyses did not reveal any consistent pattern suggestive of an increasing dose response. CONCLUSION The results of the first followup study in 1985 have shown a subtle but consistent narrowing of medical differences between the Ranch Hands and Comparisons since the Baseline Study in 1982. The 1985 examination results provide reassuring evidence that the current state of health of the Ranch Hand participants is unrelated to herbicide exposure in Vietnam. Continued close medical surveillance of these military populations is strongly indicated. This followup report concludes that there is not sufficient plausible or consistent scientific evidence at this time to implicate a causal relationship between herbicide exposure and adverse health in the Ranch Hand group. FUTURE DIRECTIONS The scope and complexity of the AFHS has required gradual refinement and correction to meet the challenges of changing technology and scientific direction, and to ensure continued participation of all enrolled members. 43 This chapter outlines some of the changes incorporated in the fifth~year followup examination and identifies several areas of future work expected to significantly augment the study. FIFTH-YEAR FOLLOWUP EXAMINATION Since the fifth-year followup examination was initiated prior to the full analysis of the data from the third-year examination, most modifications were founded upon quality control issues and the desire to make the clinical content of the examination more responsive to the medical needs of the participants. Clinical quality control enhancements were made to improve measurement techniques. The digit preference noted in systolic and diastolic blood pressure readings led to the use of automated blood pressure recording; all other parameters of the blood pressure readings (e.g., sitting position, three recordings, nondominant arm at heart level) were not changed. The problem in skin test reading was met by a rigorous quality control plan that included the following elements: refresher training for readers; a required reading of the four skin tests of all participants by both readers, each blind to the results of the other; a required reread of 10 percent of all tests by each of the readers, each blind to the previous reading; and a required weekly report citing numbers and proportions of participants with possible anergy, reversal of induration~erythema measurements, and untoward skin reactions or other reading problems (e.g., participant refusal). In addition, new skin test forms were developed to facilitate accurate recording and transcription; specific clinical criteria were formulated to require consultation by an allergist; and the skin test measurement criterion for possible anergy, consistent with current World Health Organization guidelines, was adopted for the clinical interpretation of all skin test readings. It is anticipated that this clinical quality control program will standardize both readings and interpretations, and will produce a uniformly superior data set. EXPOSURE INDEX REFINEMENTS Since the development of the Study Protocol and the analysis of the 1982 Baseline data, there has been concern among some scientists and the principal investigators over the accuracy and validity of the exposure estimates. It is unclear whether statistically significant differences in some variables between the Ranch Hand and Comparison groups, unsupported by dose-response estimates, have been due to chance, or whether true differences are obscured by an inadequate exposure index or group misclassification. In mid-1986, strong correlations between dioxin levels in fat tissue and serum were demonstrated by the CDC and other institutions. Because of these results, the Air Force is currently engaged in a collaborative study with CDC to determine whether serum dioxin levels vary significantly in the Ranch Hand population. Approximately 200 AFHS volunteers have supplied a pint of blood to be analyzed for dioxin at the CDC laboratories. If clear and meaningful 44 exposure findings are evident from this study, several additional studies are feasible: testing can be expanded to the entire study population and a meaningful exposure index based on total current TCDD body burden may be developed; and by means of archived AFHS serum samples from the Baseline study, it may be possible to calculate a reasonably precise half-life of TCDD in humans. These expanded studies will allow the estimation of body burdens of TCDD at the time of departure from SEA (assuming the absence of intervening vocational and recreational exposures). If, in fact, these potential studies become reality within the next 2 years, the fifth~year followup study data will be statistically analyzed using a more appropriate exposure index. In anticipation of this advance, the AFHS is currently collecting 280-350 ml of blood from all volunteers attendingthe fifth-year followup study. ADDITIONAL ANALYSES AND STUDIES As in the 1984 Baseline Report, not all of the measured dependentvariables were subjected to statistical analysis (e.g., prothrombin, leutinizing hormone, follicle stimulating hormone), largely because they were not within the bounds of the Air Force-prescribed analyses. Exploration of many of the unanalyzed variables is contemplated as time and resources permit. Similarly, many analytic opportunities to define possible symptom-clinical sign clusters or syndromes by multivariate analysis of variance techniques were passed over due to time and charter. Particularly challenging as an area of future work may be the changing relationships of some immunologicalvariables over time and the biological impact of these changes on the induction of diseases such as cancer. Likewise, future efforts to define shifting cardiovascular disease patterns are a logical extension of the rich longitudinal data base of the AFHS. Such efforts await future analysis and publication. The assessment of possible selection and participation bias has been addressed in a comprehensive manner in this report (see Chapter 5). The analyses and discussion suggest that statistical use of the total Comparison group (versus the Original Comparison group) is justified in this report, and that the impact of selection and participation biases have been minimal. As the followup studies continue, it is anticipated that a wealth of data on compliance-participation factors will be available for continued comprehensive bias analyses. In particular, it is hoped that more complete data will exist to examine the true differences in current health status between refusals and their replacements. As the data set grows over time, the bias analyses will become more comple~ and will have to deal with changing motivations of the participants to continue in this study. Such bias analyses and assessments will always be of great importance to this study as they ultimately set the bounds for an inference on herbicide causality. 45