The Health of Former Prisoners of War: Results from the Medical Examination Survey of Former POWs of World War II and the Korean Conflict (1992)

The creation of the medical examination survey, whose data form the basis of this report, was largely a matter of opportunity. Although data were available from repatriation exams, and questionnaire-and records based investigations had provided four decades of follow-up data, no examination data had ever been collected from the entire, combined MFUA cohort of POWs and controls. It was not until the VA's POW protocol exam program began in 1983, largely as an outreach effort, that the opportunity to examine this large group of POWs (and later, controls) presented itself. The VA's POW program was the crucial factor in the design of this survey because it provided the extensive network of physicians and other health care professionals needed to collect the examination data. Moreover, because the program had already been begun as a clinical program, there would be no additional costs in using it for research purposes.

The VA's POW examination program had not, however, been designed for research use, and here the benefits of combining it with the MFUA's longitudinal study were manifest. The earliest directive from the VA concerning the research study, Circular 10-87-138, put it thus: ''In order to adapt this [VA] program for research purposes, it is necessary to collect protocol examination data on representative groups of POWs rather than to analyze data for only men who have presented themselves for examination." It was not until later (as directed under a separate VA circular) that control groups were added, because the proposed inclusion of non-POW veterans in the VA's examination program posed some difficulties and therefore was

not immediately accepted. Eventually, however, the arguments for the inclusion of non-POW controls were persuasive: it would further strengthen the research by providing important comparison data to help interpret the findings among POWs. In sum, the combination of the VA's POW outreach program and the MFUA's research program was designed to produce a research survey with the best features of both kinds of programs.

Yet the opportunity to apply the existing mechanisms of the VA examination program to the MFUA cohort was not without its accompanying drawbacks. For example, although the widespread network of VA medical facilities provided coast-to-coast geographic coverage, it came with a built-in lack of standardization of examination procedures. Despite the publication of VA directives, the actual conduct of examinations necessarily varied from place to place and from time to time, in part as a result of local administrative customs, but mostly because of the underlying variability among examiners in the way they performed a "typical" examination. The study thus lacked the common, rigidly specified procedures of many research protocols. Yet despite these limitations, a large mass of important examination data has been accumulated.

Before discussing the analysis of these data, some discussion of their quality is in order. The overall rate of completed, coded examinations, reported in Chapter 3, was disappointingly low—around 40–50% in POWs and 10–14% in controls. Some of the large difference in POW and control examination rates, in itself disquieting, is due to the fact that a number of POWs, the so-called volunteers, came in for examinations before the formal research program began. Because the VA's POW examination program began in 1983 and the research study began in 1986, there was ample opportunity for this to occur.

In general, however, at least in the previous morbidity follow-ups in 1967 (Beebe, 1975) and 1984 (Page, 1988), it has been the case that control response rates were lower than POW response rates. As evidence for this general statement, the same disparity in POW and control response rates is again seen in their responses to the supplemental psychological questionnaire mailing sponsored by the National Institute of Mental Health: questionnaire response rates were 25–30% for POWs and only 10–25% for controls. Interestingly, only relatively minor differences are apparent among POW groups or among control groups, except for the Korean conflict POWs and controls. These groups each had the highest response rates among POWs and controls.

Chapter 3 also included brief comparisons of demographic aspects and VA hospitalization rates for respondents and nonrespondents. There were

few appreciable demographic differences between POWs who completed exams and those eligible subjects who did not. Among those with completed exams were higher proportions of men with service in the Air Corps and with a higher level of education. WE respondents had a higher proportion of men who were single at entry into service and who had been inducted.

VA hospitalization data for subjects with completed exams, eligible subjects without completed exams, and ineligible subjects were obtained for 1969x1985, predating the examination survey; these data provide evidence from a separate source about differences among the three groups. There were no appreciable differences in VA hospitalization rates among the above three comparison categories in any of the POW or control groups. These data thus suggest no overwhelming differences in health, at least as it is measured by 26 years of VA hospitalization data, between respondents and nonrespondents.

It is not unusual to find differences between survey respondents and nonrespondents, typically on measures of education (Comstock and Helsing, 1973). In an earlier study of nonresponse bias for the 1984–1985 questionnaire follow-up, Page (1991) found that nonrespondents were generally older, less well educated, and of lower military rank; in addition, they had previously reported more prison camp symptoms and slightly less weight loss. Demographic data were then used to predict depressive symptoms, and it was shown that there were only small differences between observed scores for respondents and predicted scores for nonrespondents, suggesting that nonresponse bias was not a major factor. Although a similar analysis has not yet been undertaken for the examination data from this study, the lack of obvious nonresponse bias in the 1984–1985 questionnaire follow-up, together with a similar lack of such bias indicated in the demographic and VA hospitalization data, is somewhat reassuring.

Perhaps the signal feature of the examination data collection is its magnitude—the 1,067 examinations provide diagnostic data on more than 65,000 coded medical conditions. These exam data provided the basis, in Chapter 4, for the calculation of lifetime prevalence rates—that is, the probability that a given person has ever had (or still has) some specified disease. According to the exam data, POWs had higher lifetime prevalences than controls for a number of medical conditions: infectious disease (PWK), endocrine disease other than diabetes (PWP and PWK), psychoses (all but PWP), neuroses (all), neurological disease (PWP and PWK), urogenital disease (PWE), injuries and poisonings (PWE and PWEM), acute respiratory disease (PWK), digestive disease (PWK), skin disease (PWK), and other symptoms (PWK). Both PWP and PWK had an appreciably lower lifetime prevalence

of blood diseases. In general, however, prevalence levels were similar among POWs and not strikingly different, for the most part, from control rates.

Chapter 4 also introduced the first external comparison data, showing similar lifetime prevalence rates for selected conditions that were taken from Eberly and Engdahl's (1991) study of POW examinations at the Minneapolis Medical Center. In some instances, lifetime prevalence rates were quite comparable—for example, in the case of hypertension rates for all MFUA POWs, all Minneapolis POWs, all MFUA controls, and the general population. However, in several other cases–diabetes, myocardial infarction, and cerebrovascular accident–lifetime prevalence rates were apparently lower for Minneapolis POWs than for the three other groups. Even more obvious differences could be seen between prevalence rates for psychiatric conditions in the combined data for MFUA and Minneapolis POWs: rates for depression were apparently higher for MFUA POWs and rates of posttraumatic stress disorder (PTSD) lower.

Known differences in age, length of captivity, and harshness of treatment during captivity made it reasonable to examine lifetime prevalence data for MFUA POWs and controls and for Minneapolis POWs separately, by war theater. Table 4.4 displayed lifetime prevalence rates for selected Eberly and Engdahl diagnostic categories. The MFUA data showed a rough equality across POW groups for most of the conditions not directly linked to prison camp treatment (i.e., those other than dysentery, malaria, beriberi, and frozen feet). Indeed, in regard to medical conditions, most of the noteworthy differences between MFUA POWs and controls were concentrated in the categories of prison camp-related conditions, with hypertension (among PWP) and ulcer (among PWK) the only exceptions.

The Minneapolis POWs can also be compared with the MFUA POWs. MFUA POWs had notably higher rates for intermittent claudication (all groups) and dysentery (all groups), while the MFUA PWEM, who have no comparable Minneapolis POW group, also had higher rates of hypertension and frozen feet than the Minneapolis PWE. In addition, the MFUA PWK had higher lifetime prevalences than their Minneapolis counterparts for malaria, beriberi, and frozen feet. In summary, most of the striking differences between the MFUA and Minneapolis POWs were for prison camp-related conditions.

Compared with MFUA controls, MFUA POWs have appreciably higher lifetime prevalence rates of depressive disorder and generalized anxiety disorder (across all POW groups) and PTSD (for PWP and PWK only). MFUA POW rates are also uniformly higher than Minneapolis POW rates for depressive disorder; they are uniformly lower for PTSD and roughly equal for generalized anxiety disorder, with differences across war eras not as pronounced as for some medical conditions. Finally, among MFUA

controls, the Korean combat veteran controls (WK) apparently have much higher psychiatric morbidity than other control groups. It is worth noting again that these Korean controls differ from the WW II controls in having been selected from a group of men who had been wounded and returned to action; the Chapter 4 data show that although they were not prisoners, WK probably have increased lifetime psychiatric morbidity as a result of their war experience.

Chapter 5 addressed five basic questions, each of which had been posed before the study began. These questions were based on the preliminary findings of the 1984–1985 questionnaire follow-up as well as on limited findings from other, published studies. Question A asked whether psychiatric illness was higher in POWs than in controls; the answer was that there were appreciably higher rates of illness for several specific psychiatric conditions. In particular, the rates for PTSD were appreciably higher in POWs measured independently in each of three different ways: examination, Structured Clinical Interview for DSM-III-R (SCID), or questionnaire (the Mississippi scale). Rates of depressive disorder were similarly elevated, again measured independently in three ways: examination, Center for Epidemiologic Studies depression scale (CES-D) questionnaire, or Beck questionnaire. The finding of an appreciably higher rate of PTSD was particularly important, in that the 1984–1985 survey results had led to a hypothesis that PTSD was an underlying psychiatric diagnosis or comorbid condition associated with high rates of depressive symptoms. Additional data from the 90-item Hopkins Symptom Check List (SCL-90), however, suggest that psychiatric illness is not necessarily limited to PTSD and depressive disorder. The SCL-90 showed a wide range of psychiatric pathology across many disparate indices, with appreciably higher POW scores on a global symptom index. Psychiatric illness was most pronounced in PWK, PWEM, and PWP groups.

Question B asked whether there were important differences among measurement instruments for psychiatric illness. The answer here was that there were noteworthy differences among the various ways of measuring PTSD and depressive disorder. For PTSD, the exam and structured clinical interview agreed much more closely with each other than with the questionnaire, which had been designed and tested among Vietnam-era combat veterans. For depressive disorder, one scale (the CES-D) gave appreciably higher estimates of depressive disorder (although this scale is in the strictest sense a symptom scale); the other, the Beck scale, gave estimates that were much closer to the rates of clinical diagnosis. In every case, however, there were noteworthy differences between POWs and controls, regardless of how psychiatric illness was measured.

Question C asked whether earlier hospitalization rates were comparable to recent rates of VA hospitalization. In general, the answer was that there were apparently fewer differences between POWs and controls now than in

1967. A comparison of lifetime and current prevalence rates from the present exam also showed that differences in morbidity appear to have lessened over time.

Question D asked whether examination and earlier self-reported questionnaire data provided comparable estimates of illness prevalence. The answer was that medical conditions were self-reported by questionnaire much less frequently than they were noted during examination; typically, selfreported prevalence was only one-half to one-tenth the examination-based value. Some of this disparity appeared to be the result of limitations in the questionnaire design, such as a fixed number of blanks for responses, as well as of differences between physicians and questionnaire respondents in the use of medical terminology.

Question E asked whether the special, albeit small, subset of severely malnourished European prisoners, PWEM, differed from their European theater counterparts, PWE. The answer was that PWEM had an appreciably higher lifetime prevalence of dysentery, beriberi, frozen feet, peripheral nerve disease, and gastroenteritis. The latter two conditions, which are not as clearly related to prison camp treatment as the former, were studied in further diagnostic detail. Higher PWEM rates (compared with PWE) of peripheral nerve disease were concentrated in three specific categories: mononeuritis of the upper limb, mononeuritis of the lower limb, and hereditary and idiopathic neuropathy. Higher rates of gastroenteritis in PWEM were accounted for almost entirely by noninfectious gastroenteritis and colitis and not by irritable bowel syndrome, which has already been presumptively linked to military captivity.

The results of the analyses in Chapter 5 showed that there continue to be psychological aftereffects of military captivity as long as 45 years after repatriation and that these psychological illnesses are still the most striking sequelae of military captivity among WW II and Korean conflict POWs. Less evidence was uncovered concerning nonpsychological aftereffects, possibly because these effects were not as pronounced but also perhaps because the medical conditions were analyzed in broad categories to avoid missing any important but unanticipated findings. The use of these broad categories rather than more specific ones was a choice dictated by the relative paucity of specific medical hypotheses formulated at the time the study was designed.

In Chapter 6, a different sort of analysis was undertaken to complement the analyses of Chapter 5. As a first step, a specific set of medical conditions was selected, based on published reports of POW examinations at two VA medical centers, Minneapolis and Denver; current (that is, unresolved) prevalence rates for these conditions for POWs and controls were then compared. This analysis showed higher rates for POWs for the following conditions: ischemic heart disease (PWEM only), peripheral nerve disease

(PWP and PWK), ulcer (PWK only), gastroenteritis (PWEM and PWK), depressive disorder (PWP, PWE, PWEM, and PWK), PTSD (PWP, PWEM, and PWK), and generalized anxiety disorder (PWP, PWE, PWEM, and PWK). With the exception of malaria (in which case PWP rates were lower than WP rates), conditions showing noteworthy differences in current prevalence between a POW and control group were included in the final stage of analysis (discussed below).

In the next step of the process, the selected conditions were subjected to a second screening analysis to determine whether the prevalence of any of these conditions was associated with the specific factors that measured harshness of treatment as a POW, such as percentage of body weight loss. These analyses necessarily compared POWs with one another, avoiding the problems of earlier analyses that used small control samples. The use of internal controls in this analysis was thought to be statistically sensitive, that is, more apt to identify an association between medical conditions and earlier POW experience.

The second set of screening analyses looked first at the simple associations between the prevalence of the selected medical conditions and percentage of body weight lost, as well as at associations of prevalence with the number of reported prison camp medical symptoms (see Table 6.2), an overall measure of severity of treatment. POWs who reported more than a 35% body weight loss had appreciably elevated prevalences of intermittent claudication, arterial vascular disease, peripheral nerve disease, depressive disorder, and PTSD; they also had a lower prevalence of osteoarthritis. POWs who reported more prison camp symptoms compared with other POWs had appreciably higher prevalences of intermittent claudication, arterial vascular disease, peripheral nerve disease, depressive disorder, PTSD, generalized anxiety, and schizophrenia.

When the data were analyzed separately by POW group and body weight and prison camp symptoms were treated as continuous (rather than stratified) measures, many of these associations persisted. In addition, others were seen: asthma (PWP), peripheral nerve disease (PWP and PWE), intermittent claudication (PWK), arterial vascular disease (PWK and PWE), depressive disorder (PWP, PWE, and PWK), schizophrenia (PWK), PTSD (PWK), and generalized anxiety (PWP). Finally, a preliminary multivariate analysis, analyzing all POW data jointly but controlling for the different POW groups, found intermittent claudication, arterial vascular disease, peripheral neuropathy, osteoarthritis, depressive disorder, and PTSD to be strongly associated with percent weight loss; it found cerebrovascular disease, ischemic heart disease, ulcer, asthma, depressive disorder, PTSD, and generalized anxiety disorder to be strongly associated with prison camp symptoms.

The final analysis comprised the medical and psychiatric conditions for

which there were findings of either an appreciably higher prevalence in some POW group or a noteworthy association with some military captivity factor. The prevalence data for these conditions were reanalyzed using logistic regression to determine the joint effects of weight loss and prison camp symptoms; POW group differences were also taken into account. For this analysis, prison camp symptoms were further refined into three separate measures—the presence of edema, the number of visual symptoms (such as night blindness), and the number of other symptoms—and considered along with percent weight loss.

The logistic regression analyses showed that edema was associated with a higher prevalence of ischemic heart disease and peripheral nerve disease; visual symptoms were associated with higher prevalences of cerebrovascular disease, ulcers, asthma, and posttraumatic stress disorder; and other symptoms were associated with higher prevalences of intermittent claudication, gastroenteritis, depressive disorder, and generalized anxiety. Percent weight loss was associated with a higher prevalence of intermittent claudication and arterial vascular disease and strongly associated with a lower prevalence of osteoarthritis. In most cases the odds ratios, which estimate the size of the increased prevalences, were between 1.0 and 2.0, indicating relatively moderate elevations in prevalence. It should also be noted, however, that the estimated effects of visual and other symptoms increased gradually with each additional reported symptom. Thus, the cumulative effect on the relative odds for POWs with a large number of symptoms could be quite substantial.

The effects on prevalence of POW group, per se, were mostly limited to PWK, but PWP showed an appreciably elevated prevalence of PTSD, compared with PWE. For PWK, again compared with PWE, there were noteworthy associations with a higher prevalence of ulcer, PTSD, schizophrenia, and generalized anxiety, and noteworthy associations with a lower prevalence of ischemic heart disease, arterial vascular disease, and asthma. These latter associations with lower prevalences appear to be related to age.

Beebe GW. 1975. Follow-up studies of World War II and Korean war prisoners. II. Morbidity, disability, and maladjustments. Am. J. Epidemiol. 101:400–422.

Comstock GW, Helsing KJ. 1973. Characteristics of respondents and nonrespondents to a questionnaire for estimating community mood. Am. J. Epidemiol. 97:233–239.

Eberly RE, Engdahl BE. 1991. Prevalence of somatic and psychiatric disorders among former prisoners of war. Hosp. Commun. Psychiatry 42:807–813.

Page WF. 1988. A Report of the Longitudinal Follow-up of Former Prisoners of War of World War II and the Korean Conflict: Data on Depressive Symptoms from the 1984–1985 Questionnaire. Final report to the Veterans Administration under contract V101(93)P-1088. Medical Follow-up Agency, Institute of Medicine. Washington, DC.

Page WF. 1991. Using longitudinal data to estimate nonresponse bias. Soc. Psychiatry Psychiatr. Epidemiol. 26:127–131.