potential bias due to attrition, only members of the study population who remained on active duty for at least half the study period were included in the multivariate models.
A second hospitalization study expanding on Gray et al. (1996) study compared the postwar records of Gulf War-deployed veterans (n = 552,111) and nondeployed veterans (n = 1,479,751) from DoD’s hospital-discharge data from August 1991 to April 1996 to search for admissions for unexplained illnesses in military hospitals (Knoke and Gray 1998). The study defined unexplained illnesses as diagnoses in 77 ICD-9 diagnostic categories that comprised ill-defined conditions. The study examined only first hospitalizations to avoid overcounting medical conditions that required repeated hospitalizations of the subset of patients who had at least one unexplained illness coded on a discharge summary. Up to eight discharge diagnoses were examined per hospitalization. The authors found that deployed active-duty military members were less likely to have been hospitalized for unexplained illnesses than nondeployed (RR 0.93, 95% CI 0.91-0.96) (Knoke and Gray 1998). That finding included adjustment for a variety of covariates and removed the effect of participation in the CCEP after June 1994. Participants in the CCEP were more likely to have been hospitalized only for evaluation. This study has the advantage of a large sample that allowed detection of even minimal effects. Its major limitations are its inclusion of only active-duty personnel and its inability to detect illnesses that did not warrant hospitalization.
Although the previous studies demonstrated no increase in unexplained illness among active-duty Gulf War veterans, Gray et al. (2000) sought to expand their investigation of Gulf War veterans to include reserve and separated military personnel who may not have been treated in a DoD facility. They investigated hospitalization data from DoD, VA, and nonfederal hospitals in California for 1991-1994. Hospitalization rates could not be directly compared among the three sources, because of the unreliability of state-of-residence data in DoD and VA datasets. Therefore, PMRs of hospital-discharge diagnoses (14 diagnostic categories from ICD-9) were compared for deployed vs nondeployed veterans. For VA hospitals, but not for DoD or California hospitals, the PMR for the ICD code of symptoms, signs, and ill-defined diseases was increased for deployed vs nondeployed veterans (PMR 1.24, 95% CI 1.16-1.33).
A previous IOM report (IOM 2006) has carefully described and evaluated studies of unexplained illness and increased symptom self-reporting. Cluster or factor analysis has been used by several researchers (for example, Fukuda et al. 1998) to determine whether the many symptoms reported constituted a new syndrome or a variant of an existing syndrome. However, the 2006 IOM report concluded that outcomes based primarily on symptoms or self-reports constituted “no unique syndrome, unique illness, or unique symptom complex in deployed Gulf War veterans. Veterans of the Gulf War report higher rates of nearly all symptoms or sets of symptoms than their nondeployed counterparts; 29% of veterans meet a case definition of ‘multisymptom illness,’ as compared with 16% of nondeployed veterans” (IOM 2006).
Two studies of PTSD and unexplained illness were identified. In a nested case-control study drawn from a large, population-based study of Iowa veterans of the Gulf War, Barrett et al. (2002a) investigated the relationship between PTSD and perceived physical health. Of the 53 veterans who screened positive for PTSD (37 deployed and 16 nondeployed), over 50% had symptoms that corresponded with ill-defined conditions according to the ICD-9 compared with less than 10% of the 3629 veterans without PTSD. The study was conducted by telephone interview 5 years after the Gulf War. The prevalence of PTSD among the Gulf War-deployed