qualified nosologist who coded causes of death according to the ICD-9 without knowledge of the subjects’ deployment status.
Kang and Bullman (2001) reported four analyses. The first yielded unadjusted rate ratios calculated from crude death rates. The second used Cox proportional-hazards models to account for possible confounding and effect modification by selected covariates related to veterans’ risk of dying (or hazard) from specific causes from the time they entered the cohort. Adjusted rate ratios derived from the models were used to approximate relative risk. The third analysis compared cause-specific mortality in Gulf War veterans and nondeployed veterans with expected numbers of deaths in the overall U.S. population. Finally, changes in relative mortality in Gulf War veterans and nondeployed veterans during four 20-month followup periods were examined with a chi-squared test for trend. For each of followup periods beginning in January 1993 and extending to December 1997, an adjusted rate ratio was derived from the Cox proportional-hazards model after adjustment for age, race, marital status, branch of service, and type of unit. In the analyses, there was no excess all-cause-specific mortality and no excess mortality from all external causes combined, but there was an excess risk of death from motor-vehicle accidents in Gulf War veterans compared with nondeployed veterans that decreased steadily from the first followup (rate ratio 1.32, 95% CI 1.13-1.53) to the last followup (rate ratio 1.00, 95% CI 0.82-1.22). The chi-squared value (χ2= 7.53) indicated that there was a significant decreasing trend (p = 0.0061) in the risk of motor-vehicle accidental death with time since the Gulf War in contrast with a steadily increasing trend in disease-related causes of death in Gulf War veterans compared with nondeployed veterans; deaths from disease was about the same for deployed and nondeployed veterans by the end of the last followup period. The authors concluded that 7 years after the Gulf War, mortality in both Gulf War and nondeployed veterans was half that in their civilian counterparts.
A report from the Mental Health Advisory Team-III (MHAT-III) (MHAT 2006a) issued by the Office of the Surgeon General, U.S. Army Medical Command, is included as a primary source because it used a direct measure of deployment exposure in Iraq. The MHAT-III found that in 2005 the confirmed OIF suicide rate for soliders deployed in Iraq was 19.9 per 100,000 troops. That was compared with a suicide rate in the entire Army of 12.3 per 100,000 in the same year and an average suicide rate of 11.6 per 100,000 in 1995-2005 in the Army.
Eight publications were identified as secondary studies (Adams et al. 1998; Boscarino 2005; Kaplan et al. 2007; Kramer et al. 1994; Macfarlane et al. 2000, 2005; Price et al. 2004; Writer et al. 1996). Two publications by Macfarlane et al. (2000, 2005) identified veterans from the UK who served in the Gulf War and veterans who were not deployed to the region. Data on the deployed and era cohorts were obtained from the National Health Service but not confirmed with a supplemental source. The initial study (2000) by the investigators found that the deployed cohort had higher mortality from external causes than the era cohort. However, the difference was nonsignificant and was due primarily to accidents. The second study (2005) by the investigators found no differences in external causes of death between Gulf War and era veterans at followup in June 2004. The study was considered secondary because it did not confirm cause of death.
Writer et al. (1996) used the same retrospective cohort data as Kang and Bullman (1996). Cause-specific mortality from August 1990 though July 1991 was compared for Gulf War-deployed veterans and active-duty troops deployed elsewhere. All active-duty military deaths