October 3, 2006

The authors conducted factor analysis on survey data from 1,779 Persian Gulf War veterans. Their purposes were to: 1) determine whether factor analysis identified a unique "Gulf War syndrome" among veterans potentially exposed to chemical warfare agents; 2) compare the findings of factor analysis with those from an epidemiologic analysis of symptom prevalence; and 3) observe the behavior of factor analysis when performed on dichotomous data. The factor analysis identified three factors, but they were not unique to any particular deployment group. A unique pattern of illness was not found for the larger group of veterans potentially exposed to chemical warfare agents; however, veterans who had witnessed the demolition of chemical warfare agents at the Khamisiyah site in Iraq had a greater prevalence of dysesthesia. An analysis of the performance of dichotomous variables in factor analysis showed that the standard criteria used to determine the number of relevant factors and the dominant variables within them may be inappropriate. While Gulf War veterans appear to suffer an increased burden of illness, there is insufficient evidence to identify a unique syndrome in this population of deployed servicemen and women. Furthermore, the results provide evidence that factor analysis may make a limited contribution in this area of research. • For more than 10 years, scientists have debated the nature and etiology of illness among veterans of the Persian Gulf War. • Investigators have repeatedly observed an increased prevalence of many health symptoms among troops deployed in the conflict; however, a central question has been whether these symptoms represent a distinct medical entity that can be labeled "Gulf War syndrome. • In a recent study, McCauley et al. (16) reported on the health status of veterans who had been deployed within a 50-km radius of Khamisiyah. o They noted no increased risk of current self-reported symptoms among veterans deployed in the Khamisiyah area compared with those who had been deployed to the Gulf region but not to Khamisiyah. o Within the Khamisiyah group, however, veterans close enough to witness the demolition reported significantly more of 16 different symptoms within 2 weeks of the demolition than nonwitnesses, and all but three of these symptoms were consistent with exposure to organophosphate agents. o Eight years after the demolition, these same witnesses reported a significant excess of eight health-related symptoms, some of which could plausibly be related to long-term effects of low-dose exposure to chemical warfare agents. • This paper presents the results of further analysis of the McCauley et al. (16) data • Our first purpose was to subject the self-reported health data of Khamisiyah troops to a factor analysis to determine whether a unique pattern of symptoms or factors was present and, if so, whether the pattern differed from that of non-Khamisiyah troops and troops not deployed to the Gulf region. • Our second purpose was to compare the results of the factor analysis with those reported by McCauley et al., especially as they related to the apparent increase in the presence of certain current symptoms among veterans who witnessed the detonation. • Our last purpose was to examine how factor analysis behaves when the data are dichotomous. • Study Population o veterans on active or reserve duty in US Army or Nat’l Guard during Gulf War (jan 1, 1991 – march 31, 1991). o Random sampling of three groups: Those not deployed to gulf region (non-deployed) Those deployed to southwestern Asia, but not within 50km of Khamisiyah (deployed non-Khamisiyah) Thos deployed with 50km of Khamisiyah (deployed Khamisiyah) o Telephone interviews o 91% of 3200 contacted o 500 ineligible o 1833 interviews o 1779 useable interviews for analysis 516 non-deployed 610 deployed non-Khamisiyah 653 deployed Khamisiyah • 162 witnessed munitions detonations (witness subgroup) • 405 did not (non-witness subgroup) • Study Instrument o Adapted existing survey used in study of Gulf War veterans o Includes more information on troop movements neurologic symptoms o Reliability: reported elsewhere (????) o Two checklists Health symptoms within 2 weeks of Khamisiyah detonations (not used here) Current health symptoms within past month o Unclear the format of data??? • Statistical Analysis o Factor analysis: Exploratory factor analysis per group. Principal components extraction Varimax rotation Retained factors with eigenvalues greater than 1 Dominant items: loadings greater than 0.60. o Comparison of factor analysis results with prevalence of self-reported symptoms Not able to do subgroup analysis of witnesses Used hypothesis test to see if different Number of factors per veteran determined (???) Compared distribution of “factors” with subgroup o Dichotomous Variables Simulations to see how factor analysis performs with binary items Factor analysis performed on artificial data Eigenvalues and loadings examined