Table 5.5 below reveals that each case definition is slightly different. However, the core symptoms are cognitive and are the same or would be assumed to be the same (Axelrod and Milner 1997; Hom et al. 1997) in all but the Lange et al. study (2001) in which fatigue was the exclusive symptom. The results are summarized in Table 5.5.
Of the two primary studies, David et al. (2002) categorized the veterans as either ill (n = 151) or well (n = 188) on the basis of their SF-36 physical functioning scale scores. They then recategorized them more stringently on the basis of the Center for Disease Control and Prevention working definition of Fukuda et al.(1998) for reporting (n = 65) or not reporting (n = 33) Gulf War-related symptoms (Table 5.5, column 1), analyzing each separately. (These are additional analyses from the same David et al. (2002) study identified above as comparing veterans on the basis of their deployment status.) David et al. (2002) did not find overall differences after adjusting for age, education, estimated intelligence (NART), and depression (BDI score); but they did report some cognitive test differences before making the final Bonferroni correction for multiple comparisons (column 5 of Table 5.5).
Storzbach et al. (2001) compared 239 deployed veterans who reported having at least one symptom associated with the “Gulf War syndrome” (Table 5.5) with 112 nonsymptomatic deployed veterans in a case-control study. They reported poorer performance on three neurobehavioral tests (Oregon Dual Task Procedure [ODTP] errors and latency, Digit Span Backward, and Simple Reaction Time) after adjustment for age, education, AFQT scores, and a Bonferroni correction for multiple comparisons (Table 5.5). In the analysis, Storzbach et al. identified a group of “slow” responders (12% of the symptomatic group) who had very slow latencies in choosing answers on the memory component of the ODTP, although they made few errors. Anger et al. (1999) reported on the first 100 participants in whom this slow group was identified; see Table 5.5. The slow ODTP subgroup proved to have been responsible for the statistically significant differences between cases and controls in the larger group: removing them from the analysis virtually eliminated the differences. The effect size between the slow responders and the controls on the ODTP was very high (d = 2.9). The effect sizes for the performance differences of the slow ODTP group (n = 30) vs the non-symptomatic (n = 112) ranged from 0.7 to 2.9 (mean d = 1.5) (Table 5.5, column 5).
All other studies that compared symptomatic veterans with nonsymptomatic veterans are listed as secondary studies. They are outlined in the lower six rows of Table 5.5 and are described in more detail in Table 5.7. As can be seen in Table 5.5, some categorized veterans as symptomatic on the basis of subsets of the overall symptom constellation, and some defined symptomatic in a narrower way. Some of the studies were not well described, had design flaws, asked limited questions, and studied small samples (n = 8-48). Nonetheless, as noted above, they used similar or the same neurobehavioral tests and thereby allow an evaluation of the consistency or pattern of results.
Five of the six secondary studies reported performance differences between symptomatic cases and controls or between cases and standard scores. As can be seen in Table 5.5, only two secondary studies (Axelrod and Milner 1997; Sillanpaa et al. 1997) adjusted for age and education, and only one (Lange et al. 2001) adjusted for multiple comparisons. The study that did not report a difference (Sillanpaa et al. 1997) did adjust for age, education, and an estimate of exposure (not clearly described). Each study had additional weaknesses. Axelrod and Milner (1997) used standard scores instead of controls and did not mention whether examiners were blinded. The basis of subject selection was not clear in Goldstein et al. (1996), and the issue of blinding of examiners was not addressed in the publication. The basis of subject selection was