BRITISH JOURNAL OF PSYCHIATRY (2003), 182, 391^403 REVIEW ARTICLE

Psychiatric disorder in veterans of the Persian Gulf DESERT SHIELD oror DESERTSHIELD oror GULF WAR oror GULF SYNDROME oror War of 1991 GULF WAR SYNDROME oror PERSIANPERSIAN GULF WAR oror PERSIAN GULF SYNDROME. References of identified Systematic review studies were searched for further studies. Specialist Gulf veterans’ illnesses research NICOLA J. STIMPSON, HOLLIE V. THOMAS, ALISON L. WEIGHTMAN, websites (US Department of Defense Center FRANK DUNSTAN and GLYN LEWIS for Deployment Health Research site and the Walter Reed Army Medical Center Gulf War database) and more-general Gulf web- sites were also searched for any additional Background Veterans ofthe Persian Since the end of the Persian Gulf War of references. Researchers who had expressed Gulf War of1991have reported symptoms 1991, its veterans have reported a range an interest in Gulf veterans’ illness research of health complaints attributed to service were contacted for any non-published in- attributed to their military service. during the war. These veterans report an formation. There was no restriction on the Aims ToToreview review all studies comparing increased prevalence of a whole range of identification of studies in terms of publica- common symptoms compared with other tion status or language. This search strategy the prevalence of psychiatric disorders in service personnel who were not deployed was first applied to data published up to the Gulf War veterans and in a comparison to the Gulf. end of 1998 (nn¼4156) and then repeated to group of service personnelnotdeployedto It is now widely recognised that the end of May 2001 (nn¼1231).1231). the Gulf War. exposure to combat and other wartime Studies were included if they contained experiences can have both short-term and data on veterans who had been deployed to MethodMethod Studies of military personnel long-term psychological effects. These the Gulf War on military, medical or peace- deployed to the Gulf published between psychological consequences are varied, but keeping grounds (i.e. those involved in the concept of post-traumatic stress dis- operations Desert Shield, Desert Storm, 1990 and 2001were identified from order (PTSD) has arisen to describe the Granby or Desert Peace). Any study design electronic databases.Reference lists and syndrome of intrusive thoughts, flashbacks, was eligible for inclusion provided that an websites were searched and key hyperarousal and numbing that can occur appropriate control or comparison group researchers were contacted for after exposure to any traumatic event, was included to compare the prevalence of information. AtotalA total of 2296 abstracts and including those common in wartime. psychiatric disorder. The Persian Gulf War was brief and The 5387 abstracts identified by the 409 complete articles were reviewed and there were relatively few casualties among original search were screened by N.J.S. data were extracted independently by the troops deployed on behalf of the United and the 2296 that remained eligible were two members ofthe researchteam. Nations. Nevertheless, a number of aspects examined by two members of the research of the war exposed service personnel to team to decide whether they might meet ResultsResults The prevalence of psychiatric traumatic and stressful events: these in- our inclusion criteria. Printed copies of disorderin 20 studies of Gulf WarWarveterans veterans cluded the risk of chemical and biological 409 papers were then obtained and examin- was compared with the prevalence in the warfare, exposure to combat, and dealing ed by two members of the research team to with prisoners and dead and wounded Iraqi confirm eligibility and extract data. comparison group. Prevalence of post- soldiers. This paper describes a systematic In our original search we also included traumatic stress disorder (PTSD) and review of studies that have compared the studies that compared ill and well Gulf common were higher in prevalence of psychiatric disorder in Gulf War veterans, but these were excluded the Gulf War veterans.Heterogeneity War veterans with its prevalence in a com- from the review reported here. Studies were between studies was significant, but all parison group who were not deployed to also excluded if they measured simulated the Gulf (non-Gulf veterans). exposures, if they measurednon-health- reported thisincreasedthis increased prevalence. related outcomes, or if the study population Conclusions Veterans ofthe Persian included inhabitants of the Persian Gulf METHOD states rather than deployed military, Gulf War reported an increased medical or peace-keeping personnel. prevalence of PTSD and common mental Data search All identified papers that fulfilled the disorder compared with other active Studies between January 1990 and May pre-stated inclusion criteria were cate- service personnel not deployed to the 2001 were identified from a range of elec- gorised by health outcome. Forty-nine tronic databases, including EMBASE, Med- studies included data on psychiatric dis- Gulf.These findings are attributable to the line, ASSIA, SIGLE, PsycINFO, CancerLit, order, 29 of which reported on Gulf War increase in psychologically traumatic HealthSTAR, Dissertation, Abstracts, veterans and an external comparison events in wartime. Current Contents, Health and Psychosocial group of non-Gulf War veterans. We Instruments, CINAHL and Biological further restricted the studies to those with Declaration of interest Funded bybyFunded Abstracts. Keywords used to identify the a limited range of outcomes concerned the UKMedical Research Council. studies were: DESERT STORM oror with psychiatric disorder (20 studies). The

391

Downloaded from https://www.cambridge.org/core. 02 Oct 2021 at 10:49:16, subject to the Cambridge Core terms of use. STIMPSON ET AL

outcomes we chose to include were as often difficult to establish from published 1998; Proctor et aletal, 1998; Stuart & Bliese, follows:follows: articles, and this – together with the wide 1998; Gray et aletal, 1999; Ishoy et aletal, 1999;,1999; variety of scales that were used – can intro- UnwinUnwin et aletal, 1999; Wolfe et aletal, 1999; Bar- (a)(a)PTSDPTSD diagnosed using a recognised duce difficulties in performing a quantita- tone, 2000; Kang et aletal, 2000; Steele, standardised assessment; tive synthesis. Using ratio measures to 2000; Cherry et aletal, 2001). We excluded (b)(b)commoncommon mental disorder: depression or estimate association should be less sensitive nine other studies that included data on anxiety diagnosed using a recognised to the different case definitions and psychiatric disorder in Gulf War veterans standardised assessment; or self- measures used in the constituent studies. but did not meet our inclusion criteria: five reported symptoms of depression repeated results already included, three did recorded on a checklist; RESULTSRESULTS not include any of the psychiatric outcomes (c)(c)problemsproblems related to alcohol misuse. defined above, and one compared Gulf vet- The systematic review process is shown in erans with reported illness with a compari- We have chosen to use the term ‘common Fig. 1. We identified 20 primary studies son sample (further details available from mental disorder’ (Goldberg & Huxley, that investigated the association between the authors upon request). 1992) to refer to the common symptoms deployment to the Gulf War and psychi- Table 1 summarises the studies we of depression and anxiety that are seen in atric disorder (Perconte et aletal, 1993; Sutker identified. All are best described as cross- the community and reflect the use of assess- et aletal, 1993, 1994; Stretch et aletal, 1996,1996aa,,bb;; sectional surveys. Some studies, for exam- ments such as the General Health Ques- Iowa Persian Gulf Study Group, 1997; ple those by Kang et aletal (2000), Unwin et aletal tionnaire (GHQ; Goldberg & Williams, Pierce, 1997; Stuart & Halverson, 1997; (1999) and Ishoy et aletal (1999), resemble 1988) and the Symptom Checklist (and its Goss Gilroy Inc., 1998; Holmes et aletal,, cohort studies, as the population was derivatives) (Derogatis et aletal, 1974;,1974; Derogatis, 1977; Derogatis & Spencer, 1982).1982).

Data extraction Data relating to the studies’ main hypo- theses and to methodological quality were extracted independently by two members of the research team. Information on the methodological quality of the individual studies included the response rate, the potential of selection bias in the sampling of the study participants, the potential bias in the measurement of outcomes, the availability of data on confounders, and any adjustment for such variables.

Statistical analysis Summary odds ratio and risk ratios were calculated with a random-effects model using the inverse variance method. The degree of heterogeneity was assessed using the chi-squared test within a fixed-effects model. All analyses were performed using the METAN command (Bradburn et aletal,, 1998) in Stata version 6 (StataCorp, 1999). We chose this approach because of the inherent heterogeneity in the data. In particular, we were combining studies with a variety of outcome measures. A random- effects model assumes that the studies in a meta-analysis are sampled from a distribu- tion of effect sizes, which are estimated from the data in the meta-analysis. In contrast, a fixed-effects model assumes that all the studies are sampled from a population with the same effect estimate. We chose to perform analyses on di- chotomous outcomes because the distribu- tion of scores from continuous scales is Fig. 11Fig. Flowchart of the systematic review.

392

Downloaded from https://www.cambridge.org/core. 02 Oct 2021 at 10:49:16, subject to the Cambridge Core terms of use. PSYCHIATRIC DISORDER IN GULF WAR VETERANS

defined in terms of ‘exposure’ to the and a 63% response rate in the non-Gulf Confounding Gulf War. However, these studies had War veterans sample. Goss Gilroy Inc. There was considerable variation in the little orlittleorno information on health status (1998) in the Canadian study reported extent to which the authors attempted to before deployment and therefore share response rates of 73% for Gulf War adjust for confounders. Many of the studies most ofmostofthe methodological limitations of veterans and 60% for non-Gulf War that selected from the military databases cross-sectional surveys. veterans.veterans. used a stratified sampling procedure and frequency-matched the non-Gulf veterans Sampling Measurement on some characteristics in order to adjust The sampling design of the studies varied. for confounding. Some studies included Most of the studies took place after there For example, Unwin et aletal (1999), Kang etet these variables in a multivariate model had been considerable publicity about ill- alal (2000), Goss Gilroy Inc. (1998), Ishoy when analysing their results, which was ness in Gulf War veterans. However, four et aletal (1999) and Cherry et aletal (2001) identi- probably necessary given the differential re- studies included here reported findings fied samples of service personnel from mili- sponse rate between the Gulf War veterans based upon surveys carried out within tary databases. The Unwin et aletal and CherryandCherry and non-Gulf veterans. The most thorough about a year of the end of the Gulf War: et aletal studies were of two independent adjustments were carried out by Unwin et aletal these studies were by Sutker et aletal (1993,(1993, samples drawn from the same UK military (1999). In particular, only Unwin et aletal andand 1994), Holmes et aletal (1998) and Stuart database. They employed stratified random Stuart & Bliese (1998) adjusted for marital & Halverson (1997). All reported a signifi- sampling in order to frequency-match the status. This is likely to be an important con- cant excess of psychopathological disorder characteristics of Gulf War veterans with founding variable, as single people usually within the Gulf War veterans. those who were on active duty at the have higher rates of common mental dis- Many of the studies used the Missis- time but were not deployed to the Gulf. order and were more likely to be deployed sippi scale (Keane et aletal, 1988) or modified These comparison groups are referred to the Gulf War – although not in Unwin versions thereof to assess symptoms of to as non-Gulf veterans; the proportion et aletal’s study, possibly because the UK mili- PTSD; this is a self-administered scale actually deployed to areas other than tary have fewer members who are never and it is generally assumed to be less the Gulf varied between studies. An alter- deployed on active service. Unwin et aletal valid than some of the more detailed ques- native sampling strategy used by two (1999) found that the odds ratio for being tionnaires. Some studies used their own studies, the Iowa Persian Gulf Study Group a case on the GHQ changed only from method for assessing PTSD based upon (1997) and Steele (2000), identified all 2.0 to 2.1 after adjustment, indicating that questions modelled on the DSM–III–R military service personnelwho had served there was little evidence of confounding (American Psychiatric Association, 1987) during the period of the Gulf War and by the variables identified in that study. criteria. A few studies used structured inter- who lived in one US state (Iowa and(Iowaand Results similar to these were obtained using views administered by clinicians (Sutker Kansas, respectively). Within this standard PTSD as the outcome. et aletal, 1994; Proctor et aletal, 1998; Wolfe survey design the investigators then com- et aletal, 1999), but these assessments would pared those who had been deployed to the have had the potential disadvantage of Meta-analysis Gulf with those who had not. Pierce introducing possible observer bias, as Post-traumatic stress disorder (1997) also used a military database but se- the interviewers would not have been lected only women from the US Air Force It was possible to conduct a meta-analysis masked to the participants’ deployment to study.tostudy. of 9 of the 11 studies that reported dichot- status.status. There were also more ad hocadhoc samplingsampling omous outcomes for PTSD. We were un- We identified 17 studies that included procedures that did not use the large able to use the data from Goss Gilroy Inc. data on common mental disorders. The national databases. For example, Holmes (1998) and Bartone (2000). The results self-administered questionnaire used most et aletal (1998), Gray et aletal (1999) and Sutker are summarised in Fig. 2. The overall frequently to assess common mental dis- et aletal (1993) compared Gulf War veterans summary estimate using a random-effects order, in eight studies, was the Hopkins and non-Gulf veterans within a selection model was an odds ratio of 3.17 (95% Symptom Checklist or Brief Symptom of units. Some studies also chose a small CI 2.16–4.65), indicating an increased Inventory (Derogatis et aletal, 1974; Derogatis, number of military bases without any risk in Gulf War veterans. There was signif- 1977; Derogatis & Spencer, 1982; 22 apparent justification for inclusion (Proctor icant heterogeneity (ww ¼29.4, d.f.29.4,d.f.¼8,8, Derogatis & Melisaratos, 1983). This et aletal, 1998; Wolfe et aletal, 1999).,1999). PP550.0001). In particular, the two large scale was reported either as a continuous studies by Unwin et aletal and GrayandGray et aletal dif-dif- outcome or used to define a ‘case’ of com- fered: the former found an OR of 3.5 and Response rates mon mental disorder. The other studies the latter on OR of 1.8. The summary esti- Response rates also varied considerably used a variety of methods to assess common mate for the risk ratios was 2.9 (95% CI between studies (Table 1). Of most import- mental disorder, from self-reported symp- 2.0–4.2).2.0–4.2). ance is that the response rate of the Gulf toms of depression (Proctor et aletal, 1998;,1998; War veterans was higher than that of the GrayGray et aletal, 1999; Ishoy et aletal, 1999; Kang non-Gulf veterans in studies that reported et aletal, 2000; Steele, 2000), other self-admi- Common mental disorder the response rates separately. This could nistered scales such as the GHQ (Unwin(Unwin We were able to perform a meta-analysis introduce a biased comparison. For exam- et aletal, 1999), to lengthyclinician-adminis- on 11 of the studies that reported on ple, Unwin et aletal (1999) had a 70% tered structured interviews (Wolfe et aletal,, the prevalence of common mental disor- response rate in the Gulf War veterans 1999).1999). der (Fig. 3). Two studies used the same

393393

Downloaded from https://www.cambridge.org/core. 02 Oct 2021 at 10:49:16, subject to the Cambridge Core terms of use. STIMPSON ET AL rank, because of used to control for gender, age, marital status, ethnicity, current life stress scale. NGV group from USA omitted error in table In addition to the matching, the ORs were adjusted for other confounders that were statisti- cally significantly associated with outcome Controlled for age using analysis of variance. Possibly frequency- matched for service status and parent Confounding Multiple regression was used to control for rank, gender, age, marital status, ethnicity, current life stress scale. NGV group from USA omitted because of error in table In addition to the matching, the ORs were adjusted for other confounders that were statisti- cally significantly associated with outcome Controlled for age using analysis of variance. Possibly frequency- matched for service status and parent Confounding 11 TT respectively respectively 22 TT Return rate was 31% G 03NV60.3% Response rate: GWV 73.0% NGV 60.3%NGV andand Started with 638 women, traced 525, of whom 92% and 87% responded at response bias Selection or Return rate was 31% Multiple regression was Response rate: GWV 73.0% Started with 638 women, traced 525, of whom 92% and 87% responded at Selection or response bias 22 TT or atorat 11 11 TT TT riee unlesservicemen otherwise stated riee unlesservicemen otherwise stated 3.67, 95% 3.67, 95% ¼ 1.78; 95% CI 1.78; 95% CI ¼ 2.69 (95% 2.69 (95% ¼ 22 TT :NGV1.66,GWV1.87 : NGV 1.66, GWV 1.87 : NGV 1.84, GWV, 2.02 : NGV 1.84, GWV, 2.02 22 11 Prevalence: GWV 32.1% NGV 17.3% GWV had higher levels of minor depression adjusted for rank (OR 1.51^2.11) and major depression adjusted for rank and income (OR CI 3.04^4.44) compared with NGV No significant association with alcohol misuse PTSD OR CI 1.59^4.26) adjusted for income TT or at Mean scores on depression item (no s.d. given): TT ‘No significant differences’ with any of the measures used. Unclear whether at GWV 24% NGV 15% Unclear whether at Main results Prevalence: GWV 32.1% NGV 17.3% GWV had higher levels of minor depression adjusted for rank (OR 1.51^2.11) and major depression adjusted for rank and income (OR CI 3.04^4.44) compared with NGV No significant association with alcohol misuse PTSD OR CI 1.59^4.26) adjusted for income or at Mean scores on depression item (no s.d. given): ‘No significant differences’ with any of the mental health measures used. Unclear whether at GWV 24% NGV 15% Unclear whether at Main results PTSDPTSD PTSDPTSD Common mental disorder Common mental disorder Common mental disorder Common mental disorder Common mental disorder Common mental disorder Brief Symptom Inventory. Cases defined on the Global Severity Index Minor and major depression (PRIME^MD) Alcohol misuse (measure not specified) PTSDbasedonsymptom report (PCL^M) and somatisation based on Hopkins Symptom Checklist. Depression item from RAND Questionnaire Mississippi scale Brief Symptom Inventory. Cases definedontheGlobalSeverity Index Minor and major depression (PRIME^MD) Alcohol misuse (measure not specified) PTSDbasedonsymptom report (PCL^M) Sub-scales for depression, anxiety and somatisation based on Hopkins Symptom Checklist. Depression item from RAND Questionnaire Mississippi scale Main outcomes 4years 4 years Sub-scales for depression, anxiety was 2 years was 2 years 22 11 Study period Jan. and Feb. 19931993 TT 1997 Survey June^ Dec. 1997Dec. after war and TT Jan. and Feb. Survey June^ after war and Study period Main outcomes 3113) 3113) ¼ nn 3439) 3439) ¼ nn 991991 279 279 ¼ ¼ nn nn Sample Stratified random sample, female US Air Force GWV and ‘other deployed’ veterans. Sampling frame from US DoD manpower data centre. Stratified by active duty/reserve, parent/non- parent identified from the Canadian Department of National Defense and a ‘deployed elsewhere’ control group matched for gender, age, reserve/regular ( All Canadian GWV ( Randomly selected units from all US Army National Guard and Reserves taken from the DoD manpower data centre; NGV were deployed to USA or Germany GWVGWV NGV Stratified random sample, female US Air Force GWV and ‘other deployed’ veterans. Sampling frame from US DoD manpower data centre. Stratified by active parent/non- duty/reserve, parent identified from the Canadian Department of National Defense and a ‘deployed elsewhere’ control group matched for gender, age, reserve/regular ( All Canadian GWV ( Randomly selected units from all US Army National Guard and Reserves taken from the DoD manpower data centre; NGV were deployed to USA or Germany NGV Sample 331) 331) ¼ nn 153) with those 153) with those ¼ nn GWV ( deployed elsewhere ( survey: Cross-sectional postal rs-etoa survey: Cross-sectional postal Longitudinal study comparing GWV ( deployed elsewhere ( survey: Cross-sectional postal rs-etoa survey: Cross-sectional postal Study design Studies reporting the prevalence of psychiatric disorder in Gulf War veterans (GWV) and veterans not deployed to the Gulf (NGV). All studies are of US s Studies reporting the prevalence of psychiatric disorder in Gulf War veterans (GWV) and veterans not deployed to the Gulf (NGV). All studies are of US s able 1a l e 1 Ta b l e 1Tab Samples from national military databases Pierce (1997) Goss Gilroy Inc. (1998)(1998) Stuart & Bliese (1998)(1998) Reference Samples from national military databases Pierce (1997) Longitudinal study comparing Goss Gilroy Inc. Stuart & Bliese Reference Study design

394

Downloaded from https://www.cambridge.org/core. 02 Oct 2021 at 10:49:16, subject to the Cambridge Core terms of use. PSYCHIATRIC DISORDER IN GULF WAR VETERANS Confounding Controls matched for gender, age and profession. Non- response meant GWV more likely to be older males Potential confounders (age, marital status, rank, education, employment, still serving or discharged, smoking and alcohol consumption) are adjusted for using logistic regression. The samples were frequency-matched by age, service, No adjustment made for potential confounders except for the matching variables, gender and service status service status, gender, rank and fitness. Analyses restricted to men only Confounding Controls matched for gender, age and profession. Non- response meant GWV more likely to be older males Potential confounders (age,maritalstatus, rank, education, employment, still serving or discharged, smoking and alcohol consumption) are adjusted for using logistic regression. The samples were frequency-matched by age, service, No adjustment made for potential confounders except for the matching variables, gender and service status service status, gender, rank and fitness. Analyses restricted to men only response bias Response rate: GWV 83.6% NGV 57.8% Selection or Response rate: GWV 70.4% Bosnia 61.9% Era 62.9% Responders were significantly older and more likely to be in service, but did not differ on SF^36 Response rate: GWV 75% NGV 64% Non-responders were more likely Selection or response bias Response rate: GWV 83.6% NGV 57.8% Response rate: GWV 70.4% Bosnia 61.9% Era 62.9% Responders were significantly older and more likely to be in service, but did not differ on SF^36 Response rate: GWV 75% NGV 64% Non-responders were more likely 2.6 (95% 2.6 (95% 1.61.6 ¼ ¼ 2.12.1 3.8 (95% 3.8 (95% ¼ ¼ r,Oea OR era, ORera, r,Oea OR era, ORera, onaOBsi OR Bosnia ORBosnia ona Rona OR Bosnia, ORBosnia, 4.9) v. v. v. v. v. v. v. v. Prevalence: GWV 22.6% NGV 10.4% Prevalence: GWV 39.2% Bosnia 26.3% Era 24.0% GWV (95% CI 1.4^1.8) GWV (95% CI 1.9^2.4) Adjusted ORs reported 4.7 Prevalence: GWV 13.2 Bosnia 4.7Bosnia Era 4.1 GWV Prevalence: GWV 36% NGV 22% Rate difference 14% (95% CI 13.9^14.1)13.9^14.1) Main results CI 1.9^3.4)CI1.9^3.4) GWV CI 2.8^CI2.8^4.9) Adjusted ORs reported Prevalence: GWV 22.6% NGV 10.4% Main results Prevalence: GWV 39.2% Bosnia 26.3% Era 24.0% GWV (95% CI 1.4^1.8) GWV (95% CI 1.9^2.4) Adjusted ORs reported Prevalence: GWV 13.2 Era 4.1 GWV Prevalence: GWV 36% NGV 22% Rate difference 14% (95% CI GWV Adjusted ORs reported PTSDPTSD Common mental disorder Common mental disorder Common mental disorder Common mental disorder Common mental disorder Common mental disorder 3onGHQ^12 3onGHQ^12 55 Self-reported symptom of depression or sadness during previous 12 months that started during or after Gulf War Case was (Goldberg & 1988) Williams, Items from Mississippi scale. Authors’ own case definition basedonDSM^IV Self-reported symptom of ‘depression’ Self-reported symptom of depression or sadness during previous 12 months that started during or after Gulf War Main outcomes Case was (Goldberg & 1988) Williams, Items from Mississippi scale. Authors’ own case definition basedonDSM^IV Self-reported symptom of ‘depression’ Study period Feb. 1997^Jan. 1998 Aug. 1997^11 Nov. 1998 Not mentioned Feb. 1997^Jan. 1998 Study period Main outcomes Aug. 1997^11 Nov. 1998 Not mentioned et aletal 1815 1815 ¼ nn 24082408 ¼ nn 686686 32843284 11 441 11 441 231231 9476 9476 ¼ ¼ ¼ ¼ ¼ nn nn nn nn nn NGO aid workers compared with NGV group frequency- matched on gender, age and ‘profession’ using a Danish military database GWV NGV Bosnia and era samples frequency- matched for age, service, service status, gender, rank and fitness level Different sample from Cherry Sample Era veterans Bosnian veterans (2001)(2001) Stratified random sample drawn from UK military database: GWV NGV Frequency-matched on gender and service status from the DoD manpower data centre GWV All Danish military GWV and NGO aid workers compared with NGV group frequency- matched on gender, age and ‘profession’ using a Danish military database GWV NGV Sample Bosnia and era samples frequency- matched for age, service, service status, gender, rank and fitness level Different sample from Cherry Era veterans Bosnian veterans Stratified random sample drawn from UK military database: GWV NGV Frequency-matched on gender and service status Stratified random sample taken from the DoD manpower data centre GWV rs-etoa survey: Cross-sectional postal Cross-sectional survey All Danish military GWV and Study design rs-etoa survey: Cross-sectional postal Cross-sectional survey Stratified random sample taken )) 11 (1999)(1999) continued continued 19)19)Cosscinlsurvey Cross-sectional (1999)(1999) 20)20)Cosscinlsurvey Cross-sectional (2000)(2000) (( et aletal et aletal et aletal Ishoy UnwinUnwin able 1a l e 1 Ta b l e 1Tab Reference Kang Ishoy Reference Study design Kang

395395

Downloaded from https://www.cambridge.org/core. 02 Oct 2021 at 10:49:16, subject to the Cambridge Core terms of use. STIMPSON ET AL Adjusted for stratifi- cation variables: military service (regular/National Guard) age, gender, race, branch of service and rank and Odds ratio adjusted for gender, age, income andincome education level Confounding No adjustment apart from matching variables: gender, age, service and rank Adjusted for stratifi- cation variables: military service (regular/National Guard) age, gender, race, branch of service and rank Odds ratio adjusted for gender, age, education level Confounding No adjustment apart from matching variables: gender, age, service and rank Selection or G 3NV73% Response rate: GWV 78.3% NGV 73%NGV Response rate: 65% (3138 eligible) Gulf and female veterans were more likely to respond response bias G 21NV82.1% Response rate: GWV 84.3% NGV 82.1%NGV Response rate: GWV 78.3% Response rate: 65% (3138 eligible) Gulf and female veterans were more likely to respond Selection or response bias Response rate: GWV 84.3% NGV, NGV, 0.020.02 v. v. ¼ PP 12.6%,12.6%, 9.0%9.0% 4.6% 4.6% v. v. ¼ 2.99 (95% CI 2.07^4.31) 2.99 (95% CI 2.07^4.31) ¼ ¼ (estimated from tables): (estimated from tables): Prevalence: GWV 23% NGV 9% OR Prevalence (estimated from tables): GWV 1.9% NGV 0.8% Prevalence difference in depression GWV 6.0 (95% CI 4.0^7.9) Prevalence of depression (estimated from tables): GWV NGV Prevalence of alcohol misuse 17.4% ensoeMa score: Mean score:Mean GWV 6.3 NGV 3.7 (from figure) Main results Prevalence: GWV 23% NGV 9% OR Prevalence (estimated from tables): GWV 1.9% NGV 0.8% Prevalence difference in depression GWV 6.0 (95% CI 4.0^7.9) Prevalence of depression (estimated from tables): GWV NGV Prevalence of alcohol misuse 17.4% Main results GWV 6.3 NGV 3.7 (from figure) PTSDPTSD Common mental disorder Common mental disorder Common mental disorder Common mental disorder Common mental disorder Common mental disorder 50)50) 55 PTSD Checklist ^ Military (case defined as score Self-reported ‘Feeling down or depressed’ Brief Symptom Inventory ^ case definition not stated CAGE questionnaire for alcohol misuse ^ case definition not stated efrpre symptomSelf-reported ‘Feeling unhappy and depressed’: visual analogue scale PTSD Checklist ^ Military (case defined as score Self-reported ‘Feeling down or depressed’ Brief Symptom Inventory ^ case definition not stated CAGE questionnaire for alcohol misuse ^ case definition not stated Self-reported symptom ‘Feeling unhappy and depressed’: visual analogue scale Main outcome Study period Sept. 1995^May 19961996 e.Ag 1998 Feb.^Aug. Dec. 1997^Sept. 19991999 Sept. 1995^May e.Ag 1998 Feb.^Aug. Dec. 1997^Sept. Study period Main outcome 409409 ¼ nn 8014) and NGV 8014) and NGV 1435, NGV 18961896 1435, NGV 1799 1799 ¼ ¼ ¼ nn ¼ nn nn nn (1999)(1999) 3900) frequency-matched 3900) frequency-matched ¼ nn DoD manpower data centre used to create a stratified random sample from 16 566 military personnel from Kansas. Stratified by reservist and gender GWV NGV Stratified random sample from the UK Ministry of Defence database GWV ( (( by gender, age, service and rank different sample from Unwin et aletal DoD manpower data centre used to create a stratified random sample from 28 968 military personnel from Iowa: GWV Sample DoD manpower data centre used to create a stratified random sample from 16 566 military personnel from Kansas. Stratified by reservist and gender GWV NGV Stratified random sample from the UK Ministry of Defence database GWV ( by gender, age, service and rank different sample from Unwin DoD manpower data centre used to create a stratified random sample from 28 968 military personnel from Iowa: GWV Sample telephone survey Cross-sectional survey Cross-sectional telephone survey Cross-sectional telephone survey Cross-sectional survey Cross-sectional telephone survey Study design )) continued continued (( et aletal Steele (2000) able 1a l e 1 Ta b l e 1Tab CherryCherry (2001)(2001) Iowa Persian Gulf Study Group (1997)(1997) Reference Representative samples restricted to a single geographical area Steele (2000) Cross-sectional Iowa Persian Gulf Study Group Reference Study design Representative samples restricted to a single geographical area

396

Downloaded from https://www.cambridge.org/core. 02 Oct 2021 at 10:49:16, subject to the Cambridge Core terms of use. PSYCHIATRIC DISORDER IN GULF WAR VETERANS Confounding All troops with previous combat experience were GWV GWV older than NGV No adjustment for confounding No adjusted results presented All personnel had returned to the USA and not sought mental health treatment services. GWV were more likely to be younger and higher military rank, although correlational analyses found no significant difference between these measures and scores on mood measures Confounding All troops with previous combat experience were GWV GWV older than NGV No for adjustment confounding No adjusted results presented All personnel had returned to the USA and not sought mental health treatment services. GWV were more likely to be younger and higher military rank, although correlational analyses found no significant difference between these measures and scores on mood measures Selection or response bias The 60 respondents were selected from 124 in a survey of unspecified design. Response rate uncertain Response rate: 95% (approximate) Response rate: GWV 70.3% No response rate given for NGV Selection or response bias The 60 respondents were selected from 124 in a survey ofunspecified design. Response rate uncertain Response rate: 95% (approximate) Response rate: GWV 70.3% No response rate given for NGV G .%G 4.0% GWV 15.5% NGV 4.0%NGV EDV 3.9% Main results Higher prevalence in GWV (13%). No case found in NGVNGV Prevalence: 15.1%; on GSI:onGSI: GWV 27.9% EDV 11.5% BGV 15.1%;BGV on 16.7% BDI: GWV 26.9% EDV 23.1% NGV 16.7%NGV Prevalence: The GWV who experienced ‘high war-zone stress’ scored significantly more highly on the BDI and STAI anxiety measures Main results Prevalence: GWV 15.5% EDV 3.9% Higher prevalence in GWV (13%). No case found in Prevalence: GWV 27.9% EDV 11.5% on BDI: GWV 26.9% EDV 23.1% The GWV who experienced ‘high war-zone stress’ scored significantly more highly on the BDI and STAI anxiety measures PTSDPTSD Common mental disorder Common mental disorder Common mental disorder Common mental disorder Common mental disorder Common mental disorder 0.7) or SCL^90^R; 0.7) or SCL^90^R; 1010 55 55 Cases defined on Global Severity Index (score Beck Depression Inventory case score Administered after psycho- educational presentations and discussion Mississippi scale Beck Depression Inventory; anxiety: State^Trait Anxiety Inventory; Health Symptom Checklist (includes 9 items from Hopkins Symptom Checklist) SCID^PSCID^P Main outcomes Cases defined on Global Severity Index (score Beck Depression Inventory case score Administered after psycho- educational presentations and discussion Mississippi scale Beck Depression Inventory; anxiety: State^Trait Anxiety Inventory; Health Symptom Checklist (includes 9 items from Hopkins Symptom Checklist) Depressive disorder NOS; Study period Not provided Troops were assessed 4^10 months after completion of Desert Storm 12 months after return from Desert Storm Not provided Study period Main outcomes Troops were assessed 4^10 months after completion of Desert Storm 12 months after return from Desert Storm Depressive disorder NOS; 40) 40) 20) but ‘activated 20) but ‘activated 439 439 126 126 ¼ 2626 ¼ ¼ nn ¼ nn nn ¼ nn nn EDV Selected from ‘various’ reserve units in western Pennsylvania: NGV GWV National Guard and Army Reserve units in Louisiana: 215 troops deployed to the Gulf and 60 troops from the same units who were activated but not deployed overseas grave registration duties GWV ( NGV ( remained stateside’ Sample EDV Selected from ‘various’ reserve units in western Pennsylvania: NGV GWV Sample Participants were drawn from 5 National Guard and Army Reserve units in Louisiana: 215 troops deployed to the Gulf and 60 troops from the same units who were activated but not deployed overseas 60 Army reservists assigned grave registration duties GWV ( NGV ( remained stateside’ survey survey Cross-sectional Cross-sectional Study design Cross-sectional survey Participants were drawn from 5 rs-etoa surveyCross-sectional 60 Army reservists assigned )) (1993) 19)19)Cosscinlsurvey Cross-sectional (1994)(1994) (1993) survey Cross-sectional continued continued (( et aletal et aletal et aletal Other sampling strategies PercontePerconte (1993)(1993) Ta b l e 1 Reference Sutker Sutker Other sampling strategies Ta b l e 1 Reference Study design Sutker Sutker

397

Downloaded from https://www.cambridge.org/core. 02 Oct 2021 at 10:49:16, subject to the Cambridge Core terms of use. STIMPSON ET AL No adjustment was made for potential confounders. Current life stresses were found to be strongly related to PTSD symptoms Hierarchical multiple regression used to adjust for age, rank, service branch, race, education, drinking and smoking No apparent adjustment for confounders Used logistic regression but insufficient details to decide what adjustments were made Confounding No was adjustment made for potential confounders. Current life stresses werefoundtobe strongly related to PTSD symptoms Hierarchical multiple regression used to adjust for age, rank, service branch, race, education, drinking and smoking No apparent for adjustment confounders Used logistic regression but insufficient details to decide what adjustments were made Confounding Response rate: 31% Response rate: 31% Response rate: 31% Response rate: 31% Response rate: unknown Response rate: GWV 57.2% NGV 42.3% Selection or response bias Response rate: unknown Response rate: GWV 57.2% NGV 42.3% Selection or response bias all sub-scales (and after adjustment) 1.6% Prevalence: GWV 8.6% NGV 1.6%NGV GSI mean (s.d.): GWV, male 0.7 (0.7), female 0.8 (0.7) Bosnia veterans, male 0.6 (0.6), female 0.7 (0.6) No cases in NGV 7.3% Prevalence: GWV 11.5% NGV 7.3%NGV Significant increase in mean scores on GSI in GWV Higher prevalence in GWV. Main results GWV significantly higher on all sub-scales (and after adjustment) Prevalence: GWV 8.6% GSI mean (s.d.): GWV, male 0.7 (0.7), female 0.8 (0.7) Bosnia veterans, male 0.6 (0.6), female 0.7 (0.6) Higher prevalence in GWV. No cases in NGV Prevalence: GWV 11.5% Significant increase in mean scores on GSI in GWV Main results PTSDPTSD PTSDPTSD 70 70 55 Common mental disorder Common mental disorder Common mental disorder Common mental disorder Common mental disorder Common mental disorder Research PTSD algorithm (9 items from IES) and 8itemsfromBSI Global Severity Index of the Brief Symptom Inventory on Global Severity Index of SCL^90^R Brief Symptom Inventory Case defined as score Defined using SCID^P Walter Reed Army Institute of Walter Reed Army Institute of Research PTSD algorithm (9 items from IES) and 8itemsfromBSI Global Severity Index of the Brief Symptom Inventory on Global Severity Index of SCL^90^R Brief Symptom Inventory GWV significantly higher on Case defined as score Defined using SCID^P Main outcome GWV were were studied from Nov. 1990^Mar. 1991; NGV wereNGV sampled at 2 time points; May^Sep. 1991 and Jan.^Mar. 19961996 1993 11 months after end of hostilities 19931993 Study period GWV were studied from Nov. 1990^Mar. 1991; sampled at 2 time points; May^Sep. 1991 and Jan.^Mar. 1993 11 months after end of hostilities Study period Main outcome 1254 male, 1254 male, ¼ nn 10ml,10male, 2180 male,2180 296 296 1524 1524 1524 1524 210 2512 2512 210 2512 2512 ¼ ¼ ¼ ¼ ¼ ¼ ¼ nn nn nn nn nn nn nn 182 female 184 female, serving in 182 female 184 female, serving in ¼ ¼ sampling strategy GWV nn Bosnia veterans nn Bosnia May^July 1996 NGV NGV NGV Various units from Hawaii and Pennsylvania who either deployed to the Gulf or did not deploy GWV All members of an Air National Guard Unit GWV Various units from Hawaii and Pennsylvania who either deployed to the Gulf or did not deploy GWV Sample No details provided of sampling strategy GWV Bosnia veterans onaMyJl 1996 Bosnia May^July NGV NGV NGV Various units from Hawaii and Pennsylvania who either deployed to the Gulf or did not deploy GWV All members of an Air National Guard Unit GWV Various units from Hawaii and Pennsylvania who either deployed to the Gulf or did not deploy GWV Sample Cross-sectional survey Cross-sectional survey: postal Cross-section survey: postal Cross-sectional survey: postal Cross-sectional survey No details provided of Cross-sectional survey: postal Cross-section survey: postal Cross-sectional survey: postal Study design )) continued continued (( et aletal et aletal et aletal )) )) bb aa tat&tat& Stuart &Stuart Halverson (1997) Stretch (1996(1996 Holmes (1998)(1998) Stretch (1996(1996 Ta b l e 1 Reference Halverson (1997) Stretch Holmes Stretch Ta b l e 1 Reference Study design

398

Downloaded from https://www.cambridge.org/core. 02 Oct 2021 at 10:49:16, subject to the Cambridge Core terms of use. PSYCHIATRIC DISORDER IN GULF WAR VETERANS Prevalence estimates account for Odds stratification. ratios adjusted for age, gender and education with logistic regression Adjusted for sampling design to reflect distribution of gender and reported health symptoms. No other adjust- ments were made Confounding Prevalence estimates account for Odds stratification. ratios adjusted for age, gender and education with logistic regression Adjusted for sampling design to reflect distribution of gender and reported health symptoms. No other adjust- ments were made Confounding Selection or Response rate: New England 62% New Orealns 38% NGV 51% Participants were recruited after taking part in a previous study whose response rate was 78% NGV not very comparable Response rate: GWV 30^42% NGV 51% of unit GWV non-responders tended to be younger and unmarried. The GWV groups were quite different from the NGV group response bias Response rate: New England 62% New Orealns 38% NGV 51% Participants were recruited after taking part in a previous study whose response rate was 78% NGV not very comparable Response rate: GWV 30^42% NGV 51% of unit GWV non-responders tended to be younger and unmarried. The GWV groups were quite different from the NGV group Selection or response bias . o e60for New 6.0 for New6.0 ¼ Prevalence, CAPS: New England 5% (8/148) New Orleans 8% (4/58) NGV 0% Prevalence, Mississippi: New England 8.1% New Orleans 7.6% NGV not stated GWV 6.8% NGV 1.7% England, 3.9 for New Orleans Major depressive disorder: New England GWV 6.6% New Orleans GWV 4.5% NGV 0% Prevalence: New England 22.6% New Orleans 5.8% NGV 1.6% OR Main results Prevalence, CAPS: New England 5% (8/148) New Orleans 8% (4/58) NGV 0% Prevalence, Mississippi: New England 8.1% New Orleans 7.6% NGV not stated Prevalence: GWV 6.8% NGV 1.7% England, 3.9 for New Orleans Major depressive disorder: New England GWV 6.6% New Orleans GWV 4.5% NGV 0% Prevalence: New England 22.6% New Orleans 5.8% NGV 1.6% OR Main results PTSD PTSD PTSD PTSD 90) 90) Prevalence: Common mental disorder Common mental disorder Common mental disorder Common mental disorder 55 CAPS Mississippi scale for Desert Storm war zone personnel. CAPS possibly subject to observer bias SCID used to define DSM^III^R major depressive disorder CAPS and SCID administered by trained clinicians Blindness to exposure status not mentioned Self-reported ‘frequent periods of feeling depressed’ Mississipi scale (case CAPS Mississippi scale for Desert Storm war zone personnel. CAPS possibly subject to observer bias SCID used to define DSM^III^R major depressive disorder CAPS and SCID administered by trained clinicians Blindness to exposure status not mentioned Self-reported ‘frequent periods of feeling depressed’ Mississipi scale (case Main outcome Study period 1994^19961994^1996 Spring 1994 to autumn 1996 Study period Main outcome Spring 1994 to autumn 1996 48) (same 48) (same (1998)(1998) ¼ 48) 48) nn 4)ad4)and 148) and148) 6 n n6adan 56) and an56)and ¼ et aletal ¼ nn ¼ nn nn 66), and an NGV 66),andanNGV 186) and New 186) and New ¼ ¼ nn nn Sample NGV comparison group from an ambulance unit ( GWV from New England (Fort Devens) ( New Orleans ( study as Proctor GWV from Fort Devens, New England ( Orleans ( comparison group from an air ambulance unit ( NGV comparison group from an ambulance unit ( Stratified random samples of GWV from New England (Fort Devens) ( New Orleans ( study as Proctor Stratified random samples of GWV from Fort Devens, New England ( Orleans ( comparison group from an air ambulance unit ( Sample Cross-sectional survey Cross-sectional survey Cross-sectional survey Stratified random samples of Cross-sectional survey Stratified random samples of Study design )) continued continued (( et aletal et aletal Wolfe (1999)(1999) (1998)(1998) Proctor able 1a l e 1 Ta b l e 1Tab Reference Wolfe Proctor Reference Study design

399399

Downloaded from https://www.cambridge.org/core. 02 Oct 2021 at 10:49:16, subject to the Cambridge Core terms of use. STIMPSON ET AL Statistically significant after adjustment for number of stressful life events and ‘hardiness’‘hardiness’ Confounding GWVs younger, more often male and less educated No adjustments for confounders Statistically significant after adjustment for number of stressful life events and Confounding GWVs younger, more often male and less educated No adjustments for confounders ,1970).,1970). et aletal Response rate: approximately 50% response bias Estimated overall response rate: 53% Selection or Response rate: 50% approximately Selection or response bias Estimated overall response rate: 53% , 1974); stress Clinical PTSD, Structured SCID, post-traumatic disorder; , 1974); stress Clinical PTSD, Structured SCID, post-traumatic disorder; 2.6 (95% CI 1.5^4.4) 2.6 (95% CI 1.5^4.4) et al et al ¼ GWV 13.8 (9.7) NGV (USA) 4.0 (7.2) NGV (Germany) 9.3 (9.7) GWV had statistically significantly increased scores on all 5 dimensions of the Hopkins Symptom Checklist Prevalence of self-reported depression: GWV 6.8% NGV 2.8% OR Prevalence: GWV ^ 15.2% NGV 9.0% OR 1.8 (95% CI 1.3^2.5) Mean (s.d.) Main results Main results Mean (s.d.) GWV 13.8 (9.7) NGV (USA) 4.0 (7.2) NGV (Germany) 9.3 (9.7) GWV had statistically significantly increased scores on all 5 dimensions of the Hopkins Symptom Checklist Prevalence of self-reported depression: GWV 6.8% NGV 2.8% OR Prevalence: GWV ^ 15.2% NGV 9.0% OR 1.8 (95% CI 1.3^2.5) ae G,nngvrmna raiain G,o-ufveteran; NGV,non-Gulf NOS, organisation; cale; not NGO, non-governmental otherwise econd comprisedpersonnel who were deployedin Bosnia (many of whomwere not serving econd who personnel in comprised were deployed Bosnia (many of whom were not serving PTSDPTSD PTSDPTSD Common mental disorder Common mental disorder Five dimensions of Hopkins Symptom Checklist Self-reported symptom of depression Impact of Events Scale PTSD screen (items from DSM^IV) Main outcome Five dimensions of Hopkins Symptom Checklist Self-reported symptom of depression Impact of Events Scale PTSD screen (items from DSM^IV) ,1990); DoD, Department of Defense; EDV, Europe-deployed veteran; GHQ^12,12-item General Health Questionnaire (Goldberg & , 1990); DoD, of Department veteran; & (Goldberg General GHQ^12,12-item Health Questionnaire Defense; EDV,Europe-deployed et al et al Study period Sept. 1994^June 1995 ^ otsafter 4^6 months the end of the Gulf War Study period Main outcome Sept. 1994^June 1995 ^ otsafter 4^6 months the end of the Gulf War , 1993); Primary Care PRIME^MD, of Evaluation Mental (Spitzer Disorders , 1993); Primary Care PRIME^MD, of Evaluation Mental (Spitzer Disorders et aletal 527 389389 527 970 381 (236 to USA, 970 381 (236 to USA, ¼ ¼ ¼ ¼ nn nn nn nn NGV construction workers) who remained in service after Desert and Storm/Shield were servinginoneoftwolarge Seabee centres. Selected using the DoD manpower database GWVGWV Sample GWVGWV NGV 145 to Germany) Sample NGV Active duty Seabees (naval construction workers) who remained in service after Desert and Storm/Shield were serving in one of two large Seabee centres. Selected using the DoD manpower database Six army reserve medical units NGV 145 to Germany) , 1961); CAPS, Clinician Administered PTSD Scale (Blake , 1961); CAPS, Clinician Administered PTSD Scale (Blake et al et al , 1990); SCL^90^R, 1977); Checklist (Derogatis, Symptom SF^36, 36-item Short Form Anxiety (Spielberger Health State^Trait Inventory STAI, Survey; , 1990); SCL^90^R, 1977); Checklist (Derogatis, Symptom SF^36, 36-item Short Form Anxiety (Spielberger Health Inventory State STAI, ^Trait Survey; et al et al Study design Cross-sectional survey Active duty Seabees (naval Cross-sectional survey Six army reserve medical units )) continued continued 19)19)Cosscinlsurvey Cross-sectional (1999)(1999) (( et aletal atthetimeoftheGulfWar). 1.This study included two non-Gulf veteran groups: the first, the era group, comprised personnel who were serving in the military at the time, and the s Interview for DSM^III^R (Spitzer specified; stress PCL^M, Checklist disorder Post-traumatic ^ (Weathers Military ilas 1998);Williams, Index GSI,Global Severity (Derogatis & Spencer,1982; Derogatis & 1983); Melisaratos, GWV,Gulf War veteran; IES, Impact of NGV,non-Gulf veteran; Event NOS, organisation; not Scale; NGO, non-governmental otherwise able 1a l e 1 Ta b l e 1Tab BDI, Beck (Beck Inventory Depression Reference Gray Bartone (2000) atthetimeoftheGulfWar). 1.This two study included non-Gulf veteran groups: the first, the era who personnel comprised group, were in serving the at military the time, and the s Interview for DSM^III^R (Spitzer specified; stress PCL^M, Checklist disorder Post-traumatic ^ (Weathers Military ilas 1998);Williams, Index GSI,Global Severity (Derogatis & Spencer,1982; Derogatis & 1983); Melisaratos, GWV,Gulf War veteran; IES, Impact of Event S BDI, Beck Depression Inventory (Beck Reference Study design Gray Bartone (2000) survey Cross-sectional

400400

Downloaded from https://www.cambridge.org/core. 02 Oct 2021 at 10:49:16, subject to the Cambridge Core terms of use. PSYCHIATRIC DISORDER IN GULF WAR VETERANS

Inc. (1998) stated that there was no statisti- cally significant association between alco- hol misuse and deployment. The Iowa study (Iowa Persian Gulf Study Group, 1997) reported an increased prevalence of alcohol misuse measured by the CAGE questionnaire (Ewing, 1984).

DISCUSSION

The results of our systematic review and meta-analysis indicate an increased preva- lence of PTSD and common mental disor- der in service personnel who had been deployed to the Persian Gulf War. The size of this effect was somewhat larger for PTSD, with an OR of 3.2 (95% CI 2.2– 4.7) compared with 2.0 (95% CI 1.9–2.1) Fig. 22Fig. Forest plot of the results for post-traumatic stress disorder.Odds ratios by deployment.Values above1 for common mental disorder. indicate increased risk in Gulf War veterans Publication bias We adopted a thorough search strategy but – as in all systematic reviews – may have failed to identify some studies. We are also aware that other studies on this topic are in progress and have yet to report their findings. It is difficult to assess the ef- fect of any publication or citation bias in our data, given the small number of studies that reported data in a form permitting meta-analysis. A funnel plot for the ‘com- mon mental disorder’ outcome suggested that there was an underrepresentation of small studies finding no association be- tween deployment to the Gulf and disorder. However, these small studies would not have had a major impact on the summary odds ratio, despite suggesting that it might Fig. 33Fig. Forest plot of the results for common mental disorder.Odds ratios by deployment.Values above1 be a slight overestimate. The summary esti- indicate increased risk in Gulf War veterans. mate was dominated by the two large studies.studies. sample, but one (Wolfe et aletal, 1999) re- It should be noted that the studies by ported results from the Structured Clinical KangKang et aletal (2000) and Unwin et aletal Interview for DSM–III–R (Spitzer et aletal,, (1999) accounted for 90% of the variance Sample selection 1990) and the other (Proctor et aletal,, weights in the meta-analysis. The other A critical part of these designs is the 1998) presented results from self-reported studies therefore had little influence on comparability of the deployed and non-non- symptoms of depression. The summary es- the summary estimate. deployed troops. Some of the studies used timate was an odds ratio of 2.04 (95% CI A funnel plot of the standard error of military databases and took care to ensure 1.94–2.15), irrespective of whether the the estimate against the size of effect that their sample was representative of both data from either of these studies were suggested that there were fewer small non- Gulf War veterans and the comparison excluded, indicating an increased risk of significant findings than would be expected. group. It is likely that the characteristics common mental disorder in the deployed This would not have had much influence on of troops selected for deployment systema- service personnel. Despite the variation the findings, given the presence of a number tically differed from those of other active between studies in the outcome used, of large studies. service personnel who were not deployed. there was no statistical evidence to sup- This could be less marked for the UK mili- port heterogeneity in this sample using tary service, in which almost everyone is odds ratios (heterogeneity test ww22¼9.39,9.39, Alcohol misuse likely to be deployed on active duty. Poten- d.f.d.f.¼10,10, PP¼0.5). The summary estimate There was little evidence concerning alco- tial confounding factors include gender, for risk ratio was 1.8 (95% CI 1.6–2.0). hol misuse or dependence. Goss Gilroy fitness level and marital status, along with

401401

Downloaded from https://www.cambridge.org/core. 02 Oct 2021 at 10:49:16, subject to the Cambridge Core terms of use. STIMPSON ET AL

other aspects (such as propensity for risk- depressive symptoms. Despite this variation PTSD. An increased rate of psychiatric dis- taking) that are more difficult to measure. in measurement methods, there was little orders would therefore be expected in Gulf It is also likely that within individual units, evidence of heterogeneity in the estimates War veterans, although this does not dimin- the reasons for choosing people for deploy- for common mental disorder. Studies that ish the importance of this morbidity in af- ment would lead to a greater selection bias used the longer semi-structured interviews fecting veterans many years after returning than in studies sampling from national might have introduced observer bias, given from the conflict. databases, in which whole units would the difficulty in ‘blinding’ the interviewers. What is less clear is how these findings have been selected. In contrast, there was evidence of heteroge- relate to the issue of Gulf War illnesses. It is difficult to be sure about the effect neity for the PTSD estimates. In particular, Gulf War veterans have reported a wide of selection on the results reported here. UnwinUnwin et aletal (1999) reported a larger effect variety of symptoms, aside from psychiatric Some authors have suggested a ‘healthy than did Gray et aletal (1999), although both symptoms. Unwin et aletal (1999), in the UK warrior’ effect, that the deployed have bet- reported a significant increase in prevalence study, reported an increased prevalence of ter underlying health. On the other hand, in the Gulf War veterans. Gray et aletal re-re- a whole range of symptoms after having single people, who are more likely to have stricted their sample to naval construction adjusted for the increased prevalence of been deployed (at least in some studies; workers, so the different result might common mental disorder in the Gulf War ProctorProctor et aletal, 1998), tend to have poorer merely have reflected the different experi- veterans. This supports the view that some mental health (Kessler et aletal, 1994; Jenkins ences of this group of service personnel. It other factors must be contributing to ill- et aletal, 1997). None of the studies had any in- should also be noted that the Unwin et aletal nesses in these veterans, in addition to any dependent information about the mental study used a UK military cohort in which increase in psychiatric disorder. health of participants before the Gulf War almost all the non-Gulf War veterans com- Psychiatric disorder is common, dis- and so were not able to take any account parison group would have been deployed abling and burdensome. It is an important of this factor. on active service at one time or another. source of disability after war, yet this Five of the 20 studies were carried out isisoftenoften inadequately recognised and ac- Non -response bias within 12 months of the end of the war knowledged. Developing more-effective and at a time when publicity concerning means of preventing and treating psychi- The studies that reported response rates illness in Gulf War veterans was minimal. atric disorder in service personnel is an according to deployment status all found All these studies reported a statistically important priority for future research. that the Gulf War veterans had a higher significant increase in psychopathological response rate. It is likely that the publicity disorders in Gulf War veterans. These early surrounding illnesses in Gulf War veterans ACKNOWLEDGEMENTS studies tended to be less robust from a increased the relevance of a questionnaire methodological point of view than the later We thank Simon Wessely and Matthew Hotopf for about health effects to respondents who ones: the samples were less representative, comments on an earlier draft of the manuscript.We had been deployed to the Gulf. This response rates were lower and the studies thank the Department of Information Services at differential response rate could introduce smaller in size. In contrast, the later and University of Wales College of Medicine for a systematic bias. assistance in obtaining references for this review. often more robust studies could have been Some studies have reported that non-non- subject to a reporting bias following responders tended to have poorer mental publicity about illnesses in Gulf War REFERENCES health than those who responded (Williams veterans. In conclusion, it appears unlikely & Macdonald, 1986) although Unwin et aletal American Psychiatric Association(19 (1987) 87) Diagnostic that a reporting bias could have led to the (1999) in a more intensive follow-up of and Statistical Manual of Mental Disorders (3rd edn, findings reported in the constituent studies. revised)(DSM^III^R).Washington,DC:APA.revised) (DSM^III^R).Washington, DC: APA. non-respondents did not find a statistically significant increased risk of common men- Bartone, P.T. (2000)(2000) Hardiness as a resiliency factor tal disorder. Kang et aletal (2000) also com- for United States forces in the Gulf War. In Posttraumatic Illnesses in Gulf War veterans Stress Intervention: Challenges, Issues and Perspectives pared those who responded to the later We found that veterans deployed to the (ed. D. Paton), pp. 115^133. Springfield, IL: Charles mailings with those who returned the first C. Thomas.Thomas.C. Persian Gulf War reported more PTSD mailshot. They did not find that the later and more symptoms of common mental Beck, A.T.,Ward, C. H., Mendelsohn, M., et aletal (19 61) respondents had poorer self-rated general An inventory for measuring depression. Archives ofofArchives disorder than did service personnel who health. On balance, it is unlikely that the General Psychiatry,, 44,561571.,561571. had not been deployed to this war. In- differential response rate seen in these Blake, D.,Weathers, F., Nagy, L., et al (19 9 0) Clinician creased rates of PTSD have often been studies could have explained such a large Administered PTSD Scale. Boston, MA: National Center reported after conflicts and can be attribu- for PTSD,Behavioral Sciences Division. association as that reported. ted to the increased likelihood of psycho- Bradburn, M. J., Deekes, J. J. & Altman, D. G.(19 (1998) 9 8) logically traumatising events during METAN ^ an alternative meta-analysis command. Stata Outcome measurements wartime. The increased rates of other psy- Te chnical B ulletin,, 44, 4^15.,4^15. The majority of studies relied on self-self- chiatric symptoms might just be a conse- Cherry, N., Creed, F., Silman, A., et aletal (2001) Health reported symptoms to assess the prevalence quence of the same process. There is and exposures of Gulf war veterans. of psychiatric disorder. Some of the studies evidence that psychologically traumatic Part 1:The1: The pattern and extent of ill health. Occupational and Environmental Medicine,, 58, 291^298.,291^298. used well-recognised and validated mea- events also lead to an increase in other psy- sures of psychiatric disorder, but others (in- chological symptoms, particularly anxiety, Derogatis, L. R.(19 (1977) 7 7) The SCL^90^R Manual 1: Scoring, Administration and Procedures for the SCL^90. cluding some of the larger studies) reported in addition to the symptoms more specifi- Baltimore, MD: Johns Hopkins University School of results from a single question asking about cally associated with the syndrome of Medicine.Medicine.

402402

Downloaded from https://www.cambridge.org/core. 02 Oct 2021 at 10:49:16, subject to the Cambridge Core terms of use. PSYCHIATRIC DISORDER IN GULF WAR VETERANS

__ & Melisaratos, N. (1983) The Brief Symptom Inventory: an introductory report. Psychological MedicineMedicine,, 13, 595^605. CLINICAL IMPLICATIONS

__ & Spencer, P.M.P. M. (1982) The Brief Symptom Inventory: Administration, Scoring and Procedures Manual. && Veterans of the Gulf War report more post-traumatic stress disorder and more Baltimore, MD: Johns Hopkins University School of depression and anxiety than do war veterans not deployed to that conflict. Medicine,Medicine,Clinical Clinical Psychometrics Research Unit. && __ , Lipman, R. S., Rickels, K., et aletal (19 74) The Service in a war zone leads to an increase in symptoms many years afterwards. Hopkins Symptom Checklist (HSCL): a self-report symptom inventory. Behavioural Science,, 1919,1^15. && The presence of psychiatric symptoms probably does not explain the increased Ewing, J. A.(19 (1984) 8 4) Detecting alcoholism: the CAGE prevalence of other somatic symptoms reported by Gulf War veterans. questionnaire. JAMAJAMA,, 252,1905^1907. Goldberg,D.&Huxley,P.(1992)Goldberg,D.&Huxley,P.(19 92) Common Mental LIMITATIONS Disorders: A Biopsychosocial Approach. London:.London: Routledge. && Most of the large studies were conducted after public concern about illnesses in

__ & Williams, P. (198(1988) 8) The User’s Guide to the Gulf War veterans had been voiced. General Health Questionnaire.Windsor: NFER ^ Nelson. && Goss Gilroy Inc.(19 (1998) 9 8) Health Study of Canadian Forces The studies relied upon self-reported information about psychiatric symptoms. Personnel Involved in the 19911991Conflict Conflict in the Persian Gulf && (vol.1).Ottawa: Goss Gilroy Inc. http://www.dnd.ca/ Some of the larger studies used non-standard methods of assessing psychiatric menu/press/Reports/Health/health ____ study ____ e ____ vol1____ symptoms.symptoms. TO C. htmhtmTOC.

Gray,G.Gray, G. C., Kaiser,K.Kaiser, K. S., Hawksworth, A.W.,A. W., et aletal (19(1999) 9 9) Increased postwar symptoms and psychological morbidity among US Navy Gulf War veterans. American Journal of Tropical Medicine and Hygiene,, 60,758^766., 758^766.

Holmes,D.T.,Tariot,P.N.&Cox,C.(1998)Holmes,D.T.,Tariot,P.N.&Cox,C.(19 9 8) NICOLA J. STIMPSON, PhD, HOLLIE V.THOMAS, DPhil, ALISON L.WEIGHTMAN, PhD, FRANK DUNSTAN, Preliminary evidence of psychological distress among PhD,University of Wales College of Medicine,Cardiff; GLYN LEWIS, FRCPsych,,UK reservists in the Persian Gulf War. Journal of Nervous and Mental Disease,, 186186,166^173. Correspondence: Professor Glyn Lewis,Cotham House,Cotham Hill,Bristol BS6 6JL,UK.E-mail: Horowitz, M.,Wilner,N. & Alvarex,W.(19 (1979) 79) ImpactImpact glyn.lewisglyn.lewis@@bristol.ac.uk of Event Scale: a measure of subjective stress. Psychosomatic Medicine,, 4141, 209^218.,209^218. (First received 23 May 2002, final revision 18 October 2002, accepted 30 October 2002) Iowa Persian Gulf Study Group(19 (1997) 9 7) Self-reported illness and health status among Gulf War veterans. A population-based study. JAMAJAMA,, 277, 238^245. Ishoy,T., Suadicani, P., Guldager, B., et aletal (1999)(1999) State Proctor, S. P.,White, R. F.,Wolfe, J., et aletal (19 9 8) Stuart,Stuart,J.A.&Bliese,P.D. J. A. & Bliese, P.D.(19 (1998) 9 8) The long-term of health after deployment in the Persian Gulf ^ the Health status of Persian Gulf War veterans: self- effects of operation desert storm on the psychological Danish Gulf War Study. Danish Medical Bulletin,, 46,, reported symptoms, environmental exposures and the distress of US army reserve and national guard veterans. 416^419.416^419. effect of stress. International Journal of Epidemiology,, 2727,, Journal of Applied Social Psychology,, 28,1^22.,1^22. 1000^1010. Jenkins, R., Lewis, G., Bebbington, P., et aletal (19 9 7) TheThe __ &Halverson,R.R.(1997)& Halverson, R. R. (1997) The psychological status National Psychiatric Morbidity Surveys of Great Britain: Spielberger,C. D., Gorsuch, R. L. & Lushene, R. E. of US Army soldiers during recent military operations. initial findings from the Household Survey. Psychological (1970)(1970) STAI: Manual for the State^TraitState ^Trait Anxiety Inventory.. Military Medicine,, 162162, 737^743.,737^743. MedicineMedicine,, 2727, 775^790.,775^790. Palo Alto, CA: Consulting Psychologists Press. Sutker, P. B.,Uddo, M., Brailey, K., et aletal (19 93) Kang, H. K., Mahan, C. M., Lee, K.Y., et aletal (2000) Spitzer, R. L.,Williams, J. B.W., Gibbon, M., et aletal War-zone trauma and stress-related symptoms in Illnesses among United States veterans of the Gulf War: (19 9 0) User’s Guide for the Structured Clinical Interview for Operation Desert Shield/Storm (ODS) returnees. a population-based survey of 30,000 veterans. Journal ofofJournal DSM ^ III ^ R: SCID.Washington, DC: American Psychiatric Journal of Social Issues,, 49,33^50. Occupational and Environmental Medicine,, 42,491^501., 491^501. Association.Association. __ ,, __ ,, __ ,, et aletal (19 94) Psychopathology in war-zone Keane, T. M., Caddell, J. M. & Taylor, K. L.(19 (1988) 8 8) deployed and non-deployed operation desert storm __ ,, __ ,,Kroenke,K., Kroenke, K., et aletal (19 94)94)(19 Utility of a new Mississippi Scale for Combat-Related Posttraumatic troops assigned registration duties. Journal of Abnormal procedure for diagnosing mental disorders in primary Stress Disorder: three studies in reliability and validity. PsychologyPsychology,, 103, 383^390. Journal of Consulting and Clinical Psychology,, 5656, 85^90.,85^90. care.The PRIME^MD1000 study. JAMAJAMA,, 272272,, 1749^1756. Unwin, C., Blatchley, N., Coker,W., et aletal (1999)(19 9 9) Kessler, R. C., McGonagle, K. A., Zhao, S., et aletal Health of UK servicemen who served in Persian Gulf (19 94) Lifetime and 12-month prevalence of DSM ^ III ^ R StataCorp(19 (1999) 9 9) Stata Statistical Software.College War.War. Lancet,, 353,,169^178. 169^178. psychiatric disorders in the United States. Archives ofofArchives Station,TX: Stata Corporation. General Psychiatry,, 5151,,8^19. 8^19. Weathers, F., Litz, B.T., Heran, D. S., et aletal (19 93) TheThe Steele, L.L.Steele, (2000) Prevalence and patterns of Gulf War PTSD Checklist (PCL): Reliability,Validity, and Diagnostic Lewis, G., Churchill, R. & Hotopf, M.(19 (1997) 9 7) illness in Kansas veterans: association of symptoms with Systematic reviews and meta-analysis. Psychological UtilityUtility. San Antonio,Antonio,TX: TX: International Society of characteristics of person, place and time of military Traumatic Stress Studies. MedicineMedicine,, 2727,,3^7. 3^7. service.service. American Journal of Epidemiology,, 152, 992^1002.,992^1002. Perconte, S. T.,Wilson, A. T., Pontius, E. B., et aletal Williams, P. & Macdonald, A.(19 (1986) 8 6) The effect of (19 93) Psychological and war stress symptoms among Stretch, R. H., Bliese, P.P.D., D., Marlowe, D. H., et aletal non-response bias on the results of two-stage screening deployed and non-deployed reservists following the (19 9 6 aa)) Psychological health of Gulf War-era military surveys of psychiatric disorder. Social Psychiatry,, 2121,, Persian Gulf War. Military Medicine,, 158, 516^521. personnel. Military Medicine,, 161161, 257^261.,257^261. 182^186.

Pierce,Pierce,P.F. P.F.(19 (1997) 9 7) Physical and emotional health of Gulf __ ,Marlowe,D.H.,Wright,K.M.,et aletal (19 9 6b) Wolfe,J.,Proctor,S.P.,Erickson,D.J.,et aletal (1999)(1999) War veteran women. Aviation, Space and Environmental Post-traumatic stress disorder symptoms among Gulf Relationship of psychiatric status to Gulf War veterans’ MedicineMedicine,, 68, 317^321.,317^321. War veterans. Military Medicine,, 161161, 407^410. health problems. Psychosomatic Medicine,, 61,532^540.

403

Downloaded from https://www.cambridge.org/core. 02 Oct 2021 at 10:49:16, subject to the Cambridge Core terms of use.