Context.— Gulf War (GW) veterans report nonspecific symptoms significantly more
often than their nondeployed peers. However, no specific disorder has been
identified, and the etiologic basis and clinical significance of their symptoms
Objectives.— To organize symptoms reported by US Air Force GW veterans into a case
definition, to characterize clinical features, and to evaluate risk factors.
Design.— Cross-sectional population survey of individual characteristics and
symptoms and clinical evaluation (including a structured interview, the Medical
Outcomes Study Short Form 36, psychiatric screening, physical examination,
clinical laboratory tests, and serologic assays for antibodies against viruses,
rickettsia, parasites, and bacteria) conducted in 1995.
Participants and Setting.— The cross-sectional questionnaire survey included 3723 currently active
volunteers, irrespective of health status or GW participation, from 4 air
force populations.The cross-sectional clinical evaluation included 158 GW
veterans from one unit, irrespective of health status.
Main Outcome Measures.— Symptom-based case definition; case prevalence rate for GW veterans
and nondeployed personnel; clinical and laboratory findings among veterans
who met the case definition.
Results.— We defined a case as having 1 or more chronic symptoms from at least
2 of 3 categories (fatigue, mood-cognition, and musculoskeletal). The prevalence
of mild-to-moderate and severe cases was 39% and 6%, respectively, among 1155
GW veterans compared with 14% and 0.7% among 2520 nondeployed personnel. Illness
was not associated with time or place of deployment or with duties during
the war. Fifty-nine clinically evaluated GW veterans (37%) were noncases,
86 (54%) mild-to-moderate cases, and 13 (8%) severe cases. Although no physical
examination, laboratory, or serologic findings identified cases, veterans
who met the case definition had significantly diminished functioning and well-being.
Conclusions.— Among currently active members of 4 Air Force populations, a chronic
multisymptom condition was significantly associated with deployment to the
GW. The condition was not associated with specific GW exposures and also affected
SOON AFTER cessation of Gulf War (GW) hostilities, anecdotal reports
of illness and speculation over environmental, biological, and chemical hazards
led to concerns about a "Gulf War syndrome."1-3
Subsequent studies have shown that GW veterans report numerous chronic nonspecific
symptoms, such as fatigue, neurocognitive problems, and musculoskeletal pain,
significantly more often than their nondeployed peers.4,5
No widespread disorder has been linked to that conflict, and the etiologic
basis and clinical significance of veterans' symptomatology remain unclear.6
The evidence for unique health problems among GW veterans is mixed.
Although acute gastroenteritis was frequently reported among troops,7,8 the overall occurrence of infectious
diseases was much lower than expected.9 No
specific illness is evident among the 18598 GW veterans, according to the
Department of Defense Comprehensive Clinical Evaluation Program.6
Other investigators have found no unusual increases in birth defects,10 unexplained illness,11
excess hospitalizations,12 or excess mortality13 among GW veterans. However, one study reported excess
mortality due to unintentional injuries rather than from disease.14 Evidence of an Iraqi chemical and biological weapons
program has been documented,15 and although
use of such weapons during the GW has not been confirmed,16
troops may have been exposed to chemical or biological warfare agents during
the destruction of storage bunkers. The long-term effects of such chemical
exposures are uncertain, although some investigators have suggested that wartime
exposures may have contributed to chronic neurotoxic syndromes.17,18
In December 1994, the US Secretary of Defense and the Secretary of Veterans
Affairs, and the Commonwealth of Pennsylvania asked the Centers for Disease
Control and Prevention (CDC), Atlanta, Ga, to investigate a "mystery illness"
reported among GW veterans from an Air National Guard (ANG) unit in Lebanon,
Pa.19 In the first phase of the investigation,
we interviewed and examined ill GW veterans and found that their major health
problems consisted of persistent fatigue and other chronic symptoms that began
during deployment or soon after returning from the GW, but we did not find
associated physical signs or clinical laboratory abnormalities.4
Patients' illnesses resembled those reported previously, and the classification
and etiology remain unclear.2
This article reports the second phase of our survey of the index ANG
unit and 3 comparison air force populations. Our primary objectives were to
determine whether the prevalence of symptoms was increased among currently
active members of the index unit and among GW veterans from other air force
populations in comparison with nondeployed personnel and to develop a case
definition that would allow further studies of etiology. We also report the
third phase, which was intended to characterize clinical and laboratory findings
among GW veterans who met our case definition.
Cross-sectional Questionnaire Survey
Study Population and Survey Procedures. We used study protocols that were approved by the Human Subjects Committee
of the CDC and followed the human experimentation guidelines issued by the
US Department of Health and Human Services in this study. All study participants
provided informed consent.
We surveyed the index and 3 comparison air force units between January
and March 1995. Unit A, another ANG unit in Pennsylvania, had similar demographic
features as the index unit but had a different primary mission. Unit B (US
Air Force Reserve) and unit C (active-duty air force) had similar primary
missions as the index unit's mission but were located in Florida. Any unit
member who was on base was eligible for the study. The index unit and units
A and B were each surveyed during 3 consecutive unit training weekends (held
monthly) to maximize participation rates. Unit C was surveyed during a 10-day
On each base, one of us (K.F., L.R., R.W., or D.N.) met with small groups,
explained the study, and distributed questionnaires to subject volunteers
whose identities remained anonymous. The questionnaire took 15 to 45 minutes
to complete and queried health status, demographic and military characteristics,
and potential risk factors for illness. We asked about the 35 symptoms that
had been identified during the earlier exploratory study4
including their severity (mild, moderate, or severe) and duration (<6 months
or ≥6 months).
We used 2 conceptually different approaches to develop exploratory case
definitions from the symptom data. The first exploratory case definition was
derived from clinical and epidemiological reasoning; the second used statistical
Clinical Approach. We reasoned that if an illness existed among GW veterans, case-defining
symptoms would be chronic, prominent among GW veterans, and more common among
GW veterans than nondeployed personnel. Thus, we decided that symptoms for
the clinical case definition had to be reported for 6 months or longer, by
at least 25% of GW veterans and by GW veterans at least 2.5 times more often
than by nondeployed personnel. The clinical case definition was identified
as fatigue, difficulty remembering or concentrating, moodiness, difficulty
sleeping, joint pain, and joint stiffness.
Statistical Approach. In the statistical approach, we randomly divided the 3255 participants
into 2 subsamples of 1631 and 1624 subjects (446 were excluded from analysis
because of incomplete symptom data). We conducted an exploratory principal
components analysis on the first subsample and a confirmatory factor analysis
on the second.
Exploratory Principal Components Analysis.— All 35 symptoms were submitted to a principal components analysis, an
exploratory method that can help determine the probable number and nature
of common components (or factors).20 We used
a promax rotation because it is a well-formulated procedure to extract correlated
factors21 and provides interfactor correlations.
The Kaiser-Guttman rule (an eigenvalue >1) was used to determine the number
of factors to retain for rotation. In addition, we examined the scree plot
and verified that the slope connecting the eigenvalues approached zero after
retaining the number of selected factors. After rotation, the factor pattern
was examined to determine whether an oblique rotation yielded high loadings
in only one factor. Symptoms with factor loadings of greater than 0.40 were
kept for the confirmatory factor analysis.22
The exploratory analysis yielded 10 components with an eigenvalue greater
than 1. The first (feeling depressed, feeling anxious, feeling moody, difficulty
remembering or concentrating, trouble finding words, difficulty sleeping,
and fatigue) accounted for 16.8% of the total common variance. The second
(joint stiffness, joint pain, and muscle pain) accounted for 11.9% of the
variance. The third (wheezing, shortness of breath, coughing, and chest pain)
accounted for 10.4%. The remaining components each contributed less than 10%
of the total variance.
Confirmatory Factor Analysis.— Only symptoms from the first 3 principal analysis components were used
in the confirmatory factor analysis.23 These
symptoms were submitted to an unweighted least squares factor analysis with
a Procrustes oblique rotation.24 Confirmatory
factor analysis identified 2 factors (mood-cognition-fatigue and musculoskeletal)
Preliminary Working Case Definitions. The 10 symptoms from factor analysis included all 6 symptoms identified
by the clinical approach. We used these 10 symptoms to derive 2 possible working
case definitions. First, we derived a case definition on the basis of factor
scores. The confirmatory phase factor analysis model was fit to the participants'
symptom data and a total factor score was calculated for each participant
by adding the scores of the factors. A case was defined as having a combined
factor score in the top 25th percentile.
Second, we developed an alternative working case definition based on
symptom categories. This grouped the symptoms into 3 categories: fatigue,
mood-cognition, and musculoskeletal. We separated chronic fatigue even though
it did not load as a separate factor because of the central role of chronic
fatigue in virtually all previous studies of GW veterans. We defined a case
as having 1 or more symptoms from 2 or more symptom categories.
Comparison of the Preliminary Working Case Definitions. We compared the case definitions by determining the prevalence of each
type of case among GW veterans and nondeployed personnel and calculated their
overall agreement by a κ statistic.25
Forty-seven percent of GW veterans and 15% of nondeployed were classified
as factor score cases. Of these, 25% had symptoms that included fatigue, mood-cognition,
and musculoskeletal pain; 39.2% fatigue and mood-cognition; 15.6% fatigue
and musculoskeletal pain; 6.6% mood-cognition and musculoskeletal pain. An
additional 9.9% reported only mood-cognition symptoms, and 8.1% reported only
musculoskeletal pain. Forty-five percent of GW veterans and 15% of nondeployed
were classified as symptom-category cases. Of these, 41.8% reported symptoms
of fatigue, mood-cognition, and musculoskeletal pain; 26.7% reported fatigue
and mood-cognition; 8.4% reported fatigue and musculoskeletal pain; and 23.1%
reported mood-cognition and musculoskeletal pain. The κ statistic was
0.79, which indicated substantial overall agreement between the working case
Chronic Multisymptom Illness. Since both case definitions were comparable, we chose the symptom-category
approach because it was easier to apply in a clinical setting. We defined
a case as having 1 or more chronic symptoms (present for ≥6 months) from
at least 2 of the following categories: fatigue; mood and cognition (symptoms
of feeling depressed, difficulty remembering or concentrating, feeling moody,
feeling anxious, trouble finding words, or difficulty sleeping); and musculoskeletal
(symptoms of joint pain, joint stiffness, or muscle pain). We subclassified
a case as severe if each case-defining symptom was rated as severe; otherwise,
we considered the case to be mild-to-moderate.
Cross-sectional Clinical Evaluation
Study Population and Enrollment Procedures. We conducted a third-phase, cross-sectional clinical study during April
and May 1995. Subjects were GW veterans currently in the index unit. During
the preceding cross-sectional survey, we informed the 667 participants from
the index unit of our intent to conduct a follow-up clinical study and asked
for volunteers. We also recruited participants by posting notices throughout
the base. Our solicitations stressed the need for both symptomatic and nonsymptomatic
Clinical Evaluation. Each volunteer was mailed an informed consent form, a clinical questionnaire,
and the Medical Outcomes Study Short Form 36 (MOS SF-36)26,27
to complete at home. The clinical questionnaire asked about the presence,
duration, and intensity of 35 symptoms and included questions about fatigue,
activity levels, and diarrhea.
Each subject was individually evaluated by a CDC field team at the Veterans
Affairs (VA) Medical Center in Lebanon, Pa. Subjects completed the Mississippi
Post Traumatic Stress Disorder Scale (modified for use among GW veterans).28 The 1994 revised chronic fatigue syndrome (CFS) working
case definition was used to classify cases of CFS.29
A trained interviewer privately administered selected modules on somatization
disorder, major depression, and panic disorder from the Diagnostic Interview
Schedule version of Diagnostic and Statistical Manual of
Mental Disorders, Fourth Edition, (DSM-IV)30 The Diagnostic Interview Schedule was modified, so
the onset of disorders could be referenced to periods before, during, and
after deployment to the GW. Each subject next received a screening physical
examination performed by a physician assistant. This examination systematically
covered vital signs, height, weight, skin, ears (including gross hearing),
eyes (including extraocular eye movements and retina), nasal passages, throat,
neck (including thyroid), lungs, heart, abdomen, lymph nodes (cervical, axillary,
and inguinal areas), rectal vault (including testing of stool for occult blood),
reflexes, and tests of coordination (finger to nose), strength, and sensation
(light touch, vibration, and pin prick). The personnel who performed these
evaluations were not aware of the subjects' case status.
Case Definition. We classified participants in the clinical study as cases or noncases
based on their responses to the mailed clinical epidemiological questionnaire
and the case definition derived in the cross-sectional survey.
Blood and Urine Specimen Tests. Blood specimen evaluation included a complete blood cell count, erythrocyte
sedimentation rate, serum chemistry values (alanine aminotransferase, alkaline
phosphatase, total protein, albumin, globulin, calcium, phosphorus, glucose,
electrolytes, blood urea nitrogen, and creatinine), and thyroid-stimulating
hormone level. Urine specimens underwent routine urinalysis.
Serologic Testing. To test for arboviruses and rickettsia, serum samples were tested for
antibodies to yellow fever, dengue, Sindbis, West Nile, phlebotomus fever
viruses (Naples and Sicilian), Toscana, Karimbad, and Isfahan viruses at the
Division of Vector-Borne Infectious Diseases, CDC.31
Serologic testing for antibodies to Rickettsia typhi, Rickettsia rickettsii, Coxiella burnetii, and Ehrlichia chaffeensis was performed
at the Division of Viral and Rickettsial Diseases, CDC.32
To test for parasites, serum samples were tested at the Division of
Parasitic Diseases, CDC for reactivity to Leishmania donovani and Leishmania tropica promastigote antigens,33 and also were tested for antibodies to Toxoplasma gondii, Schistosoma mansoni,Schistosoma haematobium,34
and Strongyloides stercoralis.
To test for other agents, serum samples were tested for antibodies to Helicobacter pylori at the Division of Bacterial and Mycotic
Diseases, CDC. To screen for exposure (either by vaccination or in combat)
to 2 widely discussed putative biologic warfare agents, we tested serum samples
for antibodies to toxin produced by Clostridium botulinum and Bacillus anthracis. Serum samples were
screened at the Division of Bacterial and Mycotic Diseases, CDC, for antibodies
to type A botulinum toxin. Serum samples were assayed at the US Army Medical
Research Institute of Infectious Diseases, Washington, DC, for antibodies
against anthrax protective antigen and lethal factor.
Stool Specimens. Fresh fecal specimens were examined for red and white blood cells and
were cultured for Salmonella, Shigella , Yersinia, Campylobacter, and Escherichia coli 0157:H7. Fecal specimens were examined
at the Division of Parasitic Diseases, CDC, for ova and parasites,35,36 including but not limited to Cryptosporidium parvum, Cyclospora cayetanensis , Isospora belli, and microsporidia.37
Fecal specimens also were screened for enteroviruses at Southern General Hospital,
We estimated the association between symptoms and deployment to the
GW by prevalence ratios. We assessed univariate associations of categorical
variables by odds ratios (OR) and 95% confidence intervals (95% CI). We used
the χ2, Fisher exact, and Mantel-Haenszel tests to calculate P values. We used the Student t
test and 1-way, between-subjects analysis of variance (ANOVA) to compare means
of continuous variables and compared pairs of means by the Scheffé
test. We used the PROC FACTOR procedure to perform principal components and
factor analyses. In the factor analysis, the proportion of total common variance
explained by each factor was determined by the direct contribution method.23 The PROC CATMOD procedure was used to fit a polytomous
logistic regression model and simultaneously assess associations between both
mild-to-moderate and severe cases and possible risk factors. All calculations
were performed using Statistical Analysis Systems (SAS) software.39 All statistical tests were 2-tailed.
Cross-sectional Questionnaire Survey
Survey Participation Rates. Questionnaires were completed by 4036 persons. We excluded 313 subjects;
109 were not members of the units, and 204 were younger than 17 years during
the GW. Among 3723 remaining participants, 1163 (31.2%) were GW veterans and
2560 (68.8%) had not been deployed. After exclusions, unit participation rates
were 62% (667/1083) for the index unit, 35% (538/1520) for unit A, 73% (838/1141)
for unit B, and 70% (1680/2407) for unit C.
Demographic, Military, and Health Characteristics. The median age of participants was 34 years (mean, 35 years; range,
21-60 years); 86% were men; 86% were white; and 89% were employed full-time.
Other demographic, military, and general health characteristics were similar
in the 4 units (Table 2), except
that active-duty subjects (unit C) were younger, less likely to be white,
and less likely to describe their current health as "excellent." Demographic
characteristics of survey participants were similar to overall unit demographics.
Prevalence of Symptoms. Overall, 3701 participants (99%) reported at least 1 symptom as a current
health problem. Gulf War veterans reported all symptoms, except hay fever
and other allergies, significantly more often than nondeployed (Table 3). All symptoms, except headache and difficulty sleeping,
were reported more often by GW veterans from the index unit than those from
comparison units. Chronic diarrhea was reported substantially more often by
GW veterans from the index unit. The prevalence of symptoms among nondeployed
personnel was similar among all units (data not shown).
Occurrence of Chronic Multisymptom Illness Cases. In all, 3675 subjects provided complete data on the 10 case-defining
symptoms. Among the 1155 GW veterans, 6% were classified as severe cases and
39% as mild-to-moderate cases compared with 0.7% and 14%, respectively, among
the 2520 nondeployed personnel. Deployment to the GW was the most important
risk factor for severe and mild-to-moderate illness (Table 4). Multivariate analyses (Table 5) showed that severe cases were associated with GW service,
enlisted rank, female sex, and smoking. Mild-to-moderate cases were associated
with the same variables, and increasing age and current membership in the
index unit. Illness was not associated with the number of deployments, month
or season of deployment, duration of deployment, military occupational specialty,
direct participation in combat, or self-reported locality in the Gulf region
(most were in the Rhiyadh area of Saudi Arabia).
Clinical Evaluation Study
Participation Rate and Classification of Cases. Among the index unit's 1083 members, 490 (45%) were GW veterans and
173 of those (35%) volunteered to participate in the clinical evaluation study.
Fifteen (8.7%) provided incomplete symptom data and were excluded from further
analysis. Of the remaining 158 veterans, 13 (8%) were severe cases, 86 (54%)
were mild-to-moderate cases, and 59 (37%) were noncases.
General Characteristics and Current Symptoms. Most participants were men, but 4 of the severe cases (31%) occurred
among women. The prevalence of all chronic symptoms (including those not used
to define cases) was highest among severe cases, followed by mild-to-moderate
cases, and noncases (data not shown). Diarrhea (≥3 loose stools per day)
was reported by 3% (n = 2) of noncases, by 26% (n=22) of mild-to-moderate
cases (OR, 10; 95% CI, 2.2-44.2), and by 77% (n=10) of severe cases (OR, 95;
95% CI, 14.1-642.3). All but 3 subjects who reported diarrhea indicated it
had been present for at least 6 months, and all but 2 dated onset from the
period during or immediately following their return from the Gulf region.
Only 2 subjects with diarrhea also reported weight loss.
Medical History Prior to Deployment. Severe cases reported histories of depression before deployment significantly
more often (15%) than noncases (0%) (Fisher exact test P<.05) and also were associated with self-reported sinusitis prior
to first deployment more frequently (38%) than noncases (10%) (OR, 5.8; 95%
CI, 1.4-24.5). Otherwise, there were no statistically significant differences
referable to the year before deployment to the GW between severe or mild-to-moderate
cases and noncases in terms of 11 allergy symptoms; 7 self-reported surgical,
dental, or invasive medical procedures (including blood or blood product transfusions
and mercury amalgam repair of dental cavities); or self-reported lifetime
prevalence, before deployment to the Persian Gulf, of 35 medical and psychiatric
conditions (including heart disease, hypertension, diabetes, alcohol or substance
abuse, anorexia or bulimia, migraine or severe headaches, anxiety, diarrhea,
irritable bowel syndrome, and impotence).
Physical Examination. Physical examinations were most notable for the general paucity of abnormal
findings. Severe cases were associated with slightly higher body mass indices,
a measure of weight in kilograms divided by the square of height in meters
(BMI, 29.6), than noncases (BMI, 26.5) (P<.01).
Rash was present more often among mild-to-moderate (14%) and severe (15%)
chronic multisymptom cases than among noncases (3%), but the differences were
not statistically significant. Several minor abnormalities were found on lymph
node, liver, spleen, or neurologic examination, but none was associated with
cases. A summary of physical examination data is available from the authors.
Functional Status and Well-being. As measured by the MOS SF-36, severe and mild-to-moderate cases were
associated with a significant decrease in functioning and well-being. Severe
cases were associated with a decrease on all of the 6 subscales (Table 6).
Blood and Urine Testing. The mean values of a few routine blood tests differed among cases and
noncases, but the differences were marginal and clinically unimportant. None
of the urinalysis results differed between cases and noncases. A summary of
blood and urine data is available from the authors.
Stool Specimen Testing. No fecal specimen tested positive for occult blood, white blood cells, Salmonella, Shigella, Campylobacter, Yersinia, enteropathogenic Escherichia coli, microsporidia, Cryptosporidium
parvum, Cyclospora cayetanensis, Isospora belli, or Entamoeba histolytica. Blastocystis hominis was found in stool specimens from
7 noncases (12%) and from 6 mild-to-moderate and chronic (7%) multisymptom
cases. Giardia lamblia was found in stool specimens
from 1 noncase (2%) and 1 mild-to-moderate case (1%). Enteroviruses were found
in stool specimens from 6 noncases (10%) and 8 mild-to-moderate cases (9%).
A summary of stool specimen data is available from the authors.
Serologic Testing. There was no association between seropositivity to various infectious
agents and chronic multisymptom cases. Troops are routinely vaccinated against
yellow fever virus, and 83% of subjects had antibodies against this agent,
but there were no differences between cases and noncases. Ten subjects reacted
to botulina toxin and 14 to anthrax protective antigen, but there were no
differences between cases and noncases. Eight participants (4.9%) seroreacted
to leishmanial antigens. However, no severe case was seroreactive; reactivity
rates were higher among noncases than mild-to-moderate cases; the difference
was not significant; and titers were low (range, 1:16-1:64). Nine subjects
were seropositive to phlebotomus fever viruses, but rates were similarly low
among noncases (2%), mild-to-moderate cases (9%), and severe cases (8%). Thirty
persons (19%) were seroreactive to Toxoplasma gondii,
and rates were similar between cases and noncases. Sixteen subjects had antibodies
to dengue virus and rates were similar in cases and noncases. Similar, low,
and equally distributed (among noncase and cases) seroreactivity existed to Coxiella burnetii (7 positive), R rickettsii (6 positive), E chaffeensis (5 positive),
and R typhi (1 positive). No subject had antibodies
to Sindbis, West Nile, Toscana, Karimbad, or Isfahan viruses, Schistosoma species, or S stercoralis. A summary
of serologic data is available from the authors.
Posttraumatic Stress Disorder. One subject screened positive for posttraumatic stress disorder also
was classified as a severe chronic multisymptom case.
Depression. Twenty-two subjects met the Diagnostic Interview Schedule of the DSM IV criteria for major depression occurring after their
last deployment to the GW, and 19 also reported depression within the last
year (1994-1995). Overall, major depression was more common among GW veterans
after deployment than before deployment. The prevalence of current depression
was significantly higher among severe chronic multisymptom cases (54%) and
mild-to-moderate cases (13%) than among noncases (2%).
Somatization and Panic Disorders. Four subjects (all classified as severe chronic multisymptom cases)
met DSM-III-R criteria for somatization disorder.
Three met criteria for panic disorder (2 were mild-to-moderate cases and 1
was a severe case). Panic disorder was uncommon in the periods before and
during deployment to the GW. Four of the subjects met criteria for panic disorder
during their deployment to the war, including 1 classified a as mild-to-moderate
case and 1 classified as a severe case.
Chronic Fatigue Syndrome. Eight subjects met all criteria for CFS.29
Of these, 7 (54%) also were classified as severe cases, and 1 (1%) was a mild-to-moderate
Our study of currently active members from 4 air force populations showed
a substantially higher prevalence of symptoms among GW veterans than nondeployed.
This result was similar to findings of an earlier and unrelated epidemiological
study of Pennsylvanian and Hawaiian ANG personnel,40
a population-based study from all branches of service,5
and is consistent with clinical data from GW veteran registries.6
Thus, air force GW veterans likely experience an illness similar to that reported
from other branches of the service.
The number and diversity of symptoms reported by our study subjects
and by GW veterans in other studies make analytic epidemiological study difficult.
Therefore, a major objective was to develop a case definition that captured
the chronic multisymptomatic nature of illness in GW veterans. We used 2 independent
methods to identify case-defining symptoms. An exploratory clinical approach
showed that veterans had a condition defined by fatigue, neurocognitive symptoms,
and joint pain and stiffness. Factor analysis identified 2 major factors that
included and extended symptoms identified by the clinical approach.
Besides extending the number of correlated symptoms, factor analysis
used data collected from all participants and did not bias our study toward
finding a higher case prevalence among veterans. Factor analysis also uses
rigorous empirical statistical methods to identify factors and associated
factor scores. Thus, we used information from factor analysis to derive the
final case definition. We first derived a factor score case definition that
included as cases those subjects whose total factor scores were in the upper
25th percentile of the study population. We derived a second case definition
by grouping the symptoms identified by factor analysis into 3 of the following
categories: fatigue, mood-cognition, and musculoskeletal pain. We separated
chronic fatigue, even though it did not load as a separate factor because
of the central role of chronic fatigue in all studies of GW veterans.
The κ statistic measuring agreement between the 2 preliminary
case definitions was 0.79, which represents substantial agreement.41 Disagreement between the 2 definitions occurred because
factor score–defined cases could have symptoms from only 1 factor and
symptom score–cases could have symptoms from 2 categories but not exceed
the 25th percentile of factor scores. We chose the symptom category approach
for the final working case definition because it generally identified the
same cases as the factor score approach, better captured the multisymptom
nature of illness, was clinically more understandable, and allowed us to subclassify
cases according to severity. To further evaluate the validity of defining
case subjects based on symptom categories rather than factor scores, we examined
univariate and multivariate associations (variables from Table 4 and Table 5)
with cases defined by both methods and found no meaningful differences.
Our case definition represents a unique approach toward organizing symptom
data. We intended it to provide a summary measure of illness to test for associations
with clinical abnormalities and risk factors and not as a definitive label
for a single, distinct illness. The resulting case definition needs to be
replicated by other studies of GW veterans. However, in a civilian population,
factor analysis methods identified fatigue, mood, and cognition as the most
important factors among symptoms similar to the ones in the current study.42
The key observation of this study was that air force GW veterans were
significantly more likely to meet criteria for severe and mild-to-moderate
illness than were nondeployed personnel. There was no association between
the chronic multisymptom illness we defined and risk factors specific to combat
in the GW (month or season of deployment, duration of deployment, duties in
the GW, direct participation in combat, or locality of GW service). Our finding
that 15% of nondeployed also met illness criteria was equally important and
suggests that the multisymptom illness we observed in this population is not
unique to GW service.
To characterize the clinical features of GW veterans with this chronic
multisymptom illness, we evaluated veterans from one unit. Neither mild-to-moderate
nor severe cases were associated with clinically significant physical examination
or routine laboratory test abnormalities. However, GW veterans classified
as having mild-to-moderate and severe illness had a significant decrease in
functioning and well-being compared with noncases, as measured by the MOS
We tested participants for exposure to several infectious agents that
are important health problems in the Gulf region, that may have been used
in vaccines, and that might be associated with a chronic illness. Leishmaniasis
is endemic to the region, and a syndrome termed viscerotropic
leishmaniasis has been described in GW veterans.43,44
Approximately 5% of participants seroreacted to leishmanial antigens, but
the titers were low and seroreactivity rates were not significantly different
between cases and noncases. No study subject had clinical manifestations of
visceral or cutaneous leishmaniasis. However, the manifestations of viscerotropic
leishmaniasis can be nonspecific.43,44
The serologic assay we used is a sensitive diagnostic tool for classic visceral
leishmaniasis, but there is, as yet, no reliable screening test for viscerotropic
leishmaniasis.43,44 As a surrogate
for exposure to phlebotomine sand flies, which transmit Leishmania parasites and sand fly fever viruses, we tested participants
for antibodies against Naples and Sicilian phlebotomus fever viruses. Seroprevalence
was similar to that noted with the Leishmania serologic
assay. Clearly, neither seroreactivity to leishmanial antigens nor exposure
to the phlebotomine vectors was associated with the illness we defined.
Similarly, we found no association between illness and antibody against
the other viruses, rickettsiae, parasites, or bacteria for which we assayed.
Chronic diarrhea was reported among 77% of severe cases, but we found no associations
with any infectious agents. Although we did not evaluate subjects for noninfectious
causes of diarrhea, such as malabsorption syndromes, the descriptive stool
sample characteristics and lack of weight loss suggest irritable bowel syndrome
as the predominant reason for the reported chronic diarrhea.
Our study has several limitations. Most important, the study involved
currently active air force personnel (primarily reservists) and cannot be
generalized to other branches of service or to GW veterans who have left the
service. The primary objectives were to devise a case definition, evaluate
the occurrence of illness, and clinically describe the illness. We did not
intend to test specific hypotheses, many of which have arisen subsequent to
the study; rather, we intended to establish basic parameters to aid in future
The symptom data were self-reported and, thus, may be subject to reporting
or recall bias. Other studies have attempted to categorize GW veterans as
having illness in various predefined categories (eg, CFS, multiple chemical
sensitivities, depression, and neurologic abnormalities). We chose a completely
different strategy first to document the presence, magnitude, and duration
of various symptoms reported by veterans and nondeployed personnel and then
to determine if a meaningful construct could be identified. Although a reporting
bias may have contributed to the moderately higher prevalence of symptoms
reported by GW veterans in the index unit, the main findings of this study
remain valid for the comparison units. In addition, although the participation
rate of one of the comparison units (unit A) was low, the findings within
that unit were consistent with those from the other units.
The same limitations apply to the clinical evaluation phase of the study.
We evaluated GW veterans from one unit in order to establish clinical parameters
of the illness we had defined and not to conduct a case-control study. Participants
underwent a systematic standardized physical examination by physician assistants,
who focused on documenting obvious abnormalities. Thus, this study did not
attempt to assess subtle physiologic differences. Furthermore, the number
of severe, chronic, multisymptom cases was relatively small, which potentially
limited our ability to identify significant findings associated with those
In addition, all clinically evaluated subjects were from a single ANG
unit, and only 35% of GW veterans from that unit participated. This participation
rate may not adequately represent GW veterans. The extent to which ill GW
veterans were underrepresented or overrepresented could influence the findings.
A first interpretation may suggest that ill veterans were more likely to have
volunteered, because 63% of the clinically evaluated participants were classified
as cases in contrast with the 45% prevalence of illness in GW veterans from
the cross-sectional survey. However, it was our impression that military personnel
with severe illness were less likely to participate for fear of identification
and service-related medical consequences. Moreover, it may be that the most
seriously affected GW veterans are no longer in the military.
These limitations notwithstanding, our results indicate that a substantial
proportion of currently active air force GW veterans have a chronic multisymptom
illness, which is accompanied by significant decreases in functioning and
well-being. The illness was not associated with physical examination or clinical
laboratory abnormalities or with infection by assorted agents that are found
in the region or to which troops in combat may have been exposed. Poorly characterized
illness, including fatigue, neurocognitive, and musculoskeletal complaints,
has affected veterans of many other wars.45
The empirically based approach described here provides a method for assessing
the prevalence of multiple chronic symptoms and exploring the clinical basis
of this condition. Other studies should use similarly rigorous analytic methods
to classify cases and standardized measures of function and well-being to
collect data and should enroll personnel from other branches of the military.
The elevated case prevalence among GW veterans remains unexplained by
our study. It is possible that the symptom complex is associated with GW-specific
exposures (eg, an as-yet-unidentified chemical or biological agent).16 Although in our study population the risk of illness
was not associated with the dates, season, duration, number of deployments,
or military occupational activities, we believe a more plausible explanation
is that key risk factors present among both deployed and nondeployed populations
were present at higher intensity or greater frequency among GW veterans. It
is clear that the distribution of cases among GW veterans and nondeployed
personnel in this study cannot easily be explained by risk factors unique
to Southwest Asia.
National Institutes of Health Technology Assessment Workshop Panel. The Persian Gulf experience and health. JAMA.1994;272:391-396.Google Scholar
Persian Gulf Veterans Coordinating Board. Unexplained illnesses among Desert Storm veterans: a search for causes,
treatment and cooperation. Arch Intern Med.1995;155:262-268.Google Scholar
Institute of Medicine. Health Consequences of Service in the Persian Gulf War: Recommendations
for Research and Information Systems . Washington, DC: National Academy Press; 1996.
Centers for Disease Control and Prevention. Unexplained illness among Persian Gulf War veterans in an Air National
Guard unit: preliminary report—August 1990-March 1995. MMWR Morb Mortal Wkly Rep.1995;44:443-447.Google Scholar
Iowa Persian Gulf Study Group. Self-reported illness and health status among Gulf War veterans: a
population-based study. JAMA.1997;277:238-245.Google Scholar
Department of Defense. Comprehensive Clinical Evaluation Program for Gulf War Veterans:
Report on 18,598 Participants . Washington, DC: Dept of Defense; 1996.
Hyams KC, Bourgeois AL, Merrell BR.
et al. Diarrheal disease during Operation Desert Shield. N Engl J Med.1991;325:1423-1428.Google Scholar
Hyams KC, Malone JD, Kapikian AZ.
et al. Norwalk virus infection among Desert Storm troops. J Infect Dis.1993;167:986-987.Google Scholar
Hyams KC, Hanson K, Wignall FS, Escamilla J, Oldfield EC. The impact of infectious diseases on the health of U.S. troops deployed
to the Persian Gulf during operations Desert Shield and Desert Storm. Clin Infect Dis.1995;20:1497-1504.Google Scholar
Penman AD, Tarver RS, Currier MM. No evidence of increase in birth defects and health problems among
children born to Persian Gulf War veterans in Mississippi. Mil Med.1996;161:1-6.Google Scholar
DeFraites RF, Wanant ER, Norwood AE, Williams S, Cowan D, Callahan T. Investigation of a Suspected Outbreak of an Unknown Disease Among
Veterans of Operations Desert Shield/Storm, 123rd Army Reserve Command, Fort
Benjamin Harrison, Indiana, April 1992 . Washington DC: Walter Reed Army Institute of Research; 1992.
Gray GC, Coate BD, Anderson CM.
et al. The postwar hospitalization experience of US veterans of the Persian
Gulf War. N Engl J Med.1996;335:1505-1513.Google Scholar
Writer JV, DeFraites RF, Brundage JF. Comparative mortality among US military personnel in the Persian Gulf
region and worldwide during operations Desert Shield and Desert Storm. JAMA.1996;275:118-121.Google Scholar
Kang HK, Bullman TA. Mortality among US veterans of the Persian Gulf War. N Engl J Med.1996;335:1498-1504.Google Scholar
United Nations Special Commission on Iraq. Report of the Secretary-General on the Status of the Implementation
of the Special Commission's Plan for the Ongoing Monitoring and Verification
of Iraq's Compliance with Relevant Parts of Section C of Security Council
Resolution 687 (1991) . New York, NY: United Nations; 1995.
Presidential Advisory Committee on Gulf War Veterans' Illnesses. Special Report . Washington, DC: US Government Printing Office; October 1997.
Haley RW, Kurt TL, Hom J. Is there a Gulf War syndrome? searching for syndromes by factor analysis
of symptoms. JAMA.1997;277:215-222.Google Scholar
Haley RW, Kurt TL. Self-reported exposure to neurotoxic chemical combinations in the Gulf
War: a cross-sectional epidemiologic study. JAMA.1997;277:231-237.Google Scholar
Murray-Leisure KA, Daniels MO. Visceral and skin aspects of "Persian Gulf mystery disease" associated
with sand exposure in a Pennsylvania military unit. In: Program and abstracts of the 34th Interscience Conference on
Antimicrobial Agents and Chemotherapy; October 1994; Orlando, Fla. Abstract
Basilevski A. Statistical Factor Analysis and Related Methods . New York, NY: John Wiley & Sons; 1994.
Floyd FJ, Widaman KF. Factor analysis in the development and refinement of clinical assessment
instruments. Psych Assess.1995;7:286-299.Google Scholar
Tabachnick BG, Fidell LS. Using Multivariate Statistics . New York: Harper & Row; 1983.
Harman HH. Modern Factor Analysis . Chicago, Ill: University of Chicago Press; 1976:268-270.
Gorsuch RL. Factor Analysis . Hillsdale, NJ: Erlbaum; 1983.
Agresti A. Categorical Data Analysis . New York: John Wiley & Sons; 1990.
Ware JE, Sherbourne CD. The MOS 36-item short-form health survey (SF-36): conceptual framework
and item selection. Med Care.1992;30:473-483.Google Scholar
McHorney CA, Ware JE, Rogers W, Raczek AE, Lu JFR. The validity and relative precision of MOS short- and long-form health
status scales and Dartmouth Coop charts: results from the Medical Outcomes
study. Med Care.1992;30(suppl 5):MS253-MS265.Google Scholar
Keane TM, Caddell JM, Taylor KL. Mississippi scale for combat related posttraumatic stress disorder:
three studies in reliability and validity. J Consult Clin Psychol.1988;56:85-90.Google Scholar
Fukuda K, Straus SE, Hickie I, Sharpe MC, Dobbins JG, Komaroff A. The chronic fatigue syndrome: a comprehensive approach to its definition
and study. Ann Intern Med.1994;121:953-959.Google Scholar
American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders . 4th ed. Washington, DC: American Psychiatric Press; 1994.
Tsai TF. Arboviruses. In: Murray PR, Baron EJ, Pfaller MA, Tenover FC, Yolken RH, eds. Manual of Clinical Microbiology . 6th ed. Washington, DC: American Society
of Microbiology Press; 1995:980-996.
Newhouse VF, Shepard CC, Redus MD, Tzianabos T, McDade JE. A comparison of the complement fixation, indirect immunofluorescent
antibodies, and microagglutination tests for the serologic diagnosis of rickettsial
diseases. Am J Trop Med Hyg.1979;28:387-395.Google Scholar
Camargo ME. Fluorescent antibody test for the serodiagnosis of American trypanosomiasis:
technical modification employing preserved culture forms of Trypanosoma cruzi
in a slide test. Rev Inst Med Trop Sao Paulo.1966;8:227-234.Google Scholar
Tsang VCW, Wilkens PP. Immunodiagnosis of schistosomiasis. Clin Lab Med.1991;11:1029-1040.Google Scholar
Melvin DM, Brooke MM. Laboratory Procedures for the Diagnosis of Intestinal Parasites . 3rd ed. Atlanta, Ga: US Dept of Health and Human Services, Centers
for Disease Control; 1982. Publication CDC 85-8282.
Ash LR, Orihel TC. Parasites: A Guide to Laboratory Procedures and Identification . Chicago, Ill: American Society of Clinical Pathologists Press; 1987:51.
Weber R, Bryan RT, Owen RL, Wilcox CM, Gorelkin L, Visvesvara GS. Improved light-microscopical detection of microsporidia spores in stool
and duodenal aspirates. N Engl J Med.1992;326:161-166.Google Scholar
Gow JW, Behan WMH, Clements GB, Woodall C, Riding M, Behan PO. Enteroviral RNA sequences detected by polymerase chain reaction in
muscle of patients with postviral fatigue syndrome. BMJ.1991;302:692-696.Google Scholar
SAS/Statistical User's Guide. Vols 1 and 2. Cary, NC: Statistical Analysis Systems Institute Inc;
Stretch RH, Bliese PD, Marlowe DH. Physical symptomatology of Gulf War-era service personnel from the
states of Pennsylvania and Hawaii. Mil Med.1995;160:131-136.Google Scholar
Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics.1977;33:159-174.Google Scholar
Nisenbaum R, Reyes M, Mawle AC, Reeves WC. Factor analysis of unexplained severe fatigue and interrelated symptoms:
overlap with criteria for chronic fatigue syndrome. Am J Epidemiol.1998;148:72-77.Google Scholar
Centers for Disease Control and Prevention. Viscerotropic leishmaniasis in persons returning from Operation Desert
Storm, 1990-1991. MMWR Morb Mortal Wkly Rep.1992;41:131-134.Google Scholar
Magill AJ, Grogl M, Gasser RA, Sun W, Oster CN. Visceral infection caused by Leishmania tropica
in veterans of Operation Desert Storm. N Engl J Med.1993;328:1383-1387.Google Scholar
Hyams RI, Wignall FS, Roswell R. War syndromes and their evaluation: from the U.S. Civil War to the
Persian Gulf War. Ann Intern Med.1996;125:398-405.Google Scholar