Predictors of Posttraumatic Stress Disorder and Symptoms in Adults: A Meta-Analysis

Abstract

A review of 2,647 studies of posttraumatic stress disorder (PTSD) yielded 476 potential candidates for a meta-analysis of predictors of PTSD or of its symptoms. From these, 68 studies met criteria for inclusion in a meta-analysis of 7 predictors: (a) prior trauma, (b) prior psychological adjustment, (c) family history of psychopathology, (d) perceived life threat during the trauma, (e) posttrauma social support, (f) peritraumatic emotional responses, and (g) peritraumatic dissociation. All yielded significant effect sizes, with family history, prior trauma, and prior adjustment the smallest (weighted r = .17) and peritraumatic dissociation the largest (weighted r = .35). The results suggest that peritraumatic psychological processes, not prior characteristics, are the strongest predictors of PTSD.

Full text

Predictors of Posttraumatic Stress Disorder and Symptoms in Adults:
A Meta-Analysis
Emily J. Ozer
University of California, San Francisco Suzanne R. Best and Tami L. Lipsey
Veterans Affairs Medical Center, San Francisco, California
Daniel S. Weiss
University of California, San Francisco
A review of 2,647 studies of posttraumatic stress disorder (PTSD) yielded 476 potential candidates for
a meta-analysis of predictors of PTSD or of its symptoms. From these, 68 studies met criteria for
inclusion in a meta-analysis of 7 predictors: (a) prior trauma, (b) prior psychological adjustment, (c)
family history of psychopathology, (d) perceived life threat during the trauma, (e) posttrauma social
support, (f) peritraumatic emotional responses, and (g) peritraumatic dissociation. All yielded significant
effect sizes, with family history, prior trauma, and prior adjustment the smallest (weighted r⫽.17) and
peritraumatic dissociation the largest (weighted r⫽.35). The results suggest that peritraumatic psycho-
logical processes, not prior characteristics, are the strongest predictors of PTSD.
Throughout the 20th century, interest in the psychological im-
pact of trauma has peaked during and after wartime, with the first
major study of combat-related psychological sequelae (then called
physioneurosis) published in 1941 by A. Kardiner (see Kolb, 1993,
for a more detailed accounting). Dealing with survivors of World
War II death and prisoner of war camps brought some insight (but
less than might have been expected) into the effects of extreme
trauma on psychological functioning. In 1968, the Diagnostic and
Statistical Manual of Mental Disorders (2nd ed.; DSM–II; Amer-
ican Psychiatric Association, 1968, p. 49) was published, and a
year later, the Manual of the International Statistical Classification
of Diseases, Injuries, and Causes of Death (8th ed.; ICD–8; World
Health Organization, 1969, p. 158) appeared, with reference to
“combat fatigue” in the latter and “fear associated with military
combat and manifested by trembling, running, and hiding” in the
former. Both of these clinical phenomena, however, were placed in
the category of “transient situational disturbances,” even though a
volume entitled The Traumatic Neurosis (Keiser, 1968) was pub-
lished the same year that suggested a more profound and chronic
effect of exposure to traumatic stress. A clear recognition of the
coherence of chronic, long-term posttraumatic stress reactions had
not materialized, even though these earlier nomenclatures were
explicit about the phenomenology of the short-term acute effects
of exposure to traumatic stressors.
In the early and mid 1970s, veterans of the Vietnam War were
being hospitalized in Department of Veterans Affairs hospital
psychiatry units (Haley, 1974) and receiving diagnoses of schizo-
phrenia or other psychotic disorders, even though combat-related
problems had been seen in World War II and Korean War veterans.
Moreover, at least two prescient scholars (Horowitz & Solomon,
1975) had predicted that the U.S. government would have to
respond to widespread stress-related problems in the Vietnam
veteran population, a prediction that was unfortunately all too
accurate. Among the phenomena that were most commonly re-
ported and observed in combat veterans were intrusive thoughts
and images, nightmares, social withdrawal, numbed feelings, hy-
pervigilance, and even frank paranoia, especially regarding the
government. Vivid dissociative phenomena, such as flashbacks,
occasionally characterized the symptom picture, and these symp-
toms may have centrally contributed to the misdiagnoses in the
psychotic realm.
Contemporaneously, clinicians began to recognize and write
about common patterns in the psychological sequelae of women
who had been sexually assaulted, and the term rape trauma syn-
drome entered the literature (Becker, 1982; Burgess & Holmstrom,
1974). The psychological suffering described was surprisingly
similar in terms of the processes, if not the vocabulary. These
women were observed to be avoidant, on guard, easily startled, and
flooded with memories and images of the assault that could not be
easily dispelled.
It required close to a decade, however, for investigators in both
fields of inquiry to recognize that these two separate syndromes
were more similar than different despite arising, for the most part,
in different genders and in different contexts. Increased interest in
Emily J. Ozer and Daniel S. Weiss, Department of Psychiatry, Univer-
sity of California, San Francisco; Suzanne R. Best and Tami L. Lipsey,
Mental Health Service, Veterans Affairs Medical Center, San Francisco,
California.
This article is based on a paper presented at the 14th Annual Meeting of
the International Society for Traumatic Stress Studies, Washington, DC, in
November of 1998.
Correspondence concerning this article should be addressed to either
Emily J. Ozer, who is now at the Department of Psychology and Social
Behavior, University of California, Irvine, California 92697-7085; or to
Daniel S. Weiss, Department of Psychiatry, Box 0984, University of
California, San Francisco, California 94143-0984. E-mail: eozer@uci.edu
or dweiss@itsa.ucsf.edu
Psychological Bulletin Copyright 2003 by the American Psychological Association, Inc.
2003, Vol. 129, No. 1, 52–73 0033-2909/03/$12.00 DOI: 10.1037/0033-2909.129.1.52
52

the study of posttraumatic symptoms and the consequent increase
in empirical investigations culminated in the introduction of post-
traumatic stress disorder (PTSD) into the American diagnostic
nomenclature in 1980 in the DSM–III (3rd ed.; American Psychi-
atric Association, 1980); this introduction was somewhat contro-
versial, even though the publication of the ninth edition of the
International Classification of Diseases (ICD–9; World Health
Organization, 1978) had indicated that organized psychiatry had
recognized the basic nucleus of the symptoms. In something of an
irony, the DSM–III also ushered in the elimination of any diagnosis
that encompassed the acute short-term effects of exposure to a
traumatic event. It has been only recently (4th ed.; DSM–IV;
American Psychiatric Association, 1994) that this disorder reap-
peared in the American nomenclature as acute stress disorder
(ASD), despite the knowledge of the short-term pattern of psycho-
logical reactions to traumatic stressors that had been acknowl-
edged and described for at least 50 years prior.
Since the introduction of DSM–III, the criteria for PTSD have
been revised twice, but neither revision has drastically modified
the fundamental set of symptomatic criteria. The core DSM–IV
criteria for PTSD consist of distressing symptoms of (a) reexpe-
riencing of the trauma (e.g., nightmares, intrusive thoughts), (b)
avoidance and numbing (e.g., avoiding reminders, not being able
to have loving feelings), and (c) increased arousal (e.g., difficulty
sleeping, hypervigilance, exaggerated startle response; American
Psychiatric Association, 1994). Besides the unusual feature of the
requirement of a traumatic (life-threatening) event in Criterion A,
a cardinal feature of PTSD is that the disabling aspects of the
phenomenology are linked to the event. Thus, intrusive images or
thoughts are typically of some aspect of the actual event, not just
random content that comes to mind with a distressing and intrusive
quality and cannot easily be dispelled once it has entered conscious
awareness.
Because the traumatic event is typically one of immediate life
threat and horror (e.g., sexual assault at knifepoint, torture, com-
bat, being in a flooded shelter during a hurricane and having the
water level quickly rise to one’s shoulders), there is also typically
very high adrenergic arousal that accompanies the experience.
Current neurobiological models of the acute stress response (e.g.,
Davis, 1992; LeDoux, 2000; McEwen, 1995) implicate the amyg-
dala and hippocampus as key brain areas that are involved in the
registration of potentially dangerous situations and in the later
formation of the memories of such events, and give the
hypothalamic–pituitary–adrenal (HPA) axis a central role in both
the development of PTSD and its maintenance (e.g., Yehuda,
1998). The underlying phenomenon that these elegant brain mod-
els are trying to explain is related to the long-known phenomenon
that memories formed under emotionally arousing situations be-
have differently from those that are not (Bower, 1981). There is
some evidence from preclinical and human studies (McGaugh &
Cahill, 1997) that memory formation under circumstances of emo-
tional arousal can be altered by blocking the effects of adrenalin.
This finding suggests that the degree of arousal during or imme-
diately after the traumatic event may have fundamental importance
for the development of the intrusive and hyperarousal symptoms of
PTSD through the mechanisms involved in both the formation and
recall of episodic memories. If this is so, an understanding of
which variables are good predictors of the development of symp-
toms as well as an understanding of some aspects of the disorder
itself need to examined from a more neurobiological perspective.
As well, such an understanding would help shed light on the thorny
issue of the wide heterogeneity of traumatic events that give rise to
PTSD. For example, a broadening of the types of events that some
have considered to be traumatic has led to inclusion in the PTSD
literature of studies of highly distressing events that carry infor-
mation about life threats (e.g., receiving a diagnosis of cancer) that
may or may not invoke the same adrenergic arousal that acute
life-threatening situations do. The presence or absence of adren-
ergic arousal may well become a key phenomenon that has impli-
cations for how symptoms of PTSD present and whether an event
is deemed “traumatic.”
1
If the subjective emotional and physio-
logical response to the event is overlooked, research that catego-
rizes events may not yield consistent findings that would perhaps
emerge if arousal were required. Indeed, the criteria for consider-
ing the presence of PTSD itself imply likely adrenergic arousal in
the phrase “intense fear, helplessness, or horror”in describing the
Criterion A response to exposure to the traumatic event.
Despite some suggestions that PTSD be moved from the anxiety
disorders of the DSM to the dissociative disorders, such a change
has little empirical basis because dissociative phenomena are not
routinely present in symptomatic presentations. As we shall dem-
onstrate, however, dissociative phenomena may play an important
role in predicting who develops PTSD, and we argue that the
dissociative phenomena are more closely related to activation of
the HPA axis than are predictors such as a family history of
psychopathology.
One obvious predictor of having PTSD symptoms is having met
the criteria for ASD during the month after exposure (the only
window during which it can be diagnosed), as a severe immediate
reaction might logically be implicated in having symptoms further
out from the trauma. It is clear that many who meet criteria for
PTSD some time after the exposure to the traumatic stressor do not
evidence symptoms of ASD; thus, ASD is not two-way pathogno-
monic. There is, however, less evidence and certainty about
whether having ASD is typically followed by chronic PTSD.
Because the diagnostic criteria of ASD include dissociative phe-
nomena, there may well be an important link between peritrau-
matic dissociation or other peritraumatic emotional responses and
the formation or manifestation of symptoms. Nonetheless, ASD is
beyond the scope of our efforts here.
The study of traumatic sequelae across cultures, types of trau-
matic exposure, and populations has increased dramatically in
recent years, with a burgeoning published literature (i.e., more
than 4,200 PsycINFO citations can be found using PTSD as a key
word). Of particular interest has been the search for factors that
explain why some people who are exposed to traumatic stress
develop PTSD whereas others similarly exposed do not (see
Brewin, Andrews, & Valentine, 2000), as the epidemiologic liter-
ature has made clear that PTSD is not an inevitable result of
exposure. The particular focus of our meta-analysis is a compari-
son of more static predictors to predictors more likely to be
implicated in the psychological and neurobiological processes
consequent to exposure to a traumatic stressor.
1
The conceptual discussion of the relationship between the event and
the normative response to the event as definitory of traumatic is beyond the
scope of this discussion but merits serious consideration elsewhere.
53
PREDICTORS OF PTSD: A META-ANALYSIS

Prevalence of PTSD
In the past decade, several large studies assessed the prevalence
of PTSD in adults. The National Vietnam Veterans Readjustment
Study (NVVRS; Kulka et al., 1990; Weiss et al., 1992), an exten-
sive national probability survey of male and female Vietnam
theater veterans (i.e., military service personnel, including health
professionals, who served in the Vietnam War), estimated
that 30.9% of men and 26.0% of women met the diagnostic criteria
for PTSD at some point since their service in Vietnam. The
estimate of current prevalence was 15.2% for men and 8.5% for
women (Schlenger et al., 1992). The percentage of veterans with
current clinically significant PTSD symptoms (i.e., partial PTSD)
was 11.1% for men and 7.8% for women. Combining estimates for
all current full and partial PTSD resulted in an estimate of roughly
830,000 Vietnam theater veterans who continued to experience
significant posttraumatic distress or impairment approximately 20
years after their exposure to one or more traumatic stressors
(Weiss et al., 1992).
Epidemiological research with civilian populations has found
lifetime PTSD prevalence rates to be 9.2% for adults belonging to
an urban health maintenance organization (Breslau, Davis, An-
dreski, & Peterson, 1991), 12.3% for a nationally representative
sample of women in the National Women’s Study (NWS; Resnick,
Kilpatrick, Dansky, Saunders, & Best, 1993), and 7.8% in the best
epidemiological study to date—the National Comorbidity Study
(NCS; Kessler, Sonnega, Bromet, Hughes, & Nelson, 1995)—a
nationally representative study of 5,877 people ages 15 to 45 years.
Of particular note is that the lifetime prevalence for women was
twice that for men in the NCS (10.4% vs. 5.0%). The prevalence
of PTSD is high among immigrants and refugees, particularly
those who immigrated because of armed conflict or political re-
pression, with a community study indicating that 52.0% of Central
American refugees from war or political persecution met criteria
for PTSD (Cervantes, Salgado de Snyder, & Padilla, 1989);
smaller studies with Cambodian refugees also indicate high levels
of PTSD (e.g., Carlson & Rosser-Hogan, 1993, 1994).
Another important finding from these epidemiological studies is
that the PTSD criterion in DSM–III and DSM–III–R(3rd ed., rev.;
American Psychiatric Association, 1987) referring to the charac-
teristics of the stressor (Criterion A, a necessary but not sufficient
condition for the diagnosis of PTSD) was empirically inaccurate in
that it required the event to be outside the range of normal human
experience. The NWS found that 69.0% of respondents reported
having been exposed to a traumatic event at some point in their
lives, a finding that was replicated in the NCS, which found a
lifetime prevalence of 60.7% for men and 51.2% for women for
exposure to an event that would qualify as a Criterion A event. As
a consequence, in the DSM–IV the phrase, “outside the range of
normal human experience,”was deleted and replaced with Crite-
rion A2, which requires that the person’s response to the stressor
be one of intense fear, helplessness, or horror.
The contrast between the lifetime prevalence of exposure to a
traumatic event of over 50% and the lifetime prevalence of PTSD
at roughly 7% was the disparity that led directly to the key aspect
of our investigation. Given that roughly 50%–60% of the U.S.
population is exposed to traumatic stress but only 5%–10% de-
velop PTSD, are there any systematic risk factors or correlates that
identify who will or will not develop the disorder? Moreover, if
there are, how strong are the effects? An important clue may be
found in the observation that there is considerable individual
variability in psychological response to traumatic stress.
Prior Reviews and Focus of the Present Meta-Analysis
Qualitative reviews of the PTSD literature highlighted the press-
ing need for attention to the role of other personal and environ-
mental variables in predicting PTSD (Emery, Emery, Shama,
Quiana, & Jassani, 1991; Fontana & Rosenheck, 1994; Green,
1994). Those reviews were neither comprehensive nor quantitative
(using measures of effect size [ES]) in their efforts to analyze the
adult PTSD literature.
An initial quantitative review of the predictors of PTSD in
adults was published during the period in which our work was
being conducted (Brewin et al., 2000). That meta-analysis studied
five demographic predictors (age, gender, socioeconomic status,
education, and race) and nine other variables. Though the authors
did not do so, these nine can be categorized into three groups: (a)
historical or static person characteristics such as family psychiatric
history, intelligence, childhood adversity and trauma, and other
previous trauma; (b) trauma severity; and (c) social support and
intercurrent life stress in the interval between traumatic exposure
and measurement of PTSD symptoms or the presence of the
disorder. The literature searching, study identification, and selec-
tion procedures in the Brewin et al. study yielded a set of 77
studies. Their main finding was that every predictor produced a
weighted average ES (r) that was statistically significant, but that
the effect of the predictors was quite heterogeneous: Race, the
weakest, was associated with an average ES of .05, whereas the
strongest, lack of social support, was associated with an average
weighted ES of .40. Most of the other ESs were in the range of .10
to .19. There was no particular conceptualization offered to ac-
count for the different ESs associated with the different predictors.
Another key finding of Brewin et al. (2000) was evidence of
considerable heterogeneity in effects for specific predictors, de-
pending on which trauma group was being analyzed; predictors did
not demonstrate consistent degrees of magnitude in every trauma
group in which they were studied. Moreover, the heterogeneity of
ES estimates within the studies for any individual predictor also
showed wide variability across the set of predictors, leading to
even less capability to draw generalizable conclusions. Brewin et
al. also examined the effects of sample and study characteristics on
ESs. As with the ESs themselves, these variables produced a
conflicting pattern of results. As an example, consider the com-
parison of studies with a civilian sample and studies with a military
sample. The average ES for social support was significantly dif-
ferent for the two types of samples, but the average ES for life
stress was not different in civilian and military samples. Similarly,
whether PTSD was indexed by a dichotomous diagnosis or a
continuous measure of symptoms made a significant difference on
the average ES for trauma severity but not for social support.
The main conclusion of Brewin et al. (2000) was that the set of
studies displayed remarkable heterogeneity. This heterogeneity
was both in the range of ESs across studies but within individual
predictors as well as within the sets of studies of individual
predictors, as these sets were moderated by characteristics of the
studies such as the type of sample or the method of measuring
PTSD symptoms or diagnosis. As a consequence, Brewin et al.
54 OZER, BEST, LIPSEY, AND WEISS

warned against attempting “to build a general vulnerability model
for all cases of PTSD”(p. 756), an approach that has been taken
with other mental disorders such as schizophrenia or bipolar dis-
order. Instead, they made the following suggestion:
The data may be regarded as consistent with a model in which the
impact of pretrauma factors on later PTSD is mediated by responses
to the trauma or, alternatively, with a model in which pretrauma
factors interact with trauma severity or trauma responses to increase
the risk of PTSD. In either case, [these data] suggest that it may be
more productive to investigate more proximal links in the causal
chain, such as the association between pretrauma risk factors and
immediate trauma responses. (Brewin et al., 2000, p. 756)
Quite independently of and prior to the conclusions of Brewin et
al. (2000), our focus was on only two types of predictors: (a)
person characteristics salient for psychological processing and
functioning and (b) aspects of the traumatic event or its sequelae,
just the factors that Brewin et al. suggested required greater em-
phasis and exploration. We omitted consideration of demographic
factors such as gender, education, and ethnicity, as none of these
factors is plausibly implicated in the psychological processes of
trauma response. As it turned out, Brewin et al.’s review showed
that the effects of these factors were relatively small and did not
lead to a significantly greater understanding of the development of
PTSD.
Whereas there are numerous individual and environmental fac-
tors that could be conceptually linked to the development of PTSD
symptoms, there has been sufficient empirical research on only a
small number of these predictors, and between the current study
and the work of Brewin et al. (2000), the entire universe of
predictors with even minimal empirical research has been covered.
This meta-analysis focused only on those nondemographic vari-
ables other than exposure to trauma itself that (a) could be legit-
imately conceptualized as predictors of PTSD symptoms and de-
velopment of the disorder, or that could influence the development
of PTSD symptoms before they are expressed; and (b) have been
studied adequately to generate sufficient data for meta-analytic
purposes (i.e., at least two empirical articles). Comorbid disorders
or associated symptoms occurring simultaneously with or after the
onset of PTSD were not considered as predictors (further criteria
for inclusion and exclusion are discussed in the Method section).
Because of the almost exclusively retrospective assessment meth-
odology of the PTSD literature, the large majority of predictors
studied in the meta-analysis were not assessed prior to the devel-
opment of PTSD symptoms (although this would have been far
preferable), though there are several exceptions to this rule
(Schnurr, Friedman, & Rosenberg, 1993; Shalev, Sahar, et al.,
1998).
The approach to the literature we have described led us to focus
on seven predictors of PTSD symptoms or of PTSD diagnosis that
had been sufficiently studied to include in this meta-analysis: (a) a
history of at least one other trauma prior to the index traumatic
event, (b) psychological adjustment prior to the traumatic event,
(c) family history of psychopathology, (d) perceived life threat
during the traumatic event, (e) perceived social support following
the traumatic event, (f) peritraumatic emotionality—high levels of
emotion during or in the immediate aftermath of the traumatic
event, and (g) peritraumatic dissociation—dissociative experi-
ences during or in the immediate aftermath of the traumatic event.
The meta-analysis did not include as a predictor the exposure to
the index traumatic event about which PTSD symptoms were
measured because exposure to such an event is a necessary crite-
rion for diagnosis of PTSD.
Our decision to use the conceptual term predictor rather than the
more operational term correlate, despite the largely retrospective
nature of the PTSD literature, was based on our desire to distin-
guish between factors that, if measured prospectively, could in-
deed serve as predictors from those factors that could not be
conceptualized as predictors even in prospective designs. Two
examples of the latter are comorbid diagnoses and survivor guilt.
In using the term predictor, we make no strong claim about
causality; instead, our conceptual framework is more akin to a risk
factor than to a causal factor.
Method
Sample of Studies
Relevant studies were initially identified through broad, computerized
database searches of MEDLINE, PsycINFO, and PILOTS for publications
between 1980 (the year PTSD was first included in the DSM) and 2000.
Key words entered were posttraumatic stress disorder, PTSD, trauma
(including combat trauma,rape trauma, and emotional trauma) and stress
(including stress reaction,stress response), and traumatic neurosis.We
used a range of key words to reflect the changing terminology in the
empirical literature over the course of the 2 decades covered here. In
addition, each issue of the Journal of Traumatic Stress (published by the
International Society for Traumatic Stress Studies)—beginning with Vol-
ume 1, No. 1, in January 1988 through Volume 13, No. 4, in October
2000—was reviewed for potentially pertinent research. We also reviewed
the reference sections of articles selected for study inclusion to identify any
additional relevant studies. Studies included for review were those in which
quantitative methods were used to examine the predictors of PTSD or
PTSD symptoms in adults (ages 18 and over). A predictor was defined as
any variable examined as a potential contributor to variability in PTSD
symptom levels or diagnostic status. Studies focusing on conditions co-
morbid with PTSD (e.g., substance abuse, depression) that were not as-
sessed prior to the onset of trauma were not included. Also excluded were
examinations of the longitudinal course of PTSD, as well as studies of
acute trauma response occurring prior to 1 month posttrauma. Single case
studies were also eliminated. Finally, we excluded doctoral dissertations,
reasoning that those dissertations whose positive findings would be rele-
vant would have had a high probability of being published. To address the
concern that unpublished data from dissertations or other research studies
might have findings that could alter the conclusions of our meta-analysis,
we conducted a “file drawer”calculation (Rosenthal, 1979, 1991) to test
the robustness of our findings for each ES (see below).
The search identified 2,647 articles in the English language. Our review
of the abstracts using the criteria outlined above identified 476 studies
eligible for study inclusion. In the second stage of review, additional
studies were eliminated because they did not specifically assess PTSD
symptoms (e.g., studies that reported only general symptoms) or because
they assessed some symptoms of PTSD but not all of the clusters as defined
by DSM–IV criteria (e.g., studies that assessed intrusion and avoidance and
numbing symptoms using the original Impact of Event Scale, Horowitz,
Wilner, & Alvarez, 1979, but did not assess hyperarousal). A subset of 47
studies was not included in the meta-analysis because it contained studies
that assessed the relationship between severity of exposure to traumatic
stress and PTSD symptoms only and did not assess any of the correlates we
considered. Decisions to exclude studies from the meta-analysis were
independently reviewed by Emily J. Ozer and Suzanne R. Best to ensure
that decisions were consistent with study inclusion criteria. These proce-
55
PREDICTORS OF PTSD: A META-ANALYSIS

dures yielded 68 empirical studies that we included in the meta-analysis.
Because of our reduced number of predictors and slightly different decision
rules for inclusion, our total number of studies was eight fewer than Brewin
et al. (2000). Our meta-analysis, however, included 21 studies that they did
not include. These nonoverlapping studies did not solely report on peri-
traumatic dissociation and emotional responses, the two novel predictors in
our meta-analysis. Nineteen of these studies were published before 2000,
and 2 studies were published during 2000 (although Brewin et al., 2000,
included studies published in 2000, it is likely that these two latter studies
were not possible to include in their meta-analysis). This number of
nonoverlapping studies excludes instances in which the two meta-analyses
report results for predictors from the same dataset but from published
articles with differing publication dates or authorship. We evaluated for
inclusion all studies included in the Brewin et al. meta-analysis.
Coding of Studies
All eligible studies were thoroughly reviewed by Emily J. Ozer and
Suzanne R. Best and were coded for the date of publication, country of
origin, size and demographics of study sample (i.e., age, ethnicity, gender),
type of trauma experienced, length of time elapsed between the trauma and
assessment of PTSD, clinical status of sample (i.e. community sample,
medical patients, individuals seeking mental health services), method and
measure for assessing PTSD, and a measure of ES for the relationship
between specific correlates and PTSD symptoms. In addition, we coded
methodological aspects of each study, including the following: (a) Was the
issue of missing data addressed by the study? and (b) Was PTSD assessed
as a categorical (i.e., met diagnostic criteria for PTSD) or continuous (i.e.,
symptom severity) variable? To reduce the possibility of coding errors,
Suzanne R. Best entered the codes, and then Emily J. Ozer rechecked them.
When studies neglected to provide a single degree of freedom (single-df)
ES, we used a combination of strategies consistent with established guide-
lines for meta-analytic research (Cooper & Hedges, 1994; Lipsey & Wil-
son, 2000) to derive ES estimates. When sufficient raw data were available
from the published study, we conducted ttests and chi-square analyses as
appropriate on the study data to obtain ESs. If the study did not provide
sufficient raw data or ESs but indicated the degrees of freedom and
significance level for a particular analysis, we estimated the ES as outlined
by Rosenthal (1994). If the study indicated that an analysis was not
statistically significant but provided no significance level or data (this
occurred for three ESs), we imputed an ES of zero for use in subsequent
analyses (Rosenthal & DiMatteo, 2000). Studies that did not provide
single-df ESs nor sufficient data to calculate a single-df ES were excluded
from the quantitative calculations; the findings of several epidemiological
studies that were excluded for this reason are reviewed in the Discussion
section with respect to the consistency of those studies’results with our
overall findings.
Selection and Calculation of ES Estimates
The studies reviewed for this meta-analysis were conducted using a wide
range of research designs and ESs. Because only one ES per construct per
study could be included in the meta-analysis (Rosenthal, 1994), we estab-
lished a set of decision rules consistent with standards for meta-analytic
analyses (Cooper & Hedges, 1994) for this diverse literature. First, ESs for
current rather than past PTSD symptoms were selected when both were
presented. Second, for longitudinal studies that presented multiple assess-
ments of symptoms over time, the ES representing the assessment of PTSD
symptoms closest in time to the trauma (but after 1 month—the minimum
time frame at which PTSD can be diagnosed—had elapsed) was selected.
Third, when researchers conducted analyses using both a diagnosis of
PTSD and symptom severity as outcome measures, we used the ES for
PTSD symptom severity because of the statistical advantages of continu-
ously measured variables in predictive research. Fourth, if a study reported
multiple ESs for the same general construct (e.g., separate ESs for family
history of phobia, generalized anxiety disorder, and depression), the ESs
were averaged (weighting for dfs) to calculate one overall ES for the
construct. Fifth, only single-df comparisons were appropriate for selection
in the meta-analysis (Rosenthal, 1994). Sixth, articles were reviewed
closely to ensure that the same study published in more than one article
would be counted as one study; authors, study design, and study sampling
were examined to ensure that no study contributed more than one ES per
construct.
We converted ES estimates for single-df analyses to the common metric
of Fisher’sztransformation of r(Rosenthal, 1994). ESs with artificial
dichotomization of PTSD symptoms were corrected using the adjustment
procedure recommended by Hunter and Schmidt (1990a, 1990b, 1994).
This correction takes into account the proportion of subjects in each half of
the dichotomy and increases the ES accordingly to make up for artificial
attenuation. Once appropriate adjustments were made, the mean of z
(weighted by the df associated with each ES) and 95% confidence limits
were calculated for each set of predictors using the more conservative
simple random effects approach (k⫽number of studies; Rosenthal, 1994).
Mean ESs and confidence limits were converted back to rfor ease of
interpretation. If the 95% confidence interval did not include zero, the null
hypothesis that the relationship between the specific predictor and PTSD
symptoms was zero could be rejected at the p⫽.05 level.
Cohen’s (1988) guidelines for interpreting the size of sample-weighted
average correlations were used: A small ES is r⫽.10; a medium ES is r⫽
.30; and r⫽.50 is a large ES.
Investigation of Within-Predictor Variability
We investigated the extent to which the ESs for each predictor varied
according to (a) the type of sample studied (i.e., community, medical
patients, or individuals seeking mental health services); (b) the length of
time that had elapsed between exposure to the traumatic event and assess-
ment of PTSD symptoms; (c) the type of trauma studied as the target
incident; and (d) the method used to assess either PTSD symptoms or
diagnosis. To facilitate between-groups comparisons, we categorized stud-
ies into three groups on the basis of the amount of time elapsed between
trauma exposure and PTSD assessment: 1 to 6 months, 6 months to 3 years,
and beyond 3 years. The selection of these time frames followed a rough
categorization to distinguish among acute reactions (1 to 6 months), clearly
chronic reactions (beyond 3 years), and intermediate-term reactions. With
regard to the type of trauma, events were broadly characterized as combat
exposure, interpersonal violence (i.e., human-perpetrated violence that
occurred in a civilian or nonmilitary context), accidents, or disasters.
Because of the small number of studies of disaster as the target event (three
in total and no more than one appearing in any one predictor category),
comparative analyses focused on the first three types of trauma only.
Method of assessment was dichotomized into self-report versus interviewer
or clinician assessment (in a few studies, the PTSD decision included both
self-report and interview data, and for these few studies, method of assess-
ment was assigned as missing). The significance of the difference between
the z-transformed correlation coefficients for each subgroup (divided on
the basis of type of sample, time elapsed since trauma, and type of trauma)
was then tested within each predictor (Cohen & Cohen, 1983).
To investigate how these moderator variables were related to each other
as well as to the ES they moderated, we conducted two sets of analyses.
The first was a set of chi-square analyses crossing each moderator with
every other. The second set was a series of hierarchical multiple regression
analyses focused on looking at the increment in variance accounted for by
a moderator in the presence of the other moderators.
Results
In this section we present the evidence for each predictor of
PTSD symptoms. The tables contain specific information regard-
56 OZER, BEST, LIPSEY, AND WEISS

ing each study included in the analysis for each predictor as well
as the combined ES r(unweighted and weighted for sample size)
and 95% confidence intervals for the overall unweighted and
weighted ESs.
History of Prior Trauma
The weighted average correlation from 23 studies (combined
n⫽5,308) of the relationship between a history of prior trauma
and PTSD symptoms or diagnosis was .17, a statistically signifi-
cant but small ES (see Table 1). On average, somewhat higher
levels of PTSD symptoms were reported by those who reported
experiencing a traumatic event prior to the target stressor than
those who indicated that they had not been previously exposed.
Individual ESs ranged from .00 to .46. Slightly more than half of
the studies (13) investigated the role of at least one prior trauma
that had occurred in childhood. Calculation of the average
weighted ES for this subgroup of studies yielded the identical ES
(r⫽.17) as the overall analysis produced, implying that, in
general, prior childhood trauma imparts no greater risk than prior
adult trauma.
The strength of the relationship between prior trauma and PTSD
did not differ according to the type of sample studied (i.e., com-
munity, medical patients, or individuals seeking mental health
services) nor the time elapsed since the trauma, nor the method by
which PTSD was assessed. The relationship between prior trauma
and PTSD did vary by the type of traumatic experience studied as
the target event: Having had a prior trauma was more strongly
related to PTSD when the traumatic experience involved noncom-
bat interpersonal violence (e.g., civilian assault, rape, domestic
violence; weighted r⫽.27) than when the traumatic experience
resulted from combat exposure (weighted r⫽.18; z⫽3.02, p⬍
.01) or an accident (weighted r⫽.12; z⫽2.10, p⬍.05).
Psychological Problems Prior to Target Stressor
Twenty-three studies (combined n⫽6,797) contributed to the
ES for prior adjustment problems. The weighted correlation be-
tween prior adjustment problems and PTSD symptoms or diagno-
sis was also .17, a statistically significant but small ES (see Table
2). This positive correlation shows that individuals who reported
problems in psychological adjustment prior to experiencing the
target stressor reported higher PTSD symptoms, on average, than
those who disavowed prior adjustment problems. The ESs ranged
from ⫺.13 to .47.
The types of prior adjustment problems associated with in-
creased PTSD symptoms included previous mental health treat-
ment (Carlier, Lamberts, & Gersons, 1997; Jeavons, Greenwood,
& Horne, 2000), pretrauma emotional problems (Ehlers, Mayou, &
Bryant, 1998), pretrauma anxiety or affective disorders (Blan-
chard, Hickling, Taylor, & Loos, 1995; Breslau et al., 1991;
Fauerbach et al., 1997; North, Smith, & Spitznagel, 1994; Resnick,
Kilpatrick, Best, & Kramer, 1992; Shalev, Freedman, et al., 1998;
Solomon, Oppenheimer, Elizur, & Waysman, 1990; Tedstone &
Tarrier, 1997; Ursano et al., 1999), and antisocial personality
disorder prior to military service (Cottler, Compton, Mager,
Spitznagel, & Janca, 1992). In two of the genuinely prospective
studies in the field (i.e., the predictor actually was assessed prior to
the target event), elevated precombat Minnesota Multiphasic Per-
sonality Inventory scores predicted PTSD symptoms among Viet-
nam combat veterans (Schnurr et al., 1993) and precombat neu-
roticism predicted PTSD among WWII veterans (Lee, Vaillant,
Torrey, & Elder, 1995). Preservice psychiatric problems, however,
were found to be negatively related to PTSD symptoms in one
study of WWII prisoners of war (Speed, Engdahl, Schwartz, &
Eberly, 1989).
Because a small cluster of studies assessed prior depression as
the predictor representing prior adjustment, it was possible to
examine whether studies in which depression was the prior psy-
chological problem had a different average ES from the remainder
of the studies that had examined the impact of other prior adjust-
ment problems. The results of this comparison were informative.
The weighted rfor the four studies of depression was .32, and the
comparison of this value to the remaining studies (weighted r⫽
.15) was statistically significant (z⫽3.78, p⬍.01). This was a
conservative comparison because studies that examined any prior
affective or Axis I disorder (including depression) were included
in the remainder.
The strength of the relationship between prior adjustment prob-
lems and PTSD symptoms or diagnosis did not vary according to
the type of sample studied. The relationship between prior adjust-
ment problems and PTSD did differ as a function of the type of
traumatic experience that constituted the target event, the amount
of time elapsed, and the method of assessing PTSD. Having prior
adjustment problems was more strongly related to PTSD when the
traumatic experience involved noncombat interpersonal violence
(weighted r⫽.31) or accident (weighted r⫽.28) than when the
traumatic experience resulted from combat exposure (weighted
r⫽.06; for interpersonal violence vs. combat, z⫽8.70, p⬍.01;
for accident vs. combat, z⫽4.72, p⬍.01). Stronger relationships
were found between prior adjustment problems and PTSD among
studies in which less time had elapsed between the traumatic event
and assessment of PTSD (1 to 6 months [weighted r⫽.24] vs. 6
months to 3 years [weighted r⫽⫺.02]; z⫽5.14, p⬍.01, or 3
years and longer [weighted r⫽.11]; z⫽3.35, p⬍.01). The
weighted ES for prior adjustment was higher for interview studies
(r⫽.29) than for self-report studies (weighted r⫽.15; z⫽5.31,
p⬍.01).
Psychopathology in Family of Origin
The weighted average correlation for nine studies (combined
n⫽667) of the relationship between psychopathology in the
family of origin and PTSD symptoms or diagnosis was also
significantly different from zero at .17 (see Table 3). This was a
small ES, indicating that individuals who reported a family history
of psychopathology reported higher PTSD symptoms or higher
rates of PTSD than those without such a family history. The ESs
for this predictor ranged from ⫺.06 to .43.
The strength of the relationship between family history of psy-
chopathology and PTSD symptoms or diagnosis did not vary
according to the type of sample or the time elapsed between the
traumatic event and the assessment of PTSD. The relationship
between family history of psychopathology and PTSD, however,
did vary by the type of traumatic experience that constituted the
target event and the method by which PTSD was assessed. Having
a family history of psychopathology was more strongly related to
(text continues on page 61)
57
PREDICTORS OF PTSD: A META-ANALYSIS

Table 1
Weighted Correlations of Prior Trauma With Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms
Study Sample Broad trauma type Clinical status Trauma recency Type of prior trauma PTSD measure df r
a
Andrews et al. (2000) Crime victims Interpersonal violence Medical 6 months Childhood abuse PSS 157 .21
Astin et al. (1995) Domestic violence victims Interpersonal violence Medical Variable Childhood abuse SCID (DSM–III–R) 87 .17
Blanchard et al. (1995) Motor vehicle accident
victims Accident Medical 1–4 months Prior trauma CAPS 158 .02
Bremner et al. (1993) Vietnam veterans Combat Clinical 19 years Prior trauma SCID 66 .31
b
Breslau et al. (1998) Community Variable Community Variable Mixed DIS (DSM–IV) 1922 .14
Dunmore et al. (1999) Assault survivors Interpersonal violence Community Variable Any prior trauma not including
child abuse PSS–SR 92 .30
Engel et al. (1993) Desert Storm veterans Combat Clinical 7–12 months Precombat abuse MCS for Desert Storm 292 .19
Fontana et al. (1997) Female Vietnam veterans
(NVVRS sample) Combat Community 20⫹years Adult sexual abuse MCS, CAPS 396 .27
Jeavons et al. (2000) Motor vehicle accident
victims Accident Medical 3 months Not specified PTSD–I 62 .03
Kemp et al. (1995) Female victims of
domestic violence and
verbal abuse
Interpersonal violence Variable Variable Childhood abuse MCS–CV 227 .14
Koopman et al. (1994) Fire survivors Accident Community 7–9 months Prior trauma and highly
stressful life events MCS–CV 154 .26
Kramer & Green (1991) Women treated for sexual
assault Interpersonal violence Medical Variable Previous sexual assault Semistructured interview
based on DSM–III 30 .40
Michaels et al. (1999) Motor vehicle accident
victims Accident Medical 6 months Prior abuse MCS–CV 176 .18
North & Smith (1992) Homeless Varied Community Variable Child physical abuse DIS 522 .21
Perrin et al. (1996) Female domestic violence
victims Interpersonal violence Clinical Not specified Prior battering or rape as adult MMPI–PK 69 .11
Resnick et al. (1995) Rape victims Interpersonal violence Medical 17–57 days Previous assault SCID (DSM–III–R) 37 .43
Riggs et al. (1992) Female crime victims Interpersonal violence Mixed 1 month Prior sexual assault PTSD Symptom Scale 86 .15
Roberts et al. (1998) Female domestic violence
victims Interpersonal violence Medical Variable Child abuse PTSD Checklist 161 .24
Solomon (1995) Israeli combat veterans Combat Community 18 years Childhood exposure to war IES, PTSD–I 348 .03
Speed et al. (1989) WWII prisoners of war Combat Community 40⫹years Child trauma Rating scale based on
DSM–III 62 .16
Ursano, Fullerton,
Epstein, Crowley,
Kao, et al. (1999)
Motor vehicle accident
victims Accident Medical 1 month Prior trauma of any kind SCID (DSM–III–R, DSM–
IV)122 .00
c
Van Velsen et al. (1996) Torture victims Interpersonal violence Medical Variable History of other persecution Clinical diagnosis using
DSM–III 60 .37
b
Zaidi & Foy (1994) Vietnam combat veterans Combat Clinical Variable Child physical abuse MCS 22 .46
Unweighted average r⫽.21 (CI ⫽.16, .27)
Weighted average r⫽.17 (CI ⫽.11, .22)
Note. Average rcalculated from sample-weighted correlation for each effect size. PSS ⫽Posttraumatic Stress Disorder Symptom Scale (Foa, Riggs, Dancu, & Rothbaum, 1993); SCID ⫽Structured
Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders ([3rd ed.; DSM–III; American Psychiatric Association, 1980]; Spitzer & Williams, 1983); DSM–III–R⫽Diagnostic and
Statistical Manual of Mental Disorders (3rd ed., rev.; American Psychiatric Association, 1987); CAPS ⫽Clinician Administered Posttraumatic Stress Disorder Scale (Blake et al., 1995); DIS ⫽
Diagnostic Interview Schedule (Robins, Helzer, Croughan, & Ratcliff, 1981); DSM–IV ⫽Diagnostic and Statistical Manual of Mental Disorders (4th ed., American Psychiatric Association, 1994);
PSS–SR ⫽PSS—Self-Report; MCS ⫽Mississippi Scale for Combat-Related Posttraumatic Stress Disorder (Keane, Caddell, & Taylor, 1988); NVVRS ⫽National Vietnam Veterans Readjustment
Study (Kulka et al., 1990); PTSD–I⫽PTSD Inventory (Solomon et al., 1993); MCS–CV ⫽MCS—Civilian Version (Lauterback, Vrana, King, & King, 1997); MMPI–PK ⫽Minnesota Multiphasic
Personality Inventory PTSD—Keane Scale (Keane, Malloy, & Fairbank, 1984); IES ⫽Impact of Event Scale (Horowitz, Wilner, & Alvarez, 1979); CI ⫽95% confidence interval.
a
Positive correlation indicates presence of prior trauma associated with PTSD symptoms.
b
Effect size adjusted for dichotomization of PTSD variable.
c
Effect size of zero imputed because study
indicated only that analysis was “not statistically significant.”
58 OZER, BEST, LIPSEY, AND WEISS

Table 2
Weighted Correlations of Psychological Adjustment Prior to the Traumatic Event With Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms
Study Sample Broad trauma type Clinical status Trauma recency Prior adjustment assessment method PTSD measure df r
a
Basoglu et al. (1994) Survivors of torture Interpersonal violence Community Variable Prior psychopathology SCID 55 .09
b
Blanchard et al. (1995) Motor vehicle accident
victims Accident Medical 1–6 months Depression prior to trauma CAPS 158 .35
c
Boscarino (1995) Vietnam combat and era
veterans Combat Community 3⫹years Self-reported child/adolescent delinquency DIS 2490 .06
c
Carlier et al. (1997) Police officers with and
without PTSD Interpersonal violence Community 1–6 months Previous psychiatric care SI–PTSD 262 .26
c
Cottler et al. (1992) Community survey
respondents Mixed Community Variable History of antisocial personality disorder DIS 430 .12
c
Dunmore et al. (1999) Physical assault victims Interpersonal violence Community Variable Preassault psychological problems PTSD Scale 92 .31
c
Ehde et al. (2000) Burn victims Accident Medical 1–6 months Prior positive mental health Checklist based on
DSM–III–R68 .13
Ehlers et al. (1998) Motor vehicle accident
victims Interpersonal violence Medical 1–6 months Preaccident emotional problems PSS 888 .20
Fauerbarch et al. (1997) Burn patients Accident Medical 1–6 months Preaccident anxiety or affective disorder SCID (DSM–III–R) 95 .34
c
Fontana et al. (1997) Female Vietnam veterans
(NVVRS sample) Combat Community 3⫹years Premilitary conduct disorder MCS, CAPS 396 .05
Foy et al. (1984) Vietnam combat veterans Combat Clinical 3⫹years Premilitary mental health contacts Self-report checklist 43 .07
Jeavons et al. (2000) Motor vehicle accident
victims Accident Medical 1–6 months Any prior psychiatric treatment PTSD–I 62 .31
Lee et al. (1995) Longitudinal study of
WWII veterans Combat Community 6 months to 3 years Neuroticism assessed prior to military
service Checklist based on
DSM–III 107 .20
Madakasira & O’Brien
(1987) Tornado victims Disaster Community 1–6 months History of mental illness Adapted Hopkins
Checklist 90 116 .03
c
North & Smith (1992) Homeless Mixed Community Variable Any prior psychiatric diagnosis DIS 522 .33
c
North et al. (1994) Survivors of shooting spree Interpersonal violence Community 1–6 months Predisaster depression DIS 124 .41
c
Resnick et al. (1992) Crime victims Interpersonal violence Community 3⫹years Prior Axis I disorder Adapted DIS using
DSM–III criteria 295 .47
c
Schnurr et al. (1993) Vietnam era veterans Combat Community 3⫹years MMPI Paranoia Scale assessed before
military service SCID (DSM–III–R) 91 .31
c
Shalev, Freedman, et al.
(1998) Medical patients Mixed Medical 1–6 months Depression prior to trauma MCS–CV 211 .28
Solomon et al. (1990) Combat veterans Combat Clinical 6 months to 3 years Prior combat stress reaction PTSD–I 63 .09
Speed et al. (1989) WWII prisoners of war Combat Community 3⫹years Preservice psychiatric problem Rating scale based
on DSM–III 62 ⫺.13
Tedstone & Tarrier (1997) Burn victims Accident Medical 1–6 months Prior depression Penn Inventory 45 .17
c
Ursano, Fullerton, Epstein,
Crowley, Kao, et al.
(1999)
Motor vehicle accident
victims Accident Medical 1–6 months Prior anxiety or depressive disorder SCID (DSM–III–R,
DSM–IV)122 .25
c
Unweighted average r⫽.21 (CI ⫽.13, .26)
Weighted average r⫽.17 (CI ⫽.10, .23)
Note. Average rcalculated from sample-weighted correlation for each effect size. SCID ⫽Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders ([3rd ed.; DSM–III; American
Psychiatric Association, 1980]; Spitzer & Williams, 1983); CAPS ⫽Clinician Administered Posttraumatic Stress Disorder Scale (Blake et al., 1995); DIS ⫽Diagnostic Interview Schedule (Robins, Helzer,
Croughan, & Ratcliff, 1981); SI–PTSD ⫽Structured Interview for Posttraumatic Stress Disorder (Davidson, Kudler, & Smith, 1990); DSM–III–R⫽Diagnostic and Statistical Manual of Mental Disorders (3rd
ed., rev.; American Psychiatric Association, 1987); PSS ⫽Posttraumatic Stress Disorder Symptom Scale (Foa, Riggs, Dancu, & Rothbaum, 1993); NVVRS ⫽National Vietnam Veterans Readjustment Study
(Kulka et al., 1990); MCS ⫽Mississippi Scale for Combat-Related Posttraumatic Stress Disorder (Keane, Caddell, & Taylor, 1988); PTSD–I⫽PTSD Inventory (Solomon et al., 1993); WWII ⫽World War
II; MMPI ⫽Minnesota Multiphasic Personality Inventory (Hathaway & McKinley, 1951); MCS–CV ⫽MCS—Civilian Version (Lauterbach, Vrana, King, & King, 1997); DSM–IV ⫽Diagnostic and Statistical
Manual of Mental Disorders (4th ed., American Psychiatric Association, 1994); CI ⫽95% confidence interval.
a
Positive correlation indicates poorer prior psychological adjustment associated with PTSD symptoms.
b
Effect size of zero imputed because study indicated only that analysis was “not statistically
significant.”
c
Effect size adjusted for dichotomization of PTSD variable.
59
PREDICTORS OF PTSD: A META-ANALYSIS

Table 3
Weighted Correlations of Family History of Psychopathology With Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms
Study Sample Broad
trauma type Clinical status Trauma
recency Family psychopathology
assessment method PTSD measure df r
a
Basoglu et al. (1994) Turkish torture survivors Interpersonal
violence Community Variable Family history of
psychopathology SCID 55 .36
Emery et al. (1991) Vietnam veterans Combat Clinical 20⫹years Alcohol-dependent parent Semistructured interview,
MMPI and IES 40 .34
b
Jeavons et al. (2000) Motor vehicle accident
victims Accident Medical 3 months Psychiatric or psychological
treatment in family PTSD–I 62 .05
McFarlane (1988) Firefighters who responded
to Australian bush fire Accident Community 4 months Family history of
psychopathology Clinical interview using
DSM–III criteria 45 .41
b
Reich et al. (1996) VA patients Combat Medical Variable Family history of GAD or
depression SCID (DSM–III–R) 82 .12
b
Skre et al. (1993) Twins (one proband had
received treatment for
mental disorder)
Not
specified Clinical Not
specified Twin diagnosed with
anxiety disorder SCID (DSM–III–R) 81 .43
b
Speed et al. (1989) WWII prisoners of war Combat Community 40⫹years Family history of
psychopathology Rating scale based on
DSM–III 62 ⫺.06
Ursano, Fullerton,
Epstein, Crowley,
Kao, et al. (1999)
Motor vehicle accident
victims Accident Medical 1 month Family history of
psychopathology SCID (DSM–III–R,
DSM–IV)122 .00
c
Watson et al. (1996) Vietnam veterans Combat Mixed 20⫹years Any psychiatric diagnosis
in family PTSD–I 118 .14
b
Unweighted average r⫽.21 (CI ⫽.08, .33)
Weighted average r⫽.17 (CI ⫽.04, .29)
Note. Average rcalculated from sample-weighted correlation for each effect size. SCID ⫽Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders ([3rd ed.; DSM–III;
American Psychiatric Association, 1980]; Spitzer & Williams, 1983); MMPI ⫽Minnesota Multiphasic Personality Inventory (Hathaway & McKinley, 1951); IES ⫽Impact of Event Scale (Horowitz,
Wilner, & Alvarez, 1979); PTSD–I⫽PTSD Inventory (Solomon et al., 1993); DSM–III–R⫽Diagnostic and Statistical Manual of Mental Disorders (3rd ed., rev.; American Psychiatric Association,
1987); VA ⫽Veterans Administration; DSM–IV ⫽Diagnostic and Statistical Manual of Mental Disorders (4th ed.; American Psychiatric Association, 1994); CI ⫽95% confidence interval.
a
Positive correlation indicates history of family psychopathology associated with PTSD symptoms.
b
Effect size adjusted for dichotomization of PTSD variable.
c
Effect size of zero imputed because
study indicated only that analysis was “not statistically significant.”
60 OZER, BEST, LIPSEY, AND WEISS

PTSD when the traumatic experience involved noncombat inter-
personal violence (weighted r⫽.31) than when the traumatic
experience was combat exposure (weighted r⫽.12; z⫽3.40, p⬍
.01) or accident (weighted r⫽.08; z⫽3.00, p⬍.01). The
weighted rof .28 for interview assessment was much stronger than
the weighted rof .04 for self-report (z⫽3.42, p⬍.01).
Perceived Life Threat
The weighted average correlation for 12 studies (combined
n⫽3,524) of the relationship between perceived life threat and
PTSD symptoms or diagnosis was .26, a statistically significant
effect in the small-to-medium range (see Table 4). Individuals who
perceived that their life was in danger during their index traumatic
event reported higher levels of PTSD symptoms or higher rates of
current PTSD. The ESs ranged from .13 to .49.
The strength of the relationship between perceived life threat
during the traumatic event and PTSD diagnosis or symptoms did
not vary by the type of sample studied nor by the method used to
assess PTSD. The strength of the relationship between perceived
life threat and PTSD, however, was higher among studies with
more time elapsed between the traumatic event and the assessment
of PTSD (i.e., 6 months to 3 years [weighted r⫽.44] vs. 1 to 6
months [weighted r⫽.24]; z⫽2.23, p⬍.05). Furthermore,
perceived life threat during the traumatic event was more strongly
associated with PTSD when the traumatic experience was non-
combat interpersonal violence (weighted r⫽.36) than when the
traumatic experience was an accident (weighted r⫽.20; z⫽2.44,
p⬍.05).
Perceived Support Following Trauma
The weighted average correlation for 11 studies (combined
n⫽3,537) of the relationship between perceived social support
following the trauma and PTSD symptoms was ⫺.28, a statisti-
cally significant ES in the small-to-medium range (see Table 5).
Individuals reporting lower levels of perceived social support after
the traumatic event reported higher levels of PTSD symptoms or
rates of current PTSD. The ESs ranged from ⫺.57 to .07.
The strength of the relationship between perceived social sup-
port after the traumatic event and PTSD did not vary by the type
of sample studied or by the method used to assess PTSD symptoms
or diagnosis. The inverse relationship between perceived support
and PTSD (i.e., those reporting more support showed lower rates
of meeting diagnostic criteria or lower symptom levels), however,
was strongest among studies with more time elapsed between the
traumatic event and the assessment of PTSD. That is, an incre-
mental pattern was found in which the inverse relationship was
strongest in studies in which more than 3 years had elapsed
between the traumatic event and assessment of PTSD (weighted
r⫽⫺.42), followed by studies in which 6 months to 3 years had
elapsed (weighted r⫽⫺.16), followed by studies in which 1 to 6
months had elapsed (weighted r⫽.01; z⫽2.64, p⬍.01,1to6
months vs. 6 months to 3 years; z⫽7.50, p⬍.01, 1 to 6 months
Table 4
Weighted Correlations of Perceived Life Threat With Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms
Study Sample Broad trauma type Clinical status Trauma recency PTSD measure df r
a
Bernat et al. (1998) College students Variable Community Variable IES–R 349 .21
Blanchard et al.
(1995) Motor vehicle accident victims Accident Medical 1–4 months CAPS 98 .23
David et al. (1996) Hurricane victims Accident Medical 6–12 months SCID (DSM–III–R) 61 .29
b
Dunmore et al.
(1999) Survivors of sexual and
physical assault Interpersonal violence Community Variable PSS 91 .49
b
Ehlers et al. (1998) Motor vehicle accident victims Accident Medical 3 months PSS 888 .23
b
Jeavons et al. (2000) Motor vehicle accident victims Accident Medical 3 months PTSD–I 62 .46
Kemp et al. (1995) Female victims of domestic
violence and verbal abuse Interpersonal violence Variable Variable MCS–CV 168 .15
Kilpatrick et al.
(1989) Female crime victims Interpersonal violence Community Variable DIS 294 .40
b
King et al. (1998) Vietnam veterans (NVVRS
sample) Combat Community 20⫹years MCS, DIS 1224 .26
McFarlane (1988) Firefighters of Australian bush
fire Accident Community 4 months Structured interview
based on DSM–III 44 .13
b
Michaels et al.
(1999) Accident victims Accident Medical 6 months MCS–CV 176 .18
b
Perrin et al. (1996) Domestic violence victims Interpersonal violence Clinical Not specified MMPI–PK 69 .34
b
Unweighted average r⫽.29 (CI ⫽.21, .36)
Weighted average r⫽.26 (CI ⫽.18, .34)
Note. Average rcalculated from sample-weighted correlation for each effect size. IES–R⫽Impact of Event Scale—Revised (Weiss & Marmar, 1997);
CAPS ⫽Clinician Administered Posttraumatic Stress Disorder Scale (Blake et al., 1995); SCID ⫽Structured Clinical Interview for Diagnostic and
Statistical Manual of Mental Disorders [(3rd ed.; DSM–III; American Psychiatric Association, 1980]; Spitzer & Williams, 1983); DSM–III–R⫽Diagnostic
and Statistical Manual of Mental Disorders (3rd ed., rev.; American Psychiatric Association, 1987); PSS ⫽Posttraumatic Stress Disorder Symptom Scale
(Foa, Riggs, Dancu, & Rothbaum, 1993); PTSD–I⫽PTSD Inventory (Solomon et al., 1993); MCS–CV ⫽Mississippi Scale for Combat-Related
Posttraumatic Stress Disorder—Civilian Version (Lauterbach, Vrana, King, & King, 1997); DIS ⫽Diagnostic Interview Schedule (Robins, Helzer,
Croughan, & Ratcliff, 1981); MCS ⫽Mississippi Scale for Combat-Related Posttraumatic Stress Disorder (Keane, Caddell, & Taylor, 1988); NVVRS ⫽
National Vietnam Veterans Readjustment Study (Kulka et al., 1990); MMPI–PK ⫽Minnesota Multiphasic Personality Inventory PTSD—Keane Scale
(Keane, Malloy, & Fairbank, 1984); CI ⫽95% confidence interval.
a
Positive correlation indicates perceived life threat associated with PTSD symptoms.
b
Effect size adjusted for dichotomization of PTSD variable.
61
PREDICTORS OF PTSD: A META-ANALYSIS

Table 5
Weighted Correlations of Social Support With Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms
Study Sample Broad trauma type Clinical status Trauma recency Support assessment
a
PTSD measure df r
b
Astin et al. (1993) Battered women Interpersonal violence Medical Mean of 8
months Perceived support PTSD Symptom Checklist 53 ⫺.18
Green & Berlin (1987) Vietnam veterans Combat Clinical 30 years If spoke with others about
Vietnam during first
year postwar
Figley Rating Scale 60 ⫺.26
Hough et al. (1990) Community where sniper
massacre occurred Interpersonal violence Clinical 6 months Availability of a confidant DIS 290 .07
King et al. (1998) Vietnam veterans
(NVVRS theater) Combat Community 20⫹years Functional support after
return from Vietnam MCS, DIS 1224 ⫺.43
Kramer & Green
(1991) Women treated for
sexual assault Interpersonal violence Medical Not specified Use of informal support
network to discuss the
rape
Semistructured interview
based on DSM–III
criteria and IES
30 ⫺.42
Perrin et al. (1996) Women who had
experienced domestic
violence
Interpersonal violence Clinical Not specified Interpersonal Support
Evaluation List MMPI–PK 69 ⫺.57
Perry et al. (1992) Burn victims Accident Medical 2 months Interpersonal Support
Evaluation List SCID 50 ⫺.44
c
Solomon et al. (1987) Israeli combat veterans Combat Community 12 months Perceived support during
war PTSD–I–SR 683 ⫺.18
Sutker et al. (1995) Persian Gulf War
veterans Combat Community Within 1 year Number of support
providers MCS for Desert Storm 580 ⫺.18
Weiss et al. (1995) EMS personnel Mixed Community Variable Current Social Support
Scale (NVVRS) MCS–CV 367 ⫺.25
Wolfe et al. (1998) Female veterans of
Persian Gulf War Combat Community 18–24 months Leader support during war MCS 131 ⫺.25
Unweighted average r⫽⫺.29 (CI ⫽⫺.41, ⫺.16)
Weighted average r⫽⫺.28 (CI ⫽⫺.40, ⫺.15)
Note. Average rcalculated from sample-weighted correlation for each effect size. DIS ⫽Diagnostic Interview Schedule (Robins, Helzer, Croughan, & Ratcliff, 1981); NVVRS ⫽National Vietnam
Veterans Readjustment Study (Kulka et al., 1990); MCS ⫽Mississippi Scale for Combat-Related Posttraumatic Stress Disorder (Keane, Caddell, & Taylor, 1988); DSM–III ⫽Diagnostic and Statistical
Manual of Mental Disorders (3rd ed.; American Psychiatric Association, 1980); IES ⫽Impact of Event Scale (Horowitz, Wilner, & Alvarez, 1979); MMPI–PK ⫽Minnesota Multiphasic Personality
Inventory PTSD—Keane Scale (Keane, Malloy, & Fairbank, 1984); SCID ⫽Structured Clinical Interview for DSM–III (Spitzer & Williams, 1983); PTSD–I–SR ⫽PTSD Inventory—Self-Report
(Solomon et al., 1993); EMS ⫽emergency medical service; MCS–CV ⫽MCS—Civilian Version (Lauterback, Vrana, King, & King, 1997); CI ⫽95% confidence interval.
a
Support measures all assessed self-reported, perceived support.
b
Negative correlation indicates high support associated with lower PTSD symptoms.
c
Effect size adjusted for dichotomization of
PTSD variable.
62 OZER, BEST, LIPSEY, AND WEISS

vs. 3 years or more; z⫽6.58, p⬍.01, 6 months to 3 years vs. 3
years or more). Furthermore, a stronger inverse relationship be-
tween perceived support and PTSD was found for studies of
combat trauma (weighted r⫽⫺.26) than for studies of noncombat
interpersonal violence (weighted r⫽⫺.11; z⫽2.85, p⬍.01).
Peritraumatic Emotional Responses
The weighted average correlation for five studies (combined
n⫽1,755) of the relationship between self-reports of emotional
responses at the time of the traumatic event and PTSD symptoms
was .26, statistically significant even given the small number of
studies (see Table 6). These studies showed that individuals who
described having intensely negative emotional responses during or
immediately after the index traumatic event reported appreciably
higher levels of PTSD symptoms or rates of current PTSD. The
kinds of emotional responses studied in the literature included fear,
helplessness, horror, guilt, and shame. The ESs ranged from .15 to
.55. No within-group comparisons by sample, type of trauma, or
time elapsed between trauma and PTSD assessment were possible
for this correlate because of insufficient quantity of studies in each
subgroup category. Assessment method was not analyzable be-
cause no study used an interview exclusively.
Peritraumatic Dissociation
The weighted average correlation for 16 studies (combined
n⫽3,534) of the relationship between reports of dissociative
experiences (either self-report or interviewer assessment) during
and immediately after the traumatic event and PTSD symptoms or
diagnosis was .35, a statistically significant ES in the medium
range (see Table 7). These studies showed that individuals who
described having dissociative experiences during or immediately
after the traumatic event reported appreciably higher levels of
PTSD symptoms or rates of current disorder. The ESs ranged from
.14 to .94. The coefficient of .94 was a conversion from a t-test
comparison in which the means of a modified dissociation scale
were 11.5 (SD ⫽1.6) and 1.8 (SD ⫽2.1), respectively (Bremner
et al., 1992). To assess the impact of this admittedly extreme ES,
we recalculated the weighted ES omitting the Bremner et al.
(1992) study. The result was a weighted rof .34; thus, we chose to
retain this study in the analyses because it did not unduly influence
the overall ES.
The strength of the relationship between peritraumatic dissoci-
ation and PTSD symptoms or diagnosis varied as a function of the
time elapsed between (a) the trauma and measurement of symp-
toms, (b) the type of sample, and (c) the method for assessment of
symptoms, but it did not vary by the type of event. Peritraumatic
dissociation was most strongly related to PTSD among individuals
seeking mental health services (weighted r⫽.60) as compared
with medical (weighted r⫽.33; z⫽3.81, p⬍.01) or community
samples (weighted r⫽.35; z⫽3.96, p⬍.01). With respect to the
time frame between the trauma and assessment of PTSD, the
relationship between peritraumatic dissociation and PTSD symp-
toms or diagnosis was highest among studies in which 6 months
to 3 years had elapsed between the traumatic event and the assess-
ment of PTSD (weighted r⫽.45) compared with the time frames
of 1 to 6 months (weighted r⫽.31; z⫽2.51, p⬍.05) and more
than 3 years (weighted r⫽.30; z⫽2.61, p⬍.01). For studies in
which PTSD was indexed by self-report measures, the weighted r
was .48 as compared with a weighted rof .21 for studies using
interview methods. This difference was statistically significant
(z⫽7.81, p⬍.01).
Table 6
Weighted Correlations of Peritraumatic Emotions With Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms
Study Sample Broad
trauma type Clinical status Trauma
recency Peritraumatic emotion
assessment method PTSD measure df r
a
Bernat et al.
(1998) College students Mixed Community sample Variable Initial Subjective Reaction
Emotional Scale IES–R 349 .41
Brewin et al.
(2000) Crime victims Interpersonal
violence Combination of
medical,
community
6 months Any intense fear,
helplessness, or horror PSS 138 .55
b
Ehlers et al.
(1998) Motor vehicle accident
victims recruited from
ERs
Accident Medical patient 3 months Rating of how frightening
the accident was PSS 888 .28
Epstein et al.
(1998) Military health personnel Disaster Community sample 6–18 months Initial anxiety/fright coded
from qualitative reports Items from
SCL–90–R,
DSM–III–R
306 .20
b
Roemer et al.
(1998) College students Mixed Community sample Not specified Single items assessing
fear, helplessness, and
horror
PTSD Checklist 74 .15
Unweighted average r⫽.28 (CI ⫽.10, .42)
Weighted average r⫽.26 (CI ⫽.08, .42)
Note. Average rcalculated from sample-weighted correlation for each effect size. IES–R⫽Impact of Event Scale—Revised (Weiss & Marmar, 1997);
PSS ⫽Posttraumatic Stress Disorder Symptom Scale (Foa, Riggs, Dancu, & Rothbaum, 1993); ERs ⫽emergency rooms; SCL–90–R⫽Symptom
Checklist—90—Revised (Derogatis, 1992); DSM–III–R⫽Diagnostic and Statistical Manual of Mental Disorders (3rd ed., rev.; American Psychiatric
Association, 1987); CI ⫽95% confidence interval.
a
Positive correlation indicates positive association between peritraumatic emotional response and PTSD symptoms.
b
Effect size adjusted for dichoto-
mization of PTSD variable.
63
PREDICTORS OF PTSD: A META-ANALYSIS

Table 7
Weighted Correlations of Peritraumatic Dissociation With Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms
Study Sample Broad
trauma type Clinical status Trauma
recency Peritraumatic dissociation
assessment method PTSD measure df r
a
Bernat et al. (1998) College students Mixed Community Variable PDEQ–RV IES–R 349 .35
Bremner & Brett (1997) Vietnam combat veterans Combat Medical 20⫹years Modified DES MCS 50 .52
Bremner et al. (1992) Vietnam combat veterans Combat Psychiatric 20⫹years Modified DES MCS 60 .94
Ehlers et al. (1998) Motor vehicle accident
victims recruited from ERs Accident Medical 3 months Feeling numbed or dazed
during accident PTSD Symptom Scale 888 .33
Koopman et al. (1994) Fire survivors Disaster Community 7–9 months Stanford Acute Stress
Reaction Questionnaire MCS–CV 154 .59
Lee et al. (1995) Longitudinal study of WWII
veterans Combat Community 2 years Peritraumatic symptoms
during combat Checklist based on DSM–III 107 .22
Marmar et al. (1994) Male era and theater Vietnam
veterans from NVVRS Combat Community 20 years PDEQ–RV MCS 251 .51
Michaels et al. (1999) ER injury victims Accident Medical 6 months Michigan Critical Events
Perception Scale MCS–CV 176 .24
c
O’Toole et al. (1999) Australian Vietnam veterans Combat Community 20⫹years PDEQ–RV Adapted version of SCID 636 .14
b,c
Roemer et al. (1998) College students Mixed Community Variable Single item assessing
numbing during
trauma
PTSD Checklist 74 .35
Shalev et al. (1996) ER patients (mainly accidents;
some terrorist or other
physical assault)
Mixed Medical 20⫹years PDEQ–RV MCS–CV 51 .54
Shalev, Sahar, et al.
(1998) ER patients (mainly accidents;
some terrorist or other
physical assault)
Mixed Medical 4 months PDEQ–RV CAPS 86 .22
c
Tichenor et al. (1996) Female Vietnam veterans
(NVVRS theater sample) Combat Community 20⫹years PDEQ–RV MCS 77 .18
Tucker et al. (1997) Terrorist bombing victims Interpersonal
violence Psychiatric 6 months Item assessing being on
“automatic pilot”Posttraumatic Stress Symptom
score (IES ⫹7 symptom
items)
86 .32
Ursano, Fullerton,
Epstein, Crowley,
Vance, et al. (1999)
Medical patients from motor
vehicle accidents Accident Medical 1 month PDEQ–RV SCID (DSM–III–R) 122 .27
c
Weiss et al. (1995) Two samples of emergency
services workers (disaster
exposed and variable)
Disaster Community Variable PDEQ–RV MCS–CV ⫹IES–R 367 .52
Unweighted average r⫽.43 (CI ⫽.25, .58)
Weighted average r⫽.35 (CI ⫽.16, .52)
Note. Average rcalculated from sample-weighted correlation for each effect size. PDEQ–RV ⫽Peritraumatic Dissociative Experiences Questionnaire—Rater Version (Marmar, Weiss, & Metzler,
1997); IES–R⫽Impact of Event Scale—Revised (Weiss & Marmar, 1997); DES ⫽Dissociative Experiences Scale (Bernstein & Putnam, 1986); MCS ⫽Mississippi Scale for Combat-Related
Posttraumatic Stress Disorder (Keane, Caddell, & Taylor, 1988); MCS–CV ⫽MCS—Civilian Version (Lauterback, Vrana, King, & King, 1997); ERs ⫽emergency rooms; DSM–III ⫽Diagnostic
and Statistical Manual of Mental Disorders (3rd ed.; American Psychiatric Association, 1980); NVVRS ⫽National Vietnam Veterans Readjustment Study (Kulka et al., 1990); CAPS ⫽Clinician
Administered Posttraumatic Stress Disorder Scale (Blake et al., 1995); IES ⫽Impact of Event Scale (Horowitz, Wilner, & Alvarez, 1979); SCID ⫽Structured Clinical Interview for DSM–III (Spitzer
& Williams, 1983); DSM–III–R⫽Diagnostic and Statistical Manual of Mental Disorders (3rd ed., rev.; American Psychiatric Association, 1987); CI ⫽95% confidence interval.
a
Positive correlation indicates positive association between peritraumatic dissociation and PTSD symptoms.
b
Effect size estimated from df and pvalue.
c
Effect size adjusted for dichotomization
of PTSD variable.
64 OZER, BEST, LIPSEY, AND WEISS

File Drawer Problem
To assess the robustness of our findings and address the poten-
tial problem of selective publication of studies that show statisti-
cally significant ESs for the relationship between predictors and
PTSD, we followed the recommendation of Rosenthal (1979,
1991) to estimate the number of unknown, unpublished studies
with null results that would be sufficient to reduce our overall ESs
to below the point of statistical significance (i.e., such that confi-
dence intervals would include zero). The threat to the robustness of
the meta-analysis findings can then be assessed, taking into con-
sideration the breadth of the field and the likelihood that such a
quantity of unpublished studies exists.
The results of these calculations supported the robustness of the
overall weighted ESs we have reported for five of the seven
predictors. The most robust ES was found for prior trauma, for
which 68 unknown studies with a zero ES would be needed to
generate a nonsignificant ES. Prior adjustment problems would
require 43 zero-ES studies; perceived life threat would necessi-
tate 30 zero-ES studies. Less robust but still requiring a relatively
implausible number of zero-ES studies were peritraumatic disso-
ciation, needing 17 zero-ES studies, and perceived social support,
needing 14 zero-ES studies. In contrast, family history of psycho-
pathology would require only 4 zero-ES studies in addition to the
nine studies we have identified, and peritraumatic emotional re-
sponse would require only 2 zero-ES studies, fewer than half of the
five studies we used for this predictor. It is worth noting that
interest in family history as a predictor has a much longer history
than does peritraumatic emotionality, although this latter variable
is currently receiving more attention (e.g., Brunet et al., 2001).
Interrelationship of Moderators and Unique Effects
Analyses of the effects of moderators (type of sample, type of
trauma, method of assessment, and time elapsed since the trauma)
described above indicated that more than one moderator had a
significant impact on the weighted ES for four of the six predictors
with a sufficient number of ESs to conduct these analyses. To what
extent were these moderators related or independent?
2
To explore
this question, we analyzed the data using two strategies.
Our first set of analyses systematically crossed the categories of
each moderator with those of the remaining three and tested the
contingency between the two moderators. Our chi-square results
showed that three of the six were significant: type of trauma
crossed with time elapsed,
␹
2
(9, N⫽68) ⫽57.4, p⬍.01; type of
trauma crossed with sample,
␹
2
(9, N⫽68) ⫽37.2, p⬍.01; and
time elapsed crossed with sample,
␹
2
(9, N⫽68) ⫽18.1, p⫽.05.
All three chi-squares of assessment method crossed with the other
three were not statistically significant. Thus, we concluded that
any effect of assessment method on prior psychological adjust-
ment, family history of psychopathology, and peritraumatic disso-
ciation was not likely to be explained away by looking at any of
the other three moderators. For type of trauma, time elapsed, and
type of sample, however, any effect of any one of these on prior
psychological adjustment, social support, and peritraumatic disso-
ciation could not be ruled out as being associated with either of the
other two. Thus, confounding as a general effect appeared
possible.
Our next step was to assess the overall impact of these moder-
ators on the entire sample of ESs in the meta-analysis, regardless
of predictor. We conducted four hierarchical linear multiple re-
gression analyses, with each moderator entered last in a rotating
fashion. These results did not support the presence of meaningful
confounding among the moderators. Overall, when the whole set
of 68 studies was analyzed, none of the moderator variables
accounted for a significant increment in variance at the first step of
the analysis nor at any subsequent steps (with the effects of the
other moderator variables partialed out). The overall R
2
with all
effects in the model was only .135 (shrunken R
2
⫽⫺.002), and
increments at the last step ranged from .001 for assessment method
to .090 for time elapsed. From these results, it is not clear that any
of the moderators makes a difference by themselves, and certainly
not uniquely in the presence of the others for the PTSD predictor
literature as a whole.
An argument can be made, however, that an analysis that pools
across all ESs and assumes that the moderators would have the
same effect regardless of what predictor the ES was associated
with is not a meaningful examination of the effects of the moder-
ators. Consequently, we also conducted hierarchical multiple lin-
ear regression analyses for prior psychological adjustment and
peritraumatic dissociation, the only two predictors for which (a)
the zcomparisons had yielded more than one significant moderator
and (b) there was a sufficient number of studies to conduct a
relatively stable and interpretable regression analysis given the
degrees of freedom used for the coding of the moderators. Because
of the very low statistical power associated with these analyses,
this approach was exploratory and intended to help describe the
relative independence of the predictors rather than to test for
significance per se. We included all three moderators for both
analyses.
The results for the predictor of prior adjustment revealed that the
only statistically significant finding was for method of assessment.
It explained a significant proportion of variance in prior adjust-
ment whether entered first, R
2
(1, 20) ⫽.190, p⫽.05, or second,
R
2
increment(1, 17) ⫽.207, p⫽.04, but not last. Neither the set
of variables representing time elapsed since the trauma nor the set
representing type of event yielded significant effects at any step.
The results for peritraumatic dissociation did not yield any statis-
tically significant effects, probably partly as a result of the sample
size of only 16 studies. The shrunken R
2
values at the first step
were .244 for type of sample, ⫺.069 for time elapsed, and .044 for
method of assessment.
Discussion
Replication and Extension of Salient Predictors
There are a number of seemingly obvious but nonetheless im-
portant observations that flow from the results of our meta-
analysis. First, understanding who does and does not respond to
traumatic exposure with PTSD symptoms is not purely a random
phenomenon. This general conclusion is consistent with Brewin et
al.’s (2000) initial analyses. Systematic and relatively consistent
findings emerged despite considerations that could be understood
as working against predictability. Even though nearly half of adults
2
We thank several anonymous reviewers and Scott Monroe for raising
the question of the interrelationship of the moderators and their impact on
our conclusions.
65
PREDICTORS OF PTSD: A META-ANALYSIS

report experiencing at least a single traumatic event at some point
in their lives, and despite the (a) relatively low prevalence (7.9%)
of lifetime PTSD, and (b) differing classes of events, time frames,
measurement methods, and other “noise”(cf. Meehl, 1978) that
accompany the nonexperimental psychosocial field research we
subjected to meta-analysis, replicable and relatively stable predic-
tive relationships were identified (see Table 8). This state of affairs
is encouraging and implies that there is indeed a phenomenon to be
understood and explained.
A second observation is that the sizes of the effects for the
predictors studied here are respectable and certainly not trivial.
Furthermore, in another replication and extension of Brewin et
al.’s (2000) findings, there appear to be two classes of predictors
based on the weighted ESs and their temporal proximity to the
traumatic event: (a) characteristics of the individual or his/her life
history that were more distal to the traumatic event and produced
average coefficients smaller than .20 (i.e., prior adjustment, prior
history of trauma, and family history of psychopathology in our
results; gender, age, education, socioeconomic status, IQ, and race
in Brewin et al., 2000); and (b) stronger predictive factors yielding
coefficients greater than .20 that were more proximal to the trau-
matic event (i.e., perceived life threat, perceived support, peritrau-
matic emotionality, and peritraumatic dissociation from our results
and intercurrent life stress from Brewin et al., 2000). These latter
ESs are strong enough to warrant a serious investigation of the
mechanism or mechanisms that gave rise to these results. It is
worth noting that, in most cases, for comparable predictors, the
weighted ESs we report are similar to those reported by Brewin et
al. For example, family history of psychopathology produced a
coefficient of .13 in Brewin et al.’s data compared with our
coefficient of .17, and perceived life threat had coefficients of .23
and .26, respectively. The single finding where our results diverge
from those of Brewin et al. is for the predictor of social support:
Brewin et al. reported a weighted ES of |.40| (their strongest
predictor) compared with our coefficient of |.28|, a statistically
significant difference (z⫽5.89, p⬍.01). We discuss this dis-
crepancy in more detail below.
The results also indicate, nonetheless, that a large proportion of
the variability in response is not explained by any of the predictors
that have been examined in both meta-analyses, despite the fact
that the more proximal set of predictors had been suggested in the
prior literature as potentially able to explain substantial variability
in PTSD symptoms or diagnosis. The fact that the prediction of the
variability in response to traumatic stressors is at best less than
20% of the total variance is consistent with the possibility that
factors unique to the combination of the person exposed and the
nature of the exposure are the determining factors in understanding
who becomes symptomatic and who does not. We return to this
point below as well.
Social Support as a Predictor
The main substantive difference between Brewin et al.’s (2000)
meta-analysis and ours, aside from our inclusion of peritraumatic
emotionality and peritraumatic dissociation, is that social support
was their strongest predictor (weighted r⫽.40), stronger both than
our estimate of its strength (weighted r⫽|.28|) and peritraumatic
dissociation (weighted r⫽.35), our strongest predictor. The set of
studies that constituted each analysis were obviously somewhat
different; in particular, Brewin et al. included several studies that
we did not because they did not meet our inclusion criteria (e.g.,
not having published a single-df ES). Even though this difference
in raw data produced the different results, both analyses suggest
that social support is a predictor needing further exploration.
Our analyses indicated that the strength of the relationship
between social support and PTSD symptoms and diagnosis dif-
fered according to the length of time that had elapsed since the
trauma. Social support served as a stronger predictor in studies
where the event had occurred more than 3 years prior than it was
for studies with less time elapsed. This result tends to lend some
credence to the idea that social support may function as a kind of
secondary prevention that is seen more clearly when symptoms are
most clearly symptoms of PTSD rather than common, shorter-term
reactions following traumatic exposure. Alternatively, these find-
ings may also suggest that the effects of social support are cumu-
lative over time and thus may be seen most strongly in studies that
assessed PTSD symptoms after several years had elapsed since the
index trauma.
The methodological problems associated with the retrospective
nature of the measurement of social support in most of the current
PTSD literature can be easily solved in prospective longitudinal
studies. More importantly, when we examined the measures used
to generate the 11 ESs for social support, we found that in most
studies, the emphasis was on emotional support. This suggests that
the kinds of phenomena for which individuals are receiving sup-
port likely have more to do with the psychological processing of
Table 8
Meta-Analyses of Predictors of Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms
Predictor kN r
u
rCI
(unweighted) CI
(weighted)
Prior trauma 23 5,308 .21 .17 .16, .27 .11, .22
Prior adjustment 23 6,797 .21 .17 .13, .26 .10, .23
Family history of psychopathology 9 667 .21 .17 .08, .33 .04, .29
Perceived life threat 12 3,524 .29 .26 .21, .36 .18, .34
Perceived support 11 3,537 ⫺.29 ⫺.28 ⫺.41, ⫺.16 ⫺.40, ⫺.15
Peritraumatic emotions 5 1,755 .28 .26 .10, .42 .08, .42
Peritraumatic dissociation 16 3,534 .43 .35 .25, .58 .16, .52
Note. k ⫽number of effect sizes; N⫽total number of participants in the ksamples; r
u
⫽unweighted effect
size estimate; r⫽effect size estimate adjusted for sample size (positive values reflect that the correlate is
positively associated with increased PTSD symptoms); CI ⫽95% confidence interval.
66 OZER, BEST, LIPSEY, AND WEISS

the meaning of the event or the management of the psychological
distress and pain experienced during intrusive memories or night-
mares, rather than with needs such as financial assistance, mobi-
lization of the criminal justice system, or restoration of lost pos-
sessions. Further investigation into whether this conjecture can be
empirically supported would be of interest, and could also produce
knowledge that could inform intervention to help ameliorate the
impact of exposure to traumatic stressors.
Impact of Moderators on Prediction
The impact of our moderators of type of event, time elapsed
since the event, type of sample, and method of assessment on
prediction was not uniform across predictors. In this respect, our
results also replicate Brewin et al.’s (2000), whose moderator
results were also mixed and could not be simply summarized.
Nonetheless, when approached from the perspective of compari-
sons of average weighted ES within predictor, type of event was
the most salient moderator of prediction, having an impact on five
of the six predictors for which its effect could be analyzed. Time
elapsed since the event was a significant moderator for four of the
six predictors, and method of assessment was a significant mod-
erator of three of the six.
For two predictors, interviews produced stronger average ESs
than self-report surveys, but it is not clear that this finding is an
important lead. The effect of time elapsed since the trauma also
failed to produce a clear result across predictors. For two predic-
tors (life threat and peritraumatic dissociation), the average ES was
greater for studies in which 6 months to 3 years had elapsed since
the trauma. For two other predictors, however, the pattern was
different and opposite of each other—prior adjustment was more
predictive when the shortest period of time had elapsed, and social
support was most predictive when the longest period of time (3
years or more) had elapsed.
The manner in which the type of event affected prediction, by
contrast, was nearly uniform—average ESs were stronger for prior
trauma, prior adjustment, family history of psychopathology, and
life threat if the index trauma was noncombat interpersonal vio-
lence (i.e., interpersonal violence that did not occur in a combat or
military context). The one exception to that pattern was the finding
that lower social support was in fact a stronger predictor of PTSD
when the index trauma was combat than when it was other types
of trauma. Although it is possible to construct a number of expla-
nations to account for these results, all would be ad hoc and not
supported by ancillary data. The relative uniformity of the finding
does suggest that further research into the impact of the type of
event could be fruitful.
A less optimistic perspective on the influence of moderators
might conclude that moderation of prediction need not be included
as a major phenomenon in investigating the prediction of PTSD
symptoms or diagnostic status on the basis of our results from (a)
chi-square analyses conducted across all studies indicating that the
moderators of type of event, type of sample, and time elapsed were
not independent; (b) overall multiple regression equations con-
ducted across all studies that failed to demonstrate a significant
effect for any moderator; and (c) predictor-specific regression
analyses that found a statistically significant effect only for method
of assessment. Because the appropriate literature currently avail-
able for this meta-analysis yielded a relatively small set of ESs for
predictor-specific regressions (i.e., 23 at most), definitive, multi-
variate analyses of the role of moderators for specific predictors of
PTSD will only be possible in the future with a larger body of
research from which to draw.
Strengths and Weaknesses of the Current Literature
There are a number of issues in the PTSD literature that warrant
discussion and that lend some understanding to the variety of
results within predictor categories. First, it is important to note the
heavy reliance on self-report measures and retrospective designs
that characterizes the PTSD literature on which this and Brewin et
al.’s (2000) meta-analysis are based. This naturalistic, retrospec-
tive emphasis is certainly understandable given the unpredictabil-
ity of traumatic exposure in many of the civilian populations
studied here (e.g., the unexpected nature of most natural and
human disasters), and the obvious ethical prohibitions of inflicting
exposure to highly stressful events on populations in an experi-
mental or quasiexperimental design. Longitudinal research with
special, high-risk populations, such as military personnel, emer-
gency service workers, and police officers, however, can provide
opportunities for prospective studies of the relationship of factors
studied here with expression of PTSD symptoms in the context of
traumatic exposure. In fact, a small number of studies in the
literature reviewed here were able to serendipitously provide such
prospective data by linking existing data (i.e., routine, premilitary
service psychological assessments) to later functioning. As well,
creative designs in which real-time psychophysiological data are
collected remotely from such samples during their on-duty service
may also be an important improvement on the bulk of the extant
set of studies, especially with respect to neurobiological measures.
The studies that constitute this meta-analysis used as dependent
variables either (a) PTSD diagnosis or (b) severity of PTSD
symptoms. For those studies that diagnosed PTSD, the criterion of
significant impairment in the affected individuals was necessarily
met; that cannot be said uniformly for those studies that assessed
PTSD symptoms as a continuous measure, although high mean
symptom levels in many studies certainly suggest evidence of
impairment. Both our and Brewin et al.’s (2000) findings further
indicate that the use of PTSD diagnosis versus the “continuous”
level of PTSD symptoms as outcomes does not generally have a
significant impact on the ESs in this field. Thus, although it is
possible, we believe that heterogeneity in the degree of impairment
among research participants is not a cause for serious concern in
this literature.
More specific (and fixable) shortcomings of the research liter-
ature on predictors of PTSD symptoms are these: (a) virtually total
lack of information on strategies used to address the problem of
missing data (fewer than five studies explicitly mentioned this
issue and how it was treated) and (b) insufficient information
regarding single-df ESs and degrees of freedom for findings in
multiple studies (e.g., reporting only that a finding was not statis-
tically significant). Both of these factors, as well as others such as
the routine presentation of summary data for all variables tested
rather than just those that were statistically significant, are reme-
diable editorially on a policy basis. Our review suggests such
policy changes would help advance knowledge by providing com-
plete information. The article published by the American Psycho-
logical Association’s Task Force on Statistical Inference (Wilkin-
67
PREDICTORS OF PTSD: A META-ANALYSIS

son & Task Force on Statistical Inference, 1999) has many
important suggestions in this regard.
Nonetheless, findings from several studies not included in the
meta-analysis because of the lack of published information neces-
sary to calculate single-df ESs generally support the results of the
meta-analysis and further indicate that the average ESs calculated
for prior adjustment and family history of psychopathology repre-
sent relatively conservative estimates. Prior affective and anxiety
disorders were strongly associated with higher risk of PTSD in the
NCS (i.e., affective disorders associated with at least a threefold
increased risk of PTSD; Bromet, Sonnega, & Kessler, 1998) and
among Breslau et al.’s (1991) community sample. In addition,
family history of psychopathology was associated with a higher
risk for PTSD in the same two studies.
The Fiction of Homogeneity and Fungibility
Our meta-analysis, Brewin et al.’s (2000), and almost every
other study in our meta-analysis dataset make the assumption that
different events are interchangeable with each other for the pur-
pose of categorizing and conducting analyses to determine rela-
tionships. What is often overlooked, however, is that there may be
substantial heterogeneity in each construct that is typically viewed
more homogeneously. For example, our predictor of prior trauma
makes the assumption that all the prior traumas are equal in their
effects (which they likely are not given that type of event appeared
to be one of the most robust moderators), that a single exposure is
no different from multiple exposures, and that the variety of other
characteristics by which they differ will compensate for each other
or in some other way not have a systematic impact. The hetero-
geneity of the index trauma for many of the studies in our dataset
is well described, but then conceptually overlooked with respect to
the presence of PTSD symptoms or disorder. We know that events
differ with respect to potential lethality, time elapsed since the
event, repetition of the event, developmental phase of the individ-
ual being subjected to the event, as well as the nature of the event
(e.g., natural vs. human, accidental vs. purposeful, etc.). This
heterogeneity may be as responsible as any other factor for the
limits on predictability.
What makes this issue so vexing is that the main homogeneity
in the field is in the consequences of exposure to the heterogeneity
of traumatic events—the very signs and symptoms of PTSD. From
this perspective, more attention needs to be paid to the sequelae of
exposure than to either the preexisting conditions prior to the
exposure or the aspects of the exposure itself. One way in which
such attention might manifest itself is in the recognition that
because the course of PTSD is typically chronic (Bromet et al.,
1998) and is a waxing and waning disorder more like diabetes than
like myocardial infarction, some of the heterogeneity in terms of
symptom status versus diagnostic status is that individuals who
have met criteria for PTSD prior to being studied are in partial
remission when they are interviewed. This would systematically
bias the results of studies using diagnosis in terms of underesti-
mating the impact of exposure, whereas studies using continuous
symptom measures would be less subject to these issues. Similarly,
there are individuals (see Weiss et al., 1992) who can best be
described as having lifetime partial PTSD for whom symptom
measures would make them virtually indistinguishable from those
who meet current criteria, but whose diagnostic status would be
“no history of PTSD.”Though not unique to PTSD, this issue is
especially salient given the requirement that the symptoms be
linked to a traumatic event, an event which can have occurred
years or decades previously.
Implications of the Findings for Models of the
Development of PTSD
There is a common thread between exposure to prior traumatic
stressors, family history of psychopathology, and the exposed
person’s own psychological difficulties as predictors of PTSD
symptoms. This thread is that psychological difficulties, perhaps
also manifested in or as a consequence of poorer social support,
play some role in conferring risk of developing PTSD symptoms
after exposure to a traumatic stressor. It is tempting to make an
analogy to the flu or infectious disease: Those whose immune
systems are compromised are at greater risk of contracting a
subsequent illness. Similarly, this cluster of variables may all be
pointing to a single source of vulnerability for the development of
PTSD or enduring symptoms of PTSD—a lack of psychological
resilience. The notion of resilience has a long history in measure-
ment (e.g., Block, 1965), and several researchers invoke it (Funk
& Houston, 1987; Garmezy & Rutter, 1983; Haggerty, Sherrod,
Garmezy, & Rutter, 1996; Kobasa, Maddi, & Kahn, 1982; Kobasa,
Maddi, Puccetti, & Zola, 1985; Maddi & Kobasa, 1991), though
none of those researchers has closely examined resilience after
exposure to truly traumatic events. It is possible that our results,
taken in conjunction with these other large sets of findings, are yet
another view of the same core vulnerability, but seen from differ-
ent perspectives and manifested in different ways. It would be
reasonable to suspect, however, that in circumstances where the
traumatic event is particularly horrific (i.e., on the extreme end
even for traumatic events), resilience may play less of a role. At
this time, there is little data with which to evaluate that
proposition.
How this resiliency, or lack of it, is related to both the psycho-
logical issues that working through exposure to a traumatic stres-
sor require and to peritraumatic dissociation is of particular inter-
est. Our finding that peritraumatic dissociation and, to a lesser
degree, peritraumatic emotionality are as salient predictors as any
yet identified raises the possibility that Brewin et al.’s (2000)
conclusion that a general vulnerability model for PTSD was not
warranted might require some amendment. With respect to the
former issue, the psychological issues involved in treating PTSD
(see, e.g., Goodman & Weiss, 1998; Resick, 2001) revolve, in
large part, around issues of loss. The impact of prior loss on later
risk for depression, for example, has been well documented (Ten-
nant, 1988; Tennant, Smith, Bebbington, & Hurry, 1981). It is thus
not surprising that prior trauma was found to be related to symp-
toms or diagnosis of PTSD as the overlap in symptoms between
depression and PTSD is well known. Roughly the same argument
can be made for psychological adjustment and family history of
psychopathology. A kind of convergence appears to characterize
the finding that depression was the aspect of prior adjustment that
was most strongly related to PTSD and the observation that one of
the key psychological aspects of dealing with the sequelae of
exposure to traumatic stress and PTSD is coming to terms with
loss. The degree to which depression is frequently comorbid with
PTSD may be related to one, several, or all of these factors. An
68 OZER, BEST, LIPSEY, AND WEISS

important question for future work is the relationship between the
presence of peritraumatic dissociation and the salience of loss, as
opposed to fear or horror in dealing with exposure to traumatic
stress.
It is worth noting that resiliency may also be manifest in the
individual differences in the functioning of the HPA axis, specif-
ically the manner in which high levels of physiological and emo-
tional arousal are managed in the brain structures that have been
revealed to become increasingly important in animal models (e.g.,
LeDoux, 2000) and imaging findings (e.g., Rauch et al., 2000;
Schuff et al., 2001)—the amygdala and hippocampus have now
been implicated in PTSD. It remains to be clarified whether these
effects in the brain are consequences of exposure to traumatic
stress, consequences of intrusive and hyperarousal symptoms of
PTSD, or in some way reflect preexisting vulnerabilities that
increase the likelihood of the development of PTSD when an
individual experiences severe stress.
One argument about the predictive power of the static historical
variables that we and Brewin et al. (2000) illuminated is that they
are merely a consequence of the shared variability between general
psychiatric symptoms and PTSD-specific symptoms. From the
analyses we conducted, we cannot make an empirical statement
about what would have occurred if each bivariate relationship
between predictor and PTSD symptoms had been residualized
against general psychiatric symptoms. In most of the studies, such
data were not gathered. The findings regarding peritraumatic dis-
sociation, however, do shed some light on this thesis. A number of
researchers (e.g., Shalev, Peri, Canetti, & Schreiber, 1996; Weiss,
Marmar, Metzler, & Ronfeldt, 1995) did examine the effect of the
predictors after other direct or indirect measures of psychological
adjustment were included in the multivariate model. The results
showed that after accounting for general adjustment and general
dissociative experiences, peritraumatic dissociative experiences
remained a significant predictor of PTSD symptoms. Future stud-
ies would benefit from an examination of the shared variance
between general symptoms typically present in PTSD (e.g., de-
pression, anxiety), specific symptoms of PTSD (e.g., intrusive
images, emotional numbing, hypervigilance), and the predictors
included in this meta-analysis (or other sets of predictors). Such
analyses could go a long way toward clarifying more specific
aspects of PTSD, including the conundrum of the comorbidity
between PTSD and depression.
Given these results for peritraumatic dissociation, it is difficult
not to entertain the notion that the psychological aspects of expo-
sure may be the most important, and that the in-the-moment
appraisal and meaning of the traumatic stressor may have as much
to do with explaining who develops PTSD as do the more static
factors such as adjustment, prior exposure, or concurrent psycho-
pathology. As well, the salience of peritraumatic dissociation may
also lend some empirical support to the conjecture that ASD may
indeed be an important precursor to PTSD in some people, or after
some very horrifying event, or both. Nonetheless, the strength of
the relationship of r⫽.35 clearly does not support thinking about
dissociative phenomena as an integral part of the phenomenology
of posttraumatic response. Many individuals who have high levels
of symptoms of PTSD and meet current diagnostic criteria do not
experience dissociative phenomena; thus, it would be a stretch to
argue that peritraumatic dissociation is by definition a part of
PTSD because some individuals do experience dissociative phe-
nomena. Such a position would be similar to contending that the
absence of psychophysiological arousal in response to trauma cues
(one method of assessing a part of the criteria for the diagnosis;
Orr & Roth, 2000) in a subset of those who meet criteria (Keane
et al., 1998) means that hyperarousal criteria (Criterion D) should
be eliminated from the diagnostic criteria for PTSD. Until such
time as the definitory (Meehl, 1995) criteria for PTSD are identi-
fied and understood, the approach of the DSM to use evidentiary
criteria as definitory seems to be the only viable option.
Although it is relatively easy to quantify peritraumatic dissocia-
tive experiences retrospectively, whether months or years after the
event (Weiss et al., 1995) or hours or days after the event (Shalev
et al., 1996), it is much more difficult to quantify peritraumatic
distress as it actually is occurring in the context of a traumatic
event. Leaving aside the logistic issues of being present to observe
an individual’s in-the-moment response to an ongoing traumatic
event (though accompanying emergency personnel on their runs
could be one approach), the case could be made that the subjective
experience of peritraumatic dissociation may not be accessible in
the moment and may only be understood or recognized after the
immediate danger or threat is passed. Although brain imaging is
one possible method, it would require the experimental induction
of high arousal, something that may not be possible. There are,
however, a few studies using ketamine (e.g., Chambers et al.,
1999) in volunteers that suggest this may be a possible avenue.
Ketamine is a synthesized anesthetic used primarily in veterinary
medicine and belongs to the class of dissociative anesthetics like
PCP and nitrous oxide. Its action is to block nerve paths without
depressing respiratory or circulatory functioning. At lower doses in
humans it often produces dissociative experiences, and its illicit
use has documented “trips”that may include out of body experi-
ences when used at higher doses.
Because one view of peritraumatic dissociation suggests that it
occurs when the traumatic event is so severe as to feel intolerable
(Spiegel & Cardena, 1991), it would be important to take into
account the severity of the traumatic event when evaluating the
impact of peritraumatic dissociation on PTSD symptoms. Concep-
tually, there is no true objective assessment of severity that is
totally divorced from response, because a rough assessment of the
modal response to any particular event is typically understood to
be a rough index of its severity (see Footnote 1). Nonetheless,
given a stance of “all other things being equal”with regard to the
type of event, it appears that the peritraumatic response to the
event is the most robust factor contributing to the prediction of
PTSD symptoms.
That peritraumatic dissociation was the strongest predictor of
those we examined is probably the most important finding from a
conceptual perspective as well. If we assume that the results of the
meta-analysis are not confounded by some unknown artifact in the
same way that we presume that each individual study is not
confounded by some unknown artifact, the strong implication is
that if one could bet on only one variable, the subjective psycho-
logical response to traumatic exposure is the variable on which to
place bets. It remains an open question as to how much of reported
peritraumatic dissociation is related to the level of psychophysio-
logical arousal the individual endured during the event, and how
contemporaneous versus retrospective measurement differentially
impacts the reported degree of peritraumatic dissociation. These
are important areas for further inquiry.
69
PREDICTORS OF PTSD: A META-ANALYSIS

Implications and Future Directions
Although this meta-analysis and the results of Brewin et al.
(2000) provide evidence for statistically significant predictive re-
lationships between multiple characteristics of individuals and
their experiences and PTSD symptoms, the specific processes by
which these factors may serve to influence the development of
PTSD remain largely unexamined. Further specification of the
intervening as well as directly explanatory variables would point to
areas of opportunity for intervention and possible attenuation or
prevention of the development of PTSD. Theory-based develop-
ment and evaluation of such interventions with good operational-
ization of critical variables could then provide data with which to
evaluate existing theoretical approaches, an important area of
inquiry given the ethical limitations of experimental research in
this area.
The thrust of future research needs to be focused on the more
proximal mechanisms or processes—in psychological and/or phys-
iological terms—that explain the statistical relationship between
PTSD symptoms and the more distal, static predictors such as
family history of psychopathology and prior trauma. Our exami-
nation of the PTSD literature sought to illuminate noncontingent,
simple, linear relationships. As described above, with respect to
peritraumatic dissociation, emotional and psychophysiological
arousal, and the underlying brain activities in the HPA axis and the
limbic system, it may well be that some variables operate in a
contingent fashion—heightened arousal is a necessary but not
sufficient condition for the experience of peritraumatic dissocia-
tion, so that not all those who experience heightened arousal go on
to experience peritraumatic dissociation. The mechanism by which
peritraumatic dissociation occurs may be influenced not only by
arousal, but also by temperament, prior experience, and other
genetic or environmental factors, including factors at the level of
plasticity and learning in individual neurons (Malenka, 2000). Our
examination of the literature has been able to offer only tentative
glimpses into explanatory mechanisms.
References
References marked with an asterisk indicate studies included in the
meta-analysis.
American Psychiatric Association. (1968). Diagnostic and statistical man-
ual of mental disorders (2nd ed.). Washington, DC: Author.
American Psychiatric Association. (1980). Diagnostic and statistical man-
ual of mental disorders (3rd ed.). Washington, DC: Author.
American Psychiatric Association. (1987). Diagnostic and statistical man-
ual of mental disorders (3rd ed., rev.). Washington, DC: Author.
American Psychiatric Association. (1994). Diagnostic and statistical man-
ual of mental disorders (4th ed.). Washington, DC: Author.
*Andrews, B., Brewin, C. R., Rose, S., & Kirk, M. (2000). Predicting
PTSD symptoms in victims of violent crime: The role of shame, anger,
and childhood abuse. Journal of Abnormal Psychology, 109, 69–73.
*Astin, M. C., Lawrence, K. J., & Foy, D. W. (1993). Posttraumatic stress
disorder among battered women: Risk and resiliency factors. Violence
and Victims, 8, 17–28.
*Astin, M. C., Ogland-Hand, S. M., Coleman, E. M., & Foy, D. S. (1995).
Posttraumatic stress disorder and childhood abuse in battered women:
Comparisons with maritally distressed women. Journal of Consulting
and Clinical Psychology, 63, 308–312.
*Basoglu, M., Paker, M., Ozmen, E., Tasdemir, O., & Sahin, D. (1994).
Factors related to long-term traumatic stress responses in survivors of
torture in Turkey. Journal of the American Medical Association, 272,
357–363.
Becker, J. V. (1982). The effects of sexual assault on rape and attempted
rape victims. Victimology, 7, 106–113.
*Bernat, J. A., Ronfeldt, H. M., Calhoun, K. S., & Arias, I. (1998).
Prevalence of traumatic events and peritraumatic predictors of posttrau-
matic stress symptoms in a nonclinical sample of college students.
Journal of Traumatic Stress, 11, 645–665.
Bernstein, E. M., & Putnam, F. W. (1986). Development, reliability, and
validity of a dissociation scale. Journal of Nervous and Mental Diseases,
174, 727–735.
Blake, D. D., Weathers, F. W., Nagy, L. M., Kaloupek, D. G., Gusman,
F. D., Charney, D. S., & Keane, T. M. (1995). The development of a
clinician-administered PTSD scale. Journal of Traumatic Stress, 8,
75–90.
*Blanchard, E. B., Hickling, E. J., Taylor, A. E., & Loos, W. (1995).
Psychiatric morbidity associated with motor vehicle accidents. Journal
of Nervous and Mental Disease, 183, 495–504.
Block, J. (1965). The challenge of response sets: Unconfounding meaning,
acquiescence, and social desirability in the MMPI. New York: Appleton
Century Crofts.
*Boscarino, J. A. (1995). Post-traumatic stress and associated disorders
among Vietnam veterans: The significance of combat exposure and
social support. Journal of Traumatic Stress, 8, 317–336.
Bower, G. H. (1981). Mood and memory. American Psychologist, 36,
129–148.
*Bremner, J. D., & Brett, E. (1997). Trauma-related dissociative states and
long-term psychopathology in posttraumatic stress disorder. Journal of
Traumatic Stress, 10, 37–49.
*Bremner, J. D., Southwick, S., Brett, E., Fontana, A., Rosenheck, R., &
Charney, D. S. (1992). Dissociation and posttraumatic stress disorder in
Vietnam combat veterans. American Journal of Psychiatry, 149, 328–
332.
*Bremner, J. D., Southwick, S. M., Johnson, D. R., Yehuda, R., &
Charney, D. S. (1993). Childhood physical abuse and combat-related
posttraumatic stress disorder in Vietnam veterans. American Journal of
Psychiatry, 150, 235–239.
Breslau, N., Davis, G. C., Andreski, P., & Peterson, E. (1991). Traumatic
events and posttraumatic stress disorder in an urban population of young
adults. Archives of General Psychiatry, 48, 216–222.
*Breslau, N., Kessler, R. C., Chilcoat, H. D., Schultz, L. R., Davis, G. C.,
& Andreski, P. (1998). Trauma and posttraumatic stress disorder in the
community: The Detroit area survey of trauma. Archives of General
Psychiatry, 55, 626–632.
Brewin, C. R., Andrews, B., & Valentine, J. D. (2000). Meta-analysis of
risk factors for posttraumatic stress disorder in trauma-exposed adults.
Journal of Consulting and Clinical Psychology, 68, 748–766.
Bromet, E., Sonnega, A., & Kessler, R. C. (1998). Risk factors for DSM–
III–Rpost-traumatic stress disorder: Findings from the National Comor-
bidity Survey. American Journal of Epidemiology, 147, 353–361.
Brunet, A., Weiss, D. S., Metzler, T. J., Best, S. R., Neylan, T. C., Rogers,
C., et al. (2001). The Peritraumatic Distress Inventory: A proposed
measure of PTSD Criterion A2. American Journal of Psychiatry, 158,
1480–1485.
Burgess, A. W., & Holmstrom, L. L. (1974). Rape trauma syndrome.
American Journal of Psychiatry, 131, 981–986.
*Carlier, I. V. E., Lamberts, R. D., & Gersons, B. P. R. (1997). Risk factors
for posttraumatic stress symptomatology in police officers: A prospec-
tive analysis. Journal of Nervous and Mental Disease, 185, 498–506.
Carlson, E. B., & Rosser-Hogan, R. (1993). Mental health status of
Cambodian refugees ten years after leaving their homes. American
Journal of Orthopsychiatry, 63, 223-231.
Carlson, E. B., & Rosser-Hogan, R. (1994). Cross-cultural response to
70 OZER, BEST, LIPSEY, AND WEISS

trauma: A study of traumatic experiences and posttraumatic symptoms
in Cambodian refugees. Journal of Traumatic Stress, 7, 43–58.
Cervantes, R. C., Salgado de Snyder, V. N., & Padilla, A. M. (1989).
Posttraumatic stress in immigrants from Central America and Mexico.
Hospital & Community Psychiatry, 40, 615–619.
Chambers, R. A., Bremner, J. D., Moghaddam, B., Southwick, S. M.,
Charney, D. S., & Krystal, J. H. (1999). Glutamate and post-traumatic
stress disorder: Toward a psychobiology of dissociation. Seminars in
Clinical Neuropsychiatry, 4, 274–281.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences
(2nd ed.). Hillsdale, NJ: Erlbaum.
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation
analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.
Cooper, H., & Hedges, L. V. (Eds.). (1994). The handbook of research
synthesis. New York: Russell Sage Foundation.
*Cottler, L. B., Compton, W. M., III, Mager, D., Spitznagel, E. L., & Janca,
A. (1992). Posttraumatic stress disorder among substance users from the
general population. American Journal of Psychiatry, 149, 664–670.
*David, D., Mellman, T., Mendoza, L., Kulick-Bell, R., Ironson, G., &
Schneiderman, N. (1996). Psychiatric morbidity following Hurricane
Andrew. Journal of Traumatic Stress, 9, 607–612.
Davidson, J. R. T., Kudler, H. S., & Smith, R. (1990). Assessment and
pharmacotherapy of posttraumatic stress disorder. In J. E. L. Giller (Ed.),
Biological assessment and treatment of posttraumatic stress disorder
(pp. 205–221). Washington, DC: American Psychiatric Press.
Davis, M. (1992). The role of the amygdala in fear and anxiety. Annual
Review of Neuroscience, 15, 353–375.
Derogatis, L. R. (1992). SCL–90R administration, scoring & procedures
manual—II. Towson, MD: Clinical Psychometric Research.
*Dunmore, E., Clark, D. M., & Ehlers, A. (1999). Cognitive factors
involved in the onset and maintenance of posttraumatic stress disorder
(PTSD) after physical or sexual assault. Behaviour Research and Ther-
apy, 37, 809–829.
*Ehde, D. M., Patterson, D. R., Wiechman, S. A., & Wilson, L. G. (2000).
Post-traumatic stress symptoms and distress 1 year after burn injury.
Journal of Burn Care and Rehabilitation, 21, 105–111.
*Ehlers, A., Mayou, R. A., & Bryant, B. (1998). Psychological predictors
of chronic posttraumatic stress disorder after motor vehicle accidents.
Journal of Abnormal Psychology, 107, 508–519.
*Emery, V. O., Emery, P. E., Shama, D. K., Quiana, N. A., & Jassani,
A. K. (1991). Predisposing variables in PTSD patients. Journal of
Traumatic Stress, 4, 325–343.
*Engel, C. C., Jr., Engel, A. L., Campbell, S. J., McFall, M. E., Russo, J.,
& Katon, W. (1993). Posttraumatic stress disorder symptoms and pre-
combat sexual and physical abuse in Desert Storm veterans. Journal of
Nervous and Mental Disease, 181, 683–688.
*Epstein, R. S., Fullerton, C. S., & Ursano, R. J. (1998). Posttraumatic
stress disorder following an air disaster: A prospective study. American
Journal of Psychiatry, 155, 934–938.
*Fauerbach, J. A., Lawrence, J., Haythornthwaite, J., Richter, D., McGuire,
J., Schmidt, C., & Munster, A. (1997). Preburn psychiatric history
affects posttrauma morbidity. Psychosomatics, 38, 374–385.
Foa, E. B., Riggs, D. S., Dancu, C. V., & Rothbaum, B. O. (1993).
Reliability and validity of a brief instrument for assessing post-traumatic
stress disorder. Journal of Traumatic Stress, 6, 459–473.
Fontana, A., & Rosenheck, R. (1994). Posttraumatic stress disorder among
Vietnam theater veterans: A causal model of etiology in a community
sample. Journal of Nervous and Mental Disease, 182, 677–684.
*Fontana, A., Schwartz, L. S., & Rosenheck, R. (1997). Posttraumatic
stress disorder among female Vietnam veterans: A causal model of
etiology. American Journal of Public Health, 87, 169–175.
*Foy, D. W., Sipprelle, R. C., Rueger, D. B., & Carroll, E. M. (1984).
Etiology of posttraumatic stress disorder in Vietnam veterans: Analysis
of premilitary, military, and combat exposure influences. Journal of
Consulting and Clinical Psychology, 52, 79–87.
Funk, S. C., & Houston, B. K. (1987). A critical analysis of the Hardiness
Scale’s validity and utility. Journal of Personality and Social Psychol-
ogy, 53, 572–578.
Garmezy, N., & Rutter, M. (Eds.). (1983). Stress, coping, and development
in children. Baltimore: Johns Hopkins University Press.
Green, B. L. (1994). Psychosocial research in traumatic stress: An update.
Journal of Traumatic Stress, 7, 341–362.
*Green, M. A., & Berlin, M. A. (1987). Five psychosocial variables related
to the existence of post-traumatic stress disorder symptoms. Journal of
Clinical Psychology, 43, 643–649.
Goodman, M., & Weiss, D. S. (1998). Double trauma: A group therapy
approach for Vietnam veterans suffering from war and childhood
trauma. International Journal of Group Psychotherapy, 48, 39–54.
Haggerty, R. J., Sherrod, L. R., Garmezy, N., & Rutter, M. (Eds.). (1996).
Stress, risk, and resilience in children and adolescents: Processes,
mechanisms, and interventions. New York: Cambridge University Press.
Haley, S. A. (1974). When the patient reports atrocities: Specific treatment
considerations of the Vietnam veteran. Archives of General Psychia-
try, 30, 191–196.
Hathaway, S. R., & McKinley, J. C. (1951). Manual for the Minnesota
Multiphasic Personality Inventory (Rev. ed.). New York: Psychological
Corporation.
Horowitz, M. J., & Solomon, G. F. (1975). A prediction of delayed stress
response syndromes in Vietnam veterans. Journal of Social Issues, 31,
67–80.
Horowitz, M. J., Wilner, N., & Alvarez, W. (1979). Impact of Event Scale:
A measure of subjective stress. Psychosomatic Medicine, 41, 209–218.
*Hough, R. L., Vega, W., Valle, R., Kolody, B., Griswald del Castillo, R.,
& Tarke, H. (1990). Mental health consequences of the San Ysidro
McDonald’s massacre: A community study. Journal of Traumatic
Stress, 3, 71–92.
Hunter, J. E., & Schmidt, F. L. (1990a). Dichotomization of continuous
variables: The implications for meta-analysis. Journal of Applied Psy-
chology, 75, 334–349.
Hunter, J. E., & Schmidt, F. L. (1990b). Methods of meta-analysis: Cor-
recting error and bias in research findings. Newbury Park, CA: Sage.
Hunter, J. E., & Schmidt, F. L. (1994). Estimation of sampling error
variance in the meta-analysis of correlations: Use of average correlation
in the homogeneous case. Journal of Applied Psychology, 79, 171–177.
*Jeavons, S., Greenwood, K. M., & Horne, D. J. (2000). Accident cogni-
tions and subsequent psychological trauma. Journal of Traumatic
Stress, 13, 359–365.
Keane, T. M., Caddell, J. M., & Taylor, K. L. (1988). Mississippi Scale for
Combat-Related Posttraumatic Stress Disorder: Three studies in reliabil-
ity and validity. Journal of Consulting and Clinical Psychology, 56,
85–90.
Keane, T. M., Kolb, L. C., Kaloupek, D. G., Orr, S. P., Blanchard, E. B.,
Thomas, R. G., et al. (1998). Utility of psychophysiological measure-
ment in the diagnosis of posttraumatic stress disorder: Results from a
Department of Veterans Affairs Cooperative Study. Journal of Consult-
ing and Clinical Psychology, 66, 914–923.
Keane, T. M., Malloy, P. S., & Fairbank, J. A. (1984). The empirical
development of an MMPI subscale for the assessment of combat-related
posttraumatic stress disorder. Journal of Consulting and Clinical Psy-
chology, 52, 888–891.
Keiser, L. (1968). The traumatic neurosis. Philadelphia: Lippincott.
*Kemp, A., Green, B. L., Hovanitz, C., & Rawlings, E. I. (1995). Incidence
and correlates of posttraumatic stress disorder among battered women:
Shelter and community samples. Journal of Interpersonal Violence, 10,
43–55.
Kessler, R. C., Sonnega, A., Bromet, E., Hughes, M., & Nelson, C. B.
71
PREDICTORS OF PTSD: A META-ANALYSIS

(1995). Posttraumatic stress disorder in the National Comorbidity Sur-
vey. Archives of General Psychiatry, 52, 1048–1060.
*Kilpatrick, D. G., Saunders, B. E., Amick-McMullan, A., Best, C. L.,
Veronen, L. J., & Resnick, H. S. (1989). Victim and crime factors
associated with the development of crime-related post-traumatic stress
disorder. Behavior Therapy, 20, 199–214.
*King, L. A., King, D. W., Fairbank, J. A., Keane, T. M., & Adams, G. A.
(1998). Resilience–recovery factors in post-traumatic stress disorder
among female and male Vietnam veterans: Hardiness, postwar social
support, and additional stressful life events. Journal of Personality and
Social Psychology, 74, 420–434.
Kobasa, S. C., Maddi, S. R., & Kahn, S. (1982). Hardiness and health: A
prospective study. Journal of Personality and Social Psychology, 42,
168–177.
Kobasa, S. C. O., Maddi, S. R., Puccetti, M. C., & Zola, M. A. (1985).
Effectiveness of hardiness, exercise and social support as resources
against illness. Journal of Psychosomatic Research, 29, 525–533.
Kolb, L. C. (1993). The psychobiology of PTSD: Perspectives and reflec-
tions on the past, present, and future. Journal of Traumatic Stress, 6,
293–304.
*Koopman, C., Classen, C., & Spiegel, D. (1994). Predictors of posttrau-
matic stress symptoms among survivors of the Oakland/Berkeley, Cal-
ifornia, firestorm. American Journal of Psychiatry, 151, 888–894.
*Kramer, T. L., & Green, B. L. (1991). Posttraumatic stress disorder as an
early response to sexual assault. Journal of Interpersonal Violence, 6,
160–173.
Kulka, R. A., Schlenger, W. E., Fairbank, J. A., Hough, R. L., Jordan,
B. K., Marmar, C. R., & Weiss, D. S. (1990). Trauma and the Vietnam
War generation: Report of the findings from the National Vietnam
Veterans Readjustment Study. New York: Brunner/Mazel.
Lauterback, D., Vrana, S., King, D. W., & King, L. A. (1997). Psycho-
metric properties of the civilian version of the Mississippi PTSD scale.
Journal of Traumatic Stress, 10, 499–513.
LeDoux, J. E. (2000). Emotion circuits in the brain. Annual Review of
Neuroscience, 23, 155–184.
*Lee, K. A., Vaillant, G. E., Torrey, W. C., & Elder, G. H. (1995). A
50-year prospective study of the psychological sequelae of World War II
combat. American Journal of Psychiatry, 152, 516–522.
Lipsey, M. W., & Wilson, D. B. (2000). Practical meta-analysis. Thousand
Oaks, CA: Sage.
*Madakasira, S., & O’Brien, K. F. (1987). Acute posttraumatic stress
disorder in victims of a natural disaster. Journal of Nervous and Mental
Disease, 175, 286–290.
Maddi, S. R., & Kobasa, S. C. (1991). The development of hardiness. In A.
Monat & R. S. Lazarus (Eds.), Stress and coping: An anthology (3rd ed.,
pp. 245–257). New York: Columbia University Press.
Malenka, R. (2000). Basic science in psychiatry. Neurobiology of Dis-
ease, 7, 537–539.
Marmar, C. R., Weiss, D. S., & Metzler, T. J. (1997). The Peritraumatic
Dissociative Experiences Questionnaire. In J. P. Wilson & T. M. Keane
(Eds.), Assessing psychological trauma and PTSD: A handbook for
practitioners (pp. 412–428). New York: Guilford Press.
*Marmar, C. R., Weiss, D. S., Schlenger, W. E., Fairbank, J. A., Jordan,
B. K., Kulka, R. A., & Hough, R. L. (1994). Peritraumatic dissociation
and posttraumatic stress in male Vietnam theater veterans. American
Journal of Psychiatry, 151, 902–907.
McEwen, B. S. (1995). Adrenal steroid actions on brain: Dissecting the
fine line between protection and damage. In M. J. Friedman, D. S.
Charney, & A. Y. Deutch (Eds.), Neurobiological and clinical conse-
quences of stress: from normal adaptation to post-traumatic stress
disorder (pp. 135–147). Philadelphia: Lippincott-Raven.
*McFarlane, A. C. (1988). The aetiology of post-traumatic stress disorders
following a natural disaster. British Journal of Psychiatry, 152, 116–
121.
McGaugh, J. L., & Cahill, L. (1997). Interaction of neuromodulatory
systems in modulating memory storage. Behavioural Brain Re-
search, 83, 31–38.
Meehl, P. E. (1978). Theoretical risks and tabular asterisks: Sir Karl, Sir
Ronald, and the slow progress of soft psychology. Journal of Consulting
and Clinical Psychology, 46, 806–834.
Meehl, P. E. (1995). Bootstraps taxometrics: Solving the classification
problem in psychopathology. American Psychologist, 50, 266–275.
*Michaels, A. J., Michaels, C. E., Moon, C. H., Smith, J. S., Zimmerman,
M. A., Taheri, P. A., & Peterson, C. (1999). Posttraumatic stress disorder
after injury: Impact on general health outcome and early risk assessment.
Journal of Trauma, 47, 460–467.
*North, C. S., & Smith, E. M. (1992). Posttraumatic stress disorder among
homeless men and women. Hospital and Community Psychiatry, 43,
1010–1016.
*North, C. S., Smith, E. M., & Spitznagel, E. L. (1994). Posttraumatic
stress disorder in survivors of a mass shooting. American Journal of
Psychiatry, 151, 82–88.
Orr, S. P., & Roth, W. T. (2000). Psychophysiological assessment: Clinical
applications for PTSD. Journal of Affective Disorders, 61, 225–240.
*O’Toole, B. I., Marshall, R. P., Schureck, R. J., & Dobson, M. (1999).
Combat, dissociation, and posttraumatic stress disorder in Australian
Vietnam veterans. Journal of Traumatic Stress, 12, 625–640.
*Perrin, S., Van Hasselt, V. B., Basilio, I., & Hersen, M. (1996). Assessing
the effects of violence on women in battering relationships with the
Keanne MMPI–PTSD Scale. Journal of Traumatic Stress, 9, 805–816.
*Perry, S., Difede, J., Musngi, G., Frances, A. J., & Jacobsberg, L. (1992).
Predictors of posttraumatic stress disorder after burn injury. American
Journal of Psychiatry, 149, 931–935.
Rauch, S. L., Whalen, P. J., Shin, L. M., McInerney, S. C., Macklin, M. L.,
Lasko, N. B., et al. (2000). Exaggerated amygdala response to masked
facial stimuli in posttraumatic stress disorder: A functional MRI study.
Biological Psychiatry, 47, 769–776.
*Reich, J., Lyons, M., & Cai, B. (1996). Familial vulnerability factors to
post-traumatic stress disorder in male military veterans. Acta Psychiat-
rica Scandinavica, 93, 105–112.
Resick, P. A. (2001). Cognitive therapy for posttraumatic stress disorder.
Journal of Cognitive Psychotherapy, 15, 321–329.
*Resnick, H. S., Kilpatrick, D. G., Best, C. L., & Kramer, T. L. (1992).
Vulnerability-stress factors in development of posttraumatic stress dis-
order. Journal of Nervous and Mental Disease, 180, 424–430.
Resnick, H. S., Kilpatrick, D. G., Dansky, B. S., Saunders, B. E., & Best,
C. L. (1993). Prevalence of civilian trauma and posttraumatic stress
disorder in a representative national sample of women. Journal of
Consulting and Clinical Psychology, 61, 984–991.
*Resnick, H. S., Yehuda, R., Pitman, R. K., & Foy, D. W. (1995). Effect
of previous trauma on acute plasma cortisol level following rape. Amer-
ican Journal of Psychiatry, 152, 1675–1677.
*Riggs, D. S., Dancu, C. V., Gershuny, B. S., & Greenberg, D. (1992).
Anger and post-traumatic stress disorder in female crime victims. Jour-
nal of Traumatic Stress, 5, 613–625.
*Roberts, G., Lawrence, J. M., Williams, G. M., & Raphael, B. (1998). The
impact of domestic violence on women’s mental health. Australian and
New Zealand Journal of Public Health, 22, 796–801.
Robins, L. N., Helzer, J. E., Croughan, J. L., & Ratcliff, K. S. (1981).
National Institute of Mental Health Diagnostic Interview Schedule: Its
history, characteristics, and validity. Archives of General Psychiatry, 38,
381–389.
*Roemer, L., Orsillo, S. M., Borkovec, T. D., & Litz, B. T. (1998).
Emotional response at the time of a potentially traumatizing event and
PTSD symptomatology: A preliminary retrospective analysis of the
DSM–IV Criterion A-2. Journal of Behavior Therapy and Experimental
Psychiatry, 29, 123–130.
72 OZER, BEST, LIPSEY, AND WEISS

Rosenthal, R. (1979). The file drawer problem and tolerance for null
results. Psychological Bulletin, 86, 638–641.
Rosenthal, R. (1991). Meta-analytic procedures for social research (Rev.
ed.). Newbury Park, CA: Sage.
Rosenthal, R. (1994). Parametric measures of effect size. In H. M. Cooper
& L. V. Hedges (Eds.), The handbook of research synthesis (pp. 231–
244). New York: Russell Sage Foundation.
Rosenthal, R., & DiMatteo, M. R. (2000). Meta analysis: Recent develop-
ments in quantitative methods for literature reviews. Annual Review of
Psychology, 52, 59–82.
Schlenger, W. E., Kulka, R. A., Fairbank, J. A., Hough, R. L., Jordan,
B. K., Marmar, C. R., & Weiss, D. S. (1992). The prevalence of
post-traumatic stress disorder in the Vietnam generation: A multi-
method, multisource assessment of psychiatric disorder. Journal of
Traumatic Stress, 5, 333–363.
*Schnurr, P. P., Friedman, M. J., & Rosenberg, S. D. (1993). Preliminary
MMPI scores as predictors of combat-related PTSD symptoms. Ameri-
can Journal of Psychiatry, 150, 479–483.
Schuff, N., Neylan, T. C., Lenoci, M. A., Du, A. T., Weiss, D. S., Marmar,
C. R., & Weiner, M. W. (2001). Decreased hippocampal N-acetyl-
aspartate in the absence of atrophy in posttraumatic stress disorder.
Biological Psychiatry, 50, 952–959.
*Shalev, A. Y., Freedman, S., Peri, T., Brandes, D., Sahar, T., Orr, S. P.,
& Pitman, R. K. (1998). Prospective study of posttraumatic stress
disorder and depression following trauma. American Journal of Psychi-
atry, 155, 630–637.
*Shalev, A. Y., Peri, T., Canetti, L., & Schreiber, S. (1996). Predictors of
PTSD in injured trauma survivors: A prospective study. American Jour-
nal of Psychiatry, 153, 219–225.
*Shalev, A. Y., Sahar, T., Freedman, S., Peri, T., Glick, N., Brandes, D.,
et al. (1998). A prospective study of heart rate response following
trauma and the subsequent development of posttraumatic stress disorder.
Archives of General Psychiatry, 55, 553–559.
*Skre, I., Onstad, S., Torgersen, S., Lygren, S., & Kringlen, E. (1993). A
twin study of DSM–III–Ranxiety disorders. Acta Psychiatrica Scandi-
navica, 88, 85–92.
*Solomon, Z. (1995). The effect of prior stressful experience on coping
with war trauma and captivity. Psychological Medicine, 25, 1289–1294.
Solomon, Z., Benbenishty, R., Neria, Y., Abramowitz, M., Ginzburg, K., &
Abraham, O. (1993). Assessment of PTSD: Validation of the revised
PTSD Inventory. Israel Journal of Psychiatry & Related Sciences, 30,
110–115.
*Solomon, Z., Mikulincer, M., & Hobfall, S. (1987). Objective versus
subjective measurement of stress and social support: Combat-related
reactions. Journal of Consulting and Clinical Psychology, 55, 577–583.
*Solomon, Z., Oppenheimer, B., Elizur, Y., & Waysman, M. (1990).
Trauma deepens trauma: The consequences of recurrent combat stress
reaction. Israel Journal of Psychiatry & Related Sciences, 27, 233–241.
*Speed, N., Engdahl, B., Schwartz, J., & Eberly, R. (1989). Posttraumatic
stress disorder as a consequence of the POW experience. Journal of
Nervous and Mental Disease, 177, 147–153.
Spiegel, D., & Cardena, E. (1991). Disintegrated experience: The disso-
ciative disorders revisited. Journal of Abnormal Psychology, 100, 366–
378.
Spitzer, R. L., & Williams, J. B. W. (1983). Structured Clinical Interview
for DSM–III (SCID). New York: New York State Psychiatric Institute.
*Sutker, P. B., Davis, J. M., Uddo, M., & Ditta, S. R. (1995). War zone
stress, personal resources, and PTSD in Persian Gulf War returnees.
Journal of Abnormal Psychology, 104, 444–452.
*Tedstone, J. E., & Tarrier, N. (1997). An investigation of the prevalence
of psychological morbidity in burn-injured patients. Burns, 23, 550–554.
Tennant, C. (1988). Parental loss in childhood: Its effect in adult life.
Archives of General Psychiatry, 45, 1045–1050.
Tennant, C., Smith, A., Bebbington, P., & Hurry, J. (1981). Parental loss in
childhood: Relationship to adult psychiatric impairment and contact with
psychiatric services. Archives of General Psychiatry, 38, 309–314.
*Tichenor, V., Marmar, C. R., Weiss, D. S., Metzler, T. J., & Ronfeldt,
H. M. (1996). The relationship of peritraumatic dissociation and post-
traumatic stress: Findings in female Vietnam theater veterans. Journal of
Consulting and Clinical Psychology, 64, 1054–1059.
*Tucker, P., Dickson, W., Pfefferbaum, B., McDonald, N. B., & Allen, G.
(1997). Traumatic reactions as predictors of posttraumatic stress six
months after the Oklahoma City bombing. Psychiatric Services, 48,
1191–1194.
*Ursano, R. J., Fullerton, C. S., Epstein, R. S., Crowley, B., Kao, T.-C.,
Vance, K., et al. (1999). Acute and chronic posttraumatic stress disorder
in motor vehicle accident victims. American Journal of Psychiatry, 156,
589–595.
*Ursano, R. J., Fullerton, C. S., Epstein, R. S., Crowley, B., Vance, K.,
Kao, T.-C., & Baum, A. S. (1999). Peritraumatic dissociation and
posttraumatic stress disorder following motor vehicle accidents. Amer-
ican Journal of Psychiatry, 156, 1808–1810.
*Van Velsen, C., Gorst-Unsworth, C., & Turner, S. (1996). Survivors of
torture and organized violence: Demography and diagnosis. Journal of
Traumatic Stress, 9, 181–193.
*Watson, C. G., Anderson, P. E. D., & Gearhart, L. P. (1996). Social
maladjustment indicators in PTSD patients’families of origin. Journal
of Clinical Psychology, 52, 21–25.
Weiss, D. S., & Marmar, C. R. (1997). The Impact of Event Scale—
Revised. In J. P. Wilson & T. M. Keane (Eds.), Assessing psychological
trauma and PTSD: A practitioner’s handbook (pp. 399–411). New
York: Guilford Press.
*Weiss, D. S., Marmar, C. R., Metzler, T. J., & Ronfeldt, H. M. (1995).
Predicting symptomatic distress in emergency services personnel. Jour-
nal of Consulting and Clinical Psychology, 63, 361–368.
Weiss, D. S., Marmar, C. R., Schlenger, W. E., Fairbank, J. A., Jordan,
B. K., Hough, R. L., & Kulka, R. A. (1992). The prevalence of lifetime
and partial post-traumatic stress disorder in Vietnam theater veterans.
Journal of Traumatic Stress, 5, 365–376.
Wilkinson, L., & Task Force on Statistical Inference. (1999). Statistical
methods in psychology journals. American Psychologist, 54, 594–604.
*Wolfe, J., Sharkansky, E. J., Read, J. P., Dawson, R., Martin, J. A., &
Ouimette, P. C. (1998). Sexual harassment and assault as predictors of
PTSD symptomatology among U.S. female Persian Gulf War military
personnel. Journal of Interpersonal Violence, 13, 40–57.
World Health Organization. (1969). Manual of the international statistical
classification of diseases, injuries, and causes of death (8th ed.). Ge-
neva, Switzerland: Author.
World Health Organization. (1978). Clinical modification of the World
Health Organization’s manual of the international statistical classifica-
tion of diseases, injuries, and causes of death (9th ed.). Geneva, Swit-
zerland: Author.
Yehuda, R. (1998). Psychoneuroendocrinology of post-traumatic stress
disorder. Psychiatric Clinics of North America, 21, 359–379.
*Zaidi, L. Y., & Foy, D. W. (1994). Childhood abuse experiences and
combat-related PTSD. Journal of Traumatic Stress, 7, 33–42.
Received August 2, 2000
Revision received August 8, 2002
Accepted August 9, 2002 䡲
73
PREDICTORS OF PTSD: A META-ANALYSIS