The course of PTSD in naturalistic long-term studies: high variability of outcomes. A systematic review.

Abstract Background: With a lifetime prevalence of 8% posttraumatic stress disorder (PTSD) is one of the most common mental disorders; nevertheless, its longitudinal course is largely unknown. Aims: Our aim was to conduct a systematic review summarizing available findings on the prospective, naturalistic long-term course of PTSD and its predictors. Methods: Databases MEDLINE and PsycINFO were searched. Main selection criteria were: 1) naturalistic cohort study with a follow-up period of at least 3 years, 2) adult participants with observer-rated or probable PTSD at baseline. Results: Twenty-four cohorts (25 studies) were retrieved (14 with observer-assessed, 10 with probable PTSD). In total, they comprised about 10,500 participants with PTSD at baseline that were included in the long-term follow-ups. Studies investigating patient populations with observer-assessed PTSD found that between 18% and 50% of patients experienced a stable recovery within 3–7 years; the remaining subjects either facing a recurrent or a more chronic course. Outcomes of community studies and studies investigating probable PTSD varied considerably (remission rates 6–92%). Social factors (e.g. support) as well as comorbid physical or mental health problems seem to be salient predictors of PTSD long-term course and special focus should be laid on these factors in clinical settings. Conclusions: Included studies differed notably with regard to applied methodologies. The resulting large variability of findings is discussed. More standardized systematic follow-up research and more uniformed criteria for remission and chronicity are needed to gain a better insight into the long-term course of PTSD.

T raumatic experiences are common. Lifetime exposure rates range between about 50% and 70% in the general population (1 -3). Posttraumatic stress disorder (PTSD) presumably is the core psychopathology in the aftermath of traumatic events (4, 5) and while it is one of the most frequent mental disorders with a lifetime prevalence of about 8% (6), its long-term course is largely unknown (7).
The course of PTSD has been examined in retrospective and prospective longitudinal studies that differed notably with regard to applied methods, research objectives and assessed outcomes (8,9). Prospective studies investigating various samples of traumatized populations reported PTSD remission rates that ranged between 35% and 66% after 3 -36 months (10 -15). These fi ndings suggested that a considerable percentage of subjects with PTSD remitted as time passed, while in others PTSD was a chronic condition, lasting several years.
A recent meta-analysis quantitatively summarized remission rates from PTSD that had been reported in 42 naturalistic long-term outcome studies of more than 81,000 untreated subjects with PTSD (9). The authors included studies with a minimum follow-up period of 10 months and found that on average 44% of participants with PTSD at baseline were non-cases at follow-up. There was a very large amount of variation in remission rates over all samples (8 -89%), accordingly heterogeneity was found to be extremely high ( I 2 ϭ 97%). Findings of this meta-analysis show that participants with PTSD have been extensively studied but knowledge about the course of PTSD is still scarce.
One of the reasons why fi ndings on remission from PTSD vary enormously from study to study might be related to the problem that its assessment or more generally speaking the defi nitional problem of what defi nes the characteristics of an illness and is vital for treatment planning, as it helps identifying subgroups possibly subjected to a more unfavorable course and consequently needing special attention. Ideally, the investigation of the course of a disorder begins with its fi rst occurrence or even beforehand, as otherwise its chronicity may be overestimated. However, in reality this is only rarely possible. Bearing this in mind, we did not limit our search to studies following participants from the start of their illness but rather tried to incorporate all long-term fi ndings, an approach that possibly hampers comparability of fi ndings but on the other hand paints a more realistic picture of the current state of PTSD research.
Our research questions were: How likely is the experience of a remission or recovery • versus a chronic or recurrent course in subjects with PTSD or probable PTSD over a time span of at least 3 years? Which factors have found to be associated with a • favorable or unfavorable long-term course?

Identifi cation of relevant literature
To identify relevant literature, electronic databases Medline and PsycINFO were searched up to October 1, 2014 for English language articles. Manual searches of reference lists of included studies and relevant reviews were also performed. Search terms were " PTSD " , " posttraumatic stress disorder " , " course " , " trajector * " , " remission " , " remit * " , " recover * " and " follow-up " (see Appendix).

Selection criteria
The following selection criteria were applied: 1) naturalistic cohort study with a follow-up period of at least 3 years, 2) adult participants diagnosed with PTSD at baseline, 3) study presenting at least one follow-up assessment of participants with PTSD at baseline, 4) PTSD diagnosed via interview (observer rating) or a selfreport scale with a defi ned cut-off score or algorithm for establishment of diagnosis, 5) baseline assessment taking place in 1980 or later, and 6) minimum of three participants in long-term follow-up. An important factor that might hamper comparability of PTSD rates across different studies is the use of self-rated vs. observer-rated PTSD assessments. The former often apply a cut-off score to assess PTSD caseness or simply refer to PTSD symptoms, while the latter apply (structured) interviews that are administered by trained staff members or experienced clinicians. An advantage of studies using self-rating scales is that they usually comprise much larger samples as diagnostic assessment is less time and cost-intensive. On the other hand, the sole use of self-report scales may carry a remission in PTSD and how it can be assessed properly has yet not been solved. This issue has recently been broadly and critically discussed by North & Oliver (15). Two main approaches for defi ning remission exist: 1) a symptom based defi nition (i.e. no more symptoms of the disorder are present) and 2) a threshold-based defi nition (i.e. some symptoms remain, but they have fallen beneath a diagnostic threshold). According to DSM-IV-TR (16), for a full remission the fi rst defi nition applies while for partial remission the second one applies. The assessment instrument that is used in a study usually determines the defi nition of outcome that has to be applied which in turn infl uences reporting of fi ndings. Moreover, studies applying a threshold-based defi nition usually also refer to the term remission, which in truth can be misleading and adds to the observed heterogeneity of fi ndings. Consequently, more uniform defi nitions are needed to enhance the fi eld. Nevertheless, North & Oliver (15) also state that regardless of the defi nition used, prior studies demonstrated substantial PTSD chronicity (p. 1194).
Various predictors of PTSD onset have been identifi ed, such as early childhood trauma, lower education, female gender, belonging to an ethnic minority and preexisting psychiatric psychopathology (17 -21). It has also been suggested that the severity of the traumatic event itself, initial reactions in the face of a trauma as well as peritraumatic dissociation or early hyperarousal play a role in PTSD onset (17 -19, 22 -24). In contrast, only a few factors predictive of the PTSD course have been described so far. Among them are early postdisaster symptoms (e.g. acute stress symptoms), which have been suggested to be predictive of a more chronic long-term course (25 -27). The same seems to be the case for posttraumatic circumstances like occupational, fi nancial, health or family problems as well as a lack of social support in the aftermath of a trauma (17, 28, 29).

Aims of the study
Our aim was to enlighten the long-term course of PTSD and its predictors by specifi cally focusing on true panels, i.e. studies prospectively assessing the same individuals at more than one point in time. Thus, we did not include studies merely investigating PTSD prevalence in subjects several years after the experience of a trauma but only longitudinal studies with at least one long-term follow-up measurement. Previous research suggested that PTSD can take a chronic course, lasting several years (1); moreover, prevalence rates several years after a trauma were found to be elevated within a wide range of traumatized populations (5, 30 -33). Taking these fi ndings into account, we specifi cally aimed at gaining an insight into the multiannual course of PTSD by only including studies with a follow-up period of at least 3 years. Information on the long-term course and its predictors allows an insight into was defi ned as the comeback of full PTSD symptoms after recovery had been achieved. A chronic course was defi ned as either fulfi lling the diagnostic criteria or screening positive for PTSD at both or all diagnostic assessments, or by still showing considerable residual symptoms. The term " partial remission " was dropped for the present review.

Data extraction and review process
After completing literature searches as outlined in Table A1 (see Appendix to be found online at http:// informahealthcare.com/doi/abs/10.3109/08039488. 2015. 1005023), all hits ( n ϭ 331 ϩ 12) were saved in End-Note. Two researchers (CS, MH) independently screened titles and abstracts according to the predefi ned selection criteria. All potentially relevant articles were then retrieved for full-text review ( n ϭ 106). Uncertainties regarding inclusion were discussed. A fl ow chart showing the process of study selection and reasons for exclusion is given in Fig. 1.
higher probability of over-or underestimating PTSD caseness. We decided to include both types of studies and report results separately.
For reporting long-term outcome, we used the percentages of recovery, remission, recurrence and chronic course as provided by included studies. As mentioned above, defi ning remission in PTSD can be problematic with regard to symptom versus threshold-based criteria. Furthermore, the terms recovery and remission also imply different outcomes but are sometimes used synonymously to describe improvement. Thus, for our review, we decided to apply the following defi nitions of outcome: in order to be classifi ed as recovered or fully remitted participants had to be described as (nearly) symptom-free over a certain time span (i.e. symptombased defi nition). In contrast, participants were classifi ed as remitted if they were free of a PTSD diagnosis at a certain point in time. In this defi nition, remission implies a reduced PTSD symptomatology, which could, however, merely refl ect a momentary shift in PTSD severity (i.e. a threshold-based defi nition). Recurrence time span of 8 consecutive weeks. The low recovery rate of 18% in the study by Zlotnick et al. (35) might be explained by two circumstances: all patients had a comorbid anxiety disorder and additionally all patients had chronic PTSD, i.e. PTSD with a duration of at least 1 year (assessed retrospectively) before study intake. Therefore, this sample possibly represented a group of more severely affected patients. Furthermore, Ansell et al. (37) reported a recurrence rate of 34%, which means that more than one third of the 77% who had recovered at some stage during the 7-year follow-up did experience a further episode of full PTSD symptomatology. Together with those patients who did not experience recovery at all over the 7-year period (23%), this adds up to about 50% of patients who either encountered a more chronic or a recurrent course, while the remaining 50% experienced a stable recovery with no relapse. Both studies were carried out in clinical settings and included treatment-seeking patients, but information if treatment was PTSD specifi c, adequate or suffi cient was not obtained [a recent 15-year follow-up of the study by Zlotnick et al. (35) that additionally included patients who developed PTSD during follow-up reported a very low recovery rate of 20% (39)]. The third study with clinical patients included a more homogeneous group of psychiatric outpatients that all were refugees from the former Yugoslavia and found similar rates: 71% had PTSD at both assessments, while only 29% were remitted (40).

PRIMARY CARE PATIENTS
One study examined the long-term course in primary care patients (again by the use of LIFE) and found a recovery rate of 38% after 5 years (36). However, nearly 30% of those who had recovered experienced at least one recurrence. Therefore, based on the total sample, about 27% of patients experienced a stable recovery, while the remaining 73% had either a recurrent or a more chronic course.

PARTICIPANTS FROM THE COMMUNITY
The remaining 10 community studies found PTSD remission rates that ranged considerably between 12.5% and 92% after 3 -10 years.
The worst long-term result could be seen in the only study examining the consequences of a natural disaster (41). Here, a sample of eight fi refi ghters with PTSD at baseline was followed after a serious bushfi re and only 12.5% were found to be remitted after 3 years, while 87.5% still had PTSD. This fi nding might partly be explained by the " extreme " (p. 23) nature of trauma exposure: the author refers to the bushfi re as " unusually intense and uncontrollable " (p. 23).
The most favorable outcomes were seen in a sample of 11 motor-vehicle accident survivors (42) with a remis-

Results
In all, we found 25 studies (investigating 24 different cohorts) that had been published between 1988 and 2014. Fourteen studies used observer-rated diagnostic assessments (Table 1), 10 studies used self-reports and therefore assessed probable PTSD (Table 2).

CHARACTERISTICS OF INCLUDED STUDIES
Studies with observer-assessed PTSD included about 703 subjects in their long-term follow-ups. Follow-up times ranged between 3 and 10 years, the number of assessments (including the baseline assessment) ranged between 2 and 8. Four of the 14 studies included patients (clinical or primary care setting), while 10 studies investigated subjects from the community.
We will report results separately with regard to where subjects were recruited, as participants from primary, secondary or tertiary care settings might have a different long-term prognosis than participants recruited in the community (34). Additionally, three of the four studies investigating patients (35 -37) applied a different research design than the other studies, a fact that is worth mentioning: fi rstly because they included patients who reported various kinds of traumatic events (e.g. rape, assault, serious accidents, witnessing violence) as opposed to subjects who had all undergone the same traumatic event and were then followed for a certain time span. Secondly, due to sample composition, the time between the individual trauma and baseline assessment differed for each patient. Thirdly, they assessed recovery/ full remission as described above. Also, in studies comprising patient samples most subjects received some kind of treatment (psychosocial and/or pharmacological, generally referred to as " psychiatric " by the authors) -which was, however, not systematically recorded, as all studies were observational in nature.
Generally, the time spans between the traumatic events and the baseline assessments for PTSD varied over all studies. In one study, the time that had passed since the traumatic event was approximately 50 years (38). In most of the other studies, the time span between the traumatic event and the baseline PTSD assessment was much shorter and took place within a maximum of 2 years. CLINICAL PATIENTS Three studies (35, 37, 40) investigated PTSD patients (of whom all had comorbid disorders, i.e. personality disorders, major depression or other anxiety disorders). Recovery rates varied considerably between 18% over 5 years (35) and 77% over 7 years (37). Both of these studies used the Longitudinal Interval Follow-up Evaluation (LIFE), which defi nes recovery as a Psychiatric Status Rating (PSR) Յ 2 (no or only minimal symptoms) over a  sion (i.e. were free of PTSD at all later assessments). The remaining veterans showed either a fl uctuating course or a remission at one of the later assessments. Koenen et al. (50), who studied Vietnam veterans about 9 and 23 years after the war, found that 54% were remitted at the second assessment. Solomon et al. (52) found that remission rates depended on whether the combat veterans had also been prisoners of war (14.3% vs. 50%), but again this fi nding has to be interpreted with caution, as this study consisted of only 11 veterans with PTSD at baseline. Finally, in a study following a sample of service personnel deployed to Iraq, 32% were remitted after 3 years, while 33% had probable PTSD at both times (53). The remaining 36% showed some symptomatic improvement.

STUDIES RELATING TO 9/11
Three of the four long-term studies investigating PTSD in relation to the 9/11 terrorist attacks in Manhattan found remission rates that were quite similar (54 -56). Each of these three studies comprised large samples, ranging from 458 to 6141 participants and similar lengths of follow-up (3 -5 years). Baseline assessments in these studies either took place within the fi rst year post-9/11 or 2 -3 years after the attacks. Between 38.7% and 66.9% of study participants experienced a chronic or severe chronic course, while one-third to two-thirds had a remitting course. Meanwhile, Neria et al. (57) found a remission rate of 89% in 47 primary care patients investigated 1 and 4 years after 9/11. The difference in remission rates might be because participants in the former studies had a high probability of having been more severely affected by the attacks (i.e. rescue and recovery workers, lower Manhattan residents and offi ce workers, as well as passersby).

Factors associated with the long-term course of PTSD
Ten studies reported factors that were associated with the long-term course of PTSD. As can be seen in Table 3, the most prominent negative course predictors were (comorbid) mental health problems (found by fi ve studies) as well as social factors (e.g. feeling unsupported) as found by four studies.

Discussion
As far as we know, this is one of only two studies systematically reviewing fi ndings on the prospective longterm course of PTSD in adults (9). Overall it can be said that considering the frequency and associated disease burden of PTSD (58), the number of studies eligible for inclusion was relatively small, especially as the number of subjects with PTSD included in fi ve of the 24 longterm follow-ups was quite limited (4 -11 participants). On sion rate of 81% 5 years after baseline PTSD assessment as well as in a sample of 26 North Korean defectors with a 92% remission rate after 7 years (43).
One study, investigating holocaust survivors, methodologically stands out from the others as the trauma happened about 50 years prior to baseline assessment (38). The authors found that 30% of their participants were remitted from PTSD after 10 years.
The remaining fi ve studies found remission rates that ranged between 37% and 77% over 3 -6.5 years: Two studies followed victims of a mass shooting incident over a course of 3 years and found that 75% and 43%, respectively, were free of a PTSD diagnosis at follow-up (44, 45). Three studies followed refugees or war exposed civilians that stayed in their home country over a course of 3 -6 years and found that between 80% and 67% were free of a PTSD diagnosis at follow-up (46 -48). However, of those 67%, only 37% were asymptomatic while 30% still suffered from depression. Finally, in a study with survivors of the Oklahoma City bombing, 37% were recovered 6.5 years after baseline PTSD assessment (7).

CHARACTERISTICS OF INCLUDED STUDIES
The total number of participants in studies with selfreport PTSD was considerably higher ( n ϭ 9848) than in studies with observer-rated PTSD. Follow-up times ranged between 3 and 20 years, and the number of assessments (including the baseline assessment) ranged between two and four (Table 2). Five studies investigated war veterans, four studies followed subjects exposed to the 9/11 terrorist attacks in Manhattan and one study included service personnel deployed to Iraq.

STUDIES WITH WAR VETERANS
Five studies included war veterans, i.e. veterans from WWII, the Vietnam, the Yom Kippur or the Lebanon war (32, 49 -52). Time that had passed between (the end of the) war and the fi rst PTSD assessment varied between 1 and about 50 years, which means that some men were investigated immediately others not until several decades later. Overall, between 13% and 75% of veterans remitted over the course of follow-up; thus outcomes were again hugely variable. Studies in which more than four decades had passed between the war and the baseline PTSD assessment found the lowest remission rates (13% and 32%) with a majority of veterans still having probable PTSD at follow-up 4 and 6 years later, respectively (32, 49).
One study with four assessments 1, 2, 3 and 20 years after the Lebanon war and two subsamples (veterans with or without combat stress reaction, CSR) provided more detailed course data (51). Here, only 6% (with CSR) and 25% (without CSR), respectively, showed a stable remis-PTSD diagnosis determined by an interview versus probable PTSD established through self-reports. Therefore, other methodological aspects might have played a role, as studies varied notably regarding the following factors: Follow-up length (3 -20 years); Time between the traumatic event and the baseline PTSD assessment (4 -8 weeks up to about 50 years); Kind of trauma/kind of traumatized population; Number of PTSD assessments (2 -10); Sample types (clinical, primary care, community); Defi nition of remission/recovery (symptom based versus threshold-based).
As mentioned above, studies reviewed in the present article varied with regard to symptom-based versus threshold-based defi nitions of outcome: fi ve of the studies included in our review (7,36,37,47,59) used a more symptom-based outcome defi nition (e.g. a PSR-the other hand, a notable portion of studies was based on larger samples, with the largest comprising more than 6000 participants screening positive for PTSD at baseline (54).

Methodological factors and their possible impact on fi ndings
A majority of studies investigated man-made disasters (e.g. war, terrorist attacks, mass-shootings), while one focused on a natural disaster (41) and one on traffi c accident victims (42); thus the generalizability of our fi ndings is limited with respect to natural and accidental traumas.
Results show that there is considerable variation in the percentages of participants who do or do not remit. Surprisingly, and contrary to our expectations, outcomes are equally heterogeneous irrespective of PTSD status, i.e. Table 3 . Factors found to be predictive of an unfavorable long-term course of posttraumatic stress disorder (PTSD).
ing of treatment-seeking participants (i.e. patients) might be subjected to bias (34), as it is likely that they are experiencing a more severe illness course, which led them to seek treatment in the fi rst place. Consequently, this might result in an overestimation of chronicity. Thus, it is of note that results from these studies may not generalize to non-treatment-seeking subjects. Recent studies conducted in the aftermath of 9/11 pro-4.
vided us with valuable information on the longitudinal course of PTSD. They found that chronic courses could be seen in about one third to two-thirds of subjects with PTSD.

Predictors of the long-term course
Several predictors of PTSD onset have been identifi ed while factors predicting the long-term course of PTSD are lesser known. Based on fi ndings reviewed here, it can be seen that some factors related to a more chronic PTSD course resemble predictors of PTSD onset, which was the case for female gender, belonging to a minority race, childhood trauma, trauma severity and severe initial reactions to trauma. Furthermore, some studies reviewed here support earlier research, suggesting that the amount of early postdisaster symptoms as well as negative posttraumatic events, such as adverse life events, a lack of social support or health related problems, psychological or physical in nature, might play a role in the long-term course of PTSD (17,24,29,60). More specifi cally, four studies found social factors predictive of PTSD course. Social support, in many of its connotations -be it the acknowledgement in the community, the possibility to disclose one ' s thoughts and feelings to a near person, living in a partnership, or the availability of friends -seems to be an integrative part in the long-term development of PTSD (17, 18). However, the factor marital status has to be viewed with caution in this context: while being separate, widowed or divorced was found to be a risk factor for a negative course in two studies (36, 56), prior research found that being married could also be a risk factor for an unfavorable development in the aftermath of a trauma, especially for women, while the opposite may be true for men (19). Accordingly, the same study who reported being separate, widowed or divorced could be a risk factor found that being married or cohabiting was also a negative predictor (56).
The fi ndings on course predictors have some implications for clinical practice where it is vital to identify those subjects who are at risk for an unfavorable course and supply them with suited treatments or other interventions. On the one hand the fi ndings suggest that subjects with PTSD who are embedded in supportive social structures and do not suffer from posttraumatic physical impairments or comorbid mental problems might be at level of Յ 2 [no or very mild symptoms only] over several weeks), while the remaining studies provided threshold-based rates of remission and chronicity (or failed to make it clear which criteria were applied). We investigated whether this difference could explain some of the heterogeneity in outcome rates. When only looking at outcomes of these fi ve studies, recovery rates (i.e. being free of symptoms) were 18%, 37%, 37%, 38% [27%] and 77% [50%]. Some of the rates decrease, when recurrences are taken into account (see fi gures in square brackets). As can be seen, there is still some variability, but also a tendency towards uniformly low values of stable recoveries (below 50%). This confi rms our assumption that the high variability of fi ndings is partly due to heterogeneous defi nitions of outcome.
Taken together, the methodological differences listed above impede comparability between studies and likely account for (some or most of) the found variability regarding long-term PTSD outcome. Nevertheless, despite this methodological diverseness some conclusions may be drawn: Participants whose traumatic experiences happened up 1. to several decades prior to the baseline PTSD assessment (which was the case in fi ve studies) and who were followed up over some years in later life, often showed a chronic course (46 -87%). While it remains unclear whether participants in these studies were new cases in the sense of a delayed PTSD onset or had suffered PTSD or PTSD symptoms throughout the years, the results reported here suggest that PTSD occurring in older age is rather chronic. Consistent with our fi ndings, previous research pointed towards a U-shaped curve of PTSD in long-term survivors of trauma, with symptoms being elevated shortly after a traumatic experience, declining thereafter and increasing again in older age, when age-related problems and age-related thresholdsituations appear (32). Three studies following patient samples applied quite 2.
uniformed methods, as they used at least six observerrated assessments and well-defi ned outcome criteria. Interestingly, they provided less heterogeneous data on the long-term course of PTSD suggesting that methodological aspects play an important role in investigating long-term outcome. These studies found that the long-term PTSD prognosis is favorable in only 18 -50% of patients and that PTSD can be quite fl uctuating in its course with recurrence rates ranging between 29.5% and 34%. As these studies investigated consecutive patients with (comorbid) PTSD, we have no information on the time that had elapsed between fi rst PTSD occurrence and baseline assessment. Studies with patient samples are limited in their gener-3.
alizability. When studying the long-term course of an illness it is important to consider that samples consist-know what role treatment might have played in the outcome of PTSD as it is reported here.
Summing up, the results presented here give some insight into the naturalistic long-term course of PTSD in different samples. In order to further enlighten this issue, future research should apply more than one follow-up examination as well as enhanced defi nitions of outcome.

Declaration of interest:
The authors report no confl icts of interest. The authors alone are responsible for the content and writing of the paper.
This research was supported by grants from the Dr. Karl-Wilder-Stiftung. The sponsor had no involvement in study design, in collection, analysis and interpretation of data, in writing the article, and in the decision to submit the article for publication.
lesser risk for the development of a chronic PTSD course. Those, on the other hand who are socially isolated or do not feel supported and are physically or mentally impaired seem to be at higher risk for non-remission and should therefore be identifi ed early on to prevent chronicity. Consequently, a thorough examination of individuals with PTSD should always comprise an assessment of possible comorbid mental and physical disorders as well as an evaluation of the psychosocial situation a patient lives in (60,61).

Limitations
This review has some limitations: We designed the review to present a wide range of • articles published in the research area in focus. Consequently, the selection criteria we used were broad which led to the inclusion of heterogeneous studies and fi ndings that are diffi cult to compare. Considering the large heterogeneity in studies and fi ndings, the performance of a meta-analytic evaluation did not seem suitable and therefore conclusions are based on a merely descriptive synthesis.
There is the possibility that some studies were missed • because of the way our search strategy was designed. In an attempt to attenuate this, we did not rely solely on electronic searches but manually searched the reference lists of included studies. Additionally, as we limited our search to English language articles, there is the risk of having missed some relevant studies. Moreover, as we focused on information about recovery • and remission rates from the same individuals at baseline and follow-up(s), we have -with one exception who provided a mixed approach (56) -neglected data regarding varying symptom levels over time. Studies assessing PTSD severity over time by using symptom-profi les (24, 38, 62) generally suggested that the PTSD course is better characterized by gradual levels of severity and changes within the core symptoms (i.e. hyperarousal, avoidance and intrusions) than by examination of the diagnostic status, or as Solomon et al. (62) put it, " PTSD is not a monolithic disorder " (p. 837), a view that supports the varying long-term course found in some studies reviewed here. Also, as most reviewed studies relied on two assessments only (i.e. one baseline and one follow-up measurement), they only allowed a limited view into the multiannual course of PTSD. Studies including patient samples generally applied a higher number of assessments than community studies (on average 5.6 versus 2.3) and consequently especially the latter might have missed important course information, such as recurrences.
Generally, information on whether subjects of included • studies received treatment (and if they did, what kind of treatment) was sparse or non-existent; thus we do not