Recurrence of post-traumatic stress disorder: systematic review of definitions, prevalence and predictors

Background Many people will experience a potentially traumatic event in their lifetime and a minority will go on to develop post-traumatic stress disorder (PTSD). A wealth of literature explores different trajectories of PTSD, focusing mostly on resilient, chronic, recovered and delayed-onset trajectories. Less is known about other potential trajectories such as recurring episodes of PTSD after initial recovery, and to date there has been no estimate of what percentage of those who initially recover from PTSD later go on to experience a recurrence. This systematic review aimed to synthesise existing literature to identify (i) how ‘recurrence’ of PTSD is defined in the literature; (ii) the prevalence of recurrent episodes of PTSD; and (iii) factors associated with recurrence. Methods A literature search of five electronic databases identified primary, quantitative studies relevant to the research aims. Reference lists of studies meeting pre-defined inclusion criteria were also hand-searched. Relevant data were extracted systematically from the included studies and results are reported narratively. Results Searches identified 5,398 studies, and 35 were deemed relevant to the aims of the review. Results showed there is little consensus in the terminology or definitions used to refer to recurrence of PTSD. Because recurrence was defined and measured in different ways across the literature, and prevalence rates were reported in numerous different ways, it was not possible to perform meta-analysis to estimate the prevalence of recurrence. We also found no consistent evidence regarding predictors of PTSD recurrence. Conclusion A clear and consistent evidence-based definition of recurrence is urgently needed before the prevalence and predictors of recurrence can be truly understood. Supplementary Information The online version contains supplementary material available at 10.1186/s12888-023-05460-x.


Background
Potentially traumatic events are common.Research suggests that over 70% of people will experience a potentially traumatic event (such as witnessing death or serious injury, automobile accident, life-threatening illness or injury, or violent encounter) in their lifetime [1].Understandably, these events can be very distressing in the short-term and many people will experience acute post-traumatic symptoms in the immediate aftermath of a traumatic event, including intrusive symptoms (e.g.recurrent unwanted thoughts, nightmares); avoidance symptoms (e.g.emotional numbing, social withdrawing); hyperarousal (e.g.easily startled, feeling 'on edge'); and physical symptoms (e.g.chest pain, dizziness) [2].For the majority, these symptoms will decline naturally without intervention [3], typically within the first four weeks [2].An important minority will find their symptoms persist for longer than a month.Those who continue to experience persistent re-experiencing of the traumatic event; avoidance of stimuli associated with the event; negative alterations in cognitions and mood and alterations in arousal and reactivity, causing clinical distress or functional impairment and not attributable to any other medical condition, are likely to be diagnosed with posttraumatic stress disorder (PTSD) [4].Although only a minority of people who experience potentially traumatic events will go on to develop PTSD, it remains one of the most common mental disorders with lifetime prevalence estimated to be between 8% [5] and 12% [6].PTSD is associated with reduced health-related quality of life and physical comorbidities, as well as major socio-economic costs [7].
The early 2000s saw a shift from studying PTSD itself as an outcome to studying change in symptoms as an outcome [8], with a wealth of studies using modelling approaches such as latent class growth analysis and latent growth mixture modelling to identify different trajectories of PTSD.Most of this literature identifies four trajectories, two of which are relatively stable trajectories (chronic, a stable trajectory of post-traumatic stress symptoms, and resilient, a stable trajectory of healthy functioning after an adverse event), and two which display dynamic symptom patterns (recovered, i.e. decreasing symptoms after an initial diagnosis of PTSD, and delayed-onset, i.e. increasing symptoms not meeting the diagnostic criteria for PTSD until potentially months or even years after traumatic exposure) [9].Van de Schoot et al. [10] suggest that the two trajectories which typically occur less often (chronic and delayed-onset) are at risk of being overlooked by researchers or overwhelmed within the data by the larger trajectories.There may also be other less-researched or less-understood trajectories overlooked to an even greater extent.For example, one previous review [11] identified limited evidence of another, smaller trajectory referred to as a 'relapsing' or 'recurring' PTSD trajectory, in which individuals develop PTSD, are free from symptoms for long enough to be considered 'recovered' , and then experience a recurrence of symptoms.
Recurrence is given relatively little attention in the PTSD literature, perhaps due to limitations of study methodologies and the complexities of studying recurrence.For example, Santiago et al. [11] note that few studies of PTSD follow participants for more than a year or with more than two assessments.Clearly, it would not be possible for researchers to identify recurrence of PTSD if data is only collected for two time-points: the only possible outcomes would be low symptom levels at each time-point ('resilience'), high symptoms at each time-point ('chronic'), or low level of symptoms at one time-point and a high level at the other (either 'recovery' or 'delayed-onset' depending on time-point at which symptoms were experienced).Additionally, studies which only follow up participants for a year or less are unlikely to clearly identify a recurrent trajectory of PTSD given the time needed to both recover and to experience a recurrent episode.The timing of PTSD assessment is also important: identification of PTSD recurrence relies on studies capturing the presence of symptoms during the recurrence, rather than before it occurs or after recurring symptoms have subsided.Therefore, it is perhaps unsurprising that the majority of the literature does not identify a 'recurring' trajectory of PTSD.Even studies which do identify recurrences often group these in with other trajectories: for example, Mota et al. [12] identified 'recurrent' cases of PTSD (individuals who had a lifetime diagnosis in 2002 and another post-2002 diagnosis reported in 2018), but grouped 'persistent' and 'recurrent' cases of PTSD together.Magruder et al. [13] identified a group of recurrent cases of PTSD -individuals who had lifetime PTSD pre-1992 but not a current diagnosis in 2002, who then had a diagnosis again in 2021, but these were grouped with 'chronic' cases.Karamustafalioglu et al. [14] simply include an 'other' group constituting both recurrent cases (individuals who met the criteria for PTSD diagnosis 1-3 months post-trauma and at the third follow-up 18-20 months post-trauma, but not at the second follow-up 6-10 months post-trauma) and others with delayed-onset PTSD which resolved.Boe et al. [15] identified a group of individuals with 'reactivated' PTSD who reported remission from PTSD in the first five years after the North Sea oil rig disaster of 1980 and a new episode at any point between 1985 and 2007.However, the authors suggest that there are blurred boundaries between delayed-onset and 'reactivated' PTSD, going on to include 'possible delayed cases' in their analysis of reactivated PTSD.
It is important to note that even the definitions of the more well-established trajectories of PTSD are not without their controversies.For example, Andrews et al. [16] point out the ambiguity in the criterion for delayed-onset PTSD, questioning whether 'the onset of symptoms' refers to any symptoms which might eventually lead to PTSD or only to full-blown PTSD itself.North et al. [17] comment on the ambiguities involved in the term remission (i.e.whether remission should be symptom-based or threshold-based) as well as the term onset (i.e.whether onset refers to first symptoms or first meeting diagnostic criteria).Definition of recovery also appears to differ from study to study, with some authors considering recovery to be symptom-based (i.e.no symptoms of the disorder remain) and others considering it to be threshold-based (i.e.some symptoms may remain, but they are beneath the diagnostic threshold) [18].
To date, several systematic reviews have been published which focus solely on only one PTSD trajectory.For example, previous reviews have focused on the delayed-onset trajectory [16,19]; the recovery trajectory [20]; and the resilient trajectory [21].To date there has not been a literature review examining evidence of a recurrent trajectory of PTSD.Berge et al. [22] aimed to systematically review research on relapse in veterans but found no studies reporting actual rates of relapse or recurrence.Reviews have also explored the risk of relapse of various anxiety disorders, including PTSD, after discontinuation of antidepressants [23] and after cognitive behavioural therapy [24].However, there have been no reviews attempting to quantify the risk of PTSD recurring, establish the predictors of recurrence, or quantify how much each predictive factor contributes to the risk of recurrence.The current review aimed to fill this gap in the literature by synthesising existing published data on how researchers define 'recurrence' of PTSD, recurrence rates of PTSD, and predictive factors of recurrence.
Having an appropriate understanding of recurrence is important as the concept needs to be properly understood in order to take steps to mitigate the risks of recurrent PTSD episodes.Mitigating the risk of PTSD recurring could benefit the health and wellbeing of trauma-exposed individuals and could reduce the socioeconomic costs to the wider society [7].The prevalence of recurrence is of particular importance to occupational medicine: regularly trauma-exposed organisations, for example, are often faced with decisions about when (and if ) staff who have had and recovered from PTSD should return to the frontline duties.Understanding the risk of recurrent episodes may therefore have implications for those in charge of making such decisions.The present time is also a particularly relevant time to develop our understanding of recurrence of PTSD, as it is possible that the COVID-19 pandemic could contribute to recurrence.The pandemic has been declared a potential traumatic stressor, with research suggesting that COVID-19 survivors are at elevated risk of experiencing PTSD [25] and that PTSD symptoms may also develop due to quarantine [26], concerns about the health of loved ones, or economic loss as a result of the pandemic [27].Hori et al. [28] suggest that the daily television updates regarding COVID-19 could trigger memories of surviving a previous traumatic situation, and exacerbate subthreshold PTSD symptoms.Therefore, experiencing the pandemic could potentially cause a recurrence of symptoms in people who have previously been diagnosed with PTSD.
The aim of this review was to collate literature which provides evidence of the lesser-studied 'recurrent' trajectory of PTSD and to identify: (i) the definitions of 'recurrence' used throughout the literature; (ii) prevalence of recurrence; and (iii) risk and protective factors for the recurrent trajectory of PTSD.

Method
This review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [29].Our population of interest were people who had been diagnosed with, recovered from, and experienced a recurrence of PTSD (as diagnosed by a clinician or validated PTSD assessment tool).For the aim relating to prevalence of recurrent episodes, studies needed to involve a suitable design allowing prevalence to be assessed: for example, studies involving a population of people who had recovered from PTSD, followed over time to show how many had a recurrent episode and how many did not.For the other aims (i.e., definitions of recurrence and factors associated with recurrence), a comparison group was not necessary.

Registering the review
A protocol for the current review was developed and registered with PROSPERO on March 9th 2023 (registration number CRD42023405752).The only deviation from the protocol was the addition of another quality appraisal tool, due to finding a study design (retrospective analysis of existing health data) which we had not anticipated.

Eligibility criteria
To be included in the review, studies needed to (1) be published in peer-reviewed journals, (2) be published in the English language, (3) use quantitative methodology, (4) use a standardised tool to assess PTSD and (5) present data on recurrence rates of PTSD and/or factors associated with PTSD recurrence.There were no limitations relating to publication date or location of the studies.Case studies were excluded but there were no other exclusion criteria relating to population size.

Data searching and screening
A systematic literature search was carried out to examine definitions, prevalence rates and predictors of PTSD recurrence.Four electronic databases (Embase, PsycInfo, Medline and Web of Science) were searched on 24th November 2022, using a combination of search terms relating to PTSD, recurrence, and prevalence/predictors which were combined using Boolean operators.The full list of search terms is presented in Appendix 1.The US Department of Veterans Affairs National Center for Post-Traumatic Stress Disorder's PTSDPubs database (formerly PILOTS) was searched separately on the same date using the individual terms 'recurrence' and 'recurrent' and limited to peer-reviewed articles.Reference lists of articles deemed to meet the inclusion criteria were also hand-searched.
All citations resulting from the literature searches were downloaded to an EndNote library where duplicates were removed.The titles of all citations were then screened for relevance to the review, with any clearly not relevant being excluded.Abstracts were then screened for eligibility and the full texts of all remaining citations after abstract screening were located and read in their entirety to identify studies meeting all inclusion criteria.The literature searches and screening were carried out by the first author.The two authors met regularly throughout the screening process to discuss any uncertainties about inclusion or exclusion until a decision was reached.

Data extraction
The first author carried out data extraction of all citations deemed to meet the inclusion criteria.Data were extracted to a Microsoft Excel spreadsheet with the following headings: authors, year of publication, country, study design, sampling method, inclusion/exclusion criteria, study population size, socio-demographic characteristics of participants, type of trauma exposure, time-points at which PTSD was assessed, tools for assessing PTSD, definitions of recovery and recurrence, whether any PTSD treatment was received, prevalence rates of recurrence, and factors examined as potential predictors of recurrence.

Data synthesis
For the first aim of the review (relating to definitions of recurrence), we designed a table to present data relating to how 'recurrence' was understood and defined in each study.The tools used to diagnose and measure PTSD symptoms in the first place are important in understanding how PTSD is defined, so first the assessment tools used in each study were extracted into the table.Given that we wanted to understand the length of time an individual needs to be free of PTSD in order to be considered 'recovered' , for each study we also included the time-points of PTSD assessment in the table.Next, we included the definitions of recovery and recurrence from each study, explained narratively in the table.We also added information to this table to report whether participants had received PTSD treatment during each study, as some studies focusing on interventions used 'response to treatment' in their definitions of recovery.We compared the different definitions used within the studies to establish whether there was consensus within the literature around (i) whether recovery and recurrence are symptom-based or threshold-based and (ii) how long the recovery period between initial diagnosis and recurrent episodes needs to be in order to be considered recurrent rather than chronic PTSD.
The second aim related to prevalence of PTSD recurrence.Due to the various research designs and definitions of 'recurrence' in the literature, as well as the different ways in which prevalence was reported, meta-analytic techniques could not be used.Rather, we presented the prevalence data as it was reported in each study.This sometimes meant presenting the prevalence of PTSD recurrence within an entire trauma-exposed population, including those who never experienced PTSD at any time.Other times, this meant presenting the prevalence of PTSD within a population who all had PTSD at one time-point, and other times this meant presenting the prevalence of PTSD within a group who had recovered from PTSD.
Finally, in order to explore factors associated with PTSD recurrence, all variables considered as potential covariates were recorded individually for each study.Each potential predictive factor was descriptively reported in a table, and any found to be significantly associated with experiencing PTSD recurrence were bolded to differentiate between non-significant and significant findings.Factors are also described narratively within the results section.Insights from thematic analysis [30] were used to group similar data together.For example, data relating to gender or age as predictors of recurrence were coded 'socio-demographic' and discussed together within the results.

Quality appraisal
We appraised the quality of studies using National Institutes of Health (NIH) tools: either the Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies or the Quality Assessment of Controlled Intervention Studies tool, depending on study design.Concurrent with other reviews [e.g.31] we rated quality as 'poor' if studies scored 0-4/14, 'fair' if they scored 5-10/14 and 'good' if they scored 11-14/14.One study used retrospective analysis of existing health data, and for this study we used the MetaQAT Critical Appraisal Tool [32].To keep the ratings consistent with our rating system for the studies appraised by NIH tools, we defined 'poor' quality as a score of 0-34%, 'fair' quality as a score of 35-72% and 'good' quality as a score of 78% or higher.

Results
Literature searches yielded 5,398 citations of which 1,083 were duplicates.After title and abstract screening, 4,210 citations were excluded leaving 105 citations for full-text screening.After reading full texts of the remaining citations, 75 were excluded and an additional five studies were added after hand-searching reference lists.A total of 35 citations were included in the review [15,.Figure 1 illustrates the screening process in a PRISMA flow diagram.
Table 1 provides an overview of key characteristics of all included studies.Studies originated from the United States of America (n = 13), Denmark (n = 5), Israel (n = 4), China (n = 4), Norway (n = 2), the United Kingdom (n = 2), Japan (n = 1), the Netherlands (n = 1), Switzerland (n = 1), and Turkey (n = 1).The remaining study included participants in multiple different countries across Europe and Asia.Study populations ranged from 35 to 7,918 and included military personnel (n = 15), civilian adults (n = 14), children or adolescents (n = 4) or a combination of military and civilian adults (n = 2).Only three studies were rated as 'good' quality; the majority were rated 'fair' .

Definitions of recurrence
Table 2 reports, for each study, the tools used to assess PTSD; time-points at which PTSD was assessed; definitions of recovery and recurrence; and whether the participants received PTSD treatment or not.

Terminology
The first aim of the review was to explore how 'recurrence' is defined in the literature.We found no consensus in terms of how this is defined.In fact, the studies used a variety of different terms to describe the emergence of new PTSD episodes after initial 'recovery' , including 'recurrence' [33,37,44,47,64,65]; 'relapse' [35,36,40,49,50,52,53,57]; 'reactivation' [15,60,62]; 'exacerbation/reactivation' [61]; 'relieved-worsening PTSD' [34,48,51,63]; 'response-remit' trajectory [54]; 'fluctuating course' [58]; 'intermittent cases' [43]; 'delayed increase in symptoms' [46]; and the 'relapsing/remitting' trajectory [42,55].Many others simply described recurrence as 'symptom increase' [38], 'initial declines followed by symptom increases' [56] or 'exacerbation of symptoms' [41,60].Some studies did not name the trajectory at all; rather, they presented tables or flow charts showing the number of participants with PTSD at each time-point, from which it was possible for us to identify a sub-group of participants who were described as having PTSD at  one time-point, not having it at least one follow-up, and then having it again at subsequent time-points [39,59].Similarly, Hansen et al. [45] identified and commented on a sub-group of participants who met the criteria for PTSD, did not meet the criteria at a subsequent timepoint, and then met the criteria again later, but they did not give this a name.

Criteria for recurrence
Several studies defined recurrence (or equivalent terminology such as relapse) as meeting diagnostic criteria for PTSD at a follow-up time-point after an initial 'recovery' period where they did not meet the cut-off for PTSD [33,35,37,39,43,45,46,58,65].Holliday et al. [47] referred to 'clinically meaningful change in PTSD symptoms' , which was also assumed to refer to clinical cut-off scores.Markowitz et al. [52] based the definition of relapse on similarity to baseline scores.Sungur and Kaya [64] defined recovery and recurrence as being asymptomatic and then symptomatic again, but it is not clear whether this referred to clinical cut-offs.One study defined 'reactivation' of PTSD as meeting full diagnostic criteria or being a sub-syndromal case [15].Others were more vague and did not mention cut-offs, instead referring to dramatic or steep symptom increases [34,38,56,63], fluctuating symptoms [42,55], returning to pre-treatment levels of PTSD [54], symptoms which 'decreased somewhat and increased drastically' [48], symptoms which 'decreased to a low level and increased again' [49,50] or 'steadily worsening' symptoms [36].DenVelde et al. [41] simply asked participants to self-report whether they had 'experienced remissions and exacerbations' .Martenyi et al. [53] had multiple definitions of relapse, including increases in scores on their PTSD measures or 'the clinical judgement of the investigator' .Others labelled the trajectory but did not specify the parameters of their definitions [51,[60][61][62]66].One study [57] used 'being hospitalised' as a proxy measure of PTSD recurrence, although this way of defining recurrence would obviously not capture individuals who developed recurring symptoms which were not severe enough to warrant hospitalisation; additionally, no criteria for hospitalisation were described.Similarly, Davidson et al. [40] described 'relapse' as PTSD scores reverting back to baseline or worse, or experiencing an 'untoward clinical event' including suicidality, hospitalisation, or dropping out of the study due to feeling progress was not being made.We found little consensus as to how long participants needed to be symptom-free (or have reduced symptoms) in order to be considered 'recovered' prior to recurrence.The majority of studies simply based their definitions on the time-points of the study, suggesting that recurrence was identified if participants had PTSD at baseline, did not have PTSD during at least one follow-up, and then had PTSD again at a later follow-up.The time-points of follow-ups ranged from weeks to months to years.Only four studies suggested specific timeframes: three studies claimed that participants needed to be 'recovered' for eight weeks in order for later reports of PTSD to count as 'recurrence' rather than symptom fluctuation [35,37,66] whereas Zanarini et al. [65] reported that participants needed to be not meeting the PTSD criteria for at least two years in order to be considered 'recovered' .Similarly, most studies did not clarify a time-scale for how long symptoms needed to be experienced in order to be considered a 'recurrence' .Most studies again simply based their diagnosis on the scores participants happened to report on the days they were assessed.Few studies specified a time-frame: three [35,43,65] suggested a duration of four consecutive weeks of meeting their criteria for PTSD, while Benítez et al. [37] suggested two weeks of symptoms was sufficient to identify a recurrent episode.

Prevalence of recurrence
The review's second aim was to explore PTSD recurrence rates.Table 3 presents data on the prevalence of recurrence of PTSD for each study.The second column of Table 3 presents the data that is reported in the original studies.The findings reported in this column are not easily comparable because studies reported recurrence rates in different ways.Some reported the percentage of the entire trauma-exposed sample who experienced PTSD recurrence (column 3 of Table 3).Others reported the percentage of those with PTSD who experienced recurrence (column 4 of Table 3) and the remaining studies reported the percentage of those who recovered from PTSD who experienced recurrence (column 5 of Table 3).Three studies [44,47,57] did not report the prevalence of recurrence, but were still included in the review as they included definitions and/or predictors of recurrence.One study [60] deliberately chose a sample who had all experienced recurrence; therefore, recurrence prevalence data for this study was not recorded in Table 3 as it would, by design, be 100%.
Most studies (19/35) reported the prevalence of recurrence within the entire trauma-exposed population.We would therefore expect prevalence rates to be extremely small, given that the majority of trauma-exposed people will not develop PTSD in the first place [3], let alone have recurrent episodes.However, in several studies this was not the case.Prevalence of recurrence ranged from 0.2% (for a sub-set of participants who did not directly witness the disaster in question) [45] to 57% of 63 women newlydiagnosed with ovarian cancer [43].The latter study was carried out over 27 weeks and identified 'intermittent cases' who had PTSD at one time-point, no PTSD at a later time-point, and then PTSD again later on.We note that 27 weeks is a fairly short period of time for both     recovery and recurrence to occur, and it is therefore possible that the data reflects symptom fluctuations rather than true recovery or recurrence.Overall, the mean prevalence of recurrent PTSD in trauma-exposed populations was 13.1%, and the median was 3.8%.

Definition of recurrence
Five studies presented the prevalence of recurrence within populations diagnosed with PTSD.We would expect these prevalence rates to be higher than the prevalence rates of recurrence within full trauma-exposed samples, as they are based on populations who developed PTSD only.The rates were 4.9% [39], 15.4% [66], 24.5% [36], 28% [46] and 49.6% [41].Mean and median prevalence of recurrent PTSD were both 24.5%.
Seven studies presented data on the prevalence of recurrence within sub-sets of study populations who had recovered from PTSD; therefore, the only possible trajectories for these participants would be recurrence or maintenance of recovery.Recurrence rates ranged from 5.8% (for a sub-set of participants treated with fluoxetine) [53] to 50% (for a sub-group treated with a placebo) [40].Mean prevalence of recurrent PTSD was 25.4% and the median was 22.2%.
The three studies rated highest in quality [34,47,55] did not report similar findings relating to prevalence.Holliday et al. [47] did not present prevalence data at all.Andersen et al. [34] reported that 2% of participants followed the 'relieved-worsening' trajectory, whereas Osenbach et al. [55] reported that 35% of participants followed the 'relapsing-remitting' trajectory.Notably, Andersen et al. 's [34] participants were military personnel, whilst Osenbach et al. 's [55] participants were civilian trauma survivors.For this reason, we decided to look separately at recurrence rates in military and civilian participants.We also decided to look separately at data on children as children's experiences during and after potentially traumatic events are likely to be distinct from those of adults [67].Table 4 presents the mean and median recurrence rates for different populations.

Prevalence of PTSD recurrence in military populations
Fifteen studies focused on military personnel and veterans, three of which did not provide prevalence data and one of which included only participants with PTSD recurrence.Military studies which presented rates of recurrence in trauma-exposed populations (rather than focusing on people diagnosed with PTSD only) typically found low prevalence of recurrence: seven studies found prevalence rates under 4% [34,48,51,54,61,62].Another study found a prevalence rate of 6% [38].The only higher prevalence rates were reported by Solomon & Mikulincer [59], who reported recurrence rates of 24.4% for those with combat stress reactions (people referred for psychiatric intervention during the war) and 13.2% for participants who participated in combat in the same units but without need for psychiatric intervention during the war.This study assessed participants over twenty years, which may explain its higher prevalence rate than the majority of studies which were completed within two-and-a-half years or less.However, the study period was shorter than the forty-seven years of Solomon et al. 's [62] study, which reported only a 1.6% rate of recurrence.It is unclear why Solomon and Mikulincer [59] found much higher rates of recurrence.
Two military studies reported recurrence rates for PTSD-populations.These were 24.5% [36] and 49.6% [41].We note that all of Armenta et al. 's [36] participants had comorbid depression at baseline.We also note some concerns about the reliability of DenVelde et al. 's study [41], which was a retrospective study asking participants to give complete life-history data at one time-point only.
One military study reported on the prevalence of recurrence in a sub-group of participants who had recovered.Solomon et al. [62], who reported a prevalence rate of 1.6% (out of the entire trauma-exposed sample) over the first forty-two years of the study, found in a follow-up at forty-seven years that 16.7% of those who had initially recovered experienced recurrence of PTSD during the COVID-19 pandemic.

Prevalence of PTSD recurrence in civilian adult populations
Fourteen studies focused on civilian adults.Findings relating to recurrence prevalence in entire traumaexposed samples were mixed.Two studies reported rates of under 5% [45,58] in survivors of a terrorist attack and an earthquake respectively.Sungur and Kaya [64] reported a recurrence rate of 8.9% in survivors of the Sivas disaster, a religious fundamentalist protest which resulted in civilian deaths.Higher rates of recurrence were reported for survivors of an oil rig disaster (18.8%) [15], survivors of an oil spill (32%) [56], acutely injured trauma survivors (35%) [55] and women recently diagnosed with ovarian cancer (57%) [43].

Prevalence of PTSD recurrence in combined military and civilian populations
Finally, two studies included both military and civilian participants; both of these studies were trials comparing fluoxetine to placebo treatment in people with PTSD.Davidson et al. [40] found that half of the placebo group relapsed after recovery, compared to 22.2% of the fluoxetine group.Martenyi et al. [53] reported lower rates of 'relapse': 16.1% of the placebo group and 5.8% of the fluoxetine group.The latter study followed up participants after 36 months, while Davidson et al. [40] followed up participants for a year after treatment.

Predictors of PTSD recurrence
The third and final aim of the present review was to identify factors associated with PTSD recurrence.Firstly, we note that (as shown in Table 2), participants in a number of studies had received some type of intervention during the study period, which was typically not accounted for in analyses of predictors.Many other studies did not report whether participants received treatment or not.
Having treatment, whether it be medication, therapy, or a combination, is likely to be an important factor influencing PTSD trajectory, given that there are evidence-based treatments for the condition [68], but this was typically not explored.Table 5 shows the factors considered as predictors in each study, with significant associations presented in bold.The majority of included studies (22/35) explored at least one covariate; the remaining studies either did not explore covariates or combined recurrent trajectories with other trajectories in their analyses of predictors.Of those studies which did explore covariates of recurrence, we found little consensus.

Sociodemographic factors
Gender was considered as a potential covariate by six studies; one [33] found that recurrent PTSD was associated with female gender while five studies (including two based on the same data-set) [49,50] found no significant gender association [35,36,42,49,50].None of the three studies testing age as a covariate found a significant association [35,36,57].One study of school-aged children found that children in a higher grade (i.e.older in age) were more likely to experience PTSD recurrence [33], while three studies of two cohorts [42,49,50] found no significant association between recurrence and school grade.Three studies considered race as a covariate, finding no significant association between PTSD recurrence and race [36,44,55] [65] Recurrence (defined as meeting the PTSD criteria for one month after a two-year period of remission) occurred in 40% (n = 30) of participants with borderline personality disorder and PTSD at baseline who had experienced a remission (25% by the 4-year follow-up, 36% after both six and eight years, and 40% by the 10-year follow-up).Recurrence also occurred in 6/15 axis II comparison participants who met criteria for PTSD at baseline and had a remission during the follow-up years.

Table 3 (continued)
characteristics considered included number of children in the family [42], marital status and level of education [36], none of which were found to be associated with PTSD recurrence.For military participants, there were no significant differences in service branch, service component or pay grade between the recurrent and rapid recovery groups [36].

Psychiatric history
Seven studies considered psychiatric history and concurrent diagnoses as potential covariates of PTSD recurrence, again with mixed findings.Recurrence was not found to be associated with other anxiety syndromes [36], baseline levels of anxiety [54], depressive symptoms [55], baseline levels of depression [54] or psychiatric history [55].Ansell et al. [35] found that diagnoses of a number of co-morbid mental health disorders such as major depressive disorder and personality disorders such as schizotypal personality disorder, avoidant personality disorder and borderline personality disorder were not associated with recurrence, but participants with a baseline diagnosis of obsessive-compulsive personality disorder were significantly less likely to experience PTSD recurrence.Conversely, Perconte et al. [57] found that those who experienced recurrence were significantly more likely to report obsessive-compulsive symptoms than those whose symptoms improved without recurrence.Sakuma et al. [58] found that pre-disaster treatment for mental illness was significantly associated with PTSD recurrence, but note that the results should be interpreted carefully due to the very small number of participants in the 'fluctuating symptoms' group who appeared to have experienced recurrent episodes.Perconte et al. [57] found that, versus the improved symptoms group, those with PTSD recurrence were more likely to report depression, anxiety, hostility, phobic anxiety, somaticism and psychoticism; however, previous psychiatric hospitalisations and pre-treatment ratings of global pathology on a psychiatric scale did not predict recurrence.Finally, Madsen et al. [51] found that suicidal ideation was significantly higher in the 'relieved-worsening PTSD' group than the 'low-stable' group and that suicidal ideation was in fact highest in the recurrent (termed 'relieved-worsening') group than any other.However, it should be noted that suicidality was not assessed at baseline in this study, therefore it is not clear whether suicidal ideation is a cause or a consequence of PTSD recurrence.

Physical health
Fewer studies considered physical health as a potential predictor of PTSD recurrence.One study found no association between recurrence and disabling injury/illness, somatic symptoms or bodily pain [36] and another found no association between recurrence and prior treatment for physical illness [57].However, obesity was a significant predictor of PTSD recurrence [36].In terms of health-related behaviours, Armenta et al. [36] found no association between PTSD recurrence and smoking status, alcohol problems or sleep duration.However, Perconte et al. [57] found that higher weekly alcohol intake both before and at termination of PTSD treatment predicted recurrence.

Cognitive ability
Only one study [63] explored cognitive ability as a potential covariate, finding that the participants who were in the recurrent (termed 'relieved-worsening PTSD') group had significantly lower cognitive ability scores than those in the 'low-stable' group.

Trauma history and pre-trauma experiences
The review also found mixed evidence for trauma history as a predictor of PTSD recurrence.Liang et al. [49,50] found no association between pre-disaster traumatic experience and PTSD recurrence.Armenta et al. [36] found no association between recurrence and childhood sexual abuse, childhood verbal abuse, childhood neglect, sexual assault, physical assault, or 'other life events' , but did find that participants reporting a history of childhood physical abuse were significantly more likely to experience PTSD recurrence.Holliday et al. [47] found that veterans who had experienced military sexual trauma (MST) had greater initial reductions in PTSD symptoms than those who had not experienced MST, but also N/A -the three groups who experienced 'benefits' of deployment in terms of reduction of PTSD symptoms were grouped together in regression analysis

Boe et al. (2010) [15]
Number of intrusion symptoms 5.5 months after the disaster, number of avoidance symptoms 5.5 months after the disaster, number of intrusion symptoms 14 months after the disaster, number of avoidance symptoms 14 months after the disaster, number of intrusion symptoms 5 years after the disaster, number of avoidance symptoms 5 years after the disaster When each residual symptom was considered separately, effect sizes were found for sleeping difficulties due to images or thoughts, bad dreams about the event, staying away from reminders, trying not think about the event (at 14 months) and waves of strong feelings, things that made them think of the event, thinking about the event when they didn't mean to, images popping into their mind, sleeping difficulties due to images or thoughts, being aware of feelings but not dealing with them (at 5 years)

Chopra et al. (2014) [39]
'Recurrent' groups were too small to feature in analysis of associated factors  2015) [42] Compared to the recovery group, relapsing participants experienced fewer negative life events at 6-months post-earthquake but more such events at 24 months, received less social support at 24 months No significant differences in gender, school grade, number of children in family, having a family member injured/killed/missing, house damage, property loss, directly witnessing the disaster, social support at 6 months, positive coping at 6 months, negative coping at 6 months

Gonçalves et al. (2011) [43]
N/A -intermittent and persistent cases combined Gross et al. (2022) [44] Although race was significantly associated with treatment response (with Black veterans experiencing an attenuated response over the course of the PTSD programme) and Black veterans had significantly greater PTSD symptoms at discharge, there were no significant differences between Black and White participants in terms of symptom recurrence Veterans who experienced military sexual trauma had 'modestly greater recurrence of symptoms' but this does not appear to be significant

Karstoft et al. (2015) [48]
Poor adjustment to civilian life (i.e.difficulties with community reintegration after deployment) was significantly higher for the relieved-worsening group than for all other groups Murphy & Smith (2018) [54] Response-remit group vs. resistant group: baseline depression; baseline anxiety; magnitude of reexperiencing symptoms; magnitude of avoidance symptoms; magnitude of hyperarousal symptoms; combat exposure  experienced a 'modestly greater' recurrence of symptoms than those without MST, although this difference did not appear to reach statistical significance.Zanarini et al. [65] found that the presence of childhood sexual abuse history did not significantly predict time-to-recurrence, but severity of childhood sexual abuse, adult rape history, combination of childhood sexual abuse history and adult rape history, and experiencing sexual assault during study follow-up were associated with less timeto-recurrence.Osofsky et al. [56] found that abuse, emotional abuse, domestic violence, and greater number of traumas experienced were associated with recurrence of PTSD, and Osenbach et al. [55] found that recurrent life stressors significantly increased the odds of membership in chronic, relapsing or recovery groups rather than the resilient group.For military participants, one study found combat deployment was significantly associated with recurrent PTSD [36] while others found combat exposure was not associated with recurrence [54,57].Finally, Fan et al. [42] found that compared to the recovery group, relapsing participants experienced significantly fewer negative life events 6-months post-disaster, but significantly more such events at the 24-month follow-up.Few other pre-trauma experiences were considered.An et al. [33] found that those with recurrent PTSD were significantly more likely to have experienced academic burnout than those in the recovery trajectory, although there was no difference between the recurrent and delayed trajectories.

Experiences during and immediately after the traumatic experience
The review also found mixed evidence for an association between peri-traumatic experiences and PTSD recurrence.The most consistent finding related to how stressful the traumatic experience was perceived to be at the time.For example, risk of recurrence was significantly higher in those with combat stress reactions [59] and in those with higher stress relating to the disaster they had experienced [56], as well as with greater trauma severity [49,50].However, recurrence was not found to be associated with subjective fear during the event [33]; directly witnessing a disaster [42]; property loss during the event [33,42]; property damage [42]; displacement due to property damage [58]; near-death experience [58]; or having a family member injured, killed or missing [42,58].
There was some evidence that initial post-traumatic stress symptoms immediately after the traumatic event could predict PTSD trajectory.Liang et al. [49,50], in a study of PTSD in children from two schools affected by an earthquake, found that children from one of the two schools ('School 2') were significantly more likely to experience PTSD recurrence than children from the other school ('School 1').Further investigations revealed that after adjusting for immediate post-traumatic stress symptoms the school no longer predicted relapse; those from School 2 had significantly greater post-traumatic stress symptoms immediately after the disaster, which the authors suggest might be due to School 1 providing sufficient psychological services as well as having the same students and teachers before and after the earthquake (therefore perhaps greater social support available), whereas School 2 had insufficient psychological services and consisted of teachers and students from several different schools which could not be reconstructed after the earthquake.
One study [58] considered occupational-related covariates of PTSD recurrence for disaster recovery workers.They found that having mainly disaster-related occupational duties and lack of rest due to occupational duties were not associated with recurrence, but perceived poor workplace communication did predict recurrence.

Post-trauma experiences and symptoms
An et al. [33] found that, compared to the delayed PTSD trajectory, those who experienced recurrence were less likely to have experienced post-traumatic growth after the traumatic event; however, there were no differences in post-traumatic growth between the recurrent and recovery groups.Fan et al. [42] found that neither positive coping nor negative coping six months post-disaster were associated with PTSD recurrence.In a military study, Karstoft et al. [48] found that poor adjustment to civilian life (i.e.difficulties with community reintegration after deployment) was significantly higher for the recurrent ('relieved-worsening PTSD') group than all other groups.However, it is not clear whether poor adjustment was a cause or an effect of PTSD symptoms worsening after initial improvement.
Two studies explored specific cluster symptoms.Murphy and Smith [54] found PTSD recurrence was not predicted by the magnitude of re-experiencing, avoidance, or hyperarousal symptoms.Boe et al. [15] found that the number of intrusion and avoidance symptoms five-anda-half months post-trauma did not predict recurrence, but the number of intrusion and avoidance symptoms both fourteen months and five years after the disaster did predict recurrence.

Social support
Only three studies directly considered social support as a potential covariate.Armenta et al. [36] found no association between social support and PTSD recurrence, and Perconte et al. [57] found that family support did not predict recurrence.Fan et al. [42] found that level of social support six months after experiencing an earthquake was not associated with PTSD recurrence, but those in the 'relapsing' group reported significantly less social support 24 months after the earthquake than those in the 'recovery' group.

PTSD treatment
Most of the studies investigating treatment for PTSD found that not receiving interventions, or discontinuing treatment, were associated with PTSD recurrence.For example, Osenbach et al. [55] found that those who received 'usual care' only were significantly more likely to experience recurrence than those who received interventions designed to reduce post-traumatic symptoms.Davidson et al. [40] found that those who received placebo treatment were significantly more likely to experience recurrence than those who received fluoxetine.Martenyi et al. [53] found that those who discontinued fluoxetine treatment were significantly more likely to experience recurrence, especially for those with combatrelated PTSD.However, Perconte et al. [57] found that number of weeks enrolled in treatment and number of treatment sessions attended did not significantly affect risk of recurrence.In this study, though, being hospitalised at least once since the termination of treatment was used as a proxy measure of 'recurrence' and so the findings are arguably not truly representative of actual recurrent episodes of PTSD.Overall, our findings indicated some evidence that treatment helped to avoid recurrent episodes.

Discussion
In this study, we systematically reviewed 35 studies to identify definitions and prevalence of recurrent PTSD and factors associated with recurrence.It is important to define and operationalise recurrence as the concept needs to be understood in order to make prevention efforts.The health-related, social and economic costs of PTSD can be substantial.PTSD negatively affects individuals' emotional wellbeing and physical health [7], impedes social relationships [69], limits productivity at work and increases sickness absence [70].The direct costs (e.g., medical care costs) and indirect costs (e.g., costs of unemployment or reduced productivity) of PTSD can create substantial economic burden [7,71].Determining the predictors of recurrence of PTSD (which can only be properly understood if 'recurrence' itself has a clear definition) is important for prevention efforts: identifying those most at risk for recurrent episodes would allow for the subsequent investigation of ways of mitigating or preventing the risk.However, we found little consensus as to how recurrence is defined, mixed evidence on the prevalence of recurrence and inconsistent findings relating to predictors of recurrence.This lack of clarity about what relapse or recurrence is, and is not, is a major barrier to understanding this important topic.
In a previous review exploring PTSD recurrence in veterans, Berge et al. [22] acknowledge that there is no generally accepted or used definition of recovery relating to psychological trauma.The definition of recurrence used in their review was the return of symptoms following a period of complete recovery, representing the start of a new and separate episode.However, it is not clear what length of time is covered by 'a period of complete recovery' nor what 'complete recovery' means.How many days, weeks, or months does an individual need to be free of symptoms of PTSD in order to be considered truly recovered?Is 'symptom-free' the only definition of recovery, or is 'not meeting the criteria for PTSD' enough?Our own review revealed that there is little consensus as to what recurrence means and the parameters for its definition.Even the terminology used varied across studies, with 'relapse' , 'recurrence' , 'reactivation' and numerous other terms often used to describe what essentially appeared to be the same concept.There was no consensus as to how long an individual needed to be free of symptoms in order to be considered recovered, nor for how long symptoms needed to recur in order to be considered a recurrent episode.Most studies simply defined recurrence as a change in symptoms between assessments, meaning that whether or not an individual was defined as having a recurrent episode or not very much depended on the scores they reported at arbitrary time-points.Even minor symptom fluctuations could cause someone to change from being identified as a 'case' to 'recovered' and vice versa.Because PTSD tended to be examined using prospective studies where symptoms were assessed at predetermined assessment points, it is possible that individuals may have onsets of PTSD after one assessment and then remit before the next.With no retrospective assessment between time-points, it is difficult to assess the true prevalence of recurrence.Andrews et al. [16] make a similar point in relation to delayed onset PTSD, suggesting the absence of information about symptoms outside of the predetermined time-points of studies means that estimates of delayed onset PTSD may be unreliable.
The second aim of the review was to examine the prevalence of PTSD recurrence in existing literature.Given the numerous different ways of assessing PTSD, defining initial recovery and defining recurrence, as well as the differing time-points at which PTSD was assessed across studies, we suggest that the current data on recurrence prevalence is not especially meaningful.We found very different prevalence rates reported within the literature, with data suggesting that anywhere between 0.2% and 57% of trauma-exposed populations might experience recurrent episodes of PTSD.Some of the higher percentages we found seem greater than we would expect, given that only a minority of trauma-exposed people are likely to develop PTSD in the first place -let alone suffer from it, recover from it, and experience a recurrent episode.We would expect that studies carried out over a longer period of time would find higher recurrence rates, simply because in these studies there is more time for recurrent episodes to occur.However, the highest prevalence rate (57%) was found in a study which took place over only 27 weeks [43]; the authors labelled these participants as 'intermittent cases' and it appears likely that symptom fluctuation, rather than true recovery and recurrence, occurred in this study -and potentially many others.Additionally, studies did not typically control for exposure to subsequent trauma, meaning that 'recurrences' of PTSD identified may actually be new episodes, rather than a relapse.Further research studies, especially research involving assessments over a number of years, are needed to establish the true prevalence of recurrent PTSD which also needs to be clearly defined with an agreed time period between remission and relapse.
It has been proposed that recurrence rates might increase with old age.Murray [72] suggests that PTSD can be 'reactivated' in older age because physical illnesses become more common, which can reactivate traumatic memories; increased dependence on others due to ageing can reactivate feelings of helplessness; and loss of structure and identity caused by retirement can similarly reactivate traumatic symptoms.Other factors relating to ageing such as decline of cognitive function, difficulty controlling ruminations, reminiscing, and latelife stressors such as serious illness, surgical procedures and death of spouses, siblings or close friends can either directly remind the person of their previous traumatic experience(s) or can induce similar feelings of vulnerability [73].Three studies of adults in this review did not find age predicted recurrence [35,36,57]; however, the populations trended young overall, with each of the three studies reporting the mean age of participants was under 40.We suggest, then, that more studies of older adults with lifetime PTSD are needed to establish whether this group are at increased risk of recurrence.
The third aim of this review was to understand factors associated with PTSD recurrence.Although a number of potential covariates were considered, most were not investigated by more than a few studies, and findings were varied and inconsistent.Of the covariates investigated by multiple studies, none were found to have significant associations with recurrence across all studies.It was therefore not possible to quantify the extent to which potential risk factors contribute to the risk of recurrence.One reason for the inconsistent findings might be the relatively small numbers of participants with recurrent PTSD in many of the studies.We note also that most studies did not consider either subsequent trauma or treatment impact in their analysis of predictors of recurrence.
We did not find strong evidence of an association between PTSD recurrence and comorbid psychiatric conditions.Recurrence of other mental health disorders, such as anxiety, is reportedly associated with comorbid psychiatric conditions including major depression, alcohol and substance use disorders [74].Additionally, comorbid disorders have been found to be associated with an 'unfavourable long-term course' of PTSD [18].However, in a review of predictors of developing PTSD, Brewin et al. [75] found that while psychiatric history was associated with development of PTSD, it was not a strong risk factor -factors operating during or after the traumatic exposure had greater effects than the pretrauma factors.Many studies in this review found no evidence of a relationship between PTSD recurrence and other mental health conditions; in those that did find a relationship, it was not always clear whether the other conditions pre-dated the recurrent PTSD episode or not.Overall, the most consistent evidence we found indicated that recurrence of PTSD was associated with greater stress and traumatic response at the time of the traumatic experience.
We did not find evidence to suggest that trauma type may affect recurrence.Many studies examined PTSD trajectories after a single traumatic event.Those that did include participants who had experienced various different types of trauma did not consider trauma type as a potential predictor of recurrence.Given the wide variations in methodology, it was not appropriate for us to compare recurrence rates for different trauma types within the review.Future research should include participants who have experienced different types of trauma and should consider trauma type as a potential predictor of PTSD trajectory.
Only one study assessed PTSD during the COVID-19 pandemic, with Solomon et al. [62] reporting that 16.7% of initially-recovered participants experienced recurrence during the pandemic.However, it is not clear how many of this cohort may also have experienced recurrence before the pandemic, and without being able to make that comparison, we cannot ascertain the extent to which recurrence was exacerbated by the pandemic.Additionally, the percentage (16.7%) is similar to recurrence rates in several other, non-COVID studies.Ideally, future studies will present data on PTSD recurrence rates for one cohort at regular intervals, including data collected during or after the COVID-19 pandemic, to ascertain whether the pandemic did affect recurrence rates.
In their review, Steinert et al. [18] identified older age, higher education, greater trauma severity, higher baseline symptoms, more physical/functional impairments, and poorer social support as predictors of 'unfavourable' long-term course of PTSD.These were identified as predictors due to being reported in at least two studies within their review.The current review did not find consistent evidence that age, education, trauma severity, baseline symptoms, impairments or social support predicted recurrence -although age was only considered in studies of young people.We found some evidence from treatment studies that fluoxetine reduced the risk of recurrence, as did participation in an intervention involving a combination of motivational interviewing, behavioural activation and pharmacotherapy.It is therefore difficult to make recommendations relevant to occupational health, as we had hoped to do.Managers of trauma-exposed employees who have developed PTSD may have questions around whether recovered individuals can go back to frontline work, or whether they risk experiencing a recurrence of PTSD.Our findings tentatively suggest that recurrence might be relatively rare (rates of recurrence ranged from 0.2 − 57% in full trauma-exposed samples, mean 13.1%; 4.9 − 49.6% in PTSD-only subgroups, mean 24.5%; and 5.8 − 50% for recovered subgroups, mean 25.4%) but clearer definitions and assessments of recurrence are needed to substantiate that claim.As we found no consistent evidence of predictors of recurrence, it was therefore not possible to identify which sub-groups of people might be more likely to have their PTSD recur.We did find evidence from two studies that recurrence was more prevalent in groups of PTSD patients treated with placebos compared to PTSD patients treated with fluoxetine, suggesting that medication appears at least somewhat effective in reducing the risk of recurrence.However, we found no studies looking at the impact of first-line treatments on relapse (i.e.trauma-focused cognitive behavioural therapy [76] or eye movement desensitisation and reprocessing [77,78]) which is a major gap in the literature.Whilst more, highquality studies are carried out, employers should ensure that workers get evidence-based treatments and have an occupational mental health assessment on completion of potentially traumatic work to provide an expert judgement, given that we cannot identify any clear risk factors from the literature.
The key limitation of the literature on PTSD recurrence is that it is not always easy to differentiate between recurrence and symptom fluctuation, and it is also difficult to know what 'recovery' truly means.It is not clear how many of the so-called 'recovered' participants within the reviewed studies may have been close to clinical thresholds for PTSD at the assessment points.Rather than moving from distinct 'recovered' to 'recurrent episodes' , it may be that individuals only experienced small fluctuations in PTSD symptoms, moving them above and below the symptom thresholds.Indeed, the authors of several of the included studies remarked on the difficulties in identifying PTSD trajectories.In Boe et al. 's [15] study, clinical interviews were conducted by two clinical psychologists who were trained and supervised by an experienced clinician and trauma researcher and even these experienced individuals had difficulties identifying recurrence of PTSD, with one case being recategorised from 'fullblown PTSD reactivation' to 'sub-syndromal reactivation' after discussion between the researchers.Markowitz et al. [52] pointed out that, as they defined relapse as 'loss of response (to treatment) status' , relapse might reflect barely crossing that threshold: indeed, more in-depth analysis of their six 'relapsers' showed that all but one still showed some, albeit more modest, treatment benefit relative to their baseline PTSD severity.
Sakuma et al. [58] discussed their finding of a 'fluctuating' trajectory (and lack of a delayed-onset trajectory), differing from the typical four trajectories widely accepted within the PTSD literature.They suggested the difference may be due to variations in the duration of study periods and characteristics of the study samples.The majority of studies which produce the typical four trajectories are conducted over short periods between a few months and two years [9], compared to the longer (54-month) period of Sakuma et al. 's [58] study: the trajectory commonly identified as 'delayed onset' could really be a fluctuating trajectory if examined over a longer period.Or, it could reflect a gradual accumulation of symptoms resulting in a delayed presentation of PTSD, rather than delayed onset.
The time-points of assessments could also affect reported prevalence rates.For example, Sungur & Kaya [64] pointed out that some of their 'recurrent' cases would have been considered 'recovered' if the study period had been shorter or if participants had not been reassessed at the particular time-points chosen.They also noted that symptoms across the entire participant population seemed to be higher at particular times during the study (namely, at the anniversary of the event and at the time of a disappointing result of a court hearing for compensation), suggesting that the nature and course of PTSD might be influenced by particular events which might trigger unwanted memories of the traumatic event.
In the current review, most studies assessed participants for at least a year, but not all: five [38,39,43,52,53] followed participants for less than a year.Additionally, two studies [44,47] reported assessing participants pretreatment and four months post-treatment but it was not clear how long treatment lasted.
We suggest that PTSD recurrence may not have been adequately assessed in many of the included studies.For example, Chopra et al. [39] described how, in order to minimise respondent burden, assessors were expected to stop inquiring about PTSD symptoms if participants were unlikely to meet the criteria and if they answered no to particular questions on the assessment tool.This could mean that some individuals who did have recurrent episodes of PTSD were not identified as they did not complete the full measures.Additionally, we found that a number of studies had very vague definitions of recurrence, such as 'increasing symptoms' , where it was unclear what exactly this meant.Others used hospitalisation as a proxy measure for recurrence, or simply asked participants whether they perceived their symptoms had been exacerbated and in one case used the investigator's own judgements as a way of determining recurrence.It is therefore likely that some recurrent cases may have been missed while others who never truly 'recovered' at all may have been reported to have experienced recurrence.Overall, the vague and inconsistent ways of assessing recurrence mean it is currently impossible to ascertain true recurrence rates within existing literature.
It is also possible that recurrent trajectories of PTSD appear in studies which do not identify them as such.For example, in Andrews et al. 's [16] review, the authors note that some cases of 'delayed-onset PTSD' in veterans of relatively old age with long intervals to first onset may in fact have had episodes of PTSD soon after their traumatic experiences which were undisclosed or forgotten.In other words, some cases of supposedly 'delayedonset' PTSD might actually be recurrent cases.Andrews et al. [16] also point out that many of the studies included in their review of delayed-onset PTSD did not assess whether respondents could have had onsets of PTSD and then remitted before the next assessment point -which could lead to both over-and under-estimates of delayedonset rates of PTSD.Indeed, the studies included in our own review tended to focus only on the scores at the various time-points and did not explore participants' perceptions of symptom fluctuations outside of the time-points set by the study.

Limitations
There are a number of limitations of the literature reviewed.Many did not collect data on whether participants had undergone any intervention or not, and those that did tended not to include this as a potential confounding variable.The majority of studies did not assess whether participants experienced additional potentially traumatic experiences between PTSD assessments.Many did not define the parameters of 'recovery' and 'recurrence' and it is not clear whether recurrent episodes identified were truly recurrent episodes or merely symptom fluctuations.Many did not collect data on whether or not participants received any treatment for PTSD between data collection time-points, and many of those which did ask participants whether they had received any treatment did not distinguish between types of treatment.It is therefore unclear if, and how many, participants in many studies received any evidence-based PTSD treatment or not.Additionally, the majority of studies did not collect data on the time period of any treatment received.Some studies had extremely long gaps (e.g., decades) between assessments which could mean that recurrences were missed.
There are also limitations of the review process itself.Firstly, the screening, data extraction and quality appraisal were carried out by one author.Although decisions about exclusion or inclusion were discussed with the second author, it would have been preferable to have multiple screeners.We limited the review to Englishlanguage studies only, meaning that important studies published in other languages would have been missed.We included only studies which identified 'recurrent episodes' (or equivalent terminology e.g.relapse, reactivation); studies which identified no recurrent trajectory were not reviewed.It may be that these studies did not include a sufficient number of assessments to pick up Fig. 2 Summary of review and suggested next steps on recurrent episodes, but it may also be that no participants in these studies experienced recurrence and therefore the true prevalence of recurrence may be lower than this review suggests.

Conclusions and implications
The main conclusion that can be drawn from the current review is that, moving forward, better clarity and consensus regarding the definition and identification of recurrent PTSD are urgently needed.Berge et al. [22] suggest that consistent definitions of relapse-related terms, supported by empirical research, are required in order to make studies of PTSD trajectories more robust.The findings of this review support this suggestion.Experts in the field should agree on an appropriate definition of recurrence (i.e.symptom-based or threshold-based) and should agree how long an individual needs to be 'better' for in order to be considered recovered as well as how long an individual needs to experience symptoms for in order to be considered as having a recurrent episode.Recurrence is arguably better-defined for recurrent depressive disorder, with the ICD-11 stating that recurrence is characterised by a history of depressive episodes separated by at least several months without significant mood disturbance [79].However, further clarity is still needed.How many months is 'several'?What are 'significant' symptoms?Still, we suggest this might be a useful starting point for a working definition of recurrent PTSD: a history of episodes of PTSD separated by at least several (i.e., three) months without significant (i.e., meeting diagnostic criteria) PTSD symptoms.However, further research is necessary to clarify whether these parameters (i.e. three months as a time period, symptom thresholds as a diagnostic tool) are the most appropriate to use.Using consistent terminology within the literature would make it easier to researchers in the future to understand true prevalence rates of PTSD recurrence and to compare them across studies.Further research allowing for the identification of recurrent PTSD episodes is needed.We believe the gold standard for assessing PTSD and properly identifying its trajectories, including recurrent trajectories, would be using the Clinician Administered PTSD Scale (CAPS) [80], or other validated questionnaires, at multiple specific time points over a long period of time.Figure 2 summarises the findings of the review and the proposed next steps based on our findings.
It is important to understand recurrence in order to take steps towards reducing the risk of PTSD recurring.However, due to the inconsistent findings relating to predictors of recurrence, it is difficult to draw conclusions about the best ways of preventing or minimising recurrence.We suggest that ensuring that people who develop PTSD are provided with timely, evidence-based treatments is a logical first step [68].Second, awareness of 'early warning sign' symptoms and 'triggers' might be useful, as well as awareness of effective coping strategies and how to access support.That is, if people with PTSD are able to recognise when they are struggling more and acknowledge that they need to be proactive in ensuring symptoms do not develop into full-blown PTSD again, they may be able to draw on their coping skills or reach out for formal or informal support when a recurrent episode seems imminent and may be able to stave off the recurrent episode.We also suggest that reframing the re-emergence of symptoms in a more positive way might be useful: instead of feeling defeated that symptoms have recurred, people could remind themselves that they have recovered once and therefore know that they are capable of doing so again.Within organisational settings, it is also important to foster an environment in which people who have any mental health condition, including PTSD, feel confident that asking for help will not lead to stigmatisation or increase the likelihood of inappropriate job loss.It may also be helpful to incorporate relapse prevention, understanding 'warning signs' of recurrent episodes and positive reframing into PTSD treatment programmes.

Table 1
Characteristics of included studies

Table 1 (continued) Authors (year) Country Study population Traumatic exposure Inclusion / exclusion criteria Sampling method Age and gender of participants (at base- line unless otherwise stated) Qual- ity
MilitaryInclusion: formal diagnosis of PTSD; admitted to a residential intervention; completed at least one full day of paid employment within the UK Armed Forces; if they were taking psychiatric medication at assessment, they had to remain on a stable dose Exclusion: Evidence of significant neurological impairment that would affect ability to engage in therapy; actively psychotic; actively dependent on alcohol or drugs; actively suicidal

Table 1 (
continued) Inclusion: Israeli soldiers who had fought in frontline battles in the Lebanon War; combat stress reaction participants had to have been identified as psychiatric casualties by mental health personnel and had a referral for psychiatric intervention during the war and diagnosis of combat stress reaction; comparison group needed to have participated in combat in the same units but not identified as suffering combat stress reaction Exclusion: Indication of other psychiatric disorders

Table 1 (
continued) Inclusion: Ex-Prisoners of War of the Yom Kippur War or comparable veterans matched on military background and sociodemographic variables who were not taken captive Exclusion: Physical or mental health deteriorated to the point where they could not take part aged 18-35; known or estimated IQ of 71 or higher; fluent in English Exclusion: History of, or current, symptoms of schizophrenia, schizoaffective disorder, bipolar I disorder or an organic condition that could cause psychiatric symptoms

Table 2
PTSD assessments and definitions of recovery and recurrence

Data relating to prevalence of recurrent PTSD, as presented in the studies Prevalence of recurrence in entire trauma-exposed sample (including those who never developed PTSD)
found that 17.7% of 301 earthquake survivors experienced the 'relapsing'

Table 3
Prevalence of recurrence

Data relating to prevalence of recurrent PTSD, as presented in the studies Prevalence of recurrence in entire trauma-exposed sample (including those who never developed PTSD)
. Other socio-demographic

Table 4
Mean and median recurrence rates by population

higher weekly alcohol intake both before and at termination of treatment; number
of treatment sessions attended;

pre-disaster treatment for mental illness
Baseline predictors of time-to-recurrence among patients with borderline personality disorder: presence of history of childhood sexual abuse, severity