Published on 16 September 2026
PTSD and complex PTSD: what do 54 systematic reviews say about treatment?
In brief
Two psychiatrists at University College London gathered 54 systematic reviews and meta-analyses on the treatment of post-traumatic stress disorder and its complex form, selected from 298 articles, and brought them together in a narrative synthesis. The main message will surprise no one: trauma-focused psychotherapies, trauma-focused cognitive behavioural therapy and EMDR, remain first-line, and they show the strongest effect sizes across the reviews gathered. Those magnitudes are qualitative: the publication gives no numerical effect size, pools no results, and applies neither a tool to appraise the quality of the reviews nor a grading of confidence in the evidence. On newer interventions, the picture is less clear-cut than one reads elsewhere. The five reviews on MDMA report moderate to large effects, with certainty judged very low, problems with blinding, publication bias and notable adverse effects; the US Food and Drug Administration declined the application in 2024. For ketamine, four reviews out of five report significant reductions, most often measured twenty-four hours after administration. For cannabis, three reviews suggest a reduction in symptoms, on limited evidence of safety and efficacy. The most useful figure in the consulting room may lie elsewhere: dropout during treatment, 16% in one meta-analysis of 115 randomised trials, 20.9% in another covering 85 trials. Two limitations weigh on how far these conclusions travel: reviews centred on military personnel and on refugees are excluded, and the search covered only three databases, PubMed, Medline and the Cochrane Library.
The context
The question reaches the consulting room in two forms. The first comes from the clinician: a patient presents with an established post-traumatic picture, often after repeated violence in childhood, with the disturbances in self and relationships that ICD-11 grouped under the term complex post-traumatic stress disorder. Should this patient be treated like the others, or offered something else? The second comes from the patient, who has read somewhere that MDMA cures trauma, or who asks whether ketamine, which they have heard about for depression, might help.
The UK guidelines from the National Institute for Health and Care Excellence date from 2018 and rest on the data available at that time. Since then, the literature has mostly produced systematic reviews, sometimes of the same trials, with conclusions that do not always agree. A piece of work that takes these reviews as its unit of analysis therefore answers a real need: knowing what converges and what remains contested, without rereading fifty meta-analyses. It still has to be asked only what it actually does.
The method
This is a review of reviews. The unit of analysis is neither the patient nor the trial, but the systematic review. Of 298 articles from the initial searches, 54 were retained after screening. The synthesis is narrative: the authors describe, compare and discuss, they do not aggregate. No pooled result is calculated, no heterogeneity is estimated at this level, no ranking of interventions is produced.
What this format brings: a coherent overview of a field that has become hard to read, and a reading of the current controversies, notably the one over prior stabilisation before trauma work, which the revision of the International Society for Traumatic Stress Studies guidance reopened. What it does not bring: its own estimate of effect size, a formal appraisal of the quality of the reviews gathered, a grading of confidence in the evidence. The distinction matters and the article keeps to it: when a head-to-head comparison between two psychotherapies is reported here, it comes from an included review, not from a calculation by the authors. Every figure cited below carries the limitations of the review it comes from, not the solidity of a quantitative synthesis.
The study at a glance
| Population, intervention, comparator, outcome, design | |
|---|---|
| Population | |
| PThe criteria apply to the reviews, not to patients. Included: systematic reviews on the treatment of post-traumatic stress disorder and its complex form in working age adults. Excluded: reviews focused solely or predominantly on children or older adults, on a specific population, serving or veteran military personnel, refugees and asylum seekers, or on significant comorbid conditions such as post-traumatic stress disorder and psychosis, or post-traumatic stress disorder and physical health problems. | |
| Intervention | |
| IAll psychological, social, pharmacological, physiological and physical interventions. The text covers individual, group, couple and family psychotherapies, digital, remote, intensive and virtual reality formats, pharmacotherapy, non-invasive brain stimulation, electroconvulsive therapy, neurofeedback, physical exercise and mind-body approaches, as well as MDMA, ketamine and cannabinoids. | |
| Comparator | |
| CNo comparison is calculated by this work. The comparisons reported are those of the included reviews, among them two reviews of head-to-head comparison between EMDR and trauma-focused cognitive behavioural therapy. | |
| Outcome | |
| OEfficacy on post-traumatic symptoms, tolerability, and dropout rates. Outcomes vary from one review to another, which limits how far they can be set side by side. | |
| Design | |
| DReview of reviews, narrative synthesis of 54 systematic reviews or meta-analyses published in the previous five years, selected from 298 articles. The authors claim no level of evidence, and this format corresponds to no rank on the usual scales, which reserve their top rank for systematic reviews of homogeneous randomised trials with quantitative pooling. | |
| Literature search | |
| SourcesPubMed, Medline and the Cochrane Library, searches completed on 25 July 2024, publications in English, reference lists of included articles searched by hand. Neither Embase nor PsycINFO is among the databases searched. |
The quality check
| Item | Judgement |
|---|---|
| Defined question and scope | Clear |
| FindingThe scope is explicit: treatments for post-traumatic stress disorder and its complex form in working age adults, debates included. | |
| Breadth of the search | Only three databases |
| FindingNeither PsycINFO nor Embase was searched, which is a problem for a topic with a strong psychotherapy component. The authors themselves acknowledge that they may have missed relevant reviews. The risk is not theoretical: useful reviews may never have entered the corpus. | |
| Inclusion and exclusion criteria | Stated explicitly |
| FindingThe criteria are set out in the methodology, and the full search strategy and a table describing the 54 included reviews are provided as supplementary material. What is missing is the flow diagram: the path from 298 articles to 54 is not broken down by reason for exclusion. | |
| Formal appraisal of review quality | Absent |
| FindingNo standardised tool is applied to the included reviews. The authors comment on quality as they go, but a sound review and a weak review enter the synthesis with the same weight. | |
| Grading of the evidence | None |
| FindingThe only certainty gradings cited are those produced by the included reviews, for example a “very low” certainty for ketamine as monotherapy. Readers have to reconstruct confidence for themselves, intervention by intervention. | |
| Quantitative pooling | Not performed |
| FindingA deliberate choice of narrative synthesis. Consistent with the heterogeneity of the included reviews, but it rules out any estimate of its own and any numerical hierarchy. | |
| Overlap between included reviews | Not reported |
| FindingNothing indicates to what extent the 54 reviews rest on the same primary trials. This is the Achilles heel of the format: the same trial counted ten times gives an impression of convergence that is not convergence. | |
| Heterogeneity and publication bias | Outside the calculation |
| FindingSince this work aggregates nothing, it produces neither a heterogeneity statistic nor a test for publication bias. It does, however, report them from others: publication bias flagged by the reviews on MDMA, high heterogeneity across analyses for ketamine. | |
| Funding and competing interests | None declared |
| FindingThe publication carries two explicit statements, no conflict of interest declared and no funding declared. Both authors belong to the Division of Psychiatry at University College London. | |
The findings
| Intervention | What is reported |
|---|---|
| Trauma-focused psychotherapies, TF-CBT and EMDR | Efficacy confirmed, and the strongest effect sizes in the overview |
| ReadingFirst-line position confirmed, with cost-effectiveness data. Cognitive processing therapy, cognitive therapy and prolonged exposure are described as efficacious, with large effect sizes and good long-term outcomes; EMDR and narrative exposure therapy are also supported. No numerical effect size is published in this work: magnitudes are exclusively qualitative. Conversely, the evidence remains insufficient for supportive counselling, group interpersonal therapy, psychodynamic therapy, relaxation, psychoeducation and present-centred therapy. | |
| Head-to-head comparisons of EMDR and TF-CBT | Equivalence in two reviews, of eight then fifteen trials |
| ReadingThe first review, covering eight head-to-head trials, concludes that efficacy on post-traumatic symptoms is equivalent; within it, three studies show lower depression scores and three others lower anxiety scores in the EMDR group at the end of treatment, with no difference at three and six months. The second, of fifteen randomised trials, finds no significant difference in symptom severity or in dropout. Such head-to-head comparisons remain rare: most primary trials compared psychotherapy with a waiting list. | |
| Complex PTSD | The same active therapies, with a marked effect on self-concept |
| ReadingA review and meta-analysis of 51 randomised trials reports, in participants with a complex presentation, the efficacy of TF-CBT, exposure alone and EMDR compared with usual care, with effects described as moderate-large to large on negative self-concept and moderate to moderate-large on interpersonal relationships. These magnitudes are qualitative, no value is given. An important point for interpretation: few trials report data on affect dysregulation, the third element of the triad. A meta-review of 24 meta-analyses concludes that good-quality empirical evidence exists in this population, while noting that the quality of the included meta-analyses varies. | |
| Prior stabilisation | Phased approach effective, sometimes superior, but the effect of phase 1 itself uncertain |
| ReadingThe debate stems from the expert consensus published in 2012, in which 84% of 50 expert clinicians recommended a sequential three-phase approach, stabilisation, trauma memory processing, reintegration. That guidance has since been revised towards a personalised rather than necessarily sequential approach. The data reported here are not neutral: a review of 12 studies concludes that phased treatments are effective and, in some cases, significantly more so than single-phase interventions, with the effect of phase 1 on phase 2 outcomes remaining mixed; a network meta-analysis of 116 studies concludes that multicomponent interventions are the most effective after complex trauma. Phase 3 remains the poor relation: a single review, covering 15 studies of heterogeneous designs and low overall quality, has examined it. No criterion is offered for selecting the patients who would benefit from a stabilisation phase. | |
| Pharmacotherapy | Low to medium effects as monotherapy, below psychotherapies |
| ReadingThe reviews gathered converge: low to medium effects for drug monotherapy, mixed results when added to psychotherapy, and superiority of psychological interventions when the two are compared, in efficacy, cost-effectiveness and durability of benefit, with more adverse events on the drug side. The UK guidelines do not place medication first-line and reserve a selective serotonin reuptake inhibitor or venlafaxine for cases where the patient prefers that option. For molecules repurposed from other indications, the yield is thin: rivastigmine, four studies of weak design and no difference from placebo in the randomised trial; prazosin, a meta-analysis of six studies with a large placebo effect in the largest trial and, for the authors of that review, no evident treatment difference; intranasal oxytocin, 14 randomised trials, mixed results and evidence judged insufficient. No published systematic review addresses the pharmacology of complex post-traumatic stress disorder. | |
| MDMA-assisted psychotherapy | Five reviews, moderate to large effects, very low certainty, FDA refusal in 2024 |
| ReadingAll the reviews report moderate to large effect sizes on post-traumatic symptoms, and all flag the same limitations: small samples, problems with blinding, publication bias, hence certainty judged very low. All also flag serious adverse effects, muscle tightness, jaw clenching, headaches, anxiety, low mood, nausea, loss of appetite. Blinding is practically impossible with a psychoactive substance of this kind, and expectation bias in participants as in therapists is not controlled; a novelty effect cannot be ruled out. In 2024, the US Food and Drug Administration declined approval and requested further trials addressing the methodological limitations and safety concerns. That refusal concerns an application, it does not demonstrate that the product is ineffective. | |
| Ketamine | Five reviews, four positive, effect measured mainly at twenty-four hours |
| ReadingFour of the five reviews report statistically significant reductions on post-traumatic symptom scales, but most of these effects are documented twenty-four hours after administration, with few studies measuring long-term impact. One review grades the evidence “very low” for ketamine alone and “low” in combination with psychotherapy. The review authors flag small samples and high heterogeneity across analyses, as well as adverse effects, dry mouth, dizziness, blurred vision. Neither the number of primary trials nor their sample sizes are given in this synthesis, which prevents readers from judging the available power for themselves. | |
| Cannabinoids | Three reviews, possible symptom reduction, limited safety evidence |
| ReadingThe three reviews, which bring together a variety of designs, indicate that medical cannabis may reduce post-traumatic symptoms and anxiety and improve sleep. Their authors immediately stress the limitation: evidence of safety and efficacy is currently limited. No comparison with placebo is reported in this synthesis, nor any effect size. A signal therefore exists, but its level of certainty does not make it an option to offer. | |
| Modes of delivery | Digital, telehealth, intensive format, virtual reality |
| ReadingThis is the real movement of these five years. Apps and digital interventions are feasible and acceptable, with uneven effectiveness; therapies delivered by video or telephone show efficacy comparable to in-person care, in-person delivery remaining preferable for some groups, notably in comorbid depression. Intensive formats, in a review of 11 studies, show a large impact on symptoms and a high completion rate. Three reviews support virtual reality exposure against a waiting list, but with markedly smaller effects against active treatments, variable study quality and a predominance of male and military samples. | |
| Dropout during treatment | 16% across 115 trials, 20.9% across 85 trials |
| ReadingTwo meta-analyses, two estimates. The first, of 115 randomised trials of psychotherapies, gives a pooled rate of 16%, with a 95% confidence interval of 14 to 18. The second, of 85 trials of guideline-recommended treatments, gives a mean rate of 20.9%, with a 95% confidence interval of 17.2 to 24.9. Populations and definitions of dropout differ, which explains part of the gap. This is the figure most directly transferable to a caseload: a non-negligible share of patients do not finish what they started. | |
Critical appraisal
| Domain | Risk |
|---|---|
| Source selection bias | Concern |
| FindingThree databases, without PsycINFO or Embase, and no flow diagram to follow the exclusions. The final corpus is probably incomplete in the psychotherapy field, as the authors acknowledge. | |
| Quality of included units | Not assessed |
| FindingWithout a standardised tool, the reliability of each review does not enter the balance. A conclusion shared by many weak reviews is no more solid than an isolated result from one rigorous review. | |
| Synthesis method | Narrative |
| FindingThe step from the reviews to the final message rests on the authors’ judgement. That is defensible here, but it introduces a degree of subjectivity that neither a confidence grading nor a pooled calculation keeps in check. | |
| Age of the primary data | Acknowledged |
| FindingA limitation raised by the authors themselves: the selection covers recent reviews, but the trials they contain may be old. An overview of reviews is not an up-to-date overview of trials. Work not yet taken up in a systematic review is, by construction, absent. | |
| External validity | Truncated |
| FindingReviews centred on military personnel, veterans, refugees and asylum seekers are excluded, as are those on children, older adults or significant comorbid conditions. These are among the populations most affected, and among those in which response to treatment is reputed to be hardest. In practice, this removes from scope part of the patients seen in specialist care, notably in services dedicated to victims of violence and exile. The authors say so explicitly: the interventions described may need to be adapted for these groups. | |
| Fit between conclusions and evidence | Proportionate |
| FindingThis is the real strength of the work. The authors overinterpret neither MDMA nor ketamine, carry over the reservations of each review, acknowledge the limitations of their method, and present the debates without settling them artificially. | |
| Independence | No declared interests |
| FindingNo conflict of interest and no funding declared, work from a university department. Allegiance bias remains possible in a field where teams are often attached to one psychotherapy modality, but nothing in the text suggests it. | |
Level of evidence
Confidence is high on one point, and it is the central one: the first-line place of trauma-focused psychotherapies, EMDR included, complex presentations included. This result does not rest on this work alone, it emerges consistently from the reviews gathered, and two head-to-head reviews indicate equivalence between EMDR and trauma-focused cognitive behavioural therapy. Confidence is also reasonable on a second point: the level of evidence for recent psychoactive interventions, MDMA, ketamine and cannabinoids, remains too low to make them a recommendable option today. That is the right wording. It is not an absence of signal, since the reviews report effects, but a certainty graded very low to low by those same reviews, on trials with small samples, with problems of blinding and publication bias.
Confidence is low elsewhere. On the order of magnitude of effect sizes, since no value is published here and magnitudes remain qualitative. On any fine hierarchy between psychotherapies, which cannot be derived from a narrative synthesis. On what to do about prior stabilisation, where the reviews reported lean towards a phased approach without the specific effect of the stabilisation phase being established or a patient selection criterion being offered. And on transfer to the excluded populations, where absence of data must not be read as equivalence. The scientific score reflects the gap between the breadth of the corpus and the lightness of its methodological handling: no appraisal of review quality, no confidence grading, no quantitative analysis, no count of overlap between the included reviews.
The colleague test
What an experienced colleague would say if you put this study to them in two minutes, between two consultations.
“ Well, this changes nothing in what I do: trauma-focused CBT and EMDR stay first-line, and that holds for complex presentations too. MDMA, the FDA said no in 2024 and the reviews grade the certainty at the bottom of the scale, so I wait. Ketamine, an effect at twenty-four hours and nothing solid beyond, it is too early to raise it with anyone. Cannabis, a signal but nothing to offer. What I really take away is dropout, one patient in six in one case, one in five in the other. That is where the work is, on the alliance, not on the next molecule. ”
What this means in practice: therapeutic novelty does not shift the treatment line, but it takes up the conversation with patients. Better to have the answer ready, and to put clinical energy back where it still produces results, continuity of care.
What you can do with this
- Refer first-line to a trauma-focused psychotherapy or EMDR, without steering the patient away from that route because the presentation is complex. According to the reviews gathered, disturbances in self-concept and relationships also respond to these therapies. The least documented point remains affect dysregulation, since few trials report results on it.
- Treat prior stabilisation as an individual decision, not a mandatory step. The revised international guidance now favours a personalised rather than necessarily sequential approach, and the available reviews support the efficacy of phased approaches without isolating the specific contribution of stabilisation. In the absence of a validated selection criterion, the discussion happens patient by patient, and is documented in the notes.
- Anticipate dropout from the first consultation. With 16% discontinuation in one meta-analysis of 115 trials and 20.9% in another covering 85 trials, it is worth naming the difficulty of exposure sessions from the outset, agreeing on what to do if the patient disengages, and getting back in touch rather than simply recording a missed appointment.
- Prepare a short answer on MDMA, ketamine and cannabis. What is demonstrated: nothing that reaches a level of evidence sufficient to offer these options in this indication. What is suggested: real effects but graded very low to low, measured in small samples, at twenty-four hours for ketamine, with problems of blinding and adverse effects for MDMA. What is opinion: everything else the patient has read.
- Check the regulatory framework that applies where you practise, which lies outside the scope of this synthesis. France serves as the example here, on elements verified on 13 August 2026. In France, injectable ketamine holds a marketing authorisation limited to anaesthesia, and post-traumatic stress disorder is not among its indications; a compassionate prescribing framework was opened in France on 9 March 2026 for two ketamine products at 10 mg/mL in the treatment of severe suicidal ideation in adults, an indication distinct from the one discussed here. In France, medical cannabis has fallen, since the end of the experimental programme on 31 December 2024, under a transitional framework of restricted access, as a last resort and on initial hospital prescription, and post-traumatic stress disorder is not among the situations covered. No marketing authorisation was identified in France for MDMA in this indication. Under the French framework, use outside the marketing authorisation is not freely permitted: it requires the absence of an appropriate authorised alternative. These frameworks change and should be checked in their current version at the time of the decision.
- The course of action is set out in the NICE decision tree for depression in adults.
Frequently asked questions
Does complex PTSD need a different treatment from PTSD?
According to this overview, no, not first-line: the same trauma-focused psychotherapies work, including on negative self-concept and on interpersonal relationships, with effects described as moderate to large depending on the dimension considered. That does not mean management is identical. The number of sessions, the pace and work on the relationship are not documented here, no numerical effect size is published, and affect dysregulation, the third component of the picture, remains little studied.
Is a stabilisation phase needed before trauma-focused therapy?
The question remains open, but it is not without data. The 2012 international guidance, drawn from a consensus in which 84% of 50 expert clinicians supported a three-phase approach, has been revised towards a personalised approach. The reviews gathered here indicate that phased treatments are effective and, in some cases, more so than single-phase interventions, while the specific effect of the stabilisation phase on the outcomes of the trauma memory processing phase remains mixed. The defensible position is therefore personalisation: neither systematic stabilisation, which sometimes delays active treatment indefinitely, nor immediate exposure in a patient whose safety is not assured. No validated criterion allows a choice between the two, it is a clinical judgement.
A patient asks about MDMA, ketamine or cannabis for PTSD: what should I say?
That the data exist, that they have been taken seriously, and that they are not enough. For MDMA, five reviews report moderate to large effects, but all flag small samples, blinding that is practically impossible with a psychoactive substance of this kind, publication bias and notable adverse effects; certainty is graded very low and the US agency declined the application in 2024. For ketamine, four reviews out of five report a significant reduction in symptoms, but most of the effect is measured twenty-four hours after administration, and one review grades the evidence “very low” as monotherapy. For cannabis, three reviews suggest a reduction in symptoms and anxiety, and an improvement in sleep, on evidence of safety and efficacy that their authors judge limited. In France, the example detailed above, none of these options is part of routine care for post-traumatic stress disorder, and the applicable framework, wherever you practise, should be checked in its current version.
Why does this review not rank PTSD treatments?
Because it aggregates nothing. It is a narrative synthesis of 54 reviews, without quantitative pooling, without formal appraisal of the quality of the included reviews and without grading of confidence in the evidence. It describes a landscape, it does not calculate comparisons. The only head-to-head comparisons reported are those produced by the reviews themselves, and they conclude that EMDR and trauma-focused cognitive behavioural therapy are equivalent. A full ranking would require a network meta-analysis of the trials themselves, which is a different exercise.
Do these findings apply to military veterans or refugees with PTSD?
Not directly: reviews centred on these populations are explicitly excluded from scope, as are those on children, older adults or significant comorbid conditions. This is the most important limitation for clinical readers, since part of the patients with post-traumatic disorders seen in specialist care have histories of exile or collective violence. The authors indeed direct readers to the reviews dedicated to these groups. The results should neither be transferred without reservation, nor taken to mean that treatments are less effective there: the question was not asked here.
Annotated bibliography
Source study. Billings J, Nicholls H. PTSD and complex PTSD, current treatments and debates: a review of reviews. British Medical Bulletin 2025; 156(1): ldaf015. DOI 10.1093/bmb/ldaf015. PMID 41004137. Review of reviews, 54 systematic reviews or meta-analyses, narrative synthesis. Contribution: the broadest overview available of treatments for post-traumatic stress disorder and its complex form, debates included. Limitations: no quantitative pooling, no formal appraisal of the quality of the included reviews, no confidence grading, no numerical effect size, overlap between reviews not reported, three databases searched, reviews centred on military personnel and refugees excluded.
The two dropout estimates cited. Lewis C, Roberts NP, Gibson S, Bisson JI. Dropout from psychological therapies for post-traumatic stress disorder (PTSD) in adults: systematic review and meta-analysis. European Journal of Psychotraumatology 2020; 11: 1709709. DOI 10.1080/20008198.2019.1709709. Varker T, Jones KA, Arjmand HA, et al. Dropout from guideline-recommended psychological treatments for posttraumatic stress disorder: a systematic review and meta-analysis. Journal of Affective Disorders Reports 2021; 4: 100093. DOI 10.1016/j.jadr.2021.100093.
A note on reading. The numbering of references 42 and 93 in the source publication is inconsistent: the text points to these numbers for two works different from those listed against them in the reference list. The two dropout references above were therefore identified and verified independently. The source for the result on complex post-traumatic stress disorder is printed in the publication with a journal and a year that do not match its subject; it is not cited here until its exact reference is confirmed. The guidance on the complex disorder discussed in the text is that of the International Society for Traumatic Stress Studies, an expert consensus published in 2012 and then a later position paper, the latter cited in the publication only as a file name, with no year or publisher.
What was consulted. The published title, the authors, the year, volume, issue, pagination, DOI, PMID, funding, the declaration of competing interests, and every figure and every methodological statement in this article were verified against the full text of the published version, supplemented by the publisher’s record and the DOI bibliographic registration, consulted on 13 August 2026. The supplementary material of the publication, a data extraction table describing the 54 included reviews and the full search strategy, could not be consulted: overlap between included reviews and the detail of exclusions were therefore not checked against that source. The French regulatory elements were verified on 13 August 2026 against the French public medicines database and the website of the French national agency for the safety of medicines and health products (ANSM). This article undergoes an independent double reading before publication.