Interventions for common mental disorders tested in Cochrane reviews: an analysis of quality of evidence and efficacy.

In a recent analysis1 of a random sample of 2,428 (35%) Cochrane reviews published between January 2008 and March 2021, only two of 48 interventions (4.2%) for common mental disorders were found to show a high quality of evidence, to be significantly superior to a comparator, and to be regarded as effective by Cochrane authors. The high-quality GRADE standard, however, could be too demanding. In particular, the assessment of risk of bias may be too stringent, leading to a downgrade of the evidence1. In the present analysis, we included not just a random sample, but all systematic reviews of interventions for common mental 2disorders published by the Common Mental Disorder Cochrane group between January 1, 2008 and June 24, 2025. We focused on interventions showing either a high or a moderate quality of evidence. Interventions were included if they2 were compared to placebo, no treatment or treatment as usual, in randomized or quasi-randomized trials, and received GRADE ratings for their primary outcome. The study was carried out in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses Statement (PRISMA)2. The protocol was registered at PROSPERO (CRD42024547340). For each Cochrane report, we extracted the GRADE rating of the first listed primary outcome1, 3. If this rating was not high, we inspected the GRADE ratings of up to three further primary or secondary outcomes. We included only those secondary outcomes that addressed target symptoms. For the outcome with the highest GRADE rating, we extracted additional data (e.g., number of trials, number of patients, effect size, confidence interval). As an addition to the original protocol, we also extracted information on whether the Cochrane authors recommended the intervention as effective. The identified Cochrane reports were randomly allocated to five pairs of raters. For each report, two reviewers extracted the data independently and compared the results. Discrepancies were solved by consensus or, if necessary, by involving a third rater. The Cochrane Database of Systematic Reviews included 150 reviews of interventions for common mental disorders published between January 1, 2008 and June 24, 2025. Of those, 57 were excluded for various (overlapping) reasons (no GRADE rating: n=42; including only comparisons with active treatment(s): n=20; superseded by an updated version: n=1). Therefore, 93 Cochrane reviews met the inclusion criteria, encompassing 207 eligible interventions. The included interventions were tested in adults (74/93 reports), in children and adolescents (13/93 reports), or in mixed samples of children, adolescents and adults (9/93 reports). In some reports, both adults and children/adolescents were studied separately (thus the sum exceeds 93). The 207 interventions encompassed 50 treatments for depressive disorders, 46 for anxiety disorders (including obsessive-compulsive disorder), 27 for post-traumatic stress disorder, 46 for self-harm, 6 for pathological gambling, 5 for bipolar disorder or mania, 4 for somatoform disorders, 4 for eating disorders, 3 for insomnia, and the remaining for several other conditions. Of the included 207 interventions, 11 (5.3%) were rated by the Cochrane authors as showing a high quality, 60 (29.0%) a moderate quality, 69 (33.0%) a low quality, and 66 (31.9%) a very low quality of evidence. Of these interventions, 8 (3.9%), 39 (18.8%), 20 (9.7%) and 20 (9.7%), respectively, were significantly superior to the control condition. In sum, 47 (22.7%) of all interventions were supported by a high or moderate quality of evidence and showed a significant superiority to the control condition (see also supplementary information). Harms were reported for 62 of the 207 interventions (30.0%), with 12 (5.8%) yielding significantly more harms than the control conditions. Many of these comparisons (50.2%) were based on only one (75/207) or two (29/207) studies. For pharmacotherapy (n=99 interventions) and psychotherapy (n=75 interventions), the proportions of interventions showing high or moderate quality of evidence and statistically significant results were 29.3% and 20.0%, respectively. Most pharmacological interventions showed low or very low quality of evidence (63.6%, 63/99), and the majority did not have statistically significant results (52.5%, 52/99). The same pattern was found for psychotherapy: the evidence was of low or very low quality in 66.7% (50/75) of interventions, and most of them (54.7%, 41/75) did not have statistically significant results. High or moderate quality of evidence plus statistically significant results was found in only 6.5% of all other interventions (see supplementary information). In only 6 of 207 comparisons, the quality of evidence was high or moderate, the superiority over the control condition was statistically significant, and the Cochrane authors considered the intervention as effective. These included valproate for acute mania in mixed samples of inpatients and outpatients; eszopiclone for insomnia; cognitive behavior therapy (CBT) for adult self-harm; CBT for child and adolescent anxiety disorders; sildenafil for sexual dysfunction caused by antidepressants; and amitriptyline for depressive disorders. For depressive disorders, anxiety disorders, post-traumatic stress disorder and self-harm, 32.0%, 28.3%, 14.8 and 10.9% of the interventions, respectively, showed high or moderate quality of evidence and statistically significant superiority over control conditions. In sum, of all interventions evaluated for common mental disorders, less than a quarter (22.7%) were supported by high or moderate quality evidence and were statistically significantly superior to the comparator. Our evaluation was based on the existing GRADE ratings of the Cochrane report authors, which may be considered a limitation. Errors from the Cochrane assessors cannot be excluded, but one can trust that these are probably quite accurate, as Cochrane reviews tended to have high quality in previous assessments4. Overall, our findings suggest that many often studied and clinically applied interventions for common mental disorders still lack robust, convincing evidence.