So how special is special K? A systematic review and meta-analysis of ketamine for PTSD RCTs

Post-traumatic stress disorder (PTSD) is a prevalent and burdensome condition for which current pharmacological and psychological treatments have recognised limitations, including modest efficacy, adverse effects, and high resource demands. Interest has grown in novel agents that may deliver rapid symptom relief; ketamine, an NMDA receptor antagonist with dissociative properties, has been investigated for PTSD on the rationale that its rapid antidepressant effects and putative actions on synaptic connectivity or fear-memory processing might translate into benefit for PTSD. Previous reviews and meta-analyses have been limited by mixing study designs, narrow focus (for example on pain rather than PTSD), small samples, or omission of recent large trials, leaving uncertainty about ketamine's incremental efficacy against control interventions in rigorous randomised controlled trials (RCTs). Borgogna and colleagues therefore conducted a systematic review and meta-analysis restricted to RCTs that compared ketamine to control interventions in adult PTSD samples. The main aims were to estimate the pooled between-group effect of ketamine versus controls across timepoints, to examine temporal patterns (notably rapid effects at 24 hours), and to assess bias, heterogeneity, and study features that might explain variability in findings. Because the ketamine-for-PTSD literature consists of small trials, the authors emphasised meta-analytic aggregation and bias correction as tools to obtain more reliable estimates and to guide design of future trials.

The review followed PRISMA guidelines and searched multiple databases (Medline, PsycINFO, ClinicalTrials.gov, Alt HealthWatch, CINAHL Complete, PsycARTICLES, JSTOR, PTSDpubs, PubMed, Web of Science, SpringerLINK, Gale Health and Wellness, and Scopus) using combinations of 'PTSD'/'Post-Traumatic Stress Disorder' with 'ketamine' and 'randomized controlled trial' or 'RCT'. Searches covered all years up to September 2023 and were complemented by Google Scholar and reference checking; Google Scholar results were scanned up to the first 100 items per term but not included in the formal flowchart due to volume. Inclusion criteria required adult human participants (18+), use of a psychometrically validated PTSD measure, administration of ketamine, an RCT design, and peer-reviewed publication. Exclusion criteria included participants under 18, absence of validated PTSD outcome measures, non-RCT designs, non-peer-reviewed outlets, or no ketamine administration. Titles and abstracts were screened, full texts read when potentially eligible, and data (means, standard deviations, sample sizes, demographics) extracted. Three raters performed quality coding using the NIH Study Quality Assessment of Controlled Intervention Studies; all included trials were rated as 'good' and coding discrepancies were resolved by group discussion (none remained unresolved). Extracted data sheets were deposited on the Open Science Framework. Meta-analyses were performed in Comprehensive Meta-Analysis v4 using random-effects models. Hedges' g was the primary between-group effect-size metric (adjusting for small sample sizes) with conventional benchmarks (0.2 small, 0.5 medium, 0.8 large). Cohen's davg was used for within-group repeated-measure contexts because bivariate pre–post correlations were not reported. Heterogeneity was assessed with Cochran's Q and the I2 statistic, with subgroup heterogeneity tested when k ≥ 3. Publication bias was evaluated using Egger's regression and Trim-and-Fill imputation. The authors also examined predefined temporal windows (24 hours, one week, two weeks post-initial infusion) and other outcomes such as time to relapse; where individual studies used multiple ketamine doses or cross-over designs, the authors treated distinct dose conditions separately for omnibus analyses or used pre-cross-over data as appropriate.

The systematic search yielded k = 6 RCTs totalling n = 221 participants and keffects = 52 effect-size estimates; one trial contributed a low-dose (0.2 mg/kg) condition in addition to the standard 0.5 mg/kg condition, resulting in seven condition comparisons. PTSD outcomes were measured with instruments including the PCL-5, IES-R, and CAPS-5. Three trials used active pharmacological controls (midazolam or ketorolac) and the remainder used saline (passive control). Only two trials were double-blinded, three did not have equivalent between-group baselines, and two had attrition > 20%. All trials were pre-registered and conducted in the United States; one trial had a cross-over design. An omnibus random-effects meta-analysis across all measures and timepoints found a small advantage for ketamine over control interventions (Hedges' g = 0.27, 95% CI = 0.03 to 0.51, keffects = 52). However, evidence of small-study effects was detected: Egger's test showed a significant association between effect size and standard error (Egger's t(5) = 3.41, p = .019), and Trim-and-Fill imputation suggested two missing studies; after adjustment the omnibus effect attenuated to g = 0.20 (95% CI = -0.08 to 0.48), which is not statistically significant. The 24-hour post-infusion timepoint (keffects = 9) was the only consistently reported outcome across trials. Pooled effects at 24 hours indicated a small benefit of ketamine (g = 0.35, 95% CI = 0.06 to 0.64), although between-study variability was moderate (Q(6) = 9.18, p = .164; I2 = 34.7%). Some individual effects at 24 hours were large and favoured ketamine (notably a PCL-5 and CAPS-5 effect in one study and an IES-R effect in another), while two trials reported small non-significant inverse effects. Post-hoc subgroup analyses split by control type found ketamine was significantly better than passive (saline) controls at 24 hours (g = 0.44, 95% CI = 0.03 to 0.85), but not significantly different from active pharmacological controls (g = 0.24, 95% CI = -0.30 to 0.78). At one week post-initial infusion (keffects = 6) the pooled effect approached significance but was small (g = 0.24, 95% CI = 0.00 to 0.48), with heterogeneity statistics consistent with low observed between-study variance (Q(4) = 3.43, p = .489; I2 = 0.0%). Analyses separating single versus multiple administrations yielded small non-significant effects. At two weeks (keffects = 4) the pooled effect was small and non-significant (g = 0.17, 95% CI = -0.10 to 0.44); one study reported a relatively large and significant CAPS-5 effect following three infusions, whereas other studies reported null or small non-significant effects for low- and high-dose conditions. Time-until-relapse analyses (keffects = 4) produced a large pooled effect (g = 1.01, 95% CI = 0.30 to 1.73), indicating longer time-to-relapse or lower PTSD scores at relapse in the ketamine arms in those specific comparisons. Reported mean times to relapse in the reviewed studies averaged 34.44 days (SD = 19.12) for ketamine versus 16.5 days (SD = 11.39) for controls, though the trials and contexts underlying these figures vary. When one study with suspected typographical errors in its relapse and follow-up tables was excluded, omnibus effects and several subgroup effects (including the 24-hour effect and the relapse effect) became non-significant, indicating sensitivity of pooled estimates to particular datasets. Additional concerns reported in the results include small-sample bias (smaller trials more likely to report larger ketamine benefits), possible blind penetration (dissociative effects may break blinding and amplify placebo effects), variable reporting of means and standard deviations (sometimes relegated to supplementary materials), inconsistent monitoring of longer-term outcomes, variable use and reporting of adjunctive psychotherapy, and potential conflicts of interest in some trials.

Borgogna and colleagues interpret the pooled evidence as indicating at best a small, short-lived advantage of ketamine over control interventions for PTSD that is vulnerable to bias and heterogeneity. The nominal omnibus effect across timepoints diminished after bias correction, and small trials tended to report the largest effects, suggesting a decline effect as sample sizes increased. Rapid reductions in symptoms were observed at 24 hours in some studies, but similar rapid effects were also apparent in active controls (midazolam, ketorolac) and even saline, undermining claims of a ketamine-specific mechanism and pointing instead to expectancy and placebo mechanisms amplified when blinding fails due to perceptible side effects (so-called blind penetration). The authors position these findings relative to earlier research by noting prior meta-analyses were limited by mixing designs or omitting recent trials; restricting analysis to RCTs improves rigour but also highlights the limited and inconsistent evidence base. They emphasise discrepancies between clinician-rated and self-reported outcomes (with clinician ratings sometimes showing larger ketamine effects), variable reporting of adjunctive psychotherapy, and the underreporting of adverse events in related ketamine literature—issues that complicate interpretation and generalisability. Key limitations acknowledged by the authors include the small number of RCTs (k = 6) and participants (n = 221), limited power to examine moderators (for example diagnostic comorbidity, sex, race, veteran status), all trials being US-based, heterogeneity in adjunctive treatments and blinding procedures, and potential upward bias in Hedges' g. They also note data quality concerns in individual trials (including possible typographical errors) and instances of conflict of interest among trial investigators. For future research and clinical practice, the authors recommend rigorous trial designs that address blinding (for instance multiple distinct sham controls), use appropriate and clinically relevant comparators, preregister and transparently report means, standard deviations, and adverse events, and include standardised long-term follow-up intervals (suggested 3, 6 and 12 months). They further propose RDoC-informed designs that enrol participants on the basis of biological markers purportedly targeted by ketamine rather than broad PTSD diagnoses, to better test mechanistic hypotheses and identify subgroups that might benefit. Given the small, bias-sensitive effects observed, the authors advise caution among patients, clinicians, and funders when considering ketamine for PTSD.

The study team conclude that ketamine may produce a small immediate or short-term reduction in PTSD symptoms compared with the control interventions studied, but this apparent benefit is difficult to distinguish from placebo or non-specific treatment effects in the extant RCT literature. Because effects are modest, susceptible to small-study bias, and sensitive to analytic choices, the authors advise caution in clinical use and recommend future, better-controlled trials—preferably RDoC-based—with rigorous blinding, appropriate active comparators, transparent reporting (including adverse events), and standardised long-term outcome assessments.