MDMA-assisted psychotherapy for AUD: Bayesian analysis of WHO drinking risk level and exploratory analysis of drinking behavior and psychosocial functioning at 3 months follow-up

Thurgur and colleagues situate this work within the recent resurgence of clinical research on MDMA-assisted psychotherapy, noting robust Phase III evidence for PTSD and interest in applying the approach to other disorders linked with trauma, including alcohol use disorder (AUD). Earlier open-label work by the study team established safety and tolerability for MDMA-assisted psychotherapy in AUD and reported reductions in drinking and improvements in mood up to 9 months, but questions remain about clinically meaningful drinking endpoints, broader psychosocial outcomes, and potential mediators such as craving. This paper re-analyses data from that initial feasibility trial to address two aims. First, the investigators apply a Bayesian framework to estimate the probability that participants achieved a clinically relevant harm-reduction endpoint—defined as a 2-level reduction in World Health Organization (WHO) drinking risk—at 3 months post-treatment. Second, the study reports exploratory analyses of drinking behaviour, craving, quality of life (QOL), sleep, self-compassion and empathy at 3 months versus baseline to provide preliminary evidence on broader wellbeing outcomes beyond alcohol consumption metrics.

This was an open-label feasibility study previously reported in detail; the extracted text presents the secondary-analysis methods. Fourteen adults with AUD who had recently undergone detoxification enrolled and completed an eight-week recovery-based psychotherapy course comprising ten sessions, of which two (sessions 3 and 7) were MDMA-assisted dosing sessions. The MDMA regimen for each drug-assisted session was an oral 125 mg initial dose followed by a 62.5 mg booster. Primary safety and tolerability outcomes (treatment completion, acceptance of second MDMA dose, adverse events) were described in the earlier publication; this paper focuses on secondary outcomes. Alcohol consumption was measured using the Timeline Followback (TLFB) to derive WHO drinking risk categories; treatment success for the Bayesian analysis was defined as a 2-level reduction in WHO risk from baseline to 3 months. Additional patient-reported outcome measures (PROMs) included the Penn Alcohol Craving Scale (PACS) and Obsessive Compulsive Drinking Scale (OCDS) for craving and compulsive/obsessive drinking, the SF-36 for QOL, four items derived from the Pittsburgh Sleep Quality Index (PSQI) for sleep disturbance, the Self-Compassion Scale (SCS), and the Interpersonal Reactivity Index (IRI) for empathy subscales. For the binary WHO-risk endpoint a binomial test was performed and followed by Bayesian inference using two beta priors: a flat prior (a = 1, b = 1; mean 50%) and a skeptical prior (a = 1, b = 3; mean 25%), the latter reflecting a conservative assumption about placebo-level response rates in mental health trials. Continuous PROMs were analysed with Wilcoxon matched-pairs signed rank tests given the small sample; Bonferroni correction was applied for PROMs with multiple subscales. Analyses were conducted in Jamovi and figures produced in GraphPad Prism. Effect sizes (rank biserial correlations) are reported in the appendix per the extracted text.

The study sample comprised 14 participants (8 males, 6 females; mean age 48 years), all reported as white British in the extracted text. The binomial test for the predefined 2-level WHO drinking risk reduction did not reject the null hypothesis against an expected 50% success rate (k = .64, P = .423); the 95% confidence interval for the observed success rate was 0.35–0.87. Bayesian re-analysis provided probabilistic estimates of treatment success. Using the flat prior (a = 1, b = 1), the posterior probability that participants achieved a 2-level reduction in WHO drinking risk at 3 months was 63% with a 95% credible interval of 40%–85%. Using the skeptical prior (a = 1, b = 3), the posterior probability was 55% with a 95% credible interval of 33%–77%. Measures of drinking-related behaviour and craving showed reductions at 3 months versus baseline. PACS total scores decreased (difference 5.50, 95% CI 0.5–13, P = .031, rank biserial r_b = .781). On the OCDS, the obsessive subscale decreased by 5.00 (95% CI 2.00–8.50, P = .006, r_b = .868) and the compulsive subscale decreased by 5.00 (95% CI 1.00–9.00, P = .021, r_b = .769). Broad wellbeing measures showed mixed results. Several SF-36 subscales (energy/fatigue, emotional well-being, social functioning, pain, general health, and change in reported health) showed increases at 3 months, but none remained statistically significant after Bonferroni correction. Sleep improved as indicated by a reduction in total PSQI-derived score (difference 2.00, 95% CI ~7.54e-6–3.00, P = .041, r_b = .648). Self-compassion increased substantially (SCS difference 12.1, 95% CI 9.09–14.4, P < .001, r_b = 1.00). For the IRI empathy subscales, only the personal distress subscale showed a significant reduction (difference 0.5, 95% CI 0.143–0.929, P = .009, r_b = .945), with positive mean differences reported as baseline minus 3-month scores indicating reductions over time. The extracted text notes that full statistical details and supplementary tables/figures are provided in the article's supplementary materials.

Thurgur and colleagues interpret their Bayesian re-analysis as indicating a 55%–63% probability that participants achieved a clinically meaningful 2-level reduction in WHO drinking risk at 3 months following MDMA-assisted psychotherapy. They argue this harm-reduction endpoint is relevant to regulatory and clinical contexts, noting that agencies such as the EMA accept intermediate endpoints in AUD trials and that WHO risk reductions have been associated with improved functioning and lower healthcare costs. The authors compare their estimated success probabilities with secondary analyses from large AUD trials (COMBINE and Project MATCH), which reported a 68% two-level reduction at end of treatment, and suggest the MDMA-assisted psychotherapy outcomes are broadly comparable, while emphasising the preliminary nature of the data. Beyond drinking metrics, the investigators highlight reductions in craving (PACS and OCDS), improvements in sleep and self-compassion, and a decrease in personal distress as indicators that MDMA-assisted psychotherapy may yield broader psychosocial benefits. They link these findings to the known acute prosocial and empathogenic effects of MDMA and propose self-compassion and empathy as possible transdiagnostic mediators deserving further study. The authors acknowledge several important limitations: the open-label design, small sample size, exploratory multiple secondary outcomes (risking Type I error), challenges with blinding in psychedelic research, lack of long-term follow-up in this analysis, restrictive exclusion criteria (notably excluding antidepressant users and certain comorbidities) which limit generalisability, and the sample's lack of ethnic diversity. They also note that SF-36 scores can be influenced by comorbid conditions and that the extracted text did not allow adjustment for such confounds due to limited comorbidity data. Practical and implementation barriers are discussed, including the need for manualised psychotherapeutic approaches, therapist training, resource-intensive 8-hour dosing sessions with two therapists, and overnight stays used for safety in this feasibility trial. The authors suggest that future trials should be larger, randomised, placebo-controlled, measure deblinding and expectancy, include more diverse and representative samples, assess mediation by craving, and evaluate whether more time- and cost-efficient models of MDMA delivery and therapy could be developed.

The paper extends previously reported safety and tolerability data by providing a Bayesian estimate of treatment success for MDMA-assisted psychotherapy in AUD, framing success as a 2-level reduction in WHO drinking risk and estimating a 55%–63% probability of achieving this at 3 months. Preliminary secondary findings suggest reductions in alcohol craving and improvements in sleep, self-compassion and some aspects of wellbeing, but the authors emphasise these are hypothesis-generating results from a small, open-label sample. They recommend larger, placebo-controlled trials using these measures to determine the efficacy and practical feasibility of MDMA-assisted psychotherapy for AUD.