This was a randomized double-blind study designed to compare the efficacy of MDMA-assisted therapy with that of placebo with therapy. Fifteen study sites, consisting of 11 in the United States, two in Canada and two in Israel, included both institutional sites and private clinics. Ethics approval was obtained from Copernicus Group Independent Review Board, Western Institutional Review Board, University of British Columbia Providence Healthcare Research Ethics Board, and the Helsinki Committees of Be’er Ya’akov Ness Ziona Mental Health Center and Chaim Sheba Medical Center. This clinical study was conducted in accordance with the principles of the Declaration of Helsinki. The public study protocol is available at http://maps.org/mapp1. The therapist manual is available at http://maps.org/treatment-manual.
Participants were recruited through print and internet advertisements, referrals from treatment providers, and by word of mouth. Participants were required to initiate contact with the study sites themselves, even if recommended by a provider. After providing written informed consent, participants were screened for eligibility. The criteria for inclusion consisted of meeting the Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5) criteria for current PTSD with a symptom duration of ≥6 months at screening (as assessed with the Mini International Neuropsychiatric Interview (MINI) for DSM-5), and a CAPS-5 total severity score of ≥35 at baseline. Exclusion criteria consisted of primary psychotic disorder, bipolar I disorder, dissociative identity disorder, eating disorders with active purging, major depressive disorder with psychotic features, personality disorders, current alcohol and substance use disorders, pregnancy or lactation, and any medical condition that could make receiving a sympathomimetic drug harmful due to increased blood pressure and heart rate, including uncontrolled hypertension, history of arrhythmia, or marked baseline prolongation of QT and/or QTc interval. Participants with other mild, stable, chronic medical problems (for example, type 2 diabetes mellitus or well-controlled hypertension) were eligible for enrollment if the site physician, clinical investigator and medical monitor agreed that the condition would not increase the risk associated with MDMA administration. Participants were required to comply with lifestyle modifications, including a medically supervised discontinuation of psychiatric medications for a minimum of five half-lives plus one additional week before the baseline assessments (see the study protocol for inclusion and exclusion criteria).
The study protocol was amended on three occasions during study enrollment: first, to add clarity to eligibility criteria related to comorbid medical conditions; second, to add terms of suicidal ideation and behavior as AESIs, as requested by the FDA; and third, to increase the frequency of suicidality assessments following experimental sessions, as requested by the FDA, and to add an option for some telemedicine visits following the COVID-19 pandemic. Given that the study was at full enrollment (n = 105) when COVID-19 shut down in-person interactions at most of the study sites, the FDA and sponsor concluded that a reduced sample size of 90 participants, instead of the planned 100, would maintain sufficient statistical power to meet study objectives and would avoid COVID-19 delays of experimental sessions, which might confound the assessment of treatment effects.
The study drug was manufactured in accordance with Current Good Manufacturing Practice (CGMP) standards by Onyx Scientific and compounded by Sharp Clinical Services. Assays for chemistry, manufacturing and controls were completed in accordance with the CGMP and International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) standards, and reported to the FDA, Health Canada and the Israel Ministry of Health.
Randomization, masking and bias minimization
Participants were randomized in a blinded fashion and were allocated 1:1 to either the MDMA-assisted therapy group or the placebo with therapy group. Randomization was stratified by site and occurred following enrollment confirmation (after preparatory visits). Randomization was managed via an interactive web randomization system—ITClinical IWRS, version 11.0.1 (ITClinical, LDA)—based on a centralized randomization schedule developed by an independent third-party vendor to maintain blinding. Participants, site staff and the sponsor were blinded to participant group assignment until after the database was locked.
An inactive placebo with therapy was utilized as the comparator to isolate the efficacy of the MDMA itself. Although low-dose MDMA improved blinding in phase 2 studies, it led to decreased effectiveness compared with an inactive placebo in a PTSD population, making it easier to detect a difference between the active and comparator groups15. The use of inactive placebo also allows for uncontaminated comparison of safety data between groups. Therefore, an inactive placebo was determined in partnership with the FDA as a more conservative statistical comparison, and the study utilized observer-blinded efficacy assessments to minimize bias in efficacy measurements.
An observer-blind and centralized independent rater pool was used to administer the primary and secondary outcome measures, that is, the CAPS-5 and the SDS for functional impairment, the latter of which was adapted to limit missing item-level data as per the FDA requirements and included use of the three-item mean as the total score and imputation of work-related impairment as the maximum score, if caused by PTSD. The independent rater measurements were conducted at baseline and following each experimental session via live video interviews. Independent raters did not repeatedly see the same participant and the independent rater pool was blinded to the complete study design, visit number, treatment assignment, and all data collected by the therapy team after baseline, with the exception of safety data related to suicidality. Participants were instructed to withhold their opinion on treatment group assignment from independent raters and to refrain from sharing details regarding the study design and their number of completed visits. To ensure that all site and sponsor staff were shielded from study outcome measures, primary and secondary outcome measures were collected from the blinded independent rater pool and stored in a dedicated database that was separate from the blinded, clinical database.
Following an initial phone screening, participants provided written informed consent and underwent further screening assessments for eligibility. These included the PTSD Checklist for DSM-5 (PCL-5), the MINI for DSM-5, the Structured Clinical Interview for DSM-5 Screening Personality Questionnaire and Structured Clinical Interview for DSM-5 Personality Disorders (SCID-5-SPQ and -PD), the Lifetime C-SSRS, medical history, and pre-study medications. Study staff contacted outside providers, ordered medical records, and conducted a physical examination, laboratory testing (including pregnancy and drug tests), electrocardiogram, and 1-min rhythm strip. Eligible participants were enrolled in the study and began psychiatric medication taper (Table 1) if needed, and collection of adverse events. Anticipated effects of MDMA, such as euphoria, stimulation and feelings of closeness39, were intentionally not solicited as adverse events to avoid biasing the collection of adverse event data. Participant medication taper was variable, lasting from 0 d (no taper needed) to 103 d. Clinical data were electronically captured using Medrio EDC versions R40–R40.7.
In accordance with FDA guidance, we paid special attention to a subset of adverse events, termed AESIs, relating to cardiac function that could be indicative of QT interval prolongation or cardiac arrhythmias, abuse liability, and suicidal ideation and behavior. All adverse events that included signs or symptoms potentially associated with a cardiovascular event such as palpitations or dizziness were further evaluated for reporting as a cardiovascular adverse event. To assess signs of MDMA abuse potential, any adverse event terms such as ‘behavioral addiction’, ‘drug abuser’, ‘substance abuser’, ‘dependence’, ‘intentional product misuse’, ‘overdose’ (accidental, intentional or prescribed) or ‘drug diversion’ were collected and coded as AESIs. Suicidal ideation that was judged as serious or severe by the investigator, serious ideation defined as a C-SSRS suicidal ideation score of a 4 or 5, self-harm in the context of any suicidal ideation, and any suicide attempts were reported as AESIs.
Enrolled participants underwent three 90-min preparatory sessions of therapy with a two-person therapist team in preparation for experimental sessions (Fig. 1). The preparatory sessions focused on establishing therapeutic alliance and trust, and also provided guidance on how to respond to the memories and feelings that could arise during treatment. Participants who failed to meet all eligibility criteria were withdrawn during this preparatory period. Baseline CAPS-5 assessment (to confirm PTSD diagnosis and total severity score of ≥ 35 for randomization) was performed by the independent rater pool after completion of two preparatory sessions and any necessary psychiatric medication taper to establish baseline symptom severity following removal of psychiatric medications. At the end of the preparatory period, participants were assessed for final eligibility and enrollment was confirmed prior to randomization (Fig. 1).
The treatment period consisted of three 8-h experimental sessions of either MDMA-assisted therapy or therapy with inactive placebo control, spaced ~4 weeks apart. Following a 10-h fast, experimental sessions began with a qualitative urine drug screen, pregnancy screen if applicable, and C-SSRS, as well as measurement of baseline blood pressure, body temperature and heart rate immediately before the initial drug dose. Any positive findings on the urine drug screen that could not be attributed to pre-approved concomitant medications were reviewed by the medical monitor to assess compliance with ongoing eligibility criteria and for possible AESIs. Experimental sessions were conducted following a circadian rhythm-adjusted dosing schedule for a morning (~10:00 hours) initial dose.
In each experimental session the participants received a single divided dose of 80–180 mg MDMA or placebo. In the first experimental session, an initial dose of 80 mg was followed by a supplemental half-dose of 40 mg 1.5–2.5 h after the first dose. In the second and third experimental sessions, an initial dose of 120 mg was followed by a supplemental half-dose of 60 mg. If tolerability issues emerged with the initial dose or if participants declined, the protocol permitted the supplemental dose and/or dose escalation to be withheld. There were no instances in which the supplemental dose was withheld due to tolerability issues. Six participants chose either not to take the supplemental dose (n = 3, 1 MDMA) or not to escalate to the 120 mg dose (n = 3, 2 MDMA) in a total of six experimental sessions (2.3% of the total sessions across the study). Blood pressure, body temperature and heart rate were measured before the supplemental dose was given14.
Manualized therapy was conducted in accordance with the MDMA-assisted therapy treatment manual (http://maps.org/treatment-manual). Therapy was inner-directed and designed to invite inquiry and to facilitate therapeutic effect by providing support for approaching difficult material in a manner that would not interfere with the participant’s spontaneous experience. Every therapist held a Master’s degree or above, and the protocol requirement was that one person per therapy team was licensed to provide psychotherapy in accordance with state and local requirements. Therapists were additionally required to take part in the sponsor’s five-part training process, which consisted of an online course (15 h), a training course (5 d), experiential learning (3 d), role playing (1 d), and supervision (52 h).
Blood pressure, body temperature and heart rate were measured at the end of each experimental session prior to discharging the participant.
Each experimental session was followed by three 90-min integration sessions that were spaced ~1 week apart to allow the participant to understand and incorporate their experience. The first integration session always occurred on the morning after the experimental session, and the remaining two integration sessions occurred over the following 3–4 weeks (Fig. 1).
Independent raters conducted CAPS-5 and SDS assessments ~3 weeks after each of the first two experimental sessions. The primary outcome assessment was conducted ~8 weeks after the third experimental session (18 weeks after the baseline assessment), in which the independent raters collected the final CAPS-5 and SDS assessments. Twenty per cent of independent rater assessments were randomly selected and reviewed for fidelity. Lead independent raters evaluated the fidelity of all assessments related to enrollment failures as well as an additional 20% of remaining baseline CAPS-5 assessments. Diagnostic concordance between the raters had a Cohen’s kappa coefficient of 0.94, and reliability analysis of the CAPS-5 total severity scores showed a Spearman correlation coefficient of 0.98 (P < 0.0001), demonstrating high inter-rater reliability between the independent raters. The independent raters were all mental health professionals with graduate-level training in psychology, social work or counseling, at least 1 year of experience working with trauma-exposed populations, and had previous experience administering structured assessments.
Cases of non-compliance, protocol deviations, loss to follow-up, and other reasons for participant dropout were assessed for the presence of AESIs. There were two major protocol deviations (defined as the eligibility criteria not being met by the randomized participants during the course of the study). In the first protocol deviation a participant was not compliant with drug use lifestyle modifications on study, and in the second protocol deviation a participant disclosed cannabis use at study entry but abstained for the duration of the study. There was one dosing error in which a participant in the placebo group received 80 mg placebo as an initial dose and 100 mg as a supplemental dose (n = 1). Additionally, 14 participants (10 of whom were in the MDMA arm) requested further integrative visits, as permitted by the protocol.
The primary objective of this trial was to evaluate the efficacy and safety of MDMA-assisted therapy for PTSD compared with placebo with therapy, based on comparison of CAPS-5 total severity score at baseline with that at 18 weeks after baseline. The CAPS-5 is a semi-structured interview that assesses the index history of DSM-5-defined traumatic event exposure, including the most distressing event, to produce a diagnostic score (presence versus absence) and a PTSD total severity score. The CAPS-5 rates intrusion symptoms (intrusive thoughts or memories), avoidance, cognitive and mood symptoms, arousal and reactivity symptoms, duration and degree of distress, and dissociation. The CAPS-5 is scored on a scale from 0 to 80, with moderate PTSD defined from a rationally derived severity range of 23–34 (ref. 40), and severe PTSD as ≥35.
The secondary objective of this trial was to evaluate the efficacy of MDMA-assisted therapy for PTSD compared with placebo with therapy in clinician-rated functional impairment, as measured by the mean change in SDS total scores from baseline to 18 weeks after baseline. Exploratory outcome measures included the BDI-II, the Alcohol Use Disorders Identification Test (AUDIT), the Drug Use Disorders Identification Test (DUDIT) and the Adverse Childhood Experiences (ACE) Questionnaire.
Participants agreed to be recontacted for potential enrollment in a long-term follow-up study, which will include follow-up measures to assess the durability of the treatment. These data will be published at a later date.
Statistics and reproducibility
Statistical power calculations for the initial sample size were made by fitting an MMRM of CAPS-4 data (converted to the CAPS-5 scale and pooled from the phase 2 studies) to obtain variance and covariance parameter estimates. Using the estimated effect size and variance and covariance parameters, the sample size was calculated to achieve a power of 90% at an alpha of 0.049.
The intent-to-treat (ITT) set consisted of 91 randomized participants, however, one participant declined dosing on the morning of the session and provided no additional data, and therefore it was not possible to complete this analysis. Participants were randomized in a blinded fashion with 1:1 allocation as described in the section on randomization, masking and bias minimization above. The modified intent-to-treat (mITT) set consisted of 90 randomized participants who had completed at least one blinded experimental session and at least one post-treatment assessment. The mITT set consisted of 46 participants randomized to the MDMA group and 44 participants randomized to the placebo group, with identical therapy. The per protocol set (completers) consisted of all participants who completed three experimental sessions and assessments (MDMA, n = 42; placebo, n = 37) (Fig. 1).
The SAP was guided by the ICH E9 (R1) guidelines, which describe the use of estimands and sensitivity analyses to measure the effects of the drug if taken as directed (de jure, assessment of efficacy), and the effects of the drug if taken as assigned, regardless of adherence (de facto, assessment of effectiveness). The SAP was developed in accordance with FDA requirements and was approved by the European Medicines Agency to meet the requirements for future marketing applications. The primary and secondary efficacy analyses therefore utilized a de jure estimand of the mITT set for assessing treatment efficacy from the CAPS-5 and SDS data while on the study drug. The de jure dataset did not include outcome measurements taken after treatment discontinuation in the analysis of treatment efficacy. Missing data were not imputed.
One participant in the placebo group completed only the baseline assessment, and discontinued intervention but provided CAPS data at the T4 timepoint, ~18 weeks after baseline. Given that no endpoint assessment was collected prior to treatment discontinuation, this participant is excluded from the de jure estimand (leaving n = 89) but is included in the de facto estimand sensitivity analysis (for a total of n = 90). Two additional CAPS data points at the T4 timepoint, ~18 weeks after baseline, from two participants in the placebo group who provided these data following discontinuation of treatment, were not included in the de jure estimand (Supplementary Table 1).
The de facto estimand assessed the impact of these missing data points in the mITT set. That is, the CAPS measures at the T4 timepoint, ~18 weeks after baseline for the three placebo participants who discontinued treatment but provided off-treatment outcome assessments were included in a sensitivity analysis, which determined that inclusion of these measures in the model did not significantly alter the results.
The primary and secondary efficacy analyses were carried out using an MMRM that included all outcome data from baseline and the first, second and third experimental sessions. The efficacy of treatment was tested by comparing the change from baseline to the third experimental session in CAPS-5 and SDS scores between treatment groups in two-sided tests. The fixed effects were treatment (MDMA or placebo), baseline CAPS score, dissociative subtype and investigational site, with random effect specified as study participant.
A hierarchical testing strategy was used to control for type I error, such that the hypothesis for the key secondary endpoint (SDS) would be tested only if the statistical test for the primary efficacy comparison rejected the null hypothesis. An analysis of covariance (ANCOVA) to test the effects of study participation before versus after the COVID-19 pandemic declaration by the World Health Organization indicated a non-significant interaction and therefore was not included in the primary outcome model (Supplementary Table 2). The primary outcome analysis was replicated independently by one blinded programmer and one unblinded programmer.
An independent data monitoring committee monitored adverse events for safety and conducted one administrative interim analysis, after completion of enrollment and of 60% of primary endpoints to examine the adequacy of the sample size. The data monitoring committee recommended that no additional participants should be added, based on conditional power calculations supporting 90% statistical power, but in keeping with the SAP did not provide the sponsor with any information on the conditional power or effect size. The alpha level was set to 0.05, and 2% of the alpha (0.001) was spent on the interim analysis and 98% (0.0499) was left for the final analysis.
Statistics for the primary and secondary efficacy comparisons (CAPS and SDS) are reported as P values from the results of the MMRM analysis. In exploratory analyses, additional baseline covariates of age, gender, ethnicity, prior use of SSRIs, depression as measured by the BDI-II, adverse childhood experiences, and alcohol and substance use disorders were assessed in the model, with the threshold of significance set at P < 0.05 (Supplementary Table 1). BDI-II score was also assessed as an exploratory efficacy outcome measure with a paired, two-tailed t-test. Results are reported as mean (s.d.) throughout the text. Between-group effect size was calculated with Cohen’s d, and 95% CIs are reported. SAS version 9.4 (SAS Institute) was used for analyses.
The safety analysis included all participants who were given at least one dose of the study drug or placebo. The primary safety analysis evaluated TEAEs as a participant-level analysis. An association with MDMA was determined based on the relative incidence of TEAEs with at least a twofold difference between the MDMA and placebo groups.
Further information on research design is available in the Nature Research Reporting Summary linked to this article.