Prediction of MDMA response in healthy humans: a pooled analysis of placebo-controlled studies

MDMA (3,4-methylenedioxymethamphetamine) is both a recreational drug and an investigational adjunct to psychotherapy, producing acute effects such as enhanced mood, openness, trust and empathy. Prior work has shown that responses to psychoactive substances are influenced not only by pharmacology (dose, pharmacokinetics, genetics) but also by non‑pharmacological factors often grouped as set (personality, mood, expectations) and setting (physical and social environment). Although several small studies have examined isolated predictors of MDMA response, few have jointly evaluated a broad range of pharmacological and non‑pharmacological factors while adjusting for confounders. Studerus and colleagues set out to quantify the relative contributions of multiple predictor domains to the acute physiological and psychological response to MDMA. Using pooled individual‑level data from 10 randomized, double‑blind, placebo‑controlled, cross‑over experiments performed in the same laboratory, the study aimed to test whether drug dose and exposure, demographic variables, prior drug experience, CYP2D6 genotype, personality traits and pre‑intake mood states predict acute MDMA effects. This is presented as the first comprehensive evaluation of such predictors in a single controlled dataset of healthy volunteers.

This pooled analysis combined raw data from 10 double‑blind, placebo‑controlled, cross‑over studies conducted at the University Hospital Basel between 2009 and 2018. The sample comprised 194 healthy volunteers (97 female) aged 18–45 years (mean 25.1 ± 4 years). Most participants were MDMA‑naïve or had only minimal prior use; exclusion criteria included psychiatric disorder, significant physical illness, and recent or heavy illicit drug use. All studies used washout periods of at least 7 days and the analysis used only the MDMA‑alone and placebo sessions. MDMA was administered orally as ±MDMA hydrochloride at single doses of either 75 mg or 125 mg; males and females received the same nominal doses. Dose per body weight (mean 1.7 ± 0.4 mg/kg, range 0.8–2.7 mg/kg) was treated as a covariate. Predictor variables included sex, age, dose per kg, number of prior MDMA uses, CYP2D6 genetic activity score, mood states before intake measured by the Adjective Mood Rating Scale (AMRS), personality traits from the NEO‑FFI (Neuroticism, Extraversion, Openness, Agreeableness, Conscientiousness), and trait anxiety from the STAI‑T. Primary response variables comprised pharmacokinetics (plasma MDMA concentrations sampled up to 6 h and summarised as AUC0–6h), autonomic measures (systolic/diastolic blood pressure, heart rate, tympanic temperature with peak values Emax used as outcomes), and multiple subjective measures. Subjective assessments included repeated Visual Analogue Scales (VAS) for constructs such as "any drug effect", "good/bad drug effect", "closeness" and "liking" with area under the effect‑time curve (AUEC) as the summary measure, AMRS repeated post‑drug, and the 5D‑ASC questionnaire administered 6 h post‑dose (etiology‑independent dimensions: Oceanic Boundlessness, Dread of Ego Dissolution, Visionary Restructuralization, plus constructed subscales for "impaired control and cognition" and "anxiety"). Statistical analysis used R. Missing data were multiply imputed (MICE) with 20 imputations; analyses were fitted in each completed dataset and pooled via Rubin’s rules except for LASSO models. To account for study clustering, linear mixed‑effects models with random intercepts for study were used. Predictor and response variables were z‑transformed; semi‑partial R2 quantified variance explained by fixed effects. P‑values were adjusted for multiple testing using the Benjamini‑Hochberg procedure. To identify parsimonious predictor sets and relative importance, LASSO regression (penalized package) was employed: optimal lambda was chosen via grid search and bootstrap estimation of out‑of‑bag mean squared error. LASSO models used single imputation and did not account for study clustering for simplicity.

In univariable (unadjusted) mixed models, MDMA plasma concentration (AUC0–6h) was the strongest predictor of both physiological and psychological outcomes. It was significantly associated with 16 of 24 outcome variables before multiple‑test correction and with eight after correction. The largest standardised associations (β, corrected p, semi‑partial R2) included VAS "any drug effect" (β = 0.48, corrected p < 0.001, R2 = 0.35), mean arterial pressure (β = 0.35, corrected p < 0.001, R2 = 0.25), VAS "liking" (β = 0.36, corrected p < 0.001, R2 = 0.17), VAS "good drug effects" (β = 0.35, corrected p < 0.001, R2 = 0.19), VAS "stimulated" (β = 0.30, corrected p = 0.001, R2 = 0.44), VAS "high mood" (β = 0.29, corrected p = 0.002, R2 = 0.43), AMRS "introversion" change (β = 0.24, corrected p = 0.032, R2 = 0.21), and 5D‑ASC "oceanic boundlessness" (β = 0.25, corrected p = 0.045, R2 = 0.06). Because plasma concentration is not available at the time of intake, subsequent adjusted models used drug dose per body weight as the pharmacological covariate. After adjustment and multiple‑testing correction, a genetically lower CYP2D6 activity predicted higher MDMA plasma concentration (β = -0.19, corrected p < 0.001, R2 = 0.10). Older age was associated with a smaller MDMA‑induced heart rate increase (β = -0.27, corrected p = 0.019, R2 = 0.07). Personality and mood predictors retained significant relationships with specific subjective outcomes: higher Openness predicted a stronger decrease in AMRS "general inactivation" (β = -0.25, corrected p = 0.042, R2 = 0.06), greater VAS "closeness" (β = 0.28, corrected p = 0.019, R2 = 0.08), and higher 5D‑ASC scores for "oceanic boundlessness" (β = 0.22, corrected p = 0.042, R2 = 0.06) and "visionary restructuralization" (β = 0.27, corrected p = 0.006, R2 = 0.09). Higher Neuroticism and higher trait anxiety (STAI‑T) predicted greater 5D‑ASC "dread of ego dissolution" (both β = 0.32; corrected p = 0.001 and 0.004; R2 = 0.12 and 0.11) and "impaired control and cognition" (both β = 0.26; corrected p = 0.014 and 0.042; R2 = 0.08 and 0.07). Pre‑intake AMRS "anxiety‑depressiveness" and "introversion" predicted higher 5D‑ASC "anxiety" scores (β = 0.25 and 0.23; corrected p = 0.042 and 0.042; R2 = 0.06 and 0.05). LASSO models selected, on average, 8.6 predictors per response (range 5–13). Dose per body weight was the most frequently selected predictor (identified for 21 of 25 response variables) and held the largest standardised coefficient for 12 outcomes, especially positive/neutral subjective measures and 5D‑ASC "oceanic boundlessness". By contrast, unpleasant or anxious reactions were better predicted by non‑pharmacological variables: older age, trait anxiety, pre‑intake anxiety‑depressiveness, and Neuroticism emerged as the most important predictors for "bad drug effects" and adverse 5D‑ASC subscales. Sex and prior MDMA use did not predict outcomes after adjusting for dose per kg.

Studerus and colleagues interpret the results as evidence that both pharmacological and non‑pharmacological factors influence acute MDMA effects, with pharmacokinetic exposure (plasma concentration) being the dominant predictor for many outcomes. Dose per body weight served as a practical proxy for plasma exposure and was frequently the most important predictor; CYP2D6 genotype additionally affected plasma concentration, consistent with known metabolic effects. Psychological responses showed differential patterns: higher Openness to experience amplified pleasant, prosocial and perceptual aspects of the MDMA experience (greater closeness, decreased inactivation, higher oceanic boundlessness and visionary restructuring), whereas higher Neuroticism and trait anxiety, and higher pre‑intake anxiety‑depressiveness or introversion, were associated with more anxious or challenging reactions (dread of ego dissolution, impaired control/cognition, 5D‑ASC anxiety). The authors note these relationships mirror findings from psychedelic studies and suggest an overlap in how set variables shape serotonergic drug responses. The discussion acknowledges limitations that constrain generalisability. The sample was young, healthy and largely MDMA‑naïve, so results may not translate directly to patient populations with psychiatric disorders or to heavier recreational users. Dose variety was limited (predominantly 125 mg), reducing exploration of dose–response heterogeneity. The standardised hospital research setting minimised variation in the physical and social environment, so setting effects could not be evaluated. Finally, the analysis focused on acute effects and cannot inform longer‑term outcomes. In terms of implications, the authors propose that awareness of both pharmacological factors (dose/kg, CYP2D6) and psychological traits/mood states could inform safety optimisation and preparation in MDMA‑assisted psychotherapy, for example by tailoring preparation and patient selection to reduce adverse reactions and enhance potentially therapeutic prosocial effects. They also highlight consistency with prior psychedelic research and the possibility that MDMA treatment itself may shift personality traits in therapeutic contexts.

The pooled analysis shows that MDMA response in healthy volunteers is shaped by both drug exposure and individual psychological characteristics. While dose per body weight (and resulting plasma concentration) explained much of the variance in positive and neutral effects, personality traits and pre‑intake mood—especially Openness, Neuroticism and state anxiety—contributed importantly to the likelihood of pleasant versus anxious experiences. These findings could inform the planning and safety considerations for MDMA‑assisted psychotherapy by guiding dosing decisions and psychological preparation.