Neuroimaging of chronic MDMA (“ecstasy”) effects: A meta-analysis

MDMA (3,4-methylenedioxymethamphetamine) is an amphetamine-type drug that primarily releases serotonin and norepinephrine. Recreational use of MDMA rose from the 1980s and research interest has continued, including recent clinical investigation of MDMA-assisted psychotherapy for post-traumatic stress disorder. However, concerns remain that MDMA may be neurotoxic to serotonergic neurons in humans. Prior neuroimaging studies have been numerous but heterogeneous in samples, methods and findings, and most have sampled heavy, polydrug-using participants rather than moderate users. Müller and colleagues set out to provide a comprehensive meta-analysis of human neuroimaging studies examining non-acute (chronic or post-use) brain alterations associated with MDMA. The study aimed to aggregate findings across imaging modalities (MRI, MRS, PET, SPECT, etc.), include both current and former users, and to test whether imaging alterations in serotonin transporter (SERT) density were associated with lifetime episodes of use or time of abstinence, using meta-regression where feasible.

The investigators conducted a systematic search of PubMed, Embase and Web of Science from database inception to 24 August 2018, without language restrictions. Search terms combined MDMA/ecstasy with a broad set of imaging modalities. Study selection and data extraction were performed independently by two reviewers, with a third adjudicator used for disagreements. Inclusion criteria required studies to assess non-acute MDMA effects in humans, to compare MDMA users with controls, to apply structural, functional or neurochemical neuroimaging techniques (MRI and its variants, 1H-MRS, PET, SPECT), and to report sufficient data for meta-analysis or to provide data on request. Included studies were classified as region-of-interest (ROI) or whole-brain approaches and assigned to domains such as SERT density, dopamine transporter, serotonin 2A receptors, glucose metabolism, neurochemical markers, structural measures, resting-state and task-based conditions. When multiple ROIs overlapped within a study, preference was given to larger regions to avoid double counting. The reviewers planned whole-brain meta-analysis using seed-based d mapping if at least five studies were available per modality, but insufficient studies prevented that analysis. Meta-analytical procedures therefore focused on ROI data. Extracted variables included study identifiers, imaging method, sample sizes, recruitment, demographic data and detailed drug-history metrics (age at onset, duration, lifetime tablets/episodes/mg, usual and maximum dose per occasion, time since last use). When required data were missing, the team attempted to obtain values from authors, estimated values from figures or statistical reports, or converted reported statistics to means and SDs. Lifetime MDMA exposure was standardised to "episodes of use" to improve comparability; missing episodes were estimated by dividing cumulative tablets by usual dose per episode or by using a weighted mean of 3.0 tablets per episode when necessary. Statistical analysis used Hedges' g to calculate effect sizes (appropriate for small samples) and random-effects models with restricted maximum likelihood estimation. The R package metafor and OpenMEE were used, and leave-one-out analyses assessed robustness. Meta-regression examined moderators (lifetime episodes and time of abstinence) only for domains with sufficient assessments; this criterion was met for SERT density. To account for dependent outcomes within studies, multiple ROI results were combined per study using established procedures requiring an assumed inter-regional correlation. Publication bias was assessed with Rosenberg's fail-safe N rather than funnel plots because of small numbers of studies. Study quality was appraised using the NIH Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies.

The search identified 934 records; after screening and full-text review 16 independent studies met inclusion criteria, yielding a pooled sample of 356 MDMA users and 311 controls. Meta-analysis was feasible for five 1H-MRS studies (neurochemical ratios), three cerebral blood flow (CBF) studies of basal ganglia regions, and ten PET/SPECT studies investigating SERT density across multiple regions. Across all included studies the pooled mean lifetime episodes of ecstasy use was 165.5 (median 193.5, range 2.0–344.3), indicating predominantly heavy use in sampled cohorts. Magnetic resonance spectroscopy: Five studies contributed to analysis of N-acetylaspartate/creatine (NAA/CR) and myo-inositol/creatine (MI/CR) ratios. No significant differences between MDMA users and controls were found for NAA/CR in the occipital lobe (g = -0.03, 95% CI [-0.29, 0.22], p = 0.79) or mid-frontal lobe (g = -0.22, 95% CI [-0.63, 0.19], p = 0.30). MI/CR ratios also showed no significant group differences in occipital (g = -0.03, 95% CI [-0.36, 0.30], p = 0.86) or mid-frontal regions (g = 0.06, 95% CI [-0.22, 0.34], p = 0.68). Leave-one-out sensitivity analyses did not change the significance of these findings. Cerebral blood flow: Three studies (two MRI, one SPECT) examined CBF in globus pallidus and putamen (83 users, 81 controls). No significant CBF differences were observed in globus pallidus (g = 0.14, 95% CI [-0.80, 1.07], p = 0.78) or putamen (g = -0.20, 95% CI [-0.51, 0.11], p = 0.20). Leave-one-out analyses did not materially change these results. Serotonin transporter density: Ten studies provided SERT data from PET/SPECT, encompassing 267 users and 234 controls. Significant reductions in SERT density among MDMA users were observed in eight of 13 investigated regions: amygdala (g = -0.42, 95% CI [-0.78, -0.06], p = 0.02), anterior cingulate (g = -0.58, 95% CI [-0.99, -0.17], p < 0.01), posterior cingulate (g = -1.27, 95% CI [-2.10, -0.44], p < 0.01), hippocampus (g = -0.70, 95% CI [-1.29, -0.11], p = 0.02), occipital lobe (g = -1.17, 95% CI [-1.69, -0.65], p < 0.01), parietal lobe (g = -1.12, 95% CI [-1.52, -0.71], p < 0.01), temporal lobe (g = -1.05, 95% CI [-1.59, -0.50], p < 0.01), and thalamus (g = -0.32, 95% CI [-0.56, -0.08], p < 0.01). No significant differences were found in caudate, frontal lobe, insula, midbrain or putamen. Different radiotracers were used across studies; most applied [11C]DASB, some used [11C]McN, and two used [123I]β-CIT which binds less selectively and also to dopamine transporters. Excluding the study that used β-CIT for caudate and putamen did not alter the regional results. Leave-one-out analyses showed that removal of particular studies could render the effects for some regions (amygdala, anterior cingulate, hippocampus) non-significant, indicating limited robustness for those findings. Heterogeneity and moderator analyses: Several SERT regions showed substantial heterogeneity (I2 values often > 60%), and high heterogeneity was also seen for CBF in globus pallidus. Meta-regression found no association between aggregated SERT effect sizes and lifetime episodes of MDMA use (ß = 0.001, Z = 0.27, p = 0.79). A positive association between SERT density and time of abstinence was observed (ß = 0.001, Z = 2.11, p = 0.04), which could suggest partial recovery with longer abstinence; however, this association was driven largely by a single study, and its removal abolished the effect (ß = 0.000, Z = 0.44, p = 0.66). Publication bias and study quality: Rosenberg's fail-safe N suggested that many unpublished null studies would be needed to nullify significant SERT findings in several regions (e.g. occipital lobe: 125, temporal lobe: 72), though fail-safe N was much lower for amygdala (12) and thalamus (13). Quality assessment indicated that most included studies were of poor quality, with common issues including recruitment from different populations for users and controls, widespread polydrug use among users, small sample sizes and imprecise exposure measurement.

Müller and colleagues interpret their findings as indicating reduced SERT availability in MDMA user groups in multiple brain regions, while failing to find consistent differences in neurochemical markers measured by MRS or in basal ganglia CBF. The authors note that reductions in SERT density were observed in eight of 13 regions examined, but that results for some regions were sensitive to the inclusion or exclusion of individual studies. Several possible explanations for lower SERT density are discussed. Decreases could reflect MDMA-related neurotoxic loss of serotonergic terminals, a conclusion supported by animal work, but transporter downregulation in response to repeated serotonergic stimulation is an alternative account. The absence of a relationship between lifetime episodes and SERT reductions led the authors to suggest that per-occasion dose may be more relevant than cumulative episodes. The observed positive association between abstinence and SERT density may indicate partial recovery, but this result is uncertain because it depended on a single influential study and because recovery could be regionally variable and may not reflect full restoration of normal function. The authors highlight that included samples predominantly comprised heavy users who consume substantially more MDMA than typical users, so the findings may overestimate effects in the broader MDMA-using population. They caution that these cohorts are not representative of the dosing patterns used in therapeutic MDMA studies, which typically involve few controlled doses. Strengths of the meta-analysis include comprehensive coverage of modalities and inclusion of both current and former users, plus efforts to standardise lifetime exposure across studies. Key limitations arise from the primary literature: small study sizes, observational designs, heterogeneous and often inadequate control for polydrug use, variation in time since last use, uncertainties in exposure measurement, and overall low study quality. The authors also note that techniques such as MRS and CBF are not specific measures of serotonergic neurotoxicity and have not been validated for that specific purpose. They conclude by recommending future studies of higher methodological quality, better matching of controls to user populations, more research on low-to-moderate users and former users, and efforts to disentangle neurotoxic loss from transporter regulation.

The meta-analysis concludes that, despite over two decades of neuroimaging research on MDMA, studies are heterogeneous and limited in number and quality. No consistent evidence was found for alterations in basal ganglia CBF or in occipital and frontal neurochemical markers, whereas reductions in SERT density were observed in several brain regions. The absence of an association with lifetime episodes suggests that other factors such as dose per occasion could be more important, and reductions in SERT may alternatively reflect non-MDMA factors. The authors call for better-designed, higher-quality studies, improved control for confounding (particularly polydrug use), more research on moderate and former users, and investigations aimed at clarifying whether SERT reductions reflect neurotoxicity or transporter downregulation.