MDMA pharmacokinetics: A population and physiologically based pharmacokinetics model‐informed analysis

Abstract

Midomafetamine (3,4‐methylenedioxymethamphetamine [MDMA]) is under the U.S. Food and Drug Administration review for treatment of post‐traumatic stress disorder in adults. MDMA is metabolized by CYP2D6 and is a strong inhibitor of CYP2D6, as well as a weak inhibitor of renal transporters MATE1, OCT1, and OCT2. A pharmacokinetic phase I study was conducted to evaluate the effects of food on MDMA pharmacokinetics. The results of this study, previously published pharmacokinetic data, and in vitro data were combined to develop and verify MDMA population pharmacokinetic and physiologically based pharmacokinetic models. The food effect study demonstrated that a high‐fat/high‐calorie meal did not alter MDMA plasma concentrations, but delayed T _max. The population pharmacokinetic model did not identify any clinically meaningful covariates, including age, weight, sex, race, and fed status. The physiologically based pharmacokinetic model simulated pharmacokinetics for the proposed 120 and 180 mg MDMA HCl clinical doses under single‐ and split‐dose (2 h apart) conditions, indicating minor differences in overall exposure, but lower AUC within the first 4 h and delayed T _max when administered as a split dose compared to a single dose. The physiologically based pharmacokinetic model also investigated the drug–drug interaction magnitude by varying the fraction metabolized by a representative CYP2D6 substrate (atomoxetine) and evaluated inhibition of renal transporters. The simulations confirm MDMA is a potent CYP2D6 inhibitor, but likely has no meaningful impact on the pharmacokinetics of drugs sensitive to renal transport. This model‐informed drug development approach was employed to inform drug–drug interaction potential and predict pharmacokinetics of clinically relevant dosing regimens of MDMA.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Midomafetamine capsules containing 3,4‐methylenedioxymethamphetamine (MDMA) is an investigational new drug under review for treatment of post‐traumatic stress disorder (PTSD) in adults. MDMA is characterized by nonlinear pharmacokinetics, primarily eliminated by hepatic metabolism via cytochrome P450 (CYP)2D6 with minimal renal contribution, and rapidly inhibits CYP2D6 and its own metabolism.

• WHAT QUESTION DID THIS STUDY ADDRESS?

Is MDMA pharmacokinetics meaningfully altered when administered with food? How can population pharmacokinetic (PopPK) and physiologically based pharmacokinetic (PBPK) modeling be leveraged to characterize MDMA pharmacokinetics with respect to dosing and effects of intrinsic and extrinsic factors, including DDIs, while taking into account that MDMA is a potent CYP2D6 inhibitor?

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

This analysis adds to our knowledge of MDMA pharmacokinetics, specifically that MDMA can be administered without regard to food intake; split dosing decreases plasma concentration of MDMA within the first 4 h, but not overall exposure; and MDMA is a potent CYP2D6 inhibitor, indicating care should be taken when co‐administering MDMA with other CYP2D6 substrates.

• HOW MIGHT THIS CHANGE DRUG DISCOVERY, DEVELOPMENT, AND/OR THERAPEUTICS?

Integration of PopPK and PBPK modeling in a model‐informed drug development approach provides clinically relevant pharmacokinetic information that would not be available through analysis of the clinical trial pharmacokinetic data alone.

INTRODUCTION

Midomafetamine capsules containing 3,4‐methylenedioxymethamphetamine (MDMA) are currently under review by the U.S. Food and Drug Administration (FDA) for treatment of post‐traumatic stress disorder (PTSD) in adults. MDMA in combination with psychological intervention (MDMA‐assisted therapy) was generally well tolerated and met primary end points in two phase III studies of participants with PTSD. ¹ , ² MDMA, an entactogen, is a ring‐substituted phenethylamine that increases synaptic concentration of serotonin, norepinephrine, and dopamine by inhibiting reuptake and inducing release at monoamine transporters. ³ , ⁴ MDMA is primarily eliminated by hepatic metabolism, largely via cytochrome P450 2D6 (CYP2D6) with minimal renal contribution; however, MDMA is also a strong and rapid inhibitor of CYP2D6 and thus inhibits its own metabolism. ⁵ , ⁶ , ⁷ , ⁸ Therefore, while MDMA is a precipitant (perpetrator) of potent drug–drug interactions (DDIs) via CYP2D6 mechanism‐based inhibition (MBI), it is less susceptible as a substrate as it inhibits its own metabolism. ⁹ , ¹⁰ , ¹¹ , ¹²

The proposed clinical dosing regimen consists of three separate split‐dose administrations of MDMA capsules, spaced at least 21 days apart. Maximum split doses are 120 mg MDMA HCl (80 + 40 mg) for first session and 180 mg MDMA HCl (120 + 60 mg) for second and third sessions. The second portion of the split dose is administered 1.5–2 hours (h) after the first. A full course of MDMA treatment includes three dosing sessions spread over 3 months and does not require daily dosing or steady‐state plasma concentrations to be effective. The pharmacokinetics of single‐dose MDMA were published ⁷ , ⁸ and one study examined split dosing (2 h apart) following 150 mg MDMA HCl (50 + 100 mg), which differs from doses administered in PTSD clinical trials. ¹³

A two‐period crossover food effect study (MPKF) was conducted in healthy volunteers (HVs) to determine the effect of a high‐fat/high‐calorie meal on absorption following a single 120 mg MDMA HCl dose. A population pharmacokinetic (PopPK) model of MDMA and its active minor metabolite 3,4‐methylenedioxyamphetamine (MDA) in HVs was developed using single‐dose data from MPKF and a previously published National Institute on Drug Abuse (NIDA) pharmacokinetic study. ⁷ , ⁸ The PopPK model evaluated the impact of covariates on available clinical data. Additionally, a physiologically based pharmacokinetic (PBPK) model of MDMA was developed based on in vitro and clinical pharmacokinetic data. The PBPK model simulated MDMA plasma concentration–time profiles following single‐dose and split‐dosing regimens in HVs while taking CYP2D6 and renal transporter inhibition into account. The PBPK model also evaluated the potential for CYP2D6‐ and transporter‐mediated DDIs with MDMA as a precipitant.

Using a model‐informed drug development (MIDD) approach, complimentary models were applied to better understand pharmacokinetics of clinically relevant MDMA dosing regimens and the effects of intrinsic and extrinsic factors.

MATERIALS AND METHODS

Pharmacokinetic studies

Pharmacokinetic modeling was conducted using data from two phase I studies (Figure 1): MPKF, an open‐label, randomized sequence, two‐period crossover pharmacokinetic study that assessed the effect of food on MDMA and MDA pharmacokinetics in HVs (previously unpublished, Figure S1); and the NIDA study, a randomized, balanced, double‐blind, placebo‐controlled, within‐subject crossover MDMA pharmacokinetic study in HVs, as previously described. ⁷ , ⁸ , ¹⁴ , ¹⁵ A summary of the study designs, dose, and sampling times are provided in Supplementary Materials. Both studies were designed and conducted in accordance with the Declaration of Helsinki, including the International Council for Harmonization Good Clinical Practice guidelines. All participants provided written informed consent prior to study participation.

FIGURE 1.

Pharmacokinetic studies and models' overview. HCl, hydrochloride; MDA, 3,4‐methylenedioxyamphetamine; MDMA, 3,4‐methylenedioxymethamphetamine; N, number of participants in analysis population; PBPK, physiologically based pharmacokinetic; PK, pharmacokinetic; PopPK, population pharmacokinetic.

PopPK model

Model development

Data from MPKF and NIDA studies were combined in a single modeling dataset and included dosing information, pharmacokinetic sampling information, demographics, clinical laboratory values, and other covariate information collected in the clinical studies. PopPK analysis was performed using nonlinear mixed‐effects modeling (NONMEM® version 7.4; ICON, Hanover, MD, US) using first‐order conditional estimation with interaction method. R (version 4.0.2) was used for data preparation, graphical analysis, model diagnostics, statistical summaries, and simulations. Xpose® (version 0.4.11) and Perl‐speaks‐NONMEM® (PsN version 4.8.1; Department of Pharmacy, Uppsala University, Uppsala, Sweden) were also used for model diagnostics.

For MDMA, one‐, and two‐compartment models, as well as first‐order and sequential zero‐order, followed by first‐order absorption processes were evaluated. MDA was developed as a sequential model, with one‐, two‐, and three‐compartment models tested. Inter‐individual variability (IIV) was included in model parameters using an exponential error model (Equation 1).

$${\theta }_{\mathit{ki}}={\theta }_k\times e^{{\eta }_{\mathit{ki}}}$$ (1)

where θ _ki denotes the kth parameter value for the ith patient, θ _k denotes the typical parameter value, and η _ki denotes the inter‐individual random effect for the ith patient – assumed to have mean of 0 (zero) and variance ω _k ². The estimates of IIV are provided as percent coefficient of variation calculated as √ω ² × 100 (%). Residual error was modeled as additive, proportional, or combined additive and proportional error model (Equation 2).

$$Y_{\mathit{ij}}=C_{\mathit{ij}}\times \left(1+\epsilon 1_{\mathit{ij}}\right)+\epsilon 2_{\mathit{ij}}$$ (2)

where Y _ij is the jth observed concentration for the ith subject, C _ij is the corresponding predicted concentration, and 𝜀1 _ij (proportional) and 𝜀2 _ij (additive) are the residual errors under the assumption that 𝜀~N(0, σ2). Structural model selection was based on visible improvement in the goodness‐of‐fit (GoF) plots, plausibility, and precision of the parameter estimates, and decrease in the objective function value. All parameter estimates are reported with a measure of estimation uncertainty.

Covariate model

Using the base PopPK model, a covariate analysis was performed. The following covariates were evaluated: age, body weight, alanine aminotransferase, albumin, creatinine clearance, race, sex, fed status, and study (Table S1). The subset of covariate–parameter relationships to be included was selected based on exploratory graphical analysis, mechanistic plausibility, and scientific and clinical interest. The number of participants in each categorical covariate category and any correlation between covariates were considered as criteria for inclusion in the model.

Covariates were tested in a stepwise forward addition and then backward elimination process with significance levels of 0.01 and 0.001, respectively. Continuous covariates were incorporated using a scaled structure (median value or a standard value of the covariate [e.g., 70 kg for body weight]) (Equation 3). Categorical covariates were incorporated using a proportional structure with the most common level of the covariate being the reference (Equation 4).

$$P_{\mathit{ki}}={\theta }_k\times {\left(\frac{X_{\mathit{ij}}}{M\left(X_j\right)}\right)}^{{\theta }_j}$$ (3)

$$P_{\mathit{ki}}={\theta }_k\times {\left(1+{\theta }_j\right)}^{X_{\mathit{ij}}}$$ (4)

where P _ki is the population estimate of the parameter P _k for subject i, X _ij is the value of continuous covariate X _j for subject i, or an indicator variable for subject i for categorical covariate X _j with value 1 for the nonreference category and 0 for the reference category, M(X _j ) is the median of covariate X _j in the analysis dataset, θ _k is the typical value of the parameter P _k , and θ _j is a coefficient that reflects the effect of covariate X _j on the parameter. Finally, the effect of covariates following backward elimination was assessed and those lacking in clinical relevance (i.e., <10% change) were excluded from the final model. Alternative variance–covariance structures for Ω were evaluated, including partial‐ and full‐block structures, and retained only if statistically significant (p < 0.001) and if it improved model stability.

Model evaluation and validation

A nonparametric bootstrap (n = 1000) analysis ¹⁶ was conducted (random resampling of subjects from the original dataset with replacement) to evaluate the stability of the final model and to estimate 95% confidence intervals (CIs) for the model parameters.

Visual predictive checks (VPCs, n = 1000 simulations) with prediction correction evaluated the predictive ability of the final model ¹⁷ and plots of observed data distributions were compared to simulated distributions to demonstrate the model's ability to adequately predict the data on which the model is based.

PBPK model

Model development

PBPK modeling and simulation was conducted with Version 21 of the Simcyp Population‐Based Simulator. A minimal PBPK model, which considers liver and intestinal metabolism and includes a first‐order absorption model, was developed; see Table S2 for key input parameters. The fraction of MDMA absorbed was estimated from in vitro permeability data (high in vitro solubility) and a first‐order absorption rate constant (k_a) and lag‐time (t _lag) were optimized via sensitivity analysis (to match T _max).

As MDMA is metabolized by and is a potent inhibitor of CYP2D6, the MBI parameters were optimized using a minimal base model and in vivo apparent oral clearance (CL/F) and renal clearance (CL_R). Dextromethorphan ⁹ (CYP2D6 substrate) set the MDMA CYP2D6 MBI parameters for simulations. The K _i value was fixed according to the average of available literature reports, while a manual sensitivity analysis was completed to optimize the k _inact value to best replicate the observed interaction. The final model was developed including the interaction between paroxetine (CYP2D6 MBI) and MDMA to optimize the clearance and CYP2D6‐mediated metabolism components of the MDMA model. Observed CL/F (36.3 L/h) and CL_R (9.40 L/h) data obtained from HVs following a single oral dose of 100 mg MDMA HCl ¹¹ in the absence of paroxetine determined an initial hepatic intrinsic clearance (CL_int,H) value using the retrograde calculation. Total CL_int,H and the fm_CYP2D6 were simultaneously optimized to accurately capture both MDMA PK and the observed interaction with paroxetine. ¹¹ Competitive inhibition of CYP2D6 by MDMA (in vitro enzyme competitive inhibition constant value for unbound MDMA, K _i,u ) was also incorporated along with inhibition of renal transporters OCT1, OCT2, and MATE1.

Model evaluation and validation

The model was verified via evaluation of single (MPKF and NIDA source data) and split doses ¹³ of MDMA in HVs by comparison of the simulated MDMA plasma profiles (visual check) and parameters to observed and literature data (results not shown). The contribution of CYP2D6 to the overall clearance of MDMA was verified using data from the literature by evaluating MDMA's interaction with bupropion ¹² (results not shown).

Simulations

Predictions of MDMA plasma profiles, clearance, and DDIs were performed in the Simcyp Simulator in a simulated Caucasian HV population (default parameter for Simcyp model) ¹⁸ assigned as 56.8% CYP2D6 extensive metabolizers (EM), 32.5% intermediate metabolizers (IM), 8.2% poor metabolizers (PM), and 2.5% ultra‐rapid metabolizers (UM). Two sub‐populations were established comprised of (1) solely CYP2D6 EMs and (2) only CYP2D6 EM and IM subjects. These were included in simulations as needed to match the reported clinical trial population. For all simulations, trial design and demographics were matched to the clinical study. Unless otherwise stated, 20 virtual trials for each study were run to assess variability across groups. The simulated MDMA concentration–time profiles were compared to observed data. The Simcyp HV population was used unless otherwise stated.

Simulations following single and split MDMA doses were conducted using 10 virtual trials of 10 HVs (50% female) aged 20–50 years to derive the pharmacokinetic parameters: C _max at steady state, AUC_0–44 (44 h after dose, corresponding to approximately five MDMA half‐lives), and AUC_0–4 (4 h after dose, acute exposure).

A sensitivity analysis evaluated the effect of varying the fm_CYP2D6 (0.05–0.95) on predicted DDI magnitude using atomoxetine as a representative‐sensitive CYP2D6 substrate. ¹⁹ The default Simcyp compound library file for atomoxetine was used with the clearance converted to CYP2D6 CL_int and an additional HLM CL_int via the retrograde calculation.

MDMA was a weak inhibitor of renal transporters MATE1, OCT1, and OCT2 during in vitro experiments (Table S3). To simulate pharmacokinetic parameters of metformin, a substrate sensitive to renal transport ²⁰ was selected as a victim probe of MATE1, OCT1, and OCT2. Ten virtual trials of 10 HVs (50% female) aged 20–50 years receiving a single oral 500 mg metformin dose in the absence of MDMA and with the first dose of a 180 mg split dose of MDMA HCl were generated applying the default Simcyp compound library files for metformin. To account for uncertainty in the in vitro K _i values for renal transporters OCT1, OCT2, and MATE1, sensitivity analyses up to 3‐fold (MATE1) and 15‐fold (OCTs) lower than the in vitro value were completed.

RESULTS

Pharmacokinetic studies

Participant demographics and dataset

Study participants' demographic characteristics are summarized in Table S1. MDMA and MDA concentrations were analyzed in 2012 and 1871 samples, respectively, from 65 adult HVs across studies MPKF (n = 16) and NIDA (n = 49). Results were excluded if below the limit of quantification (BLQ), had missing time, or pre‐dose concentrations. One participant terminated from the MPKF study due to vomiting within 4 h of MDMA administration (fasted) and was not included in MPKF food effect analysis, but was included in the PopPK dataset for fasted state (n = 16).

Pharmacokinetic results

The MPKF pharmacokinetic parameters for MDMA and MDA when administered under fasted (reference) and fed (test) states are presented in Table 1. MDMA and MDA plasma concentrations were not affected by a high‐fat/high‐calorie meal. Ratios of geometric least square means (LSM) for C _max and AUC_0–inf were contained within the 80%–125% interval (ranged between 0.92 and 1.04 for both analytes). A delay in time to peak MDMA concentrations (T _max) was observed when MDMA was administered in the fed state compared to the fasted state (median T _max = 4 and 2 h, respectively), but was not tested for statistical significance. MDA T _max was similar in both fed and fasted states following MDMA administration. All adverse events (AE) were mild in severity and resolved. NIDA study results were previously reported. ⁷ , ⁸ , ¹⁴ , ¹⁵

TABLE 1.

MPKF study: Summary of food effect on MDMA and MDA pharmacokinetic parameters.

MDMA plasma pharmacokinetic parameters		Fed (N = 14)	Fasted (N = 15)	LSM geometric mean ratio (90% CI)
AUC0–inf (h*ng/mL)	Geo. Mean (Geo. CV%)	3318 (43.3)	3388 (38.4)	0.99 (0.94–1.04)
	Median [Min, Max]	3338 [1918, 8923]	3729 [1815, 7881]
C max (ng/mL)	Geo. Mean (Geo. CV%)	227 (21.7)	238 (24.7)	0.97 (0.92–1.01)
	Median [Min, Max]	230 [163, 395]	234 [160, 440]
T max (h)	Median [Min, Max]	4.0 [2.0, 6.0]	2.1 [2.0, 8.0]	–

MDA plasma pharmacokinetic parameters		Fed (N = 14)	Fasted (N = 15)	LSM geometric mean ratio (90% CI)
AUC0–inf (h*ng/mL)	Geo. Mean (Geo. CV%)	369 (33.3)	330 (37.6)	0.99 (0.93–1.06)
	Median [Min, Max]	343 [233, 710]	358 [189, 666]
C max (ng/mL)	Geo. Mean (Geo. CV%)	12.1 (22.6)	12.4 (25.4)	0.96 (0.90–1.03)
	Median [Min, Max]	11.9 [8.1, 19.3]	12.5 [7.0, 20.1]
T max (h)	Median [Min, Max]	7.0 [4.0, 12.0]	6.0 [2.1, 8.0]	–

Abbreviations: AUC_0–inf, area under the plasma concentration–time curve from time 0 to infinity; C _max, maximum observed plasma concentration; CV, coefficient of variation; LSM, least square means; MDA, 3,4‐methylenedioxyamphetamine; MDMA, 3,4‐methylenedioxymethamphetamine; N, number of participants in analysis population; T _max, time of observed maximum concentration.

PopPK model

PopPK model analysis

The final models for MDMA and MDA were one‐ and two‐compartment models, respectively, both with linear clearance. MDMA had a parallel first‐ and zero‐order absorption model and MDA data were modeled sequentially based on individual post hoc estimates from the final MDMA model. For the MDMA model, IIV was included on CL/F and apparent central volume of distribution (V2/F) with an omega block to capture this correlation. The final MDA model included IIV on metabolite CL, Q, and peripheral volume (V4) with an Omega block to capture the correlation.

Statistically significant covariates in the MDMA model were body weight, fed/fasted status, and study effect. There was an increase in MDMA CL/F and V/F with increasing body weight. Since the NIDA study subjects received a light breakfast, and not a high‐fat/high‐calorie meal as in the MPKF study, the final model included the NIDA subjects as a separate category from MPKF fed and fasted states. There was 49% slower absorption in fed subjects and 193% faster absorption in fasted subjects in MPKF compared to subjects in the NIDA study receiving a light breakfast. However, the relative standard error (RSE) for the fasted effect was 72.3%; therefore, this result should be interpreted with caution. Due to observed pharmacokinetic differences between studies, an additional empirical study effect was added on clearance and volume parameters. There was 36% lower MDMA CL/F and 26% lower MDMA V2/F for MPKF subjects (compared to NIDA). Finally, due to apparent study differences, separate additive residual errors were included for each study. Statistically significant covariates in the MDA model were body weight and sex: an increase in MDA CL and Q with increasing body weight and 18% lower V4 (MDA peripheral volume of distribution) in female subjects (compared to male).

The final PopPK model parameters for MDMA and MDA are provided along with the bootstrap estimates (Table 2). Prediction‐corrected VPCs (pcVPCs) for the final MDMA and MDA PopPK models (Figure S2) demonstrate the final PopPK models predicted the observed median and 5th and 95th percentiles (p5 and p95) of observed concentrations with good accuracy and the median profiles are fully captured within the 5–95th prediction interval (PI) of the median of simulations for MDMA and MDA across both studies.

TABLE 2.

PopPK model: parameters for MDMA and MDA.

Parameter	MDMA pharmacokinetic parameters and random effects
	Estimates	%RSE	Bootstrap median (95% CI)	Random effects	Estimates (%CV)
CL/F (L/h)	41.5	4.70	41.8 (37.7, 45.8)	IIV on CL/F	30.5 a
V2/F (L)	462	3.46	461 (429, 496)	IIV on V2/F	19.9 a
ka (1/h)	1.17	8.26	1.17 (0.969, 1.43)	IIV on KA	60.2
ALAG1 (h)	0.322	5.15	0.323 (0.273, 0.358)
D1 (h)	0.304	18.2	0.271 (0.141, 0.485)	IIV on D1	153
fmet	0.1 FIXED	–	–
Prop. Error (ng/mL)	0.207	9.24	0.208 (0.164, 0.25)
Add. Error NIDA (ng)	8.11	20.8	7.69 (4.47, 13.3)
Add. Error MPKF (ng)	2.5 FIX	–	–
Fed status effect on ka	−0.484	18.6	−0.496 (−0.645, −0.239)
Fasted status effect on ka	1.93	72.3	1.08 (0.277, 4.31)
Body weight effect on CL/F	1.03	17.2	1.05 (0.627, 1.4)
Study effect on CL/F	−0.36	16.40	−0.359 (−0.474, −0.23)
Body weight effect on V2/F	0.885	13.1	0.884 (0.6, 1.11)
Study effect on V2/F	−0.263	14.40	−0.267 (−0.342, −0.19)

Parameter	MDA pharmacokinetic parameters and random effects
	Estimates	%RSE	Bootstrap median (95% CI)	Random effects	Estimates (%CV)
CLM (L/h)	34.9	3.7	36.1 (33.5, 38.8)	IIV on CLM	32.0 b , c
V3 (L)	20.8	22.6	20.8 (11.9, 28.4)
QM (L/h)	79.6	10.3	83.9 (68.8, 103)	IIV on QM	46.8 b , d
V4 (L)	213	7.66	213 (181, 246)	IIV on V4	35.2 c , d
Prop. Error (ng/mL)	0.122	32.8	0.118 (0.0334, 0.17)
Add. Error (ng)	0.776	19.6	0.769 (0.594, 1.08)
Body weight effect on CLM	0.580	26.2	0.585 (0.209, 0.879)
Body weight effect on QM	1.00	29.8	1.02 (0.408, 1.63)
Sex effect on V4	−0.181	38	−0.18 (−0.321, −0.0256)

Abbreviations: Add, additive; ALAG1, absorption lag time; CI, confidence interval; CL/F, apparent central clearance; CLM, apparent central clearance for MDA; corr, correlation; CV, coefficient of variation; D1, duration of zero order input; f_met, fraction of MDMA metabolized to MDA; IIV, inter‐individual variability; k_a, absorption rate constant; MDA, 3,4‐methylenedioxyamphetamine; MDMA, 3,4‐methylenedioxymethamphetamine; Prop, proportional; QM, intercompartmental clearance for MDA; r, correlation; RSE, relative standard error; V2/F, apparent central volume of distribution; V3, apparent central volume of distribution for MDA; V4, peripheral volume of distribution for MDA.

^a Correlation of CL/F, V2/F: r = 0.67.

^b Correlation of CLM, QM: r = 0.82.

^c Correlation of CLM, V4: r = 0.86.

^d Correlation of QM, V4: r = 0.70.

Covariate effects on MDMA and MDA exposures

To assess the clinical significance of covariates, MDMA and MDA exposure were evaluated by means of simulations presented in covariate effects plots (Figure 2). Clinical significance was defined in the initial analysis plan as a 20% decrease or a 20% increase in exposure relative to a reference individual. MDMA is effective at doses as low as 75 mg HCl and tolerated at doses as high as 225 mg HCl (split dose). ²¹ , ²² Therefore, these are the therapeutic bounds subsequently established as clinically meaningful. The hypothetical reference individual for covariate comparison was defined as a white male from the NIDA study population, weighing 70 kg, aged 25 years, with baseline creatine clearance of 113 mL/min, and fed a light meal prior to receiving MDMA. For MDMA, study and body weight were identified as the only two covariates to meet this benchmark (>20% effect on exposure); neither fed status nor sex were clinically significant. However, the effect of body weight was most evident at the extremes (5th and 95th percentiles) and not considered clinically meaningful as they remain within the therapeutic bounds. The effect of study was an idiosyncrasy of the combined MPKF and NIDA datasets and not considered meaningful for the patient population. MPKF subjects had a 50% higher AUC_0–44 and 40% higher MDMA C _max as compared to NIDA subjects. Subjects in the 5th percentile of body weight had an overall 30% higher MDMA AUC_0–44 and C _max, while subjects in the 95th percentile of body weight had an overall 28% lower MDMA AUC_0–44 and 25% lower C _max. No covariates were found to be clinically significant for MDA.

FIGURE 2.

PopPK Model: Covariate effects on (a) MDMA and (b) MDA. For all covariate scenarios, all other covariates were maintained at values for a reference individual. The reference individual was defined as a white male from the NIDA study population, weighing 70 kg, aged 25 years, with baseline creatine clearance of 113 mL/min, and fed a light meal. Left side of the plot: Number to the right of body weight corresponds to the 5th and 95th percentiles of the population in kg. Numbers to the right of categorical covariates correspond to numbers of subjects in each category (Nonreference: Reference). Right side of the plot. Values in boxes represent data shown in the graph, summarizing the median and 95% confidence interval for AUC (left box) and C _max (right box). AUC refers to AUC_0–44. AUC, area under the plasma concentration–time curve; CI, confidence interval; C _max, maximum observed plasma concentration; CV, coefficient of variation; MDA, 3,4‐methylenedioxyamphetamine; MDMA, 3,4‐methylenedioxymethamphetamine.

PBPK model

PBPK model analysis

The simulated profile of MDMA was comparable to observed clinical data from MPKF and NIDA studies for verification of the model (Figure S3). The simulated arithmetic mean AUC_0–inf and C _max values for MDMA were within 1.25‐fold of observed values.

Pharmacokinetic simulations of dosing scenarios

Mean plasma concentrations following single and split 120 and 180 mg MDMA HCl doses were simulated (Figure 3, Table S5). Split‐dosing simulations were generated with an interval of 1.5 or 2 h between the first and second parts of the split dose, but as there was no significant difference between the two intervals, only the 2 h interval is reported. The split‐dose regimen showed a slightly lower C _max (≤5% difference) and marginally higher overall exposure (AUC_0–44, ~1%) compared to single dose. However, across the acute exposure interval, a decrease in AUC_0–4 (~15%) was observed for the split dose compared to single dose.

FIGURE 3.

PBPK model: Simulated MDMA exposure following single and split dose of MDMA. (a) MDMA AUC from 0 to 44 h post‐first dose; (b) MDMA AUC from 0 to 4 h post‐first dose; (c) MDMA C _max. Box plot area, dark line, and whiskers represent the interquartile range, median, and range of simulated values; black circles are outlier values. AUC, area under the curve; C _max, maximum observed concentration; MDA, 3,4‐methylenedioxyamphetamine; MDMA, 3,4‐methylenedioxymethamphetamine.

DDI simulations

A sensitivity analysis was completed to evaluate the change in interaction potential (AUC and C _max ratios) with changes in the fm_CYP2D6 of a sensitive CYP2D6 substrate based on atomoxetine. A fm_CYP2D6 = 0.85 best recovered the observed interaction and was used in the model. A summary of predicted AUC and C _max geometric mean ratios (GMRs) is presented in Figure 4. When administered with the second part of a split dose of 180 mg MDMA HCl (2 h interval), a mild interaction (≥1.25‐fold) by AUC ratio was predicted starting at fm_CYP2D6 = 0.05, with risk increasing to moderate (≥2‐fold) at fm_CYP2D6 = 0.20 and strong (≥5‐fold) at fm_CYP2D6 = 0.75.

FIGURE 4.

PBPK model: Predicted change in AUC and C _max for a theoretical substrate over a range of fm_CYP2D6 values following split dose of MDMA. Depicted are simulated AUC (teal) and C _max (orange) ratios over the range of fm_CYP2D6 = 0.05 to 0.95 following a single oral dose administered with the second dose of a split 180 mg dose of MDMA HCl (2 h interval between doses). The dashed horizontal lines represent the FDA cutoff values for a weak (ratio ≥1.25), moderate (2≤ ratio <5), or strong (ratio ≥5) interaction. AUC, area under the curve; C _max, maximum observed concentration; fm_CYP2D6, fraction metabolized by CYP2D6; GMR, geometric means ratio; MDMA, 3,4‐methylenedioxymethamphetamine.

The simulated geometric means for AUC_0–72 and C _max values and corresponding GMRs were evaluated for metformin pharmacokinetics in the presence and absence of MDMA (Figure 5, Table S6). The interaction was evaluated using the determined in vitro K _i values for the renal transporters (MATE1, OCT1, and OCT2) and using reduced K _i values based on prior experience modeling in vitro‐to‐in vivo effects of transporter inhibition. The simulation using the K _i as determined in vitro (MATE1 K _i , 2.21 μM; OCT1 K _i , 0.121 μM; and OCT2 K _i , 0.127 μM) indicated no interaction. When the simulation was completed with applied reductions in the K _i of all three transporters (3‐fold reduction in MATE1 K _i and 15‐fold reduction in the K _i values for OCT1 and OCT2), a weak interaction was observed.

FIGURE 5.

PBPK Model: Simulated metformin exposure in the absence of MDMA and with the first dose of a split dose of 180 mg MDMA HCl. (a) Metformin AUC from 0 to 72 h post‐first dose and (b) metformin C _max. Box plot area, dark line, and whiskers represent the interquartile range, median, and range of simulated values; symbols are outlier values. AUC, area under the curve; C _max, maximum observed concentration; K _i , inhibitor constant; MDMA, 3,4‐methylenedioxymethamphetamine.

DISCUSSION

Through collection of new pharmacokinetic data (MPKF study) and integration of existing pharmacokinetic data (NIDA study) and in vitro data, an MIDD approach using PopPK and PBPK modeling was leveraged to describe and predict clinically relevant aspects of MDMA pharmacokinetics. This included characterization of food effect and simulation of split‐dosing regimen, along with other intrinsic and extrinsic factors.

MDMA pharmacokinetics following oral administration was well characterized by a one‐compartment PopPK model with sequential zero‐ and first‐order absorption and linear elimination, and for MDA, a two‐compartment PopPK model with linear elimination. Although several covariates were found to be statistically significant in the final model, only body weight and study effect met the benchmark of ±20% exposure, which was at the extremes (5th and 95th percentiles) of body weight. Subsequent to these analyses, the range of MDMA plasma concentrations deemed efficacious and tolerable were determined to span wider than the ±20% exposure benchmark used in this analysis. The study effect was an idiosyncrasy of this specific dataset that assessed the MPKF and NIDA study data side by side and potentially due to differences in washout period, drug use history, or other factors between the two studies. Therefore, neither body weight nor study effect warrants a modification of the dose in the intended patient population and as such, is not clinically meaningful. Of note, the effect of fasting status on exposure was not clinically significant and the MPKF study demonstrated no effect of food on MDMA C _max and AUC, although T _max was delayed 2 h following administration with food. Furthermore, a pooled analysis of 10 clinical studies (n = 194) examining variables predicting physiological and psychological response to MDMA in HVs aligned with our finding of lack of clinically significant covariates. ²³ That study reported MDMA response was most dependent on MDMA plasma concentration, but unaffected by age and sex (following adjustment for body weight).

The PBPK model captured the observed MDMA pharmacokinetics within 1.25‐fold for all studies evaluated and within the 2‐fold bound of prediction accuracy and model fidelity. ²⁴ As MDMA is a strong CYP2D6 MBI (time dependent), ⁹ , ¹⁰ , ¹¹ , ¹² a PBPK modeling approach was applied to simulate clinical split‐dosing regimens, taking CYP2D6 and transporter inhibition into account. Although clinical trials evaluated the safety and efficacy of MDMA‐assisted therapy using a split‐dose regimen for the treatment of PTSD, ¹ , ² the pharmacokinetics of split dosing were not characterized in clinical studies. Simulated pharmacokinetics for the proposed 120 and 180 mg MDMA HCl clinical doses suggest minor differences when the dose is administered as a split dose over 1.5–2 h compared to a single dose as overall exposure remained similar. However, split‐dose administration of MDMA reduced AUC within the first 4 h and T _max was slightly delayed (~1 h), relative to single dose. MDMA peak effects occur 70–90 min after the first dose, persisting for 4–6 h, and the presence of peak effects can be extended without increasing total duration of physiological or subjective effects by administering a split dose. ¹³ , ²⁵ The split‐dose modality is widely used when conducting therapeutic sessions to facilitate a more gradual onset and subsidence of MDMA effects. ²⁵ A lower AUC over the first 4 h following a dose may reduce severity of adverse effects, such as elevated blood pressure and heart rate, which mostly occur within the first 4 h and are dose dependent. ¹³ Thus, split doses were used throughout the MDMA development program. ¹ , ²

The PBPK model also investigated CYP2D6‐mediated DDIs with MDMA as a precipitant. Clinical DDI studies evaluated interactions of MDMA and CYP2D6 substrates such as dextromethorphan, paroxetine, and bupropion. ⁹ , ¹⁰ , ¹¹ , ¹² , ²⁶ Using atomoxetine as a sensitive CYP2D6 substrate in the PBPK model, a sensitivity analysis was completed to predict AUC and C _max GMR across a range of fm_CYP2D6 (0.05–0.95). A weak interaction was predicted at fm_CYP2D6 = 0.05, moderate interaction at fm_CYP2D6 = 0.20, and strong interaction for substrates with fm_CYP2D6 ≥ 0.75. This illustrates that MDMA is a potent CYP2D6 inhibitor and care should be taken when co‐administering MDMA with other CYP2D6 substrates (even substrates with low fm_CYP2D6), particularly if the compound has a narrow therapeutic index.

The potential for MDMA to perpetrate DDI with metformin, a drug sensitive to renal transport, was simulated in the PBPK model as follow‐up to the observation that MDMA inhibits OCT1, OCT2, and MATE1 with IC₅₀ values that exceed anticipated clinical MDMA plasma concentrations (Table S4). ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ When using the reported in vitro K _i values, there was no effect of MDMA on metformin pharmacokinetics. However, a sensitivity analysis in which the K _i was reduced for MATE1, OCT1, and OCT2 suggests MDMA may cause a weak DDI resulting in AUC and C _max elevations of >1.5‐fold, but <2‐fold. Additionally, MDMA and MDA are substantially excreted unchanged (<20%) by the renal route. ¹⁴ , ¹⁵ Collectively, data suggest MDMA may have no meaningful impact on the pharmacokinetics of drugs sensitive to renal transport.

CONCLUSIONS

Based on observed pharmacokinetic data in study MPKF, MDMA can be administered without regard to food intake. The PopPK model identified no covariates associated with clinically meaningful effects on MDMA pharmacokinetics. Simulations of clinically relevant dosing regimens using the PBPK model indicate split dosing decreases AUC during a critical therapeutic window (0–4 h) while maintaining the same overall exposure in comparison to the same amount of MDMA administered as a single dose, which may facilitate a more gradual onset and decrease the severity of adverse events. The PBPK model also indicates the primary MDMA DDI liability is as a precipitant when taken in combination with sensitive CYP2D6 substrates.

AUTHOR CONTRIBUTIONS

N.B.M., C.L., and B.Y‐K. wrote the manuscript; M.A.H, W.B.S., C.L., R.B., E.S., and B.Y‐K. designed the research; M.H., W.S.B., A.V., and E.S. performed the research; C.L., E.S., and A.V. analyzed the data.

FUNDING INFORMATION

The MPKF study (NCT05147402) and PK modeling were funded by the Multidisciplinary Association of Psychedelic Studies (MAPS) and sponsored by Lykos Therapeutics (formerly MAPS PBC).Data used in the preparation of this article were obtained from the Neurobiology and Pharmacokinetics of Acute MDMA Administration Study (NCT01148342) supported by the Intramural Research Program, National Institute on Drug Abuse (NIDA), National Institutes of Health; the PI of this study was Dr. Marilyn Huestis at the time when she was a U.S. federal employee and principal investigator at the NIDA Intramural Research Program. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Article processing fees were funded by Lykos Therapeutics.

CONFLICT OF INTEREST STATEMENT

MAH, WBS, AV, CL, and RB received payment from Lykos Therapeutics for support with the research. AV is a contractor of Lykos Therapeutics. NBM is an employee of and has stock/stock options in Lykos Therapeutics. BY‐K is an employee of and has stock/stock options in Lykos Therapeutics where she is the Chief Scientific Officer, was previously an employee of MAPS, and has received support from Lykos Therapeutics and MAPS for attending meetings/travel. All other authors declared no competing interests in this work. No authors were compensated for authorship activities associated with this article.

Supporting information

Data S1.

Huestis MA, Smith WB, Leonowens C, et al. MDMA pharmacokinetics: A population and physiologically based pharmacokinetics model‐informed analysis. CPT Pharmacometrics Syst Pharmacol. 2025;14:376‐388. doi: 10.1002/psp4.13282