Population pharmacokinetic/pharmacodynamic modelling of the psychedelic experience induced by N,N-dimethyltryptamine - implications for dose considerations

Depressive and anxiety disorders affect hundreds of millions worldwide, and a substantial fraction of patients do not respond to available treatments. Renewed interest in classic serotonergic psychedelics has been driven by preliminary evidence of lasting therapeutic effects after single or few doses. N,N-dimethyltryptamine (DMT) is a naturally occurring psychedelic found in ayahuasca and is believed to act primarily via 5-HT2A receptor agonism; however, DMT is rapidly metabolised (principally by monoamine oxidase A (MAO‑A)) and, despite growing clinical investigation, its human pharmacokinetic (PK) and pharmacodynamic (PD) properties remain poorly characterised. Eckernäs and colleagues set out to characterise population PK of intravenously administered DMT and to quantify the PKPD relationship between plasma DMT concentrations and the psychedelic experience as measured by real‑time subjective intensity ratings. The stated aim was to provide a model that could support dose selection for future clinical studies, including decisions about doses producing sub‑psychedelic versus fully psychedelic experiences and the design of extended administration regimens.

The analysis used data from a placebo‑controlled clinical trial in 13 healthy volunteers (seven males; median age 33 years, range 22–48) conducted at an NIHR clinical research facility. The trial employed a fixed‑order design: each participant received placebo on their first visit and an intravenous bolus of DMT fumarate one week later. Individual subjects received one of four dose levels: 7 mg (n=3), 14 mg (n=4), 18 mg (n=1) or 20 mg (n=5). For PK sampling, nine blood draws per subject were taken via an indwelling catheter up to 60 minutes post‑dose. Subjective psychedelic intensity was recorded every minute during the first 20 minutes on a 0–10 scale (0 = no effect, 10 = most intense imaginable). Plasma concentrations of DMT, DMT N‑oxide and indole 3‑acetic acid (IAA) were quantified by a validated liquid chromatography–tandem mass spectrometry method. Calibration ranges were 0.25–200 nM (DMT), 15–200 nM (DMT N‑oxide) and 500–5000 nM (IAA); lower limits of quantification were 0.25, 15 and 500 nM respectively. IAA was reported as change from baseline due to high endogenous levels. Samples above the upper limit were diluted and re‑analysed. Samples below the LLOQ were excluded (n=3 for DMT, n=9 for IAA) and one sample with an implausibly high DMT concentration was removed. Population PK and PKPD modelling was performed using nonlinear mixed‑effects methods in NONMEM (FOCE with interaction), with Pirana and PsN for workflow and R for diagnostics and plots. A sequential modelling strategy was used: a population PK model for DMT was developed, extended to include IAA, and finally a PKPD model was built using a PPP&D approach (population PK parameters fixed while individual PK parameters and PD parameters were estimated). Between‑subject variability (BSV) was modelled with exponential terms (log‑normal). For the PK stage, one‑ and two‑compartment models with first‑order elimination were evaluated; IAA was modelled as a one‑compartment primary metabolite with the metabolic fraction (fm) fixed to 1. For PD, effect‑compartment models with Emax or sigmoid Emax relationships were considered. The PD model used a logit transform to constrain predicted intensity to the 0–10 scale. Model selection used changes in objective function value (ΔOFV), diagnostic plots, visual predictive checks (VPCs) and sampling importance resampling to derive parameter precision and 95% confidence intervals. Simulations were performed in R/mrgsolve for five bolus doses (1, 4, 7, 14 and 20 mg), each simulated in 100 virtual subjects, to predict distributions of maximum achieved intensity.

A total of 93 DMT and 87 IAA plasma concentration observations were included (distributed across the four dose groups), together with 273 subjective intensity ratings. No measurable DMT concentrations or subjective effects were observed after placebo, so placebo data were not modelled. PK findings: DMT concentrations were best described by a two‑compartment model with first‑order elimination that was assumed to form IAA as the primary metabolite. A two‑compartment fit improved the model substantially over a one‑compartment alternative (ΔOFV = -41.64). The estimated apparent plasma clearance of DMT was high (reported as 26 L/min). Between‑subject variability was included on clearance; IAA was described by a one‑compartment model with BSV on apparent volume. Proportional residual error models were used for both compounds. Parameter precisions met pre‑specified criteria (%RSE ≤ 30% for fixed effects and ≤ 50% for BSV) except for intercompartmental clearance (Q), for which %RSE was 37%. Goodness‑of‑fit plots and VPCs indicated acceptable predictive performance of the final PK model. PKPD findings: A short delay between plasma DMT and reported intensity led to selection of an effect‑compartment model with a sigmoid Emax relationship. The effect‑site equilibrium rate constant (ke0) was estimated at 1.38 min‑1, implying an approximate 2‑minute equilibration between blood and the effect compartment. Emax was fixed to 10 and baseline intensity was zero for all subjects. The effect‑site EC50 (EC50,e) was estimated at 95 nM and the Hill coefficient (γ) was approximately 3, indicating a steep concentration–effect slope in the mid‑response range. BSV was estimated at 39% for EC50,e and 77% for γ. A high correlation (93%) between the BSV terms was observed in exploratory fits but was not retained because it produced model ill‑conditioning. VPCs and GOF diagnostics supported the final PKPD model. Simulations: Using the final model, simulated median maximum intensity ratings at doses of 1, 4, 7, 14 and 20 mg were 0, 2, 4, 8 and 9 respectively. The proportion of simulated subjects achieving a maximum rating above 5 rose across doses: 0% (1 mg), 4% (4 mg), 42% (7 mg), 92% (14 mg) and 100% (20 mg).

Eckernäs and colleagues present what they describe as the first population PKPD characterisation linking plasma DMT concentrations to real‑time subjective intensity ratings. The investigators interpret the two‑compartment PK model and the PKPD effect‑compartment with a sigmoid Emax as consistent with a short distribution delay between blood and the presumed brain effect site, and with steep concentration dependence of subjective intensity within the mid‑response range. A notable finding is the very high estimated plasma clearance of DMT (26 L/min), well above average cardiac output. The authors suggest this implies elimination processes that are not limited by organ blood flow and hypothesise that widespread MAO‑A expression (including in blood vessels) could contribute; they caution that further work is required to confirm mechanisms and refine clearance estimates. The modelling assumed complete conversion of DMT to IAA (fm set to 1) to permit metabolite parameter estimation; the authors acknowledge that DMT metabolic pathways after intravenous dosing are incompletely understood and that other metabolites (notably DMT N‑oxide) were absent in these plasma samples. On the PD side, the effect‑site equilibrium rate constant (ke0 ≈ 1.38 min‑1) corresponds to about two minutes for blood–brain equilibration, and the Hill coefficient (~3) denotes a steep concentration–effect relationship where small concentration changes within 20–80% of maximal response produce large effect changes. The EC50,e of 95 nM is reported to be in the range of in vitro EC50 values for 5‑HT2A receptor activation reported elsewhere, which the authors suggest is supportive of 5‑HT2A mediation of the psychedelic effects. The authors highlight substantial between‑subject variability in EC50,e and γ, and note the limitations imposed by the small sample (13 subjects) which prevented a formal covariate analysis. They also describe modelling choices and limitations related to the subjective intensity ratings: the integer 0–10 scale was treated as continuous and constrained by a logit transform, which cannot predict exact 0 or 10 but can approximate boundary values closely. Simulations using the model are presented as practical guidance: a 14 mg bolus yields a median maximal intensity of 8 with 92% of subjects predicted above an arbitrarily chosen intensity threshold of 5, whereas doses ≤7 mg are more compatible with sub‑psychedelic or modest intensity responses. The authors suggest the model may therefore aid dose selection for clinical development, including choices about sub‑psychedelic regimens and parameters for extended administration, but they emphasise the need for larger studies to explore covariates and refine these findings.