Population pharmacokinetic-pharmacodynamic modeling of co-administered N,N-dimethyltryptamine and harmine in healthy subjects

Ayahuasca combines N,N-dimethyltryptamine (DMT) with monoamine oxidase A (MAO-A) inhibitors such as harmine; the MAO-A inhibition enables sustained systemic exposure to DMT by preventing rapid oxidative deamination. Previous research has shown that co-administration of harmine increases DMT bioavailability, plasma half-life and duration of subjective effects, but non-compartmental analyses (NCA) used previously do not fully characterise inter-individual variability or allow mechanistic PK/PD predictions that incorporate covariates such as body weight or differences in metabolic capacity. Äbelö and colleagues set out to develop a population pharmacokinetic–pharmacodynamic (PopPK/PD) model for buccally administered (oromucosal) DMT co-administered with harmine. The stated aim was to quantify how harmine modulates DMT exposure and the concentration–effect relationship for subjective psychedelic intensity, to characterise between-subject variability, and to provide a framework for model-based dose simulations that could inform precision dosing strategies in future clinical applications.

The analysis used data from a previously published single-blind, randomised, factorial, dose-finding clinical trial performed in 16 healthy subjects at the Psychiatric University Hospital Zurich. The trial tested seven dose combinations of buccal DMT and harmine across six study days (dose combinations included 90:0, 90:60, 90:120, 90:180, 0:120, 60:120, 120:120 mg DMT:harmine). Doses were given in three increments spaced 20 minutes apart. Blood samples for pharmacokinetics and serial subjective intensity ratings (psychometric scale 0–10) were collected at multiple time points up to 1440 minutes after the first administration; mild transient adverse events were reported in the primary publication. Plasma concentrations of DMT, DMT-N-oxide, harmine and harmine metabolites were quantified by LC–MS/MS with LLOQs of 0.5 ng/mL for the principal analytes. Nonlinear mixed-effects modelling was performed in NONMEM 7.5.1 using first-order conditional estimation with interaction. The analysis followed a sequential approach: a population PK model for DMT was established first, then a PK/PD model was developed using a “Population PK Parameters and Data” approach in which population PK parameters were fixed while individual PK parameters and PD parameters were estimated. For PK model development the investigators tested one- and two-compartment disposition models with first-order elimination, various absorption models including lag time and transit compartments, and applied allometric scaling of apparent clearance (CL/F) and volume (V/F) to body weight (exponents 0.75 and 1). Bioavailability (F) was included with inter-individual variability. Between-subject variability (BSV) was modelled exponentially and residual error used combined additive and proportional components. For PD, direct concentration–response models (Emax and sigmoidal Emax) were evaluated to describe subjective intensity ratings. The sigmoidal Emax model was parameterised by baseline (E0), Emax, EC50 and the Hill coefficient (γ); Emax was fixed at 10 (the scale maximum). BSV was assessed for EC50 and γ. The authors evaluated alternative transformations (e.g. logit) and explored indirect response/effect-compartment models but did not identify a clinically meaningful delay between plasma concentrations and observed subjective effects in this dataset. Model selection used objective function value (OFV) comparisons with a chi-squared threshold (ΔOFV ≥ 3.84 for p < 0.05 for one degree of freedom). Final model evaluation included goodness-of-fit plots, visual predictive checks (VPCs), and precision assessment via sampling importance resampling (SIR) with 5000 samples/1000 resamples. Simulations of dosing scenarios were performed in NONMEM for 100 virtual subjects to explore the impact of harmine dose on peak subjective intensity at fixed DMT doses (60, 90, 120 mg).

The extracted text reports that a dataset of DMT plasma concentrations and subjective intensity ratings from the 16 participants was analysed; the text explicitly states a total of 847 DMT plasma concentration observations were included. The exact count of subjective intensity ratings in the extraction is unclear due to formatting artefacts; however, serial ratings across the same sampling schedule were analysed. Pharmacokinetics: A one-compartment model with first-order elimination and delayed absorption implemented via three transit compartments (NN = 3) best described the DMT concentration–time profiles following oromucosal dosing. Allometric scaling to body weight improved model fit (ΔOFV = -20) and was retained. Between-subject variability was implemented on CL/F, V/F and mean transit time (MTT). Residual error was modelled with a combined additive and proportional structure. Harmine dose was a significant covariate: including harmine dose as a proportional effect on bioavailability (F) reduced OFV substantially (ΔOFV = -238), and additionally modelling harmine's effect as a saturable increase in mean transit time further reduced OFV (ΔOFV = -416). The estimated maximum increase in MTT (MTMAX) was 207 minutes (relative standard error 9.8%), and the harmine dose producing half-maximal MTT effect (MT50) was 126 mg (22.5% RSE). Parameter precisions met the prespecified criteria (≤30% RSE for fixed effects; ≤50% for BSV parameters) according to the authors. The authors note that CL/F was large (reported as approximately 404 L/min in the text) but interpret this in light of extravascular administration and reduced bioavailability; they discuss that assuming a modest bioavailability (for example ~10%) yields true CL estimates more comparable to IV literature values. Pharmacodynamics: The relationship between DMT plasma concentration and subjective intensity was best described by a sigmoidal Emax model; this choice produced a significantly better fit than a simple Emax model (the extraction reports OFV improvements of ΔOFV = -104 in one place and ΔOFV = -577 in another, both indicating consistent superiority of the sigmoidal model). Emax was fixed to the scale maximum (10) and baseline ratings were zero. Between-subject variability was estimated for EC50 and the Hill coefficient (γ), with reported coefficients of variation of approximately 54% and 50% respectively, indicating substantial individual differences in sensitivity and slope of the concentration–effect relationship. The authors report that no concentration–effect delay could be robustly estimated in this dataset; direct linkage between plasma DMT and subjective ratings fitted the observed time-course. Simulations: Using the final PK/PD model, simulations in 100 virtual subjects showed a dose-dependent increase in maximum subjective intensity with rising harmine doses at fixed DMT doses (60, 90, 120 mg). The relationship between harmine dose and median peak intensity followed a sigmoidal pattern with evidence of a plateau at higher harmine doses. Simulations illustrated that harmine co-administration elevates and prolongs DMT plasma concentrations, which coincided with more pronounced and sustained subjective intensity ratings. The authors highlight that between-subject variability in clearance and bioavailability leads to wide distributions of predicted peak intensities under identical dosing.

Äbelö and colleagues interpret their findings as providing a first population-based PK/PD framework for an oromucosal DMT–harmine formulation. They emphasise that harmine markedly increases apparent DMT bioavailability and prolongs absorption (increasing mean transit time), which produces higher and more sustained plasma concentrations and, in turn, enhanced and prolonged subjective psychedelic intensity. The one-compartment model with transit compartments is presented as physiologically plausible for oromucosal dosing, capturing a delayed absorption phase with observed Tmax values around 100–150 minutes. The authors compare their extravascular CL/F and volume estimates to intravenous literature and explain discrepancies by the influence of bioavailability on apparent clearance, differences in model structure (one- versus two-compartment), and sparse late-phase data leading to imprecise BSV estimates. Potential biological contributors to inter-individual variability are discussed: MAO-A activity outside the liver, CYP2D6-mediated harmine metabolism (with known "fast" and "slow" phenotypes), and possible MAO-A genetic polymorphisms. In this context the investigators argue that population modelling is advantageous because it can incorporate covariates to explain variability, and they retained allometric scaling despite a narrow weight range because it improved model fit. Regarding pharmacodynamics, the sigmoidal Emax model yielded EC50 estimates comparable to prior IV DMT modelling but a substantially lower Hill coefficient, which the authors interpret as a gentler concentration–effect slope under harmine co-administration. They suggest this gentler slope arises from the slower, prolonged exposure produced by harmine, which may broaden the therapeutic window by avoiding abrupt on–off subjective responses. The considerable between-subject variability in EC50 and γ (CV% ~50–54) is highlighted; potential mediators include receptor density/affinity, downstream signalling, anatomical distribution, psychological factors (e.g. set and setting), and prior psychedelic experience. The authors note that harmine’s predictable enhancement of DMT bioavailability makes it a useful tool for probing the mechanisms of between-subject differences when combined with neuroimaging or electrophysiology. The modelling work is used to illustrate practical implications: simulated dose–response surfaces may inform selection of DMT–harmine combinations that target desired distributions of peak subjective intensity, supporting precision dosing in clinical settings. However, the authors caution that higher combined doses increase total drug exposure and could raise the risk of adverse events, particularly given high inter-individual variability. They therefore recommend cautious application and further investigation. Key limitations acknowledged by the investigators include the small sample size (N = 16) and limited demographic diversity, which restrict generalisability and parameter precision; the subjective nature of intensity ratings and the decision to treat an ordinal 0–10 scale as continuous; limited covariate exploration to avoid overfitting, meaning genetic or phenotypic determinants (e.g. CYP2D6 or MAO-A polymorphisms) were not assessed; and that the model was fitted to a specific dosing schedule and may not generalise to other administration protocols without re-evaluation. The authors suggest future larger, more diverse studies, routine genotyping and incorporation of harmine concentration-time data into an interaction model as valuable next steps.

The extracted conclusion states that population-based PK/PD modelling of oromucosal DMT co-administered with harmine demonstrates that harmine substantially enhances DMT bioavailability and prolongs its absorption, producing sustained subjective effects. The authors present the model as a proof-of-concept framework that quantifies key sources of variability and can inform model-based dose simulations to support more precise dosing strategies in future clinical development. The extraction truncates the final sentence(s), so the full concluding text is not available in the provided material.