N,N-dimethyltryptamine affects EEG response in a concentration dependent manner - a pharmacokinetic/pharmacodynamic analysis

DMT (N,N-dimethyltryptamine) is a classic serotonergic psychedelic that produces profound alterations in perception, cognition and affect when given exogenously. Prior EEG and MEG studies of classic psychedelics, including DMT (often studied as part of ayahuasca), have consistently reported suppression of alpha-band power and increases in signal diversity, but results linking those electrophysiological changes to subjective experience and clinical outcomes have been inconsistent. Crucially, earlier work has not quantified the quantitative relationship between drug exposure (plasma concentrations) and EEG effects using pharmacokinetic/pharmacodynamic (PKPD) modelling, leaving uncertainty about dose–response characteristics and suitable biomarkers for clinical development. Eckernäs and colleagues set out to quantify the relationship between DMT plasma concentrations and longitudinal EEG measures using a population PKPD approach. Using data from a placebo-controlled pilot study in which thirteen healthy volunteers received a single intravenous bolus of DMT fumarate at one of four dose levels (7 mg, 14 mg, 18 mg or 20 mg), the study investigated three EEG endpoints—alpha power, beta power and Lempel–Ziv complexity (LZc, a measure of signal diversity)—and sought to characterise exposure–response parameters such as EC50/IC50 and the time delay between plasma concentration and effect.

This analysis used data from a placebo-controlled pilot study in which thirteen healthy subjects received a single intravenous bolus of DMT fumarate at one of four dose levels: 7 mg (n=3), 14 mg (n=4), 18 mg (n=1) or 20 mg (n=5). Nine blood samples per subject were collected up to 60 minutes after dosing; plasma was stored at –80°C and DMT quantified by a previously described LC–MS/MS method. The investigators used a previously developed population PK model (two-compartment disposition with first-order elimination and between-subject variability on clearance) as the pharmacokinetic input for the PKPD analysis. Electroencephalography was recorded with a 32-channel system at 1000 Hz, with a 0.1 Hz high-pass and 450 Hz anti-aliasing filter. Preprocessing used the FieldTrip toolbox: data were band-pass filtered at 1–45 Hz, visually inspected for artefacts, and segments with gross artefacts or concurrent intensity-rating issues were excluded. Spontaneous signal diversity was summarised with Lempel–Ziv complexity (LZc). For modelling, EEG outcomes (alpha power, beta power, LZc) were averaged across channels and summarised as mean values per minute. Population PKPD modelling was performed in NONMEM (first-order conditional estimation with interaction) using a ‘‘Population PK Parameters and Data’’ approach: the previously developed population PK parameters were fixed while individual PK parameters and PD parameters were estimated simultaneously. Effect compartment models were used to characterise the short delay between plasma concentration and EEG response; for alpha and beta power the drug effect was modelled with inhibitory Imax or sigmoid Imax functions, and for LZc with linear, Emax or sigmoid Emax functions. Between-subject variability (BSV) and between-occasion variability (BOV) were modelled as exponential (log-normal) random effects, with Box–Cox transformations applied when distributions appeared skewed. Residual unexplained variability (RUV) was tested as additive, proportional or combined error structures. Model selection and evaluation relied on objective function value (ΔOFV threshold −3.84 for p=0.05), parameter plausibility and precision, goodness-of-fit plots, individual predictions and visual predictive checks (VPCs). Sampling importance resampling (5,000 samples/1,000 resamples) was used to derive parameter confidence intervals; %RSE thresholds of ≤30% for fixed effects and ≤50% for BSV parameters were used as guides. Finally, simulations using the final models predicted effects over time for five dose levels (1, 4, 7, 14 and 20 mg) in 100 simulated subjects, incorporating BSV.

Alpha power was best described by a sigmoidal Imax effect-compartment model. Imax was fixed to 1 (complete suppression) in the final model because estimates were close to unity when allowed to float. The estimated IC50,e for alpha power was 71 nM with between-subject variability (BSV) in IC50,e of 29% CV. Baseline alpha power (R0) showed substantial variability between individuals (125% CV) and between occasions (32% CV). Residual error for the alpha model was described as proportional. The estimated ke0 values across endpoints were short (0.59–1.2 min−1), indicating a small delay between plasma concentration and EEG effect. Beta power was also modelled with a sigmoidal Imax effect compartment model but did not reach full suppression in these data; the estimated Imax for beta was approximately 0.7 (partial suppression). The IC50,e for beta was estimated at 137 nM but was associated with large variability (BSV reported as high, and the abstract reports ~75% CV for the EC50,e analogue). A correlation between baseline beta power and IC50,e was observed (higher baseline associated with lower IC50,e). Residual variability for beta was described using a proportional error; attempts to include additive error improved OFV but caused precision issues, so the final model retained proportional error. Signal diversity (LZc) was best described by a sigmoidal Emax model with a Hill coefficient (γ) that substantially improved fit. The estimated EC50,e for LZc was 54 nM, with a maximum relative increase in LZc of about 10% compared with baseline. BSV was estimated for R0, EC50,e and Emax; the EC50,e BSV was large (~77% CV as reported in the abstract and discussion), while Emax and R0 had lower variability (discussion reports 42% and 5.2% CV, respectively). An observed correlation between R0 and Emax (92%) could not be retained in the final model due to ill conditioning. Residual error for LZc was described as additive. Visual predictive checks and individual goodness-of-fit plots (figures referenced in the text) were reported to show acceptable model fit, and simulation results illustrated dose–response relationships over time for the evaluated dose range. Across models, the estimated Hill coefficients were relatively high (around 4–5), implying steep concentration–effect slopes over the mid-range of observed effects. Simulations suggested that doses above ~10 mg would be required to approach full suppression of alpha power, that substantial beta effects were largely limited to the highest administered dose, and that LZc increases would generally track alpha suppression for much of the population but with greater interindividual variability.

Eckernäs and colleagues interpret their findings as evidence of a systemic concentration–response relationship between intravenous DMT and several EEG endpoints, most robustly for suppression of alpha power and also for increases in signal diversity as measured by LZc. Their PKPD analysis found a short delay between plasma concentration and EEG effect that was well captured by effect compartment models. Alpha suppression achieved full effect in the studied dose range (Imax fixed to 1), with an IC50,e around 71 nM and the least between-subject variability among endpoints, suggesting alpha power is a stable electrophysiological marker of DMT action. The authors caution that beta power showed only partial suppression (Imax ≈ 0.7), a higher IC50,e (≈137 nM) and very large between-subject variability, making it a poor candidate as a reliable individual-level biomarker in this dataset. Signal diversity (LZc) increased with DMT (EC50,e ≈54 nM; maximal relative increase ≈10%), and parameter estimates for LZc were similar to those for alpha power, indicating that LZc responses may often co-vary with alpha suppression but with greater variability in EC50 estimates. Key limitations acknowledged include the small sample size (thirteen subjects), large fluctuations in baseline EEG values (baseline was summarised over one minute which may have increased apparent variability), limited PK data preventing estimation of variability in volume of distribution, and consequently potential inflation of individual EC50 variability. The authors note that no covariate analyses were performed to explain between-subject variability because of the limited sample size. They also recognise that higher temporal resolution of EEG averaging might yield different ke0 estimates and that higher dose ranges would be needed to confirm whether true maxima were observed for beta and LZc. In terms of implications, the investigators suggest that quantifying exposure–response relationships could inform dose selection and choice of EEG biomarkers in future clinical development of DMT. They highlight that alpha power suppression is a relatively robust, low-variability response and could be a useful endpoint to guide dosing, potentially even in continuous-infusion paradigms if DMT tolerance is absent. The authors also discuss links between EEG markers and depression (for example, increased alpha power observed in some depressed populations and associations between signal diversity and depressive states), but they stop short of claiming clinical efficacy—emphasising that further research is required to confirm these PKPD relationships and to test whether acute EEG changes predict clinically relevant outcomes.