Pharmacokinetics and pharmacodynamics of sublingual microdosed lysergic acid diethylamide in healthy adult volunteers

Abstract

Introduction:

Microdosing is the practice of taking psychedelic drugs at doses that produce no or minimal perceptible subjective or behavioural effects. This study investigated the pharmacokinetics and pharmacodynamics of microdosed lysergic acid diethylamide (LSD).

Methods:

This was a Phase 1 double-blind placebo-controlled parallel-groups trial with 80 healthy male volunteers (four withdrawals due to anxiety). Plasma samples were taken at 0.5, 1, 2, 4 and 6 h after 10 µg sublingual LSD and analysed with liquid chromatography-tandem mass spectrometry (LC-MS/MS). LSD pharmacokinetics were modelled. Population analyses were performed using nonlinear mixed effects models. Heart rate and a visual analogue scale (‘feel effect’) were used to describe LSD pharmacodynamics. The effect of the relevant cytochrome P450 (CYP) genotype on LSD pharmacokinetics was qualitatively assessed. Plasma and serum levels of brain-derived neurotrophic factor (BDNF) were evaluated.

Results:

A one-compartment model best described LSD pharmacokinetics. Mean (95% confidence interval): elimination clearance = 7.78 L/h/70 kg (6.75–8.77), central volume of distribution = 32.9 L/70 kg (30.1, 36.0). Maximal concentration (0.20 µg/L), time to maximal concentration (1.51 h) and elimination half-life (3.08 h). The maximal increase in heart rate and visual analogue scale was small (<15%) compared to baseline estimates limiting the modelling. Two of the participants withdrawn from the study due to anxiety had intermediate-weak CYP2D6 activity. CYP2D6, CYP1A6, CYP2B6 and CYP2C9 qualitatively appeared to influence concentration. No evidence of alterations of peripheral BDNF with microdosing was found.

Conclusion:

This study provides a population pharmacokinetic model and LC-MS/MS assay that can inform clinical and bioequivalence studies. Relevant CYP genotypes should be studied in larger samples as combined potential biomarkers of response. Microdose-sensitive and reliable pharmacodynamic measures are needed.

Introduction

The pharmacology of lysergic acid diethylamide (LSD) has been well characterised since its discovery in the 1940s. LSD has a complex pharmacological profile interacting at numerous receptor-binding sites. It is typically described as a mixed 5-hydroxytryptamine (5-HT) 2/5-HT1 partial receptor agonist, although it also has activity at other serotonergic, dopaminergic and histaminergic receptor sites (Passie et al., 2008; Rickli et al., 2016). Activity at the 5-HT2A receptor in particular is thought to mediate its psychedelic effects (Holze et al., 2021b).

Microdosing is the practice of taking psychedelic drugs such as LSD at doses that produce no, or minimal perceptible subjective or behavioural effects (Murphy et al., 2024). A typical LSD microdose is 10 µg, whereas a psychedelic LSD (macro)dose might be in the range of 100–200 µg. Microdosing has become an increasingly popular community practise with users claiming a range of benefits including improvements in mood and cognition and reductions in symptoms of depression, anxiety, post-traumatic stress disorder, attention-deficit hyperactivity disorder and substance misuse (Lea et al., 2020).

Clinical trials in healthy volunteers have shown dose-dependent (5–26 µg) changes in blood pressure, sleep, neural connectivity, social cognition, mood, and the perception of pain and time (Bershad et al., 2019; Hutten et al., 2024; Holze et al., 2021a). The potential therapeutic effects of microdosing LSD for clinical disorders have yet to be studied (see (Murphy et al., 2024) for a review). Recently in Murphy et al. (2023), we reported the results of the Microdosing LSD (MDLSD) trial, the first clinical trial of LSD microdosing which allowed participants to microdose in their home environments. This demonstrated individual variability after 10 µg LSD microdoses on both positive mood but also adverse events with several participants reporting feeling over-stimulation and anxiety.

Two studies have previously investigated the pharmacokinetics of LSD microdoses (Family et al., 2020; Holze et al., 2021a). There are no published population pharmacokinetic models for sublingual microdoses of LSD. A population pharmacokinetic model and knowledge of the LSD time–concentration and concentration–effect relationships could facilitate the selection of an LSD dose that achieves a concentration associated with a desired therapeutic effect.

The primary objective of this study was to describe LSD pharmacokinetics and pharmacodynamics after LSD microdoses (10 µg) were administered to healthy adult volunteers (Murphy et al., 2023) in a laboratory environment. Only male (sex) participants were included because of the known and profound effects of the menstrual cycle on pharmacodynamic measures used in the study, such as electroencephalography (Murphy et al., 2023).

A secondary objective was to present a high sensitivity and validated liquid chromatography-mass spectrometry assay for quantification of low concentrations of LSD in plasma along with the major metabolite 2-oxo-3-hydroxy-LSD (O-H-LSD) and commonly occurring diastereomer iso-LSD.

We had two further secondary objectives. The first was based on a small study showing that polymorphisms at CYP2D6 demonstrated a relationship between enzyme activity and LSD concentration (Vizeli et al., 2021); poor metabolisers (no functional CYP2D6 activity) were found to have longer half-lives and approximately 75% higher parent drug plasma concentrations following macrodoses. Non-functional metabolisers also had greater and longer subjective effects of LSD (measured with VAS). Metabolism of LSD is via cytochrome P450 (CYP) enzymes 1A2, 2C9, 2D6, 2E1 and 34A, and UDP-glucuronosyltransferases (Luethi et al., 2019; Vizeli et al., 2021; Wagmann et al., 2019). To extend findings from Vizeli et al. (2021) in this study, we attempted to relate CYP2D6 genotype information and adverse events/withdrawals recorded during the phase of the trial where participants microdosed at home. Other relevant CYP enzymes available on the panel used (Lois et al., 2021) were qualitatively interpreted and plotted as exploratory measures.

A further secondary objective of this work was to examine plasma and serum concentrations of brain-derived neurotrophic factor (BDNF) before and after LSD microdosing. Pre-clinical research has demonstrated that the administration of various psychedelics (e.g. LSD, psilocybin, 2,5-dimethoxy-4-iodoamphetamine (DOI)) promotes neuroplasticity (see Calder and Hasler (2023) for review). BDNF is involved in mediating synaptic plasticity and because it can be measured in peripheral blood samples, concentrations of peripheral BDNF may be a biomarker of central plasticity processes being upregulated after administration of psychedelics. In terms of LSD, results are largely mixed as to the viability of this biomarker (Hutten et al., 2021; Becker et al., 2023; Holze et al., 2022; Ley et al., 2023; Straumann et al., 2023). Given the relatively large sample size of the current study compared to the previous microdosing study (Hutten et al., 2021, it was of interest to explore BDNF levels acutely after microdosing and after a 6-week regimen of microdosing in both plasma and serum samples.

Methods

A phase 1 trial with double-blind placebo-controlled parallel groups was used to generate pharmacokinetic, pharmacodynamic and BDNF observations for analysis. The trial protocol was prospectively published (Murphy et al., 2021), and primary results were reported (Murphy et al., 2023). The trial was registered at ANZCTR (trial ID = 381476). All participants provided written information consent, and ethical approval was provided by the New Zealand Health and Disability Ethics Committee (reference number: 19/STH/91). Lab visits in the study took place at the Clinical Research Centre at the University of Auckland.

Participants

Healthy male volunteers (N = 80) were enrolled in the trial with 40 participants randomised into the LSD group and 40 into the inactive treatment group (see Figure S1 for CONSORT diagram). The inclusion criteria were male sex between the ages of 25 and 60 years. See Murphy et al. (2023) for full inclusion/exclusion information as well as justification for criteria such as sex. Key exclusion criteria included: resting blood pressure over 160/90, body weight <50 kg or >120 kg, renal or hepatic impairment, unstable medical or neurological conditions, lifetime history of depression/schizophrenia and psychotic disorders. Exclusion criteria also included a current diagnosis of anxiety or eating disorders, suicidality, first-degree relatives with a psychotic disorder, substance use disorder and use of psychotropic medication. Individuals who had used a serotonergic psychedelic in the last year and any lifetime history of psychedelic microdosing were not included. A urine drug test was performed at screening, and a breathalyser test was performed at each study visit. Demographic characteristics of the population are shown in Table S1.

Study intervention and procedures

The study drug (d-lysergic acid diethylamide) and placebo were formulated by being first dissolved in a small amount of ethanol with distilled water added. Participants self-administered 1 mL oral syringes which contained either 10 μg of LSD in distilled water (0.2% ethanol), or only distilled water. LSD is tasteless and colourless so these tasted the same. The drug was formulated from GMP LSD base API (Onyx Scientific Limited, Sunderland, UK) by Biomed NZ Ltd., under a GMP manufacturing license issued by Medsafe NZ following a site inspection and audit. Syringes were provided in sealed lightproof bags and separate stability batches provided shelf-life information. Participants self-administered doses sublingually and held them there for 30 s before swallowing. A titration regimen was used in the trial where the dose was decreased by 5 µg and then increased by 1 µg incrementally as the participant felt able to avoid overstimulation. The titration regimen was introduced to six participants, one was on placebo and five on LSD. Four of the five on LSD were consequently withdrawn.

The study consisted of four visits including a screening visit, a baseline measurement visit, a first treatment visit, 6 weeks of microdosing at home followed by a final visit (see (Murphy et al., 2023)). Only the first treatment visit procedures are relevant to the current report. All first treatment visits commenced in the morning (typically 8 am or 10 am). After re-confirming consent and eligibility, an intravenous cannula was placed in the antecubital fossa. Drug administration took place approximately 1 h after arrival.

Participants were asked to arrive having consumed their usual meal, caffeine and nicotine intake (if any, to prevent withdrawal effects) and to have abstained from alcohol the day before. Participants were provided a light meal between the 240 and 360 min sample points.

Blood samples were obtained at 30, 60, 120, 240 and 360 min after drug administration using 10 mL lithium heparin tubes with actual time recorded at each sample. Blood samples were centrifuged immediately at 1500 g for 15 min after they were obtained, pipetted into 500 μL plasma aliquots and stored at −80°C. In the same time period (±3 min), the blood sample vital signs (heart rate, blood pressure) and adverse events were recorded, and the participants completed 16 VAS scales. The VAS scales were completed on an iPad with anchor points ranging from ‘not at all’ to ‘extremely’. The specific questions were ‘Do you feel a drug effect right now?’, ‘Do you like any of the effects you are feeling right now?’, ‘Do you dislike any of the effects you are feeling right now?’, ‘Are you high right now?’, ‘Would you like more of the drug you took right now?’, ‘Do you feel sick’, ‘Do you feel sleepy’, ‘Do you feel dizzy’, ‘Can you see movement in still objects?’, ‘Does your body feel different or changed?’, ‘Do your surroundings appear different or changed?’, ‘Do you feel unusually sad or negative?’, ‘Do you feel unusually happy or positive?’, ‘Do you feel fearful?’, ‘Do you feel stimulated?’, ‘Are you feeling unusual thoughts?’

On the first treatment day prior to drug administration and at the final visit after the treatment protocol was completed (see (Murphy et al., 2023)), separate plasma and serum samples were taken. Whole blood was collected into K₂EDTA BD Vacutainer tubes for plasma BDNF and genetic DNA analysis. Samples were centrifuged immediately, and the plasma was pipetted into aliquots for BDNF analysis while the buffy coat was pipetted into ~1 mL aliquots for genetic analysis. Serum samples were held at room temperature for 1 h prior to centrifugation and then aliquoted for serum BDNF analysis. All aliquots were stored at −80°C until extraction.

Statistical analysis of pharmacodynamic measures

Linear mixed-effects modelling of the pharmacodynamic measurements was undertaken using the lmerTest package in R (Kuznetsova et al., 2017) with Group (LSD, placebo) and Time (0, 30, 60, 120, 240, 360 h) being treated as fixed effects with dummy coding, and participants as a random effect, with the primary estimates of interest being the group × time interaction effect.

Mass spectrometry assay

A liquid chromatography-tandem mass spectrometry (LC-MS/MS) method for quantitative determination of lysergic acid diethylamide (LSD) in human plasma was developed and validated. For further information, see Supplemental Materials.

Calibrated solutions (1 mg/mL in acetonitrile) of LSD, LSD-D3 (internal standard, IS), iso-LSD and 2-oxo-3-hydroxy-LSD (O-H-LSD) were purchased from Lipomed (https://www.lipomed-shop.com/de). LC-grade water (Millipore^®, Milli-Q system) and acetonitrile (LC-MS grade) were used for the mobile phase.

Chromatographic separations were achieved on an Agilent 1200 series Infinity LC system (Agilent Technologies, CA, USA) using an Alltima HP C18 (3.0 mm i.d. × 150 mm, 5 μm; Alltech, Grace) column with a guard column (C18, 4.6 mm i.d. × 10 mm, 5 µm). Column and autosampler were maintained at 20°C and 5°C, respectively. Mobile phase component A was 0.1% formic acid in LC-grade water and mobile phase component B was 100% acetonitrile.

Mass spectrum analysis was carried out using a positive (+) mode electrospray ionisation method and multiple reaction monitoring (MRM) modes on an Agilent 6410 triple quadrupole tandem mass spectrometer to monitor ion transitions of LSD, iso-LSD, O-H-LSD (a major urinary metabolite of LSD) and LSD-D3 (IS). Agilent MassHunter Data Acquisition software (version B.08.02) was used to control the equipment.

LSD could be quantified in all samples collected in the study. Two samples from each participant were re-analysed on a separate day as the original analysis to demonstrate the reproducibility of the bioanalytical method. The percentage difference among the initial and repeated concentrations of the samples was all within 15%.

Pharmacokinetic–pharmacodynamic model

Sublingual LSD pharmacokinetics were investigated using one- and two-compartment models with first-order elimination and first-order absorption. Pharmacokinetic models were parameterised in terms of elimination clearance (CL) from the central compartment, inter-compartment clearances (Q), central and peripheral volumes of distribution (e.g. V1, V2), absorption half-life (T_{1/2 ABS}) and a lag-time (T_LAG). Differences in size between individuals were described using theory-based allometric scaling of total body weight (equation (1)):

$$F_{\mathrm{size}}={\left(\frac{WTi}{W{T}_{\mathrm{STD}}}\right)}^{\mathrm{Exp}}$$ (1)

where F_size is the fractional difference in size when scaled using allometry, WT is the weight (kg) in the ith individual, and WT STD is a standard weight of 70 kg. The allometric theory-based exponent was fixed at ¾ for clearance parameters and 1 for distribution volumes (Anderson and Holford, 2008).

The delay in the LSD effect was described using an effect compartment quantified using an equilibration rate constant (keq). This was parameterised as an equilibration half-time (T_1/2 eq) (equation (2)):

$$T_{1/2}eq=\frac{\ln (2)}{k\mathrm{eq}}$$ (2)

LSD pharmacodynamics (heart rate and VAS ‘feel effect’) were described using a sigmoid maximal effect (E_MAX) model (equation (3)). VAS ‘feel effect’ observations greater than zero were retained in the model and used to estimate pharmacodynamic parameters:

$$\mathrm{Effect}=E_0+\frac{E_{\mathrm{MAX}}\times C{e}^{\mathrm{Hill}}}{C_{50}^{\mathrm{Hill}}+C{e}^{\mathrm{Hill}}}$$ (3)

where E₀ is the baseline heart rate or VAS, E_MAX is the maximal drug effect, Ce is the concentration of drug in the effect compartment, Ce₅₀ is the concentration of drug in the effect compartment at 50% of E_MAX and the Hill exponent describes the steepness of the concentration–response curve. Placebo effects on heart rate and VAS ‘feel effect’ in the inactive treatment groups were described using an exponential placebo model (equation (4)):

$$\mathrm{Effect}=P_{\mathrm{MAX}}\times \left(1+\mathrm{EXP}\left(-\mathrm{KPL}\times \mathrm{Time}\right)\right)$$ (4)

where P_MAX is the maximal placebo effect and KPL is the rate constant for the onset of the placebo effect.

PKPD model computation and model selection

Population parameter estimates were obtained using nonlinear mixed-effects models (NONMEM 7.5.1 ICON Development Solutions, USA) with first-order conditional estimation and a convergence criterion set to three significant digits. Population parameter variability (PPV) was described using an exponential model for the random effect variables (η); these variables were assumed to have a mean of zero and variance denoted by ω² (equation (5)):

$$P_i=P_{\mathrm{TV}}{e}^{\eta i}$$ (5)

where P is the parameter (e.g. CL) for the ith individual, P_TV is the typical value for that parameter and η is the random effects variable. Residual unidentified variability (RUV) was modelled using additive (θ_{RUV_SD}) and proportional (θ_{RUV_CV}) error models for LSD PK and effect. Between-subject variability in the residual error model (η_{PPV_RUV}) was estimated for each observation (equation (6)):

$$\mathrm{SD}ij=\sqrt{\left({\left({\mathrm{Obs}}_{ij}.{\mathrm{\theta }}_{\mathrm{RUV}\_\mathrm{CV}}\right)}^2+{\left({\mathrm{\theta }}_{\mathrm{RUV}\_\mathrm{SD}}\right)}^2\right).}{e}^{\eta {\mathrm{PPV}}_{\mathrm{RUV}i}}$$ (6)

where Obs _ij is the observation (concentration or effect measure) in the ith individual at the jth time. Individual predictions of concentration and effect were calculated using equation (7) with the random effects ( $\epsilon$ ) fixed to 1:

$$Y={\mathrm{Obs}}_{ij}+\mathrm{SD}ij\times \epsilon$$ (7)

A decrease in the value of the objective function (OFV) by 3.84 points provided by NONMEM indicated an improvement to the pharmacokinetic model. The likelihood ratio test with alpha = 0.05 was used to assess a significant improvement in the fit. A visual predictive check (VPC) for the PKPD model was generated based on simulations of model predictions including random effects. The biological plausibility of parameter estimates and inspection of VPC plots served as a guide during model building. Bootstrap methods were used to evaluate parameter uncertainty (Efron and Tibshirani, 1986). A total of 100 bootstrap replications were used to estimate parameter averages and confidence intervals (CI). Results from the population models are presented as parameter estimates, together with their 95% CI. PPV is expressed as an apparent coefficient of variation obtained from the square root of the variance estimate.

CYP genotyping

Extracted DNA was analysed using Agena MassARRAY (Agena Bioscience, San Diego, CA, USA). The Veridose^® Core Panel (Lois et al., 2021) was used to assess CYP2D6, CYP2C19, CYP1A2, CYP2B6, CYP2C9 and CYP3A4. This was paired with the Veridose^® CYP2D6 copy number variation (CNV) Panel (Everts et al., 2019) to permit the assessment of the total number of functional CYP2D6 haplotypes as well as CYP2D6 copies. Those with more than two copies are referred to as ‘hybrid’. Additional genes on the panel not of interest were masked (not able to be interpreted).

The terminology includes ‘extensive’ to describe the average or normal activity of the CYP enzyme. ‘Intermediate’ refers to slower than normal and ‘poor’ even slower or absent enzymatic activity. Intermediate-weak was added to describe slower than intermediate, but activity above poor. ‘Ultra-rapid’ metabolism is faster than normal. Note participants identified as having ‘hybrid’ CYP2D6 CNV are unable to have their precise activity reliably estimated if they have two differing alleles determining their diplotype and are instead identified as ‘hybrid’.

Activity scores for CYP2D6 were calculated (Caudle et al., 2020) from the Agena Bioscience^® software, and CYP2C19 activity status was determined. Both CYP2D6 and CYP2C19 were assessed according to current (2024) Clinical Pharmacogenetics Implementation Consortium (CPIC) guidelines (https://www.pharmgkb.org/). For CYP2D6, additional distinction within intermediate metabolisers was made (Molden and Jukic, 2021) to identify weaker (activity score 0.5) intermediate metabolisers.

CYP1A2 (rs2069514, rs762551, rs12720461, rs56107638 and rs72547513), CYP2B6 (rs28399499 and rs3745274), CYP2C9 (rs1799853, rs1057910, rs56165452, rs28371686, rs9332131, rs7900194, rs28371685, rs9332239, rs72558187, rs1304490498 and rs72558190) and CYP3A4 (rs55785340, rs4987161 and rs35599367) were classified as wild-type or variant carriers.

BDNF quantification

Plasma and serum were analysed using R&D Systems Quantikine ELISA, Human Free BDNF Immunoassay (R&D Systems 2024). Plasma was centrifuged for 10 min at 10,000 g and 10 µL pipetted from the middle of the tube to avoid precipitated platelets. Plasma and serum samples were diluted 1:20 and steps were carried out according to the manufacturer’s protocol. Samples were analysed in duplicate. Outliers and samples below limit of detection (LOD) were rerun to confirm values. Inter- and intra-assay values were ⩽6.1%. Mean plasma intra-assay = 4.5%; inter-assay = 5.3%. Mean serum intra-assay = 4.1%; inter-assay = 3.3%.

Results

Mass spectrometry and drug concentration data for PK

Blood samples were obtained from 79 enrolled participants. Samples could not be obtained from one individual due to failure to fit a cannula. All blood samples were obtained as intended at the post-intervention time points except for one sample at the 120-min time point and one at the 360-min time point. All these samples were from the LSD group. Therefore, the quantification rate for samples that were intended to have LSD in plasma was 96.5% (193/200). Raw concentration values are shown in Figure 1(a). Using our LC-MS/MS method, LSD was quantified in 100% of the samples that we successfully collected in the active treatment group.

Figure 1.

(a) Mean plasma LSD concentration for n = 39(/40) participants in the MDLSD trial LSD cohort. All individual values are in grey. Error bars represent standard deviation. (b) Individual plasma concentration curves of all participants with intermediate-weak metabolisers of CYP2D6 (n = 2) are indicated in red and hybrid allele carriers in blue, extensive metabolisers are in grey and SNP fails are not shown. (c) Same as (b) except with participants (n = 4) withdrawn from the trial indicated in red. (d) Same as (b) except with participants (n = 6) withdrawn from the trial indicated in red.

LSD: lysergic acid diethylamide; SNP: single nucleotide polymorphism.

PKPD model

LSD pharmacokinetics were best described using a one-compartment model. A two-compartment model resulted in a statistically significant decrease in OFV compared to a one-compartment model (ΔOFV 11.4, p = 0.003); however, this model was over-parametrised indicated by poorly estimated intercompartmental clearance and peripheral volume of distribution. Additionally, there was no difference in visual predictive check plots (VPCs) between the one- and two-compartment models. Allometric scaling of total body mass was used to account for size-related changes in LSD pharmacokinetics. Visual predictive checks for the pharmacokinetic model are shown in Figure 2 with the final pharmacokinetic parameter estimates and their PPV shown in Table 1. The pharmacokinetic parameters clearance and volume of distribution were used to predict the C_max (0.20 µg/L ± 0.13), T_max (1.51 h ± 0.66) and T_1/2 (3.09 h ± 0.37).

Figure 2.

VPC for the final LSD PK model.

Plots show median (solid) and 90% intervals (dashed lines). The left-hand plot shows observed LSD concentrations. The right-hand plot shows percentiles (10%, 50%, and 90%) for observations (red dashed lines) and predictions (black dashed lines) with 95% confidence intervals for prediction percentiles.

PK: Pharmacokinetic; LSD: lysergic acid diethylamide; VPC: Visual predictive check.

Table 1.

Parameter estimates for the LSD pharmacokinetic model.

Parameter	Units	Estimate	Bootstrap median	95% Confidence interval	PPV
CL	L/h/70 kg	7.78	7.66	6.75, 8.77	0.37
V1	L/70 kg	32.9	32.8	30.1, 36.0	0.33
T ABS	h	0.47	0.46	0.36, 0.57	0.39
Proportional residual error	%	14.6	13.5	7.1, 17.5	0.16
Additive residual error	µg/L	0.001	0.003	0.001, 0.009

Parameter estimates are displayed as individual estimates and bootstrap medians. The 95% confidence intervals and median parameter estimates are from 100 bootstrap replications. Clearance and volumes of distribution are confounded by bioavailability (F) which is unknown.

PPV%: Population parameter variability; CL: Clearance; V1–V2: volumes of distribution; intercompartmental clearance Q.

Pharmacodynamic VAS measurements are presented in Figure 3. The time course of heart rate and blood pressure changes are shown in Figure 4. The final pharmacodynamic parameter estimates and their PPV are shown in Table S3. Visual predictive checks for heart rate and VAS ‘feel effect’ are shown in Figures 5 and 6.

Figure 3.

VAS ratings of subject effects were measured from 16 VAS ratings (0–100) corresponding to each PK sampling point.

Due to the wide variability in the mean and range within responses across different scales, different y scales are used for each. Means and standard deviations are presented with data presented as change from the baseline (time = 0) score. Linear mixed models showed interaction effects for body different (120 min, p = 0.0173), dizzy (60 min, p = 0.043), feel effect (240 min, p = 0.036), see movement (60 min, p = 0.049), unusually sad (60 min, p = 0.045) and want more (60 min, p = 0.050; 360 min, p = 0.027) (all uncorrected).

PK: pharmacokinetic; LSD: lysergic acid diethylamide; VAS: visual analogue scale.

Figure 4.

Vital signs (diastolic blood pressure/systolic blood pressure/heart rate) were measured at each PK sampling point.

Means and standard deviations are presented with data presented as change from the baseline (time = 0) score. Heart rate showed an interaction effect at 60 min (p = 0.028).

PK: pharmacokinetic.

Figure 5.

VPC for the final LSD pharmacodynamic (PD) model for changes in heart rate.

The upper panel shows heart rate changes in the active treatment group (heart rate, HR), and the lower panel shows changes in the inactive treatment group (heart rate placebo, HRP). Plots show median (solid) and 90% intervals (dashed lines). The left-hand plots show heart rate observations. The right-hand plot shows percentiles (10%, 50% and 90%) for observations (red dashed lines) and predictions (black dashed lines) with 95% confidence intervals for prediction percentiles.

BPM: beats per minute, LSD: lysergic acid diethylamide; VPC: visual predictive check.

Figure 6.

VPC for the final LSD PD model for changes in ‘feel effect’ measured using VAS.

The upper panel shows VAS in the active treatment group (VAS feel effect, VASF), and the lower panel shows changes in the inactive treatment group. Plots show median (solid) and 90% intervals (dashed lines). The left-hand plots show VAS observations. The right-hand plot shows percentiles (10%, 50% and 90%) for observations (red dashed lines) and predictions (black dashed lines) with 95% confidence intervals for prediction percentiles.

BPM: beats per minute, LSD: lysergic acid diethylamide; VPC: visual predictive check; VAS: visual analogue scale.

CYP Genotyping

A complete summary of the CYP genotype results is provided in Table S3. Activity score calculation from CYP2D6 diplotypes revealed 21/40 were extensive metabolisers, 13/40 were intermediate metabolisers and 0/40 were identifiable as poor or ultra-rapid metabolisers. While one participant did have a *3/*4 diplotype consistent with poor metabolism, they also had a CNV which meant their diplotype could not be used to classify their CYP2D6 activity. This lack of poor or ultra-rapid metaboliser substantially limited our further analysis.

There were two single nucleotide polymorphism (SNP) failures. 5/40 had CNVs including three exon 9 exchanges (one of which was also a SNP failure) and two with *68. In general, no further interpretation of these hybrid alleles could be made and so they may be poor, intermediate, normal or ultra-rapid metabolisers. The one exception was a *10/*10 so where a duplication occurred at exon 9 they could be classified as at most 0.5 a weak intermediate metaboliser (Turner et al., 2023).

Recognising the phenotypic complexity of the intermediate metaboliser category (Molden and Jukic, 2021), participants with 0.5 activity scores (2/13) versus 1 activity scores (11/13) were differentiated (Figure 1(b)). Hybrid and intermediate-weak (activity score = 0.5) metaboliser plasma concentration curves are identified in Figure 1(b). CYP2C19 had 10 extensive metabolisers and 29 intermediate metaboliser with 1 SNP fail. Metabolism status had no apparent effect on pharmacokinetics (Figure 7).

Figure 7.

Mean plasma LSD concentration for n = 39(/40) participants in the MDLSD trial LSD cohort classified by CYP2C19 activity status

Blue indicates intermediate metabolisers and red extensive metabolisers. Error bars represent standard deviation.

LSD: lysergic acid diethylamide.

Considering the role of CYP1A2, 2C9, 2D6 and 34A enzymes in the metabolism of LSD (Luethi et al., 2019; Wagmann et al., 2019), concentration–time curves are plotted and described considering wild-type and variant carriers. CYP1A2 rs762551 had 25 A/A, 14 C/A and 1 C/C. CYP1A2 rs2069514 had 3 A/A, 2 G/A and 35 G/A. rs12720461, rs56107638 and rs72547513 had no incidence of variant phenotypes in our sample. The curves are as would be predicted if on average some of the rs762551 A/A carriers had induced metabolism of LSD, with on average lower concentrations (Figure 8(b)). CYP1A2 activity is induced in A/A carriers in smokers (Dobrinas et al., 2011), heavy caffeine users (Djordjevic et al., 2008) and with some medicines (Granfors et al., 2004). While only caffeine use was permitted in our sample this is supported by Figure 8, which highlights the two participants with heavy daily caffeine use (greater than five servings). Most participants drank coffee daily. No participants were smokers.

Figure 8.

(a) CYC2D6 blood concentrations over time, plotted according to activity score indicating normal extensive metabolism in red. Or intermediate-weak metabolisers with an activity score of 0.5 in blue. Or where participants where participants had copy number variations, indicated as hybrids in grey. (b) CYP2C19 is plotted according to activity score classifying them as extensive (red) or intermediate (blue) metabolisers. (c–e) Wild types are in red, variant heterozygotes are in blue and variant homozygotes are in grey. While 40 participants had SNPs tested, one participant did not have pharmacokinetic blood samples. Thus, each panel (a)–(e) will have 39 samples unless there was a failure to determine the SNP for an individual in which case n < 39.

SNP: single nucleotide polymorphism.

CYP2B6 rs3745274 had 20 G/G, 18 G/T, 1 T/T and 1 SNP failure. Qualitatively, the curves in Figure 8(d) are largely what would be predicted based on studies on propofol metabolism and this genotype (Poma et al., 2022). G/T participants have higher concentrations on average than G/G participants. The T/T participant has an average rather than even higher concentration, demonstrating the complicated pharmacokinetics of LSD. Rs28399499 had no incidence of variant phenotypes in our sample.

CYP2C9 rs1057910 had 36 C/C and 4 T/C. Rs179985 had 36 A/A, 3 C/A and 1 C/C. Again, there is qualitative evidence of predicted differences in concentration based on research on propofol (Poma et al., 2022), the three T/C participants have higher concentrations on average than C/C (Figure 8(f)). Rs56165452, rs28371686, rs9332131, rs7900194, rs28371685, rs9332239, rs72558187, rs1304490498 and rs72558190 had no incidence of variant phenotypes in our sample.

CYP3A4 had rs35599367 38 GG and 2 A/G. Rs55785340 and rs4987161 had no incidence of variant phenotypes in our sample.

Participants withdrawn from the trial are shown in Figure 1(c) and participants who entered into the titration protocol in Figure 1(d).

BDNF quantification

There were 214/240 (89%) plasma BDNF observations amenable for statistical analysis (five samples were not collected, 18 samples were below the level of quantification (1250 pg/mL) and three outliers with values >10,000 pg/mL were removed prior to analysis).

Linear mixed models showed no interaction effect at either the 6 h timepoint (t = 0.14, p = 0.89) or the 43-day timepoint (t = 0.40, p = 0.68). There was an effect of time at the 6 h timepoint (t = 2.15, p = 0.03) but not at the 43-day timepoint (t = 0.79, p = 0.43), nor was there a group effect (t = −0.48, p = 0.63) (see Figure 9(a)). These results were not changed by imputing the value 1250/2 for samples below the level of quantification.

Figure 9.

(a) Plasma BDNF concentration for each group in the MDLSD trial prior to the intervention (Hour 0), 6 h after first dosing (Hour 6) and 2 days after completion of the treatment regimen (Day 43) and (b) same as a but for serum BDNF.

BDNF: brain-derived neurotrophic factor.

For serum BDNF analysis, of the potential 240 samples, five samples were not collected, four samples were below the level of quantification and seven samples were outliers with values <10,000 pg/mL which were removed prior to analysis. This left 224/240 (93%) samples available for statistical analysis. Linear mixed models showed no interaction effect at the 6 h timepoint (t = 0.234, p = 0.81) and an interaction effect at the 43-day time point (t = 2.45, p = 0.016). There was no group effect (t = −1.66, p = 0.10) and no effect of time at 6 h (t = 0.541, p = 0/58) or 43 days (t = 0.05, p = 0.96). The means of the interaction effect at Day 43, however, suggest it was caused by a low mean baseline in the placebo group rather than an increase in serum BDNF in the LSD group (placebo pre-mean = 26,496 (6676) pg/mL; placebo post-mean = 28,528 (6661) pg/mL; LSD pre-mean = 28,412 (6284) pg/mL; LSD post-mean = 28,482 (5767) pg/mL) (see Figure 9(b)).

Discussion

The potential for microdoses of LSD to be developed as medicines for mental health disorders has renewed interest in the PK and PD of LSD given in different formulations. In addition, understanding the factors that may mediate response and tolerance to LSD may pave the way to precision medicine approaches to using microdosing of LSD in treating mental illness.

This study validates the findings of Murphy et al. (2023) by providing the accompanying blood sample concentration of LSD. We successfully collected 96.5% of the samples we intended, to quantify LSD concentrations. We were able to quantify LSD in 100% of the analysed samples. We present the high-sensitivity LC-MS/MS assay we used for the detection of low concentrations of LSD in plasma along with validation data. The assay is suitable for use in microdosing studies. Furthermore, the assay demonstrated the detection of OH-LSD. Although it was unlikely that OH-LSD metabolite would be detected in plasma samples following a microdose of LSD, this method could be used as a starting point for future plasma/urinary analyses that require the quantification of this metabolite.

We demonstrate that a one-compartment model was found to best describe the pharmacokinetics of microdosed sublingual LSD, consistent with previous studies that used single or non-compartment modelling for microdosing with LSD (Family et al., 2020; Holze et al., 2021a). One previous study on intravenous LSD used a two-compartment model (Levy et al., 1969). We had insufficient samples of the elimination phase to consider a two-compartment mode as in Levy et al. (1969). Unfortunately, even if we had these samples the overall lack of detectable effect at low concentrations of LSD (in our VAS and vital sign measures of PD) would still limit further inference on the nature of the PK compartments (in physiological terms).

Our analysis extends those in the literature by accounting for size-related changes in LSD pharmacokinetics using theory-based allometric scaling of total body weight and quantifying the influence of inactive treatment on effect (Anderson and Holford, 2008). Pharmacokinetic parameters such as clearance and volume of distribution are consistent with previous modelling studies (Dolder et al., 2015; Holze et al., 2019; Holze et al., 2021a). The C_max of 204 pg/mL was consistent with Holze et al., (2021a), who also reported a 10 µg dose (their C_max was 151 pg/mL). T_max at 1.5 h sat midway between the studies led by Liechti and his group at the University Hospital of Basel (Dolder et al., 2015; Holze et al., 2019; Holze et al., 2021a; Holze et al., 2024) (Ms = 1.1–1.7 h). By contrast, one other group has reported fast T_max = 50.4 min (Family et al., 2020). T_1/2 of 3 h was consistent with several previous papers (Dolder et al., 2015; Family et al., 2020; Holze et al., 2019; Holze et al., 2021a; Holze et al., 2024).

We used heart rate and VAS scales (specifically whether the participants were able to ‘feel effect’) to describe LSD pharmacodynamics. The maximal increase in heart rate and VAS were small (<15%) compared to baseline estimates; the lack of change is likely attributed to the low dose of LSD administered and the consequent concentrations. Furthermore, one noticeable difference between the VAS data we report compared (Figure 3) to the data presented in Holze et al. (2021a) is a much stronger response in the placebo group. This could be due to several, possibly interacting factors. Firstly, our study used a parallel-groups trial design which is better able to blind participants compared to a crossover trial where participants can use their psychological responses in previous sessions to help them guess the identity of subsequent intervention sessions (Muthukumaraswamy et al., 2021). Secondly, demographic factors may be important as Holze et al. (2021a) used a young (M = 23 years, SD = 3) mixed-sex cohort where all participants had previous psychedelic use experience, whereas the MDLSD trial had a wider age range (M = 36, SD = 7) of male participants of whom 30% in each group were psychedelic naïve.

Incorporating covariates (such as genotypes or phenotypic traits) that can be used to predict response in the individual patient is one of the most sought-after outcomes of modelling in addition to determining PD measures that are sensitive to microdoses. CYP2D6 has been studied extensively for the potential usefulness of classifying metabolism status by activity score (Caudle et al., 2020). A poor, intermediate or ultra-rapid status has been reported as potentially useful in guiding clinical decision-making (Arnone et al., 2023). Individuals with CYP2D6 poor metaboliser status have increased LSD plasma concentrations and decreased LSD clearance (Vizeli et al., 2021).

The only participant with a CYP2D6 activity score of 0 (indicating a poor metaboliser) from their core panel-derived diplotype also had a copy-number variation meaning they could not be accurately classified from their diplotype. We did not incorporate activity scores into the modelling because of these low numbers. However, there were two participants in our cohort who were intermediate-weak (activity score = 0.5) metabolisers and they were both among the four participants who were withdrawn from the study due to over-stimulation/anxiety (Figure 1(b)). This finding is potentially relevant to the design of LSD microdosing therapeutic interventions if the tendency for overstimulation to occur in people with activity scores 0–0.5 was replicated in larger samples. In such cases, CYP profiles could be added to models.

It is also possible in the future that potential patients entering into an LSD microdosing regimen could be genotyped for CYP2D6 genotype status prior to starting a treatment regimen and be started at a lower dose to reduce adverse effects if they are found to be poor metabolisers. Visual inspection of Figure 1(c) and (d) also reveals the importance of LSD plasma concentration in terms of generating over-stimulation effects during home microdosing protocols. This suggests that dose titration/escalation protocols should be used to reduce adverse effect incidence and allow patients to find an individually optimised dose—an approach we are taking in ongoing Phase 2 trials (Donegan et al., 2023).

Overall, it is apparent that more than CYP2D6 activity scores are needed to predict LSD concentration given the wide variability in individual concentration, including around Figure 1(b) intermediate-weak metabolisers. From our results, CYP2C19 clearly had no effect on the concentration following the first microdose of LSD. However, CYP2C19, CYP1A2 and CYP3A4 are induced and so in a less controlled sample (i.e. in a clinical or recreational setting), these genotypes are likely to become more important, particularly 1A2 and 3A4 (Luethi et al., 2019; Wagmann et al., 2019). CYP1A2 rs762551 variant carriers did have qualitatively higher concentrations of LSD (Figure 8) which may reflect some induction.

CYP2B6 rs3745274 and CYP2C9 variant carriers showed either a less clear spread, or low numbers, respectively, so were less compelling, but were trending in predicted directions (Luethi et al., 2019; Wagmann et al., 2019). CYP3A4 had two extreme values in the variant carriers (the highest AUC and the fifth lowest). As summarised in Table S4, individual participants had a complex profile of CYP genotypes. The data support larger studies with similar panels, where the average contribution of different SNPs to drug concentration in plasma could be quantified, and it could be determined if it was useful to inform titration regimens for LSD microdosing. Luethi et al. (2019) report that CYP2E1 is also important to LSD metabolism and should be included in future studies.

The current study found no evidence to support the hypothesis that LSD microdosing causes an increase in peripheral BDNF concentrations 6 h after a 10 µg dose nor after completing a 6-week regimen of microdosing. One previous study (Hutten et al., 2021) found increases in plasma BDNF concentrations in the hours following microdosing. However, despite that trial containing 24 participants, many samples were not available for assay. Statistical analyses were conducted using a complete case analysis of n = 10 for the 5 µg dose, n = 9 for the 10 µg dose and n = 8 for the 20 µg dose with a 37% data availability rate. The lack of available data may have increased the bias within the study and consequently, results may not reflect the true population time course of BDNF. Conversely, this study had a larger sample size n = 40 with high data availability (~90%). From the data, it is clear that there is high variability even at baseline (Figure 9)—reinforcing the need for high numbers and within-subject analyses when assessing peripheral BDNF. Furthermore, more recent studies with macrodosing have largely failed to see an increase in plasma or serum BDNF following 100–200 µg doses (Becker et al., 2023; Holze et al., 2022; Ley et al., 2023; Straumann et al., 2023).

This study measured BDNF concentrations in both plasma and serum as it is not clear which sample type might be most sensitive to psychedelics. Most peripheral BDNF is stored in platelets (which cannot cross the blood–brain barrier) and as such plasma BDNF might be a better proxy of acute drug effect changes (Gejl et al., 2019). The coagulation processes that occur during serum sample processing causes BDNF to be released from platelets into serum allowing this pool of BDNF to be accessed. Plasma does still contain some platelets and as yet no studies with psychedelics have measured platelet-poor plasma in which these extra platelets have been removed by additional centrifugation (Gejl et al., 2019).

The current study did observe plasma BDNF concentrations increased only at the 6-h timepoint in both groups, an effect which might be due to some non-specific study effect. Particularly because in males, plasma BDNF is expected to decrease from morning to afternoon (Choi et al., 2011; Piccinni et al., 2008), and serum BDNF is expected to remain stable (Choi et al., 2011), which it did. Despite this study not identifying the effects of LSD microdosing on BDNF concentrations, clinical populations such as those with depression (rather than healthy volunteers) may exhibit a different time course of BDNF since that population has been shown to have lower peripheral BDNF concentrations (Cavaleri et al., 2023).

Additional strengths, limitations and future research

The current paper provides a population pharmacokinetic model, however, compared to works such as Holze et al. (2021b), we would have benefited from additional blood samples during the elimination phase, as well as additional data with varying doses. The strength of this work is likely to be realised with implementation in further studies with high numbers of variables that may influence PD. As well as a direct investigation of factors that may influence or be influenced by PK, such as regimen and formulation choices.

A strength of the current study was the number of participants who provided LSD concentration–effect data to derive a pharmacokinetic–pharmacodynamic model. However, a major limitation is the lack of CYP2D6 variant/activity score representation. Specifically, we had low numbers of intermediate metabolisers and no poor or ultra-rapid metabolisers. Copy number variation was found in five participants including three with exon 9 exchanges (one SNP fail) and two with *68. They are potential ultra-rapid or poor metabolisers. A strength of identifying these hybrids, however, is that without assessing copy number variations, we potentially would have wrongfully classified them based on their diplotype determined in the core panel alone. For example, one of the hybrid participants would be classified as a poor metaboliser had we not assessed copy number variations.

Studies of a larger size are required to assess any pharmacogenomic effects, for example, the influence of CYP2D6 genotype status on LSD clearance. The current data provide a rationale for continued study into whether activity scores and CYP SNPs could be used to inform future prescribing of microdoses of LSD.

Future studies are warranted to find a PD metric that is informative on the physiological effects and mechanisms of microdoses of LSD. Such information would be useful in establishing the effects of different formulations and if future therapeutic uses are found for microdosing, PKPD models would be additionally useful in determining effective concentrations and establishing rationalised optimal doses, regimens and formulations. Modelling could also indicate an effective mechanism.

As described in Murphy et al. (2023), this study only included males. While Holze et al. (2019) demonstrate there appear to be no PK differences in females vs males, to the author’s knowledge, PD differences have not been investigated. Diurnal effects of BDNF have also been found to vary with sex, with plasma concentrations decreasing across the day in males but not females (Choi et al., 2011; Piccinni et al., 2008), and not for either sex in serum (Choi et al., 2011). Future studies should study females, not just by including them, but by accounting for the effect of being female on the study variables in the design and statistical analysis. This means considering the menstrual cycle phase and life stage (such as reproductive, perimenopausal and postmenopausal).

Conclusion

With growing interest in the potential therapeutic applications of microdosing psychedelics comes the need for renewed study into the PK and PD of candidate drugs such as LSD. This study provides a population PKPD model, and LC-MS/MS assay that can be taken forward into clinical and bioequivalence studies. CYP2D6 activity score remains worthy of continued large studies into whether poor or intermediate-weak metabolism may serve as a biomarker of response. Further studies ought to find sensitive and reliable PD measures that can be densely sampled and coupled with PK modelling.