Self-blinding citizen science to explore psychedelic microdosing

There is renewed scientific and clinical interest in psychedelic drugs such as LSD and psilocybin, principally in the context of psychedelics-assisted psychotherapy where a few large doses are given alongside psychotherapy. An alternative pattern of use, "microdosing", has become popular; it generally refers to repeated low doses (commonly 10-20% of a typical full dose) taken one to three times per week. Anecdotal reports and uncontrolled observational studies have suggested benefits for well-being, creativity and cognition, but these designs are vulnerable to expectancy, confirmation-bias and placebo effects, and randomised controlled laboratory studies to date have been small and focused on single acute doses rather than repeated use. The combination of (i) uncontrolled naturalistic evidence and (ii) underpowered controlled trials leaves uncertainty about whether reported benefits of microdosing are pharmacological or placebo-driven. Szigeti and colleagues set out to address these limitations by developing a low-cost, large-sample, placebo-controlled citizen-science approach they call "self-blinding". The study aimed to test whether repeated psychedelic microdosing produces superior accumulative, acute and post-acute effects compared with placebo on psychological state and cognitive function. The primary endpoint was after a core 4-week dosing period (week 5 of a 10-week protocol), and the investigators hypothesised that improvements from baseline would correlate with the number of microdoses taken and that acute/post-acute outcomes would be better on/after microdose days compared with placebo days. The design combined prospective online assessments and cognitive testing with randomisation and participant-implemented blinding to obtain a large, ecologically valid sample at minimal cost.

The study used a three-arm, participant-randomised, self-blinding design. Participants prepared opaque capsules at home: a set of capsules containing their chosen microdose substance and a matching set of empty placebo capsules. Weekly capsule sets were sealed into envelopes with non-human-readable QR codes; a constrained semi-random draw produced one of three possible sequences corresponding to the three study groups (placebo-only, half/half, or microdose-heavy). Participants opened one envelope per week across a 4-week dose period and took capsules according to the envelope schedule. Scanning the QR code allowed the study team to reconstruct allocation after the fact without revealing it to participants. Adults (>18) with prior psychedelic experience who intended to microdose were recruited online and consented via a study website. Inclusion criteria included English proficiency, prior psychedelic use, and agreement not to use other psychedelics during the study period; no biochemical verification of substances or doses was performed. Participants were free to use any psychedelic and were instructed to use their usual microdose amount; reported average doses were ~13 ± 5.5 µg for LSD/analogues and 0.2 ± 0.12 g for psilocybin-containing mushrooms (psilocybin doses were converted to LSD-equivalents for some analyses). Recruitment resulted in an educated, largely male sample (mean age 33.5 ± 9.4; 70% male). Outcomes were organised by timescale. Accumulative outcomes (baseline, week 5 primary endpoint, optional week 9 follow-up) included Ryff's Psychological Well-being (RPWB), the Cognitive and Affective Mindfulness Scale (CAMS), Satisfaction With Life (SWL), Green Paranoid Thought Scales (GPTS), Big Five personality traits (B5) and a composite cognitive performance score (CPS) derived from six Cambridge Brain Sciences tasks, with learning effects removed. Acute outcomes were assessed weekly 2–6 hours after capsule ingestion and included PANAS (positive/negative affect), visual analogue scales (VAS) for drug intensity, mood, energy, creativity, focus and temper, and cognitive tasks. Post-acute outcomes were collected 48–72 hours after the last capsule of a week. Statistical analyses used mixed-effect repeated-measures models (SAS PROC MIXED) for accumulative outcomes with change from baseline as the dependent variable, and mixed linear models for acute/post-acute outcomes with subject as a random effect. Models adjusted for significant baseline covariates tested from a prespecified list (age, sex, education, baseline score, dose, short suggestibility scale, expectation score, psychiatric history, current psychiatric medications, lifetime macrodose experiences and months microdosing). Planned comparisons included within-group change from baseline to week 5 and week 9, and between-group comparisons (placebo vs half/half and placebo vs microdose-heavy). Participants were also asked to guess daily and weekly whether they had taken a microdose or placebo; guess data were incorporated into additional models and a 2x2 stratified analysis (condition x guess) to explore blinding integrity and expectancy effects.

Enrollment and completion numbers reported in the extracted text were n = 240 at baseline, n = 191 at the week 5 primary endpoint, and n = 159 at week 9. The sample was predominantly male (70%), middle-aged (mean 33.5 ± 9.4) and pro-psychedelic in attitude. Most participants microdosed LSD (n = 147, 61%) or LSD analogues (n = 33, 14%); 57 participants (24%) used psilocybin-containing mushrooms and three used other psychedelics. Average reported microdoses were ~13 ± 5.5 µg LSD-equivalent for LSD/analogues and 0.2 ± 0.12 g for mushrooms. Randomisation produced no significant baseline differences across groups. Accumulative (within-group) results: In the microdose-heavy group (MD) from baseline to week 5, self-reported psychological outcomes improved significantly: RPWB increased by 4.2 ± 3.9 (p=0.04), CAMS by 2.4 ± 1.1 (p<0.001), SWL by 1.2 ± 1.2 (p=0.04), and GPTS decreased by −5.0 ± 1.7 (p<0.001). Personality changes included reduced neuroticism (−1.3 ± 0.9, p<0.01) and increased openness (0.9 ± 0.8, p=0.03). However, similar improvements (notably in mindfulness and reduced paranoia) were observed in the placebo (PL) and half/half (HH) groups: CAMS increased and GPTS decreased significantly in all groups, and neuroticism also fell in PL (−1.0 ± 1.0, p=0.04). Cognitive performance (CPS) did not improve in the MD group, and between-group comparisons for accumulative outcomes at week 5 showed no significant differences between PL and MD. Acute and post-acute results (initial models without guess): Acute measures taken on dose days showed higher scores under MD compared with PL for PANAS (acute emotional state; adj. mean difference 2.2 ± 1.4, p<0.01), and for VAS items drug intensity (12.5 ± 3.0, p<0.001), mood (4.6 ± 2.9, p<0.001), energy (5.3 ± 2.7, p<0.001) and creativity (4.7 ± 2.6, p<0.001). Effect sizes were small (Cohen's d < 0.3) for all scales except drug intensity (d = 0.58). Among post-acute measures, only trait-state anxiety (STAIT) differed (−1.4 ± 1.3, p=0.03) favouring microdose weeks. Blinding integrity and the role of guess: Participants were asked to guess daily and weekly; the overall break-blind rate (proportion of correct guesses) was 0.72 ± 0.18, higher than a simple random-guess expectation (reported comparison: random 0.62). Specificity (true negative rate for placebo) was 0.82 ± 0.16 and sensitivity (true positive rate for microdose) 0.45 ± 0.30. Break-blind rate correlated with reported dose (higher doses associated with better guessing), and an estimated detection threshold where guessing exceeded random was ~12 µg LSD-equivalent. When guess was added to acute/post-acute models, the condition effect (PL vs MD) lost significance for all scales except acute drug intensity (adjusted mean difference 3.4 ± 2.0, p<0.001). The guess*condition interaction was non-significant except for drug intensity. Stratified 2x2 analyses (condition by guess) showed that when guess was held constant there were no significant differences between PL and MD (except drug intensity), whereas when condition was held constant, outcomes were consistently better when participants believed they had taken a microdose (21 of 22 comparisons across acute and post-acute scales). Cognitive tasks and CPS, which are less subjective, showed no effects of either drug condition or guess. These patterns indicate that participants’ beliefs about taking a microdose largely explained the observed improvements, with a small genuine drug-related effect detectable only on subjective drug intensity.

Szigeti and colleagues interpret the findings as supporting the hypothesis that reported psychological benefits of repeated microdosing in a real-world sample are attributable primarily to placebo or expectancy effects rather than to a robust pharmacological action of low-dose psychedelics. Although the microdose group improved from baseline across a range of self-reported measures, parallel improvements occurred in placebo and half/half arms and no significant between-group differences were observed for accumulative outcomes. The initial acute and post-acute differences also disappeared after accounting for participants' guesses about capsule content, except for a small drug-related increase in subjective drug intensity. Cognitive outcomes, which are less vulnerable to subjective expectation, did not improve under microdosing. The authors place these results in the context of prior small controlled studies and uncontrolled observational work, suggesting that uncontrolled designs are especially vulnerable to expectancy bias in this area. They highlight the practical challenge that doses sufficient to produce detectable pharmacological effects may also compromise blinding, and note that participants’ ability to detect active capsules rose with dose, with an estimated psychoactive detection threshold around 12 µg LSD-equivalent in this sample. The discussion stresses that high break-blind rates, combined with positive attitudes toward psychedelics in the recruited sample, likely amplified placebo-like improvements. Key limitations acknowledged by the investigators include the inability to verify substance identity, purity or exact dose (participants sourced their own drugs), the lack of objective confirmation of adherence to the self-blinding procedure or capsule ingestion, and the observational nature of recruitment which limits clinical inference. They also note that guess was recorded after assessments, so causal direction between performance and guess cannot be established definitively, although many participants reported thinking about whether they had taken a microdose prior to being asked to guess. The sample was largely healthy with high baseline well-being, reducing scope for improvement; two post-hoc analyses in more vulnerable subsamples (lowest 25% well-being and highest 25% neuroticism) did not reveal between-group benefits. Finally, the authors advocate the self-blinding citizen-science method as a low-cost way to implement randomisation and placebo control at scale, useful as a screening or exploratory tool for trending interventions. They recommend future trials consider active placebos or different dosing strategies and place emphasis on assessing blinding integrity, especially given the interaction between expectancy and reported effects. Despite limitations, the investigators conclude that their largest placebo-controlled psychedelic study to date provides evidence that the positive self-reported effects of naturalistic microdosing are largely explained by expectancy/placebo mechanisms rather than direct pharmacological benefit.