Control Conditions in Randomized Trials of Psychedelics: An ACTTION Systematic Review

Psychedelics have attracted substantial interest as potential therapeutics because clinical trials report preliminary signals of benefit across diverse conditions. However, methodological concerns have been raised about control conditions in randomized psychedelic trials, notably the risk of unblinding from powerful subjective drug effects and the difficulty of separating drug-specific effects from non-specific effects arising from intensive interpersonal support during treatment sessions. The similarity between psychedelic subjective effects and psychotherapeutic processes further complicates the selection of appropriate control arms, and different trial goals (assessment of the combined intervention, isolation of drug mechanisms, or regulatory efficacy evidence) place divergent demands on control design. Nayak and colleagues set out to systematically review the control conditions used in all available randomized human trials of classic psychedelics through May 2020. The review aimed to document what drug and non-drug comparators have been used, how blinding has been handled and assessed, whether psychological support elements were controlled or compared, and to use these findings to produce methodological recommendations to strengthen future trials.

The investigators conducted a systematic review following PRISMA guidance and preregistered the project on PROSPERO. They searched PubMed, PsycINFO and EMBASE without language limits for randomized trials of classic psychedelics in humans from 1940 through May 2020, supplemented by hand searches of other sources and prior reviews. Duplicate removal and part of the extraction process were managed with the Covidence platform. Full-text extraction occurred between 28 August 2020 and 30 April 2021. Two authors (S.M.N. and M.K.B.) independently screened and coded studies, with a third author (B.A.K.) arbitrating unresolved disagreements. Inclusion criteria required randomisation and dosing of human participants with a classic psychedelic (operationalised as a 5-HT2A agonist); there were no restrictions on participant population or language. Extracted items (when available) included study design and sample size, demographics, blinding type and any blind-integrity assessment, the psychedelic and dose, drug control and dose, non-drug control condition, number of dosing sessions, recruitment source, and for therapeutic trials additional details such as monitor assessment and number/hours of preparatory and integration sessions. Outcome data and formal study quality assessments were not collected. Analysis was descriptive only, limited to calculation of means, standard deviations, and percentages.

The search initially returned 1,350 records plus 7 manually identified studies; after automatic duplicate removal and screening, 126 articles were extracted. Those 126 articles represented 86 unique randomized studies (several datasets were reported across multiple articles and two articles each described two separate randomized studies). Across the 86 studies (conducted between 1963 and 2020) the mean (SD) number randomised was 34.5 (34.9), range 3–176. Seventy-nine studies (91.9%) reported sex; among these males comprised 71.3% of participants. Geographic distribution was concentrated in the United States (31 studies; 36%) and Switzerland (26; 30.2%), with smaller numbers in Spain, the UK, Germany, Brazil, Canada and several other countries. Only 12 studies (14%) reported racial composition; within those, White participants comprised 73.8% of the sample. Forty-seven studies (54.7%) reported prior psychedelic use among participants, and in those studies 53.8% of participants had previously used psychedelics. Regarding designs, 53 studies used within-subject crossover designs, 24 were between-subjects, and 9 combined randomised parallel groups with subsequent crossover. The psychedelics studied included psilocybin (36 studies; 41.9%), LSD (33; 38.4%), ayahuasca (10; 11.6%), non-ayahuasca DMT (6; 7.0%), DPT (2; 2.3%), mescaline (1; 1.2%) and 6-HDMT (1; 1.2%). Of 80 studies that included a drug control condition, 49 (57.0%) used an inert placebo, 16 (18.6%) used active comparators, 12 (14.0%) used both inert and active comparators, and 3 (3.5%) used only different active psychedelic doses as the control. Seven studies employed a low-dose active psychedelic as an active placebo (four low-dose psilocybin, two low-dose LSD, one low-dose DMT). Eighteen studies used differing active doses of the study psychedelic as comparators. Six studies used pretreatment manipulations to block psychedelic effects as a control, five of which used ketanserin (a 5-HT2A antagonist); other pretreatments reported included haloperidol, risperidone and agents such as buspirone or ergotamine in isolated examples. Fifteen studies used non-drug controls such as treatment as usual, no-intervention, waitlist, or differing psychosocial interventions (for example hypnosis or varying degrees of spiritual/psychological support). Within the therapeutic subset, 21 studies targeted defined medical conditions (predominantly substance use disorders or psychiatric symptoms); 12 of these 21 recruited from inpatient settings and one recruited recently released prisoners. Substance use disorder trials included several studies treating alcohol dependence and other SUD populations; non-SUD trials addressed anxiety or depression in medical illness, obsessive-compulsive disorder and one study of treatment-resistant depression. Blinding was reported in 81 original studies: 60 double-blind, 21 single-blind, and 5 unblinded. However, only 14 of the 81 blinded studies (17.3%) included any assessment of blind integrity, and methods for assessing blinding varied widely. When blind integrity was assessed, success was generally poor with many participants and/or monitors correctly identifying treatment assignment in several trials. Representative examples from extracted reports include therapist correct-identification rates of 95% in one older LSD study and participant correct-identification rates of 96% for LSD in a recent study; one psilocybin alcohol-dependence trial reported participants guessing correctly in 93.6% of first sessions with high certainty. Only 2 of 21 therapeutic studies examined whether individual monitors differed in effect; both reported no monitor effects. Reporting of preparatory and integration sessions was inconsistent: among the 21 therapeutic trials, 10 reported numbers of preparatory sessions and hours, and only 5 reported integration session counts or hours.

The authors conclude that randomized psychedelic trials have underused methodological elements that would strengthen inference about drug-specific effects. Key deficits identified include infrequent use of active drug controls, rare assessment of blind integrity, and the near absence of trials that manipulate or minimise psychological support to disentangle pharmacologic from non-pharmacologic effects. Blinding assessments that were performed tended to show poor maintenance of the blind, but success varied across studies and appeared influenced by participant psychedelic experience and expectations. Nayak and colleagues discuss trade-offs for different control strategies. Active comparators can improve masking but their success depends on expectations and participant experience and some active agents may themselves have therapeutic effects (for example certain NMDA antagonists), complicating efficacy interpretation. Low-dose active controls or dose–response designs are highlighted as plausible approaches that can both aid masking and support regulatory aims. Non-drug controls (treatment as usual, waitlist, minimal support) can address the non-pharmacologic components of psychedelic therapy but are difficult to blind and may introduce nocebo effects; only a small number of older studies trialled minimal psychological support conditions. The authors recommend routine, prompt assessment of blinding belief, certainty and reasoning for guesses among both participants and monitors, preferably on the day of dosing to avoid confounding by later efficacy. They advocate dose–response trials, measurement of therapeutic expectancies (for example the Credibility/Expectancy Questionnaire), use and further study of active placebos (such as oral THC), and complementary open-label pragmatic or comparative efficacy trials. Trials that compare the full psychedelic therapy package (preparation, supported dosing, integration) to minimally supportive drug conditions could clarify the contribution of psychotherapeutic elements, though the authors note ethical and safety considerations. Recruiting psychedelic-naive and clinician-referred patients is suggested as a strategy to improve masking and reduce expectancy bias. Limitations acknowledged by the reviewers stem from the primary literature: inconsistent reporting of participant characteristics (including prior psychedelic use and race), sparse reporting of blind-integrity assessments and inadequate description of non-drug controls, heterogeneity of study designs, and the possibility that some eligible randomized studies were missed. The review itself did not assess efficacy outcomes because outcome data were not extracted.