Lysergic acid diethylamide (LSD) for alcoholism: meta-analysis of randomized controlled trials

Alcohol causes substantial global harm and remains difficult to treat: alcoholism (alcohol dependence) contributes notably to mortality and disability and many patients fail to achieve recovery with existing interventions. Earlier clinical investigators reported that single, supervised doses of lysergic acid diethylamide (LSD), given alongside psychosocial treatments, produced profound subjective experiences that were claimed to reduce relapse, but these reports had not been synthesised with modern systematic review methods. Krebs and colleagues therefore set out to perform a quantitative meta-analysis of randomized controlled trials to evaluate whether LSD, administered as a single dose within alcoholism treatment programmes, affects subsequent alcohol misuse. The aim was to pool trial data to estimate efficacy, examine follow-up intervals, and assess adverse events and risk of bias across the trials.

The investigators searched PubMed and PsycINFO without language restrictions using terms for LSD (including lysergic, lysergide, psychedelic*, hallucinogen*) combined with alcohol-related terms (alcohol*, addict*, dependence). Search results were screened by title and abstract, full texts of potentially relevant publications were retrieved, reference lists inspected, and experts contacted. The study protocol pre-specified inclusion and exclusion criteria and primary/secondary outcomes. Included studies were randomized controlled trials of LSD for alcoholism; control conditions could be any type of treatment and active controls of up to 50 mcg LSD were allowed. Participants with schizophrenia or psychosis were excluded from the meta-analysis. Two reviewers independently extracted trial data and assessed risk of bias using Cochrane domains (sequence generation, allocation concealment, blinding, incomplete outcome data, selective reporting). Extracted items included intervention details (dose, control, additional treatments), participant characteristics, outcome measures and timing, and trial features (year, location, funding). Primary outcomes were measures of alcohol misuse at the first reported follow-up; secondary outcomes were alcohol misuse at short-term (~3 months), medium-term (~6 months) and long-term (~12 months) follow-up. Abstinence and adverse events were also extracted where reported. Categorical alcohol outcomes were dichotomised into 'improved' versus 'not improved' based on categories indicating clear, substantial improvement. Continuous and categorical data were pooled using the generic inverse variance method with a random effects model to calculate pooled odds ratios (ORs) and 95% confidence intervals (CIs) in Review Manager 5.0. Heterogeneity was quantified with the I2 statistic (the percentage of total variation across trials attributable to between-trial heterogeneity). Participants lost to follow-up were conservatively counted as not improved. In a post hoc analysis the authors computed pooled benefit differences (risk differences) and derived the number needed to treat (NNT = 1/benefit difference) where dichotomised data were available. Description of the included trials was integrated into the methods. Six randomized trials (with additional reports on three) were included, totalling 536 adults (325 assigned to full-dose LSD, 211 to control). Most participants were male inpatients recruited from alcohol treatment programmes. Single oral LSD doses ranged approximately from 210 mcg (3 mcg/kg) to 800 mcg, with a median around 500 mcg. Preparatory, supportive and integration procedures varied markedly between trials: some provided minimal orientation and simple observation during the drug session, while others offered more clinical guidance, music and comfortable surroundings; review sessions after the drug session were generally limited. Outcome measures included maintained abstinence, alcohol use rating scales and composite alcohol/social adjustment scales; one trial reported a continuous percentage change in time abstinent.

Six eligible randomized controlled trials were pooled. For the primary outcome—improvement in alcohol misuse at the first reported follow-up—the pooled odds ratio comparing LSD to control was 1.96 (95% CI 1.36–2.84; p = 0.0003). Using dichotomised data from five trials, 185 of 315 (59%) LSD recipients versus 73 of 191 (38%) control patients were classified as improved at first follow-up; the pooled benefit difference was 16% (95% CI 8%–25%; p = 0.0003), corresponding to a number needed to treat of six. When analysed by follow-up interval, LSD showed a significant effect at short-term follow-up (2–3 months): pooled OR 1.85 (95% CI 1.14–3.00; p = 0.01), and at medium-term follow-up (6 months): pooled OR 1.66 (95% CI 1.11–2.47; p = 0.01). The effect was not statistically significant at long-term follow-up (approximately 12 months): pooled OR 1.19 (95% CI 0.74–1.90; p = 0.47). Between-trial heterogeneity was negligible for the primary, short-term and medium-term analyses (I2 = 0%); heterogeneity was low at long-term follow-up (I2 = 15%). Among three trials reporting maintained abstinence, LSD was associated with higher odds of abstinence at the first reported follow-up (OR 2.07; 95% CI 1.26–3.42; p = 0.004) and at short-term follow-up (OR 1.80; 95% CI 1.07–3.04; p = 0.03), but not at medium-term follow-up (OR 1.42; 95% CI 0.65–3.10; p = 0.38). Heterogeneity for abstinence was negligible at first and short-term follow-up (I2 = 0%) and moderate at medium-term follow-up (I2 = 44%). Five trials reported eight acute adverse reactions to LSD without lasting harm. Reported acute events included two participants who 'acted bizarrely', agitation in one participant, a grand mal seizure in one patient who had a prior history of alcohol-withdrawal seizures and very short alcohol abstinence, and other unspecified transient adverse reactions; mild nausea, vomiting and agitation were also reported and were generally relieved by supportive measures. One trial noted that around one-third of LSD recipients briefly experienced vivid perceptual or thought experiences on one or a few occasions within a year after LSD, typically after drinking alcohol. No trials reported lasting detrimental effects on psychosocial functioning; two of three trials reporting employment outcomes found improved employment after LSD. Risk of bias assessments found no trials with high risk for sequence generation or allocation concealment, though methods were often underdescribed. Two trials had high risk for inadequate blinding because allocation was concealed only until the possible LSD session; four trials used double-blind designs with active placebos. Outcome assessment blinding was low or unclear risk in most trials; several used allocation-blind interviewers. Two trials had high risk due to incomplete outcome data because participants who did not complete intended treatment or who received additional LSD doses were excluded. Two trials were judged at high risk for selective outcome reporting, and one trial showed baseline imbalance on some sociodemographic variables though groups were matched on baseline alcohol misuse. Sensitivity analyses showed the primary outcome remained statistically significant when excluding any two of the four larger trials or when excluding trials judged to have high risk on specific bias domains. Excluding the two trials with non-blinded control conditions increased the LSD effect. Secondary outcomes were more sensitive: removing the single trial with the most favourable result in each secondary analysis rendered those outcomes non-significant (p ≥ 0.06); secondary analyses were based on only three to five trials each.

Krebs and colleagues interpret their pooled analysis as evidence that a single supervised dose of LSD, given within a range of alcoholism treatment programmes, is associated with reduced alcohol misuse at first follow-up and up to about 6 months, with effects that were not statistically significant at 12 months. The beneficial effect on maintained abstinence was seen at early follow-up (1–3 months) but not at 6 months. Authors note the pattern—initial benefit that may wane by 12 months—is consistent with earlier non-randomized and open-label reports and with a quasi-randomized study showing benefit at 3 months. Possible reasons the approach has been overlooked are discussed: the individual randomized trials were underpowered and often non-significant on their own; investigators may have discounted moderate or transient effects; early non-randomized studies were poorly reported creating confusion about the evidence base; and sociopolitical barriers complicated regulatory approval for LSD research. The authors state that a single dose of LSD compares favourably with commonly prescribed pharmacotherapies for relapse prevention (naltrexone, acamprosate, disulfiram) based on pooled benefit-difference calculations cited in the paper. On mechanism and subjective effects, the investigators remark that many participants reported insights, increased self-acceptance and motivation following the LSD experience; they also note similarities in subjective and neurobiological effects between LSD and other classic psychedelics (mescaline, psilocybin, dimethyltryptamine) and ethnographic claims that such substances have been used to support sobriety. Regarding safety, the authors state that LSD is widely regarded as physically safe but can produce acute psychiatric adverse events such as anxiety or confusion; they recommend administration in comfortable settings with informed participants. Key limitations acknowledged by the authors include limited trial descriptions of populations and diagnostic methods, too few trials to explore dose-response or treatment-context effects, the possibility of unpublished or missed trials, variable preparation and support across trials (including some trials that concealed the possibility of LSD or provided minimal preparation), difficulties with blinding (some trials used low-dose LSD as active placebo which may attenuate observed differences), and heterogeneity in primary outcome measures. The authors recommend further research, including trials testing repeated LSD doses, stratified analyses to identify subgroups with differential benefit or risk, combinations with contemporary relapse-prevention treatments, and evaluation of shorter-acting psychedelics such as mescaline, psilocybin or dimethyltryptamine.