A comparison between psilocybin and esketamine in treatment-resistant depression using number needed to treat (NNT): A systematic review

Major Depressive Disorder (MDD) is common and often fails to respond to standard monoamine-based treatments, producing rates of treatment-resistant depression (TRD) estimated at 30–60%. Clinical definitions from regulatory agencies require failure of at least two adequate antidepressant trials for a diagnosis of TRD. Because TRD is associated with substantial disability and healthcare costs, there is interest in mechanistically novel treatments. Esketamine, an NMDA receptor antagonist, is an approved glutamate-based therapy for TRD, while psilocybin, a 5-HT2A agonist, is investigational. Preliminary studies suggest both agents produce rapid antidepressant effects that may converge on pathways involved in synaptogenesis and cellular plasticity (for example, BDNF/mTOR signalling), but direct head-to-head comparisons are lacking. Wong and colleagues set out to compare the clinical meaningfulness of psilocybin and esketamine for adults with TRD using Number Needed to Treat (NNT) and Number Needed to Harm (NNH) metrics. In the absence of randomized head-to-head trials, the authors performed a systematic review of randomized controlled trials (RCTs) that used the Montgomery–Åsberg Depression Rating Scale (MADRS) to quantify depressive symptoms, and then derived NNTs and NNHs from the included placebo-controlled studies to inform relative efficacy and tolerability.

The investigators conducted a systematic search of PubMed and BIOSIS up to 25 July 2023 for randomized controlled trials of psilocybin or esketamine in adults with TRD, using the search terms (Psilocybin OR Esketamine) AND (Treatment-Resistant Depression) AND ("Randomized Controlled Trial" [Publication Type]). No language or date restrictions were applied. Eligible trials had to be randomized, enrol participants diagnosed with TRD without psychotic features, and report outcomes using the MADRS. Excluded reports included abstracts without full text, case reports, preclinical studies, and non-RCT designs. Data were manually extracted from each included trial. Extracted items comprised treatment group sizes, participant mean age (when reported), dose/regimen, proportion of participants achieving a predefined clinical response on the MADRS at specified time-points, and rates of treatment-emergent adverse events (TEAEs). The authors required MADRS-based response data to standardise NNT calculations across studies. Clinical relevance was assessed by calculating NNT as the inverse of the absolute risk reduction (ARR) between active treatment and placebo (NNT = 1 / [response rate in treatment arm − response rate in placebo arm]). The authors defined a clinically relevant NNT as <10. NNHs were computed similarly from differences in TEAE rates between arms, and an NNH <10 was judged clinically relevant; TEAEs considered for NNH were those reported in ≥5% of the investigational group and with higher incidence than placebo. Confidence intervals for ARRs and derived NNT/NNH values were calculated using the Wald interval; intervals that included infinity were treated as not statistically significant. Each trial was analysed separately rather than pooled, and the authors produced a forest plot in RStudio. Risk of bias for included trials was assessed with the Revised Cochrane Risk-of-Bias Tool for Randomized Trials (RoB 2); one reviewer (S.W.) judged all included studies to be at low risk of bias.

The search yielded 791 records; after screening and eligibility assessment, five randomized controlled trials met the inclusion criteria and were analysed, comprising a total of 1,065 participants with TRD. The five trials included one psilocybin study and four esketamine trials. Studies varied in dose regimens, populations (including one trial restricted to participants aged ≥65 years), response definitions, and follow-up durations. Psilocybin (Goodwin et al.): 233 participants were randomised 1:1:1 to 1 mg (active control), 10 mg, or 25 mg single oral doses with psychological support. MADRS was assessed at baseline, day 2, and weeks 1, 3, 6, 9 and 12; the primary endpoint was change to week 3 with clinical response defined as ≥50% decrease in MADRS. At 3 weeks the response rates were 18% (1 mg), 19% (10 mg) and 37% (25 mg). The derived NNT for 25 mg versus 1 mg was 5.3 (95% CI = 3.1, 18.5), while the 10 mg comparison yielded an NNT of 100 (95% CI = −7.5, 8.9). At 12 weeks response rates were lower and less favourable to active dosing (1 mg 10%, 10 mg 5%, 25 mg 20%), producing NNTs that were not clearly significant. Esketamine trials: Four trials of intranasal esketamine plus an oral antidepressant were included; dosing strategies differed (fixed-dose versus flexible dosing) and primary endpoints were typically at 28 days. - Popova et al. (flexible dosing): Across 227 participants, the 28-day response rate was 10.8% for placebo plus oral antidepressant versus 16.5% for esketamine plus oral antidepressant, yielding an NNT of 30.3 (95% CI = −10.0, 29.6). At 2 days post-dose the NNT was 17.5 (95% CI = −6.6, 26.7). - Ochs‑Ross et al. (elderly, flexible dosing): Among participants aged ≥65 years, response at 28 days was 13.3% for placebo and 27% for esketamine, giving an NNT of 7 (95% CI = −3.6, 436.7); the confidence interval was very wide and included non-significance. - Fedgchin et al. (fixed-dose 56 mg and 84 mg): In a 1:1:1 randomised trial of 342 participants, 28-day response rates were 38.9% (placebo), 54.1% (56 mg) and 53.1% (84 mg). The calculated NNTs for both active doses were 7, with 95% CIs reported as 56 mg: 3.5 to 46.7 and 84 mg: 3.6 to 142.2. - Chen et al. (flexible dosing, China): In 126 participants randomised to esketamine or placebo with an oral antidepressant, 28-day response rates were 34% (placebo) and 35.5% (esketamine), producing an NNT of 67 (95% CI = −7.0, 8.9), not statistically significant. Overall, fixed-dose esketamine in Fedgchin et al. produced single-digit NNTs at 28 days, while most flexible-dosing studies did not show statistically significant NNTs or produced wide CIs. Across trials the definition of response varied (most used ≥50% MADRS reduction; Chen et al. used ≥30%). Adverse events and NNHs: For single-dose 25 mg psilocybin, nausea on day 1 yielded an NNH of 5.0 (95% CI = 3.4, 9.2). Beyond day 1 and up to week 3, commonly reported psilocybin TEAEs did not reach clinically relevant NNH thresholds and incidence rates fell below 5% after the primary endpoint. For esketamine, commonly reported TEAEs with single-digit NNHs included headache, nausea, dizziness and dissociation; however, incidence rates and NNHs varied between studies. The reported TEAEs were generally transient, did not meet criteria for serious adverse events, and did not typically result in treatment discontinuation.

Wong and colleagues interpret the findings as preliminary evidence that a single 25 mg dose of psilocybin produced a clinically meaningful reduction in MADRS score at the 3-week primary endpoint, reflected by an NNT of approximately 5. They note that this result derives from a single trial and therefore requires replication. The authors emphasise that psilocybin-associated adverse events were mostly transient; only nausea on day 1 reached a clinically relevant NNH (NNH = 5), and event rates declined thereafter. For esketamine, fixed-dose intranasal regimens (56 mg and 84 mg) yielded NNTs of about 7 at 28 days in the Fedgchin et al. trial, indicating potential clinical benefit, although confidence intervals were wide. Flexible-dosing esketamine trials yielded inconsistent results and, in several cases, NNT estimates were not statistically significant. Common esketamine TEAEs—headache, nausea, dizziness and dissociation—often produced single-digit NNHs but were described as transient and not leading to discontinuation. The authors acknowledge several limitations that constrain interpretation. Only five RCTs were available, with substantial heterogeneity across studies in dosing regimens, participant characteristics and response definitions; only one psilocybin RCT was identified, precluding pooled estimates for that agent. The exclusive use of MADRS was intentional to standardise comparisons but may omit trials using other scales. Methodological limitations of the analysis are also discussed: Wald intervals for NNT CIs can be unreliable with small samples or when NNTs exceed 10, and inverse-transformations of ARRs can produce CIs that approach infinity and complicate interpretation. Sample size limitations, participant dropout, and time-point selection are additional concerns; for example, other work suggests psilocybin may peak at 5–6 weeks post-dose, a window not sampled in the Goodwin et al. trial, so the calculated NNTs may not capture maximum effect. Implications highlighted by the authors include the need for replication, larger trials, and longer-term studies to assess maintenance therapy; they note that esketamine currently has multiyear efficacy data and regulatory approval, whereas psilocybin remains investigational with multiple Phase III trials underway. Finally, the authors remark that cost-effectiveness and implementation planning should consider not only NNT and NNH but also infrastructure and personnel requirements, and that further high-quality evidence is required before widespread clinical adoption.