Effects of psychoplastogens on blood levels of brain-derived neurotrophic factor (BDNF) in humans: a systematic review and meta-analysis

Preclinical work established that several fast-acting compounds — including ketamine and classic psychedelics — promote structural and functional neuroplasticity in cortical circuits and have rapid antidepressant or therapeutic effects in animals. This group of molecules, often termed psychoplastogens, has been associated in animal models with dendritogenesis, synaptogenesis and increased neurotrophic signalling, notably involving brain-derived neurotrophic factor (BDNF). In humans, peripheral measures of BDNF (serum or plasma) are frequently used as an accessible proxy for brain neuroplasticity, but peripheral BDNF is biologically noisy and prior human studies of psychoplastogens have produced mixed results. Calder and colleagues set out to test whether the rapid upregulation of neuroplasticity reported in preclinical studies is detectable in humans using peripheral BDNF. They conducted a preregistered systematic review and meta-analysis across a broad set of psychoplastogens (ketamine, esketamine, LSD, psilocybin, ayahuasca/DMT, MDMA, scopolamine, rapastinel and related compounds), examined multiple timepoints, and planned moderator analyses to investigate factors such as blood fraction (plasma versus serum), timing of measurement, dose, assay specificity, study design and risk of bias.

The review was preregistered on PROSPERO and followed PRISMA guidance. The authors searched PubMed, Web of Science and PsycINFO (initial search 25 May 2022, update 13 June 2023) for clinical studies in adults reporting peripheral BDNF after administration of a psychoplastogen. Inclusion criteria required peer-reviewed English-language studies since 1982 that measured BDNF and included a comparison condition (placebo and/or baseline). Studies that systematically combined a psychoplastogen with another pharmacological or neurophysiological intervention in every treatment arm were excluded, while trials that incorporated psychotherapy or permitted continuing psychiatric medication were eligible. Screening proceeded in two phases: identification of human psychoplastogen trials by title/abstract, then manual searching of full texts for BDNF-related terms. Two reviewers performed selection and resolved discrepancies by discussion. Risk of bias was assessed in duplicate using Cochrane tools for randomized parallel and crossover designs and ROBINS-I for non-randomized trials. Data extraction captured sample sizes, means and standard deviations for BDNF values and all available effect sizes (different drugs, doses and timepoints). When numerical data were absent, the authors contacted study teams or extracted values from plots using WebPlotDigitizer. Additional variables extracted included participant age, sex, diagnoses, blood fraction (serum or plasma), study design, percentage of missing data, timing of BDNF sampling, single versus repeated dosing, and immunoassay specificity for mature BDNF (mBDNF) versus pro-BDNF or total BDNF. Statistical analysis used Hedges' g as the standardized mean difference (SMD) in mixed effects meta-analytic models with random effects for study and effect size, and cluster robust variance estimation for confidence intervals and tests. The primary meta-analysis pooled all psychoplastogens, with separate analyses for drugs represented by at least three independent studies (ketamine, LSD, psilocybin) and a pooled analysis of classic psychedelics. Subgroup analyses contrasted plasma versus serum and healthy versus depressed samples. Meta-regressions evaluated moderators: timing of measurement (hours), average age, control type (baseline only, placebo only, both), study design (parallel, crossover, nonrandomized), single versus repeated dosing, immunoassay specificity (mBDNF/pro-BDNF/total), overall risk of bias, percentage of missing data, and dose where applicable. Publication bias was assessed using Egger's test and funnel plots; analyses were performed in R with the metafor and meta packages. The search yielded 13,094 records after duplicate removal, of which 1,170 were clinical psychoplastogen studies; 29 studies met inclusion criteria and provided 104 effect sizes for meta-analysis. Most included studies administered ketamine (17 studies); others included LSD (7), psilocybin (3), ayahuasca (2), MDMA (2), and single studies of esketamine, DMT, mescaline, rapastinel and scopolamine. Fourteen studies involved healthy participants and 16 involved patients (12 depression, 3 bipolar depression, 2 anxiety disorders, 1 OCD). Most BDNF measurements occurred within 48 hours post-treatment (25 studies); six ketamine studies measured later (7–26 days). Information on assay specificity was available for 23 studies: 13 measured total BDNF and 10 (from the same group) used an assay specific to mBDNF. Risk-of-bias assessment found that of 12 randomized crossover trials, one was high risk and the remainder low risk. Among nine randomized parallel trials, four were low risk, three raised some concerns, and two were high risk. All eight non-randomized trials were rated high/serious risk, largely due to baseline confounding and selective reporting concerns.

Pooling all psychoplastogens across 29 studies and 104 effect sizes produced no significant overall effect on peripheral BDNF (SMD = 0.024, p = 0.64). Heterogeneity was relatively low (I2 = 22.64%) and there was no evidence of small-study/publication bias by Egger's test (p = 0.15). Subgroup analyses showed no significant change in BDNF when analysing plasma or serum separately, nor when analysing healthy participants or patients with depression alone. Re-analyses that matched short-term plasma measures (<7 days) and long-term serum measures (≥7 days) did not alter the null result. In meta-regressions, timing of measurement had a statistically significant but very small association with effect size: later sampling times were associated with minimally larger effects (SMD increase per hour = 0.0005, p = 0.038). Other moderators such as average participant age, study design, immunoassay sensitivity (mBDNF/pro-BDNF/total) and overall risk-of-bias rating did not show significant associations in the all-drug analysis. Several methodological factors were associated with larger reported effects. Studies using both placebo control and change-from-baseline reported smaller effect sizes than studies using baseline only (difference SMD = -0.31, p = 0.031). A higher percentage of missing BDNF values correlated with larger effect sizes (SMD = 0.77, p = 0.029). Studies with a high overall risk-of-bias rating reported greater effect sizes than low-risk studies (difference SMD = 0.26, p = 0.017). Analyses limited to ketamine (17 studies) similarly found no significant effect (SMD = 0.03, p = 0.65). For ketamine, randomized parallel trials reported significantly smaller effect sizes than non-randomized open-label trials (difference SMD = -0.26, p = 0.03), and trials with high risk-of-bias ratings reported larger effects than low-risk trials (difference SMD = 0.32, p = 0.04). Later sampling times again associated with minimally larger effects (SMD per hour = 0.0007, p = 0.02). No significant effects were observed for plasma versus serum or for healthy versus depressed subgroups within ketamine studies. When classic psychedelics were pooled, there was no significant effect on peripheral BDNF (SMD = 0.048, p = 0.62). Individual analyses of LSD (SMD = 0.01, p = 0.93) and psilocybin (SMD = 0.007, p = 0.97) also showed no significant changes. For LSD there was a non-significant trend that studies with higher risk of bias reported larger effects (difference SMD = 0.34, p = 0.06). Overall, the null findings were robust across drug classes, blood fraction, diagnostic groups and most moderators, while indicators of lower methodological rigour (missing data, nonrandomized designs, high risk-of-bias) were associated with larger reported BDNF effects.

Calder and colleagues interpret their findings as evidence that peripheral BDNF does not show a rapid increase after administration of psychoplastogens in humans. They highlight that this is the first meta-analysis to include the full range of psychoplastogens and timepoints, to make particular efforts to capture null results, and to model all reported effect sizes while testing multiple moderators. Two principal explanations are offered for the discrepancy between strong preclinical evidence of rapid brain BDNF upregulation and the null peripheral BDNF signal in humans. First, peripheral BDNF may be an unreliable or invalid marker of short-term changes in brain BDNF: commercial assays show high inter-assay variability, peripheral levels are influenced by many pre-analytical and behavioural factors (sample handling, time of day, exercise, smoking, hormones), and most circulating BDNF is stored in platelets and may largely originate from megakaryocytes rather than the brain. The authors note that plasma might, in theory, be more sensitive to rapid changes than serum, but existing data did not support such a difference. Distinguishing mBDNF from pro-BDNF is proposed as a potential refinement, since these forms may relate differently to plasticity and psychiatric symptoms. Second, translational differences between animal models and human subjects may explain the gap. Psychoplastogen effects observed in animals are commonly tested in standard laboratory housing, an unstimulating environment that can lower baseline neuroplasticity. Humans typically live in relatively enriched settings, which could produce ceiling effects that make further drug-induced plasticity harder to detect. The authors also acknowledge species-specific neurophysiological differences may play a role. Methodological findings reinforce caution: more rigorous designs tended to produce smaller effects. Studies with higher percentages of missing data, nonrandomized open-label designs, baseline-only controls and higher risk-of-bias ratings reported larger BDNF changes. These patterns suggest that biased study designs or selective reporting may inflate apparent effects. Given the limitations of peripheral BDNF, Calder and colleagues recommend alternative human methods to probe drug-induced neuroplasticity. These include structural and molecular neuroimaging (MRI, PET targeting SV2A), functional connectivity measures (fMRI, MEG, EEG), and stimulation-based electrophysiological paradigms such as paired associative stimulation or sensory-induced long-term potentiation, which index the brain’s capacity for synaptic plasticity more directly. They cite some evidence that neuroimaging and electrophysiological metrics have shown region-specific or baseline-dependent changes after psychoplastogen treatment in small studies. The authors acknowledge limitations of their meta-analysis: incomplete reporting of sample handling and assay procedures in primary studies, limited variability in some moderators (notably dose for ketamine and age in LSD/psilocybin samples), predominance of short-term (<48 h) measurements, few studies with multiple infusions, and inability to analyse treatment responders separately in most patient studies. They also note that grouping diverse psychoplastogens assumes shared downstream mechanisms, and that only ketamine, LSD and psilocybin had sufficient studies for separate meta-analyses, with small numbers for the latter two. In sum, the investigators conclude that peripheral BDNF offers little support for the hypothesis that psychoplastogens rapidly increase neuroplasticity in humans, and they recommend more specific, brain-focused measures for future translational research.

The authors conclude that the rapid neuroplasticity increases seen in animal models after psychoplastogen administration are not reflected in peripheral BDNF measurements in humans. Psychoplastogens as a group, classic psychedelics, and individual drugs including ketamine, LSD and psilocybin showed no significant effect on peripheral BDNF. Calder and colleagues therefore argue that peripheral BDNF is an extremely limited biomarker for rapid, drug-induced neuroplasticity in humans and suggest that neuroimaging and stimulation-based electrophysiological methods will be more informative for future translational studies.