Biomarkers of ketamine's antidepressant effect: An umbrella review

This umbrella review followed PRISMA and PRIOR recommendations and searched Scopus, PubMed and the first ten pages of Google Scholar from database inception to July 2022. No restrictions were applied for language, timespan, document type or publication status. The search terms combined “systematic review” or “meta‑analysis” with “ketamine” and “biomarker”. Studies were eligible if they were peer‑reviewed systematic reviews or meta‑analyses that investigated biomarkers of ketamine's antidepressant effects in people with depression. Exclusion criteria were animal or in vitro studies, non‑full papers (e.g. abstracts), unclear methodology, or reviews that included fewer than two primary studies with the required characteristics. Two independent reviewers screened records in two stages (title/abstract then full text). From each included review the authors extracted standard bibliographic and methodological information (first author, year), the number of included primary studies and patients, the biomarkers assessed and measurement methods, follow‑up timing, ketamine dose and route, quality assessment approaches used by the review, and reported meta‑analytic results where applicable. To appraise review quality the team used AMSTAR‑2, a 16‑item tool for assessing the methodological rigour of systematic reviews; based on AMSTAR‑2 major and minor domains each review was rated from high to critically low quality.

The search identified 894 records; after deduplication and screening 5 systematic reviews/meta‑analyses met inclusion criteria. Across those reviews the authors reported a total of 108 primary studies including 4,912 participants. Most primary studies (76/108) employed a single ketamine infusion. Ketamine dosing in the included literature ranged from 0.2 to 0.5 mg/kg given intravenously, with many studies using 0.5 mg/kg. Most primary study samples were treatment‑resistant by common criteria (at least two failed antidepressants). Quality assessment of the five included reviews using AMSTAR‑2 produced scores ranging as low as 3/16, with an average score of 8.2; the authors therefore characterised the overall review quality as moderate to low. Inflammatory biomarkers: Three reviews summarised blood‑based inflammatory markers. Kopra and colleagues reviewed six studies (n = 277) and reported that five of six studies observed decreases in at least one inflammatory marker after ketamine. Frequently measured markers included IL‑6 (n = 6), TNF‑α (n = 5) and IFN‑γ (n = 3). Significant reductions were reported for IL‑1β, TNF‑α, IL‑6 and several other cytokines in subsets of studies; most changes were short‑lived (up to 240 minutes), although some studies observed reductions lasting 24–72 hours. Medeiros and colleagues conducted a meta‑analysis of six studies (N = 260) examining pro‑inflammatory markers; baseline levels of pro‑inflammatory factors were non‑significantly lower in responders compared with non‑responders and CRP showed the strongest (but non‑significant) baseline association with response (SMD = -0.28, 95% CI -0.67 to 0.10, p = 0.15). The same meta‑analysis found no statistically significant longitudinal changes in inflammatory markers in responders versus non‑responders following infusion. Tryptophan–kynurenine pathway: Ten studies of TRP–KYN metabolites were synthesised (most commonly TRP, KYN, kynurenic acid and quinolinic acid). The reviews found largely inconsistent evidence: only one of seven studies showed a significant association between baseline pathway measures and antidepressant response (lower baseline KynA/QA ratio associated with greater improvement, p = 0.016). Some studies reported post‑infusion increases in KYN and KynA and decreases in IDO at 24–72 hours, but other investigations found no changes over time. Neuroimaging biomarkers: Two reviews of neuroimaging studies reported that patients with MDD had reduced global brain connectivity at baseline. Post‑ketamine, increases in connectivity were reported in regions including lateral and ventromedial prefrontal cortex (PFC), dorsal caudate, hippocampus and networks such as the central executive network; reduced connectivity was observed in left cerebellum, left amygdala and substantia nigra. Certain connectivity changes correlated with symptom domains (e.g. amygdala–SN decreases predicted anxiety improvements; dorsal caudate–vmPFC and hippocampal connectivity increases predicted reductions in anhedonia). Antidepressant effects commonly emerged within 2 hours, peaked around 24 hours and could persist up to one week after a single infusion. Neurotrophic markers (BDNF/VEGF): Rossi and others summarised studies of BDNF and FGF‑23; BDNF was measured at multiple early and later post‑infusion timepoints. Three of seven studies reported significant BDNF increases after ketamine; one study reported higher BDNF at 240 minutes in responders versus non‑responders and only one trial found BDNF changes that correlated with symptom reduction. Medeiros et al. synthesised 15 BDNF studies and 5 VEGF studies and performed a meta‑analysis on ten BDNF studies (N = 332) and three VEGF studies (N = 154). The pooled analyses found no statistically significant association between baseline BDNF or VEGF and response. The authors also reported an observed increase in BDNF among responders that they described as not significant (reported SMD 0.19, 95% CI 0.04 to 0.66 in a subset of three studies). Ketamine and metabolite blood levels: Thirteen studies examined whether plasma levels of ketamine, norketamine, DHNK or hydroxynorketamines (HNKs) related to antidepressant response. Ten studies found no significant associations. Two studies reported higher ketamine concentrations linked to greater symptom improvement, while two studies observed that lower levels of (2R,6R)‑HNK were associated with better outcomes. Amino acids and derivatives: Five studies evaluating baseline or longitudinal amino acid measures (e.g. glutamate, GABA precursors and others) produced no reproducible, statistically significant associations with ketamine response. Additional findings: Several investigations linked sleep‑related changes (increased slow‑wave sleep/slow‑wave activity) and EEG/MEG measures of cortical plasticity with ketamine's antidepressant effects. Structural and functional imaging markers were variably associated with outcomes, for example lower baseline left hippocampal volume and greater right caudate connectivity were reported as predictors in specific studies, but findings were not uniform across the literature.

Meshkat and colleagues interpret the assembled evidence as indicating multiple promising but not yet clinically validated biomarker signals across immune, neurotrophic, neuroimaging and electrophysiological domains. The reviews collectively suggest that ketamine can exert an anti‑inflammatory effect in at least a subset of patients, with reductions reported for IL‑6, TNF‑α, CRP and other cytokines shortly after infusion in several studies. The authors note, however, that the temporal brevity of many inflammatory changes (often not persisting beyond 24 hours) and the lack of consistent longitudinal associations with clinical response raise uncertainty about whether cytokine modulation mediates ketamine's antidepressant effect. Neuroimaging evidence is highlighted as pointing to the anterior cingulate cortex—particularly the subgenual ACC—and prefrontal networks as key loci of ketamine's action. Studies reporting normalisation or increases in PFC global connectivity in responders provide a mechanistic link to glutamatergic circuitry and synaptic plasticity. Nonetheless, the authors emphasise inconsistent imaging findings (for example divergent sgACC metabolic results across PET studies), the influence of scan timing relative to acute dissociative effects, and the need for larger, better‑powered imaging trials. Regarding neurotrophic markers, BDNF findings were mixed: some studies showed post‑infusion increases and occasional correlations with symptom change, but pooled analyses did not support a reliable baseline or longitudinal predictor. The discussion expands to other candidate markers such as sleep slow‑wave activity, EEG/MEG indices of plasticity, hippocampal volume and caudate connectivity, which have shown associations with clinical outcomes in isolated studies. Limitations acknowledged by the authors include moderate to low methodological quality of the included reviews (per AMSTAR‑2), small sample sizes in many primary studies, heterogeneity in biomarkers assessed and study designs, and a predominance of correlational analyses rather than prespecified predictive biomarker testing. The authors caution that inflammatory markers or other correlates may modify but not mediate response, implying that markers linked to depression biology are not necessarily the same as predictors of ketamine benefit. They conclude that larger, higher‑quality, real‑world and mechanistically informed studies—potentially integrating multimodal and multiomic datasets and machine‑learning approaches—are needed to replicate candidate biomarkers and move towards clinically useful biosignatures.

Multiple translational ("transomic") biomarkers have been evaluated in relation to ketamine treatment outcomes, but no single marker or biosignature has sufficient validation for routine clinical use. Current evidence suggests promising signals in neuroimaging, neurophysiological and possibly cognitive domains, and there are reproducible short‑term anti‑inflammatory effects in at least some patients. The authors stress that most studies to date have reported associations rather than prespecified predictive markers, and that future work should use larger samples, a priori biomarker hypotheses, replication cohorts and integrative analytic methods (including machine learning) to determine clinically actionable biomarkers.