Neurocognitive Effects of Ketamine and Association with Antidepressant Response in Individuals with Treatment-Resistant Depression: A Randomized Controlled Trial

Ketamine is a high-affinity, noncompetitive NMDA glutamate receptor antagonist that has demonstrated rapid antidepressant effects in patients with treatment-resistant depression (TRD). Although single-dose ketamine (commonly 0.5 mg/kg over a slow 40-minute infusion) produces acute dissociative and cognitive effects that typically resolve within hours, there is limited data on its delayed or persistent neurocognitive impact in clinical mood-disorder populations. Prior work by the same group and others reported acute memory impairments following ketamine in healthy volunteers and in an open-label TRD sample, and an earlier observation that slower baseline processing speed was associated with greater antidepressant response at 24 hours. Murrough and colleagues designed the present two-site, double-blind, randomised controlled trial to extend those findings. The study aimed to (1) characterise the effect of a single sub-anesthetic ketamine infusion on neurocognitive performance measured 7 days post-treatment, and (2) examine whether baseline neurocognitive functioning predicts rapid antidepressant response to ketamine. The primary hypothesis was that ketamine would not worsen neurocognitive functioning at 7 days and that slower baseline processing speed would predict greater symptom improvement at 24 hours.

Participants were adults aged 21–80 with a primary diagnosis of major depressive disorder and inadequate response to at least three antidepressant trials; they had recurrent or chronic major depressive episodes and required a score of 32 or greater on the Inventory of Depressive Symptomatology–Clinician Rated at screening and 24 hours before randomisation. Exclusion criteria included lifetime psychotic or bipolar disorder, recent substance abuse, unstable medical illness, prominent suicidal or homicidal risk, cognitive impairment (Mini-Mental State Examination score <27), and contraindicated medications. Antidepressant medications were washed out before enrollment (≥1 week, or 4 weeks for fluoxetine). Medical and toxicology screening were performed prior to treatment. The trial was conducted with participants free of concomitant antidepressants and other psychotropic medications (except a stable non-benzodiazepine hypnotic). Under double-blind conditions, eligible participants received a single intravenous infusion of ketamine 0.5 mg/kg or midazolam 0.045 mg/kg administered over 40 minutes in an inpatient research setting. Randomisation used a 2:1 allocation favouring ketamine; midazolam served as an active control to approximate nonspecific sedative effects. Symptom ratings were obtained during and up to 240 minutes after infusion, and participants had outpatient evaluations at 48, 72 hours and 7 days post-infusion. The primary clinical outcome was Montgomery–Åsberg Depression Rating Scale (MADRS) score at 24 hours; secondary outcomes included MADRS at 48 and 72 hours and 7 days, and responder proportions at those time points. Categorical response was defined in the protocol as a ≥50% reduction in MADRS score relative to baseline. Neurocognitive testing was performed within one week before infusion and repeated once at 7 days using alternate test forms; neuropsychology raters were masked to treatment allocation and same-day side effects. Neurocognition was assessed using a subset of the MATRICS Consensus Cognitive Battery (MCCB): Trails A, WMS Spatial Span, BACS Digit Symbol, Letter–Number Sequencing, Hopkins Verbal Learning Test, Brief Visual Memory Test, NAB Mazes, and Category Fluency. T-scores (mean 50, SD 10), corrected for age and sex by the MCCB program, were averaged to yield domain scores for processing speed, working memory, verbal learning, visual learning, and reasoning/problem solving. Statistical analysis included summary statistics and univariate comparisons of treatment groups at baseline. Repeated-measures ANOVA models examined effects of time (baseline vs 7 days), treatment condition, and responder status on cognitive domains, with change in depression severity from baseline to 7 days included as a covariate. To identify baseline cognitive predictors of antidepressant response (MADRS change baseline to 24 h), the investigators used backwards stepwise linear regression with age, baseline MADRS, and the five neurocognitive domain scores; logistic regression was used for categorical response outcomes. Separate models evaluated responder status at 24 hours and at 7 days.

Seventy-three individuals were randomised and 72 received study medication (47 ketamine, 25 midazolam) in the parent trial. Due to missing or incomplete neurocognitive data (four ketamine, six midazolam), the neurocognitive sub-study included 62 patients: 43 randomised to ketamine and 19 to midazolam. One domain (visual learning) had additional missing data (n = 34 ketamine, n = 17 midazolam). Treatment groups were similar on age, sex distribution, baseline depression severity and degree of treatment resistance. At the 7-day assessment, repeated-measures ANOVA revealed a significant main effect of time across both treatment arms: participants improved from baseline on processing speed (F(1,59) = 6.58, p = 0.013), verbal learning (F(1,59) = 6.80, p = 0.012), and visual learning (F(1,48) = 6.48, p = 0.014) while controlling for change in depression severity. No significant changes were observed for working memory or reasoning/problem solving. There was no main effect of treatment condition (ketamine versus midazolam) on any neurocognitive domain, no effect of antidepressant responder status at either 24 hours or 7 days on cognitive performance, and no significant time-by-treatment or time-by-responder interaction effects. Baseline cognitive performance predicted antidepressant response to ketamine at 24 hours. In a backwards stepwise linear regression including age, baseline MADRS, and the five domain scores, the final model retained only processing speed (F(1,37) = 5.3, p = 0.027; beta = 0.43 ± 0.19; t = 2.3, p = 0.027), indicating that poorer processing speed at baseline was associated with greater improvement in MADRS score at 24 hours. A backwards stepwise logistic regression for categorical response yielded a model containing processing speed and visual learning (χ2 = 11.82, p = 0.003), but only processing speed was a significant predictor in that model (Exp(beta) = 0.823, Wald = 7.76, p = 0.005). On univariate comparisons, ketamine responders had lower baseline processing speed T-scores (43.37 ± 8.78) than ketamine non-responders (49.24 ± 10.1; F(1,45) = 4.36, p = 0.043). There was no association between baseline processing speed and baseline depression severity (r = −0.167, p = 0.164). No significant relationships between baseline cognitive measures and antidepressant response were observed for midazolam (data not shown).

Seven days after a single sub-anesthetic ketamine infusion in patients with TRD, the investigators found no evidence of deleterious effects on neurocognitive performance compared with an active midazolam control. Improvements observed over time in processing speed, verbal learning, and visual learning occurred across both treatment conditions and persisted after controlling for changes in depression severity, which the authors interpret as likely reflecting non-specific practice or learning effects from repeated testing. The study therefore replicates prior open-label findings suggesting absence of persistent cognitive impairment following single-dose ketamine in depressed patients. Murrough and colleagues also replicated their earlier observation that slower processing speed at baseline predicted a more robust rapid antidepressant response to ketamine at 24 hours. The authors note that this association was independent of baseline depression severity and was not observed for midazolam. They situate this finding within a mechanistic framework: processing speed and attention have been linked to dopaminergic functioning in prefrontal–subcortical circuits, and preclinical and some human data indicate that ketamine modulates dopamine signalling and rapidly enhances synaptic plasticity (BDNF-dependent) in cortical and subcortical regions. Nevertheless, the authors acknowledge that no study to date has directly established a causal relationship between dopamine signalling and ketamine’s antidepressant effects in humans, and they call for in vivo imaging and other mechanistic studies to clarify these pathways. Key limitations noted by the investigators include the single-dose design and the absence of immediate post-infusion cognitive testing in this trial (so acute cognitive effects at peak drug exposure were not assessed here), limited ability to generalise to repeated or long-term treatment regimens, and a sample size that—while comparatively large for this literature—may be insufficient to identify robust biological or cognitive biomarkers. They also highlight differences in neurocognitive assessment batteries across studies (this study used the MCCB and did not include a dedicated attention measure), which may complicate replication. The authors recommend that future trials employ standardised, multi-modal biomarker assessments (potentially combining neurocognitive testing, neuroimaging, qEEG and peripheral markers), explore dose and frequency parameters, and evaluate the cognitive effects of repeated ketamine administrations over longer time frames. In sum, the trial provides evidence that a single 0.5 mg/kg ketamine infusion is not associated with short-term adverse neurocognitive effects at 7 days in patients with TRD, and that reduced processing speed at baseline may be a candidate pre-treatment predictor of rapid antidepressant response; further, larger and more mechanistically oriented studies are required to confirm and extend these findings.