Neurocognitive effects of subanesthetic serial ketamine infusions in treatment resistant depression

Major depressive disorder affects a large global population and around 30% of patients do not remit after two or more adequate antidepressant trials, meeting criteria for treatment-resistant depression (TRD). Subanesthetic ketamine produces rapid antidepressant effects in many patients, but those effects are typically short-lived and the cognitive consequences of repeated, clinically administered ketamine remain incompletely characterised. Prior clinical work has reported mixed findings: some studies identify transient cognitive deficits immediately after ketamine, whereas others report improvements in specific domains (for example, processing speed, visual memory) within days to weeks. It is also unresolved whether any cognitive changes following ketamine are secondary to mood improvement or reflect effects on partly distinct brain systems. Zavaliangos-Petropulu and colleagues conducted an exploratory, naturalistic investigation to clarify how four subanesthetic ketamine infusions affect multiple neurocognitive domains in people with TRD, and how cognitive change relates to antidepressant response. Using the NIHToolbox Cognition Battery, assessments were made at baseline, 24 hours after the first and fourth infusion, and at 5 weeks post-treatment; the study tested whether serial ketamine would produce no cognitive harm, would improve fluid cognitive abilities (for example, working memory) with durable effects at 5 weeks, and whether cognitive gains would be mediated by improvement in depressive symptoms. Healthy controls were included in supplementary analyses to help interpret practice effects and longitudinal change.

The study enrolled 66 adults meeting DSM-5 criteria for a major depressive episode who had failed at least two adequate antidepressant trials and had been depressed for at least six months; three participants had bipolar disorder while the remainder had unipolar major depressive disorder. Inclusion required age 20–64, English as a primary language, a 17-item Hamilton Depression Rating Scale (HDRS) score ≥ 17, and stable antidepressant or mood stabiliser therapy for at least six weeks. Exclusion criteria included dementia, major neurological or medical illness, active suicidality, recent substance abuse, psychotic disorders, recent neuromodulation or ketamine exposure, pregnancy, and other ketamine contraindications assessed by laboratory testing and ECG. A complementary sample of 39 healthy controls (17 of whom were retested after ~14 days) underwent the same assessments but received no ketamine. Open-label ketamine (0.5 mg/kg IV diluted in 60 ml saline over 40 minutes) was administered two to three times per week for a total of four infusions across approximately 14 days. Neurocognitive testing and clinical assessments (17-item HDRS) were performed at baseline (TP1), 24 hours after the first infusion (TP2), 24 hours after the fourth infusion (TP3), and at 5-week follow-up or earlier at relapse (TP4). Attrition resulted in 66 participants at TP1, 44 at TP2, 58 at TP3, and 31 at TP4. Patients were allowed to continue stable psychotropic medication regimens; benzodiazepines were stopped >24 hours before visits. Cognition was measured with seven NIHToolbox tests covering language (picture vocabulary, oral reading), attention/inhibition (Flanker), working memory (List Sorting), executive function (Dimensional Change Card Sort), processing speed (Pattern Comparison), and episodic memory (Picture Sequence). Composite crystallised scores averaged language measures, while composite fluid scores averaged flanker, list sorting, card sort, processing speed, and picture sequence. All scores were age- and sex-corrected Z-scores per NIHToolbox norms. Analyses were performed in R using general linear mixed models (nlme). Model 1 tested the main effect of time across TP1–TP3 for each cognitive measure and HDRS (Assessment ~ time + random(participant)). Bonferroni correction for seven cognitive tests set significance at p<0.005. Post-hoc pairwise comparisons included TP4. Pearson correlations assessed relationships between percent change scores in cognition and HDRS (100*((TP3−TP1)/TP1)), and whether baseline cognition predicted clinical outcome. A mediation-style test (Model 2: Cognitive Assessment ~ time + ΔHDRS + random(participant)) examined whether cognitive change over time remained significant after accounting for percent change in HDRS; loss of significance would imply mediation by antidepressant response. Supplemental analyses compared baseline cognition between patients and healthy controls and assessed practice effects in the retested controls; primary analyses were repeated excluding the three bipolar participants to test robustness.

Sample completion varied by timepoint: 66 at baseline, 44 at 24 hours after the first infusion, 58 at 24 hours after the fourth infusion, and 31 at the 5-week follow-up. Mixed-effect models across serial ketamine treatment found significant main effects of time for several cognitive measures: List Sorting (working memory) F=12.10, p<0.001; Pattern Comparison (processing speed) F=56.85, p<0.001; Picture Sequence (episodic memory) F=22.82, p<0.001; and the fluid composite F=80.98, p<0.001. Moderate improvements in picture vocabulary and flanker inhibition were observed (F=4.94, p=0.03 and F=7.75, p=0.01 respectively) but these did not survive the Bonferroni-corrected threshold. Pairwise comparisons showed improvements from TP1 to TP2 in list sorting (p=0.006), flanker inhibition (p=0.02), processing speed (p=0.005), picture vocabulary (p=0.007) and fluid composite (p<0.001). From TP2 to TP3, processing speed (p<0.001) and fluid composite (p=0.002) further improved. Comparisons from TP1 to TP3 indicated gains in list sorting (p=0.003), processing speed (p<0.001), picture sequence (p<0.001), and fluid composite (p<0.001). Durability analyses including TP4 showed TP1→TP4 improvements for picture vocabulary (p=0.008), flanker inhibition (p=0.001), processing speed (p<0.001), and fluid composite (p<0.001); there were no significant changes between TP3 and TP4. Depression severity (HDRS) improved across time (F=190.3, p<0.001) but post-hoc tests indicated HDRS scores tended to revert toward baseline at follow-up: TP3 to TP4 showed reversion (p=0.004) and TP2 to TP4 was non-significant (p=0.54). Results were similar when excluding the three bipolar cases. Correlational analyses found a modest relationship between change in flanker inhibition and change in HDRS after serial infusion (R=−0.20, p=0.045), indicating greater improvement in inhibition associated with greater symptom reduction. Baseline picture sequence performance (episodic memory) correlated moderately with HDRS improvement (R=−0.27, p=0.037), suggesting better baseline episodic memory predicted larger clinical benefit. Mediation-style mixed models that included percent HDRS change as a covariate showed that list sorting, processing speed, picture sequence and the fluid composite remained significantly associated with time (e.g. processing speed F≈56.57, p<0.001), indicating these cognitive improvements were statistically independent of antidepressant response. Supplemental analyses identified practice effects in healthy controls for processing speed and flanker inhibition, but the magnitude of cognitive change in patients exceeded that in controls, arguing against practice effects fully explaining the patient improvements.

Zavaliangos-Petropulu and colleagues interpret their findings as evidence that serial subanesthetic ketamine infusions are associated with improvements in multiple domains of neurocognitive function in TRD, notably processing speed, episodic memory, working memory and overall fluid cognition. The investigators emphasise that some domains (processing speed and the fluid composite) continued to improve between the first and fourth infusion and that several gains persisted at least 5 weeks post-treatment, while antidepressant effects showed partial reversion by that follow-up. Although a modest correlation emerged between improvements in inhibitory control and reduction in depressive symptoms, mediation analyses indicated that gains in fluid cognition, working memory, processing speed and episodic memory were independent of change in HDRS. The authors therefore suggest that ketamine may engage overlapping but at least partially distinct neural systems for mood and cognition. This interpretation is aligned with some prior clinical reports showing procognitive effects and with preclinical evidence that subanesthetic ketamine induces synaptogenesis and neuroplastic changes that could affect cognitive circuitry. The study team situates these results within the existing literature, noting concordance with other single- and serial-infusion studies that observed improvements in processing speed and working memory, and with large esketamine studies showing stable or improved cognitive performance. They also acknowledge discrepant reports where cognitive gains appeared mediated by mood change, and propose that differences in measures (cold versus hot cognition) and sample characteristics may account for variability. The authors recommend future work linking cognitive outcomes to functional imaging to determine whether distinct brain networks underlie procognitive versus antidepressant effects, and suggest examining "hot" cognitive processes (emotionally valenced tasks) that may map more closely to mood response. Key limitations acknowledged include the naturalistic, open-label design without placebo control, potential practice effects (observed in some tasks), the relatively small number of infusions (four) and limited follow-up, and uncertainty about long-term cognitive safety with repeated exposure. The investigators note that while practice effects were present in controls, patient improvements exceeded those changes, but caution that cumulative or longer-term effects—positive or adverse—remain to be established. They call for future studies using functional imaging, more diverse cognitive batteries (including hot cognition), and longer-term follow-up to clarify mechanisms and durability.

The study reports that four subanesthetic ketamine infusions in patients with TRD were associated with cognitive safety and procognitive effects across several domains, with some improvements sustained at least 5 weeks after treatment. Improvements in inhibitory control were moderately related to antidepressant response, whereas gains in processing speed, episodic memory, working memory and composite fluid cognition appeared independent of mood improvement, suggesting ketamine perturbs partly distinct functional brain systems underlying cognitive and antidepressant outcomes.