Neurocognitive effects of six ketamine infusions and the association with antidepressant response in patients with unipolar and bipolar depression

Ketamine, an N-methyl-D-aspartate (NMDA) receptor antagonist long used as an anaesthetic, has in recent decades been shown to produce rapid antidepressant effects in treatment-resistant depression (TRD) and to reduce suicidal thinking. A single subanesthetic intravenous infusion (0.5 mg/kg) produces robust but typically short-lived benefit, prompting increasing use of repeated infusions to prolong response. However, ketamine has also been associated with cognitive impairment in healthy volunteers and in chronic, heavy recreational users, raising concern about possible neurocognitive harms from repeated therapeutic infusions. Prior small studies in depressed samples produced inconsistent findings: some reported transient memory impairments immediately after single infusions, whereas others found no decline or even improvements after serial infusions, but sample sizes were limited. Zhou and colleagues set out to clarify two linked questions in a larger clinical sample: first, whether six serial subanesthetic ketamine infusions (0.5 mg/kg) over 12 days produce short-term neurocognitive changes during a 26-day observation period in patients with unipolar or bipolar depression; and second, whether baseline neurocognitive performance predicts antidepressant response to this six-infusion regimen. The authors hypothesised that six infusions would not produce lasting neurocognitive harm and that lower baseline cognition would predict a greater antidepressant response.

This was a single-arm, open-label, 26-day study (including follow-up) conducted at the Affiliated Brain Hospital of Guangzhou Medical University. The trial enrolled adults aged 18–65 diagnosed by structured interview with a major depressive episode as part of major depressive disorder (MDD) or bipolar disorder (BD), required treatment resistance (≥2 failed antidepressant trials) or significant suicidality, and a 17-item Hamilton Depression Rating Scale (HAMD-17) score ≥17. Key exclusions included current or past substance use disorder, positive urine toxicology, unstable medical/neurological illness, moderate/severe traumatic brain injury, psychotic disorders, pregnancy or active imminent suicide risk. The study was approved by the local ethics committee and registered in the Chinese Clinical Trials Registry. - Inclusion criteria: aged 18–65; diagnosis of MDD or BD without psychotic features; treatment resistance or suicidality; HAMD-17 ≥17; able to complete scales and provide informed consent. - Exclusion criteria: current/prior substance abuse/dependence; positive urine toxicology; unstable medical/neurological illness; moderate/severe traumatic brain injury; psychotic disorders; pregnancy/lactation; imminent suicide risk. All participants maintained stable concurrent psychiatric medications for at least four weeks prior and during the study (except heavy sedatives); electroconvulsive therapy and other physical interventions were disallowed. The active regimen consisted of six intravenous ketamine hydrochloride infusions at 0.5 mg/kg diluted in 40 mL saline, given over 40 minutes on a Monday–Wednesday–Friday schedule across 12 days. Vital signs, oxygen saturation and mental status were monitored at frequent intervals during and after each infusion. Participants remained under observation for at least 30 minutes post-infusion and could withdraw for adverse effects or lack of willingness to continue. Clinical outcomes were measured one day before the first infusion (baseline, day 0), one day after the last infusion (day 13) and two weeks later (day 26). Depression severity was assessed with HAMD-17. Four cognitive domains from the MATRICS Consensus Cognitive Battery (MCCB) were administered at each time point and converted to T-scores (mean 50, SD 10): speed of processing, working memory, visual learning and verbal learning. Inter-rater reliability for clinical ratings was reported as excellent-to-good (intra-class correlations >0.90). Analyses included only participants who completed all six infusions and had baseline and day 13 assessments. Demographic comparisons used chi-squared tests and t-tests. Change over time in HAMD and MCCB domains was examined with linear mixed models. Mediation analyses tested whether changes in depressive symptoms (HAMD) mediated neurocognitive changes, using mixed models with the Sobel test to assess the significance of indirect effects (the Sobel test evaluates whether a mediator carries the effect of an independent variable to a dependent variable). Finally, logistic regression examined whether baseline cognition predicted response (defined as ≥50% reduction in HAMD at day 13); the model adjusted for baseline HAMD and then entered patient characteristics and baseline MCCB domain scores. Statistical significance was set at p<0.05 and analyses were performed in SPSS v22.

Of 185 patients screened, 100 met inclusion criteria and were enrolled. Three withdrew before receiving any infusion and 13 discontinued during the series (five after one infusion, five after two, three after four), leaving 84 completers whose data formed the primary analysis set; of these, 10 (11.9%) did not complete the day 26 follow-up. The analysed sample comprised 68 patients with MDD and 16 with BD, 47.6% male, with a mean age of 34.8±12.1 years. The MDD and BD subgroups were similar on age, education, BMI, age at onset and prior psychiatric hospitalisation, although BD included a higher proportion of males. Antidepressant outcomes: On day 13 (one day after the sixth infusion) 46 of 84 completers (54.8%) met response criteria (≥50% reduction in HAMD) and 26 (31.0%) met remission criteria (HAMD <7); response and remission rates did not differ between MDD and BD subgroups. A linear mixed model for HAMD showed a significant main effect of time (F=124.223, p<0.001), with significant decreases from baseline to day 13 and to day 26 (both p<0.001). Neurocognitive outcomes: Two MCCB domains showed significant time effects in the whole sample. Speed of processing improved over time (linear mixed model F=9.344, p<0.001), with significant improvements at day 13 (t=-0.277, p=0.001) and day 26 (t=-0.273, p=0.001) versus baseline. Verbal learning also improved (F=5.647, p=0.004), with a significant improvement at day 13 (t=-0.347, p=0.013) compared with baseline. Working memory and visual learning showed no significant change across assessments. Similar patterns occurred within MDD and BD subgroups and there were no significant group effects between diagnoses. Mediation analyses examined whether improvements in cognition were accounted for by reductions in depressive symptoms. When change in HAMD was included as a mediator in the mixed models, the time effect lost significance for speed of processing and several domains, whereas verbal learning retained a significant time effect (F=9.199, p<0.001). The Sobel test (used to assess the significance of mediation) indicated significant indirect effects of time on cognitive change mediated by HAMD reductions. The extracted text reports Sobel test statistics for speed of processing and verbal learning (Sobel test=3.573, p<0.001 for speed of processing; Sobel test=6.649, p<0.001 for verbal learning), indicating that improvements in depressive symptoms significantly mediated the observed improvements in these cognitive domains. Prediction of antidepressant response: In logistic regression adjusting for baseline HAMD, higher baseline visual learning scores predicted response at day 13 (B=0.118, p<0.001). The absence of psychiatric comorbidity also predicted response (B=-1.822, p=0.021). No other baseline cognitive domains were reported as significant predictors. Attrition occurred primarily for concerns about side effects or dissatisfaction with efficacy; 13 patients discontinued during the infusion course and were excluded from the primary analyses.

Zhou and colleagues interpret their findings as providing evidence that six subanesthetic ketamine infusions given over 12 days do not produce deleterious neurocognitive effects in patients with unipolar or bipolar depression. Rather than cognitive decline, small improvements were observed in verbal learning at day 13 and in speed of processing at day 26. Importantly, mediation analyses suggested these cognitive gains were largely explained by concomitant reduction in depressive symptoms: changes in HAMD scores significantly mediated the relationship between time and improvements in speed of processing and verbal learning. The authors therefore suggest that the apparent cognitive improvements reflect recovery from depression rather than a direct procognitive action of ketamine. The investigators situate their results within mixed prior literature. Acute or chronic high-dose ketamine exposure in healthy volunteers and recreational users has been associated with cognitive impairment, but studies using subanesthetic doses for depression have generally found minimal or reversible effects; small previous studies of serial infusions reported no long-term deficits and in some cases improved memory. Zhou and colleagues argue that their larger sample strengthens the evidence that a six-infusion, 0.5 mg/kg regimen carries minimal cognitive risk in the short term. Mechanistic discussion notes that although NMDA antagonism can impair synaptic processes relevant to cognition, ketamine's antidepressant effects may engage mechanisms such as increased brain-derived neurotrophic factor (BDNF) and hippocampal plasticity, which could counteract or reverse depression-associated cognitive deficits. Regarding predictors of response, the study found that better baseline visual learning (and absence of psychiatric comorbidity) predicted antidepressant response to six infusions. This contrasts with some prior reports that poorer baseline cognition predicted greater ketamine benefit. The authors suggest that higher baseline cognitive function, larger hippocampal volume and higher BDNF levels may indicate greater neuroplastic capacity and thus better potential to benefit from antidepressant interventions, but they acknowledge that the discrepancy with earlier findings remains unresolved. The authors acknowledge several limitations: attrition (13 patients discontinued during infusions, potentially introducing selection bias), the open-label single-arm design without a control group (allowing placebo or rater-expectancy effects), pooling MDD and BD participants (reducing sample homogeneity), and continuation of background psychotropic medications (which might mask or interact with ketamine effects). They recommend randomised double-blind controlled trials, studies in medication-free samples, and further work to identify robust predictors or moderators of ketamine response. Overall, the study team concludes that six ketamine infusions at the tested regimen appear safe with respect to short-term neurocognitive performance, and that baseline visual learning warrants further investigation as a potential predictor of response.

The authors conclude that six serial infusions of ketamine at 0.5 mg/kg did not produce neurocognitive deterioration in this sample of patients with unipolar or bipolar depression; instead, modest improvements in verbal learning (by day 13) and speed of processing (by day 26) were observed, although these improvements were mainly mediated by reductions in depressive severity. Greater baseline visual learning performance predicted antidepressant response across the six infusions. The investigators suggest that, within the limits of this open-label study, ketamine appears safe regarding short-term neurocognitive outcomes and that further controlled studies are needed to confirm cognitive safety and to identify clinically useful predictors of response.