Effects of Ketamine Versus Midazolam on Neurocognition at 24 Hours in Depressed Patients With Suicidal Ideation

Keilp and colleagues situate the study in the context of evidence that subanesthetic ketamine produces rapid antidepressant and antisuicidal effects within hours to days in some patients. They note a gap in knowledge about ketamine's effects on neurocognitive functions in depressed patients with clinically significant suicidal ideation, despite prior findings linking certain cognitive deficits (for example, slowed processing, impairments in cognitive control) to suicidal behaviour. Previous work in healthy volunteers has shown short‑term detrimental cognitive effects of ketamine across attention, memory, working memory and executive function that are dose dependent and tend to resolve within days; by contrast, some studies in depressed samples have reported cognitive improvements in the days after infusion, but results have been mixed and understudied in suicidal patients. This randomised, double‑blind trial therefore aimed to measure changes in neurocognitive performance from baseline to 24 hours after a single intravenous infusion of ketamine versus an active comparator, midazolam, in depressed inpatients with clinically significant suicidal ideation. The investigators hypothesised global neurocognitive improvement in the period corresponding to ketamine's clinical effects, and they also planned to examine whether baseline cognitive slowing predicted clinical response and whether changes in cognition correlated with changes in depression, suicidal thinking, or general mood. A secondary objective examined a subgroup of midazolam nonremitters who later received open‑label ketamine to determine whether cognitive effects seen after blinded ketamine would be reproduced after open treatment.

This was a double‑blind, randomised comparison of a single intravenous infusion of racemic ketamine (0.5 mg/kg) versus midazolam (0.02 mg/kg) administered over 40 minutes to adult inpatients with DSM‑5 major depressive disorder and clinically significant suicidal ideation. The analytic sample for the neuropsychological study included 78 participants (39 per arm) drawn from a larger clinical trial; two participants from the full trial were excluded from neurocognitive analyses because of limited English or lack of cooperation. Inclusion thresholds were a 17‑item HDRS score ≥ 16 and a Beck Scale for Suicidal Ideation (SSI) score ≥ 4. Benzodiazepines were stopped at least 24 hours before infusion and testing; other psychotropic medications were continued if doses were stable. Neurocognition was assessed with a standard battery targeting processing speed, attention, memory, concentration, language and impulse control. Key tasks included Choice Reaction Time (Choice RT), a computerized Stroop interference measure, WAIS‑III Digit Symbol, Buschke Selective Reminding Test (SRT) immediate and delayed recall, continuous performance measures, and verbal fluency (letter and category). Primary clinical outcomes were HDRS, SSI, and the Profile of Mood States (POMS). Tests were converted to age/sex/education‑adjusted z scores using published or laboratory normative data; alternate forms and counterbalancing were used where possible to reduce practice effects. Procedures comprised a baseline neurocognitive assessment 1–3 days pre‑infusion (day −1), blinded infusion (day 0), and reassessment approximately 24 hours post‑infusion (day 1). Participants who failed to meet a predefined remission criterion for suicidal ideation at day 1 (a ≥ 50% SSI decline and SSI < 4) had their blind broken; midazolam nonremitters were offered open‑label ketamine the following day and, when the protocol allowed, were reassessed about 24 hours after that infusion (referred to as day 3). Because the open‑ketamine assessment was introduced later, 21 midazolam nonremitters completed the third neuropsychological assessment. Statistical analysis used a general linear model omnibus approach testing main effects of assessment point (baseline vs day 1), drug assignment, and test (10 primary outcome measures), plus interactions. Significant omnibus effects led to univariate follow‑ups on individual tests (hierarchical approach, with uncorrected individual test comparisons conditional on omnibus significance). Correlations were computed between baseline cognitive measures and clinical change (HDRS, SSI, POMS), and between change in cognition and change in clinical measures. For the midazolam nonremitters who later received open ketamine, repeated‑measures comparisons across three time points were examined; a covariate for baseline medication status was tested in sensitivity analyses.

Baseline demographic and clinical characteristics were comparable across the ketamine and midazolam groups; both groups had an average estimated intelligence in the superior range. Clinical severity measures improved overall after the blinded infusion: HDRS scores improved in both groups (assessment point effect F1,76 = 63.27, P < .001) without a significant drug × time interaction for HDRS (P = .092). SSI scores improved in both groups but more with ketamine than midazolam (assessment point × drug interaction F1,76 = 7.31, P = .008). POMS total mood scores also improved in both groups and favored ketamine (assessment point × drug interaction F1,74 = 9.91, P = .002). Omnibus analysis of the neuropsychological battery from baseline to day 1 showed a significant main effect of assessment point (F1,76 = 15.83, P < .001), a marginal assessment point × drug interaction (F1,76 = 3.58, P = .062), and a significant assessment point × drug × test interaction (F9,684 = 2.75, P = .004). Overall cognitive performance improved across the battery at day 1 regardless of drug, but changes were not uniform across tests. Relative improvements favouring ketamine emerged on three measures: Choice Reaction Time (assessment point × drug F1,75 = 4.21, P = .044), Stroop Interference (F1,76 = 4.43, P = .039), and Buschke Delayed Recall (F1,76 = 5.91, P = .017). In pairwise testing, Choice RT and Stroop interference improved after ketamine (Choice RT t37 = 3.85, P < .001; Stroop t38 = 2.85, P = .007) with no significant change after midazolam (Choice RT t38 = 0.82, P = .419; Stroop t38 = 0.28, P = .780). For Buschke Delayed Recall, improvement after ketamine was non‑significant (t38 = 1.66, P = .106) and there was a trend toward worsening after midazolam (t38 = 1.80, P = .081). Adjusting for baseline medication status did not alter the significant assessment point × drug × test interaction. Correlational analyses found one nominally significant association among many tested: better baseline Buschke SRT immediate recall correlated with greater HDRS decline (r = −0.39, P = .016), but this was 1 of 30 correlations and no consistent pattern emerged. Baseline slowed processing speed did not predict change in HDRS, SSI or POMS. Change in neuropsychological performance after blinded ketamine was not significantly correlated with change on any clinical severity measure. In the subgroup of midazolam nonremitters who received open ketamine and completed three assessments (n = 21), clinical scores improved significantly on HDRS (F2,40 = 19.94, P < .001), SSI (F2,40 = 20.59, P < .001) and POMS (F2,40 = 24.31, P < .001) by day 3. Neurocognitively, the omnibus comparison across the three time points showed overall improvement (F2,40 = 18.40, P < .001). Mean z score across tests improved from −0.22 ± 0.76 at baseline to −0.04 ± 0.80 at day 1 and +0.30 ± 0.63 at day 3 (day 3 vs baseline and day 1 both P < .001). Specific measures that improved on day 3 relative to day 1 and baseline included Choice RT (F2,40 = 10.86, P < .001), WAIS‑III Digit Symbol (F2,40 = 10.18, P < .001), Stroop Interference (F2,40 = 5.40, P = .008), Letter Fluency (F2,40 = 7.50, P = .002) and Category Fluency (F2,40 = 10.13, P < .001). Buschke SRT immediate and delayed recall did not show improvement after open ketamine. As in the blinded comparison, improvements in neuropsychological test performance after open ketamine were not correlated with clinical improvement.

Keilp and colleagues interpret the findings as showing a modest, general improvement in cognition 24 hours after infusion in both treatment groups, with selective, drug‑related benefits of ketamine on measures of response speed (Choice RT) and interference processing/cognitive control (computerized Stroop). These two measures are sensitive to prefrontal function and have been associated in prior research with past suicide attempts, which led the authors to suggest that ketamine may exert an independent therapeutic effect on neurocognition that is orthogonal to its effects on mood and suicidal ideation. The observed relative change on delayed memory appeared to reflect a small decline after midazolam rather than robust improvement after ketamine, consistent with known acute benzodiazepine effects on memory despite midazolam's short half‑life. Overall, the data suggest that most acute negative cognitive effects of the study drugs had dissipated by 24 hours, while specific enhancements in cognitive control were apparent after ketamine. The investigators emphasise that changes in neurocognition were not correlated with changes in depressive symptoms, suicidal thinking, or general mood after a single infusion, aligning with several prior studies though conflicting with at least one report involving multiple infusions. They propose that improvements in interference processing and cognitive control may represent an additive mechanism by which ketamine could reduce suicide risk beyond mere reductions in ideation, but they acknowledge that causality and the durability of these effects are uncertain. Several limitations noted by the authors are pertinent: the lack of an untreated or inert placebo group limits separation of practice effects from treatment effects, although they cite prior data suggesting little practice effect for the Stroop interference measure over similar intervals. The sample had a high estimated intelligence, which may have attenuated some associations and reflect recruitment via online advertising rather than clinician referral. Concurrent medications were common, but inclusion of a baseline medication covariate did not alter key interactions. Finally, conclusions apply to a single infusion; effects of repeated, long‑term therapeutic ketamine are unknown and warrant study given evidence of persistent neurocognitive impairment with heavy chronic recreational use. The authors recommend further research to clarify mechanisms, durability, and clinical implications of ketamine's neurocognitive effects.