Ketamine versus midazolam in bipolar depression with suicidal thoughts: A pilot midazolam-controlled randomized clinical trial

Earlier research has suggested that sub‑anesthetic ketamine infusions produce rapid antidepressant effects and reductions on suicide-related items in depressed patients, including small randomized trials in bipolar disorder patients maintained on lithium or valproate. Those trials did not, however, specifically enrol patients meeting a threshold for suicidal ideation and relied on saline controls and depression-scale items rather than a dedicated instrument for suicidal thoughts, leaving uncertainty about ketamine's specific anti‑suicidal effects in actively suicidal bipolar depression. Grunebaum and colleagues conducted a pilot, double‑blind, randomized, add‑on trial comparing a single 0.5 mg/kg ketamine infusion with a low‑dose midazolam control (0.02 mg/kg) in patients with bipolar depression and clinically significant suicidal ideation. The primary aim was to assess feasibility, safety, and the effect on suicidal ideation at day 1 post‑infusion, with secondary evaluations of global depression, neurocognition, cortisol awakening response, plasma ketamine and metabolites, and serum brain‑derived neurotrophic factor (BDNF). The authors hypothesised that ketamine would produce a greater reduction in suicidal ideation than midazolam at day 1.

This was a randomised, double‑blind, parallel‑group pilot trial. Participants were adults aged 18–65 with DSM‑IV bipolar disorder currently in a major depressive episode, scoring ≥16 on the HDRS‑17 and ≥4 on the Scale for Suicidal Ideation (SSI). Sixteen participants were randomised and all completed the infusion phase. Interventions comprised a single intravenous infusion of racemic ketamine hydrochloride 0.5 mg/kg or midazolam 0.02 mg/kg, each delivered in 100 mL normal saline over 40 minutes. Concomitant psychotropic medications were permitted except benzodiazepines within 24 hours of infusion. Vital signs and oxygenation were monitored every 5 minutes during infusions; a psychiatrist certified in Advanced Cardiac Life Support supervised each infusion and an anaesthesiologist was available by phone. Non‑responders at day 1 could be unblinded and offered an open‑label ketamine infusion the following day; responders were unblinded after completing follow‑up. After the day 1 assessment participants received clinical treatment for up to 6 months. Clinical assessments were performed by PhD‑ or masters‑level psychologists using standardised instruments. The primary outcome was SSI score at day 1 post‑infusion, with SSI also measured at screening, baseline (day before infusion), 230 minutes post‑infusion, and weekly for 6 weeks. Response was defined as ≥50% reduction in SSI from baseline to day 1 with day 1 SSI <4; remission was day 1 SSI = 0. Depressive symptoms were measured with HDRS‑17 and BDI; mania with YMRS; anxiety by a 5‑point self‑rated Likert item; and global status with CGI. Adverse effects were monitored with SAFTEE, CADSS and the positive symptom subscale of the BPRS. A neurocognitive battery assessed reaction time, processing speed, attention, memory, working memory, pattern separation, language fluency and impulsivity at baseline and day 1. Exploratory biological measures included saliva samples for cortisol awakening response (CAR) pre‑infusion and on day 1, and, in subsequently enrolled participants, post‑infusion plasma ketamine, norketamine and dehydronorketamine (N=11 assayed) and pre/post‑infusion serum BDNF (measured by ELISA). Randomisation used a permuted blocked design (1:1) with block sizes 4–6 and stratification by concurrent psychiatric medication use and baseline SSI <8 versus ≥8. Participants, treating psychiatrists and assessors were blind to assignment; blinding guesses were collected at day 1. The primary analysis was intention‑to‑treat. The main hypothesis was tested using linear regression of day 1 SSI with baseline SSI and treatment as predictors (SSI_day1 ~ SSI_baseline + treatment). Secondary analyses included treatment effects on response and remission, analogous regression models for other clinical outcomes, a generalized least squares model for POMS, and repeated‑measures ANCOVA including the 230‑minute and day 1 SSI assessments. Correlations between clinical change and neurocognitive and biomarker measures were explored. Safety analyses focused on dissociative, psychotomimetic and cardio‑respiratory effects.

All 16 randomised participants completed the infusion and were included in the intent‑to‑treat analysis. Baseline characteristics did not differ between groups on age, sex, race, frequency of past suicide attempts, SSI, HDRS‑17 or BDI. Most participants were taking antidepressants (13/16); other baseline medications included antipsychotics (7/16), lithium (5/16) and anticonvulsants (4/16). Primary outcome: Adjusting for baseline SSI, mean SSI at day 1 was almost 6 points lower in the ketamine group than the midazolam group. This difference corresponded to an estimated treatment effect of 5.84 (SE=3.01), t=1.94, P=0.074, 95% CI = −0.65 to 12.31, representing a trend but not meeting conventional statistical significance in this pilot sample. The number needed to treat (NNT) reported for response (SSI <4 and ≥50% reduction) was 2.2 and for remission (SSI=0) was 3.2. Depressive symptoms: HDRS‑17 scores were nearly 6 points lower at day 1 after ketamine versus midazolam, and the BDI showed an 11‑point differential reduction favouring ketamine; these differences were not statistically significant in this small sample but were described by the authors as consistent with potentially large effects. Follow‑up: Paired t‑tests indicated SSI was lower than baseline at 230 minutes, day 1, week 1 and week 5 in paired comparisons across the follow‑up weeks 1–6. The paper reports mean SSI plotted over 6 weeks for those randomised to ketamine. Neurocognition: At baseline the sample showed below‑average performance in reaction time, memory and language fluency despite above‑average estimated verbal IQ. On day 1 both groups improved in reaction time. The strongest neurocognitive finding was a correlation between memory improvement on the Selective Reminding Test and reduction in SSI after ketamine (Spearman ρ = −0.89, P=0.007, as reported in the abstract). Ketamine tended to blunt improvement in reaction time and language fluency relative to midazolam acutely. Biomarkers: In the assayed subsample serum BDNF decreased pre‑ to post‑infusion in both groups (P>0.8). However, the decrease in BDNF correlated with reduction in SSI from baseline to day 1 after ketamine (n=5, ρ=0.90, P=0.037) but not after midazolam (n=6, ρ=0.09, P=0.087). Baseline saliva cortisol and CAR did not correlate with SSI or HDRS‑17 (P>0.9), and neither baseline CAR nor treatment predicted day 1 CAR in regression models. In a subset of participants who received randomized or open‑label ketamine (n=10), plasma ketamine, norketamine and dehydronorketamine were not associated with day 1 SSI or HDRS‑17. Blinding: Participants correctly guessed their infusion at day 1 in five of seven ketamine cases and seven of nine midazolam cases; clinical assessors guessed correctly after four of seven ketamine and five of nine midazolam infusions. Safety and adverse events: There were no serious cardio‑respiratory infusion‑related adverse events. Ketamine produced transient increases in blood pressure: mean systolic BP rose 23 ± 12.8 mmHg after ketamine versus 3 ± 10.2 mmHg for midazolam (t=3.49, df=14, P=0.004), and mean diastolic BP rose 16 ± 8.8 mmHg versus 4 ± 9.6 mmHg (t=2.46, df=14, P=0.028); BP typically returned to baseline within 4–6 minutes after ketamine. Dissociative symptoms were higher immediately after ketamine (Wilcoxon P=0.001) but not different at 230 minutes or day 1. BPRS positive symptom scores were essentially zero except one case of very mild conceptual disorganisation immediately post‑infusion. Serious adverse events during months 2–6 of follow‑up included six events: four inpatient admissions for increased suicidal thoughts (three admissions for one subject), one suicide attempt by overdose in month 2, and one episode of increased suicidal ideation with preparatory behaviour occurring the evening after an open ketamine infusion; that latter participant's symptoms remitted within 24 hours, which the authors describe as possibly a delayed ketamine response. Mania and anxiety: There were no differential treatment effects at day 1 on manic symptoms (P>0.4) or anxiety (P>0.8), and no post‑infusion manic episodes were observed.

Grunebaum and colleagues conclude that a single sub‑anesthetic ketamine infusion in bipolar depressed patients with clinically significant suicidal ideation is feasible to deliver in a randomised, double‑blind design and is associated with a generally acceptable safety profile in this pilot sample. The authors note that suicidal thoughts were nearly 6 points lower at day 1 after ketamine versus midazolam when adjusted for baseline, but that this difference did not reach statistical significance, a result they attribute primarily to insufficient power in the small pilot sample despite effect sizes and NNTs consistent with large clinical effects. The investigators position their findings alongside prior ketamine trials in bipolar depression and mixed diagnostic samples, emphasising that this study uniquely required a pre‑specified level of suicidal ideation for eligibility. They highlight the concordant reductions in depressive symptom scores and the observed correlations between memory improvement and reduction in suicidal thoughts after ketamine, drawing parallels to prior work in suicidal depression linking memory change to suicidal ideation independent of overall depressive symptom change. Regarding mechanisms, the authors discuss the exploratory biomarker findings cautiously: although peripheral serum BDNF decreased across groups, the decrease correlated with SSI reduction after ketamine in a very small subsample, leaving the relevance unclear given inconsistent literature. No relationships were found for cortisol awakening response or post‑infusion plasma ketamine/metabolites with clinical outcomes in the available subsamples. Key limitations acknowledged by the authors are the small pilot sample size and the exploratory nature of biomarker and cognitive analyses. They report several serious adverse events in the follow‑up period and recommend that, given one SAE occurring close to an infusion, administering ketamine in an inpatient setting is a reasonable precaution for this population. The authors call for an adequately powered RCT to test the promising trend‑level clinical effects and for larger mechanistic studies to elucidate ketamine's neurobiological actions.