A Randomized Add-on Trial of an N-methyl-D-aspartate Antagonist in Treatment-Resistant Bipolar Depression

Bipolar disorder (BPD) remains difficult to treat pharmacologically in part because most available agents were developed for other illnesses and because the neurobiology of bipolar depression is incompletely understood. Recent lines of evidence implicate glutamatergic dysfunction—particularly abnormalities of the NMDA‑receptor complex—in BPD. Postmortem studies report altered NMDA complexes, genetic associations have been observed for GRIN1 and GRIN2B subunits, and glutamatergic modulators such as riluzole show antidepressant effects and influence AMPA receptor trafficking. Preclinical data and clinical work in major depressive disorder also indicate that NMDA antagonists can produce rapid antidepressant effects within hours. Diazgranados and colleagues designed a proof‑of‑concept trial to test whether directly targeting the NMDA receptor with a single intravenous infusion of ketamine would produce a rapid antidepressant response in patients with treatment‑resistant bipolar I or II depression. The primary aim was to determine efficacy and safety of a 0.5 mg/kg ketamine infusion, given as an add‑on to therapeutic lithium or valproate, with measurement of depressive symptoms over minutes to 14 days postinfusion.

Participants were recruited from inpatient psychiatric units, referrals and the internet, and studied at the NIMH Mood Disorders Research Unit between October 2006 and June 2009. Eligible inpatients were aged 18–65 with DSM‑IV bipolar I or II depression without psychotic features, a current major depressive episode of at least 4 weeks, and a Montgomery‑Åsberg Depression Rating Scale (MADRS) score ≥20 at screening and at infusion start. Subjects were required to have failed at least one adequate antidepressant trial and to have failed a prospective open trial of lithium or valproate for a minimum of 4 weeks at therapeutic serum levels. Patients had to be medically stable, free from substance abuse or dependence for at least 3 months, not at high imminent suicide risk, and not receiving other psychotropic medications (washout of 2 weeks, 5 weeks for fluoxetine). Comorbid anxiety disorders were permitted if not the primary focus of treatment. The study used a single‑centre, double‑blind, randomized, placebo‑controlled, crossover add‑on design. After the open mood stabilizer trial and a 2‑week drug‑free period (except for lithium or valproate), subjects received two infusions two weeks apart in randomized order: 0.5 mg/kg ketamine hydrochloride or saline placebo, each administered intravenously over 40 minutes by an anaesthesiologist using identical syringes and infusion pumps. Patients remained on lithium or valproate throughout; levels were monitored weekly. Vital signs and oximetry were monitored during and for 1 hour after infusion, with baseline and end‑of‑study ECG and laboratory testing. The MADRS was the primary outcome, rated 60 minutes before infusion and at 40, 80, 110 and 230 minutes postinfusion, then days 1, 2, 3, 7, 10 and 14. Secondary measures included the Hamilton Rating Scale for Depression (17‑item), Beck Depression Inventory, visual analog scales, Hamilton Anxiety Rating Scale, Brief Psychiatric Rating Scale, Clinician Administered Dissociative Scale, and Young Mania Rating Scale. Ketamine and norketamine plasma levels were measured at several early postinfusion time points. The primary statistical approach used linear mixed models for repeated measures with fixed effects for time and treatment and random intercepts for participants; intent‑to‑treat analyses included all available data. Response was defined as ≥50% MADRS improvement and remission as MADRS <10. Carryover effects were examined and multiple comparisons were adjusted using Bonferroni/Hochberg procedures; Cohen's d was reported for effect sizes.

Fifty‑five subjects were screened and 18 met criteria and were randomized. Thirteen of 18 (72%) completed both phases; overall, 13 of 17 (76%) completed the ketamine phase and 15 of 16 (94%) completed the placebo phase. Five participants dropped out (three in phase one, two in phase two), most during the ketamine phase; four of the five dropouts were taking valproate. Mean pre‑infusion lithium levels were 0.77 mEq/L (SD 0.14) and valproate 74.38 μg/mL (SD 13.02); no patient deviated from therapeutic levels. On the primary outcome (MADRS) the linear mixed model showed a significant time×drug interaction (F10,245 = 3.22, P < .001). Post hoc tests indicated significantly lower depressive symptoms with ketamine versus placebo from 40 minutes through 3 days postinfusion (P < .001); differences at baseline and at days 7, 10 and 14 were not significant after correction. Effect sizes (Cohen's d) in the intent‑to‑treat analysis were 0.52 (95% CI, 0.28–0.76) at 40 minutes, 0.67 (95% CI, 0.42–0.91) at day 1, and peaked at day 2 with d = 0.80 (95% CI, 0.55–1.04); by day 14 the effect was small (d = 0.22; 95% CI, −0.03 to 0.48). Similar results were obtained when analyses were restricted to completers. Responder and remission rates favoured ketamine in the early period. No patients on placebo met response or remission criteria in the first 3 days; after ketamine, 9 of 16 (56%) responded and 2 of 16 (13%) remitted at 40 minutes, while at day 1 7 of 16 (44%) responded and 5 of 16 (31%) remitted. Across the trial, 12 of 17 (71%) patients responded to ketamine at some point versus 1 of 16 (6%) to placebo. Median time to initial response was 40 minutes; the mean duration of response after initial improvement was 6.8 days (SE 1.4), with four patients responding for 1 week and three maintaining response ≥2 weeks. Secondary depression measures (Hamilton Scale, Beck Depression Inventory, visual analog scales) showed similar drug×time interactions, with significant ketamine effects from 40 minutes through 2–3 days. Anxiety symptoms decreased with ketamine on Hamilton Anxiety Rating Scale and visual analog scales. Mania ratings showed a transient increase at 40 minutes with ketamine but significantly lower Young Mania Rating Scale scores on ketamine from 80 minutes through day 2 and at day 14. Dissociative and positive psychotic‑like symptoms increased markedly at 40 minutes with ketamine on the Clinician Administered Dissociative Scale and Brief Psychiatric Rating Scale positive items, but these did not correlate with antidepressant response. Plasma ketamine and norketamine levels at 40 minutes did not correlate with MADRS change. An item‑level MADRS analysis found significant improvement on 8 of 10 items after correction; sleep and suicidal thoughts did not change significantly. No serious adverse events occurred. Common adverse events (≥10%) during infusions included feeling woozy/loopy, lethargy, cognitive impairment, fear/anxiety, nausea, dizziness, odd sensations, blurred vision and headache. Events more specific to ketamine (≥10%) included dissociation, feeling strange, dry mouth, transient tachycardia and increased blood pressure; the cardiovascular changes normalised within minutes. No significant ECG, respiratory or laboratory abnormalities were observed at study endpoints.

In their commentary on the findings, Diazgranados and colleagues report that a single subanesthetic intravenous infusion of ketamine produced a rapid and robust antidepressant effect in treatment‑resistant bipolar depression, with measurable improvement as early as 40 minutes postinfusion. They note this is, to their knowledge, the first controlled demonstration of such rapid antidepressant effects of an NMDA antagonist in a bipolar depressed sample. The effect was observed across multiple clinician‑ and self‑rated depression scales and for anxiety measures as well. The investigators compare these results with their earlier ketamine study in major depressive disorder, pointing out differences including faster onset (40 versus 110 minutes), higher response at 40 minutes but lower response at day 1, similar day‑1 remission rates, and shorter duration of effect in the bipolar sample. Possible reasons for these differences include concomitant use of lithium or valproate in the current study and greater treatment resistance of the bipolar sample. Mechanistically, the authors reiterate preclinical models whereby initial NMDA antagonism by ketamine leads to disinhibition of glutamatergic neurons, increased glutamate release and subsequent increased AMPA relative to NMDA throughput, a sequence thought to underlie rapid antidepressant actions. Strengths highlighted include the inpatient setting with prolonged characterisation and prospective mood‑stabilizer trials prior to randomization, and the randomized, placebo‑controlled crossover design. The authors acknowledge several limitations: the small sample size, the highly refractory and later‑stage illness of participants limiting generalisability, and potential confounding by ongoing lithium or valproate despite arguments against this explanation (mean duration on stabilizer ≈7 weeks and prior nonresponse history). They also emphasise concerns about maintenance of blinding because ketamine produces transient dissociative effects; blinding was not formally assessed by asking raters or patients to guess treatment assignment and this could have influenced placebo response and reporting. Although placebo produced some modest improvement and adverse events in a minority, ketamine effects were substantially greater in the early postinfusion period. Finally, the authors stress that this study was intended to probe rapid antidepressant response related to NMDA receptor antagonism rather than to evaluate relapse prevention. They conclude that the results support targeting the NMDA receptor complex for inducing rapid antidepressant effects in bipolar depression and recommend future work to address blinding, include active comparators, and develop strategies to sustain ketamine's rapid benefit over the longer term.