Identification of an optimal dose of intravenous ketamine for late-life treatment-resistant depression: a Bayesian adaptive randomization trial

Treatment-resistant depression (TRD) in older adults is associated with poor outcomes—disability, cognitive decline and higher suicide risk—and evidence-based pharmacological options are scarce. Earlier research has established some benefit for augmentation strategies such as aripiprazole and mixed support for lithium, combination therapies, rTMS and ECT, but remission rates in pragmatic trials have generally been under 30%. Ketamine, an NMDA receptor antagonist with rapid antidepressant effects in younger adults, is a promising candidate for late-life TRD, yet data on its safety and efficacy in older populations are very limited and concerns persist about potential cognitive harms observed in recreational users and animal models. Prior reports in older adults consist largely of small case series and one small randomized trial of subcutaneous ketamine; esketamine has mixed evidence and practical concerns relating to cost and bioavailability. Oughli and colleagues undertook a pilot clinical trial to evaluate intravenous (IV) ketamine in people aged 60 and older with TRD. The primary aims were to assess acceptability (completion rates), tolerability and safety (adverse events, blood pressure changes, dissociation, craving), with secondary objectives to characterise clinical benefit on depressive symptoms and effects on cognition—particularly executive function (EF) and overall fluid cognition—using the NIH Toolbox. The extracted text does not describe any Bayesian adaptive randomization procedures despite that term appearing in the paper title; the methods reported here describe an open-label, multisite pilot design instead.

This multisite pilot study enrolled 25 community-dwelling adults aged 60 years and older across five sites between October 13, 2020 and November 6, 2021. Inclusion required a DSM-5 diagnosis of major depressive disorder without psychotic features confirmed by SCID-5, failure of at least two adequate antidepressant trials in the current episode as assessed by the ATHF, and baseline depressive symptom severity of PHQ-9 ≥15. Key exclusions were dementia or short‑blessed test score ≥10, schizophrenia-spectrum or bipolar disorder, recent substance use disorder, lifetime recreational use of ketamine or similar dissociatives, use of contraindicated medications (eg. naltrexone, memantine), uncontrolled medical conditions that would affect ketamine safety, and baseline systolic BP >150 mm Hg or diastolic BP >90 mm Hg. Concurrent benzodiazepines and antipsychotics were permitted under dose constraints and clinician review. Participants continued a stable oral antidepressant for at least 4 weeks prior to and during the infusion protocol. The acute treatment phase comprised IV ketamine infusions at 0.5 mg/kg body weight administered over 40 minutes twice weekly for 4 weeks. Those achieving either MADRS <10 or ≥30% reduction in MADRS from baseline after the acute phase were eligible for a continuation phase of once-weekly infusions for an additional 4 weeks. Vital signs were monitored during and after infusions, and dissociative symptoms were assessed using the first six items of the Clinician-Administered Dissociative States Scale (CADSS) pre‑infusion and at 40 and 90 minutes after infusion start. Clonidine (up to 0.6 mg pretreatment, up to 0.3 mg during infusion) was available prophylactically or reactively to mitigate dissociation or hypertensive responses. Primary safety and tolerability measures included adverse events (AEs), serious adverse events (SAEs), treatment-emergent hypertension (SBP >160 or DBP >100), dissociation, and craving (participant-rated Craving Scale). Efficacy measures were MADRS response (≥50% reduction) and remission (MADRS <10) assessed at baseline, end of acute phase and end of continuation phase. Cognitive outcomes were assessed at the same three timepoints using the NIH Toolbox Cognition Battery: Dimensional Change Card Sort (DCCS), Flanker test, List Sorting Working Memory, Picture Sequence Memory, and Pattern Comparison Processing Speed, with a Fluid Cognition Composite score derived from these tests. Analyses were performed on participants with complete data at relevant timepoints (complete case analysis). Descriptive statistics summarised baseline characteristics, while paired t‑tests with 95% confidence intervals and Cohen's d assessed within-subject changes in MADRS and cognitive measures. A Spearman correlation (averaged across subjects via bootstrap) examined trends in CADSS over infusion number. Linear models evaluated associations between changes in EF measures and responder/remitter status, adjusting for baseline EF, education and gender; age was not included because NIH Toolbox scores are age-adjusted. The extracted methods do not describe randomisation or placebo control.

Participant characteristics: the 25 enrolled participants had a mean (SD) age of 71.5 (4.9) years, 13/25 (52%) were female, and all reported European descent. Educational attainment was high (mean 16.2 (1.9) years), medical comorbidity moderate (CIRS‑G mean 8.1 (5.4)), and baseline depressive severity was moderately severe (MADRS mean 24.4 (7.9)). The mean ATHF score for adequate trials was 3.3 (SD 1.4). Completion and retention: 22/25 (88%) completed the acute 4‑week, twice‑weekly infusion schedule. Three participants withdrew during the acute phase (two due to perceived poor response, one without stated reason). Fifteen participants met the pre‑specified criterion (≥30% MADRS reduction or MADRS <10) and all 15 entered and completed the 4‑week continuation phase (100% completion among eligibles). Safety and tolerability: there were no treatment‑related serious adverse events. Two participants (8%) experienced mild, manageable nausea with vomiting and headache. Five participants (25%) had infusion‑related transient hypertension, and eight (32%) received clonidine as pretreatment for blood pressure concerns; one clonidine recipient reported dry mouth. No infusions were discontinued due to elevated blood pressure or other adverse events. Dissociative symptoms occurred in all participants at least once, peaking around 40 minutes after infusion start and resolving by 90 minutes. CADSS scores appeared to decline over repeated infusions, with an average Spearman correlation of approximately −0.28 (bootstrap 95% percentile CI −0.43 to −0.12). Two participants (8%) reported substantial craving (scores 50–100) on part of the Craving Scale, and four (16%) reported milder preoccupation about the next dose. Clinical outcomes: across the full sample, the mean change in MADRS from baseline to end of the acute phase was a decrease of 9.4 points (95% CI 6.46 to 12.32), corresponding to a large within-subject effect size (Cohen's d ≈1.19) and a paired t statistic indicating p <0.01. At the end of the acute phase, 12/25 (48%) met criteria for response (≥50% reduction) and 6/25 (24%) met remission (MADRS <10). Among the 15 who entered continuation, the mean MADRS change between the start and end of continuation was an increase (worsening) of 3.5 points (95% CI 0.38 to 6.56; paired t p = 0.03), with 7/15 (47%) meeting response and 4/15 (27%) meeting remission at continuation end. Cognitive outcomes: 23 participants completed cognitive assessments. During the acute phase there was improvement in the Fluid Cognition Composite and in the three EF measures (Flanker, DCCS, List Sorting), with the Fluid Composite showing a medium-to-large effect (Cohen's d ≈0.61). Linear models found no significant association between changes in EF measures and responder or remitter status, nor between EF change and change in MADRS. During the continuation phase there were no statistically significant changes in EF measures; effect sizes for continuation change were small (Cohen's d range 0.07–0.33). Overall, the age-corrected fluid cognition composite improved during the acute phase and showed no evidence of further change during continuation.

Oughli and colleagues draw four main conclusions from this pilot study. First, IV ketamine was acceptable to older adults with TRD, as indicated by high completion rates in both acute and continuation phases. Second, the regimen appeared generally well tolerated and safe in this small sample: adverse events were uncommon, manageable, and did not necessitate treatment discontinuation; clonidine was used to manage infusion‑related blood pressure elevations. Third, the intervention showed preliminary clinical benefit, with 48% response and 24% remission after 4 weeks of twice‑weekly infusions, and similar proportions among those completing continuation treatment. Fourth, executive function improved during the acute phase and this improvement was maintained during the continuation phase; importantly, no deleterious effects on overall fluid cognition were observed over the 8‑week course. The authors situate these findings with prior work in younger adults, noting that sympathomimetic and psychotomimetic effects are typically transient and that prior reports have not shown cognitive decline with short-term clinical dosing. They also acknowledge preclinical and observational concerns about NMDA antagonist‑related cognitive vulnerability—effects that may be age-dependent and more pronounced with chronic, high‑dose exposure—yet highlight that this study did not detect cognitive impairment over the 8‑week course used. Strengths noted include the multisite design, use of both acute and continuation phases, and systematic measurement of fluid cognition and executive function at three timepoints using a computerized battery. Important limitations are emphasised by the investigators. The study was small and uncontrolled (no randomisation or placebo/comparison arm), and the sample lacked racial or ethnic diversity. There was no adjustment for multiple comparisons, and participants were in the ‘‘young‑old’’ range (mean age 71.5), limiting generalisability to older or more medically complex cohorts. The absence of a control group means practice effects could account for some cognitive improvement; the authors argue against a pure practice effect because some cognitive domains did not improve, but they treat the cognitive findings as preliminary and hypothesis‑generating. Finally, the extracted text does not provide any description of Bayesian adaptive randomization methods implied by the paper title. The investigators conclude that a larger, adequately powered randomised controlled trial is needed to confirm tolerability, safety and efficacy in late‑life TRD, to probe mechanisms, and to assess whether benefits persist beyond 8 weeks.