Effect of ketamine on anxiety: findings from the Ketamine for Adult Depression Study

The paper frames anxious distress as a common and clinically important feature of major depressive disorder (MDD), noting that comorbid high levels of anxiety are associated with poorer antidepressant outcomes. Earlier studies and case reports have suggested that ketamine can produce rapid reductions in depressive and anxiety symptoms, but the literature is limited by small samples, open-label designs, variable dosing and mixed patient populations. Few trials have specifically examined ketamine's anxiolytic effects in people with treatment-resistant depression (TRD), and prior TRD studies that reported anxiety change used single infusions or small samples, leaving uncertainty about dose, durability and whether anxiety improvements are independent of mood improvement. Mills and colleagues set out to examine whether subcutaneous racemic ketamine administered twice weekly for 4 weeks reduces anxiety in people with TRD. Using data from the Ketamine for Adult Depression Study (KADS), a multicentre randomised, double-blind, active (midazolam) controlled trial, the authors report secondary analyses with the Hamilton Anxiety Rating Scale (HAM-A) as the primary anxiety measure and MADRS item 3 (inner tension) as a supplementary measure. Specific hypotheses were that ketamine would reduce HAM-A and MADRS item 3 scores versus midazolam, that effects would be larger in participants with comorbid anxiety disorders independent of antidepressant effects, and that higher ketamine doses would produce greater anxiolytic effects.

KADS was a multi-centre, double-blind randomised controlled trial conducted across seven sites in Australia and New Zealand. Adults with treatment-resistant major depressive disorder were randomised to receive subcutaneous racemic ketamine or an active psychoactive placebo, midazolam, administered twice weekly for 4 weeks. The trial operated in two sequential cohorts: cohort 1 used a fixed low ketamine dose (0.5 mg/kg; midazolam 0.025 mg/kg), whereas cohort 2 used flexible, response-guided titration (ketamine 0.5–0.9 mg/kg; midazolam 0.025–0.045 mg/kg) with up-titration guided by MADRS response. Ethics approvals and written informed consent were obtained; the protocol and cohort procedures have been published previously. Anxiety was assessed with the 14-item HAM-A (total score and psychic versus somatic subscales) at baseline, end of treatment (4 weeks) and 4 weeks after treatment end. MADRS item 3 (inner tension) was measured at the same time points and was analysed as an ordinal outcome. Presence of comorbid anxiety disorders was determined by DSM-5 diagnosis during a psychiatrist-conducted screening interview. The authors treated these anxiety analyses as secondary and exploratory. Statistical approaches included mixed effects repeated measures models (MRMM) for continuous HAM-A outcomes with fixed effects for Time, Group and their interaction and baseline HAM-A as a covariate; participants were included as a random effect. A cumulative link mixed model (CLMM) was used for ordinal MADRS item 3. Mediation analyses estimated the indirect effect of change in total MADRS score (baseline to end treatment) on HAM-A outcomes, reporting average causal mediation effects (ACME), average direct effects (ADE) and proportion mediated. Subgroup analyses examined psychic versus somatic HAM-A factors and outcomes in participants with versus without comorbid anxiety disorders. Dose–response exploratory analyses compared ketamine-treated participants across cohorts and used linear regression to test associations between ketamine dose (mg/kg and mg) and change in HAM-A, adjusting for potential confounders. Analyses were conducted in R; results are reported uncorrected for multiple comparisons and labelled exploratory.

Of 184 participants randomised, 174 received at least one dose of study drug and 167 had HAM-A assessed at both baseline and end treatment. Baseline HAM-A scores averaged roughly 20–22, corresponding to the upper end of moderate anxiety. Across the combined cohorts and in cohort 2, MRMMs showed a significant difference favouring ketamine over midazolam in change in total HAM-A across end treatment and 4 weeks after treatment end (combined cohorts P = 0.0071; cohort 2 P = 0.0058). No significant group difference was observed in cohort 1. Post hoc testing indicated the ketamine–midazolam difference was present at end treatment but not at 4 weeks after treatment end. In the combined cohorts the end-treatment difference was significant (P = 0.0093; Cohen's d = 0.53) but absent at 4 weeks (P = 0.80; Cohen's d = 0.05). In cohort 2 the end-treatment difference was also significant (P = 0.004; Cohen's d = 0.77) and non-significant at 4 weeks (P = 0.46; Cohen's d = 0.20). Thus, anxiolytic effects were short-lived without further treatment. Mediation analyses examined whether changes in depression (total MADRS) explained the HAM-A effect. In the combined cohorts the indirect effect via MADRS change was significant (ACME = -1.2, 95% CI [-1.9, -0.5], P < 0.0001) while the ADE (direct effect) was non-significant (0.1, 95% CI [-1.4, 1.7], P = 0.90). Cohort 2 showed a similar pattern (ACME = -1.4, 95% CI [-2.5, -0.6], P < 0.0001; ADE = -0.5, 95% CI [-2.7, 1.8], P = 0.68). Linear regression also found that reductions in HAM-A were associated with reductions in total MADRS during the RCT (β = 0.48, s.e. = 0.090, P < 0.0001). The authors conclude that much of the ketamine–anxiety relationship is mediated by antidepressant benefit, although group effects remained significant in some models after controlling for MADRS change. Analyses of MADRS item 3 (inner tension) found no significant difference in cohort 1 or in the combined cohorts across end treatment and 4 weeks after treatment end, but a significant advantage for ketamine in cohort 2 (P = 0.026), driven by end-treatment differences (P = 0.0027). Subscale analyses of HAM-A showed a significant Group effect for the psychic subscale in the combined cohorts (P = 0.011), driven by end-treatment differences (P = 0.011; Cohen's d = 0.54), whereas the somatic subscale showed no significant Group effect. In participants with a comorbid anxiety disorder, ketamine also outperformed midazolam on total HAM-A across end treatment and 4 weeks after treatment end (P = 0.019), with a post hoc end-treatment difference (P = 0.016; Cohen's d = 0.82) that did not persist at 4 weeks. Mediation in this subgroup again showed a significant indirect effect via MADRS change (ACME = -1.3, 95% CI [-2.8, -0.3], P = 0.008) with a non-significant ADE. MADRS item 3 differed between groups in those with comorbid anxiety (P = 0.016), driven by end-treatment differences (P = 0.018). When MADRS item 3 was removed from the total MADRS score among ketamine recipients, depression scores did not differ significantly between those with and without comorbid anxiety (P = 0.13). Exploratory dose analyses found a larger mean HAM-A decrease in cohort 2 ketamine recipients versus cohort 1 ketamine recipients, but the cohort comparison within the ketamine arm did not reach statistical significance (β = -4.00, P = 0.14). Regression analyses testing maximum ketamine dose (mg/kg or mg) against HAM-A change were non-significant (P = 0.49 and P = 0.51 respectively), suggesting no clear dose–response in these exploratory tests. Overall, anxiolytic effects were apparent at end treatment in adequately dosed participants but were not maintained at 4 weeks after treatment cessation.

Mills and colleagues interpret their findings as supporting an anxiolytic effect of subcutaneous ketamine in people with treatment-resistant depression when administered at adequate doses. The effect was evident at the end of the 4-week treatment course in the combined sample and in cohort 2 (response-guided, higher dosing), but not in cohort 1 (fixed lower dose), and was not maintained at the 4-week post-treatment follow-up without further intervention. While mediation analyses indicated a substantial indirect effect of ketamine on anxiety via reduction in depressive symptoms, the group effect on HAM-A remained significant in models controlling for MADRS change, suggesting antidepressant response does not entirely account for anxiolytic effects. The reduction was concentrated in the HAM-A 'psychic' factor and was mirrored by improvements on MADRS item 3 in cohort 2. The authors place their findings in context with prior ketamine and esketamine studies that reported anxiolytic effects, noting the literature is limited by small samples and methodological variability. They highlight strengths of the present analysis: a relatively large randomised sample, active blinding with midazolam, repeated anxiety measurements, multiple anxiety outcome measures and control for mood change. Limitations acknowledged include that these analyses were secondary and exploratory with uncorrected multiple comparisons, relatively small numbers in some subgroups (notably participants with comorbid anxiety in certain comparisons), and the sequential cohort design that complicates inference about dose effects. The authors also note that the lack of durability at 4 weeks could relate to initial dosing strategy and time to reach adequate dose in cohort 2. For future research the study team recommends hypothesis-driven trials to replicate these findings, longer treatment durations with individualised titration, investigation of maintenance strategies and trials combining ketamine with psychotherapies such as cognitive-behavioural therapy to prolong benefits. They caution that, given the exploratory nature of the present analyses, confirmation in subsequent trials is required before changes to clinical practice or guideline recommendations are considered.