Do sleep changes mediate the anti-depressive and anti-suicidal response of intravenous ketamine in treatment-resistant depression?

Sleep disturbance is common in major depressive disorder and bipolar disorder and is linked to broader functional problems such as cognitive impairment. Circadian rhythm dysregulation—indexed clinically by delayed sleep phase, prolonged sleep latency and delayed awakening—has been associated with greater depressive symptom severity, and sleep problems have also been related to suicidal ideation and attempts independent of depression severity. Treatments that target circadian disruption (for example, melatonin receptor agonists, bright-light therapy, and sleep-deprivation therapy) can remit depressive symptoms, but a substantial minority of patients do not respond to conventional monoaminergic antidepressants or chronotherapies, leaving a need for novel mechanistic approaches for treatment-resistant depression (TRD). Preclinical and clinical data suggest repeated-dose ketamine and esketamine have antidepressant effects, and there is convergent evidence that ketamine influences circadian genes and sleep architecture (including reductions in sleep latency, increases in slow-wave sleep, and changes in REM and total sleep), raising the possibility that some of ketamine's clinical benefit could be mediated through sleep-related changes. Rodrigues and colleagues set out to evaluate whether self-reported changes in sleep symptoms mediate the antidepressant and anti‑suicidal effects of repeated intravenous (IV) ketamine in adults with TRD. The study specifically examined whether improvements in four sleep domains (insomnia, night‑time restlessness, early morning waking and hypersomnia), and an aggregate sleep score derived from the QIDS‑SR16, mediated changes in total depressive symptoms and suicidal ideation (SI) across four ketamine infusions delivered in a community clinic setting.

This retrospective, exploratory analysis used clinical data from adults treated at the Canadian Rapid Treatment Center of Excellence (CRTCE) in Mississauga, Canada, between July 2018 and November 2020. Treatment resistance was defined as failure of two or more adequate trials of different antidepressant classes. Participants received four IV ketamine infusions over 1–2 weeks: the first two infusions were typically 0.5 mg/kg administered over 40–45 minutes; if a participant had ≤20% reduction in QIDS‑SR16 total score after those doses, dose optimisation to 0.75 mg/kg was offered for the remaining two infusions. Three infusions were scheduled in the evening (18:00–21:00) and one during daytime (09:00–15:00). A clinic psychiatrist followed up approximately 7–14 days after the fourth infusion. Participants completed the QIDS‑SR16 30 minutes before each infusion and at follow-up; this scale (range 0–27) includes one SI item (0–3) and five sleep‑related items from which four sleep questions and a total sleep score were derived. Analyses were conducted in SPSS v26. Change over time in each sleep item across the four infusions was assessed with one‑way mixed models using a compound symmetry covariance matrix and Restricted Maximum Likelihood (REML); models controlled for age, sex, primary diagnosis and concomitant antidepressant treatment. Mediation analyses used the Process macro (v3.4) for SPSS: infusion number was the predictor (X), the mean score for each sleep question (and the total sleep score) served as mediators (M), and outcomes (Y) were QIDS‑SR16 total score and the QIDS‑SR16 SI item. Within mediation models, path a was the unstandardised regression coefficient between predictor and mediator, path b the coefficient between mediator and outcome, c the total effect and c' the direct effect after accounting for the mediator; bootstrapping provided 95% confidence intervals. Unstandardised betas and kappa squared (κ2) effect sizes were reported. Finally, binomial logistic regression models tested whether change in each sleep variable (controlling for age, sex and primary diagnosis) predicted likelihood of being a responder (≥50% reduction in QIDS‑SR16 from baseline to post‑treatment) or a remitter (QIDS‑SR16 ≤5 at post‑treatment). Significance was set at two‑sided p = .05.

Of 332 patients who received IV ketamine during the study period, nine were excluded for missing QIDS‑SR16 data at all four infusions, leaving 323 patients for analysis. The mixed models confirmed that IV ketamine was associated with reductions in QIDS‑SR16 total score and SI (previously reported). There was a significant main effect of infusion on insomnia (F(4,729) = 6.9, p < .001), night‑time restlessness (F(4,730) = 3.5, p = .007) and early morning waking (F(4,729) = 2.5, p = .04), indicating these self‑reported sleep problems improved across infusions. Hypersomnia showed a nonsignificant trend (F(4,729) = 2.0, p = .10). Mediation analyses found a significant total effect of infusion number on QIDS‑SR16 total score (c = -1.47, 95% CI [-1.7, -1.2]); the extracted text reported a t statistic that appears corrupted and is not clearly reported. Small but significant partial mediation was observed for each sleep domain: insomnia, night‑time restlessness, early morning waking and hypersomnia produced small κ2 effect sizes between 0.01 and 0.08, while the mediation model using the total sleep score yielded a medium effect size (κ2 = 0.09). For suicidal ideation, there was a significant total effect of infusion number on the QIDS‑SR16 SI score (c = -0.149, t(945) = -6.8, 95% CI [-0.19, -0.11], p < .001). Insomnia, night‑time restlessness, early morning waking and hypersomnia each acted as significant partial mediators of the relationship between number of infusions and SI; the total sleep score was also a significant partial mediator. Overall, the authors report that improvements in sleep contributed a small but statistically significant portion of the reductions in depressive symptoms and SI associated with IV ketamine. For the responder/remitter analysis, 59 patients met endpoint criteria (56 responders, 28 remitters; all but two remitters were also responders). A further 142 patients did not meet responder or remitter criteria at endpoint, and 131 patients were excluded from logistic regression due to unavailable endpoint data. In models predicting responder/remitter status, change in insomnia symptoms was a significant predictor (χ2(4) = 19.89, p < .001; Nagelkerke R2 = 14.5%), with greater reductions in insomnia associated with higher odds of response/remission (OR 2.15, 95% CI 1.47–3.13). Reduction in early morning waking also predicted response/remission (χ2(4) = 13.50, p = .009; Nagelkerke R2 = 10%; OR 1.65, 95% CI 1.22–2.23). Change in hypersomnia did not predict outcome (χ2(4) = 2.23, p = .693). The overall sleep total score model was the strongest predictor (χ2(4) = 28.49, p < .001; Nagelkerke R2 = 20%), with greater improvement in total sleep associated with markedly higher odds of achieving response/remission (OR 3.29, 95% CI 2.00–5.41).

Rodrigues and colleagues interpret their findings as evidence that subjective improvements in multiple sleep domains partially mediate the antidepressant and anti‑suicidal effects of repeated IV ketamine in patients with TRD. Although mediation effect sizes were generally small, the results indicate that reductions in insomnia, night‑time restlessness and early morning waking—and improvements in an overall sleep score—make a statistically significant but limited contribution to reductions in depressive symptoms and SI. Improvement in hypersomnia was less consistently associated with outcome and did not predict responder/remitter status. The authors situate these findings within broader literature linking sleep, chronobiology and depression: ketamine's known influence on circadian gene expression and sleep architecture (for example, phase‑advancing sleep and increasing slow‑wave sleep) offers a plausible mechanistic pathway connecting glutamatergic modulation, sleep normalisation and mood improvement. They also note possible links between sleep disturbance, systemic inflammation and depressive phenomenology; the discussion cites evidence that sleep problems correlate with inflammatory markers and that inflammatory status may moderate ketamine response in some patients. In terms of suicidality, the authors highlight that worsening insomnia has been independently correlated with increased SI in prior work and suggest ketamine's anti‑suicidal effects may in part operate through sleep improvement. Key limitations are acknowledged. The analysis is retrospective and uncontrolled, and all patients were aware they were receiving ketamine, so expectancy effects cannot be excluded. Sleep was assessed only via four self‑report items from the QIDS‑SR16 rather than comprehensive sleep questionnaires or objective measures (for example, actigraphy or polysomnography), limiting the breadth and objectivity of sleep characterisation. Data on duration of concomitant antidepressant treatments were unavailable, preventing full assessment of their chronotherapeutic contributions. The follow‑up interval reported in the extraction is incomplete, but the methods state follow‑up occurred approximately 7–14 days after the final infusion. Finally, missing endpoint data led to exclusion of a substantial subset from the responder/remitter analyses, which may affect generalisability. The authors conclude that while sleep improvements contribute to ketamine's clinical effects, they account for only a modest portion of the antidepressant and anti‑suicidal response, and the findings should be interpreted in light of the study's methodological constraints.