Predictive value of heart rate in treatment of major depression with ketamine in two controlled trials

Meyer and colleagues frame the study against a background in which ketamine has shown rapid antidepressant effects, but objective predictors of which patients will benefit remain scarce. Previous work has explored genetic markers (for example BDNF polymorphisms), functional neuroimaging correlates, MRS glutamate metrics and sleep-architecture features as candidate predictors, yet clinical implementation has been limited. The authors note that ketamine exerts marked effects on the autonomic nervous system (ANS), typically increasing heart rate (HR) via sympathetic activation, and highlight prior literature linking reduced heart rate variability (HRV) to depression severity and to differential antidepressant response, suggesting that ANS measures could carry predictive information. This study set out to evaluate whether HR and HRV-derived parameters, measured before, during and after a single intravenous ketamine infusion, discriminate responders from non-responders in major depressive disorder (MDD). The principal aim was to test the predictive and discriminative power of simple ECG-derived markers for ketamine treatment outcome, with the hypothesis that higher baseline sympathetic tone (higher HR, greater HRV power and increased low-frequency power) would be associated with larger symptomatic improvement after ketamine.

Two consecutive controlled trials conducted between 2010 and 2015 were pooled for this secondary analysis. Fifty-one patients with MDD aged 18–65 were enrolled; diagnoses were confirmed with the M.I.N.I. and baseline severity required a MADRS score of at least 20. Most participants remained on stable antidepressant medication (no dose changes four weeks prior to study); augmentation with lamotrigine, lithium, antipsychotics and MAO inhibitors was not permitted. Exclusion criteria included suicidality, major psychiatric comorbidity, significant unstable medical or neurological illness and any cardiac condition or cardiac medication (for example beta-blockers). The first trial was double-blind and randomized (placebo versus ketamine), while the second was a single-blind, fixed-sequence design (placebo first, then ketamine for all). For the present analysis only data from the ketamine (verum) cohorts were used. Ketamine was administered intravenously as racemic ketamine hydrochloride with a loading dose of 0.27 mg/kg over 10 minutes, followed by 0.27 mg/kg over 20 minutes, yielding a total of 0.54 mg/kg over 30 minutes. Placebo infusions used matched saline infusion times. Ketamine and norketamine blood levels were sampled 10 and 30 minutes after infusion start. Depressive symptoms were measured with the Montgomery-Åsberg Depression Rating Scale (MADRS) at baseline and at 24 h, 72 h and 7 days post infusion. Response definitions included the conventional 50% symptom reduction and a modified 30% reduction at 24 h; analyses also considered response at any time up to 7 days. ECG was recorded in a semi-recumbent, eyes-closed resting state using a BrainScope amplifier, with a baseline window limited to 5 minutes before infusion and 10-minute segments available at the start and end of the ketamine bolus; 36 subjects had an additional 10-minute ECG 24 hours later. Sampling rates were 250 Hz (Study 1, N = 27) and 1000 Hz (Study 2, N = 24), the latter downsampled to 250 Hz for analysis. RR intervals were processed in Kubios to derive HR, total HRV power (0.04–0.4 Hz) and normalized low-frequency power (nLF; 0.04–0.15 Hz); normalized high-frequency power was not reported separately to avoid redundancy. Statistical analysis pooled the two trials and excluded placebo arms. Baseline comparisons used t-tests or Fisher's exact test. The primary inferential approach was linear mixed modelling with fixed effects for group (responders versus non-responders), time (repeated measures), and group-by-time interaction, plus random intercepts and a random effect for participant; age and sex were included as covariates. Schwarz's Bayesian criterion selected the covariance structure. Post hoc tests were Bonferroni corrected. Cohen's d was calculated from raw ECG parameters for effect-size estimation. Sensitivity analyses restricted models to the first three ECG timepoints when 24-hour recordings were missing. Logistic regression and receiver operating characteristic (ROC) analyses assessed discriminative/predictive performance of HR and HRV measures, using either baseline parameters alone or all available ECG-derived parameters.

ECG data suitable for analysis were available from 47 of 51 participants (92%); four recordings were discarded for artefact or insufficient length. The analysed sample had mean age 43.0 (SD 12.3) and 20 males (43%). Mean baseline MADRS was 24.0 (SD 6.3). Antidepressant treatments in the sample most commonly included sertraline/escitalopram (36% combined), venlafaxine (28%) and mirtazapine (17%). After ketamine infusion, 9 patients (19%) met the conventional 50% MADRS reduction criterion at 24 h, 9 were in remission (MADRS < 10), and 21 (45%) met the modified 30% reduction criterion at 24 h. Across the cohort, HR increased significantly from baseline to during infusion and further post infusion, then returned to baseline by 24 hours (all p < 0.001). Total HRV power showed a significant increase during infusion compared to 24 hours post infusion (p < 0.009). Normalized low-frequency power (nLF) rose significantly immediately post infusion versus baseline (p < 0.035) and versus 24 hours (p < 0.026). There were no significant differences in ECG measures between the two original studies. Mixed-model analyses differed by response definition. Using the 50% reduction at 24 h (n = 9 responders) there was no significant group effect on HR. Applying the 30% criterion at 24 h (n = 21 responders) produced a significant group effect for HR (F = 10.86, df = 147.65, p = 0.001) with additional significant covariate effects for sex and age; Cohen's d for all HR measures was 0.47, and 0.61 when considering baseline HR alone. No significant time or time-by-group interactions were found. Analyses restricted to the first three ECG intervals produced similar results, as did analyses using the 30%/50% response window up to 7 days (30% criterion: F = 11.32, p = 0.001). For HRV power the mixed models showed a significant group effect for the 30% at 24 h criterion (F = 6.65, p = 0.011) and time effects; nLF analysis yielded only a significant time effect. ROC analyses indicated that baseline-only HR and HRV parameters produced an area under the curve (AUC) of 0.68 with sensitivity 72% and specificity 64%, with pre-infusion HR contributing most strongly. When logistic regression included HR and HRV measures from all available recordings (up to four timepoints), the reported AUC rose to 0.96 (sensitivity 94%, specificity 93%), with total HRV power during infusion and post-infusion nLF contributing most. The extracted text, however, also states that including parameters after the first infusion "has no predictive value," and that point is not fully clear from the available extraction. Correlation analyses showed that ketamine and norketamine blood levels correlated positively with HR and negatively with total HRV power; only the negative correlation between ketamine at 30 minutes and total HRV power at 24 hours survived Bonferroni correction. No serious adverse events were reported.

The investigators interpret their findings to mean that ANS activity profiles, particularly higher HR and greater HRV power, differ between ketamine responders and non-responders and that baseline HR carries medium-sized predictive information for clinical improvement. A simple elevated HR before infusion was associated with better symptomatic outcome using the 30% MADRS reduction threshold. The observed increase in HR with ketamine is consistent with known sympathomimetic effects and with ketamine's pharmacology; the authors discuss a possible dissociation in which central nervous system arousal decreases while ANS activity remains high or increases. Meyer and colleagues situate their results alongside prior work showing reduced HRV in depression and earlier studies linking HR/HRV changes to antidepressant response to other agents; they acknowledge that elevated ANS activity in responders might not be specific to ketamine but could mark a broader susceptibility to treatment. The correlation between drug blood levels and HR could suggest a pharmacokinetic contribution, but the authors note that baseline HR was already higher in responders before ketamine exposure, arguing against a purely dose-dependent explanation. Whether higher HR represents a stable trait or a situational state (for example anticipatory arousal) remains unresolved and is highlighted as an area for future work. Key limitations recognised by the authors include the low number of strict (50%) responders at 24 h, which limited statistical power for that outcome; the pooling of two trials with different blinding designs and the fact that HRV analyses were not the primary endpoints of those trials; exclusion of placebo-arm data from the present analysis, which limits inferences about specificity to ketamine; and the exclusion of suicidal patients, which constrains generalisability. Concurrent antidepressant medication in participants is also acknowledged as a potential confounder though the authors argue the setting reflects real-world practice. They recommend prospective studies that pre-specify ANS markers, compare different treatment modalities, and combine multiple biomarkers rather than relying on a single parameter. The authors conclude that further work should assess whether ANS-derived electrophysiological profiles can be used to individualise fast-acting treatment or augmentation strategies in MDD.

The authors conclude that ECG-derived measures of autonomic nervous system regulation, notably heart rate and HRV power, show promise as markers for identifying subgroups of patients more likely to respond to ketamine. They recommend prospective studies to test whether such electrophysiological profiles can guide rapid treatment or augmentation strategies and thereby help personalise therapy for major depressive disorder.