In a randomized, double-blind crossover study of 36 people with treatment-resistant depression (TRD) and 25 healthy volunteers, ketamine altered the temporal dynamics of sleep EEG spectral power in TRD—producing earlier-night increases in delta power and later-night increases in alpha and delta. These spectral changes occurred without effects on sleep macroarchitecture (e.g. WASO, TST, REM latency), did not mediate ketamine’s antidepressant or anti‑suicidal effects, and baseline TRD showed lower total sleep time and shorter REM latency.
Papers cited by this study that are also in Blossom
Wilkinson, S. T., Ballard, E. D., Bloch, M. H. et al. · American Journal of Psychiatry (2017)
The glutamatergic modulator ketamine is associated with changes in sleep, depression, and suicidal ideation (SI). This study sought to evaluate differences in arousal-related sleep metrics between 36 individuals with treatment-resistant major depression (TRD) and 25 healthy volunteers (HVs). It also sought to determine whether ketamine normalizes arousal in individuals with TRD and whether ketamine’s effects on arousal mediate its antidepressant and anti-SI effects. This was a secondary analysis of a biomarker-focused, randomized, double-blind, crossover trial of ketamine (0.5 mg/kg) compared to saline placebo. Polysomnography (PSG) studies were conducted one day before and one day after ketamine/placebo infusions. Sleep arousal was measured using spectral power functions over time including alpha (quiet wakefulness), beta (alert wakefulness), and delta (deep sleep) power, as well as macroarchitecture variables, including wakefulness after sleep onset (WASO), total sleep time (TST), rapid eye movement (REM) latency, and Post-Sleep Onset Sleep Efficiency (PSOSE). At baseline, diagnostic differences in sleep macroarchitecture included lower TST (p = 0.006) and shorter REM latency (p = 0.04) in the TRD versus HV group. Ketamine’s temporal dynamic effects (relative to placebo) in TRD included increased delta power earlier in the night and increased alpha and delta power later in the night. However, there were no significant diagnostic differences in temporal patterns of alpha, beta, or delta power, no ketamine effects on sleep macroarchitecture arousal metrics, and no mediation effects of sleep variables on ketamine’s antidepressant or anti-SI effects. These results highlight the role of sleep-related variables as part of the systemic neurobiological changes initiated after ketamine administration. Clinical Trials Identifier: NCT00088699.
Ballard and colleagues situate their study within evidence that ketamine, a glutamatergic modulator, produces rapid reductions in depressive symptoms and suicidal ideation alongside transient changes in neurophysiological and molecular biomarkers. Previous work links ketamine to shifts in awake-state electrophysiology (for example gamma power) and to sleep-related measures such as increased slow wave activity in the first non-rapid eye movement (NREM) episode. Because sleep disturbances and nocturnal hyperarousal (indexed by higher nocturnal alpha and beta power and reduced delta power) are common in major depressive disorder (MDD) and have been implicated in suicide risk, the authors argue that sleep-related electrophysiology may be an important component of ketamine's system-level actions and a potential mediator of its clinical effects. This study therefore set out to characterise sleep-related arousal in individuals with treatment-resistant depression (TRD) versus healthy volunteers (HVs), to test whether a single 0.5 mg/kg intravenous ketamine infusion normalises arousal-related sleep metrics in TRD, and to evaluate whether ketamine-induced changes in sleep metrics mediate its antidepressant and anti-suicidal effects. The investigators emphasised a combined approach using both traditional macroarchitecture sleep measures (for example wakefulness after sleep onset, total sleep time, REM latency, and post-sleep onset sleep efficiency) and functional data analysis of spectral EEG power over time (alpha, beta, delta bands) to retain temporally resolved electrophysiological information.
The analysis used data from a previously published randomised, double-blind, placebo-controlled, crossover clinical trial of intravenous ketamine (0.5 mg/kg) versus saline. Participants included unmedicated adults aged 18–65 with recurrent MDD without psychotic features who met a minimal treatment-resistance definition (failure to respond to at least one medication during the current episode) and healthy volunteers with no Axis I disorders and no first-degree relatives with Axis I disorders. After medication washout, TRD participants (n = 36; 61% female; mean age 36.4 years, SD = 10.29) and HVs (n = 25; 64% female; mean age 34.0 years, SD = 10.6) underwent polysomnography (PSG). Five PSGs were excluded for technical problems, yielding the final sample reported. Each participant had an adaptation night; PSGs were recorded the night before (baseline) and the night after (post-infusion) each infusion. For ketamine-specific hypotheses, analyses were restricted to the TRD participants' baseline and post-infusion PSGs. PSG was performed in a sleep laboratory with standard EEG, electrooculogram and submental electromyogram channels, and 30-second epochs were scored blind to diagnosis and treatment using Rechtschaffen and Kales criteria. Macroarchitecture measures derived were wakefulness after sleep onset (WASO; hours awake between sleep onset and Lights-On), total sleep time (TST; hours asleep after sleep onset until Lights-On), REM latency (minutes from sleep onset to first REM epoch), and post-sleep onset sleep efficiency (PSOSE; TST divided by time between sleep onset and Lights-On). Functional spectral analyses measured normalised alpha, beta and delta power across 605 30-second epochs (approximately 5.04 hours) beginning at the first NREM Stage 2 epoch (defined as sleep onset here). For time-resolved spectral analyses, the investigators used multilevel function-on-scalar regression (FoSR) to model EEG power as a smooth function of time and to preserve temporal dynamics across the five-hour window; inference employed the Fast Univariate Inference (FUI) approach with 95% joint confidence bands (implemented via the R package fastFMM). Macroarchitecture comparisons used mixed-effects models: linear mixed models for WASO (log-transformed due to right skew) and TST (transformed by subtracting from eight then log-transforming), a beta-distributed generalised linear mixed model with logit link for PSOSE, and a Cox proportional hazards model for REM latency with robust standard errors to account for two nights per person. Ketamine effects in TRD were tested with analogous within-person mixed models including covariates (age, sex, infusion number, period-specific and average baseline values). Mediation analyses followed the Baron and Kenny approach: first testing ketamine's effect on outcomes (Day 1 suicidal ideation and MADRS*), then ketamine's effect on EEG spectral functions, and finally whether spectral functions predicted clinical outcomes via scalar-on-function mixed-effects regression. No correction for multiple comparisons was applied; all analyses were performed in R.
Baseline diagnostic comparisons. Function-on-scalar regression suggested a general pattern of lower alpha, beta and delta power in the TRD group during the earlier part of the five-hour post-sleep onset interval, but 95% joint confidence bands included zero throughout and these spectral differences were therefore not statistically significant. In contrast, macroarchitecture showed significant diagnostic differences: total sleep time was lower in TRD than HVs (b = 0.36, SE = 0.12, p = 0.006; approximated Cohen's f2 = 0.14). REM latency was shorter in TRD with a greater event hazard (b = 0.48, SE = 0.20, p = 0.04); the median time to first REM epoch in TRD was 55 minutes, while the extracted text did not clearly report the corresponding HV median. No baseline diagnostic differences were observed for WASO (b = 0.005, SE = 0.19, p = 0.98) or PSOSE (b = 0.01, SE = 0.18, p = 0.94). Ketamine effects on spectral dynamics in TRD. Comparing post-ketamine to post-placebo nights within TRD participants, multilevel FoSR analyses with 95% joint confidence bands supported significant increases in EEG power across all three frequency bands, with temporally specific patterns. Delta power showed recurring significant increases at multiple intervals across the five-hour window. Alpha power increases were detected primarily at the end of the five-hour period. Beta power increases were observed in the middle portion of the night. These spectral effects were detected despite no concurrent effects on macroarchitecture. Macroarchitecture, clinical response and mediation. Ketamine produced the expected clinical effects in this subset: Day 1 suicidal ideation decreased (p = 0.01) and Day 1 MADRS* scores declined (p < 0.01). However, ketamine did not produce detectable changes in macroarchitecture measures (WASO, TST, PSOSE, REM latency) relative to placebo in the TRD sample, and interaction terms testing whether post-infusion macroarchitecture mediated ketamine's clinical effects were non-significant (p > 0.05 for all). Mediation testing of functional spectral variables likewise did not support mediation of ketamine's antidepressant or anti-suicidal effects by post-infusion alpha, beta or delta power functions; there was a trend for delta power mediating antidepressant effects (p = 0.06) but this did not reach conventional significance. The extracted text reports no significant mediation effects for macroarchitecture measures either.
Ballard and colleagues interpret their findings as evidence that ketamine produces temporally specific changes in sleep-related electrophysiology in individuals with TRD without appreciably altering conventional sleep macroarchitecture measures in the night following infusion. The study replicated baseline features of depression in PSG macroarchitecture—reduced total sleep time and shortened REM latency—but baseline spectral differences were smaller than hypothesised and directionally inconsistent with some expectations, a result the authors note was not statistically significant. They highlight that ketamine increased delta power at recurring intervals during the recorded night and produced later-night increases in alpha and mid-night increases in beta power; these latter spectral effects were contrary to the initial hypotheses but are framed as plausible indicators of altered cortical excitability or plasticity. The authors position the temporal spectral findings relative to prior literature showing ketamine-linked increases in slow wave activity during the first NREM episode and parallels with recovery sleep after sleep deprivation, suggesting possible shared mechanisms between ketamine and sleep-deprivation therapy. They also note relationships with circadian and clock-gene effects reported for both ketamine and sleep deprivation, arguing for a systemic, 24-hour perspective on ketamine's neurobiology. Methodologically, the investigators emphasise the value of functional data analysis for preserving temporally rich EEG information across the night, arguing that this approach captures dynamics lost when PSG is reduced to stage-based summary metrics. Key limitations acknowledged by the authors include sample size and power constraints (the parent trial was powered to detect large effects), the inpatient status of TRD participants versus greater schedule flexibility among HVs (a potential confound with group), the absence of comprehensive self-report sleep measures beyond insomnia items in depression scales, and the focus on a severe TRD sample which limits generalisability. The authors further observe that mediation and moderator analyses of ketamine response face statistical challenges and that larger samples may be required to detect small-to-moderate neurobiological mediators. Strengths highlighted include the placebo-controlled, double-blind, crossover design with within-person comparisons, medication washout prior to PSG in TRD participants, and inclusion of HV PSG data for contextualising diagnostic differences. Finally, the authors recommend replication of these FDA-based findings, investigation of whether the spectral signatures are ketamine-specific or common to rapid-acting antidepressant interventions, and extension to broader clinical samples and complementary self-report measures of sleep quality.
Participants were drawn from a previously published randomized clinical trial (NCT00088699) of intravenous ketamine (0.5 mg/kg) compared to saline placebo. All patient participants were 18-65 years old and diagnosed with recurrent MDD without psychotic features using the Structured Clinical Interview for Axis I DSM-IV Disorders (SCID)-Patient Version. Treatment resistance was characterized as not responding to at least one psychiatric medication during the current depressive episode. Participants were free from psychiatric medications for at least two weeks (five weeks for fluoxetine, three weeks for aripiprazole) prior to baseline assessment and were required to have a score of 20 or more on the Montgomery-Asberg Depression Rating Scale (MADRS)before each ketamine or placebo infusion. The participant sample also included HVs, who did not meet criteria for any Axis I disorders on the SCID and had no first-degree relatives with Axis I Disorders. Participants were voluntary inpatients over the course of the clinical trial. All participants were considered to be in good physical health based on medical history and physical exam. For the current study, a subset of HVs and participants with TRD were selected who had at least one PSG recording for five hours following sleep onset (defined as the first 30-s epoch of stage 2 NREM sleep) (TRD: n = 36; 61% F; mean age = 36.4 yrs (SD = 10.29); HVs: n = 25; 64% F; mean age = 34.0 yrs (SD = 10.6)). See Tablefor the demographic and clinical characteristics of the sample. The study was approved by the National Institutes of Health (NIH) Combined Neuroscience Institutional Review Board (NCT00088699), and all methods were performed in accordance with relevant guidelines and regulations. All participants provided written informed consent.
Baseline diagnostic differences. Baseline diagnostic (MDD vs HV) differences in WASO and TST were tested using a linear mixed model that took the general form: where i represents the participant, j represents infusion number (or period) (1 or 2), b i is a random intercept, and ϵ ij is a normally distributed error term. X T ij represents the row vector of predictor variables; for the baseline models, these included infusion number, age, diagnosis, and sex (the latter three did not depend on j). β is a column vector that contains the estimated coefficient corresponding to each variable in X ij . WASO was right-skewed and log transformed, and because TST was left-skewed, it was subtracted from eight (so that all values remained positive) and then log transformed. Baseline diagnostic differences in PSOSE were assessed using a betadistributed generalized linear mixed model with a logit link (beta GLMM)and a random intercept per person. The model was chosen because the beta distribution is a priori the most appropriate distribution for responses that are essentially continuous but strictly bounded between 0 and 1. For REM latency, diagnostic differences in time to first post-sleep onset REM epoch were modeled using a Cox proportional hazards model with robust standard errors to account for the two nights per person. Age, sex, and infusion number were also included as covariates in the PSOSE and REM latency models. Ketamine's effects in TRD. The effects of ketamine (relative to placebo) on WASO (log transformed as above) and TST in TRD were tested using linear mixed models (as described above). For PSOSE, a beta-distributed generalized linear mixed model (as described above) was used to test the effect of drug, and for REM latency, drug differences in time to first post-onset REM epoch were tested using a Cox proportional hazards model with robust standard errors (as above). For all models, covariates included age, sex, infusion number, period-specific baseline values, and within-person average baseline (to avoid cross-level bias). Mediation. Linear mixed models (see Eq. ()) that each included a drug x post-infusion sleep macroarchitecture measure were used to test the hypothesis that treatment-related effects on macroarchitecture measures mediated ketamine's therapeutic effects on SI at post-infusion Day 1. Models also included age, infusion number, sex, period-specific baseline SI, and average baseline SI as covariates, as well as a random intercept per person. An identical approach was used for MADRS* total score. Due to the exploratory nature of this study, no correction for multiple comparisons was made, and the risk of false negatives was minimized at the expense of incurring a higher false positive risk. All statistical analyses were performed in R.
Create a free account to open full-text PDFs.
Nugent, A. C., Ballard, E. D., Gould, T. D. et al. · Molecular Psychiatry (2018)
Vande Voort, J. L., Ballard, E. D., Luckenbaugh, D. A. et al. · Journal of Clinical Psychiatry (2017)
Murrough, J. W., Iosifescu, D. V., Chang, L. C. et al. · American Journal of Psychiatry (2013)
Rodrigues, N. B., Mcintyre, R. S., Lipsitz, O. et al. · Journal of Sleep Research (2021)
Medeiros, G. C., Gould, T. D., Prueitt, W. L. et al. · Molecular Psychiatry (2022)
Price, R., Kissel, N., Baumeister, A. et al. · Molecular Psychiatry (2022)
This study evaluated sleep-related arousal at baseline and following ketamine administration in unmedicated individuals with TRD and HVs enrolled in a randomized, placebo-controlled, crossover trial. At baseline, the hypothesized shorter REM latency and TST were observed in unmedicated individuals with TRD relative to HVs. Contrary to our hypotheses, however, unexpectedly lower baseline alpha, delta, and beta power were observed in the TRD group relative to HV group, though these findings were not statistically significant. Ketamine's impact on sleep-related brain electrophysiology included temporally specific increases in delta, alpha, and beta power; the latter two effects were contrary to our hypotheses. Specifically, post-ketamine (vs placebo) delta power was significantly greater at regularly occurring intervals during the five hours under study, while post-ketamine alpha power was higher at the end of the night, and post-ketamine beta power was higher in the middle of the night. Ketamine exerted no detectable effects on any of the macroarchitecture measures such as PSOSE or WASO. None of the sleep variables mediated ketamine's impact on depression or SI. Overall, our results underscore the impact of ketamine on the sleeping brain in TRD, as well as the use of functional data analysis (FDA) for understanding sleep electrophysiology following rapid-acting antidepressant treatment. Collectively, the findings support the use of FDA to represent the sleep process as a temporally dynamic spectral power signal. The sensitivity of this approach may be maximized because information is not collapsed over time or within the confines of a defined state, thereby gaining proximity to the symphonic neural sleep processes managing the sleep-arousal balance. All EEG epochs are included in the analysis, including brief movements and more extended wakefulness periods that might have been excluded from more traditional sleep staging approaches. Also, inand) that tended to be inverted relative to temporal patterns in beta power. Additionally, delta power findings in the present study were in line with previous findings of ketamine increasing slow wave activity in the first NREM episode. Other areas of brain research have also supported the use of FDA, including studies using anatomic magnetic resonance imaging (MRI) and functional MRI measures. Because our application of FDA in a study of drug effects on sleep EEG is novel, additional work is needed to replicate these findings in other datasets and treatments to determine whether the potential effects are ketamine-specific or represent a signature of antidepressant response to treatment. Our findings are also particularly relevant for understanding ketamine's neurobiological mechanisms relative to sleep deprivation therapy as a non-pharmacologic rapid-acting antidepressant treatment. Specifically, the ketamine-related increases in alpha and delta power observed here are consistent with reports of alpha and delta power increases in recovery sleep following sleep deprivation in samples of HVs and individuals with seasonal affective disorder. Such increases in alpha and delta power may reflect cortical excitability and brain plasticity processes that are hypothesized to underlie rapid-acting antidepressant mechanisms. Ketamine has also been linked to changes in TRD wrist actigraphy markers of circadian rhythms. Because individuals with MDD have been shown to have altered clock gene expression, it is relevant to note that both ketamine and 36-hour sleep deprivation therapy have been linked to changes in clock gene expression, including Ciart, Per2, Npas4, Dbo, and Rorb. This further supports the possibility of shared mechanisms of action between ketamine and sleep deprivation therapy, including the role of clock gene expression. Because daytime, rather than nocturnal, biomarkers are often the focus of ketamine research, the current findings reinforce the need to take a systemic, 24hour approach to studying ketamine's antidepressant effects. The baseline diagnostic spectral power function comparisons warrant further discussion, as the results for each of the three frequency bands were inconsistent with our hypotheses. First, the lower but nonsignificant group difference in delta power functions over time was qualitatively consistent with our hypothesis (lower delta in individuals with TRD vs HVs), and previously published literature supporting lower delta power-reflecting lower sleep drive-in depression. It is not clear if our failure to reject the null was due to sample effects, factors related to inpatient psychiatric hospitalization, or other methodological issues. Baseline diagnostic differences in beta and alpha power were also non-significant and qualitatively reversed relative to what the original hypotheses. While this could be due to methodological issues such as those mentioned above, an alternative explanation lies in reports supporting a general lowering of power in depression, even in frequency bands associated with wakefulness. Further clarification of the nature and scope of the EEG signature of sleep and sleep drive in depression is recommended. The absence of mediation effects highlights the challenges of identifying neurobiological moderators or mediators of clinical response to rapid-acting interventions. While other trials of repeated ketamine administration in individuals with TRD found that improvements in depression and SI were partially mediated by improvements in self-reported insomnia symptoms, these trials relied on subjective indicators of sleep quality rather than PSGbased measures. It is possible that sleep changes may be implicated in the durability of response to ketamine; however, this analysis was not designed to answer this consideration, and informal correlations between sleep macroarchitecture variables and MADRS scores seven days post-ketamine administration did not support a relationship. In general, the field has struggled to link neurobiological markers with clinical antidepressant response to ketamine. For instance, meta-analyses across ketamine trials found few consistent clinical moderators or neurobiological mediators of antidepressant response to ketamine. Much larger sample sizes may be required to power these types of moderation or mediation analyses, especially given that some trends in our analysis did not achieve statistical significance (e.g., the relationship between antidepressant effect and delta power) as well as the fact that changes in sleep-related arousal are likely only one correlate of the larger, systemic response to ketamine. Limitations of the study include using data from a clinical trial powered to detect large effect sizes, which restricted our ability to detect small or medium effects of a specific biomarker. An additional limitation is that the TRD participants were primarily inpatients over the course of the study while the HV participants were members of the local community who spent most of their time on pass/off the unit and thus had more flexibility in terms of choosing a bedtime and adhering to their usual sleep patterns as requested. This difference in daily schedules is inextricably confounded with group and may have affected the results in a manner that cannot be disentangled from diagnosis. Fortunately, actigraphy data were also collected for 19 HVs, and actimeter (Actiwatch AW64; Philips, Amsterdam, the Netherlands) data could be screened to examine potentially unusual sleep behavior (e.g., staying up all night) in the three to four night(s) before baseline PSGs. However, when the six individuals who appeared to have alterations to their usual sleep patterns prior to PSG were removed, the same pattern of diagnostic differences in the macroarchitecture variables was observed. This suggests that more time off the unit was not necessarily a primary factor driving our diagnostic results. Additionally, there were no self-reported sleep measures beyond the insomnia items from the depression rating scales; future analyses should use scales such as the Pittsburgh Sleep Quality Indexto identify the association between oscillations in alpha, delta, and beta power with selfreported sleep quality. Lastly, this sample comprised individuals with severe TRD; future analyses should consider including a range of depression severity and/or other psychiatric diagnoses implicated in both sleep and suicide risk, including bipolar disorder and post-traumatic stress disorder. Although not a limitation, per se, it should also be noted that the FDA approach used in the present study allowed the activity of alpha, beta, and delta frequency bands to be evaluated as functions over time, independent of REM/NREM sleep stages. Therefore, results should not be considered an effort to replicate our previous slow wave activity (SWA) findingsbecause those depended on boundaries of REM/NREM episodes to calculate SWA power. A primary strength of this study is that its data were drawn from a placebo-controlled, double-blind, crossover ketamine trial and allowed each TRD participant to serve as their own treatment control. In addition, because all TRD participants underwent an inpatient medication washout prior to their first PSG, there were no confounding effects of antidepressant medication. Finally, HVs with PSGs provided important context for the inherent differences in sleep quality in individuals with TRD. In summary, this study used an innovative FDA approach to PSG data and found that ketamine was associated with temporallydependent changes in alpha, delta, and beta power. While these sleep changes did not mediate ketamine's antidepressant effects, the results reinforce the need to consider sleep as a critical variable in understanding the impact of ketamine across neurobiological dynamics and systems.