Sleep architecture and cognitive changes in olanzapine-treated patients with depression: A double blind randomized placebo controlled trial

A 15% change in the primary outcome measure, SWS, was selected to be a clinically significant and meaningful change; this effect size was used to calculate sample size, indicating that 30 participants (15 placebo, 15 treatment) were needed to have enough power to detect an effect of olanzepine on SWS. To account for drop-outs, we aimed to enroll 40 participants. The trail was stopped, however, before reaching recruitment goals (May 2009) due to an inability to recruit participants. Only thirty-one participants signed informed consent to participate in this study.

Participants were screened and enrolled into the study (beginning October 2007) by blinded clinicians (nurses) and recruited from tertiary care mood disorders units, general practitioners offices, and from the community. All were 18 years of age or older, and met DSM-IV-TR criteria [44] for MDD, Bipolar I Disorder, Bipolar II Disorder or Bipolar Disorder NOS, as confirmed by the Mini International Neuropsychiatric Inventory (MINI) [45]. At enrollment, participants were experiencing a major depressive episode (MDE), defined as a Hamilton Depression Rating Scale-17 item (HDRS-17) score of >15, and not a mixed episode, defined as a Young Mania Rating Scale Score (YMRS) of ≤ 12 [46]. Excluded criteria were current or past diagnosis of schizophrenia or dementia, substance abuse within three months of enrolment (excluding caffeine or nicotine), imminent risk of suicide or danger to themselves or others, known intolerance for olanzapine, a serious or inadequately treated medical illness, a history of seizures, previous enrollment in the study or enrollment in another treatment study within 4 weeks prior. Participants could not be taking any other antipsychotic medication at the time of enrolment and must have been on a stable dose of all medications for 4 weeks prior to enrolment. Benzodiazepines and other sleep aids were discontinued if a stable dose had not been achieved for four weeks prior to enrolment (number of participants on each concomitant drug: Placebo: 1 Remeron, 2 Prozac, 1 Valproic Acid, 2 Clonazepam, 1 Lorazepam, 1 Trazodone, 1 Citalopram, 1 bupropion, 1 fluvoxamine, 1, cipralex, 1 oxycontin, 1 Ativan, 1 Topamax, 1 fluvianzine; Olanzepine: 2 Remeron, 2 Prozac, 1 doxepine, 1 amitriptyline,1 oxazepam, 5 Clonazepam, 1 ritalin, 1 Lorazepam, 1 lithium, 1 immovane 1 Citalopram, 2 bupropion, 1 fluvoxamine, 2 cipralex, 2 Effexor). Four participants failed baseline screening (1 had a HAMD score that was too low, 1 was already receiving olanzepine treatment [and also had a too-low HAMD score], and 2 failed to show at the screening appointment) and two participants withdrew (refused to continue to participate) between randomization and day 2–4, and were not included in the analysis. Thus, 25 participants were included in the analysis. Three participants from the olanzapine-treated group terminated from the study before day 28–31, due to a worsening of original mood symptoms. Participants were monitored for sleep apnea and/or hypopnea throughout the course of the study as the presence of these conditions at any point warranted exclusion from the study due to the confounding nature of these events on sleep architecture. No participants needed to be removed due to obstructive sleep apnea/hypopnea.

Participants were assessed at three time-points by a clinician blinded to treatment group: baseline (before randomization and administration of study medication), 2–4 days, and 28–31 days after the administration of study medication. Each clinical assessment consisted of the HDRS-17 [47], the Montgomery Asberg Depression Rating Scale (MADRS) [48], the Hamilton Anxiety Rating Scale (HARS) [49, 50], the YMRS [46], the Pittsburgh Sleep Quality Index (PSQI) [51], Visual Analogue Scale (VAS) for sleep quality [52], and the Epworth Sleepiness Scale (ESS) [53]. At baseline, the MINI and the Clinical Global Impression-Severity (CGI-S) [54], scales were administered. During day 28–31, the Clinical Global Impression-Improvement (CGI-I) was administered. Baseline blood work, physical exam, and pregnancy test (in women of childbearing potential) were performed.

At each visit, participants completed cognitive testing using three Cambridge Neuropsychological Test Automated Battery (CANTAB) tasks, including: Spatial Span (SSP), Spatial Working Memory (SWM) and Reaction Time (RTI) [55, 56] (Cambridge Cognition, Cambridge, UK; http://​www.​camcog.​com). One participant did not complete the CANTAB testing.

Participants were randomly assigned to either placebo or olanzapine (orally disintegrating, zydis, formulation) conditions. A randomization table was generated by a statistician to randomize for order that then placed sealed envelopes containing a single allocation (A:Placebo, B:Olanzapine) into a box in the order specified by the randomization table. Following screening, the participant was instructed to pull the next envelope in the box. This allocation was given to an unblinded pharmacist who filled a prescription that read ‘placebo/olanzapine.’ Medication dosing started at 2.5 mg on day 1 and increased to 5 mg at day 2; during day 2–4, the dosing was titrated up or down, in increments of 2.5 mg to a maximum of 20 mg. The mean dose of olanzapine was 6.67 mg at the end of the study; doses ranged from 5 mg to 10 mg. Eight participants in the placebo-treated group and twelve in the olanzapine-treated group were taking at least one antidepressant as concomitant medications; concomitant antidepressants included amitriptyline, bupropion, escitalopram, citalopram, doxepine, venlafaxine, fluoxetine, remeron, and trazodone. One participant in the placebo-treated group was taking only a benzodiazepine; one participant in the ola\nzapine-treated group was only taking one mood stabilizer; one participant in the placebo-treated group and two in the olanzapine-treated group were not taking any other psychotropic medications. All concomitant medications were stable throughout the duration of the study.

At each time-point, an overnight sleep polysomnograph (PSG) was performed by a clinician blinded to treatment group at the participants’ home, using the MediPalm Personal Recording Device (Braebon Medical Corporation, Kanata, Ontario, Canada). The overnight sleep PSG included four electroencephalogram channels (C4-A1, C3-A2, O2-A1, O1-A2 [57]), an electro-oculogram (two channels), a submental electromyogram (EMG), a finger pulse oximetry, an oronasal airflow (oronasalthermistor), a chest and abdominal movement belt (respiratory inductance plethysmography), a vibration snore sensor and an anterior tibialis EMG. A position sensor was used to monitor position continuously (Ultima Body Position Sensor; Braebon Medical Corporation, Carp, Canada). Participants were not monitored in person. Sharpley et al. [31, 58, 59] have demonstrated that the use of home sleep recordings provides a reliable means of detecting the effects of medications on sleep architecture. Once the PSG was applied (approximately 1900 hrs each study night), a timer was set to record for eight hours following usual sleep time or until the participant rose in the morning. Usual sleep time was verbally confirmed for each participant. Participants were asked to retire and rise at their usual time and were to refrain from alcohol on study nights; however, normal caffeine and nicotine intake was maintained. While participants did not all go to bed at the same time, eight hours or less was sufficient sleep time for those enrolled. Clinicians returned in the morning (following the participants’ usual wake time) to retrieve the PSG equipment. Sleep latencies were calculated from lights out. PSGs were manually scored in 30-second epochs according to standardized criteria of American Academy of Sleep Medicine [60], using Pursuit Advanced Sleep System software (Braebon Medical Corporation, Carp, Canada). Ruehland et al. [61] have demonstrated that measures of sleep architecture are not affected by the number of EEG electrode placements used in scoring the PSG. Technicians were blinded to treatment status. Only one PSG recordings was excluded from analysis due to technical problems. Obstructive apneas and hypopneas, defined as partial airway obstructions leading to a 50% reduction in thoracoabdominal movement lasting for at least 10 seconds [62], were scored using the criteria from the American Academy of Sleep Medicine Task Force [63]. Arousals were scored based on American Sleep Disorders Association criteria, 1992 [57]. The respiratory disturbance index (RDI), which included apneas, hypopneas, and snore arousals for the number of events per hour of sleep, were calculated.

PSG recording and clinical measures (with the exception of the CGI) were analyzed using 2-way and 1-way repeated measures analysis of variance (ANOVA). The design included two treatment groups (placebo × olanzapine) across three time-points (baseline, Day 2–4 and Day 28–31). Independent-samples t-tests examined post-hoc comparisons. Hedge’s g was used to calculate the effect size of the primary outcome measure, changes in SWS, between groups on Day 28–31. Missing PSG, clinical and CANTAB data were replaced with the last observation carried forward. One participant did not complete any of the CANTAB and was not included for this analysis. One-tailed distributions were used for all clinical measures, while for polysomnographic and neurocognitive measures, two-tailed distributions were used. Additionally, demographic comparisons between groups were analyzed using Fisher’s exact tests or ANOVAs (see Table 1). For repeated-measures ANOVA, trend analyses (contrasts) were performed and reported. Neither sphericity of variance nor a significant main effect of the within-subject variable is an assumption of running trend analysis.

Table 2 summarizes all primary and secondary polysomnographic outcome measures for sleep architecture, in both olanzapine and placebo groups, across baseline, day 2–4, and day 28–31. Figure 1 shows the percentage of TST spent in SWS and latency to SWS in both olanzapine- and placebo-treated groups. A 2-way repeated measures ANOVA did not yield significant changes in percentage of TST spent in SWS between the olanzapine- or placebo-treated group (F[1,23] =2.36; p = 0.138; Hedge’s g = 0.53) across time (F[1,23] =0.256; p = 0.775). Latency to SWS significantly decreased with olanzapine treatment by the end of the trial (no main effect of treatment, F[1,23] = 2.49, p = .128; significant interaction of time x treatment, F[1,23] = 5.49, p = 0.028). Comparing the two treatment groups at each time point indicated that at the beginning of the trial latency to SWS was not different (t[23] = -.192, p = .849) but by Day 28–31, the olanzapine treated group reached SWS faster than the placebo group (t[23] = 2.42, p = 0.024; hedges g = 0.97). Latency to sleep however, was not significantly different and could not account for decreased latency to SWS in the olanzapine treated group (no main effect of treatment F[1,23] = 2.17, p = 0.154 or time F[1,23] = 0.297, p = .591). Duration of SWS was also not significantly affected by the olanzapine treatment compared to placebo (F[1,23] =0.581; p = 0.454) over time (F[1,23] = 0.00; p = 0.984; Hedge’s g = 0.13).

TST and Sleep efficiency are shown in Figure 2 for both olanzapine- and placebo-treated groups. Two two-way repeated-measures ANOVA showed an increase in TST (trend for main effect of treatment, F[1,23] = 3.69, p = 0.067) and sleep efficiency (main effect of treatment, F[1,23] = 5.10, p = 0.034) of the olanzapine-treated group compared to the placebo-treated group. Both TST and sleep efficiency demonstrated significant time x treatment interactions (TST: F[1,23] = 4.58; p = .043; sleep efficiency: F[1,23] = 5.27, p = 0.031). Post-hoc independent-samples t-tests were conducted to analyze the nature of these interactions and revealed no significant difference in TST or sleep efficiency at baseline (TST: t₂₃ = -0.325, p = 0.798; Sleep efficiency: t₂₃ = -0.460, p = 0.65) but by Day 28–31 olanzapine treatment significantly increased both (TST: t₂₃ = -2.21; p = 0.037; Sleep efficiency: t₂₃ = -3.30; p = 0.003).

As well as an increase in TST, olanzapine treated participants experienced significantly fewer awakenings (Table 2) and less overall time awake compared to treatment with placebo (2-way repeated measures ANOVA; main effect of treatment, number of awakenings, F[1,23] = 10.0, p = 0.004; awake time, F[1,23] = 4.8, p = 0.037). Post-hoc analysis of the significant interaction of time by treatment in awake time (F[1,23] = 4.93, p = 0.037) indicated that at baseline, awake times did not differ between placebo and olanzapine-treated participants (t₂₃ = 0.531, p = 0.601) but by Day 28–31, olanzapine treatment was significantly decreasing amount of awake time participants experienced (t₂₃ = 2.26, p = 0.003).

Two-way repeated measures ANOVA also analyzed the percentage of TST spent in R and duration of R (Table 2). There was neither a main effect of treatment or time on percentage of TST spent in R (F[1,23] = .796, p = 0.382; F[1,23] = .025, p = 0.875, respectively. There was a significant effect of olanzapine treatment on duration of R (F[1,23] = 4.78, p = 0.039) however there was neither a significant effect of time (F[1,23] = .003, p = 0.957) or a significant interaction of treatment by time (F[1,23] = .696, p = 0.413). Furthermore, Table 2 means indicate that the olanzapine group showed longer R than the placebo group even at baseline, suggesting that the difference in R sleep was not due to the addition of olanzapine.

Clinical measures are included as results of a secondary objective of the study. Figure 3 shows the mean MADRS total scores for both olanzapine- and placebo-treated participants. Two-way repeated ANOVA showed no significant differences between olanzapine and placebo treatment for MARDS or HDRS scores (F[1,23] = 1.09, p = .306; F[1,23] = 0.269, p = .609, respectively). However, both MARDS and HDRS scores significantly decreased with time (F[1,23] = 6.03, p = .022; F[1,23] = 20.0, p = <.001, respectively). One-way repeated measures ANOVA showed a significant decrease in total MADRS score from baseline to the end of the trial (day 28–31) for olanzapine-treated patients (main effect of time: F[1,14] = 6.03, p = .022), where as placebo treatment did not show a decrease (F[1,9] = 0.529, p = .486). HDRS scores also significantly decreased over time in the olanzapine-treated group (F[1,14] =19.6, p = .001), but the placebo-treated group did not (F[1,9] =4.67, p = .059).Table 3 shows clinician administered and self report rating scale scores for both the olanzapine- and placebo-treated groups.

Neurocognitive measures, evaluated as mean CANTAB testing scores, are shown in Table 4 and were assessed as a secondary objective of the study. The between errors and strategy score of the CANTAB SWM task of both the olanzapine- and placebo-treated groups did not significantly change between groups (F[1,22] = 0.3, p = 0.590; F[1,22] = 0.658, p = 0.426, respectively) or across time (F[1,22] = 1.4, p = 0.242; F[1,22] = 0.713, p = 0.407, respectively). The mean spatial span length recalled on the CANTAB SSP task was also not significantly different between treatment groups or across time (2-way repeated measures ANOVA, F[1,22] = 0.102, p = 0.753; F[1,22] = 0.036, p = 0.852, respectively). Finally, the mean reaction time and movement time from the CANTAB RTI task for both the olanzapine- and placebo-treated groups also showed no significant changes as a result of olanzapine treatment (2-way repeated measures ANOVA reaction time and movement: treamtment, F[1,22] = 0.089, p = .768; F[1,22] = 0.15, p = .702, respectively) or over time (2-way repeated measures ANOVA reaction time and movement: time, F[1,22] = 2.28, p = .145; F[1,22] = 2.57, p = .123, respectively).