- Top of page
- Financial disclosure
It has been suggested that increase in delta sleep ratio (DSR), a marker for the relative distribution of slow wave activity (SWA) over night time, is associated with clinical response to antidepressant treatment. We examined this index and its relationship to rapid eye movement (REM) suppression before and during long-term treatment with nefazodone, which does not suppress REM sleep, and paroxetine which does. The effect of serotonin (5-HT2A) receptor blockade on the evolution of SWA during treatment was also investigated. In a double-blind, randomised, parallel group, 8-week study in 29 depressed patients, sleep electroencephalograms were performed at home at baseline, on night 3 and 10, and at 8 weeks of treatment with either paroxetine or nefazodone. SWA was automatically analysed and a modified DSR (mDSR) was derived, being the ratio of amount of SWA in the first 90 min of sleep to that in the second plus third 90-min periods. At baseline, the pattern of SWA over night time was similar to other reports of depressed patients. mDSR improved over the course of treatment; there was no difference between remitters and non-remitters but there was a significant drug effect and a significant drug × time effect with paroxetine patients having a much higher mDSR after treatment, regardless of clinical status. SWA and REM during antidepressant treatment appear to be interdependent and neither of them alone is likely to predict response to treatment. Higher mDSR did not predict therapeutic response. 5-HT2A blockade by nefazodone does not increase SWA above normal levels.
- Top of page
- Financial disclosure
Delta activity in deep sleep, also called slow wave activity (SWA), is generated by pace-making thalamocortical neurons, and then spreads throughout the telencephalon, especially the anterior cingulate and the orbitofrontal cortex. Although its functional significance is still debatable, this widespread and coherent activity is considered critical for the adaptation of behaviour to environmental pressures (Maquet et al., 1997). SWA appears during non-rapid eye movement (nREM) sleep and gradually increases with increasing depth of sleep; when it reaches a certain amplitude and predominates over other activity in the electroencephalogram (EEG) during a night’s sleep recording then sleep is scored visually as stage 3 or 4, slow wave sleep (SWS), using the recognised international criteria (Rechtschaffen and Kales, 1968). Specific automated methods (e.g. Fast Fourier Transform or Digital Period Analysis) can be applied to the same recording to quantify the overall power of SWA (Armitage, 1995). SWA is a marker of the homeostatic drive to sleep, thus the amount of SWA is greatest in the first sleep cycle when sleep propensity is high and gradually diminishes in subsequent cycles as sleep debt is made up and sleep drive diminished.
The total amount of SWS and the percentage of time spent in SWS over the whole of the night (% SWS) appear to be decreased in depression compared with normal controls (Benca et al., 1992), and the amplitude of slow waves is also reduced (Armitage, 1995). These reductions may be related to decreased regional cerebral blood flow seen in the orbitofrontal and anterior cingulate cortex during SWS in depression (Maquet et al., 1997) and it may be a consequence of the hypofrontality described in this condition (Kimbrell et al., 2002). In addition, these changes in SWS probably reflect the insomnia that is frequently associated with depression. Further, SWA appears to be distributed differently through the night in depression, with the patients showing a later peak than normal subjects (Kupfer and Ehlers, 1989). This is usually expressed as a lower delta sleep ratio (DSR), which is the quotient of SWA in the first to the second nREM period (cycle) of sleep, and may reflect disruption of the homeostatic sleep processes in depression.
During treatment with antidepressants, a number of changes are seen in the sleep parameters of patients. Although some of these phenomena may be side effects of the drugs, for a long time researchers have sought to identify specific indices that could predict response to treatment, help to determine the length of drug maintenance period after recovery, or even elucidate the mechanism by which these compounds exert their therapeutic effect. REM suppression had thus been suggested as a possible common underlying mechanism (Kupfer et al., 1981), but it was found that not all antidepressants share this property (Ehlers et al., 1996).
On the other hand, it has been suggested that SWA and DSR are normalised by successful antidepressant treatment (Ehlers et al., 1996). In that study, the balance of SWA was shifted towards the beginning of the night in the responders to clomipramine, in particular during the first nREM cycle of sleep, thus increasing the DSR. As this index differentiated responders from non-responders, it was hypothesised that changes in SWA rather than in REM sleep may be associated with the clinical effect of antidepressants. If this were true, DSR could prove to be the elusive laboratory marker predicting response to treatment.
However, the length of a sleep cycle is determined by the occurrence of bouts of REM sleep, and most antidepressants including clomipramine, as in the study mentioned above (Ehlers et al., 1996), delay the onset of REM, thus lengthening the first sleep cycle. Landolt et al. (2001) have shown that if REM sleep is suppressed entirely by the monoamine oxidase inhibitor, phenelzine, then familiar pattern of SWA with a large peak at the beginning of the night and a decline thereafter is maintained. Thus increasing the length of the first sleep cycle will allow more SWA to occur. A high correlation between DSR and REM latency has been observed (Kupfer et al., 1990). Therefore, in order to dissociate the predictive properties of the DSR from those of REM suppression, we set out to compare the effects of two different antidepressants on SWA, DSR, and clinical response: one that suppresses REM (paroxetine) and one that does not (nefazodone) (Sharpley et al., 1996).
Paroxetine is a potent selective inhibitor of the re-uptake of serotonin (5-HT) while nefazodone is a weak one (Tatsumi et al., 1997). The selective serotonin re-uptake inhibitors (SSRIs), like paroxetine, suppress REM and decrease sleep continuity (Wilson et al., 2000). On the other hand, nefazodone may have some advantage over the SSRIs early in treatment, in that it improves sleep, even before the onset of antidepressant action (Hicks et al., 2002; Rush et al., 1998). It is thought that the main therapeutic action of nefazodone is via blockade of the postsynaptic 5-HT2 receptor (Cusack et al., 1994). These postsynaptic receptors have also been implicated in the control of SWS (Sharpley and Cowen, 1995). Ritanserin, a selective 5-HT2 receptor antagonist, was found to increase slow wave sleep SWS in a dose-dependent manner in healthy volunteers (Idzikowski et al., 1991), and meta-chlorophenylpiperazine (mCPP), an agonist at 5-HT2 receptors, to decrease it in a dose-dependent manner (Katsuda et al., 1993). However, no increase in SWS has been reported with nefazodone in depressed patients so far (Hicks et al., 2002; Rush et al., 1998). This may be because nefazodone is more selective for 5-HT2A receptors (Davis et al., 1997), while ritanserin is a mixed 5-HT2A/2C antagonist and mCPP acts predominantly on 5-HT2C receptors (Fiorella et al., 1995). Alternatively, the effect of nefazodone on SWA may have not been apparent with standard staging procedure. Therefore, we decided to use computerised analysis to examine the subtle evolution of SWA over the course of the night.
We examined the above by using data from a study of effects on sleep architecture of the two aforementioned antidepressants in depressed outpatients (Hicks et al., 2002), powered to detect differences in sleep efficiency. We found that nefazodone significantly increased objective sleep efficiency and total sleep time and improved subjective sleep on days 3 and 10. Paroxetine decreased sleep efficiency early in treatment but this effect had resolved at week 8; however, the number of awakenings and stage 1 sleep remained significantly higher at week 8 with paroxetine. Paroxetine but not nefazodone produced marked suppression of REM sleep. The sample studied here is small for analysis of SWA, nevertheless this kind of data is difficult to obtain and therefore we thought it worth reporting. Although nefazodone was withdrawn from the market in 2003 in several countries, there are other 5-HT blocking agents that are still being used in depression (e.g. trazodone) and there are more in the pipeline, so this study may be of interest for the future.
- Top of page
- Financial disclosure
This was a double-blind, randomised, parallel group, 8-week study in patients with moderate to severe depression without psychotic features. The local Ethics Committee approved the study, and all subjects gave written informed consent. The objective of the main trial was to compare the effects of nefazodone and paroxetine on sleep and mood. Forty (40) depressed patients (Hamilton Depression Rating Scale – HAMD ≥ 18) were randomised to take paroxetine 20–40 mg per day or nefazodone 400–600 mg per day over a period of 8 weeks. Sleep EEGs were performed on the patients at home, using the Medilog 9200 (Oxford Instruments Medical, Old Woking, UK) ambulatory monitoring system, at baseline, on night 3 and 10, and at 8 weeks of antidepressant treatment. Assessments of clinical change were performed at regular intervals including on the days preceding the sleep recordings. Patients were classified as remitters if their HAMD score was ≤ 8 at week 8. All patients were medication-free for at least 2 weeks before the study (5 weeks if on fluoxetine) and more clinical information is given in Table 1. More details of the methodology concerning the patient population, the study design and medication, the assessments and the treatment outcome are given elsewhere (Hicks et al., 2002).
Table 1. Baseline clinical characteristics
|Age||45 ± 9||45 ± 10|
|Length of current episode (weeks, mean ± SD)||60 ± 72||48 ± 66|
|Previous episodes (number of patients)||8||6|
|Previous antidepressant (number of patients)||13||10|
|HAMD score at baseline||21 ± 3||22 ± 3|
|Remitters at 8 weeks (number of patients)||7||8|
The Medilog system uses the following algorithm to detect SWA. First, the raw EEG signal is filtered using a 4-pole band-pass filter set for 0.5–2.5 Hz: slow waves are detected if their amplitude is greater than 75 μV peak to peak. Baseline crossings are detected, and a value output for each 2-s mini-epoch during the night. This effectively is a period analysis technique. After computer analysis, these slow wave values are summed over each 30-s epoch, giving an index, which represents the percentage of each epoch occupied by slow (delta) waves between 0.5 and 2.5 Hz and greater than 75 μV in amplitude. The index was manually set to 0 for each 30-s epoch scored as waking or REM in the visual analysis, and all epochs containing movement or interference artefacts were identified visually and also removed from the analysis. Although this period analysis method is different from the usual spectral analysis methods used by others, it gives more information than traditional sleep staging, in that it allows a continuous measure of all high amplitude slow waves occurring in all nREM sleep, which follows a very similar pattern to that reported using other methods (see Fig. 1). However, it does not account for any waves lower than 75 μV in amplitude.
Figure 1. Evolution of slow wave activity (SWA) over the night during treatment with paroxetine. SWA summed for each 5-min period and mean ± SEM shown for all patients, mean only for remitters and non-remitters.
Download figure to PowerPoint
Of the 36 patients that completed the main study, this SWA Index (SWAI) was analysed in 29 patients (16 on nefazodone and 13 on paroxetine) whose sleep recordings provided technically valid data for nights 0, 3 and 10. Data were also available for 22 of the 29 patients at week 8, thus giving a total of 109 nights studied.
Because of the marked REM-delaying properties of paroxetine, the first sleep cycle after treatment with this drug occupied almost the whole night in the early treatment recordings. Therefore, DSR following the strict definition given in the introduction could not be derived; instead a modified DSR (mDSR) was calculated. We used the ratio of the SWAI total for the first 90 min of sleep (starting from the onset of stage 2 sleep) to the SWAI total for the second and third 90-min periods (91–270 min). We felt that this ratio approximated the DSR as well as possible in this context. mDSR was compared at baseline between drug groups and at subsequent time points both between drug groups and between remitters and non-remitters using Student’s t-test. Analysis of variance was performed over the four assessment periods with time as the within-subjects effect and drug, remission, drug × time and drug × remission as between-subjects effects, and covariates of age and gender. As SWA was strongly correlated with age in previous reports, we also performed a correlation between mDSR and age in our subjects at baseline to exclude this potential confounding factor.
- Top of page
- Financial disclosure
Figs 1 and 2 show the evolution of SWAI over the night. Simple linear regression lines have been superimposed to aid comparison. At baseline (Table 2) the pattern of SWAI over the night was similar to other reports of depressed patients (Ehlers et al., 1996). During treatment the two drug groups were very different, with a much more ‘normal’ looking pattern of clear decline over the night in the paroxetine group, and rather unclear cycles and a shallower slope in the nefazodone group.
Figure 2. Evolution of slow wave activity (SWA) over night during treatment with nefazodone. SWA summed for each 5-min period and mean ± SEM shown for all patients, mean only for remitters and non-remitters.
Download figure to PowerPoint
Table 2. Sleep architecture following Rechtschaffen and Kales (1968) scoring
|Variable||Drug||Baseline||Day 3||Day 10||Week 8|
|Total sleep time (min)||n||377 ± 45||396 ± 62||385 ± 64||385 ± 50|
|P||366 ± 69||354 ± 81||352 ± 52||403 ± 67|
|Wake after sleep onset (min)||n||60 ± 25||37 ± 45||42 ± 41||57 ± 64|
|P||62 ± 59||61 ± 32||75 ± 60||55 ± 24|
|Stage 1 (min)||n||39 ± 44||39 ± 23||27 ± 19||25 ± 17|
|P||62 ± 42||48 ± 31||48 ± 28||60 ± 37|
|Stage 2 (min)||n||182 ± 59||201 ± 52||174 ± 44||192 ± 58|
|P||177 ± 56||206 ± 54||190 ± 48||185 ± 47|
|SWS (min)||n||55 ± 38||52 ± 37||58 ± 50||74 ± 46|
|P||62 ± 44||65 ± 53||63 ± 47||67 ± 50|
|REM (min)||n||87 ± 31||101 ± 35||108 ± 40||95 ± 28|
|P||83 ± 34||36 ± 42||46 ± 26||93 ± 33|
|Sleep onset latency (min)||n||27 ± 26||30 ± 28||37 ± 48||20 ± 10|
|P||35 ± 27||65 ± 54||34 ± 23||33 ± 12|
|REM onset latency (min)||n||79 ± 39||58 ± 30||54 ± 32||62 ± 37|
|P||78 ± 53||291 ± 124||218 ± 120||180 ± 55|
|Sleep efficiency (%)||n||81 ± 9||85 ± 11||83 ± 12||83 ± 12|
|P||78 ± 12||72 ± 13||78 ± 10||82 ± 6|
|REM onset latency <60 min (number of points)||n||5||6||7||5|
Because of the REM-delaying properties of paroxetine, the first sleep cycle with this drug is very long, as predicted, and the effects of the two drugs on REM sleep for each hour of sleep are shown in Table 3 to illustrate this. The amount of waking during the night is also shown in Table 3, and the effects of increased waking after paroxetine, particularly in the third hour of sleep may be seen on the graphs in Fig. 1.
Table 3. Distribution of REM sleep and waking (mean ± SEM) through the night after nefazodone or paroxetine
|Occasion||Drug||Hour 1||Hour 2||Hour 3||Hour 4||Hour 5||Hour 6|
|REM sleep (min)|
| Baseline||Nefazodone||4 ± 1||9 ± 2||12 ± 3||16 ± 3 ||15 ± 3||16 ± 4|
|Paroxetine||4 ± 1||10 ± 2||12 ± 3||6 ± 2||20 ± 4||12 ± 3|
| Day 3||Nefazodone||5 ± 1||18 ± 4||10 ± 3||18 ± 3||12 ± 3||20 ± 4|
|Paroxetine||0||2 ± 2||0||6 ± 3||7 ± 4||6 ± 3|
| Day 10||Nefazodone||11 ± 32||16 ± 4||12 ± 3||15 ± 3||16 ± 5||20 ± 4|
|Paroxetine||0||1 ± 1||4 ± 2||6 ± 3||13 ± 4||13 ± 4|
| Week 8||Nefazodone||6 ± 2||15 ± 5||16 ± 4||18 ± 4||9 ± 3||21 ± 5|
|Paroxetine||0||2 ± 2||9 ± 4||16 ± 5||13 ± 3||10 ± 4|
| Baseline||Nefazodone||6 ± 3||11 ± 4||6 ± 2||7 ± 4 ||7 ± 3||10 ± 4|
|Paroxetine||9 ± 4||3 ± 1||6 ± 2||11 ± 5||9 ± 4||15 ± 5|
| Day 3||Nefazodone||6 ± 3||4 ± 1||2 ± 1||5 ± 1||5 ± 2||9 ± 5|
|Paroxetine||1 ± 1||7 ± 2||11 ± 5||5 ± 1||11 ± 5||12 ± 5|
| Day 10||Nefazodone||4 ± 1||2 ± 1||3 ± 1||10 ± 5||9 ± 4||9 ± 3|
|Paroxetine||3 ± 1||7 ± 3||19 ± 6||9 ± 3||6 ± 1||12 ± 4|
| Week 8||Nefazodone||1 ± 1||4 ± 1||2 ± 1||8 ± 5||10 ± 5||12 ± 6|
|Paroxetine||5 ± 2||9 ± 2||8 ± 2||8 ± 1||9 ± 4||10 ± 5|
Modified DSR was not correlated with age at baseline. Fig. 3 illustrates the mDSR at baseline and during treatment for remitters versus non-remitters as well as irrespective of clinical outcome, for each drug group separately. There was no significant baseline difference between the remitters and non-remitters (mean ± SEM: remitters = 0.80 ± 0.12, non-remitters = 0.73 ± 0.13), between the two drug groups (nefazodone = 0.71 ± 0.11, paroxetine = 0.84 ± 0.14), or within each drug group for remission. Taking all patients together, the mDSR improved over the course of treatment (anova with time as within-subjects effect: F = 3.98, df: 3.72, P = 0.011) but there was no difference between remitters and non-remitters, and when age was introduced as a covariate there was no significant overall time effect. However, when data were analysed per drug group, with age and gender as covariates, there was a significant drug effect (F = 12.0, df: 1.24, P = 0.002) and a significant drug × time effect (F = 3.7, df: 3.72, P = 0.015), with paroxetine patients having a much higher mDSR after treatment. There was no significant drug × remission effect, indicating that the paroxetine group had increased mDSR regardless of therapeutic effect.
Figure 3. Delta sleep ratio (DSR) during treatment. Modified DSR is defined as the ratio between total slow wave activity (SWA) during the first 90 min after sleep onset and total SWA during 91–270 min.
Download figure to PowerPoint
Paroxetine was associated with a decrease in total sleep time (Hicks et al., 2002), and there was a possibility that a greater mDSR in the paroxetine arm might simply reflect greater homeostatic sleep drive in these patients during therapy, brought on by increased 24-hour wakefulness. Therefore we tested to see if there was a correlation between change in mDSR and change in total sleep time at day 10 (Spearman’s rank correlation), and there was none (r = −0.019, P = 0.92, n = 29).
In order to explore whether mDSR might be a useful early predictor of antidepressant effectiveness, we carried out individual Spearman’s rank correlations between mDSR (at baseline and day 3) and illness state at the four time points. There was no correlation between mDSR (baseline or day 3) and change in HAMD rating of depression at any of the time points, therefore no indication of a predictive effect of mDSR. We also performed similar correlations between mDSR (baseline and day 3) and sleep architecture parameters (REM onset latency, total SWS – SWS and wake after sleep onset), none of which was significant. We also calculated the change in total SWA in relation to remission in the two drug groups. Numbers were small but the results are shown in Fig. 4, and these may indicate different effects in remitters in the paroxetine group.
Figure 4. Change in total slow wave activity (SWA) during treatment in remitters and non-remitters. This is obtained by summing SWA in all epochs and subtracting the value from the baseline night.
Download figure to PowerPoint
- Top of page
- Financial disclosure
In our group of depressed patients we found that the mDSR changed significantly with treatment, the balance of SWA shifting towards the beginning of the night. This change was not associated with clinical improvement, but it was associated with the type of medication used. While mDSR increased in patients treated with paroxetine, it was largely unaffected in those treated with nefazodone, despite the fact that there was no drug group difference in the clinical outcome. This increase of mDSR seen with paroxetine in our study cannot be an artefact of the increased length of the first sleep cycle caused by SSRI-induced REM suppression, as we compared the first 90 min of sleep on each occasion. The lack of association with clinical outcome contrasts with the previous report (Ehlers et al., 1996), where DSR changes predicted response or non-response to clomipramine. The mDSR was higher in remitted paroxetine patients than non-remitters only on night 3, but the difference did not reach significance (see Fig. 3), perhaps because there were only five non-remitters in the paroxetine group. It is possible that a bigger sample would have uncovered an existing difference. However, this appears very unlikely in the nefazodone group.
Further, our findings in the paroxetine group, which had marked but not total REM suppression, are in accordance with the results of the study by Landolt et al. (2001) where phenelzine completely suppressed REM but did not alter total SWA or the exponential decline of SWA during the night. Patients in the phenelzine study had all responded to treatment, but not all our patients did. There is a hint that paroxetine but not nefazodone treatment may have increased SWAI in the remitters.
The EEG sleep profiles of depressed patients differ during the course of the illness, being more abnormal early on and normalising later (Dew et al., 1996). Our group had a mean duration of illness longer than 4 months. Unfortunately, Ehlers et al. (1996) do not provide such information, so we cannot establish whether a substantial difference between the two patient populations in this area may have accounted for the discrepancy in the results. Of note though, in our sample there was no correlation between baseline mDSR and length of episode.
The importance of the abnormal distribution of the SWA in depression and the resulting reduction in DSR may lie in areas other than prediction of response to treatment. Kupfer et al. (1990) showed that it was a robust predictor of relapse, following the discontinuation of imipramine from a group of patients who had been treated successfully with a combination of this drug and psychotherapy. Similarly, patients who were successfully treated with this combination and then continued on the medication alone were more likely to relapse if they had lower amounts of delta sleep (Kupfer et al., 1993). Reduced delta sleep (but not DSR) predicted recurrence in patients treated with interpersonal psychotherapy (Buysse et al., 1997). Higher baseline DSR had good prognostic value for psychotherapy (Buysse et al., 1992; Kupfer et al., 1990; Thase et al., 1998) and sleep deprivation (Nissen et al., 2001).
However, one study showed that following treatment with cognitive behaviour therapy, decreased DSR and decreased SWA remained stable into remission, thus indicating a trait marker for depression (Thase et al., 1998). Further, family studies (Giles et al., 1989; Lauer et al., 1995; Linkowski et al., 1991) suggest that these SWA abnormalities that were observed post-treatment might be under genetic influence rather than a biological scar resulting from the depressive episode per se.
Studies in healthy volunteers (Sharpley et al., 1996) and depressed patients (Hicks et al., 2002; Rush et al., 1998) show that nefazodone improves sleep maintenance. Given that this is often compromised in depression, its promotion is clinically significant even if nefazodone is not a sedative drug. SWA, as a complex phenomenon, is affected by various inputs, and 5-HT2 receptors are very likely to be involved in this process (Landolt et al., 1999). Antagonism of these receptors by compounds like nefazodone might restore the maintenance of sleep without necessarily enhancing SWA beyond normal levels.
The main limitation of this study is that it was not powered to detect changes in SWA in relation to clinical response; in particular, the number of available remitted and non-remitted patients with spectral analysis data in the paroxetine group was small. In addition, there were marked differences in sleep between our two drug groups at baseline, which occurred entirely by chance as allocation was random, and are unexplained by other known factors. Finally, our modification of the DSR makes direct comparison with other studies more difficult.
In summary, in our depressed group, REM and early SWA seem to be interdependent; in the paroxetine group, as REM latency increased to 3–4 h, more SWA accumulated into the first 90 min of sleep and mDSR increased, but the sample was too small to detect if this corresponded to clinical improvement. It may be true for some antidepressants, especially those that have their main action through blocking the re-uptake of 5-HT (SSRIs and clomipramine), that both an increase in DSR and a decrease in REM are necessary for their therapeutic efficacy.