Antidepressants and their effect on sleep
Andrew G. Mayers1,2 & David S. Baldwin1
1. University of Southampton; 2. Southampton Solent University
Perinatal Mental Health
The Lodge, Tatchbury Mount
Calmore, Southampton, SO40 2RZ
Tel: +44 (0)23 8087 4330
Fax: +44 (0)23 8087 4360
Given the relationship between sleep and depression, there is inevitably going to be an effect of antidepressants on
sleep. Current evidence suggests that this effect depends on the class of antidepressant used and the dosage. The extent
of variation between the effects of antidepressants and sleep may relate to their mechanism of action. This systematic
review examines randomised-controlled trials (RCTs) that have reported the effect that antidepressants appear to have
on sleep. RCTs are not restricted to depressed populations, since several studies provide useful information about the
effects on sleep in other groups. Nevertheless, the distinction is made between those studies, because the participant’s
health may influence the baseline sleep profiles and the effect of the antidepressant. Insomnia is often seen with
monoamine oxidase inhibitors (MAOIs), with all tricyclic antidepressants (TCAs) except amitriptyline, and all selective
serotonin reuptake inhibitors (SSRIs), as well with venlafaxine and moclobemide. Sedation has been reported with all
TCAs except desipramine, with mirtazapine and nefazodone, the TCA-related maprotiline, trazodone and mianserin,
and with all MAOIs. REM sleep suppression has been observed with all TCAs except trimipramine, but especially
clomipramine, with all MAOIs and SSRIs and with venlafaxine, trazodone and bupropion. However, the effect on sleep
varies between compounds within antidepressant classes, differences relating to the amount of sedative or alerting
(insomnia) effects, changes to baseline sleep parameters, differences relating to REM sleep, and the degree of sleep-
related side effects.
Key words: Antidepressants, sleep, review, randomised-controlled trials
The review exercise was undertaken by exploring the Ovid® database, searching the CINAHL
(1982 - May 2005), EMBASE (1980 – May 2005), Ovid MEDLINE® (1966 – May 2005) and
PsychINFO (1985 – May 2005). A search strategy was undertaken to improve the likelihood of
including high quality randomised controlled-trials (RCTs) that used a double-blind randomisation
of participants into groups of at least 5 (per group), included in a baseline and follow-up
examination of the effect of antidepressants on sleep, where those antidepressants were compared to
placebo (placebo-controlled trials) and/or to other antidepressants (comparator trials). Papers were
selected regardless of the nature of the participants. Antidepressant effects on sleep may vary with
the current health of the participant and it is important to make that distinction. Careful
consideration is also paid to the dose of antidepressant as that may explain some of the variation
between studies in similar participant groups. A more general overview is also presented on the
mechanisms of action of differing classes of antidepressants that might explain the effect they
appear to have on sleep.
Following exclusions, 120 papers were examined, 53 of which included placebo. Those papers are
presented in Table 1. The following section presents general findings for each antidepressant class,
and indicates the mechanisms that might be responsible for those effects. Within each class some of
the more specific findings for each antidepressant are examined. Rather than duplicate the data from
Table 1, only the most important aspects are described.
Several mechanisms are important in the effects of antidepressant treatment on sleep. Increases in
the availability of serotonin and noradrenaline appear to be associated with the suppression of REM
sleep, but also with increases in sleep fragmentation (Wilson and Argyropoulos, 2005). The
pathways responsible for these actions vary across antidepressant class and with individual
medications, but generally refer to action on pre-synaptic autoreceptors, post-synaptic 5HT receptor
sites (such as the 5-HT1A and 5-HT2 receptors), α1- and α2-adrenoceptors and histamine H1
receptors. 5-HT1A stimulation may be associated with REM sleep suppression; 5-HT2 agonism may
be related to sleep disturbance. Inhibition of α2-adrenoceptors autoreceptors increases availability of
noradrenaline, and therefore may be associated with fragmentation of sleep. Blockade of the other
receptor sites (α1-adrenoceptors and histamine H1) may facilitate sleep promotion (Wilson and
Tricyclic Antidepressants (TCAs)
There is much variation between TCAs in the effect on sleep architecture, and with regard to
sedating and alerting properties. The British Association for Psychopharmacology (BAP) guidelines
(Anderson et al., 2000) suggest that sedation is ‘relatively common or strong’ with amitriptyline,
dothiepin and clomipramine, while this ‘may occur or is moderately strong’ with imipramine,
desipramine and nortriptyline. Sedation may be useful in depressed patients with insomnia, but
might not be welcome in those patients wishing to avoid daytime sleepiness.
The mechanisms thought to be responsible for sleep effects in TCAs vary with specific compounds.
Most TCAs inhibit the reuptake of both serotonin and noradrenaline, but the relative extent that they
do this varies, and may explain some of the differences in sedation and REM sleep suppression. All
TCAs except lofepramine block histamine H1 receptors, and all but desipramine block α1-
adrenoceptors. The blockade of histamine H1 receptors may be related to sleep promotion (Haas and
Panula, 2003), but the evidence for an effect on REM sleep or SWS is weak (Wilson and
Argyropoulos, 2005). Antagonism of α1-adrenoceptors is more likely to explain the sedative
properties of TCAs, as might the 5-HT2 blockade action, as seen with amitriptyline and
trimipramine (which are particularly associated with sedation).
(Staner et al., 1995) found that amitriptyline (150mg) produced more alerting effects than
paroxetine (30mg). (Kerkhofs et al., 1990) demonstrated that amitriptyline (150mg) and fluoxetine
(60mg) both produced significant REM sleep suppression. (Casper et al., 1994) showed that
patients presented better improvement in early morning awakening, and nocturnal wakings with
amitriptyline (100-150mg) than imipramine (100-150mg); although this was only for those who had
responded to treatment. (Kerr et al., 1993) observed that amitriptyline (75mg) was associated with
significantly shorter sleep latency, but more drowsiness, than fluoxetine (20mg) on the Line
Analogue Rating Scale for Sedation (LARS) scale. However (De Ronchi et al., 1998) found no
between-group differences for patients in respect of Leeds Sleep Evaluation (LSEQ) scores between
amitriptyline (50-100mg) and fluoxetine (20mg).
Other patient groups
(Mertz et al., 1998) found that amitriptyline (50mg) reduced REM sleep in gastroenterology
patients, compared to placebo, while (Carette et al., 1995) demonstrated fewer changes in REM
sleep parameters in fibromyalgia patients (dosage, 25mg). This is just one example where the dose
may be a significant factor in contrasting findings. For fibromyalgia patients (Hannonen et al.,
1998), subjective sleep ratings were significantly improved from baseline with amitriptyline (25-
37.5mg), compared to placebo. In a study of cancer patients with neuropathic pain (Mercadante et
al., 2002), it was found that drowsiness was significantly higher with amitriptyline (25-30mg) than
placebo. In another study (Mertz et al., 1998), amitriptyline (50mg) was associated with poorer
sleep efficiency for patients with functional dyspepsia, compared to placebo. In a study of patients
with chronic pain (Versiani et al., 1999), amitriptyline (50-250mg) was associated with better
improvements in Hamilton Rating Scale for Depression (HAMD; (Hamilton, 1960)) sleep scores
than fluoxetine (20mg), although daytime drowsiness was a significantly greater problem with
(Rosenzweig et al., 1998) found that subject-rated alertness and behaviour upon waking was
significantly poorer with amitriptyline (50mg) than placebo. This hangover effect was confirmed by
(Hindmarch et al., 2000) who demonstrated that sedation and trouble waking were significantly
worse for amitriptyline (50mg), compared to placebo.
Clomipramine may be associated with sedation, but has also been linked with insomnia (Anderson
et al., 2000). While most TCAs suppress REM sleep to some extent, clomipramine appears to be
the most marked in this respect (Winokur et al., 2001). Clomipramine is associated with the most
potent serotonin reuptake inhibition of all the TCAs (Wilson and Argyropoulos, 2005).
(Lepine et al., 2000) demonstrated no differences between clomipramine (50-150mg) and sertraline
(50-200mg) on LSEQ and HAMD sleep scores, but both showed significant improvements on all
four LSEQ factors (Ease of getting to sleep (EGS); perceived quality of sleep (QOS); ease of
awakening (EOA); and behaviour following wakefulness (BFW)).
(Lacey et al., 1977) found that clomipramine (25-75mg) was associated with slightly longer
nocturnal awakenings than placebo, and almost completely suppressed REM sleep.
(Sonntag et al., 1996) demonstrated that imipramine (50-200mg) significantly increased sleep
latency, while trimipramine (50-250mg) was associated with a non-significant decrease; imipramine
was associated with significantly less total sleep time, and significantly more nocturnal awakenings
than trimipramine. (Volkers et al., 2002) found that imipramine (mean dose 220mg) was associated
with significantly more nocturnal restlessness than fluvoxamine (mean 201mg).
Other patient groups
In a study of patients reporting panic disorder or agoraphobia, (Cassano et al., 1994) imipramine
(25-250mg) was associated with more sedation than placebo (although less than alprazolam; 1-
10mg), but significantly more insomnia than placebo and alprazolam. (Sonntag et al., 1996) found
that imipramine (50-200mg) was associated with decreased total sleep time, while this was
increased with trimipramine (50-250mg); sleep efficiency was significantly more improved with
trimipramine but wakings were significantly more frequent with imipramine.
(Wolf et al., 2001) showed that trimipramine (150mg) was associated with improved sleep
efficiency, longer sleep, and fewer nocturnal arousals, compared to fluoxetine (20mg).
Other patient groups
(Riemann et al., 2002) found that trimipramine (mean 100mg) was not associated with REM sleep
suppression, when compared to placebo with insomnia patients. Unlike other TCAs, which are
associated with REM suppression, trimipramine is not associated with the reuptake inhibition of
serotonin (Wilson and Argyropoulos, 2005).
(Kupfer et al., 1991) demonstrated that desipramine (100-200mg) significantly reduced sleep
latency after just one day of treatment, but this significantly increased again within a week and
throughout the remainder of the 4-week study. Desipramine was associated with shorter sleep
latency than fluvoxamine (200mg), and presented better sleep efficiency. In another study (Shipley
et al., 1985), desipramine (50-250mg) was associated with more nocturnal waking, shorter sleep,
and less efficient sleep than amitriptyline (50-150mg). Unlike other TCAs, desipramine is not
associated with α1-adrenoceptor blockade (Wilson and Argyropoulos, 2005), which may explain
why it does not promote sleep as well. It is also associated with less serotonin reuptake inhibition
than most other TCAs.
(Reynolds, III et al., 1997) demonstrated that nortriptyline (80-120mg) was associated with longer
sleep latency than placebo. Nortriptyline also showed initial suppression of REM sleep, with
prolonged REM latency and reduced REM proportion, but this rebounded in later REM periods to
show greater REM production and density than placebo.
Other patient groups
In a study of patients with skin complaints (Hammack et al., 2002), total sleep time improved for
those treated with nortriptyline (100mg), compared to placebo. However, daytime sleepiness was
reported as a problem in the treatment group.
(Stephenson et al., 2000) demonstrated that drowsiness side effects were more common with
dothiepin (150mg) than fluoxetine (20mg). (Ferguson et al., 1994) found that HAMD sleep scores
were significantly reduced with dothiepin (150mg), compared to placebo (but were similar to
doxepin). (Blacker et al., 1988) showed that dothiepin (75-150mg) was associated with more
immediate improvement of EGS and QOS perceptions on LSEQ than amitriptyline (75-100mg) or
mianserin (30-75mg), although was similar to trazodone (150mg). LSEQ perceptions of BFW were
poor during the first week for all the comparator compounds, but improved thereafter.
(Ramaekers et al., 1995) found that dothiepin (75-150mg) was associated with increased trouble in
waking, but longer total sleep time than placebo. (Wilson et al., 2002) demonstrated that dothiepin
(75-150mg) was associated with poorer sleep efficiency than placebo (and fluoxetine 20mg), but
shorter nocturnal awakenings than fluoxetine; REM sleep latency was significantly shorter for
dothiepin than for fluoxetine. (Wilson et al., 2000) showed that dothiepin (100mg) was associated
with longer TST, shorter nocturnal disturbances, better sleep efficiency, and better sleep quality
than fluvoxamine (100mg).
(Ferguson et al., 1994) found that clinician-rated HAMD sleep scores were significantly reduced
with doxepin (150mg), compared to placebo, while (Feighner et al., 1986) showed that doxepin
(100-225mg) was related to significantly better improvements on these scores than bupropion (300-
Other patient groups
Sleep efficiency and sleep quality were significantly improved for insomnia patients taking doxepin
(25-50mg), compared to placebo (Hajak et al., 2001), while doxepin (25mg) was associated with
significantly increased total sleep time, and significantly reduced sleep latency and length of
nocturnal awakenings, compared to placebo with insomnia patients and healthy volunteers (Hajak et
Monoamine oxidase inhibitors (MAOIs)
MAOIs have been associated with increased sleep latency, poorer sleep efficiency, and increased
nocturnal disturbances (Winokur et al., 2001). Insomnia has been reported for phenelzine,
tranylcypromine and isocarboxazid (Anderson et al., 2000), while significant REM sleep
suppression has been noted with phenelzine and tranylcypromine (Winokur et al., 2001). However,
REM rebound is noted subsequent to the withdrawal of medication (Kupfer and Bowers Jr, 1972).
There is a paucity of RCTs with MAOIs. Moclobemide, a reversible MAOI, has been associated
with less REM sleep suppression than traditional MAOIs (Winokur et al., 2001). Sedation is not
reported with moclobemide, although minor insomnia has been noted (Anderson et al., 2000).
MAOIs increase the availability of monoamines, but REM suppression often appears later than with
TCAs and SSRIs (Wyatt et al., 1971).
(Nolen et al., 1993) found that tranylcypromine (20-100mg) significantly increased REM sleep
latency and almost completely suppressed REM sleep overall. Sleep latency was also increased, but
patients reported deeper and more refreshed sleep than with brofaromine (50-250mg).
(Giller et al., 1982) demonstrated that isocarboxazid (20mg) did not differ from placebo on HAMD
sleep scores, but treatment responders tended to sleep better overall with isocarboxazid than with
(Sogaard et al., 1999) found that moclobemide (300-450mg) was associated with poorer BFW
scores on LSEQ than sertraline, while sleep was observed to better with moclobemide (450mg) than
with toloxatone (100mg; (Lemoine and Mirabaud, 1992)).
Other patient groups
(Hannonen et al., 1998) demonstrated that moclobemide (450-600mg) was associated with poorer
subjective sleep satisfaction and fatigue (not assessed with a specific scale) than amitriptyline (25-
37.5mg) in patients with fibromyalgia.
Two trials involving moclobemide with healthy participants ((Dingemanse et al., 1992), 450mg;
(Ramaekers et al., 1992), 200mg) suggest that moclobemide has no effect on sleep, when compared
to placebo or other antidepressants.
Selective serotonin reuptake inhibitors (SSRIs)
SSRIs are frequently associated with insomnia (Anderson et al., 2000); around one-quarter of
depressed patients in clinical trials report insomnia (Winokur et al., 2001). Less well documented is
that SSRIs may cause daytime somnolence, particularly at higher doses (Beasley Jr et al., 1992).
EEG studies of sleep confirm that SSRIs immediately suppress REM sleep, and continue to do so
throughout treatment; REM parameters return to normal once the SSRI is discontinued (Winokur et
The observed effects on sleep of SSRIs are thought to be due to the effects of increased levels of on
5-HT1A and 5-HT2 receptors. Activation of 5-HT1A receptors is probably responsible for REM
suppression (Gillin et al., 1994), but is unlikely to mediate sleep fragmentation. This is more likely
to be due to stimulation of 5-HT2 receptors (Lawlor et al., 1991). By definition, SSRIs block
serotonin reuptake, but some also block noradrenaline reuptake. Both actions have been associated
with REM suppression and sleep disruption (Wilson and Argyropoulos, 2005).
(Mendels et al., 1999) found that citalopram (20-80mg) was associated with significant
improvements in HAMD sleep scores, relative to placebo; although daytime sleepiness was a
significantly greater problem for those taking citalopram than for placebo. (Rosenberg et al., 1994)
demonstrated that citalopram (10-60mg) was associated with significantly better HAMD sleep
scores (from baseline), but did not differ from imipramine (50-100mg). (Leinonen et al., 1999)
showed that subjective ratings for all LSEQ factors significantly improved with citalopram (20-
60mg), although not as quickly as with mirtazapine (15-60mg).
Escitalopram is a relatively new antidepressant in the SSRI class. It has been developed from one of
the isomers of citalopram, so whilst chemically identical, it may be more beneficial than citalopram
if the efficacy elements reside in that single isomer; it may also possess less side effects than the
original combination. There are currently no RCTs that specifically examine escitalopram in
placebo or comparator trials. In a recent pooled analysis (Lader et al., 2005), which compares data
from RCTs involving citalopram and escitalopram, it was shown that escitalopram (10-20mg)
showed significantly better improvements on the Montgomery-Asberg Depression Rating Scale
(MADRS; Montgomery and Åsberg, 1979) item 4 (sleep) at all time points (weeks 1, 4, 6 & 8);
citalopram (20-40mg) was only significantly better at week 6. The proportion of patients with sleep
problems (at baseline MADRS item 4 ≥ 4) improving by endpoint (MADRS item 4 ≤ 1) was
significantly higher with escitalopram than citalopram. However, prospective RCTs specifically
examining sleep are required.
(Jindal et al., 2003) found that sertraline (mean 142mg) suppressed REM sleep and increased sleep
latency (although not significantly), compared to placebo. (Lepine et al., 2000) showed that
sertraline (50-200mg) and clomipramine (50-150mg) significantly improved LSEQ (all factors) and
HAMD sleep scores, but there were no between-group differences. (Bennie et al., 1995)
demonstrated that sertraline (50-100mg) was associated with fewer reports of trouble in sleep
initiation than fluoxetine (20-40mg), but with poorer perceptions on waking. Although overall
LSEQ scores were significantly improved for both groups, they differed on individual items:
sertraline showed better EGS scores than fluoxetine, but poorer EOA and BFW.
(Paul et al., 2002) found that sertraline (50-150mg) was associated with significantly more
insomnia than with placebo.
(Rush et al., 1998) found that sleep was significantly less efficient, and nocturnal awakenings were
significantly greater, with fluoxetine (20-40mg) when compared to nefazodone (100-500mg).
Fluoxetine significantly suppressed REM sleep, while nefazodone significantly increased the time
spent in REM sleep. (Wolf et al., 2001) demonstrated that fluoxetine (20mg) was associated with
less efficient, shorter and more disrupted sleep than trimipramine (150mg); fluoxetine suppressed
REM sleep, whereas trimipramine did not. (Satterlee and Faries, 1995) showed that HAMD sleep
scores tended to show better improvement for fluoxetine (20mg) than placebo, but this was not
significant. (Winokur et al., 2003) found no differences between fluoxetine (20-40mg) and
mirtazapine (15-45mg) in respect of HAMD sleep scores; both showing significant improvements.
However, improvements in sleep latency and total sleep time were not as marked for fluoxetine as
they were for mirtazapine, which resulted in more efficient sleep and less nocturnal disturbances
Other patient groups
(Wolfe et al., 1994) found that self-reported sleep quality perceptions were significantly better with
fluoxetine (20mg) than placebo for patients with fibromyalgia.
(Vasar et al., 1994) demonstrated that fluoxetine (20mg) increased REM sleep latency and reduced
overall REM proportion, increased sleep stages 2 and 3, increased sleep latency and worsened sleep
efficiency, compared to placebo.
(Volkers et al., 2002) found that fluvoxamine (mean 201mg) was associated with more fragmented
sleep than imipramine (mean 220mg), while (Kupfer et al., 1991) demonstrated greater sleep
disruption for fluvoxamine (200mg) than desipramine (100-200mg). (Perez and Ashford, 1990)
showed that fluvoxamine (100-300mg) was associated with poorer EGS ratings on the LSEQ than
mianserin (60-180mg) but fluvoxamine was related to better BFW ratings. While fluvoxamine
(100mg) and fluoxetine (20mg) did not differ in their effect on sleep in the first month of treatment,
after that HAMD sleep scores were significantly better for fluvoxamine (Dalery and Honig, 2003).
(Silvestri et al., 2001) found that fluvoxamine (100mg) was less disruptive to sleep than paroxetine
(20mg), but tended to be associated with greater REM sleep suppression. (Wilson et al., 2000)
demonstrated that fluvoxamine (100mg) was associated with shorter and more disrupted sleep than
with dothiepin (100mg) or placebo. Although poorer subjective sleep quality was reported for
fluvoxamine than dothiepin, perceptions upon waking were better.
(Dunbar et al., 1993) found that HAMD sleep scores were significantly more improved with
paroxetine (10-50mg) than placebo. (Staner et al., 1995) showed that paroxetine (30mg) was more
alerting than amitriptyline (150mg). Sleep quality was rated significantly more poorly with higher
doses of paroxetine (40mg vs. 20mg) than with amitriptyline (75mg) or placebo (Robbe and
O'Hanlon, 1995). (Schatzberg et al., 2002) demonstrated that HAMD sleep scores were poorer with
paroxetine (20-40mg) than mirtazapine (15-45mg). (Hicks et al., 2002) found that sleep time was
less, and disruption greater, for paroxetine (20-40mg) compared to nefazodone (400-600mg). REM
sleep was shown to be significantly more suppressed with paroxetine than nefazodone, and
subjective sleep ratings showed greater improvements with nefazodone.