does trazodone work for sleep — what the evidence actually shows

a 50 mg tablet at bedtime, prescribed by a doctor who did not want to give you a sleeping pill. it is one of the most prescribed sleep aids in the country and it has never been approved for sleep.

here is what the record holds. what the label approves it for, what the seven placebo-controlled trials found when they were pooled, what the cochrane review made of the same trials, what the sleep society's guideline says and how strongly, what the drug does to the next morning, and the side-effect numbers with the caveat the label itself attaches to them. ten sources, funding printed where the record states it.

approved for depression, prescribed for sleep

every trazodone label in the united states lists one indication: “the treatment of major depressive disorder”, at a starting dose of 150 mg a day and a maximum of 400 mg^[9]. it carries the antidepressant boxed warning that these drugs “increased the risk of suicidal thoughts and behaviors in pediatric and young adult patients”^[9]. there is no insomnia indication and no insomnia data on the label. the sleep use is off-label, and it is the majority use: by 2017 a systematic review could open with the sentence that trazodone's “off-label use of this medication for insomnia has surpassed its usage as an antidepressant”^[7]. a 2023 appraisal calls it “one of the most commonly used prescription medications for insomnia” in the same breath as noting that “some recent clinical guidelines do not recommend its use”^[10].

the pooled trials: seven studies, 429 people

the 2018 meta-analysis in sleep medicine pooled every randomised placebo-controlled trial it could find: seven, with 429 people between them^[1]. on the objective primary outcome, sleep efficiency, the share of time in bed actually asleep, trazodone did nothing measurable (standardised mean difference 0.09, 95% confidence interval −0.19 to 0.38, P = 0.53)^[1]. on the other primary outcome, how people rated their own sleep, it did slightly better than placebo, and only just: standardised mean difference −0.41, interval −0.82 to −0.00, P = 0.05, with moderate heterogeneity between the trials^[1]. among the secondary outcomes, the number of awakenings fell (−0.51, −0.97 to −0.05); sleep latency, total sleep time and time awake after sleep onset did not differ from placebo^[1]. dropouts for side effects did not differ either^[1].

the authors' summary is worth reading whole: “trazodone was effective in sleep maintenance by decreasing the number of early awakenings and it could significantly improve perceived sleep quality, although there were no significant improvements in sleep efficiency or other objective measures”^[1]. that is the honest shape of the evidence. people wake less and say they slept better; the machine does not see more sleep.

what cochrane made of the same trials

the 2018 cochrane review of antidepressants for insomnia found 23 randomised trials in all, seven of them trazodone^[2]. three trazodone studies (370 people) could be pooled, “indicating a moderate improvement in subjective sleep outcomes over placebo” (standardised mean difference −0.34, −0.66 to −0.02)^[2]. two studies measured sleep with polysomnography and “found little or no difference in sleep efficiency” (1.38 percentage points, −2.87 to 5.63; 169 people), graded low quality^[2]. there was “low quality evidence from two studies of more adverse effects with trazodone than placebo (i.e. morning grogginess, increased dry mouth and thirst)”^[2].

the review's conclusion: “there may be a small improvement in sleep quality with short-term use of low-dose doxepin and trazodone compared with placebo. the tolerability and safety of antidepressants for insomnia is uncertain due to limited reporting of adverse events”, and “there was no evidence … for long-term antidepressant use for insomnia”^[2]. one of the review's authors declares research grants and advisory roles with nine drug companies; the first author declares none^[2]. neither changes the arithmetic, which is the same as the meta-analysis above from a different team.

the guideline: suggested against, weakly

the american academy of sleep medicine reviewed the trials drug by drug for its 2017 guideline. its recommendation on this one is a single sentence: “we suggest that clinicians not use trazodone as a treatment for sleep onset or sleep maintenance insomnia (versus no treatment) in adults”^[4]. the recommendation is graded WEAK, and the guideline says what that word means: a weak recommendation “reflects a lower degree of certainty in the outcome and appropriateness of the patient-care strategy for all patients, but should not be construed as an indication of ineffectiveness”^[4]. read it as the society saying the trials did not justify recommending the drug, not as the society saying the drug does nothing.

the recommendation has not settled the field. when a 2023 appraisal put the statement “trazodone should never be used as a first-line medication for insomnia” to practising clinicians and to a seven-member expert panel, most of the clinicians disagreed with it and most of the panel agreed with it, on the same published evidence^[10]. that gap, between what the trials support and what the prescriptions do, is the whole story of this drug.

the next morning

the best-controlled look at what trazodone does after you wake up is small: sixteen people with polysomnography-confirmed primary insomnia took 50 mg or placebo thirty minutes before bed for seven nights in a double-blind crossover, funded by the national institutes of health^[5]. the drug did what the pooled trials say it does, fewer awakenings, less stage-1 sleep, fewer reports of difficulty sleeping, and by night seven more slow-wave sleep and less measured daytime sleepiness^[5]. it also “produced small but significant impairments of short-term memory, verbal learning, equilibrium and arm muscle endurance across time-points”^[5]. the authors call this “a modest caveat”; a 2005 review had put it more sharply, naming “sedation, dizziness, and psychomotor impairment, which raise particular concern regarding its use in the elderly”, and calling the efficacy evidence “very limited”^[6].

the side effects, from the label, with the label's caveat

the prescribing information reports adverse reactions from the controlled depression trials in two columns, inpatients (142 on trazodone, 95 on placebo) and outpatients (157 and 158)^[9]. drowsiness: 24% against 6% in inpatients, 41% against 20% in outpatients. dizziness or light-headedness: 20% against 5%, and 28% against 15%. dry mouth: 15% against 8%, and 34% against 20%. blurred vision: 6% against 4%, and 15% against 4%^[9]. the warnings section adds orthostatic hypotension and syncope, an increased risk of bleeding with aspirin, NSAIDs and anticoagulants, priapism (“cases of painful and prolonged penile erections and priapism have been reported. immediate medical attention should be sought”), activation of mania, a discontinuation syndrome, and the “potential to impair judgment, thinking, and motor skills”^[9].

the caveat, which the label states itself: those rates come from depression trials at 150 to 400 mg a day, and “cannot be directly compared to rates in the clinical trials of another drug and may not reflect the rates observed in practice”^[9]. the sleep trials used 25 to 100 mg^[1],[5]. nobody has a label-quality side-effect table for the dose most people are actually taking.

against the alternatives

the one large trial on this page that placed trazodone beside other treatments did so as a second step. in a 2020 sequential trial of 211 adults with chronic insomnia, behavioural therapy and zolpidem produced equivalent first-stage results (45.5% and 49.7% responders)^[8]. people who did not remit on zolpidem and were switched to trazodone went from 31.4% to 49.4% in remission, an odds ratio of 2.13 whose interval (0.91 to 5.00) crosses no effect; the sequence was among those the authors singled out as better for people with a comorbid anxiety or mood disorder^[8]. behavioural therapy for insomnia, which every guideline puts first, is the comparison the trazodone literature mostly lacks^[2],[4].

questions worth taking to whoever prescribed it

• is this for sleep or for mood? the label approves one and the evidence for the other is small and short-term^[1],[2],[9]
• has cognitive behavioural therapy for insomnia been offered? the guideline that suggests against trazodone puts it first^[4]
• what is the plan for stopping, given that there is no trial of long-term use^[2]
• if i am over 65, or drive early, was next-morning impairment weighed^[5],[6]
• if a dose above 100 mg has been suggested, is that a depression dose, and does the side-effect table now apply^[9]