Melatonin vs. Off-Label Trazodone for Sleep Disruption in ASD
A genuine evidence base for melatonin in autism-related sleep disruption, but no FDA approval for the use and a real alternative-agent question when melatonin resolves one part of the problem but not another.
By 2 a.m. most nights, the only two people awake in the house are Y.F. and, increasingly, one exhausted parent taking a turn sitting with her. Y.F. is 8, autistic, and has never reliably slept through a night in her life, but the pattern has worsened over the past year into something more specific: she falls asleep without much difficulty around 9, wakes fully between 1 and 3 a.m., and simply cannot get back to sleep for two to three hours regardless of what her parents try. A sleep diary kept over the past month confirms what the family already suspected — this isn't difficulty initiating sleep, which is the piece melatonin is best evidenced for, but a middle-of- the-night awakening pattern that a purely sleep-onset intervention may not actually address.
Her parents have already tried melatonin, started at a low dose and titrated up over six weeks under her pediatrician's guidance, with a clear effect on how quickly she falls asleep but no measurable change in the 2 a.m. awakenings that are actually driving the family's exhaustion. Both parents work full-time, and the sleep disruption has begun affecting Y.F.'s own daytime functioning at school, where she has started falling asleep during independent work — new this semester, and something her teacher flagged directly.
What the sleep diary captures that a simple parent description might not is the shape of what happens during those middle-of-the-night hours: Y.F. doesn't cry or call out when she wakes, which is part of why the pattern went undercounted for as long as it did. She lies quietly, sometimes humming to herself, sometimes rocking gently, awake and apparently comfortable but simply unable to return to sleep, for two to three hours before drifting off again just before her usual wake time.
Her parents only fully understood the pattern's true length once one of them began sitting with a notebook rather than relying on memory the next morning, since a wide-awake but quiet child is easy to underestimate in retrospect. That quietness is also, in its own way, a small mercy and a real cost at once — it means she isn't distressed during the awakenings themselves, but it also means the exhaustion accumulating on both sides of her door has gone unaddressed far longer than a more disruptive pattern likely would have.
When melatonin's real evidence doesn't match this child's actual sleep problem
The melatonin literature in autism, including Gringras and colleagues' randomized pediatric trial, is genuinely strong on sleep-onset latency and total sleep time — but it wasn't designed to, and doesn't, address maintenance insomnia the way Y.F. actually has it. Before adding a second medication, I'd want to optimize the melatonin timing and formulation first, since a prolonged-release preparation timed differently across the night sometimes helps a maintenance pattern that immediate-release dosing at bedtime doesn't reach.
That's worth trying, but I'd flag directly that trazodone is the option most families and clinicians reach for next when melatonin's onset benefit doesn't extend to maintenance sleep, and its evidence base for this specific use — pediatric autism, maintenance insomnia — is genuinely thinner than melatonin's. It's off-label here in every sense melatonin technically also is, but with less pediatric-specific trial support behind the extrapolation from adult insomnia use.
I'd push back only on the framing that thinner evidence means it shouldn't be considered at all — it means it should be a considered second step after a real melatonin optimization attempt, not a reason to withhold the option indefinitely from a family this exhausted and a child whose daytime functioning is now visibly affected.
Given the daytime functional impact already documented at school, I wouldn't extend the melatonin-optimization trial indefinitely before considering trazodone — a defined two-to- three-week window for the prolonged-release trial, with a specific plan to add low-dose trazodone if the middle-of-the-night awakenings haven't measurably improved by then, gives both approaches a real, time-bounded chance rather than open-ended sequential trials.
Agreed: switch to prolonged-release melatonin with retimed dosing for a defined three-week trial, sleep diary continued throughout, with trazodone named explicitly as the contingent next step if maintenance awakenings haven't improved by the follow-up visit.
Not agreed: how much evidentiary weight trazodone's thinner pediatric-specific record should carry if it does become necessary — the sleep medicine physician's preference to exhaust further melatonin strategies first was noted as a standing view rather than settled by the current time-bounded plan.