Sleep outcomes in the populations that have been randomised
Evidence basis: randomised trials
Randomised trials agree that weighted blankets are well tolerated and disagree completely about sleep, because each tested a different population.
Population: 120 adults with major depressive disorder, bipolar disorder, generalised anxiety disorder or ADHD and insomnia in a four-week randomised trial with a subsequent open-label follow-up; 73 autistic children with severe sleep problems in a randomised placebo-controlled crossover trial using actigraphy; 94 children with ADHD (mean age 9.0 years, 57.4% boys) in a randomised controlled crossover trial. No retrieved randomised trial enrolled healthy adults without a sleep or psychiatric diagnosis.
Research details
Summary
Three randomised trials test weighted blankets against a lighter blanket, and their results diverge by population rather than by chance. In 120 adults with major depressive disorder, bipolar disorder, generalised anxiety disorder or ADHD and comorbid insomnia, a weighted metal chain blanket produced a large improvement in Insomnia Severity Index ratings over a light plastic chain blanket at four weeks (Cohen's d 1.90, P < .001), with better sleep maintenance, higher daytime activity and reduced daytime fatigue, depression and anxiety symptoms; the authors conclude that weighted chain blankets are an effective and safe intervention for insomnia in these patients. In autistic children with severe sleep problems, the result is null on every objective measure: actigraphy showed the weighted blanket did not increase total sleep time once baseline was adjusted for, there were no group differences in any other objective or subjective measure of sleep including behavioural outcomes, and the authors state plainly that the blanket did not help the children sleep longer, fall asleep faster or wake less often - yet parents and children preferred it. In 94 children with ADHD, the effects were small and did not cover sleep onset: total sleep time increased by 7.72 minutes, sleep efficiency by 0.82 percentage points and wake after sleep onset fell by 2.79 minutes, with sleep-onset latency unchanged (p = 0.432). The three trials measured different outcomes in different populations over different durations, so the honest synthesis is a directionless one: the evidence supports a substantial benefit in one clinical group, a small partial benefit in another and no benefit at all in a third, and pooling them into a single estimate would hide exactly the finding a reader needs.
Why this certainty
Rated Limited and recorded as Mixed evidence. The designs are randomised, which is why this is not graded lower, and each trial is internally credible about the population it studied. It is not Moderate because the trials disagree with each other and the disagreement is not statistical noise but a difference in who was studied and what was measured; because the largest effect is a self-reported outcome that participants could not be blinded to; because the null trial and the small-effect trial used objective actigraphy that a preference cannot influence; and because no trial has been conducted in the general adult population the product is sold to. Pooling the three into one universal estimate would produce a number that describes none of them, so the entry states the heterogeneity instead of averaging it away.
Limitations
The trial populations are not interchangeable, so no single effect estimate applies: one enrolled adults with psychiatric diagnoses, one autistic children and one children with ADHD, and the outcomes measured across them differ (a subjective insomnia questionnaire in the first, actigraphy in the second, actigraphy in the third). The largest effect is also the most subjective: the adults' result rests on the Insomnia Severity Index, a self-reported scale, in a trial that could not blind participants to whether their blanket was heavy, and expectancy is a plausible part of a d of 1.90. The null result in autistic children came from a crossover design whose washout and habituation behaviour with a heavy blanket are difficult to control, and preference findings were collected from parents and children who could tell which blanket was which. The ADHD trial's effects are small in absolute terms - under eight minutes of total sleep time - and it did not change the latency most insomniacs care about. Every trial is short (four weeks to a few months), none measured falls, injury, overheating or long-term use, and the trials excluded people with significant respiratory, circulatory or skin conditions, which is also why the safety claims in this literature are narrower than they read. No trial randomised healthy adults sleeping normally, so the general marketing claim is untested.
Effect estimate
Adults with psychiatric diagnoses and insomnia (n = 120, 4 weeks): Insomnia Severity Index advantage for the weighted chain blanket over a light blanket, Cohen's d 1.90, P < .001. Autistic children (n = 73, crossover): no increase in actigraphically measured total sleep time after adjustment for baseline, and no group difference in any other objective or subjective sleep measure; parents and children preferred the weighted blanket. Children with ADHD (n = 94, crossover): total sleep time +7.72 min (p = 0.027, d = 0.24), sleep efficiency +0.82% (p = 0.038, d = 0.23), wake after sleep onset -2.79 min (p = 0.015, d = -0.27), sleep-onset latency not significant (p = 0.432). No pooled estimate is reported because the populations and outcomes are not comparable.
- A randomized controlled study of weighted chain blankets for insomnia in psychiatric disorders.
- Weighted blankets and sleep in autistic children--a randomized controlled trial.
- The efficacy of weighted blankets for sleep in children with attention-deficit/hyperactivity disorder-A randomized controlled crossover trial.