Melatonin Use Linked to Small REM Sleep Reduction in Children

Outpatient melatonin use was associated with a modest reduction in rapid eye movement (REM) sleep among children referred for polysomnography (PSG), according to a propensity score–matched study published in JAMA Network Open.¹ The association held across multiple sensitivity analyses but attenuated when psychiatric comorbidities were added to the model, and the authors caution that its clinical significance remains uncertain.

Study Draws on Large Pediatric Sleep Database

Melatonin is the most widely used sleep aid among children, yet no large study had previously examined its relationship to objectively measured sleep architecture. Using the Nationwide Children's Hospital Sleep DataBank, researchers analyzed 3392 children evaluated at a single academic pediatric sleep laboratory between 2017 and 2019, comparing 346 outpatient melatonin users with 3046 nonusers.

Because melatonin users differed substantially from nonusers at baseline, including higher rates of epilepsy, attention-deficit/hyperactivity disorder (ADHD), anxiety, and depression, the investigators used 1:1 propensity score matching on 6 covariates to control for confounding by indication. This yielded 342 matched pairs, with all covariates achieving standardized mean differences below the prespecified threshold of 0.10.

REM Sleep Percentage Lower Among Melatonin Users

The prespecified primary outcome, percentage of REM sleep, was lower among matched melatonin users than nonusers (median, 16.7% vs 19.0%; Hedges g, −0.22; 95% CI, −0.32 to −0.11; false discovery rate [FDR]–corrected P = .003). None of the 14 exploratory PSG outcomes, including total sleep time, sleep efficiency, non-REM sleep stages, respiratory indices, arousal index, and periodic limb movements, differed significantly between groups after correction for multiple comparisons.

The REM finding proved directionally consistent across 7 prespecified subgroup analyses and 4 post hoc sensitivity analyses, including 1:2 and 1:3 matching ratios and models using an expanded set of diagnosis-based covariates, with effect sizes remaining within 0.06 of the primary estimate in each case. The association was among the strongest in the epilepsy subgroup (Hedges g, −0.21; n = 114 pairs), a pattern consistent with research findings that sleep problems in children with epilepsy track more closely with co-occurring neurodevelopmental conditions than with seizure activity itself, underscoring how comorbidity burden can complicate sleep outcomes in this population.²

When the propensity model was adjusted for ADHD, anxiety, and depression, the association persisted but weakened (Hedges g, −0.17; 95% CI, −0.27 to −0.06).¹ A further model adding insomnia and 2 sleep medication classes attenuated the effect to a point where its CI bound approached the null (Hedges g, −0.11; 95% CI, −0.22 to 0.01).¹

An exploratory age-stratified analysis suggested the REM reduction was largest among children aged 6 to 12 years and smaller at the youngest and oldest ends of the pediatric range, although the interaction term did not reach statistical significance.

Authors Urge Caution on Clinical Interpretation

The study authors calculated an E-value of 1.74 for the association's confidence bound closest to the null, indicating that an unmeasured confounder of moderate strength could account for the observed relationship. They also noted that the cross-sectional design cannot rule out reverse causation, since children with intrinsically lower REM sleep may be more likely to be prescribed melatonin in the first place.

Because the exposure definition relied on outpatient prescription records rather than confirmed use, dose, timing, or formulation, the authors could not verify that melatonin was taken on the night of the PSG, nor could they capture over-the-counter use among children classified as nonusers. The referred clinical population, drawn from a single center with elevated rates of sleep-disordered breathing, epilepsy, and neurodevelopmental disorders, also limits generalizability to community samples.

The authors concluded that prospective studies with documented dosing, timing, and adherence are needed to determine whether the REM association reflects a pharmacologic effect, residual confounding, or some combination of the 2.

“The REM association was attenuated when psychiatric diagnoses were added to the propensity model,“ they wrote. “Residual confounding cannot be excluded, and prospective studies with documented dose, timing, and adherence are needed to clarify the directionality of this association.”

References

1. Juginovic A, Rodman L. Melatonin use and polysomnographic sleep architecture in children. JAMA Netw Open. 2026;9(8):e2626983. doi:10.1001/jamanetworkopen.2026.26983

2. Steinzor P. Investigators examine role of neurodevelopmental characteristics in mediating sleep problems in pediatric epilepsy. AJMC^®. July 14, 2023. Accessed September 24, 2026. https://www.ajmc.com/view/investigators-examine-role-of-neurodevelopmental-characteristics-in-mediating-sleep-problems-in-pediatric-epilepsy