We provide an umbrella review of the reported polysomnographic changes in patients with neuropsychiatric diseases compared with healthy controls.

An electronic literature search was conducted in EMBASE, MEDLINE, All EBM databases, CINAHL, and PsycINFO. Meta-analyses of case–control studies investigating the polysomnographic changes in patients with neuropsychiatric diseases were included. For each meta-analysis, we estimated the summary effect size using random effects models, the 95% confidence interval, and the 95% prediction interval. We also estimated between-study heterogeneity, evidence of excess significance bias, and evidence of small-study effects. The levels of evidence of polysomnographic changes in neuropsychiatric diseases were ranked as follows: not significant, weak, suggestive, highly suggestive, or convincing.

We identified 27 articles, including 465 case–control studies in 27 neuropsychiatric diseases. The levels of evidence of polysomnographic changes in neuropsychiatric diseases were highly suggestive for increased sleep latency and decreased sleep efficiency (SE) in major depressive disorder (MDD), increased N1 percentage, and decreased N2 percentage, SL and REML in narcolepsy, and decreased rapid eye movement (REM) sleep percentage in Parkinson's disease (PD). The suggestive evidence decreased REM latency in MDD, decreased total sleep time and SE in PD, and decreased SE in posttraumatic stress disorder and in narcolepsy.

The credibility of evidence for sleep characteristics in 27 neuropsychiatric diseases varied across polysomnographic variables and diseases. When considering the patterns of altered PSG variables, no two diseases had the same pattern of alterations, suggesting that specific sleep profiles might be important dimensions for defining distinct neuropsychiatric disorders.