How does cognitive behavioural therapy for insomnia work? A systematic review and meta-analysis of mediators of change

Insomnia is prevalent and debilitating, comprising sustained difficulties initiating or maintaining sleep. Cognitive-behavioural therapy for insomnia (CBT-I) is a multicomponent intervention recommended as the firstline treatment, but the mediators of change remain unclear. This systematic review and meta-analysis aimed to synthesise and evaluate the evidence for potential mediators of CBT-I. Searches were performed for studies published until February 2021, reporting on mediation analyses with CBT-I. Seventeen unique samples of adults with insomnia (20 studies, N = 3125) were included. Two-stage structural equation modelling was applied to the available data, where 7 studies examined the cognitive mediator, 'dysfunctional beliefs about sleep', 5 studies examined a hyperarousal mediator, and 3 studies examined the behavioural mediator, 'time in bed'. There was evidence in support of changes in dysfunctional beliefs as a cognitive mediator of insomnia symptom improvement following CBT-I. There was more limited evidence in support of changes in hyperarousal-related mediators, and no significant effect of time in bed as a mediator. Numerous studies recorded diary-based measures of potential behavioural mediators, but analyses of these variables were not typically conducted. The most serious limitation of the evidence base is that the temporal precedence of mediator changes cannot be established. Future studies should consider: i) using mid-treatment measurements of mediator changes; ii) reporting on mediator psychometric properties; and iii) explicitly stating analyses as pre-specified or exploratory.

Automated summary

CBT-I may improve insomnia symptoms by changing dysfunctional beliefs about sleep, with limited evidence for hyperarousal-related mediators and no significant effect of time in bed as a mediator. Future studies should consider mid-treatment measurements of mediator changes, report on mediator psychometric properties, and explicitly state analyses as pre-specified or exploratory.