Ten RCTs were included in the review (n=1000). Study quality was generally poor. Only two trials involved randomisation carried out by an independent party, none of the trials reported adequate concealment of treatment allocation or blinding of outcome assessors. Drop-out rates ranged from 0 to 25%. Four trials used an intention-to-treat analysis.
Self help interventions were more effective than waiting list controls for all but one of the sleep outcomes assessed as well as for the overall combined outcome (d=0.36, 95% CI: 0.16, 0.57 based on nine trials, 12 comparisons). There was significant heterogeneity between studies (I2=44.5%, p<0.05). The exception was total sleep time which did not increase (d<0.01, 95%CI: -0.15, 0.19). There was no evidence of heterogeneity for this analysis (I2=18.8%, p>0.05).
Sleep onset latency was worse in patients allocated to self-help treatments compared to those allocate to face-to-face therapy (d=-0.37, 95% CI: -0.73, -0.02, based on three trials, four comparison groups). There was no evidence of heterogeneity between studies (I2=0%, p>0.05). All other outcomes showed small, non-significant (p>0.05) negative effects of the self-help interventions compared to face-to-face therapy.
Anxiety and depression both improved following self-help interventions (d=0.51, 95% CI: 0.11, 0.91 from five trials for depression and 0.28, 95% CI: 0.05, 0.50 for anxiety based on four trials). There was significant heterogeneity for depression (I2=76.5%, p<0.01) but not for anxiety (I2=26.4, p=0.25).
None of the subgroup analyses conducted showed any association with the treatment effect.
The funnel plot suggested the presence of publication bias with smaller studies showing the most positive results. The adjusted effect size following the trim and fill procedure was 0.12 (95% CI: -0.11, 0.35) which was no longer statistically significant.