Long-term clinical and cost-effectiveness of a fully automated digital cognitive behavioural therapy for insomnia: 2-year follow-up of a single-blind, superiority, randomised controlled trial

Background We evaluated the long-term clinical and cost-effectiveness of fully automated digital cognitive behavioural therapy for insomnia (dCBT-I) or online patient education (PE) in the Norwegian general population.Methods A parallel-group, participant-blinded, superiority randomised controlled trial in self-referred adults with significant insomnia symptoms (Insomnia Severity Index [ISI] 12).Participants completed automated screening prior to randomisation and outcomes were assessed over 2 years.The primary outcome (9-week clinical effectiveness) has been published.We now report intention-to-treat analyses from the 6-month and 2-year followups.Incremental cost-effectiveness ratio (ICER), with non-parametric bootstrapping, summarized the betweengroup societal costs in 2019 Euros, and quality-adjusted life years (QALYs).The trial was preregistered at ClinicalTrials.gov(NCT02558647) and followed a prespecified protocol.Findings 1720 participants (1167 [67.8%] female; mean age 44.4) were randomised between February 26, 2016, and July 1, 2018 (867 to dCBT-I vs. 853 to PE).The final follow-up included 587/1720 (34.1%) participants (dCBT-I = 315; PE = 272), a median of 28.3 months (IQR 22.6 to 34.0) after baseline.At 2-year follow-up the mean ISI was 10.7 (SD 5.9) in the dCBT-I group and 13.4 (5.9) in the PE group (estimated difference: -1.77 [95% CI: -2.65 to -0.90]; Cohen's d= -0.43).The dCBT-I group reported -278 [95% CI: -1413 to 858] lower costs and 0.025 QALYs gained [95% CI: 0.010 to 0.041] compared with PE, yielding a 94% probability of cost-effectiveness at 0 willingness-to-pay (ICER: -10,973 per QALY gained).No harms or adverse events were reported.Interpretation Compared with online PE, fully automated, low-threshold dCBT-I demonstrates greater short-term (9 week) and longer-term improvements (2 years) in insomnia severity and is likely cost-effective over the 2-year horizon.However, effects attenuated over time and long-term certainty is limited by attrition.