Sleepio to treat insomnia symptoms

The clinical evidence comprises 28 studies, 12 of which are randomised controlled trials

3.1 The EAC assessed 25 full text publications, an unpublished study and 2 abstracts. Twelve of the studies were randomised controlled trials (including 6 secondary analyses of randomised controlled trials). Six were non-randomised studies and there was 1 abstract. Also, the company provided the EAC with 2 unpublished, real-world evidence studies. For full details of the clinical evidence, see section 3 of the assessment report.

The 12 randomised controlled trials are relevant to the decision problem and show that Sleepio reduces symptoms of insomnia

3.2 There is good quality evidence that Sleepio improves sleep in people with self-reported insomnia symptoms (according to DSM-5 [Diagnostic and Statistical Manual of Disorders 5], SCI [Sleep Condition Indicator] and ISI [Insomnia Severity Index] measures). The most robust evidence for Sleepio comprises 12 randomised controlled trials, 10 of which used intention to treat analyses to control for high drop-out rates. The studies are small relative to the potential reach of Sleepio but are adequately powered and well reported.

The UK population is well represented in the evidence for Sleepio

3.3 The UK population is well represented in the evidence base for Sleepio, which includes 7 UK studies and 4 multinational studies that included UK populations. Four of the studies done in the UK were randomised controlled trials (Espie et al. 2012, Freeman et al. 2017, Denis et al. 2020 [pilot study], Kyle et al. 2020); all concluded that Sleepio was more effective in reducing insomnia symptoms than the comparator (standard care, waiting list, placebo or attention control).

The evidence is heterogenous

3.4 The studies included in the assessment varied in design, population, outcome measures and comparators. Study participants included people with difficulty sleeping with or without medical and mental health comorbidities, and different durations of insomnia. The comparator differed between studies and often the description of standard care lacked clarity. It was unclear whether standard care included aspects of cognitive behavioural therapy for insomnia (CBT‑I), the hypnotic medication prescription, or both, and there was little information about what was offered as sleep hygiene education.

There is no evidence comparing Sleepio with face-to-face CBT or other forms of digital CBT‑I

3.5 The company acknowledged that the lack of evidence comparing Sleepio with face-to-face CBT or digital CBT‑I was a limitation. It said that face-to-face CBT‑I for insomnia is not routinely available on the NHS and is not scalable to the UK NHS population. There is an indirect meta-analysis (Soh et al. 2020) that indicated that digital CBT-I is non-inferior to face-to face CBT-I. There are currently no studies that compare Sleepio with other digital CBT‑I technologies.

The company used a single cohort spreadsheet model to compare the cost of Sleepio with treatment as usual and face-to-face CBT‑I

3.6 The company submitted 12 economic studies relevant to the economic assessment. The EAC found 3 of them met the decision problem. The company's economic analysis modelled a population of adults with insomnia symptoms. The model compared Sleepio with 2 comparators: treatment as usual (which includes sleep hygiene and sleep medication), and face-to-face CBT‑I. The company's analysis estimates the overall cost of providing Sleepio to a large population. The company assumed that:

3.7 The cost impact and proportion of patients using Sleepio are based on data from Sampson et al. 2021. The key costs were:

For full details of the cost evidence, see section 4 of the assessment report.

The EAC made changes to the percentage uptake parameter

3.8 The EAC considered that the company's assumption that 1% of the population will start session 1 of Sleepio was an overestimate. The EAC changed it to 0.58% (13,920 people in a population of 2.4 million) based on data for the Buckinghamshire region reported in Sampson et al. (2021). The company provided uptake data from the rollout of Sleepio in North Hampshire, where a 0.84% uptake rate was observed using the standard implementation model. New data from years 2 and 3 in Buckinghamshire suggest that the uptake rate is maintained in subsequent years, although the rate is significantly lower than the 1% assumed by the company.

The EAC's updates to the cost model make Sleepio cost incurring compared with treatment as usual but cost saving compared with face-to-face CBT-I

3.9 With the updated cost parameters, the EAC's base case shows that after 3 years:

3.10 The EAC used sensitivity analyses to explore how the percentage uptake affects the cost of Sleepio compared with treatment as usual. These showed that cost savings fall as the proportion of users (uptake) reduces. The breakeven point for the first year cohort was an uptake percentage of 0.666%.

The EAC used sensitivity analyses to explore costs associated with subsequent cohorts being included in the modelling

3.11 The company modelled the costs associated with using Sleepio in a single cohort over a 3 year time horizon. The EAC did additional analyses to quantify the rolling cost of Sleepio considering subsequent cohorts of patients using it. The EAC modelled 2 scenarios. In both scenarios, the overall cost rises over time:

The EAC concluded that the statistical analysis outlined in Sampson et al. (2021) is robust

3.12 The committee asked the EAC to review the statistical analysis described in Sampson et al. (2021) and explore if it was possible to link the NHS data with the data from Sleepio to better understand the outcomes associated with its use. Patient-level data was made available to the EAC, who replicated the multilevel generalised linear model described in the paper. It was not possible to link the NHS data to the data available from Sleepio users about usage and weekly sleep score. The EAC also investigated adding an individual patient level to the generalised linear model, the impact of seasonal adjustment, and relevant comorbidities. It found that the resource use saving results from the statistical model did not change significantly from those reported in the study (£6.64 compared with £5.53 per patient per year in the EAC model). It concluded that the Sampson et al. (2021) results are robust enough for use in the economic modelling for Sleepio.

The EAC re-ran the economic modelling using the company's alternative cost model

3.13 The company proposed an alternative Sleepio cost model (described as the Scotland cost model), which is tiered and based on the anticipated annual treatment volumes. Treatment reduces in cost as volume increases, from £80 per patient for 1 to 1,000 patients having treatment to £60 per patient for more than 10,000 patients having treatment. Using this costing approach, Sleepio's cost in the EAC's base case reduced from £155.17 to £66.11 per patient. With the Scotland costing model applied to the data in Sampson et al. (2021), the EAC's base case showed that after 3 years: