- Design
- Multicentre, double-blind, sham-controlled randomised clinical trial over 12 months
- Population
- 93 adults with major depressive disorder or bipolar disorder randomised 1:1; 80 in the modified intention-to-treat analysis (38 active, 42 sham)
- Primary outcome
- Number of recurrent mood episodes per participant during 12 months of follow-up
- Effect
- Incidence rate ratio 3.39 (95% CI 1.86-6.17) favouring active; episode-days per person-year rate ratio 2.76 (1.19-6.40); time to recurrence hazard ratio 3.03 (1.58-5.81)
Digital mental health tools are usually tested against waiting lists, which measures enthusiasm as much as effect. This trial did it properly: 93 adults with major depressive disorder or bipolar disorder randomised 1:1 for 12 months to an active app giving individualised three-day mood forecasts and machine-learning-generated circadian feedback from passive sensor data, or to a visually identical sham app whose dummy algorithm produced non-actionable feedback designed not to influence circadian behaviour. Double-blind, multicentre.
In the modified intention-to-treat sample of 80, the sham group had 3.39 times the recurrence rate of the active group (95% CI 1.86-6.17). Cumulative recurrent episode-days per person-year were 2.76 times higher with sham (1.19-6.40), and time to recurrence favoured the active app (hazard ratio 3.03, 1.58-5.81). No significant adverse effects.
The sham design is what makes this worth acting on. Both groups received a daily app, both believed they might be receiving active feedback, and the difference between them is the circadian content rather than the attention. That is a stronger claim than almost any digital mental health trial makes.
The caution is size: 93 randomised, 80 analysed, and an incidence rate ratio of 3.39 with an interval reaching 6.17 rests on a modest number of episodes. Thirteen participants were excluded from the modified intention-to-treat set, which is a meaningful fraction. Replication in a larger trial is needed before this becomes standard.
What it supports today is the underlying clinical principle, which needs no app: circadian rhythm stabilisation is an active treatment in mood disorders, not sleep hygiene advice. Where the app is not available - which is everywhere outside the trial - the deliverable version is a fixed rising time, consistent meal timing, morning light and a regular routine, reviewed as seriously as a drug level.
- Treat circadian stabilisation as active relapse prevention, not as generic sleep advice
- Fixed wake time, consistent meal times and morning light are the deliverable components
- Review rhythm regularity at every maintenance visit alongside mood and adherence
- Note this is 80 analysed patients - replication is needed before it becomes standard care
- The sham comparator makes this stronger than most digital mental health evidence; look for that design
The statistics, in plain English
An incidence rate ratio of 3.39 means the sham group had roughly three and a half times as many recurrences per unit of follow-up; the interval, 1.86 to 6.17, is wide because the number of episodes is small even though the direction is clear. Reporting the comparison this way round - sham versus active - makes the number look larger than the equivalent statement that active reduced recurrence by about 70%. Thirteen of 93 randomised participants did not reach the modified intention-to-treat analysis, and in a trial this size their outcomes could matter.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for psychiatry, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free