A smartphone app that provides personalized circadian rhythm feedback based on passively collected sensor data reduced the rate of mood episode recurrence by approximately two-thirds in patients with major depressive disorder and bipolar disorder, according to a multicenter double-blind randomized trial published in the American Journal of Psychiatry.
The study provides the strongest evidence to date that a purely digital chronotherapeutic intervention, one that adjusts daily behavior timing rather than delivering medication or light therapy, can meaningfully alter the course of recurrent mood disorders.
The intervention
Researchers at five Korean university hospitals developed and tested the Circadian Rhythm for Mood (CRM) app. The application runs on a smartphone, collects data passively from the phone’s light sensor and a paired Fitbit wearable (activity, heart rate, sleep timing), and has users complete a brief daily symptom check on an integrated mood chart.
A machine learning algorithm processes these digital phenotype variables to generate individualized three-day mood forecasts, indicating low, moderate, or high risk of an impending episode, and delivers personalized behavioral feedback. The feedback includes recommended wake-up times, optimal light exposure schedules, and activity timing strategies designed to stabilize the circadian system.
The control condition was a sham app that was visually identical but provided nonactionable feedback from a dummy algorithm specifically designed not to influence circadian behavior. Neither participants nor investigators knew which version was assigned. All participants were euthymic (symptom-free) for at least two weeks before enrollment and had not experienced a mood episode in the preceding two years.
What they found
Of the 93 participants randomized, 80 completed sufficient follow-up for the modified intention-to-treat analysis (38 in the active CRM group, 42 in the sham group). Over the 12-month study period:
- The sham group had a 3.39-fold higher rate of recurrent mood episodes compared to the active CRM group (incidence rate ratio 3.39; 95% CI 1.86-6.17). Translated to raw numbers: 41.1 recurrences per 100 person-years in the CRM group versus 139.3 per 100 person-years in the sham group, a reduction of roughly 70%.
- When episodes did occur, they were shorter in the CRM group. Cumulative episode-days per person-year were 2.76 times greater in the sham group (duration rate ratio 2.76; 95% CI 1.19-6.40).
- Time to first recurrence also favored the active app. The hazard ratio was 3.03 (95% CI 1.58-5.81), meaning the sham group faced about three times the risk of experiencing a recurrence at any given point during follow-up.
- Both depressive and hypomanic episodes were reduced. Manic episodes were too few (four total) to draw conclusions.
- No significant adverse effects were reported in either group.
Why it matters
Mood disorders carry a high risk of recurrence, approximately 50-80% of patients with MDD will experience at least one additional episode, and bipolar disorder is almost universally recurrent. Maintenance pharmacotherapy reduces this risk but is limited by side effects, incomplete adherence, and residual symptoms.
Circadian rhythm disruption has long been recognized as both a trigger and a consequence of mood episodes. Social rhythms therapy, interpersonal and social rhythm therapy (IPSRT), and bright light therapy have demonstrated efficacy but require trained clinicians and sustained patient effort. A smartphone-based intervention that automates circadian stabilization could reach many more patients at low marginal cost.
The study’s double-blind, sham-controlled design is a methodological strength that sets it apart from most digital mental health trials, which are typically open-label or compare against treatment-as-usual. Showing a large effect size in a fully blinded comparison defuses the concern that digital interventions work only through nonspecific factors such as increased attention or regular self-monitoring.
Limits
The sample was relatively small (93 enrolled, 80 analyzed), and attrition may have introduced bias, though the modified ITT analysis mitigates this. The study was conducted exclusively in South Korea, and it is unclear whether the app’s algorithm, which was trained on Korean populations, would generalize to other chronotypes and cultural contexts. The trial used only Fitbit devices and Android smartphones, leaving open questions about cross-platform compatibility.
The corresponding author, Professor Heon-Jeong Lee of Korea University, is a co-founder of Huseo Circadian, the company that developed the CRM platform. The authors report that the company provided the app platform but did not participate in study design, data collection, analysis, or interpretation.
Bottom line
A smartphone app that delivers personalized circadian rhythm feedback reduced mood episode recurrence by approximately 70% in patients with MDD and bipolar disorder over 12 months, in a double-blind, sham-controlled trial. The results suggest that digital chronotherapy, delivered at scale through passive sensing and machine learning, warrants consideration as an adjunct to standard pharmacologic maintenance treatment.
Source
Yeom JW, Jeong J, Moon E, Park YM, Lee MS, Yoon HK, Shin C, Yoon Y, Seo JY, Jeon S, Choi M, Cho CH, An H, Lee T, Lee JB, Lee HJ. “Circadian Rhythm Stabilization App to Prevent Mood Episode Recurrence in Patients With Mood Disorders: A Multicenter, Double-Blind, Sham-Controlled, Randomized Clinical Trial.” American Journal of Psychiatry, 2026 Jun 24:appiajp20251008. DOI: 10.1176/appi.ajp.20251008. PMID: 42337416.