Passively collected sleep and activity data detected early warning signs of depression relapse weeks before a full episode emerged, results of a longitudinal cohort study showed.

In a 1- to 2-year study of 93 adults with remitted major depressive disorder (MDD), increasingly irregular sleep patterns nearly doubled the risk for relapse. The strongest predictor was a dampened circadian rhythm, marked by a reduced contrast between daytime activity and nighttime rest.

“Our findings are consistent with prior research showing that sleep disturbance functions both as a symptom of an acute depressive episode and as an early warning signal of future relapse,” lead author Benicio Frey, MD, PhD, professor of psychiatry at McMaster University in Hamilton, Ontario, Canada, told Medscape Medical News.

“The data suggest that the biology may extend beyond sleep itself, pointing to broader disruptions in circadian rhythms,” he added.

Frey said that these early changes in daily rhythms could create a window for action, giving patients and clinicians time to reinforce sleep hygiene, adjust routines, or seek clinical follow-up before symptoms fully return.

The study was published online on February 11 in JAMA Psychiatry.

Tracking Sleep and Mood Over Time

MDD affects roughly 5% of adults worldwide each year and frequently follows a recurrent course. More than 40% of patients who recover from an initial episode will experience a recurrence within 2 years. After two episodes, the risk for recurrence within 5 years is approximately 75%.

Most existing predictive models rely primarily on symptom scales and clinical history — approaches that have demonstrated limited accuracy. Frey said longer-term passive actigraphy may provide a more objective biological signal.

Investigators conducted the study at five sites through the Canadian Biomarker Integration Network in Depression (CAN-BIND). They followed 93 adults (mean age, 39 years; 62% women) with MDD diagnosed using the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) who were in remission at baseline, indicated by a Montgomery-Åsberg Depression Rating Scale (MADRS) score of 14 or lower.

Throughout the follow-up, participants completed clinician assessments every 2 months and wore a research-grade wrist actigraphy device continuously for up to 2 years, generating nearly 32,000 days of sleep and activity data.

Relapse was defined prospectively. Participants met criteria if their MADRS score rose to 22 or higher for at least 2 consecutive weeks or if they required hospitalization, developed suicidal intent or behavior, or needed intensified antidepressant treatment. Each case was reviewed and confirmed by an independent panel of psychiatrists.

The study showed that individuals whose sleep became less regular and whose nighttime wakefulness increased were more likely to relapse. Poorer sleep efficiency and higher overnight activity showed similar patterns.

Notably, reduced relative amplitude — a weaker distinction between daytime activity and nocturnal rest — emerged as a robust predictor of recurrence. Participants with lower relative amplitude had nearly double the risk for relapse (hazard ratio, 0.45), even after researchers accounted for subtle fluctuations in depressive symptoms.

In the weeks leading up to recurrence, sleep-wake schedules grew progressively more erratic. By contrast, individuals who maintained remission showed far more stable daily rhythms.

An Important Advance

Reached for comment, Benjamin W. Nelson, PhD, senior clinical research scientist in digital biomarkers at Verily Life Sciences and adjunct professor at Harvard Medical School and Beth Israel Deaconess Medical Center in Boston, said the study advances digital monitoring research in two important ways.

First, investigators paired continuous, yearlong actigraphy with structured clinician assessments every 2 months using MADRS — an FDA-accepted endpoint widely used in clinical trials. That design provides stronger temporal resolution than studies relying on brief monitoring windows or infrequent symptom check-ins.

“High-resolution monitoring like this avoids the need to average continuous digital data across long intervals, which removes the strength of continuous collection,” Nelson said.

Second, participants had formal DSM-5 diagnoses of MDD rather than relying solely on self-reported questionnaire cutoffs, strengthening diagnostic validity.

Nelson cautioned that the modest sample size and relatively homogeneous population may limit generalizability. The study also relied exclusively on actigraphy without additional physiologic sensors, such as photoplethysmography or temperature monitoring, that could provide more detailed sleep-staging data.

If replicated, he said, wearable-derived sleep and activity metrics could eventually support relapse-monitoring strategies, particularly for individuals with recurrent depression.

The study was funded by the Ontario Brain Institute through an Ontario Research Fund-Research Excellence grant, with additional support from Janssen Research & Development as part of CAN-BIND. Several authors reported receiving consulting fees or research funding from or holding advisory roles in pharmaceutical and biotechnology companies, as detailed in the published report. Nelson reported receiving salary and equity from Verily Life Sciences.