You have begun to emerge from the fog that had settled over your life. You are back at work, seeing friends again, and perhaps even sleeping a little better. Yet, in the background, is there a persistent worry that the clouds might return.

For many people living with major depressive disorder (MDD), this concern is not misplaced. Even with adequate treatment, approximately 30-50% of people with MDD experience relapse within five years of remission (Kessler & Bromet, 2013). Such recurrent episodes may gradually compromise social functioning, work performance, and quality of life (Verduijn et al., 2017).

Sleep problems are closely connected with depression (Sullivan, 2026). Disturbances in sleep and circadian rhythm – the body’s internal clock – are central to this condition, linked with lower remission rates (Edinger et al., 2023), higher suicidality (Harris et al., 2020), and an increased risk of relapse (Matcham et al., 2024). Importantly, changes in sleep and rest-activity patterns may be detectable before a depressive episode has fully emerged (Solelhac et al., 2024).

Wearable devices are an exciting, objective way to study sleep and activity in real-world settings. Previous studies using actigraphy (wristwatch-like device used to monitor daily rest-and-activity cycles) have posited a bidirectional relationship between depressive symptoms and disrupted sleep or rest-activity rhythms (Smagula et al., 2022). However, most previous studies have been brief (2-16 weeks), relied on self-reported depressive symptoms, and/or used consumer-grade wearables.

This new study by Tonon et al. (2026), published in JAMA Psychiatry, aimed to address these gaps by asking: can specific changes in sleep and rest-activity rhythms, derived from actigraphy, help predict who will experience a relapse? 

Methods

Adults with remitted MDD from five Canadian outpatient clinics participated in an up to two years (median 46 weeks) observational study as part of the CAN-BIND Wellness Monitoring programme.

Participants wore a GT9X Link actigraphy device continuously throughout the study. Sleep and rest-activity metrics were derived from the accelerometery data and averaged over 2‑week epochs. Metrics included:

• Sleep regularity index (SRI): day-to-day consistency of sleep timing.
• Relative amplitude (RA): the contrast between daytime activity and night-time rest.
• Wake after sleep onset (WASO): time spent awake after initially falling asleep.
• Composite phase deviation (CPD): variability in sleep timing relative to a person’s typical sleep schedule.

Participants also attended in-person assessments every 8 weeks, which included clinician-rated measures such as the Montgomery-Åsberg Depression Rating Scale (MADRS).

The primary outcome was relapse, defined as one or more of the following, verified by a panel of five board-certified psychiatrists: MADRS ≥22 for at least 2 consecutive weeks; hospitalisation; risk of suicide (based on intent/behaviour); changes/escalation to treatment. Non-relapsing participants were further classified as ultrastable (MADRS ≤14 throughout) or unstable (periods of MADRS >14 without meeting relapse criteria). Thus, there were three clinical groups: ultrastable (n=39), unstable (n=27), and relapse (n=28).

Results

A total of 102 participants met the inclusion criteria and completed the baseline assessment; 96 remained after early dropouts, and 93 (mean age 39.1 years; 62% female) provided usable actigraphy data. Together, they contributed an impressive 31,898 actigraphy days, with the median monitoring period lasting 46 weeks. The median time to relapse was 33 weeks (range 6-94 weeks).

Baseline predictors of relapse

Cox proportional hazards regression was conducted, adjusting for age, sex, season, and baseline MADRS scores. Several actigraphy measures were associated with future relapse risk:

• Lower SRI (HR=0.46, 95%CI [0.28 to 0.74], p=.002)
• Lower RA (HR=0.45, 95%CI [0.29 to 0.70], p<.001)
• Higher WASO (HR=1.77; 95%CI [1.12 to 2.80]; p=.01).

Thus, participants with more irregular sleep-wake patterns, lower contrast between daytime activity and nighttime rest, and greater time spent awake after sleep onset were more likely to experience relapse.

Also associated with relapse risk were reduced sleep efficiency, higher night-time activity, and as expected, higher baseline MADRS scores.

Time-varying models

The authors then examined whether changes in these markers over time were associated with relapse using time-dependent Cox models.

In the primary time-varying model (adjusted for the same covariates) two actigraphy metrics stood out as predictors of relapse risk:

• Higher CPD (HR=1.76, 95%CI [1.04 to 2.98], p=.04).
• Lower RA (HR=0.45; 95%CI [0.21 to 0.97]; p=.046).

Again, a weaker, less distinct day-night activity contrast remained a consistent predictor of relapse. Greater variability in sleep timing relative to one’s typical schedule, was also associated with increased relapse risk, implying that disrupted sleep-wake rhythms and day-night patterns may be an important marker of vulnerability.

The authors then ran a second time-varying model restricted to the two weeks before each MADRS assessment. In this more stringent model, lower RA and higher concurrent MADRS scores remained associated with a higher risk of relapse.

Trajectories over time

Finally, longitudinal analyses using linear mixed-effects models showed that compared with the ultrastable group, the relapse group consistently showed lower SRI. There was also some evidence of lower RA and a less steep decline in sleep phase variability over time.

Similar trends were observed when comparing the unstable and relapse groups:

• SRI (β=-0.57; SE=0.25; p=.03)
• RA (β=-0.69; SE=0.24; p=.006)

Interestingly, there were no statistically significant differences in these longitudinal results between unstable and ultrastable participants. This suggests that actigraphy may help distinguish individuals at risk of imminent relapse from those who remain well, potentially reflecting underlying physiological processes specifically linked to relapse risk in MDD.

Conclusions

Tonon et al. (2026) concluded that specific, differentiated:

actigraphy-derived sleep and rest-activity rhythms were associated with MDD relapse.

These markers, measured concurrently (e.g., SRI, RA) and over time (e.g., SRI, CPD), were able to differentiate individuals who relapsed from those who did not, including stable patients and those with fluctuating symptoms still in remission.

These findings support actigraphy as a promising digital biomarker for detecting early physiological signs of relapse, which could enhance traditional clinical assessments and support the development of more personalised treatment approaches in MDD.

Strengths and limitations

Strengths

The study’s primary strength is its design. Unlike many wearable studies that rely on short monitoring periods, this research implemented continuous actigraphy for up to two years, offering a more comprehensive and reliable picture of sleep and activity patterns over time. Additionally, with around 32,000 days of actigraphy data and an independent panel confirming each relapse event, the outcome assessment was exceptionally robust for a real-world clinical cohort.

Secondly, using continuous wrist-worn actigraphy, the researchers could examine potential predictors of relapse without placing excessive additional demands on participants. This feels especially important from the perspective of my own work using actigraphy with autistic children and their caregivers, where families are often already managing substantial cognitive, emotional, and practical demands, and minimising burden on participants’ limited time and energy is essential.

In addition, to avoid simply capturing the very early stages of an episode already underway, the researchers excluded data collected after relapse and the two weeks immediately preceding it. I consider this a very deliberate methodological decision, because it allows the analysis to genuinely assess whether sleep and rest-activity patterns can predict relapse risk before symptoms worsen significantly.

Limitations

At the same time, there are a few limitations. Some of the longitudinal associations the conclusions rest on appear as trends rather than consistently significant effects.

Additionally, the Sadeh-based sleep scoring used, shares the familiar weaknesses of employing actigraphy: high sensitivity to movement but relatively low specificity for wake detection, which tends to underestimate WASO and overestimate sleep efficiency (Conley et al., 2019). However, the authors are explicit about this, and it is not a flaw unique to their work; similar weaknesses in the use of actigraphy have been reported in other samples (e.g., in children; Meltzer et al., 2012).

Another limitation is that the participants were treatment-responsive, recruited via clinics, and were able to engage with a demanding long-term protocol and wear a device continuously. The sample was also predominantly White (just above 80%). Thus, people from minoritised groups, difficult-to-treat depression, precarious housing situations, or limited access to specialist care are likely underrepresented, yet may be significantly susceptible to the risk of relapse.

It is also important to interpret the findings in light of the study’s funding and affiliations, including substantial support from the Ontario Brain Institute and Janssen, as well as multiple authors with industry affiliations; nonetheless, the authors are transparent about these connections.

Taken together, these factors suggest that, while this work represents a valuable and methodologically rigorous contribution, its conclusions would be strengthened by independent replication in larger, more diverse, and representative cohorts.

Implications for practice

The findings potentially strengthen the case for integrating routine, low‑burden monitoring of sleep and daily rhythms into ongoing care for people in remission from MDD, particularly those with a history of recurrent episodes. Once replicated across more diverse samples and settings, with consistently significant patterns, this kind of monitoring could become part of standard relapse prevention.

Importantly, even after accounting for depressive symptom scores (MADRS), the authors found that objective disruptions in sleep timing and day-night activity patterns provided information beyond what clinicians can obtain from symptom scales and routine clinical interviews alone. They further noted that most current relapse prediction models focusing on symptom severity and dimensions have limited predictive accuracy. They proposed that actigraphy-derived measures, which may reflect underlying biological processes, might be more effective in identifying specific targets to lower relapse risk, such as cognitive behavioural therapy for insomnia, addressing comorbid sleep disorders, and implementing more structured sleep hygiene and chronotherapy strategies.

For me, the key implication for research is that this paper also sets the stage for the next step: interventional studies that use actigraphy-derived markers to guide more tailored and timely support for MDD, and then assess whether this approach prevents relapse. There are also limits to actigraphy, both in terms of accuracy (it tends to under- or over‑estimate certain sleep metrics) and practical issues such as gradual non‑adherence to wearing the device over time (as seen in other CAN-BIND work, e.g., Slyepchenko et al., 2023). Before actigraphy can be considered part of standard relapse prevention, evidence is needed that these markers are robust and reliable across more diverse populations at risk of MDD relapse, remain informative and acceptable in long-term use, and that interventions guided by them genuinely reduce the likelihood of relapse.

On a more personal note, this is something many of us recognise intuitively; that small, gradual disruptions to sleep and daily structure are often the first sign that something is wrong. Tonon et al.’s (2026) results quantify and name that pattern: objective changes in sleep and rest-activity rhythms become a shared language among patients, clinicians, and researchers; a way to notice that the weather is turning before the storm fully breaks. The question the paper leaves me with is a hopeful one: if we learn to trust and act on these early signals, might we help the clouds thin just enough for a ray of sun to break through?

Statement of interests

Rhea Varghese has no involvement in the CAN-BIND programme or the study by Tonon et al (2026), and does not know the authors personally. She has no financial relationships with Janssen Research & Development, the Ontario Brain Institute, or other funders mentioned in the paper.

Her own work is in the field of developmental psychology and includes using actigraphy to measure sleep in autistic children and parents, which gives her an interest in this methodology as a way to predict long-term outcomes, but no stake in these specific findings.

Editor

Edited by Éimear Foley. ChatGPT assisted with language refinement and formatting during the editorial phase.

Links

Primary paper

Andre Tonon, Adile Nexha, Jasmyn Cunningham et al. (2026). One-Year Actigraphy Study of Sleep and Rest-Activity Rhythms as Markers of Relapse in Depression. JAMA psychiatry, 83(4), 379–388. https://doi.org/10.1001/jamapsychiatry.2025.4453

Other references

Conley, S., Knies, A., Batten, J., Ash, G., Miner, B., Hwang, Y., Jeon, S., & Redeker, N. S. (2019). Agreement between actigraphic and polysomnographic measures of sleep in adults with and without chronic conditions: A systematic review and meta-analysis. Sleep medicine reviews, 46, 151–160. https://doi.org/10.1016/j.smrv.2019.05.001

Edinger, J. D., Smith, E. D., Buysse, D. J., Thase, M., Krystal, A. D., Wiskniewski, S., & Manber, R. (2023). Objective sleep duration and response to combined pharmacotherapy and cognitive behavioral insomnia therapy among patients with comorbid depression and insomnia: a report from the TRIAD study. Journal of Clinical Sleep Medicine, 19(6), 1111-1120.

Harris, L. M., Huang, X., Linthicum, K. P., Bryen, C. P., & Ribeiro, J. D. (2020). Sleep disturbances as risk factors for suicidal thoughts and behaviours: A meta-analysis of longitudinal studies. Scientific Reports, 10(1), 13888. https://doi.org/10.1038/s41598-020-70866-6

Kessler, R. C., & Bromet, E. J. (2013). The Epidemiology of Depression Across Cultures. Annual Review of Public Health, 34(Volume 34, 2013), 119-138. https://doi.org/10.1146/annurev-publhealth-031912-114409

Matcham, F., Carr, E., Meyer, N., White, K., Oetzmann, C., Leightley, D., Lamers, F., Siddi, S., Cummins, N., Annas, P., De Girolamo, G., Haro, J., Lavelle, G., Li, Q., Lombardini, F., Mohr, D., Narayan, V., Penninx, B., Coromina, M., . . . Hotopf, M. (2024). The relationship between wearable-derived sleep features and relapse in Major Depressive Disorder. Journal of Affective Disorders, 363, 90-98. https://doi.org/10.1016/j.jad.2024.07.136

Meltzer, L. J., Montgomery-Downs, H. E., Insana, S. P., & Walsh, C. M. (2012). Use of actigraphy for assessment in pediatric sleep research. Sleep Medicine Reviews, 16(5), 463-475. https://doi.org/10.1016/j.smrv.2011.10.002

Slyepchenko, A., Uher, R., Ho, K., Hassel, S., Matthews, C., Lukus, P. K., Daros, A. R., Minarik, A., Placenza, F., Li, Q. S., Rotzinger, S., Parikh, S. V., Foster, J. A., Turecki, G., Müller, D. J., Taylor, V. H., Quilty, L. C., Milev, R., Soares, C. N., . . . Frey, B. N. (2023). A standardized workflow for long-term longitudinal actigraphy data processing using one year of continuous actigraphy from the CAN-BIND Wellness Monitoring Study. Scientific Reports, 13, 15300. https://doi.org/10.1038/s41598-023-42138-6

Smagula, S. F., Zhang, G., Gujral, S., Covassin, N., Li, J., Taylor, W. D., Reynolds, C. F., 3rd, & Krafty, R. T. (2022). Association of 24-Hour Activity Pattern Phenotypes With Depression Symptoms and Cognitive Performance in Aging. JAMA psychiatry, 79(10), 1023–1031. https://doi.org/10.1001/jamapsychiatry.2022.2573

Solelhac, G., Imler, T., Strippoli, M. F., Marchi, N. A., Berger, M., Haba-Rubio, J., Raffray, T., Bayon, V., Lombardi, A. S., Ranjbar, S., Siclari, F., Vollenweider, P., Marques-Vidal, P., Geoffroy, P., Léger, D., Stephan, A., Preisig, M., & Heinzer, R. (2024). Sleep disturbances and incident risk of major depressive disorder in a population-based cohort. Psychiatry Research, 338, 115934. https://doi.org/10.1016/j.psychres.2024.115934

Sullivan, E. (2026, May 11). How sleep changes across later life, and what it means for mental health – National Elf Service. National Elf Service.

Verduijn et al. Verduijn, J., Verhoeven, J. E., Milaneschi, Y., Schoevers, R. A., van Hemert, A. M., Beekman, A. T. F., & Penninx, B. W. J. H. (2017). Reconsidering the prognosis of major depressive disorder across diagnostic boundaries: Full recovery is the exception rather than the rule. BMC Medicine, 15, Article 215. https://doi.org/10.1186/s12916-017-0972-8

Photo credits

• Illustration by the iop on Unsplash
• Photo by Kinga Howard on Unsplash
• Photo by Brett Jordan on Unsplash
• Photo by Megan te Boekhorst on Unsplash
• Photo by Miles Peacock on Unsplash
• Illustration by Gilang Fahmi on Unsplash