Sleep-Based Model Helps Distinguish Unipolar and Bipolar Depression: Study
France: A non-invasive predictive model combining self-reported sleep measures, actigraphy, and polysomnography (PSG) effectively differentiated unipolar depression from bipolar depression during a major depressive episode, according to researchers. The findings suggest that sleep-related measures may help improve diagnostic discrimination between the two conditions.
The study, published in Depression and Anxiety by Jeanne Leseur, Department of Psychiatry and Addictology, AP-HP, GHU Paris Nord, DMU Neurosciences, Bichat–Claude Bernard Hospital, Paris, France, and colleagues, addressed the challenge of distinguishing unipolar depressive disorder (UDD) from bipolar depressive disorder (BDD). The researchers noted that sleep disturbances differ between the conditions, with insomnia more commonly associated with UDD and hypersomnia observed more often in BDD.
The analysis included 159 patients experiencing a major depressive episode, comprising 116 with UDD and 43 with BDD. Participants underwent clinical assessments and completed sleep-related questionnaires, while objective sleep and circadian patterns were assessed using actigraphy and PSG. The researchers then used backward stepwise logistic regression to identify the combination of variables that best differentiated the two depressive disorders.
Key findings included:
- Patients with UDD reported poorer sleep quality and more severe insomnia, along with lower sleep efficiency measured through actigraphy.
- Those with BDD had longer total rest time over 24 hours and lower activity levels during both their least active five-hour period and most active 10-hour period.
- PSG showed that patients with BDD had a greater proportion of N2 sleep and total non-REM sleep.
- Six sleep and circadian variables were retained in the final predictive model, which explained 49.4% of the variance.
- The model demonstrated strong discriminatory performance, with an area under the curve (AUC) of 0.926, sensitivity of 88.2% and specificity of 89.5%.
- The model also achieved a positive predictive value of 88.3% and negative predictive value of 89.5%.
The researchers said the findings support a multidimensional approach in which subjective sleep complaints are combined with objective measures of sleep architecture and circadian activity. Importantly, no individual sleep marker was sufficiently accurate on its own, indicating that the diagnostic value came from combining multiple features.
Limitations included the inability to account for medication effects because antidepressant and mood-stabilizer use was closely linked to diagnostic status. Actigraphy and PSG were also unavailable for some participants because of missed appointments, technical issues, or limited access. The model explained about half of the observed variance, indicating that additional predictors are needed. PSG accessibility may also limit routine clinical application, although ambulatory PSG could improve feasibility.
The researchers concluded that integrating subjective, actigraphic, and polysomnographic measures could offer a promising approach for differentiating UDD from BDD. They highlighted N2 sleep percentage as a particularly relevant marker and suggested that future multimodal and AI-based approaches could further improve diagnostic classification.
Reference:
Leseur, J., Geoffroy, P. A., Perozziello, A., Abderrahmane, D., Rach, H., Lejoyeux, M., Frija, J., & Maruani, J. (2025). A Sleep and Circadian Biomarker-Based Predictive Model for Differentiating Unipolar and Bipolar Depression. Depression and Anxiety, 2026(1), 5556144. https://doi.org/10.1155/da/5556144
Next Story