Circadian rhythm parameters differentiate euthymic, manic and depressive mood states in bipolar disorders - an explorative pilot study.

Abstract

Background: Bipolar disorders (BD) pose significant therapeutic health challenges due to recurrent and largely unpredictable depressive and (hypo)manic episodes. Traditional self-report methods for symptom monitoring are limited by their dependence on patient adherence which is frequently diminished during symptomatic phases. Circadian movement patterns, measured via actigraphy, have emerged as promising digital biomarkers for distinguishing mood states in BD. This study examined the utility of circadian rhythm parameters in differentiating euthymic, depressive, and (hypo)manic states.

Methods: This study analyzed data from 27 BD patients (mean age = 46 years, 16 female) monitored over 12 months as part of the BipoSense project. Wrist-worn accelerometers continuously recorded physical activity, while mood state was assessed using daily self-reports and biweekly expert evaluations. Circadian rhythm parameters included interdaily stability (IS), intradaily variability (IV), mean activity difference (MeanDiff), and circadian form difference (FormDiff). IS and IV reflect rhythm stability and fragmentation, while MeanDiff and FormDiff quantify overall activity and deviations in circadian rhythm form. Multilevel models were used to predict categorical mood states (depressive, (hypo)manic, euthymic) and dimensional symptom severity.

Results: Physical activity data from 23 patients yielded 2,669 valid days for analysis. In multilevel logistic models, lower MeanDiff (B = -.02, P < .001), reflecting reduced overall activity, lower IS (B = -.80, P = .009), indicating less stable circadian rhythms, and higher FormDiff (B = .03, P < .001), denoting a more rigid circadian activity pattern, were significantly associated with increased odds of depressive days compared to euthymic days. Conversely, higher MeanDiff (B = .02, P = .007) was linked to higher odds of (hypo)manic days. Dimensional linear mixed models showed a similar pattern: lower MeanDiff (β = -.11, P < .001), IS (β = -.06, P = .001), and IV (β = -.06, P = .002), together with higher FormDiff (β = .10, P < .001), predicted increased depressive symptom levels. Conversely, higher MeanDiff (β = .10, P < .001), IS (β = .04, P = .024), IV (β = .07, P < .001), and lower FormDiff (β = -.07, P = .001) were associated with heightened (hypo)manic symptoms.

Conclusions: Circadian rhythm parameters can effectively differentiate mood states in BD, highlighting their potential as clinical markers for episode transitions. Although the study was explorative by nature, the findings emphasize the potential value of integrating circadian biomarkers into digital phenotyping for mood state monitoring. Future studies should explore extended monitoring periods, larger samples, and real-time feedback systems to improve early intervention and personalized treatment strategies in BD.