Circadian rhythm modulation in heart rate variability as potential biomarkers for major depressive disorder: A machine learning approach

Abstract

Major depressive disorder (MDD) is associated with reduced heart rate variability (HRV), but its link to circadian rhythm modulation (CRM) of HRV is unclear. Given that depression disrupts circadian rhythms, assessing HRV fluctuations may better capture the CRM and the related autonomic nervous system (ANS) alterations, potentially enhancing our understanding of the pathophysiological mechanisms of MDD. This study aimed to explore the relationship between CRM of HRV and MDD, and to identify potential biomarkers for MDD using machine learning (ML). A total of 165 MDD patients and 60 healthy controls (HCs) were enrolled in the study, with each participant completing 24-h Holter electrocardiogram (ECG) monitoring and psychological scale assessments prior to receiving antidepressant treatment. The circadian rhythm of HRV was quantified using a cosine regression model, and seven typical ML models were employed to distinguish MDD from HCs. MDD patients exhibited a significant decrease in average diurnal HRV indices, particularly during night-time, along with reductions in the parameter M of HRV circadian rhythms compared to HCs. Depression severity was negatively associated with the parameters M of RMSSD, PNN50, HF, while positively associated with the parameter M of LF/HF ratio. Furthermore, the gradient boosting machine (GBM) model demonstrated the best performance in classifying MDD (accuracy 0.823, AUC 0.868), and a final GBM model was developed with 12 selected features. This study provides new insights into the relationship between circadian rhythm abnormalities and MDD, highlighting the potential of using CRM of HRV as novel biomarkers for MDD pathophysiology and clinical applications.