Decoding brain-heart dynamics: Effective connectivity predictors of heart rate variability

Autonomic dysregulation characterizes neuropsychiatric and somatic disorders, often reflecting disrupted brain-heart communication mediated by the Central Autonomic Network (CAN). The CAN integrates visceral inputs and cortical control to maintain autonomic balance. Heart rate variability (HRV) provides a peripheral index of CAN regulation, yet the causal dynamics underlying HRV-brain interactions remain poorly understood. We investigated effective connectivity (EC) within a core (C-CAN), extended (E-CAN) and non-canonical CAN (N-CAN) to characterize bidirectional brain connectivity-heart dynamics at rest. Resting-state fMRI and photoplethysmography were acquired from 232 adults (164 females; mean age = 47.8 ± 18.9 years). PPG-derived HRV metrics (time, frequency, and entropy) were extracted and EC was estimated via regression dynamic causal modeling across 100 brain regions, including 42 C-CAN nodes. Predictive modeling was implemented using cross-validated ridge regression and bidirectional interactions were modeled using HRV as a driving input. The E-CAN EC model best predicted entropy metrics (ApEn: r = 0.22, SampEn: r = 0.21). The C-CAN model improved predictive performance of these metrics (ApEn: r = 0.23), SampEn: r = 0.27). Non-CAN EC predictions aligned with that of the E-CAN (SampEn: r = 0.17). Analyses also revealed HRV-driven influences on distributed cortical and subcortical regions. Our findings show that EC predicts HRV through integrative brain networks beyond canonical CAN nodes. Entropy-based HRV measures emerged as sensitive indicators of central influence on heart dynamics, while bottom up cardio-autonomic signals causally influenced key brain regions supporting neurovisceral integration. Collectively, these results highlight that the complexity of causal brain connectivity-heart interactions, reflected in HRV dynamics, mirrors that of ROI-to-ROI connectivity patterns across canonical and extended CAN.