Unveiling Sleep Dysregulation in Chronic Fatigue Syndrome with and without Fibromyalgia Through Bayesian Networks

2025-02-10

medRxiv

AI Summary

This study addresses sleep architecture abnormalities specific to chronic fatigue syndrome (CFS) and comorbid fibromyalgia (CFS+FM). We propose the first intervention-driven Bayesian network model characterizing second-order Markov transition dynamics across sleep stages in female patients. Validated across multiple independent datasets, the model achieves 70.6% sleep-stage prediction accuracy (cross-domain range: 60.1–69.8%) and 75.4% AUROC for disease classification. Key contributions include: (i) the first identification of pathology-specific anomalies—spanning stage proportions, dwell times, and first-/second-order transition patterns—that collectively form a dynamic sleep biomarker discriminative of healthy controls, CFS-only, and CFS+FM patients; and (ii) causal intervention simulations revealing fundamental mechanistic differences in sleep regulation between CFS and CFS+FM. These findings establish a novel paradigm for differential diagnosis and mechanism-informed therapeutic targeting.

Technology Category

• Cognitive Modeling & Cognitive Systems: Other Foundations of Cognitive Modeling & Systems
• Reasoning under Uncertainty: Relational Probabilistic Models
• Application Domains: Humanities & Computational Social Science

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Social Networks and Social Media: Fairness and bias in social network and social media analysis
• User Modeling, Personalization and Recommendation: Psychology-informed user models and recommender systems

Abstract

Chronic Fatigue Syndrome (CFS) and Fibromyalgia (FM) often co-occur as medically unexplained conditions linked to disrupted physiological regulation, including altered sleep. Building on the work of Kishi et al. [7], who identified differences in sleep-stage transitions in CFS and CFS+FM females, we exploited the same strictly controlled clinical cohort using a Bayesian Network (BN) to quantify detailed patterns of sleep and its dynamics. Our BN confirmed that sleep transitions are best described as a second-order process [14], achieving a next-stage predictive accuracy of 70.6%, validated on two independent data sets with domain shifts (60.1-69.8% accuracy). Notably, we demonstrated that sleep dynamics can reveal the actual diagnoses. Our BN successfully differentiated healthy, CFS, and CFS+FM individuals, achieving an AUROC of 75.4%. Using interventions, we quantified sleep alterations attributable specifically to CFS and CFS+FM, identifying changes in stage prevalence, durations, and first- and second-order transitions. These findings reveal novel markers for CFS and CFS+FM in early-to-mid-adulthood females, offering insights into their physiological mechanisms and supporting their clinical differentiation.

Problem

• Quantifying sleep patterns in CFS and CFS+FM using Bayesian Networks
• Differentiating healthy, CFS, and CFS+FM individuals via sleep dynamics
• Identifying sleep alterations specific to CFS and CFS+FM diagnoses

Innovation

• Bayesian Networks analyze sleep-stage transitions
• Second-order process models sleep dynamics accurately
• BN differentiates healthy, CFS, and CFS+FM individuals