Network Analysis to Risk Stratify Patients With Exercise Intolerance.

Area of Science:

• Cardiopulmonary disease research
• Exercise physiology
• Biostatistics and data analysis

Background:

• Current clinical risk assessment for exercise intolerance in cardiopulmonary disease patients relies on limited traditional variables.
• There is a need for alternative strategies that incorporate a broader spectrum of prognostic factors.

Purpose of the Study:

• To utilize unbiased analyses to identify variables associated with clinical risk in patients experiencing exercise intolerance.
• To develop a novel risk-stratification calculator for improved patient management.

Main Methods:

• Retrospective analysis of invasive cardiopulmonary exercise testing data from 738 patients.
• Construction of a correlation network of exercise parameters (|r|>0.5) to identify key variables.
• K-mean clustering based on a subnetwork of 10 variables, including peak oxygen consumption (pVo2), to define patient clusters.
• Development and validation of a risk-stratification calculator using the derived clusters.

Main Results:

• A network analysis identified 4 novel patient clusters with significant differences in exercise measurements (P<0.01).
• The network model was less redundant and identified more distinct clusters compared to a probabilistic model.
• Cluster assignment predicted subsequent clinical events, with higher risk clusters showing increased hospitalization hazard (e.g., 4.3-fold increase).
• The developed risk calculator demonstrated consistent performance across independent cohorts.

Conclusions:

• Network analyses successfully identified novel exercise groups and enabled the creation of a point-of-care risk calculator.
• This approach expands the clinical utility of variables beyond pVo2 for predicting hospitalization in patients with exercise intolerance.
• The findings offer a more comprehensive approach to risk stratification in this patient population.