Exercise dysfunction is highly prevalent across the global spectrum of medical diseases, and is a principal cause of morbidity and increased healthcare cost burden. Abnormalities in multiple organ systems often underlie exercise dysfunction. However, current methods for analyzing exercise test results often rely on a narrow subset of variables involving a single organ system. This reductionist approach has, in turn, hampered consensus on the parameters that define different forms of exercise dysfunction and the development of patient-specific treatments.

To address this dilemma, data were assembled from a large cohort of patients (N=738) referred to BWH (2011-0215) for invasive cardiopulmonary exercise testing (iCPET), which provides comprehensive measurements of pulmonary function, cardiopulmonary hemodynamics, and skeletal muscle oxygen extraction at rest and peak exercise (Figure 1). A correlation network of functionally distinct iCPET variables was assembled that contained 98 pairwise correlations, 39 nodes, and 101 edges (Figure 2). We focused on a 10-variable subnetwork to group patients into 4 distinct clusters. Clustering depended on contributions from all 10 variables in the subnetwork, but was independent of traditional exercise diagnoses (e.g., heart failure). The clusters were associated with distinct clinical profiles, variable exercise performance, and significant differences in hard clinical end-points (Figure 3).

This systems-based method for classifying exercise dysfunction has important implications on point of care risk stratification in patients, and the development of exercise subtype-specific therapies. There is substantial opportunity for repurposing this approach to other complex clinical phenotypes for which enhanced diagnosis, prognosis, and patient-specific treatments are needed.

For more information about Dr. Maron’s research, please contact Partners HealthCare Innovation by clicking here.