Physical activity phenotypes in endometriosis using unsupervised learning via functional mixture models

Abstract Background Endometriosis is a chronic condition associated with severe pelvic pain, dysmenorrhea, infertility, and worsening quality of life. Regular physical activity (PA) is effective for pain management and reducing chronic disease symptoms, yet individuals with endometriosis are more likely to be insufficiently active. This study investigated latent profiles of daily PA trajectories in this population via clustering. Methods We analyzed 171 adults (4,795 person-level days) with a confirmed diagnosis of endometriosis enrolled in the All of Us Research Program. PA data were collected from participants using Fitbit wrist-worn trackers. We used 30 consecutive days of data from each individual, allowing up to 10 days of missingness, imputed using multiple imputed chained equations. Functional mixture models (FMMs) were used to identify latent PA trajectory clusters using daily step counts as the outcome variable. The optimal number of clusters was selected via Bayesian Information Criterion (BIC). Exploratory analyses of PROMIS pain and fatigue surveys were conducted in a subset of 129 participants who completed the surveys after their PA time windows. Results FMM-identified profiles differed both with respect to PA volume and variability. Combinatory model fit indices supported a 4-cluster (K = 4) solution. The “High Active” phenotype exhibited the highest volume and variability of daily step counts and moderate-to-vigorous PA (MVPA) minutes over the sampling period (Steps: Mean (SD) = 12918.8 (5606.4); MVPA: Mean (SD) = 75.2 (64.6)). The “High Moderate” phenotype exhibited the second highest activity (Steps = 9283.9 (3661.2); MVPA = 58.2 (59.6)), followed by “Low Moderate” (Steps = 6234.0 (2515.8); MVPA = 18.6 (32.3)), and “Insufficiently Active” (Steps = 4317.1; MVPA = 17.2 (28.9)). Exploratory analyses revealed that higher-activity phenotypes tended to report lower pain scores. However, the “High Active” phenotype had the highest proportion of individuals reporting severe to moderate fatigue. Conclusion This is the first study to investigate and report distinct PA profiles among a nationally-representative sample of individuals living with endometriosis using objectively-estimated PA. Identifying phenotypes based on within- and between-individual variance may help identify those at risk and inform the development of personalized interventions aimed at promoting PA and improving health outcomes in this population.