Beyond MVPA minutes: physical activity phenotypes are associated with metabolic profiles in children with obesity

Pediatric obesity is associated with early cardiometabolic risk, yet MVPA alone may not capture meaningful variation in movement behavior. We applied DDRTree to accelerometry data to identify multidimensional physical activity phenotypes in children and adolescents with obesity and assessed their associations with body composition and cardiometabolic. This cross-sectional study included 91 children and adolescents with obesity (43% girls), aged 6-17 years, who wore a thigh-mounted accelerometer for 7 days. Multidimensional accelerometry features were extracted and PA phenotypes identified using DDRTree, an unsupervised reversed graph-embedding method that maps high-dimensional data onto a low-dimensional branching manifold. DDRTree identified 3-PA phenotypes. Anthropometry, metabolic outcomes, and a continuous metabolic syndrome risk score were compared across phenotypes using ANCOVA adjusted for age and sex. DDRTree identified three PA phenotypes: Sedentary (24%), Light Activity (55%), and Higher Activity (21%); only three participants met MVPA recommendations. Compared with the Sedentary Pattern, the Higher Activity Pattern accumulated 122 min/day less sedentary time and 16.5 min/day more MVPA (p < 0.001) and showed lower fat mass percentage, higher fat-free mass percentage, higher insulin sensitivity (SPISE), and a lower metabolic risk score (p ≤ 0.05).Conclusion: Multidimensional PA phenotypes derived from accelerometry identify clinically meaningful differences in body composition and metabolic risk beyond MVPA duration, supporting phenotype-based approaches for pediatric obesity risk stratification and intervention design.