Phenotypic Heterogeneity of Movement Behaviors and Its Association With Cardiometabolic Health Using Data From the...

Movement behavior is inherently complex and heterogeneous, making it difficult to capture using a single indicator. To better identify meaningful movement behavior phenotypes, robust dimensionality reduction techniques are needed that integrate multiple behavioral dimensions while preserving their continuous nature. This study aimed to identify distinct movement behavior phenotypes and to investigate how variations in movement behavior phenotypes relate to cardiometabolic health biomarkers.

This cross-sectional study included 1009 Luxembourg residents aged 25-79 years, each with at least 4 valid days of triaxial accelerometry data. A reversed graph embedding method was used to reduce the complexity of movement behaviors into a 2-dimensional tree structure, enabling the identification of distinct phenotypes. Separate linear regression models were used to assess associations between the derived tree dimensions and both movement behavior indicators and cardiometabolic health biomarkers.

Four distinct movement behavior phenotypes were identified: inactive, irregularly active, regularly active, and regularly low active. Both tree dimensions were favorably associated with cardiometabolic health outcomes, suggesting that higher overall activity volume, regardless of its regularity, may offer health benefits. Average 24-hour acceleration and total step count emerged as the most influential indicators contributing to phenotypic variation.

This study demonstrates that a novel dimensionality reduction approach effectively captures the complexity of movement behaviors, identifies key movement behavior indicators, and distinguishes between meaningful phenotypes. The findings provide new insights into behavioral heterogeneity and highlight key movement dimensions linked to cardiometabolic health.