Sex-specific, subject-specific modeling reveals anatomical drivers of hip adductor moment arms

The hip adductors play critical roles in locomotion and stability, yet how anthropometric features influence muscle moment arms, and whether these relationships differ by sex, remains unclear. This study developed a subject-specific musculoskeletal modeling framework to quantify the influence of anatomical scaling features and sex on adductor moment arms. Using MRI data from 80 healthy adults across a range of body sizes, subject-specific hip adductor models were constructed using automated segmentation and k-nearest neighbor-based muscle path generation. Model outputs were validated against two independent experts, with pipeline performance falling within inter-expert variability. Fixed-effect and mixed-effect models were used to evaluate relationships between adductor moment arms and global, regional, and local features. Global anthropometrics exhibited weak associations with moment arms and significant sex-specific differences, suggesting that global scaling approaches would require sex-specific corrections to remain accurate. In contrast, regional and local features showed stronger, muscle-specific associations with fewer sex-specific differences remaining, indicating that locally informed scaling may reduce the need for sex-specific adjustments Residual analyses identified sex-specific effects for select muscle-feature pairs not captured by geometry alone, and mixed-effects modeling confirmed that including sex as a covariate improved model fit beyond feature-only models. These findings demonstrate that hip adductor moment arm variations across individuals and sexes are best explained by local and regional morphology. Notably, while local features inherently capture sex-specific differences, global anthropometrics do not; thus, in the absence of subject-specific anatomical data, incorporating sex-specific modeling approaches remains necessary for accurate predictions and rigorously addressing sex as a biological variable.