Consistency among musculoskeletal models: Caveat utilitor

Musculoskeletal simulation software and model repositories have broadened the user base able to perform musculoskeletal analysis and have facilitated in the sharing of models. As the recognition of musculoskeletal modeling continues to grow as an engineering discipline, the consistency in results derived from different models and software is becoming more critical. The purpose of this study was to compare eight models from three software packages and evaluate differences in quadriceps moment arms, predicted muscle forces, and predicted tibiofemoral contact forces for an idealized knee-extension task spanning -125 to +10 of knee extension. Substantial variation among models was observed for the majority of aspects evaluated. Differences among models were influenced by knee angle, with better agreement of moment arms and tibiofemoral joint contact force occurring at low to moderate knee flexion angles. The results suggest a lack of consistency among models and that output differences are not simply an artifact of naturally occurring inter-individual differences. Although generic musculoskeletal models can easily be scaled to consistent limb lengths and use the same muscle recruitment algorithm, the results suggest those are not sufficient conditions to produce consistent muscle or joint contact forces, even for simplified models with no potential of co-contraction.