Subject-specific regression equations to estimate lower spinal loads during symmetric and asymmetric static lifting.

Workplace safety assessment, personalized treatment design and back pain prevention programs require accurate subject-specific estimation of spinal loads. Since no noninvasive method can directly estimate spinal loads, easy-to-use regression equations that are constructed based on the results of complex musculoskeletal models appear as viable alternatives. Thus, we aim to develop subject-specific regression equations of L4-L5 and L5-S1 shear and compression forces during various symmetric/asymmetric tasks using a nonlinear personalized finite element musculoskeletal trunk model. Kinematics and electromyography (EMG) activities of 19 young healthy subjects were collected during 64 different symmetric/asymmetric tasks. To investigate the reliability and accuracy of the musculoskeletal model and regression equations, we compared estimated trunk muscle activities and L4-L5 intradiscal pressures (IDPs) respectively with our own electromyography data (EMGs) and reported in vivo pressure measurements. Although in general, six independent rotation components (three trunk T11 rotations and three pelvic S1 rotations) are required to determine kinematics along the spine, only two surrogate variables (trunk flexion and its asymmetric angles) satisfactorily predicted all six rotation components (R2 > 0.94). Regression equations, developed based on subject-specific inputs, predicted spinal loads in satisfactory agreement with IDP measurements (R2 = 0.85). Predicted muscle activities in the personalized musculoskeletal models were in moderate to weak agreements with our measured EMGs in 19 participants. Based on dominance analysis, trunk flexion and its asymmetry angle, hand-load weight, hand-load lever arm, and body weight were the most important variables while the effects of body height and sex on spinal loads remained small. Crown