Sensitivity of musculoskeletal model-predicted loads to spinal alignment individualization in adult spinal deformity

Spinal alignment is a critical biomechanical risk factor for mechanical complications after adult spinal deformity (ASD) surgery. Although spinal loads are substantially affected by alignment, musculoskeletal (MSK) models often simplify its reconstruction. This study aimed to assess the radiographic conformity of alignment reconstruction methods and to evaluate their effects on spinal load predictions in an ASD cohort. A second objective was to investigate load deviations under a coronal-symmetry assumption in relation to coronal deformity. Patient-specific thoracolumbar MSK models of 554 ASD patients (age: 52.5 ± 21.2yrs) were generated from biplanar radiograph annotations using six alignment reconstruction strategies with increasing level of detail in the sagittal and coronal planes: from few regional measures to individual vertebral positions and orientations. Geometric conformity of MSK models was assessed from vertebral orientation and centroid position errors relative to biplanar data. Inverse-static simulations were performed to estimate intervertebral forces. Models based on regional measures produced large geometric errors, with vertebral inclination deviations up to 29° in the sagittal and 57° in the coronal plane, and position errors up to 6 cm anteroposteriorly (AP), 13cm craniocaudally, 17cm mediolaterally. The most detailed alignment representation (SC₃) showed spinal load differences of 2-17% in compression and 10-36% AP shear from simplified alignment models. Load discrepancies generally increased with coronal deformity severity; however, no trend was identified, as the effect of omitting coronal alignment was already evident in mild deformities. Spinal load predictions in ASD are highly sensitive to alignment reconstruction assumptions, supporting vertebral-level, bi-planar alignment individualization whenever detailed radiographic data is available.