The T4-L1-Hip Axis Objectifies the Roussouly Classification Using Continuous Measures.

BACKGROUND The Roussouly classification is a popular system for the categorization of spinal alignment, although the categorization of continuous measures may compromise efforts toward a precision-medicine approach to sagittal alignment in spine surgery. Vertebral-pelvic angles provide continuous measures of sagittal alignment without the risk of misclassification. METHODS We performed a cross-sectional study of asymptomatic adult volunteers with normal spines (no evidence of disc degeneration or scoliosis). Full-spine radiographs were obtained, and radiographic parameters were collected, including pelvic incidence (PI), sacral slope, lumbar lordosis, the apex of lordosis, the L1-pelvic angle (L1PA), and the T4-pelvic angle (T4PA). All spines were classified as Roussouly Type 1, 2, 3, or 4 on the basis of sacral slope and the apex of lumbar lordosis. Associations between the L1PA and PI, the L1PA and T4PA, and the T4-L1PA mismatch and PI were assessed for the whole cohort and when stratified by Roussouly type. A multinomial logistic regression model was fit to estimate Roussouly type based on PI, the L1PA, and the T4PA. Agreement (weighted κ), accuracy, and area under the receiver operating characteristic curve (1 type versus the rest) were computed. A subanalysis assessed potential variations in the relationships when Roussouly Type-3 spines were further classified as Type 3A (anteverted) versus Type 3. RESULTS The 320 included volunteers had a median age of 37 years (interquartile range [IQR], 27 to 47 years), and 193 (60%) were female. By self-reported race or ethnicity, the highest percentage of patients were Caucasian (White, 38%) or East Asian (36%), followed by Arabo-Bèrbère (16%). Spines were classified as Roussouly Type 1 in 18 (6%) of the volunteers, as Type 2 in 63 (20%), as Type 3 in 161 (50%), and as Type 4 in 78 (24%). The L1PA was strongly associated with PI across Roussouly types (weakest in Roussouly Type-1 spines). A multinomial logistic regression model estimating Roussouly type by PI, the L1PA, and the T4PA showed strong agreement (weighted κ, 0.84), excellent discrimination, and overall accuracy of 0.82. CONCLUSIONS The T4-L1-Hip axis is conceptually aligned with the description of spinal shapes in the Roussouly classification but with the advantage of utilizing continuous measures of spinal alignment. Goals of surgical realignment incorporating the T4-L1-Hip axis will be comparable with alignment planning using the Roussouly classification but with improved accuracy and precision. LEVEL OF EVIDENCE Diagnostic Level II. See Instructions for Authors for a complete description of levels of evidence.