Lordosis distribution index for predicting mechanical complications after long-level fusion surgery: comparison of Global Alignment and Proportion score and Roussouly classification

OBJECTIVE Both the Global Alignment and Proportion (GAP) score and Roussouly classification account for the lordosis distribution index (LDI), but the LDI of the GAP score (G-LDI) is typically set to 50%–80%, while the LDI of the Roussouly classification (R-LDI) varies depending on the degree of pelvic incidence (PI). The objective of this study was to validate the ability of the G-LDI to predict mechanical complications and compare it with the predictive probability of R-LDI in patients with long-level fusion surgery. METHODS A total of 171 patients were divided into two groups: 93 in the nonmechanical complication group (non-MC group) and 78 in the mechanical complication group (MC group). The mean age of the participants was 66.79 ± 8.56 years (range 34–83 years), and the mean follow-up period was 45.49 ± 16.20 months (range 24–62 months). The inclusion criteria for the study were patients who underwent > 4 levels of fusion and had > 2 years of follow-up. The predictive models for mechanical complications using the G-LDI and R-LDI were analyzed using binomial logistic regression and receiver operating characteristic analyses. RESULTS There was a significant correlation between R-LDI and PI (r = −0.561, p < 0.001), while there was no correlation between G-LDI and PI (r = 0.132, p = 0.495). In reference to G-LDI, most patients in the non-MC group were classified as having alignment (72, 77.4%), while the MC group had an inhomogeneous composition (aligned: 34, 43.6%; hyperlordosis: 37, 47.4%). The agreement between the G-LDI and R-LDI was moderate (κ = 0.536, p < 0.001) to fair (κ = 0.383, p = 0.011) for patients with average or large PI, but poor (κ = −0.255, p = 0.245) for those with small PI. The areas under the curve for the G-LDI and R-LDI were 0.674 (95% CI, 0.592–0.757) and 0.745 (95% CI, 0.671–0.820), respectively. CONCLUSIONS The R-LDI, which uses a PI-based proportional parameter, enables individual quantification of LL for all PI sizes and has been shown to have a higher accuracy in classifying cases and a stronger correlation with the risk of mechanical complications compared with G-LDI.