Which sagittal plane assessment method is most predictive of complications after adult spinal deformity surgery?

Purpose Different methods of sagittal alignment assessment compete for predicting adverse events after adult spinal deformity (ASD) surgery. We wanted to study which method provides greater benefit. Methods Retrospective study of 391 patients operated for ASD, with > 6 instrumented levels, fused to the pelvis, and 2 years of follow-up. Three alignment methods were analyzed 6-week postoperatively: (1) Roussouly mismatch; (2) GAP score/GAP categories; (3) T4-L1-Hip axis. Binary logistic regression generated models that best predict the following adverse events: mechanical complications (MC): in general and isolated (PJK, PJF, rod breakage); reinterventions (in general and after MC); and readmissions. ROC/AUC analysis was also implemented. In a second regression round, we added different variables that were selected on univariate analysis-demographic, surgical, and radiographic-to complete the models. Results The best predictor parameters in most models were T4-L1PA mismatch and GAP score; we could not prove a predictive ability of the Roussouly mismatch. The T4-L1PA mismatch best predicted general MC, PJK, PJK + PJF, and readmission, while the GAP score best predicted PJF and reinterventions (for MC and for any complication). However, the variance explained by these models was limited (Nagelkerke's R2 = 0.031-0.113), with odds ratios ranging from 1.070 to 1.456. ROC curves plotted an AUC between 0.57 and 0.70. Introducing additional variables (demographic, surgical, and radiographic) improved prediction in all the models (Nagelkerke's R2 = 0.082-0.329) and allowed predicting rod breakage. Conclusion The T4-L1-Hip axis and GAP score show potential in predicting adverse events, surpassing the Roussouly method. Despite partial efficacy in complication anticipation, recognizing postoperative sagittal alignment as a key modifiable risk factor, the crucial need arises to integrate diverse variables, both modifiable and non-modifiable, for enhanced predictive accuracy.