Puberty changes the natural history of idiopathic scoliosis: three prediction models for future radiographic curve severity from 1563 consecutive patients

Purpose: Understanding idiopathic scoliosis (IS) natural history during growth is essential for shared decision-making between patients and physicians. We developed a retrospective model with the largest available sample in the literature and we aimed to investigate if using three peri-pubertal growth periods provides better prediction than a unique model. Methods: Secondary analysis of a previous study on IS natural history data from radiographs before and at the first consult. Three groups: BEFORE (age 6–10), AT (age 11-Risser 2) and AFTER (from Risser 3) the pubertal growth spurt. Available predictors: Cobb angle, curve type, sex, observation time, and Risser score. We used linear mixed-effects models to predict future Cobb angles in each group. We internally validated prediction accuracy with over 100 patients per group (3 to 5-fold cross-validation). Results: We included 1563 participants (275 BEFORE, 316 AFTER, 782 females and 190 males AT). Curves increased over time mostly in AT, importantly in BEFORE, but also in AFTER. All models performed better than the general one. In BEFORE, 74.2% of the predictions were within ± 5^o, 71.8% in AFTER, 68.2% in AT females, and 60.4% in males. The predictors (baseline curve, observation time also squared and cubic, and Risser score) were similar in all the models, with sex influencing only AFTER. Conclusion: IS curve severities increase differently during growth with puberty stages. Model accuracy increases when tailored by growth spurt periods. Our models may help patients and clinicians share decisions, identify the risk of progression and inform treatment planning.