Prognostic model development for risk of curve progression in adolescent idiopathic scoliosis: a prospective cohort study of 127 patients

Background and purpose: The study’s purpose was to develop and internally validate a prognostic survival model exploring baseline variables for adolescent idiopathic scoliosis curve progression. Methods: A longitudinal prognostic cohort analysis was performed on trial data (n = 135) including girls and boys, Cobb angle 25–40°, aged 9–17 years, remaining growth > 1 year, and previously untreated. Prognostic outcome was defined as curve progression of Cobb angle of > 6° prior to skeletal maturity. 34 candidate prognostic variables were tested. Time-to-event was measured with 6-month intervals. Cox proportional hazards regression survival model (CoxPH) was used for model development and validation in comparison with machine learning models (66.6/33.3 train/test data set). The models were adjusted for treatment exposure. Results: The final primary prognostic model included 127 patients, predicting progress with acceptable discriminative ability (concordance = 0.79, 95% confidence interval [CI] 0.72–0.86). Significant prognostic risk factors were Risser stage of 0 (HR 4.6, CI 2.1–10.1, P < 0.001), larger major curve Cobb angle (HRstandardized 1.5, CI 1.1–2.0, P = 0.005), and higher score on patient-reported pictorial Spinal Appearance Questionnaire (pSAQ) (HRstandardized 1.4, CI 1.0–1.9, P = 0.04). Treatment exposure, entered as a covariate adjustment, contributed significantly to the final model (HR 3.1, CI 1.5–6.0, P = 0.001). Sensitivity analysis displayed that CoxPH maintained acceptable discriminative ability (AUC 0.79, CI 0.65–0.93) in comparison with machine learning algorithms. Conclusion: The prognostic model (Risser stage, Cobb angle, pSAQ, and menarche) predicted curve progression of > 6° Cobb angle with acceptable discriminative ability. Adding patient report of the pSAQ may be of clinical importance for the prognosis of curve progression.