Height Gain After Spinal Fusion for Idiopathic Scoliosis: Which Model Fits Best?

Abstract

Patients will often inquire about the magnitude of height gain after scoliosis surgery. Several published models have attempted to predict height gain using preoperative variables. Many of these models reported good internal validity but have not been validated against an external cohort. We attempted to test the validity of 5 published models against an external cohort from our institution. Models included were Hwang, Van Popta, Spencer, Watanabe, and Sarlak models. We retrospectively queried our institution's records from 2006 to 2019 for patients with adolescent idiopathic scoliosis treated with posterior spinal fusion. We recorded preoperative and postoperative variables including clinical height measurements. We also performed radiographic measurements on preoperative and postoperative radiographic studies. We then tested the ability of the models to predict height gain by evaluating Pearson correlation coefficient, root mean square error, Akaike Information Criterion for each model. A total of 387 patients were included. Mean clinical height gain was 3.1 (±1.7) cm.All models demonstrated a moderate positive Pearson correlation coefficient, except the Hwang model, which demonstrated a weak correlation. The Spencer model was the only model with acceptable root mean square error (≤0.5) and was also the best fitting with the lowest Akaike Information Criterion (-308). The mean differences in height gain predictions between all models except the Hwang model was ≤1cm. Four of the 5 models demonstrated moderate correlation and had good external validity compared with their development cohorts. Although the Spencer model was the best fitting, the clinical significance of the difference in height predictions compared with other models was low. The Watanabe model was the second best fitting and had the simplest formula, making it the most convenient to use in a clinical setting. We offer a simplified equation to use in a preoperative clinical setting based on this data-ΔHeight (mm)=0.77*(preoperative coronal angle-postoperative coronal angle). Not Applicable.