132. A patient specific approach to alignment and proximal junctional kyphosis risk assessment in adult spinal deformity surgery: development and validation of a predictive tool

Abstract

<h3>BACKGROUND CONTEXT</h3> While adult spinal deformity surgery has clinical benefit, the prevalence of proximal junctional kyphosis (PJK) remains between 10% and 40%. Currently accepted alignment targets adjust for age or pelvic incidence, but do not consider other risk factors for PJK. <h3>PURPOSE</h3> To develop a predictive model for PJK and PJK severity that adjusts for known preoperative factors and modifiable surgical alignment. <h3>STUDY DESIGN/SETTING</h3> Retrospective cohort of adult deformity patients undergoing instrumented fusion at a single institution with minimum 2-year follow-up. <h3>PATIENT SAMPLE</h3> This study included 145 adult spinal deformity patients undergoing surgery with minimum 2-year radiographic and clinical follow-up. <h3>OUTCOME MEASURES</h3> Proximal junctional kyphosis, proximal junctional failure, proximal junctional kyphosis severity. <h3>METHODS</h3> Patients undergoing adult deformity surgery between 2009 and 2017 with 2-year follow-up were retrospectively reviewed. Using the Hart-ISSG PJK severity score as an ordinal outcome, a proportional odds regression model was fit to estimate probability of PJK and PJK severity. Predictor variables included preoperative Charlson comorbidity index, vertebral Hounsfield units at the UIV (+/- 4 vertebra), pelvic incidence, T1 pelvic angle, and postoperative regional lumbar lordosis (L1-L4 and L4-S1). Predictor effects were assessed using adjusted odds ratios and a nomogram was constructed for computing probability of PJK. Model performance was assessed using robust internal validation with bootstrap resampling. <h3>Results</h3> Of 145 patients, 47 (32%) developed PJK with a median PJK severity score of 6 (IQR, 4 to 7.5). After adjusting for all predictors, Charlson comorbidity index, vertebral Hounsfield units, preoperative T1 pelvic angle, and postoperative L1-L4 and L4-S1 lordosis were significantly associated with PJK severity (P < 0.05). After adjusting model performance measures for potential overfitting, the predictive model showed acceptable discrimination (C-statistic (AUC) = 0.75) and overall accuracy (Brier score = 0.10). <h3>Conclusions</h3> We have developed a model for predicting probability of PJK while adjusting for preoperative alignment, comorbidity burden, vertebral bone density, and modifiable postoperative L1-L4 and L4-S1 lordosis. This approach may allow alignment targets to be selected based on patient-specific risk of PJK. <h3>FDA DEVICE/DRUG STATUS</h3> This abstract does not discuss or include any applicable devices or drugs.