Predictive Model for Cervical Alignment and Malalignment Following Surgical Correction of Adult Spinal Deformity.

Abstract

Retrospective review of prospective multicenter database. Use predictive modeling to identify patient characteristics, radiographic, and surgical variables that predict reaching an outcome threshold of suboptimal cervical alignment after adult spinal deformity (ASD) surgery. Cervical deformity (CD) after ASD correction has been defined with the following criteria: T1S-CL>20°, C2-C7 SVA>40 mm, and/or C2-C7 kyphosis >10°. While studies have analyzed CD predictors, few have defined and identified predictors of optimal cervical alignment after thoracolumbar surgery. Inclusion criteria were surgical ASD patients with baseline and 2-year follow-up. Postoperative cervical alignment (CA) and malalignment (nonCA) at 2 years was defined with the following radiographic criteria: 0°≤T1S-CL≤20°, 0 mm≤C2-C7 SVA≤40 mm, or C2-C7 lordosis >0°. Three thresholds classifying malalignment were defined: (T1) missing 1 criterion, (T2) missing 2 criteria, (T3) missing 3 criteria. Multivariable logistic stepwise regression models with bootstrap resampling procedure were performed for demographic, surgical, and radiographic variables. The model was validated with receiver operative characteristic and area under the curve. Two hundred twenty-five surgical ASD patients were included. At 2 years 208 patients (92.4%) were grouped as CA in T3, while 17 (7.6%) were nonCA. Patients were similar in age (CA: 56.10 vs. nonCA: 55.78 years, P = 0.150), BMI (CA: 26.93 vs. nonCA: 26.94 kg/m, P = 0.716), and sex (CA: 76.5% vs. nonCA: 87.0%, P = 0.194). The final predictive model included C2 slope, C2-T3 CL, T1S-CL, C2-C7 CL, Pelvic Tilt, C2-S1 SVA, PI-LL, and Smith-Peterson osteotomies number. In this model (area under the curve 89.22% [97.49-80.96%]), the following variables were identified as predictors of nonCA: increased Smith-Peterson osteotomies use (OR: 1.336, P = 0.017), and C2-T3 angle (OR: 1.048, P = 0.005). This study created a statistical model that predicts poor 2-year postoperative cervical malalignment in ASD patients. T3 (patients not meeting all three alignment criteria) was the most effective threshold for modeling nonCA, and included increased baseline C2-T3 angle and increased Smith-Peterson osteotomies during index. 3.