Study design:

Retrospective, single-center case study.

Objective:

Postoperative cervical imbalance with cervical sagittal vertical axis (cSVA) >4 cm can be predicted in 3-level to 5-level anterior-only cervical multilevel fusion surgery (ACMS).

Summary of background data:

Previous studies established correlations between cervical kyphosis (CK) correction and postoperative balance (cSVA ≤4 cm) with improved clinical outcomes. Understanding of what influences restoration of cervical lordosis (CL) in patients with degenerative disease with mild to moderate CK subjected to ACMS is important. To achieve a better understanding of geometric changes after ACMS, this study examines factors predicting perioperative alignment changes and regional interdependencies.

Materials and methods:

Analysis of patients with ACMS. Analysis included patient baseline characteristics, demographics and complications, and focused on radiographic measures including CL C2-7, fusion angle (FA), C7-Slope (C7S), T1-slope (T1S), T1-CL mismatch, and cSVA (cSVA ≤4 cm/>4 cm). We aimed to predict postoperative imbalance (cSVA >4 cm) and conducted a multivariable logistic regression analysis.

Results:

Inclusion of 126 patients with 3-level to 5-level ACMS, mean age was 56 years and 4 fusion levels on average. Preoperative CK was present in 9%, mean FA-correction was 8 degrees, maximum 46 degrees. Postoperatively, 14 patients had cSVA >4 cm. A neural network model for prediction of cSVA >4 cm was established including preoperative cSVA, preoperative CL and correction of FA. The model achieved high performance (positive predictive value=100%, negative predictive value=94%, specificity=100%, sensitivity=20%). Also, variables such as nonunion, chronic lumbar pain or thoracolumbar multilevel fusion influenced the postoperative cSVA >4 cm rate. Alignment analysis highlighted strong correlations between C7S/T1S and cSVA/C2-tilt (r=0.06/r=0.7, P<0.0001). A formula was established to transfer cSVA data into C2-tilt data.

Conclusion:

This study identified independent variables predicting postoperative cSVA >4 cm including FA, which can be influenced by the surgeon. Our model supports the decision-making process targeting a postoperative cSVA ≤4 cm.