H. Koller, W. Hitzl, M. C. Marks, P. O. Newton


September 2019, Volume 28, Issue 9, pp 1987 - 1997 Original Article Read Full Article 10.1007/s00586-019-06000-6

First Online: 24 June 2019

Accurate prediction of spontaneous lumbar curve correction following posterior selective thoracic fusion in adolescent idiopathic scoliosis using logistic regression models and clinical rationale

Introduction

Accurate prediction of spontaneous lumbar curve correction (SLCC) after selective thoracic fusion (STF) remains difficult. This study sought to improve prediction accuracy of SLCC. The hypothesis was preoperative and intraoperative variables could predict SLCC < 20°.

Methods

A multicenter observational prospective analysis was conducted to determine predictors of SLCC in AIS patients that had posterior STF. Curve types included major thoracic curves (Lenke 1, 3–4).The primary outcome variable was to establish prediction models, and a postoperative lumbar curve (LC) ≤ 20° was defined as the target variable. Multivariate logistic regression models were established to study the relationship between selected variables and a LC ≤ 20° versus a LC > 20° at ≥ 2-year follow-up. Single and dual thresholds models in perspective of clinical rationales were applied to find models with the highest positive/negative predictive values (PPV/NPV). The secondary outcome measure was SRS scores at ≥ 2-year follow-up.

Results

410 patients were included. At ≥ 2-year follow-up 282 patients had LC ≤ 20°. These patients had better SRS-22 scores than those with LC > 20° (P = 0.02). The postoperative LC and LC ≤ 20° were predicted by preoperative LC and LC-bending Cobb angle (P < 0.01, r = 0.4–0.6). Logistic regression models could be established to identify patients at risk for failing the target LC ≤ 20°.For preoperative LC and LC-bending, the prediction model achieved a NPV/PPV of 80%/72%. If the postoperative main thoracic curve is combined with the preoperative LC and a gray area for difficult decisions was allowed, model accuracy could even be improved (NPV/PPV = 96%/81%).

Conclusion

An accurate prediction model for postoperative SLCC was established based on a large analysis of prospective STF cases. These models can support prediction and understanding of postoperative SLCC aiding in surgical decision making when contemplating a selective thoracic fusion.

Graphical abstract

These slides can be retrieved under Electronic Supplementary Material.[Figure not available: see fulltext.]


Read Full Article