Development of a LASSO dynamic prediction system for interbody cage subsidence following OLIF surgery

Cage subsidence after oblique lumbar interbody fusion (OLIF) frequently causes poor outcomes, yet existing predictive models lack accuracy and applicability. This study leveraged the Least Absolute Shrinkage and Selection Operator (LASSO) regression to efficiently identify key predictors of subsidence from quantitative CT vertebral bone quality and endplate-related factors. We analyzed 337 OLIF surgical segments (674 endplates; May 2017-May 2024), observing subsidence in 45.70%. LASSO selected intervertebral disc height correction, intraoperative endplate injury, inferior endplate morphology, and volumetric bone mineral density of the inferior vertebra/endplate as top predictors. These variables informed a multivariable logistic regression model, visualized as a digital nomogram. The model demonstrated excellent predictive performance in training and validation cohorts via ROC, precision-recall, and calibration curves. Decision curve analysis confirmed high clinical utility across risk thresholds, enabling personalized preoperative risk assessment and prevention strategy optimization for OLIF patients.