Development and validation of a clinical predictive model for the risk of recurrence after unilateral biportal...

To identify the independent risk factors for postoperative recurrence following unilateral biportal endoscopic discectomy (UBED) for lumbar disc herniation (LDH), and to develop and validate a recurrence risk prediction model.

To analyze the risk factors contributing to postoperative recurrence in patients with LDH treated by UBED, and to construct and validate a predictive nomogram for recurrence risk after UBED.

A total of 257 patients with single-level LDH who underwent UBED were retrospectively enrolled, among whom 24 presented postoperative recurrence, corresponding to a recurrence rate of 9.34%. The cohort was split into a training set (n = 179) and a validation set (n = 78) at a 7:3 ratio. Thirty clinical and radiological parameters were collected. Least Absolute Shrinkage and Selection Operator (LASSO) regression combined with multivariate logistic regression was applied to screen independent risk factors, and a nomogram prediction model was subsequently established. Model performance was assessed via receiver operating characteristic (ROC) curves, Hosmer-Lemeshow (H-L) test, calibration curves, and decision curve analysis (DCA).

Twelve independent risk factors for postoperative recurrence were identified. The area under the ROC curve (AUC) reached 0.976 in the training set and 0.953 in the validation set. The P values of the H-L test were both >0.05 in the two datasets, indicating favorable model calibration. The model yielded significant clinical net benefit within threshold probabilities of 0-0.92 (training set) and 0-0.89 (validation set).

The established nomogram incorporating 12 clinical and radiological variables exhibits satisfactory discrimination and calibration for predicting recurrence after UBED. It serves as a quantitative tool for rapid identification of patients at high recurrence risk in clinical practice. Nevertheless, limitations including single-center retrospective design, relatively small sample size, and an events-per-variable (EPV) ratio of approximately 2:1 (far below the recommended threshold) introduce potential overfitting and limited generalizability. This model is only applicable for preliminary screening of high-risk individuals and preoperative patient counseling, rather than acting as the sole basis for clinical decision-making. Multicenter, large-sample prospective cohort studies are warranted for further external validation and model refinement in the future.