Spine Journal - 2026-03-29 - Journal Article
A nomogram prediction model for lumbar disc herniation recurrence after percutaneous endoscopic lumbar discectomy: a multicenter retrospective study.
Lin K, Yang J, Liu T, Wei J, Zhang B, Chu L, Gu Y, Abudouaini H
Topics
Key Takeaway
A six-variable nomogram predicts recurrent LDH after PELD with AUC 0.755 (training) and 0.831 (external validation), identifying Modic type II changes (OR 5.755) as the strongest independent predictor.
Summary Depth
Choose how much analysis to show on this article page.
Summary
This multicenter retrospective study developed and externally validated a nomogram to predict rLDH after PELD at L4-L5 or L5-S1 using LASSO-selected variables and multivariate logistic regression. Six independent predictors were identified: younger age, Modic type I (OR 1.650) and type II (OR 5.755) changes, smaller multifidus CSA (OR 0.900), excessive lumbar lordosis (OR 1.033), and larger facet joint orientation (OR 1.047). The model achieved AUC 0.755 in the training cohort and 0.831 in the external validation cohort with satisfactory calibration and positive decision curve analysis.
Key Limitation
The absence of intraoperative variables — specifically annular defect size and extent of nucleus removal — omits factors with established mechanistic links to reherniation that could substantially alter model performance.
Original Abstract
BACKGROUND CONTEXT
Recurrent lumbar disc herniation (rLDH) is a common postoperative complication after percutaneous endoscopic lumbar discectomy (PELD), with recurrence rates reported between 5% and 21%. Identifying patients at risk remains a critical clinical need, yet few predictive models exist to guide prevention.
PURPOSE
To develop and externally validate a predictive nomogram model for rLDH following PELD, based on preoperative clinical and imaging parameters.
STUDY DESIGN/SETTING
A multicenter retrospective cohort study using data from four hospitals in China.
PATIENT SAMPLE
A total of 3,610 patients who underwent PELD for L4-L5 or L5-S1 disc herniation were included. The training cohort comprised 2,436 patients from two centers between January 2012 and December 2021. The external validation cohort included 1,174 patients from two additional centers between January 2012 and December 2021.
OUTCOME MEASURES
The primary outcome was the occurrence of rLDH, defined as symptom recurrence at the operated level after a pain-free interval of at least two weeks. Predictive accuracy was assessed using area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis (DCA).
METHODS
Independent risk factors for rLDH were identified using multivariate logistic regression after variable selection with the LASSO algorithm. A nomogram was constructed based on the final model. Internal validation was performed using bootstrap resampling. External validation was conducted in a separate patient cohort. This study received no external funding. The authors declare that no financial support was obtained for the research, authorship, and/or publication of this article.
RESULTS
Younger age (OR per year increase 0.972, 95% CI 0.956-0.988), Modic type I (OR 1.650, 95% CI 1.075-2.607) and type II (OR 5.755, 95% CI 3.702-9.202) changes, smaller multifidus cross-sectional area (CSA) (OR 0.900, 95% CI 0.864-0.937), excessive lumbar lordosis (OR 1.033, 95% CI 1.019-1.049), and larger facet joint orientation (OR 1.047, 95% CI 1.031-1.064) were independent predictors of rLDH. The nomogram demonstrated good discriminative ability in both the training cohort (AUC = 0.755) and validation cohort (AUC = 0.831), with satisfactory calibration and clinical utility supported by DCA.
CONCLUSION
We developed and externally validated a nomogram model to predict rLDH after PELD using readily available preoperative variables. This tool may assist spine surgeons in early risk stratification and tailored postoperative management; however, as this was a retrospective study, prospective validation and inclusion of intraoperative variables are still needed.