Prediction Model for Recurrence After Endoscopic Sinus Surgery in Chronic Rhinosinusitis With Nasal Polyps

Abstract

Background

Recurrence of chronic rhinosinusitis with nasal polyps (CRSwNP) following endoscopic sinus surgery (ESS) is common, with reported rates varying considerably depending on follow-up duration. A clinically practical and reliable model for predicting long-term recurrence risk remains an unmet need.

Objective

To identify clinical predictors of recurrence and develop a prognostic model for estimating recurrence-free survival at 2, 5, 10, and 15 years after ESS.

Methods

A retrospective, single-institution cohort study of 437 patients with CRSwNP who underwent ESS was analyzed for time-to-event recurrence. Candidate predictors, including age, sex, smoking status, asthma, NSAID hypersensitivity, symptom duration, blood eosinophil count (BEC), modified Lund-Kennedy (MLK) score, and Lund-Mackay (LM) score, were entered into a LASSO penalized Cox model for variable selection. A nomogram was constructed to estimate recurrence-free survival at predefined time points. Model performance was assessed using time-dependent area under the ROC curve (AUROC), Brier scores, calibration curves, internal validation via 1000 bootstrap resamples, and clinical utility through decision curve analysis (DCA).

Results

Recurrence occurred in 54.0% of patients. The LASSO-penalized Cox model identified age, NSAID hypersensitivity, asthma, symptom duration, BEC, MLK, and LM scores as significant predictors. The nomogram demonstrated strong discrimination, with AUROCs of 0.878, 0.870, 0.886, and 0.873 at 2, 5, 10, and 15 years post-ESS, respectively. Corresponding Brier scores were 0.150, 0.147, 0.135, and 0.138, indicating low prediction error. Internal validation confirmed the model's stability, with AUROCs of 0.873, 0.866, 0.879, and 0.864 at the same time points. Calibration plots showed good agreement between predicted and observed outcomes across all time horizons. DCA demonstrated greater net benefit compared to treat-all or treat-none strategies across the 0.1 to 0.9 threshold range.

Conclusion

The nomogram developed using a LASSO-penalized Cox model offers a robust, well-calibrated, and clinically applicable tool for individualized long-term recurrence risk prediction in patients with CRSwNP following ESS.