Risk factors for postoperative recurrence of sinonasal inverted papilloma: a retrospective analysis of 171 cases

BACKGROUND: Sinonasal inverted papilloma(SNIP) is a benign tumor with a potential of malignant transformation but has a certain recurrence. OBJECTIVES: A retrospective analysis of risk factors for SNIP was conducted. MATERIALS AND METHODS: Univariate analysis and multivariate logistic regression were performed to identify the independent predictors of SNIP recurrence. Nomogram and ROC curve were constructed for the prediction of the final model. Machine-learning interpretation was further provided by SHAP analysis, and a decision-tree algorithm was built to validate the predictive performance. RESULTS: A total of 171 patients were enrolled, 30 patients in recurrent group, and 141 patients in non-recurrent group, yielding a recurrence rate of 17.54%. Multivariable logistic regression identified Krouse stage and peripheral monocyte count as independent predictors of recurrence, the same as SHAP analysis. Nomogram constructed from these variables and produced an area under the ROC curve of 0.75, with 93.3% sensitivity and 83.8% specificity. A decision-tree algorithm demonstrated absence of management of the tumor attachment site, 'osteitis', and 'convoluted cerebriform pattern' achieved an predictive accuracy for recurrence. CONCLUSION AND SIGNIFICANCE: Krouse stage and peripheral monocyte count are the independent risk factors for recurrence of SNIP. For recognition of 'osteitis' and 'convoluted cerebriform pattern', the management of the tumor attachment site can substantially reduce the recurrence.