High-Risk Clinical Characteristics and Risk Model for Unexpected Cervical Cancer Diagnosed After Surgery in Patients...

Women with colposcopy-directed biopsy results of low-grade or high-grade squamous intraepithelial lesion (LSIL/HSIL) who are subsequently found to have cervical cancer after surgery likely represent missed preoperative malignancy. This study aims to identify risk factors associated with these missed diagnoses and to develop a preoperative risk prediction model.

Patients with LSIL/HSIL on colposcopy-directed biopsy who underwent cervical conization or hysterectomy at a single center were retrospectively included. Postoperative cervical cancer was the adverse outcome. Baseline characteristics were compared using chi-square tests. Predictors were identified by univariate and multivariable logistic regression to build a logistic regression model. Internal validation applied machine learning with five-fold cross-validation. Performance was evaluated by receiver operating characteristic (ROC) curve (AUC) and accuracy, with discrimination, calibration, and clinical utility assessed using ROC curves, calibration plots, and decision curve analysis (DCA).

A total of 1271 patients were included, and 79 were diagnosed with cervical cancer after surgery. Risk factors significantly associated with postoperative cervical cancer diagnosis included age (OR, 1.0; 95% CI, 1.0-1.0, p=0.045), gravidity (OR, 1.2; 95% CI, 1.1-1.3, p <0.001), human papillomavirus 16 or 18 infection (OR, 2.6; 95% CI, 1.4-4.6, p=0.002), colposcope biopsy site numbers ≥4 (OR, 0.2; 95% CI, 0.1-0.3, p <0.001), ThinPrep Cytologic Test indicates high-grade squamous intraepithelial lesion (OR, 3.1; 95% CI, 1.1-8.5, p=0.028), and neutrophil-to-lymphocyte ratio (OR, 1.1; 95% CI, 1.0-1.3, p=0.015). In five-fold cross-validation, the machine learning algorithms yielded an AUC of 0.72-0.81 and an accuracy of 0.93-0.94 for the model. ROC curves, calibration curves, and DCA analysis demonstrate the model's excellent predictive capability and performance.

This study developed and internally validated a clinical prediction model to estimate the preoperative risk of occult cervical cancer in patients with LSIL/HSIL on colposcopy-directed biopsy. This model may support risk-stratified surgical decision-making and help reduce missed diagnoses.