A Clinical Prediction Model for Pathologic Upgrade to Invasive Carcinoma Following Conization of Cervical High-Grade Squamous Intraepithelial Lesions

Abstract

Objective: To explore the risk factors associated with the pathological progression to invasive carcinoma following the conization of cervical high-grade squamous intraepithelial lesions (HSIL) and to construct a risk prediction model to guide preoperative risk assessment and optimize the selection of surgical approaches.

Methods: A retrospective analysis was conducted on the clinical data of 3337 patients who underwent cervical conization for HSIL at Hunan Provincial Maternal and Child Health Care Hospital from December 2016 to March 2022. The patients were categorized into the pathological progression group (398 cases) and the nonprogression group (2939 cases) based on postconization pathology results. Statistical significance factors were selected by least absolute shrinkage and selection operator regression and then multivariate logistic regression was utilized to build predictive models, which were presented as a nomogram and evaluated for discriminability, calibration, and decision curves. The Bootstrap method was utilized for internal validation. A total of 277 patients were enrolled from April 2022 to October 2022 for external validation.

Results: The percentage of pathologic upgrades to invasive carcinoma following cervical conization was 11.9%. The predictive model included age, contact bleeding symptoms, HPV16/18 infection, HSIL cytology, cervical biopsy pathology diagnosis level, suspicious stromal infiltration in the biopsy pathology diagnosis, and endocervical curettage HSIL. The model demonstrated good overall discrimination in predicting the risk of HSIL progression to early invasive cancer, and internal validation confirmed its reliability (C-index = 0.787). Area under the curve analysis indicated good model discriminability across external datasets. The decision curve analysis also suggested that this model is clinically useful.

Conclusion: We developed and validated a nomogram incorporating multiple clinically relevant variables to better identify cases of HSIL progressing to early cervical cancer, providing a basis for individualized treatment and surgical approach selection.

Keywords: HSIL; cervical cancer; nomogram; pathological progression; predictive models.