Prediction model for residual high-grade cervical intraepithelial lesions

• DOI: 10.23736/s2724-606x.22.04991-0
• Publication Date: 2022-02-02T14:56:44Z
• Authors: Suna YILDIRIM KARACA; Mehmet ADIYEKE; Onur INCE; Alper ILERI; Tayfun VURAL; Alkim G. ŞAHINGÖZ...; Ayşe R. ŞENKAYA; Sertaç BULUT; Sercan KANTARCI; Muzaffer SANCI
• Keywords: Neoplasm, Residual; Papillomavirus Infections; Electrosurgery; Uterine Cervical Neoplasms; Margins of Excision; Cervix Uteri; Uterine Cervical Dysplasia; 3. Good health; 03 medical and health sciences; 0302 clinical medicine; Humans; Female; Neoplasm Recurrence, Local

The aim of the study was to evaluate risk factors associated with high-grade cervical intraepithelial lesions (HSIL) in patients undergoing a second cervical excision procedure due to positive surgical margins and to create a prediction model for residual disease.This study included patients with HSIL positive surgical margins following loop electrosurgical excision procedures (LEEP) between March 2015 and August 2019. HSIL in the second cervical excision pathology in these patients was accepted as residual disease. For residual disease prediction, a multivariate logistic regression and stepwise elimination analysis of 14 variables including demographic characteristics, clinical characteristics, pathology results and HPV genotypes of the patients was performed.Second cervical excision procedures were performed in 290 patients 85(29.4%) of these patients had CIN 2 (cervical intraepithelial neoplasia) and 205 (70.6%) had CIN 3. In the second excision procedure, 166 patients (57.2%) had ≤CIN 1, 124 patients (42.8%) had ≥CIN2. The prediction model of residual disease includes only 3 variables out of the 14 different clinical characteristics (AUC=0.605 [0.539-0.671]). These variables are gravida (adjusted OR: 1.15 [0.97-1.38], P=0.107), CIN2-3 presence in the endocervical canal in the first LEEP specimen (adjusted OR: 1.52 [0.94-2.47], P=0.091) and the presence of HR-HPV except 16/18 lesions (adjusted OR: 0.64 [0.38-1.06], P=0.083).A prediction model was designed with our data, from variables reported to be risk factors for residual disease in previous studies. While this model was statistically significant, it was poor at distinguishing residual disease. A prediction model can be designed to guide clinicians with future studies.