Risk Prediction Models Of Occurred VAIN In Patients Accepted Conization For Cervical HSIL Within 2 Years

Objective:Vaginal intraepithelial neoplasia(VAIN)is a group of atypical hyperplastic lesions of vaginal squamous epithelium closely related to high-risk human papillomavirus(HR-HPV)infection,which is considered as a precursor of invasive vaginal carcinoma.In recent years,with the popularization of cervical cancer screening and the progress of colposcopy,the detection rate of VAIN has gradually increased,but the clinical awareness of VAIN is still inadequate.Cervical lesions are closely related to vaginal intraepithelial neoplasia.Studies of VAIN after conization are rare.The purpose of this study was to investigate the risk factors of occurred VAIN in patients accepted conization(including cold knife conization and loop electrosurgical excision procedure)within 2 years for high-grade squamous intraepithelial lesion(HSIL)in cervix and established clinical predictive models to predict the risk probability of occurrence of VAIN and provide decision-making basis for follow-up,diagnosis and treatment of these patients after conization.Methods:We collected the clinical data of patients who underwent cervical conization(including cold-knife conization and loop electrosurgical excision procedure)for high-grade cervical intraepithelial neoplasia within 2 years.The risk factors related to the occurrence of vaginal intraepithelial neoplasia were analyzed,including age,menopause status,gravidity,parity,contraceptive method preoperative cervical thin rip cytologic test(TCT)results,preoperative human papillomavirus infection,cervical transformation area type,pathological results,glands involvement,cervical cytological results and HPV infection and follow-up after operation in 6 months and 12 months after respectively.The binary logistic regression model was used for univariate analysis of the involved factors,and the OR,95%CI and P values were recorded.P<0.05 meant that the difference was statistically significant.A clinical prediction model was constructed by the binary logistic regression analysis,and was divided into training set and test set in a ratio of 4:1 for internal verification.The model differentiation was tested by receiver operating curve(ROC),calibration was tested by calibration curve and Brier score(BS),clinical decision curve was used to evaluate the clinical value of the model.Finally,a personalized nomogram prediction model was used to present the model.Results:We collected clinical data of 5125 patients who underwent cervical conization due to cervical HSIL in our hospital from January 1,2013 to December 31,2019.A total of 99 patients occurred VAIN within 2 years after cervical conization.The incidence of occurred VAIN within 2 years after conization for HSIL was about 1.93%.A total of 495 patients were randomly matched from the remaining patients for follow-up according to 1:5.A total of 93 patients with lost follow-up,hysterectomy immediately after surgery and no postoperative review were excluded,and 402 patients did not occur VAIN within 2 years after conization.Univariate analysis screened out 5 preoperative factors and 3 postoperative follow-up factors were related to the occurrence of VAIN within 2 years after conization,including age≥ 46 years,menopause,type 3 transformation area,combined with VAIN,condom contraception,HR-HPV positive within 1 year after conization,the result of TCT is atypical squamous cells of undetermined significance and even worse(TCT≥ASC-US)within 1 year after conization,HR-HPV positive or TCT≥ASC-US within 1 year after conization.Using these predictors,a total of 4 clinical prediction models were established,in which only preoperative factors were included in model 1,preoperative factors and HR-HPV positive within 1 year after conization were included in model 2,preoperative factors and TCT≥ ASC-US within 1 year after conization were included in model 3,preoperative factors and HR-HPV positive or TCT≥ASC-US within 1 year after conization were included in model 4.R software was used to draw the receiver operating characteristic curve and calculate the area under the curve to evaluate the model differentiation.The area under the curve(AUC)of models 2-4 was significantly improved compared with model 1,and no matter in the training set or test set,model 4 with HR-HPV positive or TCT≥ASC-US combined screening results within 1 year after surgery has the best AUC.The sensitivity and specificity were 90.9%and 85.1%respectively.The calibration graph was used to evaluate the model calibration.All models were close to the solid line with slope of 1,and Brier score<0.25 showed no significant difference between the predicted probability and the actual probability of all models.Finally,the decision curve analysis was used to analyze and evaluate the clinical value of the model.No matter in the training set or test set,the area under the curve of model 2-4 was larger than that of model 1 and model 4 had the largest area under the curve and the best clinical value.All the four models could well predict the probability of VAIN occurrence within 2 years after conization,and the prediction ability of model 2-4 with the addition of postoperative follow-up factors was stronger than that of model 1.Among them,model 4 with the combined screening results of HPV and TCT within 1 year after surgery had the strongest prediction ability,and the model showed good performance in calibration degree and clinical value.Finally,the model is presented by a personalized nomogram diagram.Conclusion:1.In this study,statistically significant high risk factors related to the outcome of VAIN within 2 years after conization are age≥46 years,menopause,type 3 transformation area,combined with VAIN,non-condom contraception or no contraception,HR-HPV positive within 1 year after conization,TCT≥ASC-US within 1 year after conization,HR-HPV positive or TCT≥ASC-US within 1 year after conization.2.In this study,a total of 4 prediction models for VAIN occurrence within 2 years after cervical conization were established,which could better predict the probability of VAIN occurrence 2 years after conization.3.The predictive ability of models 2-4 was significantly improved compared with model 1.No matter what postoperative screening method was adopted,HPV detection and/or cytological detection should be carried out within 1 year after surgery.4.Model 4 has the strongest predictive ability,and the combined screening of HPV and cytological detection is the most valuable in postoperative follow-up,which can make up for the shortcomings of single screening method and improve the predictive ability.