Nomogram-based prediction of cervical dysplasia persistence/recurrence.

The widespread introduction of screening methods allow to identify cervical dysplasia before having invasive cancer. The risk of developing cervical dysplasia persistence/ recurrence following conization represent a major health issue. Although several studies tried to identify predictors for cervical dysplasia persistence/recurrence, no previous study has been conducted to develop a risk calculator. The current study aimed to identify predictors of cervical dysplasia persistence/recurrence among women undergoing primary conization. We aimed to build nomograms estimating the risk of developing cervical dysplasia recurrence. Data of consecutive women with diagnosis of high-risk human papillomavirus (HPV) undergoing conization were retrospectively evaluated (1503 patients). The risk of developing cervical dysplasia persistence/recurrence was assessed with Kaplan-Meier and Cox's hazard models. Additionally, two nomograms were built to estimate likelihood of cervical dysplasia recurrence: the first based on baseline and operative parameters and the second focusing on type-specific HPV detected. The performance of the above nomograms was assessed using concordance index. A total of 1503 patients were analyzed. After a mean (SD) follow-up of 48.6 ( ± 17.5) months, 84 (5.6%) patients required secondary conization. By multivariate analysis, HIV infection [hazard ratio (HR): 7.78; 95% confidence interval (CI): 2.77-21.81; P < 0.001], positive margins (HR: 26.2; 95% CI: 14.1-48.71; P < 0.001) and persistence of HPV (HR: 6.82; 95% CI: 4.15-11.21; P < 0.001) correlated with cervical intraepithelial neoplasia 2+ persistence/recurrence. The importance of those variables was corroborated by our first nomogram. The second nomogram suggested the impact of type-specific HPV infection in predicting cervical dysplasia persistence/ recurrence. HPV16, HPV18, HPV33, HPV35 and HPV45 were the HPV types most commonly associated with cervical dysplasia persistence/recurrence. The concordance index was greater than 0.70 for both nomograms, thus suggesting the reproducibility of our models. We developed the first two nomograms predicting this risk. The findings of this study require external validation. Once validated our data might be useful to plan a tailored postoperative surveillance of women receiving primary conization.