Risk-stratified surveillance after LEEP: a nomogram integrating HPV persistence, margin status, and clinical factors to predict CIN2+ recurrence

Background Cervical intraepithelial neoplasia (CIN) recurrence after loop electrosurgical excision procedure (LEEP) remains a clinically consequential barrier to cervical cancer prevention, and risk stratification tools tailored to real-world practice are limited in China. This study developed and internally validated a clinical prediction nomogram for histologically confirmed CIN2+ recurrence after LEEP. Methods A retrospective single-center cohort was assembled of women treated with LEEP for CIN2+ between January 2018 and October 2024. Candidate predictors included demographic and reproductive factors, smoking, HPV vaccination, prior cervical treatment, transformation zone type, LEEP pathology (including adenocarcinoma in situ [AIS] and margin status), pre-/post-treatment high-risk HPV measures, and neutrophil-to-lymphocyte ratio (NLR). Time-to-recurrence was analyzed using Cox regression with hierarchical domain modeling. A nomogram was constructed from the final multivariable model and evaluated for discrimination and calibration. Results Among 2,230 women (median follow-up 31.8 months, IQR 19.6–43.5), 334 developed CIN2+ recurrence (15.0%), with a median time to recurrence of 15.6 months (IQR 8.2–24.3). Persistent HPV infection occurred in 50.6% of women with recurrence versus 23.1% without recurrence ( p < 0.001). Persistent HPV infection (same genotype pre-/post-LEEP) was the strongest independent predictor (adjusted hazard ratio [aHR] 2.51, 95% CI 1.99–3.16). Additional independent predictors included unvaccinated status (aHR 1.54, 95% CI 1.08–2.20), multiple positive margins (aHR 1.52, 95% CI 1.08–2.14), AIS versus CIN2 (aHR 1.48, 95% CI 1.03–2.12), prior cervical treatment (aHR 1.38, 95% CI 1.04–1.84), single positive margin (aHR 1.38, 95% CI 1.02–1.87), and higher NLR (per one-unit increase: aHR 1.21, 95% CI 1.02–1.44). Model discrimination increased across hierarchical models from 0.516 (model 1) and 0.562 (model 3) to 0.619 in the final model. Risk stratification separated low-, intermediate-, and high-risk groups with observed 24-month recurrence rates of 6.2%, 14.8%, and 31.5%, respectively ( p for trend