Prognostic Survival Model Following Primary Radical Surgery for Early-Stage Cervical Cancer

Background: Radical hysterectomy with pelvic lymphadenectomy is a standard surgical procedure for early-stage cervical cancer. However, even with optimal treatment, some patients still experience disease recurrence. We aimed to develop and validate a prediction model to classify patients according to the risk of recurrence which can better assist clinicians to tailor the postoperative treatment and improve patient outcomes. Methods: Data of women diagnosed with early-stage cervical cancer who underwent radical hysterectomy were collected and analyzed. The primary outcome was recurrence-free survival (RFS). A prediction model based on Cox proportional hazard regression was developed by using the backward elimination procedure. Internal validation was performed by bootstrapping. The model's discriminative ability was demonstrated by the concordance index (C-index). The model's calibration was examined through a calibration plot. The final prognostic model was presented as a nomogram and a web-based calculator, which were further used to categorize patients into low, moderate, and high-risk groups for clinical application. Results: Among the 1309 patients, 115 (8.8%) experienced a recurrence. The median follow-up was 72.2 months. The 3-year and 5-year RFS rates were 93.0% (95% CI, 91.5-94.6%) and 90.7% (95% CI, 88.9-92.5%), respectively. Tumor size, histological subtype, number of positive lymph nodes, lymphovascular space invasion, and platelet-to-lymphocyte ratio were significantly associated with RFS. These factors were employed to construct a prediction model. The model exhibited a good overall fit with minimal overfitting and good calibration. The model's discriminative performance, as measured by the C-index, was 0.73. Conclusions: Our proposed survival model offers a potentially valuable tool for therapeutic decision-making in patients with early-stage cervical cancer. This model demonstrates robust discriminative performance and predictive calibration. Nevertheless, external validation across diverse datasets should be conducted to assess the reproducibility and applicability of this predictive model across a broader spectrum of patients.