Abstract
OBJECTIVE: This study aimed to develop a preoperative nomogram based on clinical and pathological characteristics to provide a more individualized and accurate estimation of lymph node metastasis (LNM) in patients with early-stage cervical cancer. METHODS: A total of 7,349 early-stage cervical cancer patients with pathologically confirmed between 1988 and 2015 were obtained from the Surveillance, Epidemiology, and End Results (SEER) database. All the patients were divided into training (n = 5,500) and validation (n = 1,849) cohorts randomly. A cohort of 455 patients from multicenter was used for the external validation. We established a multivariate logistic regression model based on preoperative clinicopathological data, from which a nomogram was developed and validated. A predicted probability of LNM < 5% was defined as low risk. RESULTS: From multivariate logistic regression analysis, age at diagnosis, histologic subtype, tumor grade, tumor size and FIGO stage were identified as preoperative independent risk factors of LNM. The nomogram incorporating these factors demonstrated good discrimination and calibration (concordance index = 0.723; 95% confidence interval (CI), 0.707-0.738). In the validation cohort, the discrimination accuracy was 0.745 (95% CI, 0.720-0.770) and 0.747 (95% CI, 0.690-0.804), respectively. The nomogram was well calibrated with a high concordance probability. We also established an R-enabled Internet browser for LNM risk assessment, which tool may be convenient for physicians. CONCLUSIONS: We developed an effective preoperative nomogram based on clinical and pathological characteristics to predict LNM for early-stage cervical cancer. This model could improve clinical trial design and help physicians to decide whether to perform lymphadenectomy or not.