Establishment and validation of a prediction model for early-stage epithelial ovarian cancer based on LASSO regression

Abstract

Objective: To establish and validate a prediction model for early-stage epithelial ovarian cancer based on least absolute shrinkage and selection operator (LASSO) regression. Methods: A total of 509 cases ovarian mass patients who underwent surgical treatment in Tianjin Medical General Hospital from January 2018 to March 2023 were retrospectively analyzed. The patients were randomly divided into modeling group [n=356, M(Q1,Q3) for age were 43 (31, 61) years] and internal validation group [n=153, age 42 (31, 60) years] by 7∶3 ratio. In addition, 86 patients [age 44 (33, 61) years] who underwent surgical treatment for ovarian mass in Tianjin Medical University General Hospital from April to November 2023 were collected as external validation group. The variables were screened by LASSO regression. The nomogram model was established and plotted by multivariate logistic regression. Internal and external validation were then conducted. The model performance and clinical applicability were evaluated using receiver operating characteristic (ROC) curve, calibration curve and decision curve. Results: Five variables including age (OR=1.040,95%CI:1.000-1.050,P=0.002), carbohydrate antigen 125 (CA125) (OR=1.001, 95%CI: 1.000-1.010, P=0.017), human epididymis protein 4 (HE4) (OR=1.020, 95%CI: 1.000-1.030, P=0.002), carbohydrate antigen 199 (CA199) (OR=1.001, 95%CI:1.000-1.020, P=0.023) and lactate dehydrogenase (LDH) (OR=1.020, 95%CI: 1.010-1.022, P=0.001) were screened as risk factors for early-stage epithelial ovarian cancer. The nomogram model was constructed based on these above five risk factors to predict early-stage epithelial ovarian cancer. ROC curves showed the area under curve (AUC) were 0.915(95%CI:0.910-0.932)for modeling group, 0.891(95%CI:0.874-0.905) for internal validation group, and 0.924(95%CI:0.907-0.942) for external verification. The calibration curves and clinical decision curves showed the model exhibited good consistency and clinical practicability. Conclusions: The nomogram model built includes age, CA125, HE4, CA199, and LDH. It can effectively predict early-stage epithelial ovarian cancer and has strong clinical practicability.