Abstract
OBJECTIVES: To explore the factors influencing hyperuricemia in breast cancer patients based on the National Health and Nutrition Examination Survey (NHANES) database. METHODS: The univariate and multivariate generalized linear regression were used to screen the influencing factors of hyperuricemia. Logistic and XGBoost algorithms were used to rank the importance of influencing factors. Receiver Operating Characteristic (ROC) curves and Decision Curve Analysis (DCA) curves were used to assess the predictive performance and clinical benefit. Trend analysis, Restricted cubic spline (RCS) analysis, and generalized additive model were used to explore the relationship between key factor and hyperuricemia. RESULTS: A total of 359 patients with breast cancer were included, of whom 99 patients had hyperuricemia. Among all variables collected, BMI, total calcium, creatinine, hypertension, and gout were found as independent factors of hyperuricemia (all p < 0.05). Among them, Both the 2 algorithms indicated that importance of creatinine on hyperuricemia ranked first. Further, BMI and creatinine levels had higher area under the curve than other variables (BMI: 0.626 [95%CI: 0.574-0.685]; creatinine: 0.722 [95%CI: 0.674-0.777]), but prediction performance difference between them was insignificant (P for Delong test = 0.051). DCA next indicated that creatinine achieved better clinical net benefit than BMI. Further, a detailed positive association between creatinine and hyperuricemia was determined (P for trend<0.001), with a linear relationship (P for non-linear = 0.428). CONCLUSION: Creatinine was identified as the most important factor of hyperuricemia in breast cancer patients, as it had independent association with hyperuricemia and favorable prediction performance.