Prediction of hyperuricemia in people taking low-dose aspirin using a machine learning algorithm: a cross-sectional study of the National Health and Nutrition Examination Survey

Bin Zhu; Li Yang; Mingfen Wu; Qiao Wu; Kejia Liu; Yansheng Li; Wei Guo; Zhigang Zhao

doi:10.3389/fphar.2023.1276149

Prediction of hyperuricemia in people taking low-dose aspirin using a machine learning algorithm: a cross-sectional study of the National Health and Nutrition Examination Survey

Front Pharmacol. 2024 Jan 19:14:1276149. doi: 10.3389/fphar.2023.1276149. eCollection 2023.

Authors

Bin Zhu¹, Li Yang¹, Mingfen Wu¹, Qiao Wu², Kejia Liu³, Yansheng Li³, Wei Guo^{2

4}, Zhigang Zhao¹

Affiliations

¹ Department of Pharmacy, Beijing Tiantan Hospital, Capital Medical University, Beijing, China.
² Emergency Department, Beijing Tiantan Hospital, Capital Medical University, Beijing, China.
³ DHC Mediway Technology Co., Ltd., Beijing, China.
⁴ Beijing Hospital of Traditional Chinese Medicine, Capital Medical University, Beijing, China.

Abstract

Background: Hyperuricemia is a serious health problem related to not only gout but also cardiovascular diseases (CVDs). Low-dose aspirin was reported to inhibit uric acid excretion, which leads to hyperuricemia. To decrease hyperuricemia-related CVD, this study aimed to identify the risk of hyperuricemia in people taking aspirin. Method: The original data of this cross-sectional study were obtained from the National Health and Nutrition Examination Survey between 2011 and 2018. Participants who filled in the "Preventive Aspirin Use" questionnaire with a positive answer were included in the analysis. Six machine learning algorithms were screened, and eXtreme Gradient Boosting (XGBoost) was employed to establish a model to predict the risk of hyperuricemia. Results: A total of 805 participants were enrolled in the final analysis, of which 190 participants had hyperuricemia. The participants were divided into a training set and testing set at a ratio of 8:2. The area under the curve for the training set was 0.864 and for the testing set was 0.811. The SHapley Additive exPlanations (SHAP) method was used to evaluate the performances of the modeling. Based on the SHAP results, the feature ranking interpretation showed that the estimated glomerular filtration rate, body mass index, and waist circumference were the three most important features for hyperuricemia in individuals taking aspirin. In addition, triglyceride, hypertension, total cholesterol, high-density lipoprotein, low-density lipoprotein, age, race, and smoking were also correlated with the development of hyperuricemia. Conclusion: A predictive model established by XGBoost algorithms can potentially help clinicians make an early detection of hyperuricemia risk in people taking low-dose aspirin.

Keywords: Aspirin; Hyperuricemia; National Health and Nutrition Examination Survey; eXtreme Gradient Boosting; machine learning.

Grants and funding

The authors declare that no financial support was received for the research, authorship, and of this article.