Large-Scale Signal Detection and Personalized Risk Prediction for Cardiovascular Drug Adverse Reactions in Elderly Patients: Real-World Evidence from Western China.

Abstract

BACKGROUND: In the context of global ageing, the incidence of adverse drug reactions (ADRs) with use of cardiovascular drugs in elderly patients is increasing, with important implications for the health and well-being of elderly people as well as the societal burden on healthcare systems. There is an urgent need for systematic analysis of cardiovascular ADR data and the identification of risk signals to enhance personalized risk prediction and guidance for the rational use of cardiovascular drugs in elderly patients. METHODS: We collected data from the ADR reporting system of the Shaanxi Provincial Centre for Adverse Drug Reaction Monitoring in China. We conducted statistical analysis using 21,501 ADR data records related to the use of cardiovascular drugs by elderly patients from 2018 to 2023. We applied disproportionality analysis to identify drug risk signals without performing a case-by-case analysis. We used restricted cubic spline analysis to quantify the relationship between patient age and the risk of ADRoccurrence. Additionally, association rule mining was applied to analyse multidimensional association patterns among patient characteristics, drugs, and ADRs. We explored the fundamental causes of ADR occurrence and proposed corresponding medication recommendations and preventive measures. RESULTS: Signal mining results showed that 102 effective risk signals were identified at System Organ Class level. The strongest signal was associated with endocrine disorders caused by Class III antiarrhythmics (reporting odds ratio = 1023.41, information component = 5.11); nitrates and nitrites for angina pectoris causing nervous system disorders occurred most frequently (2819 cases). We identified 321 effective risk signals at High-Level Term level. Angiotensin-converting enzyme (ACE) inhibitors causing cough occurred most frequently (1104 cases); statins caused various ADRs such as hepatic dysfunction and muscular injury. Restricted cubic spline analysis showed that the risk of patients experiencing headaches when using nitrates and nitrites was highest at age 70.28 years (p < 0.001). The risk of coughassociated with use of ACE inhibitors was high after age 71.58 years (p < 0.05). Through multidimensional association rule mining, four significant strong association rules were identified; that regarding female patients experiencing coughing when using ACE inhibitors had the highest confidence (0.676) and lift (11.226), indicating a greater risk of ADRs. CONCLUSION: We comprehensively analysed cardiovascular ADR data in elderly patients, revealing potential drug safety issues and establishing individualized ADR warnings based on patient characteristics. It should be noted that the disproportionality analysis is a hypothesis generating approach and insufficient for causal inference. Our findings will help optimize clinical treatment plans for cardiovascular drugs and promote the rational use of cardiovascular drugs in elderly patients to achieve healthy ageing.