Minji Kwon 1, Jongdae Han 2, Sumin Chae 3, Bo-Hyung Kim 1
1 Department of Clinical Pharmacology and Therapeutics, Kyung Hee University College of Medicine, Kyung Hee University Hospital (, Republic of Korea), 2 Department of Statistics and Data Science, Korea National Open University (, Republic of Korea), 3 Department of Surgery, Kyung Hee University College of Medicine, Kyung Hee University Hospital (, Republic of Korea)
Introduction / Objectives
Using data from the U.S. Food and Drug Administration Adverse Event Reporting System (FAERS), a comprehensive assessment of drugs reported for myotoxicity, nephrotoxicity, and hepatotoxicity identified lipid-lowering agents among the most frequently implicated medications. Lipid-lowering therapies, including statins, are widely prescribed worldwide for the management of dyslipidemia and for cardiovascular risk reduction. Although these agents are generally regarded as safe and well tolerated, adverse event reports related to these toxicities continue to accumulate in pharmacovigilance databases. Given their widespread use, even rare but severe toxicities may have substantial public health implications. Identifying predictors of serious myotoxicity, nephrotoxicity, and hepatotoxicity among patients receiving lipid-lowering therapy is therefore of considerable clinical and regulatory relevance.
Methods
This retrospective pharmacovigilance study utilized the FAERS database. Cases were included if a lipid-lowering agent was reported as the primary suspect drug. Drug names were standardized and mapped to Anatomical Therapeutic Chemical (ATC) codes to ensure consistent identification of active ingredients. The included agents were atorvastatin (C10AA05), rosuvastatin (C10AA07), simvastatin (C10AA01), pravastatin (C10AA03), fluvastatin (C10AA04), pitavastatin (C10AA08), lovastatin (C10AA02), ezetimibe (C10AX09), fenofibrate (C10AB05), and gemfibrozil (C10AB04).
Myotoxicity, nephrotoxicity, and hepatotoxicity were defined as outcomes using Standardised MedDRA Queries (SMQs). Myotoxicity was captured using the rhabdomyolysis/myopathy SMQ (20000002). Nephrotoxicity was defined using the SMQs for acute renal failure (20000003) and chronic kidney disease (20000213). Hepatotoxicity was defined using the SMQs for liver-related investigations, signs and symptoms (20000008), cholestasis and jaundice of hepatic origin (20000009), hepatitis, non-infectious (20000010), and Hepatic failure, fibrosis and cirrhosis, and other liver damage-related conditions (20000013).
Covariates included demographic variables (age and sex), reporter characteristics, and medication-related factors. Concomitant drugs were categorized according to their role in the report (e.g., secondary suspect and interacting drugs), and summary measures, such as the total number of reported drugs, concomitant medications, and interacting medications were constructed to capture polypharmacy burden and clinical complexity. Indications were grouped into clinically meaningful categories (e.g., lipid disorders, cardiovascular disease, diabetes and metabolic conditions, renal disease, hepatic/biliary disease, autoimmune disorders, and cancer) to account for underlying comorbidities.
To avoid drug-specific attribution, individual lipid-lowering drug ingredients were excluded from the primary models as predictors. The dataset was temporally split into a training set (2004–2022) and a validation set (2023–2025). Logistic regression, least absolute shrinkage and selection operator (LASSO), decision tree, and random forest models were applied. Model performance was evaluated using the area under the receiver operating characteristic curve and precision-recall-based metrics in the validation set.
Results
A total of 122,420 FAERS cases were included in the analysis. Based on SMQ definitions, 9,096 cases were identified with myotoxicity, 5,399 with nephrotoxicity, and 12,529 with hepatotoxicity; 21,890 cases met at least one outcome definition. Older age and a higher burden of polypharmacy, reflected by increased numbers of concomitant and interacting medications, were consistently associated with myotoxicity, nephrotoxicity, and hepatotoxicity across models. Indicators of overall clinical and reporting burden, including the total number of reported drugs and reaction terms, were also strongly associated with adverse outcomes.
In the temporal validation period, event proportions were higher than in the earlier training period. Overall model performance demonstrated moderate discrimination across outcomes. Among the evaluated models, random forest consistently achieved the best predictive performance for all three outcomes. Precision–recall analysis also indicated predictive ability beyond baseline event rates.
Conclusions
Among FAERS reports involving lipid-lowering agents, SMQ-defined myotoxicity, nephrotoxicity, and hepatotoxicity occurred frequently and were moderately predictable from case-level demographic and pharmacological features. Models excluding individual lipid-lowering drug ingredients suggest that polypharmacy burden and interacting medications may contribute more to the prediction of serious adverse event reports. These findings highlight the importance of comprehensive medication review and interaction assessment for risk stratification in pharmacovigilance settings.
References:
[1] Morris R, Bu K, Han W, Wood S, Hernandez Velez PM, Ward J, Crescitelli A, Martin M, Cheng F. The association between statin drugs and rhabdomyolysis: an analysis of FDA Adverse Event Reporting System (FAERS) data and transcriptomic profiles. Genes (Basel).
[2] Zhang J, Guo Y, Wei C, Yan Y, Shan H, Wu B, Wu F. A pharmacovigilance study of chronic kidney disease in diabetes mellitus patients with statin treatment by using the US Food and Drug Administration adverse event reporting system. Frontiers in Pharmacology. 2024;15.
[3] Zhou L, Wu B, Bian Y, Lu Y, Zou Y, Lin S, Li Q, & Liu C (2025). Hepatotoxicity associated with statins: A retrospective pharmacovigilance study based on the FAERS database. PLoS One, 20(7):e0327500
Reference: PAGE 34 (2026) Abstr 12032 [www.page-meeting.org/?abstract=12032]
Poster: Methodology – AI/Machine Learning