(175) Distinguishing Patterns Associated with Statin-Related Myopathy Using Korea Adverse Event Reporting System Database: Unsupervised Machine Learning Methods

Publication Number: 506

Background: Several studies have reported that drug-drug interactions are related to the risk of statin-related myopathy. However, the evidence in Korea is limited.

Objectives: We aimed to identify relevant patterns for patients with statin-related myopathy using Korea Adverse Event Reporting System (KAERS) database in Korea.

Methods: We performed an unsupervised machine learning methods using Korea Adverse Event Reporting System (KAERS) database between 2016 and 2020. We included reports including any statin. We defined myopathy cases using ‘myopathy’ and ‘rhabdomyolysis’ among MedDRA SMQ, and mapped World Health Organization Adverse Reactions Terminology (WHO-ART). We conducted a hierarchical clustering including reports characteristics, that is, gender, age, statin type, contraindicated drugs, and concomitant drugs in the individual case safety reports (ICSRs). The reporting characteristics and myopathy patterns in accordance with clusters were described.

Results: Among 724 spontaneous reports, three clusters were identified. Cluster 1 (n =203) included reports involving contraindicated co-medications and a high proportion of pitavastatin and pravastatin. Cluster 2 (n=205) included no contraindicated comedications and showed a high proportion of rosuvastatin. Cluster 3 (n=316) was mainly composed of atorvastatin. Mean age was lowered in cluster 1 (62.9 years) than other clusters (65.2 and 65.0 years in cluster 2 and cluster 3). Rhabdomyolysis was reported more frequently in cluster 2 compared to other clusters (4.3%, 8.6%, 7.7% in cluster 1, 2, 3, respectively). Cluster 3 showed a higher frequency of consumers or other healthcare professionals (4.4%, 14 reports) than other clusters. PTs that reported more than 10% in each cluster were ‘blood creatine phosphokinase MM increased’, ‘blood creatine phosphokinase increased’, ‘hypercreatininaemia’ in cluster 1, ‘muscular weakness' in cluster 2, ‘chromaturia’, ‘hypercreatininaemia’, and ‘muscular weakness’ in cluster 3, and 'acute kidney injury' in all three clusters.

Conclusions: Using hierarchical clustering, we found three distinct clusters based on reporter characteristics. It is expected that these findings can be used as evidence for the usefulness of approaching clusters in the KAERS database. However, because other factors including demographics and comorbidities may affect AEs, cautious interpretation was required.