Analyzing Medicare fee-for-service claims (2017–2019) from 70,130 statin-naive patients aged 65+ hospitalized for myocardial or cerebral infarction, a claims-based pharmacovigilance framework using sequential target trial emulation identified five primary adverse drug event signals for atorvastatin initiation versus initiating a different medication: valve disorders (sHR 1.71), sprains and strains (sHR 1.79), general sensation/perception symptoms (sHR 1.23), abnormal findings without diagnosis (sHR 1.55), and prediabetes (sHR 1.71). Sensitivity analyses additionally flagged hemorrhagic stroke, posthemorrhagic anemia, varicose veins, and other circulatory conditions.
The prediabetes signal carries particular clinical weight — statin-associated dysglycemia is an established pharmacological concern, and detecting it through automated claims analysis validates this methodology against known biology. More provocative are the valve disorder and hemorrhagic stroke signals, which warrant mechanistic scrutiny before clinical action. Statins' pleiotropic effects on coagulation and vascular tone could plausibly underlie circulatory findings, yet confounding remains a serious concern even after inverse probability weighting in this observational design.
Critically, this is a preprint posted on medRxiv and has not yet undergone peer review — effect estimates and signal classifications could change substantially. The real contribution here is methodological: demonstrating that target trial emulation, previously used for causal estimation, can be systematically scaled across 552 outcome categories for proactive signal detection. For older adults chronically underrepresented in drug trials, this framework could meaningfully complement passive spontaneous reporting systems. Incremental as an atorvastatin finding; potentially paradigm-shifting as a surveillance architecture.