Active pharmacovigilance that moves beyond spontaneous reporting can detect signals but requires rigorous confirmation. This study applied sequential target‑trial emulation methods to confirm or refute previously detected adverse drug event (ADE) signals associated with atorvastatin initiation among older adults. The objective was to use refined outcome definitions, expanded exclusions, within‑outcome false discovery control, and probabilistic quantitative bias analysis to produce robust, pre‑specified confirmation within a sequential emulation framework.
The analysis used Medicare fee‑for‑service claims from 2017–2019. Eligible participants were statin‑naïve beneficiaries aged 65 years or older who were hospitalized with a primary diagnosis of myocardial infarction or cerebral infarction, had a length of stay of at least 3 days, and were discharged home. In total 70,130 eligible patients were identified: 39,948 initiated atorvastatin and 19,182 initiated another new outpatient medication.
Up to 14 nested daily trials (Trials 0–13) were constructed beginning on each patient’s hospital discharge date. Eligibility, treatment assignment, and follow‑up were synchronized at each trial origin to remove immortal time bias. The primary analysis stacked all eligible trials to estimate effects across the sequence. A pre‑specified sensitivity analysis restricted inference to Trials 0 and 1, which produced superior covariate balance (maximum standardized mean difference < 0.1).
Three treatment strategies were encoded: initiation of atorvastatin (strategy A1), initiation of any other new outpatient medication (strategy A2), and no new medication (strategy A0), although A0 was retained only to preserve sequential eligibility. The study estimated per‑protocol effects after applying inverse‑probability‑of‑treatment and inverse‑probability‑of‑censoring weights. Artificial censoring handled treatment deviation with a 30‑day grace period; death was treated as a competing risk and modeled with Fine–Gray subdistribution hazard models.
Previously detected signals were re‑specified using more granular and clinically coherent outcome definitions within outcome families: hemorrhagic events, cardiac valve disorders, musculoskeletal injuries, sensory symptoms, abnormal laboratory findings, and hyperglycemic events. Definitions relied on Clinical Classifications Software Refined categories and independently validated Sentinel or published algorithms. Incident events required absence of relevant baseline diagnostic codes.
Confirmation required two simultaneous criteria: (1) a within‑outcome Benjamini–Hochberg q ≤ 0.05 with an estimated subdistribution hazard ratio (sHR) > 1.0; and (2) robustness to probabilistic quantitative bias analysis such that both the median and the 2.5th percentile of the bias‑adjusted sHR remained > 1.0 across 5,000 Monte Carlo draws. The bias analysis varied an unmeasured confounder’s confounder–outcome risk ratio from 1.25 to 3.00 and the prevalence difference from 0.05 to 0.25.
The study reported absolute risks, risk differences, and numbers needed to harm (NNH) for confirmed associations. Stratified analyses examined three post‑discharge time windows (1–30, 31–91, and 92–182 days) and subgroup effects by age, sex, and race.
After weighting, baseline covariates were closely balanced between treatment groups. The primary stacked‑trial analysis and the Trials 0–1 restricted sensitivity analysis were both used to assess robustness; Trials 0–1 achieved the best covariate balance.
Several ADE signals met the dual confirmation criteria. Notable confirmations included:
Acute hemorrhagic cerebrovascular disease overall (sHR 1.43; 95% CI 1.00–2.04), with a reported risk difference of 0.5% and an NNH of 205. The association was stronger in the first 30 days after discharge (sHR 2.20; 95% CI 1.35–3.58) and persisted when inference was restricted to Trials 0 and 1 (sHR 1.50; 95% CI 1.02–2.20). Related early intracranial hemorrhage signals were also confirmed.
Cardiac valve disorders: nonrheumatic and unspecified valve disorders were confirmed in days 92–182 (sHR range 1.48–1.58), and cardiac valve intervention was confirmed overall (sHR range 1.74–1.83).
Musculoskeletal injuries: composites including sprains and strains were confirmed among men (sHR range 1.66–1.94).
Sensory symptoms: general sensation/perception symptoms and dizziness were confirmed among non‑White patients (sHR 1.40–1.43) in the unrestricted set of trials.
Hepatobiliary events: acute hepatic failure was confirmed overall (sHR range 1.61–1.72), and biliary tract disease was confirmed among women (sHR range 1.45–1.49).
For every confirmed association, the proportion of bias‑adjusted Monte Carlo draws that remained above the null was 1.00, indicating robustness under the specified bias scenarios.
Multiple prior signals did not meet the dual confirmation criteria when subjected to refined definitions, multiplicity control, and bias analysis. Examples reported include prediabetes and acute posthemorrhagic anemia, which failed the confirmation thresholds.
The study shows that sequential target‑trial emulations with refined outcome definitions, within‑outcome multiplicity control, restriction to optimally balanced early trials, and probabilistic quantitative bias analysis can confirm several ADE signals associated with atorvastatin initiation in older adults. The most consistent findings were early hemorrhagic cerebrovascular events, late cardiac valve disorders and interventions, musculoskeletal injuries in men, and selected hepatobiliary events. Absolute excess risks were modest but considered clinically relevant in a high‑risk post‑infarction population. The authors suggest a two‑stage active pharmacovigilance paradigm—signal detection followed by rigorous confirmation—and recommend heightened clinical vigilance for the confirmed events while calling for external validation in independent populations and data sources.
This report is a preprint and has not been peer reviewed; the authors note the need for external validation. Data were obtained under a Centers for Medicare & Medicaid Services Data Use Agreement and are not publicly available due to privacy and contractual restrictions; qualified researchers may request access via ResDAC and execute their own data use agreements. The manuscript reports that relevant ethical approvals were obtained and that the authors declared no competing interests.