Beyond Disproportionality: Systematic Prioritization of Actionable Levothyroxine Safety Signals in FAERS.

BACKGROUND: Levothyroxine is widely prescribed, yet its real-world safety profile remain unclear. Although disproportionality analysis (DA) can detect potential safety signals, reporting raw findings without systematic prioritization may overemphasize statistical artifacts rather than clinically meaningful hypotheses. RESEARCH DESIGN AND METHODS: We analyzed FAERS Individual Case Safety Reports (2004-2023), identifying levothyroxine as the primary suspect drug. Signal detection used four DA measures: Reporting Odds Ratio (ROR), Proportional Reporting Ratio (PRR), Information Component (IC), and Empirical Bayes Geometric Mean (EBGM) with standard thresholds, including a dose-stratified comparison (≥100 µg vs <100 µg). Detected drug-event pairs were evaluated using a structured Signal Prioritization Framework integrating statistical strength, clinical seriousness, biological plausibility, and novelty to assign high, moderate, or low priority. Inter-rater reliability was assessed using weighted Cohen's κ. RESULTS: We identified 291 signals (22 labeled; 269 unexpected) and 21 dose-dependent signals. Fifteen signals were classified as high priority , while high-dose therapy identified additional moderate-priority signals, including acute kidney injury.Inter-rater agreement was substantial (κ = 0.63; SE = 0.092; 95% CI 0.45-0.81). CONCLUSION: A structured approach combining FAERS-based signal detection with systematic prioritization distinguished clinically plausible levothyroxine safety signals from statistical noise, supporting targeted pharmacoepidemiologic evaluation and improved clinical interpretation.