The authors applied four disproportionality metrics to FDA Adverse Event Reporting System data covering 2014 to 2026 (18.37 million reports), across five gemcitabine exposure groups relevant to biliary tract cancer and pancreatic ductal adenocarcinoma. Myelosuppression was the most consistent signal (ROR 3.35 to 7.74), and thrombocytopenia in the disease-specific subset had an ROR of 19.36, reproduced at 19.70 after deduplication. The triple-combination liver toxicity signal (ROR 10.32) collapsed to 1.62 with zero cases when restricted to primary-suspect reports.
Why it is interesting: A worked example of how fragile spontaneous-reporting signals are once case definitions are restricted, in a cancer setting where toxicity bears on treatment tolerability.