Impact of safety alerts on measures of disproportionality in spontaneous reporting databases - The notoriety bias

Background: Disproportionality analysis of spontaneous reporting is increasingly used, but it may be influenced in unknown ways by safety alerts (notoriety bias). Objective: To explore the consequences of safety alerts on reporting disproportionality. Methods: Within the French national pharmacovigilance database, disproportionality of reporting was tested, using the reporting odds ratio (ROR) and its 95% confidence interval, before and after four safety alerts: valvulopathies with pergolide; tuberculosis with infliximab; strokes with atypical antipsychotics; and rhabdomyolysis with HMG-CoA reductase inhibitors (statins) [after cerivastatin withdrawal]. Results: No cases of valvulopathy were reported in association with pergolide before the safety alert and 63 cases were reported after the alert, (ROR 9400; 95% CI 4300, 20 000), of which five had occurred before the alert. Twenty-five reports mentioned rhabdomyolysis associated with statins (not including cerivastatin) before the safety alert (ROR 5.8; 95% CI 3.8, 9.0), and 63 did so after the alert (ROR 9.4; 95% CI 7.0, 13.0). Approximately 280 cases involving cerivastatin were reported after its withdrawal. There were two reports of tuberculosis associated with infliximab before the alert (ROR 1500; 95% CI 130, 18 000) and seven after the alert (ROR 430; 95% CI 110, 1700). There was one report of a stroke in association with atypical antipsychotic treatment before the safety alert (ROR 0.10; 95% CI 0.01, 0.63) and 16 after the alert (ROR 1.10; 95% CI 0.70, 1.90). After excluding events involving treatment with anticoagulant agents, the RORs for stroke in association with atypical antipsychotic treatment were 0.14 (95% CI 0.02, 1.00) before the alert and 2.0 (95% CI 1.2, 3.4) after the alert. Conclusion: Disproportionality in spontaneous reporting databases increases after a safety alert because of increased reporting of the event of interest, including reports of such events that occurred before the alert. This may overflow to increased reporting of the event in association with other drugs.