Impact of notoriety bias on reporting of pancreatitis with antidiabetics: the case of exenatide

Background. The notoriety bias (i.e. increased reporting of adverse drug events following safety alerts) may significantly impact on signal detection in spontaneous reporting systems (Pariente et al., 2007). This study aimed at evaluating the contribution of this bias on reporting of pancreatitis with antidiabetics, with exenatide as the main drug of interest. Methods. A case-control study was performed on the FDA_AERS (Adverse Event Reporting System) database to analyze the association between antidiabetic drugs and pancreatitis. The analysis covered the period 2005 (exenatide was launched on the market in April 2005) through 2009. The base cohort consisted of Adverse Drug Reactions (ADRs) for antidiabetics. Cases were represented by reports of pancreatitis according to the Medical Dictionary for Regulatory Activities (MedDRA), whereas all other ADRs reported with antidiabetics were considered controls. Disproportionality analysis was conducted by calculating Reporting Odds Ratios (RORs) with corresponding 95%CI, thus estimating the strength of the association. Significant disproportionality was formally defined when the lower limit of the 95%CI was >1. The contribution of notoriety bias was evaluated by performing a quarter-by-quarter analysis. Results. Exenatide was reported in 709 cases of pancreatitis, followed by metformin (630), insulin (349), glimepiride (148), pioglitazone (146), glibenclamide (142), sitagliptin (128), rosiglitazone (126) and glipizide (114). The cumulative disproportionality analysis (2005-2009) found significant disporportionality only for exenatide (ROR=1.71; 95%CI=1.57-1.87) and sitagliptin (ROR=1.82; 95%CI=1.52-2.18). For exenatide, the quarter-by-quarter analysis revealed that disproportionality signal appeared in the fourth quarter (Q4) of 2007, rapidly increased with a first peak in Q3 of 2008, remained statistically significant and reached a second peak in Q4 of 2009. Therefore, reporting was strongly influenced by the FDA alert first issued in October 2007 and updated in August 2008 on the risk of pancreatitis. Notably, the number of total reports for exenatide (an indirect marker of drug consumption) showed a rapid increase from its marketing approval, with a first peak in Q1 of 2006 (“Weber effect”, Hartnell et al. 2004) and a strong decrease after the updated FDA alert. The parallel decrease in the total number of reports may reflect the attitude towards reduced prescriptions due to increased risk perception by physicians. Conclusion. In the FDA_AERS, the notoriety bias significantly contributed to increased reporting frequency of pancreatitis with exenatide. Awareness of this important bias should not affect clinician’s vigilance on this serious drug-event association.