Notoriety bias in a database of spontaneous reports: the example of osteonecrosis of the jaw under bisphosphonate therapy in the French national pharmacovigilance database.

PURPOSE

The purpose of this study was to identify a notoriety bias in a database of spontaneous reports and its consequences on the calculation of the reporting odds ratio (ROR).

METHODS

We used the case/noncase methodology to calculate the ROR for bisphosphonates and osteonecrosis of the jaw (ONJ) in the French national pharmacovigilance database (from 1985 to 2013). To evaluate notoriety bias, drug-related risk factors for ONJ [as specified in the summary of product characteristics (SPC) of bisphosphonates] were systematically scanned for notifications of reports of ONJ occurring under bisphosphonate therapy. When a risk factor was present, the ONJ was considered as not due to bisphosphonates, and a second ROR was calculated under the hypothesis of maximum bias.

RESULTS

In total, 148 cases of ONJ were reported (143 with bisphosphonates and five without). The raw ROR was 3448 (95% confidence interval 1413-8417). After analysis of the reports, only 86 had no mention of a risk factor for ONJ. The ROR under the maximum bias hypothesis was 87 (95% confidence interval 63-121). Among ONJ where chemotherapy was being administered simultaneously to bisphosphonates, 27 reports did not consider the chemotherapy to be implicated, despite seven of these occurring in cases where ONJ was mentioned in the summary of product characteristics.

CONCLUSIONS

The existence of a notoriety bias has an impact on measures of disproportionality. The detection of pharmacovigilance signals might be delayed. It is advisable to list all drugs being taken when an adverse drug reaction occurs, and not only those known to be associated with the observed reaction.