The Powers and Perils of Post-Marketing Data Analysis: Quantification and Mitigation of Biases in the FDA Adverse Event Reporting System

The Food and Drug Administration Adverse Event Reporting System (FAERS) is the primary source for postmarketing pharmacovigilance. Though potentially highly useful, the database reflects reporting biases, stimulated reporting, and suffers from lack of standardization and the use of multiple drug synonyms. These biases can suggest adverse drug reactions (ADRs) where none exist, and can obscure others that do exist. To decrease the noise in FAERS, and to reinforce important associations, we mapped over 750,000 drug identifiers in FAERS to the normalized chemical structures of their ingredients. This illuminated associations that would not otherwise be apparent, and also allowed a time-resolved analysis of ADR reporting. It also revealed similarities between drugs and adverse events across therapeutic classes, enabling unbiased classification of adverse events, indications, and drugs with similar clinical profiles. For instance, comparison of two selective cyclooxygenase-2 inhibitors, celecoxib and rofecoxib finds distinctive FAERS profiles after time-resolved analysis. We also investigated key idiosyncrasies, such as confusion between drug indications and drug ADRs, which can tar a drug treating a life-threatening disease, like thalidomide’s use against myeloma, with a deadly ADR that is likely the result of the disease itself, multiplications of the same report, which unjustifiably increases its apparent importance, and the correlation of reported ADRs with public events, regulatory announcements, and with publications. Comparing the pharmacological, pharmacokinetic, and clinical ADR profiles of methylphenidate, aripiprazole and risperidone, and of kinase drugs targeting the VEGF receptor (VEGF-R2), demonstrates how underlying molecular mechanisms can emerge from ADR co-analysis. The precautions and methods we describe may enable investigators to avoid confounding chemistry-based associations and reporting biases in FAERS, and illustrates how comparative analysis of ADRs can reveal underlaying mechanisms. INTRODUCTION Safety assessment of drug candidates is crucial for drug discovery, enabling the development of medicines that achieve the desired therapeutic effects with the least risk of adverse side effects. Preclinical regulatory investigations and clinical trials are designed to address safety of drug candidates and eliminate those that do not meet risk-benefit expectations1. However, limited access to large, diverse patient populations in clinical trials, untested drug co-administrations, and development of ADRs associated with chronic treatment, often results in post-marketing labeling and occasional withdrawals2–4. Thus, postmarketing pharmacovigilance is essential to track ADRs and ultimately reduce the over 1 million serious drug related side effects that occur each year in the USA. Between 5-10% of these ADRs are fatal5, and many others cause patient suffering, hospitalization, and increased health system burden6. Indeed, the fatality rate attributed to ADRs puts them among the top causes of death in the USA (over 40,000 in 2011), similar to suicide-related mortality7. Determinant tools in post-marketing pharmacovigilance are databases that aggregate ADR reports. Key among these is the FDA Adverse Event Reporting System (FAERS), which is perhaps the most extensive, and among the most widely accessible of these databases, currently containing over 8.5 million reports and rapidly growing8. FAERS and related databases, such as those of the EMEA and of Health Canada, can provide specific ADR phenotypes typical for either individual drug classes or specific indications, and can be accessed either directly8 or by APIs9,10. These large-scale adverse event databases enable analysis to relate clinical phenotypes and compounds11, and they have been widely used by the clinical community with much impact12–15. It is an attractive proposition to exploit the sheer scale of FAERS to detect drug-ADR associations that would otherwise not be apparent. A challenge in doing so has been the heterogeneous data sources and data conflation in the database. FAERS, while providing a solid frame for reporting, contains redundancies, biases, and conflations that affect its analysis and interpretation16. Our ability to even correlate drugs with their effects is obscured by something as simple as the tangle of drug synonyms in FAERS on average 16 different names for medicines containing each active drug ingredient which can obscure associations. In this work, we investigate the effects that these data . CC-BY-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/068692 doi: bioRxiv preprint first posted online Aug. 10, 2016;