Under-ascertainment and underreporting of adverse events in clinical trials

Abstract

Introduction: Clinical trials are widely regarded as the “gold standard” for evaluating different interventions’ adverse events (AEs). However, numerous cases have suggested that AEs are underreported in clinical trials due to inadequate data collection methodology and inconsistent reporting criteria. This study examined 1) the AE ascertainment methodologies used for the marketing approval of new drugs and 2) the diversity and consistency of AE reporting criteria used in trial reports.

Methods: We screened drugs approved by the US Food and Drug Administration (FDA) in 2018-2019 and collected publicly available trial documents for all pivotal trials. From these documents, we examined ascertainment methods of adverse events of special interest (AESIs) and newly signaled post-marketing AEs. We also assessed the association between trial characteristics and the AE ascertainment approach using binary logistic regression. Additionally, from the obtained reports, we examined the characteristics of reported AEs and the usage of AE reporting criteria. Then, we assessed the consistency of the number of reported AE types and reporting criteria used across trial publications and other important sources of trial results.

Results: 322 AESIs were identified from trial documents for 64 trials reporting 31 drugs approved in 2018-2019. 71% were systematically ascertained, mainly using diagnostic measurement tools and laboratory assessments. 10% were non-systematically ascertained. The ascertainment method of 19% was unclear. The regression analysis did not reveal statistically significant associations between trial characteristics and the use of a systematic ascertainment approach for AESIs. Of the six examined newly signaled post-marketing AEs, one was systematically ascertained. The examined sources utilized various criteria to report both serious and non-serious AEs. Frequency criteria were the most commonly used AE reporting criteria. Furthermore, the examined sources inconsistently reported serious AEs and inconsistently utilized reporting criteria.

Conclusions: We were unable to identify the ascertainment methodology for some AEs, even with access to underlying trial documents. Additionally, trial reports applied various criteria that potentially resulted in only a subset of AEs recorded during the trial being reported. The study suggests room for improvement in AE data collection and reporting to aid unbiased harm-benefit assessments of study interventions and informed treatment decisions.