(C-167) Toward exposure-wide and outcome-wide pharmacoepidemiology for generating high-quality real-world evidence

Background: The traditional template for evaluating the effectiveness of medical products typically focuses on the effects of one or a few treatments on one or a few outcomes. However, this approach can be subject to several limitations, including inefficiencies in time and resources, slower accumulation of knowledge, and potential investigator bias in selecting models and reporting results.

Objectives: To formalize exposure-wide and outcome-wide pharmacoepidemiology as strategies to generate comprehensive, credible, and high-quality real-world evidence.

Methods: Over the past decade, two newly proposed templates—exposure-wide epidemiology and outcome-wide epidemiology—have received increasing attention. Briefly, exposure-wide epidemiology (Ioanndis 2016 Stat Med) examines the associations between a broad spectrum of exposures and a single outcome, whereas outcome-wide epidemiology (VanderWeele 2017 Epidemiology, and VanderWeele et al. 2020 Statistical Science) suggests exploring the associations between a single exposure and a broad spectrum of outcomes. Although originally and primarily utilized in genome-wide association studies and psychosocial epidemiology respectively, these two templates hold wider relevance, implications and utility in the field of pharmacoepidemiology, especially given the growing interest in and availability of real-world data that contain extensive and detailed information on a wide range of exposures/treatments and outcomes/diseases.

Results: The template of exposure-wide pharmacoepidemiology is particularly useful for drug-repurposing (e.g., alcohol use disorder) and pharmacovigilance for specific adverse events. The template of outcome-wide pharmacoepidemiology, on the other hand, can play an important role in evaluating the effectiveness of blockbuster treatments (e.g., GLP-1 agonists, COVID-19 vaccines). Furthermore, when implemented within the “target trial emulation” framework using large healthcare databases (e.g., electronic health records of a healthcare system), these two templates can mimic master protocols from randomized trials: exposure-wide approaches emulate umbrella trials, whereas outcome-wide approaches emulate basket trials.

Conclusions: The exposure-wide and outcome-wide templates are promising approaches for pharmacoepidemiologic research, offering the potential to enhance research efficiency, expedite knowledge and real-world evidence generation, and mitigate investigator bias. Their successful implementation, however, requires coordinated efforts among experts and stakeholders across diverse fields, highlighting the need for continued exploration and research in this area.