Methods used to generate safety-related evidence in real-world oncology settings: A systematic literature review.

Abstract

e18769 Background: With the increasing availability of real-world data (RWD) as a valuable healthcare data source, there have been large shifts in how a drug’s properties can be evaluated over its life cycle. RWD can be used for many evidence generation purposes, including the evaluation of a drug’s safety in real-world settings outside of clinical trials. Traditional methods (e.g., cohort and case-control studies) are often used to analyze RWD, but novel methods have emerged as the field advances with access to more diverse data sources. The extent to which these newer methods are being used is unclear. The aim of this systematic review is to identify methods that have been used in recent years to generate safety-related evidence in real-world oncology settings. Methods: A search of literature published between 1/1/2016-8/10/2021 (Embase, Ovid) was performed to identify studies describing methods used to generate evidence related to the real-world safety of oncology products. Each article was independently screened by two researchers in two levels (title/abstract and full text screening). Data extracted included, but was not limited to, study design, cancer type, data source, and statistical methods. Results: Out of 845 publications identified, 44 were included for review (742 excluded after level 1 screening; 59 excluded after level 2 screening). The most common method used was the retrospective cohort study (n = 26), followed by prospective cohort study (n = 8), literature review (n = 4), cross-sectional study (n = 3), artificial intelligence (AI) methods (n = 3), data mining analysis (n = 2), case-series, and case-control study (n = 1 each). Some publications used multiple methods. While traditional methods (e.g., cohort study, literature review) were the most common throughout the search period, newer methods (e.g., AI methods) were identified in the latter part of the search period (2019-2020). The majority of studies that used traditional methods relied on medical records or claims data (e.g., Optum), which often provided structured RWD to evaluate physician-documented drug safety data. In contrast, articles that used newer methods like AI analyzed large amounts of patient-reported data (e.g., chatbot conversations; social networks), which allowed researchers to evaluate patients’ own experiences with drug safety. Conclusions: The most common evidence generation methods identified involved traditional methods, but novel methods were also identified, especially in recent years. This indicates a trend towards increased use of innovative methods like AI that allow for the analysis of data in ways not possible before. Taken together with more traditional methods and data sources, these findings can provide clinicians with a more comprehensive and patient-focused understanding of drug safety, which will ultimately improve patient care.