Treatment Effect Estimation Using Observational Cohort and Nonrandomized Trial Data

Observational cohort (noninterventional) studies and nonrandomized (interventional) studies are often used for generating hypotheses or serving as an alternative to interventional randomized controlled trials (RCTs) when RCTs are not available yet. Unlike the treatment effect assessment from RCTs, estimating the treatment effect from observational cohort or nonrandomized studies can easily introduce biases because of factors such as missing data, unbalanced baseline covariates, unmeasured confounders, and treatment heterogeneity. Thus, it is critical to appropriately conduct, design, and analyze observational and nonrandomized studies. In this chapter, we will discuss the following three themes: (1) the regulatory guidance and the strategies of applying observational cohort and nonrandomized trial data in the development of therapeutic drugs and biologics, (2) the advantages and disadvantages of different data sources, and (3) the common statistical methods for minimizing the potential biases in observational cohort and nonrandomization studies.