PROJECT SUMMARY Large confirmatory or pragmatic clinical trials enroll patients from multiple centers to obtain high-quality evidence useful for decision-making in diverse healthcare systems. The design of multicenter trials is particularly appealing to policymakers because many tend to be pragmatic and offer access to a large sample of diverse patient populations. Decision-makers trying to use the results of multicenter trials have specific target populations in mind. Yet, selection at both the center and individual participant level challenges the ability of even well-designed and conducted multicenter trials to draw inferences about any meaningful target population. These selection mechanisms, even in impeccably conducted multicenter trials, result in samples of trial participants that are not representative of any specific target population. When treatment effects are heterogeneous over covariates that are differentially distributed across centers, the overall trial average treatment effect estimate will not apply to any one of the populations underlying the participating centers or the target population. Consider for example, the National Lung Screening Trial (NLST), a large multicenter trial of 33 centers (53,456 individuals) across the United States, which motivated this proposal. Because the positive benefits of screening are mainly concentrated in individuals at high-risk for lung cancer it is critical that individuals are appropriately targeted for screening recommendations. Our long-term objective is to optimize the applicability of trial results into clinical practice. We will take steps toward this objective by achieving the following specific aims: (1) Develop robust and efficient statistical methods, that can be combined with modern machine learning techniques, to reinterpret a multicenter trial in the context of each of the participating centers and to transport inferences from a multicenter trial to a new target population; (2) Evaluate the performance of methods developed in Aim 1 in simulation studies that reflect real-world data by considering scenarios with different sample sizes, number of centers, treatment effects, and outcome or selection mechanisms; (3) Apply the methods in the NLST to reinterpret the trial in the context of one of the participating centers and to transport trial findings to a new target population using baseline data from the National Health and Nutrition Examination Survey, a nationally representative survey of the United States general population. The proposed work is innovative for analyzing multicenter trials because it develops new methods, clarifies the assumptions needed for obtaining valid inference, and empirically assesses method performance. This work is significant because as a result of applying these methods, policymakers will be able to learn about treatment effects in different underlying populations and produce high-quality evidence of interventions in diverse populations that are ofte...