In this talk, we will explore various techniques for estimating treatment effects by leveraging different causal measures and integrating multiple data sources, including randomized controlled trials (RCTs) and real-world observational data. We will begin by discussing generalization methods that combine RCT data with observational datasets to predict treatment effects in populations different from those studied in the trial. We will then examine how the choice of causal measures (e.g., Risk Ratio, Odds Ratio) affects the validity and robustness of these generalizations. Next, we will address scenarios where multiple clinical trials are available and explore how causal federated learning can be used to aggregate evidence across these sources.