Understanding the potential increase in the risk of uncommon adverse events associated with exposure to medical products requires large sample sizes, only seldom achieved in typical development programs. Applicability of most programs to real world use can also be limited. The use of large pragmatic trials can potentially answer these questions in a less biased way than observational studies, however, challenges can arise from treatment discontinuation or switching. The ITT analysis addresses the effects of offering treatment. Per-protocol analysis may introduce selection bias. Applying methods for causal inference might address questions such as whether the relative risk is less than a specified value among compliers. This talk will outline these issues, present examples, and introduce (at a high level) some potential approaches to analyses that might be helpful in these situations.