Concerns over safety of a medical product can arise at any time in a product's development. Companies will be asked for additional data by regulators if there is "too much" uncertainty in the risk, either overall or for a specific group. "Too much" will clearly depend on current perceptions of risk and benefits. So data from Real World Evidence such as registries, passive surveillance systems such as VAERS and FAERS, and active surveillance systems such as Sentinel are often used to study or confirm suspected risks. Some methods using these same resources are designed to find unanticipated risks using a data mining approach. No premarket study can provide a complete picture since some events can be rare and some can be the result of uses not studied (eg drug-drug interactions). I will talk about 21st Century Cures and the recent user fee renegotiations and what is said about real world evidence. Then I will highlight some FDA efforts in this arena and point to some opportunities for methodological advances.