Physicians and patients must routinely make important decisions using limited and imperfect evidence. This includes decisions about new drugs and devices, which require a regulatory determination by the US Food and Drug Administration (FDA) that they are safe and effective for their intended use. For drugs, FDA approval is typically based on rigorous randomized clinical trials; for devices, requirements for approval depend on the potential risks and similarity with existing versions of devices on the market. Unsurprisingly, such evidence is often not ideal for “real-world” decision-making. Premarket evidence has limited ability to resolve uncertainty regarding long-term outcomes, effectiveness in different practice settings, and benefits and risks in populations poorly represented in trials. These limitations are particularly important for personalized medicine, where information is needed about the effects of a new drug or device in specific subgroups of patients based on preferences, genomics, and other clinical factors.