Off-label drug use is common in oncology, due in part to significant unmet medical need, the rarity of many cancers, and the difficulty of conducting randomized controlled trials (RCTs) to support labeling of every drug in every disease setting. As new drugs are developed for use in tumors defined by genomic aberrations, it may be scientifically reasonable to expect that a targeted anti-cancer agent with efficacy in a biomarker-defined population within one tumor type may also have activity in another tumor type expressing the same biomarker. Such expectations also fuel off-label prescribing. However, the current approach to prescribing targeted agents off-label does not capture patient outcomes, thus missing an opportunity to gather data that could validate this approach. We explore the potential for collecting such data, highlight two proposals for oncology-specific patient registries, and put forward considerations that should be addressed to move toward better evidence development around off-label use.