INSP-01. What to do when you cannot randomize? Leveraging Historical Data in Efficient Study Designs for Pediatric Neuro-Oncology.

BACKGROUND: Randomized phase III studies represent the gold standard in clinical trial design for many good reasons. They control bias and the effects of known and unknown covariates on outcomes of interest. Recent examples from Children’s Oncology Group Medulloblastoma studies have demonstrated their utility in providing insights that likely would not have been possible otherwise. However, experience with these trials also reaffirmed that large, randomized studies often take too long to keep up with the ever-changing landscape in pediatric Neuro-Oncology, and the rarity of these tumors is a significant barrier in utilizing such designs effectively. METHODS: Recent global efforts in pediatric Neuro-Oncology have led to rich, well annotated repositories that contain patient-level data. While these data suffer from the well-known limitations when used as sole comparison cohorts for ongoing studies, they also offer an opportunity to design more efficient studies in ultra-rare patient populations that trialists in pediatric neuro-oncology often face. Therefore, there is renewed effort in the statistical community in devising methodologies that can effectively utilize external data in the design of prospective studies. These approaches include incorporating external data as a supplement to a small fraction of patients randomized to standard of care arms and prospectively assessing similarity with an intent to minimize overall sample size. Others focus on patient selection methodologies from external controls with an intent to optimize matching between the retrospective and prospective cohorts to control for known covariates. Additional considerations include incorporating arms into the study that retain standard of care treatments to capture the magnitude of drift in outcome over time due to improved supportive care. CONCLUSIONS: While there are important limitations to designs based on external controls, judicious choice of design parameters and careful selection of controls could provide a viable alternative when rarity of patient populations make randomized designs infeasible.