Historical borrowing in nAMD: the present as a "prior" for designing more efficient future trials

Anti-vascular endothelial growth factor (aVEGF) treatment has been the standard of care for the treatment of nAMD since the approval of ranibizumab in 2006. A wealth of clinical historical data on aVEGF is now available and by utilizing that appropriately, the efficiency of future clinical trials could be improved by reducing the sample size needed or improving their statistical power. While virtual trials and model-based simulations are increasingly recognized within modeling and simulation discussions, the statistical nuances of employing virtual patients have not been thoroughly explored. In this talk, we will discuss how to incorporate historical knowledge on standard of care ranibizumab in the analysis of a new trial, blending modeling and Bayesian statistics. Historical data of ranibizumab treatment arms from 6 Phase 2, 3, and 4 clinical trials were used to develop a non-linear mixed effect model that relates the clinical endpoint (i.e. Best Corrected Visual Acuity, BCVA) with ranibizumab exposure over time. Relevant patient characteristics at baseline were incorporated. The model was validated and could thus be used to simulate the clinical endpoint in a new trial design. These simulations represent our current knowledge on ranibizumab before a new trial starts. Using the Bayesian paradigm, this prior knowledge is integrated and updated based on emerging ranibizumab data collected as part of the new trial. We found that, through our borrowing approach, the sample size of a future trial can be reduced to two thirds with a minimal loss of power and bias in the estimation of the treatment effect.