Workshop sheds light on important statistical advances for the biopharma industry

Statistics and AI/ML have empowered modern clinical trials to rigorously leverage external information to obtain clinical evidence that can be used for regulatory decisions. Randomized clinical trials could be made more efficient with Bayesian adaptive designs. This could result in a higher statistical power with the same sample size or, assuming the same power, a smaller expected sample size or shorter expected duration. 

Borrowing information from external controls, such as natural history or real-world data (RWD), could potentially reduce the need for large control groups, especially in rare diseases or when a placebo arm presents ethical concerns. Proper statistical approaches make single-arm trial design possible and feasible. In addition, statistical frameworks combining machine learning and causal inference can learn from external information while addressing confounding and bias properly. Such approaches have shown promise in clinical trials to estimate treatment effect in a principled, data-adaptive way. As a result, a smaller trial is possible given the same statistical power.

The workshop raised important considerations about statistical advancements, supported by AI/ML, to address many of the questions and challenges clinical trials and drug development face. By leveraging statistical advancements, biopharmaceutical companies can apply broader and deeper insights to the design and conduct of clinical trials and to more efficient drug development.