Agent-based modeling (ABM) and randomized clinical trials (RCTs)

Here is a nice opinion piece in PNAS by Ross Hammond and Shari Barkin that makes a strong case for combining ABM and RCTs to alleviate the intrinsic limitations of RCTs, particularly when interactions between individuals may affect outcomes and treatment effects. Social context, peer influence, diffusion of information are all elements that can have a disproportionate impact on treatment effects. ABM is sometimes called Individual-Based Modeling (mostly in ecology, to the best of my knowledge, please comment if I am wrong), which expresses its ability to model individual differences and heterogeneity. The temporal dimension is also important in that subjects may adjust their response, consciously or not, learn, feel better, or worse, over time.

I would argue that the case made in the article goes beyond ABM -even though I do love ABM 💖 . What the authors describe is the practical utility of mechanistic models that aim at the heart of the problem being addressed: mechanisms of action and hypotheses about them. Such models would serve as the basis for more adaptive trials, leveraging adaptive design of experiments, extra-powered by statistical learning/machine learning ML and AI.