In Rosemont, Illinois, on May 11, Eli Lilly’s Shankar Vaidyaraman presented on work that he and colleagues are doing to incorporate common cause variability into Design Space modeling. The talk, “Design Space Mapping Accounting for Model Parameter Uncertainty and Common Cause Variability: Methodology and Case Study,” was part of the DynoChem 2011 North American users group meeting.
Drug development teams, Vaidyaraman noted, have become fairly adept at deterministic Design Space modeling. “A deterministic Design Space is always a good starting point,” he said. “It gives a good feel for the size and shape of the region that defines constraints.”
However, manufacturers must now begin to complement (but not replace) this work by also conducting modeling that takes into consideration input variability and uncertainty. The idea of developing Design Spaces with model uncertainty and common cause variability is still somewhat new, he noted—John J. Peterson of GlaxoSmithKline has been one pioneer (see here for a 2010 article)—but is a next logical step in DS modeling.
Design Space under uncertainty, he said, “is the combination of input variables that result in outputs which satisfy constraints within the specified probability.” What matters is “not whether or not a constraint is met, but the probability of whether or not a constraint is met.”
Vaidyaraman finished his talk by illustrating how the concept of uncertainty has been incorporated into Lilly’s DS models to address yield and solvent constraints. He showed how using common cause variability in Design Space modeling causes the region of operability to shrink, and thus “incorporating model uncertainty reduces the size of the Design Space.”