@BenWestphalen and others have asked my thoughts about the PEGPH20 trial results, nicely summarized here. https://twitter.com/BenWestphalen/status/1287304669421854720 This trial grew directly from studies using the KPC mouse, a model that I have championed. 1/

I have some thoughts. Every failed clinical trial is a tragedy. It is a much greater tragedy if we do not learn something new in the process. Post-clinical study of unsuccessful trials is a DUTY. With IPI-926, we learned that some elements of the stroma can restrain tumors. 2/

Models are not humans and even humans require interpretation, as the Phase 2 data show. The art of tumor modeling is not in the engineering, it is in the interpretation. Our ability to interpret models improves every time there is a clinical outcome with which to benchmark. 3/

So what do we learn? One lesson from IPI-926 was that the KPC model appears best at modeling short-term responses. But tumors adapt. We learned that short-term loss of Hedgehog-responsive CAFs led PDACs to adopt a more aggressive state in the long term. 4/

Mouse models are rarely treated for more than a few weeks. Patients are treated for months, sometimes years. If a drug extends survival by a few weeks in a mouse, this may double overall survival on therapy. How to interpret? Clearly not as a doubling of OS in humans. 5/

The increased RR observed in the PEGPH20 trial hints at the idea of improved (very) short-term responses. But they were not durable. They were also not durable in the mouse experiments. Every mouse died. Frank regressions were modest. 6/

So how do we utilize the KPC model? We require large regressions and durable responses, at a minimum. And we ALSO do longer-term studies starting with tiny tumors to examine chronic treatment response. “But that’s hard!” Yes. So is PDAC. 7/

We also demand a higher burden of proof in early clinical trials. To those who say that mice are not predictive of humans, I point out that with the history of Phase 1/2 studies in PDAC, humans are not predictive of humans! Only, let us not repeat our mistakes. Learn. 8/8