So I did a bit of a deep dive on this retrospective-prospective model. It's amazing to me how widely adopted and used OnctotypeDX was before the published results of TAILORx were available. It's one of the most important decision points in a person's life and you're basing a it on a retrospective analysis of the tissue bank of a super old clinical trial, using a different cut off point (more favorable) to define low-risk? It's not like the harm of undertreating was local recurrence of DCIS, it was the potential development of metastatic breast cancer. Granted, there were also more severe risks to overtreatment than with radiation as well. Obviously, everything worked out fine for OncotypeDX, but it was seemingly ubiquitous before the prospective validation trial hit the NEJM in 2015.