Tuck,
Thanks for filling me in on the p2 trial for av fistulas in hemodialysis. My dd is still incomplete. I've been a natural resources junkie for the past 3 years, and so I'm just catching up on this sector.
"do you have any insight on its ability to precisely measure the 75% occlusion ratio, differentiating well between patients with 74% occlusion and 76%?"
OK, I see where you're coming from, but the use of 75% is just using the language of interventional cardiologists and vascular/ct surgeons. All other trials that ask questions about critical stenosis are conducted in this fashion. Why?
The human vascular system can accomodate some remarkable amount of occlusion, but somewhere at around 75% occlusion a stenosis is considered hemodynamically "critical". Along with this designation, there is a recognition that the next level of intervention is usually required. It doesn't quite follow your standard fluid laws (poiseuille's), because of some unique characteristics of vessels, non laminar flow, non-newtonian fluid propery of blood, collateral flow, etc.
A reduction in cross sectional area of around 75% usually means than it is unlikely that any amount of compensatory/collateral mechanisms can supply blood to the area distal to that stenosis. So in terms of what can be observed on a QuantitativeCA, no other endpoint makes sense if you're asking a basic question about graft failure.
But there's another question you're asking- and that's about the statistics involved in this question.
"Do the QCA machines have the accuracy to reliably descriminate between 74 and 76%?"
I'll give it a shot, but if you know any board statisticians, send them a post. At first blush, however, it seems that it will all come out clean in the statistical wash. QCA is good - the new standard of care IMO, and far better than the old days when CT surgeons had the enormous responsbility of watching an angiography movie and deciding "LAD stenonsis. Put him on the schedule." However, QCA does not have perfect accuracy or perfect precision. However, whatever variability a QCA machine has is not going to bias the data one way or the other, but just cause a little more variability than say a 99.9% accurate measurement. So this variability will actually make a successful trial a slightly higher hurdle. In any case, it's impossible to avoid, and CGTK's approach is the best possible. I'll ask a statistician tomorrow if my thinking is right. |
