Hello, my name is Mark Conaway.
I'm a professor in the Division of Translational Research and Applied Statistics
in the Department of Public Health Sciences at the University of Virginia.
This talk is on Patient Heterogeneity and Dose-Finding trials.
Part two focuses on trials conducted in more than two groups.
There's also a part one that focuses on trials done in exactly two groups.
The outline for this talk is that I'll first provide
motivating examples for trials done in three or more groups,
and then I'll discuss some of
the existing statistical methods for the design of trials in three or more groups.
First, was published by Yuan and Chappell in 2004,
and then I'll talk about a couple of papers recently
published that discuss completely or partially ordered groups,
and I'll define what I mean by
completely or partially ordered groups in the context of dose-finding.
A study done in four groups presented by Ramanathan et al in 2008.
Prior to assigning doses,
patients were grouped into categories based
on the degree of liver dysfunction at baseline.
Patients were grouped into either none,
mild, moderate, severe levels of liver dysfunction.
A similar group was used in a dose-finding study reported by LoRusso et al in 2012.
In this case, the groups are what we call 'completely
ordered' in the sense that if you fix a dose level j,
you expect that the probability that a DLT at dose level dj among patients that have
no baseline liver dysfunction would be
less probability of a DLT at dose dj for patients with mild,
or moderate, or severe liver dysfunction.
Another way of thinking about completely ordered groups is you can say,
"Well, if you fix a dose,
you know completely the ordering of any two DLT probabilities from the different groups."