On Tue, 09 Jul 2002 05:31:08 GMT, "jsnell55" <[EMAIL PROTECTED]> wrote:
> I am designing a clinical trial to test the efficacy of a new therapeutic > modality of bone fusion. I plan to run a double blind superiority study to > prove clinical efficacy. In my preparation for estimating the study size I > have calculated that the efficacy of this device will be around 10% better > than a placebo. You believe, placebo has *efficacy* for BONE FUSION? I think you need to check your terminology, or the literature.... I do believe in "Pbo effects" -- but they are less and less likely as you move to the strictly-physiological. Bone fusion sure does sound to me like something physiological. > Furthermore, there will be some variability in the data due > to drug effects (which I can identify, but not control for). Previous > similar studies have used between 220 - 350 patients. > > My question is this: If I assume that a placebo group (no treatment) > demonstrates improvement in bone growth among 40% of the patients, will I What is this "no treatment" in parentheses beside "placebo group"? That does not follow the definitions asserted anywhere, by anyone, so far as I know. To add to the confusion: If you were saying that the Pbo effect is ZERO in efficacy, equal to no treatment, you could not immediately attribute to it "improvement" in 40%. Improvement sounds like a useful thing to measure. Compared to what? > need a larger sample size to power the study, than if I assume 65% of the > placebo group will demonstrate improvement in bone growth? > When I guess at what the question is, I try to forget entirely about your categories. Then, I come up with a distinction of detecting a difference 40% vs 50%, or 65% vs 75%. "Are similar sample-Ns needed?" Practically speaking, there is not much difference. Set up the corresponding 2x2 tables and manipulate the total N (increasing all the cells) until you compute a chisquared that is significant: That is where the "power" is 50%. (The Ns will be slightly smaller for the second case.) It might be that I have seriously misconstrued this question, since I don't understand why Elliot C. posted the answer that he did. But I am pretty sure that the *question* is not asked very well. -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
