[Minutes, Advisory Committee to the Director GP-160 Meeting (1992: Bethesda, Maryland)]

139

1 greatest chance to benefit have the lower progression
2 rates.
3 So we're going to have to study many more
4 patients in the people who actually have the greatest
5 chance to benefit in order to detect any differences
6 there.
7 Next slide. All right. Well, this is
8 what you've all been waiting for, the sample sizes in
9 thousands. As you can see, the numbers depend very
10 heavily on how long you follow patients up because, as
11 any of you who have been involved in sample size
12 calculations know, sample sizes are driven by the
13 number of events you observe, more so than the number
14 of patients that you enter.
15 So if we were willing to, say, enter
16 patients in a year and then follow them for 5 years
17 until a result, we would need 14,000 patients. But if
18 we wanted a result after 2 years, we would need as
19 many as 30,000 patients.
20 I've put the assumptions on the bottom of
21 this slide. This is for a trial with 90 percent power
22 over the 5 percent two-sided significance level, 33
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1325 RHODE ISLAND AVENUE, N.W.
[phone] WASHINGTON, D.C. 20005 [phone]