Is It Possible to Claim Efficacy if One of Two Trials is Significant While the other Just Shows a Trend?
- PDF / 368,997 Bytes
- 5 Pages / 504 x 719.759 pts Page_size
- 46 Downloads / 196 Views
Drug Information Juunral. V d 16. pp. 875 879. 2002 Printed in the USA. All rights reserved.
Copyright Q 2002
Drug Information Association
Inc.
IS IT POSSIBLE TO CLAIM EFFICACY IF ONE OF TWO TRIALS IS SIGNIFICANT WHILE THE OTHER JUST SHOWS A TREND? GERDROSENKKANZ, PHD Biostatistics and Statistical Reporting. Novartis Pharma AG. Basel. Switzerland
This note explores the possibility of allowing cin efficacy claim for a new drug based upon the results of fico pivotal clinical trials if one trial i s significant in the usual sense while the other just shokcs a trend in the right direction. We show that it is possible to select the type I errors rhat enable the individual trials to achieve this goal without a substantial increase in the overall type I error. Key Words: Combined analysis of clinical trials; Strength of evidence; Trend; Confirmatory results
INTRODUCTION THE USUAL REQUIREMENT for more than one clinical trial reflects the need for independent substantiation of scientific results that are based upon experiments. The traditional position of the United States Food and Drug Administration (FDA) has, therefore, been that one well-controlled clinical trial might not be enough to provide sufficient evidence for the efficacy of a new drug (1). European regulators never explicitly requested more than one controlled study with statistically compelling and clinically relevant results, however, a recent Points to Consider document (2) lists many reasons why it is usually prudent to plan for more than one phase 3 study. Although both American and European regulators are willing to accept one well-controlled trial for approval under certain circumstances, it seems that more
Reprint address: Gerd Rosenkranz, Biostalistics and Statistical Reporting, Novartis Pharma AG, CH-4002 Basel, Switzerland.
than one pivotal trial is still needed in most circumstances. Therefore, some thoughts on the combined analysis of two or more clinical trials might be useful. There will be little doubt about the strength of evidence of a drug’s efficacy if two welldesigned and well-controlled clinical trials show a statistically significant difference in favor of the experimental drug. More formally, if p I and p 1 denote the p-values from the tests utilized to compare two treatments in the respective trials, we say that both tests are significant on a level a if pI I a and p: I a.Under this condition, the probability that both trials erroneously show a significant effect is given by:
Thus, the overall significance level y = a’ is much lower than for each individual trial. If, for example, the level for one test was set to a = 0.025, we obtain y= 0.000625. On the other hand, the probability that at least one trial shows a difference by chance is given by:
875
Downloaded from dij.sagepub.com at NEW YORK UNIV LIBRARY on April 24, 2015
8 76
Gerd Rosenkrcinz
Pr[pl Ia U p. I a] =1 =
-
Pr(p, > a
1 - (1
-
n p? > a]
correspond to one-sided tests such that a small p-value indicates a trend in favor of the test drug.
a)? RESULTS
= 2
Data Loading...
