|
Activity Number:
|
232
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #307884 |
|
Title:
|
Why There Are So Many Contradicted or Exaggerated Findings in Highly Cited Clinical Research
|
|
Author(s):
|
Ying Yuan*+ and Valen Johnson
|
|
Companies:
|
The University of Texas M.D. Anderson Cancer Center and The University of Texas M.D. Anderson Cancer Center
|
|
Address:
|
1515 Holcombe Blvd, Houston, TX, 77054,
|
|
Keywords:
|
test statistics ; Bayesian model ; p value
|
|
Abstract:
|
It is common for reports of clinical research regarding the effectiveness of new interventions to be contradicted in subsequent trials. This is particularly disturbing when high-impact research is involved. We analyzed 42 highly-cited (>1000 citations) controlled randomized clinical studies published between 1990 and 2003 and previously analyzed in Ioannidis (2005). Seven (18.5%) out of 38 positive studies were contradicted or were found to have overstated effects in subsequent studies. To understand this fact, we extracted the test statistics from the original articles and analyzed their values using a simple Bayesian model. Based on this model, we relate the reported p-values from these studies to posterior probabilities of associated hypotheses. Conclusions from this study are somewhat surprising and have important implications for consumers of classical testing methods.
|
