Statistical methods to detect fraud and errors in clinical trials

Published on September 29, 2016   40 min
0:00
Hello, my name is Marc Buyse, I'm the Chief Scientific Officer of CluePoints, a company devoted to central statistical monitoring of clinical trials and also a professor of Biostatistics at Hasselt University in Belgium. Today, I'm going to talk about methods to detect fraud and errors in clinical trials. Why is this problem important?
0:23
If we look at the number of submissions to the FDA between 2000 and 2012, they were 332 such submissions of which exactly half failed, 151 failed, at the first cycle, and the other half was approved at the first cycle. So that is a high failure rate. And of those applications who failed, 80 were never approved, so half of them were never granted approval. So this is a high failure rate. And why do so many applications fail? The next slide shows you some reasons for failures.
0:58
If we look at the reasons, study conduct, and, in particular, missing data or data integrity problems caused about 7% to 9% of the applications to fail either during the first cycle or during any cycle. And again, that is probably too high a percentage to be acceptable. And if we look at the clinical data submitted to support the application, again about 24% to 36% of the applications fail either during the first cycle or at any cycle because of inconsistent results between trials, between sites, or between endpoints. And what I'm going to concentrate on today is differences in clinical data submitted by different sites in a clinical trial. And again, this is important because it is the cause of between a quarter and a third of the rejections of new applications for approval.
Hide

Statistical methods to detect fraud and errors in clinical trials

Embed in course/own notes