Please Note: This seminar will be given in person.
|
Statistics and Biostatistics seminar series
Stephen
Walter Room: M3 3127 |
Some potential biases in the meta-analysis of clinical trials data
This talk will review two recent issues concerning the meta-analysis of clinical trials. First, we will examine the impact of stopping a trial early at an interim analysis. The standard literature has focused on maintaining the nominal type-I error rate for associated significance tests, but there has been less attention paid to the bias that arises in the estimation of a treatment effect from interim data.
Second, we will discuss a recent finding, that the conventional methods for weighting studies involve an inherent bias, something that has apparently not been noticed for decades. I will briefly present some very recent work on developing appropriate correction factors to counteract this bias.
Both of these problems require a little theory to solve them, but there are important practical implications.