Combining estimates of the odds ratio: the state of the art.

Medical research commonly relies on the combination of 2 x 2 tables of counted data for making inferences about treatment effects or about the causes of disease. This article reviews point estimation and interval estimation for a common odds ratio. Traditional methods for providing these estimates face special challenges, and sometimes break down, when the data are sparse. Recent research provides practical alternatives to the traditional methods, and uses new computer algorithms to make them readily available in microcomputer software packages. This article illustrates and compares the various methods and offers recommendations for using the newer methods when analysing data sets of moderate size.