A0213
Title: A unified framework for meta-analysis with the five-number summary
Authors: Jiandong Shi - Hong Kong University of Science and Technology (Hong Kong) [presenting]
Abstract: For clinical studies with continuous outcomes, if the data are skewed, researchers often report the whole or part of the five-number summary rather than the sample mean and standard deviation. Most existing methods for meta-analysis, however, cannot handle the normal and skewed data simultaneously. By incorporating the recent advances in data transformation, we develop a unified framework for meta-analysis when some studies are reported with the five-number summary. Specifically, we first develop a new testing method, using only the five-number summary, to check whether or not the underlying distribution of the data is skewed away from normality. If the skewness test is not rejected, we then apply the transformation methods to recover the sample mean and standard deviation from the five-number summary. Otherwise, it is suggested to either exclude the skewed studies from the meta-analysis for normal data, or apply a subgroup analysis that separates the normal and skewed studies.
