Presented by Malcolm Price, Richard Riley, Ian White, Dan Jackson and Jamie Kirkham
The use of meta-analysis to synthesise effect estimates from multiple studies is now well established in evidence based medicine. Many studies have more than one outcome of interest, such as disease-free survival and overall survival, and researchers currently meta-analyse each outcome separately. However, such multiple outcomes are often related to each other, i.e. they are correlated. For example, a patient’s time to recurrence of disease is generally associated with their time of death. By meta-analysing each outcome independently, researchers ignore this correlation and thus lose potentially valuable information. However, a multivariate meta-analysis can analyse correlated outcomes together and utilise their correlated information to get the most out of the available data. This course provides a firm introduction to multivariate meta-analysis methods and gives a full demonstration of its implementation in STATA.
The course begins by revising the well-known methods for univariate meta-analysis. Next it introduces the multivariate meta-analysis model for both fixed and random effects. This includes the rational, mathematical form, and a description of its advantages and disadvantages as compared to the univariate approach. It explains how to calculate within-study and between-study correlations, and show why they lead to 'borrowing of strength' across outcomes. It details estimation methods to fit the multivariate model, and provide a detailed demonstration of the MVMETA package in STATA. Next the presenters discuss some of latest developments in the field, including a multivariate equivalent of the I-squared statistic, and outcome reporting bias. Finally, the course gives a more in depth description of one particular application of multivariate meta-analysis – the synthesis of summary evidence from diagnostic test studies.
For more details and to register, please visit www.rss.org.uk/courses
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|