Generalized least squares for assessing trends in cumulative meta-analysis with applications in genetic epidemiology

Abstract

Objective: Cumulative meta-analysis allows the evaluation of a study's contribution to the combined effect of the preceding research. It accrues evidence, gradually adding studies one at a time and provides updated estimates along with confidence intervals whenever new evidence emerges. In many research areas, a temporal evolution of the effect size (ES) is present, leading to diminishing effects and would be advantageous to have methods capable of detecting it. Study Design and Setting: We propose a simple regression-based approach for detecting trends in cumulative meta-analysis. We use the combined ES of studies published up to a particular time, as dependent variable and the rank of the published studies as independent variable, in a weighted linear regression to detect a possible trend over time. The correlation between successive ESs used in the regression, is dealt by introducing a first-order autoregressive coefficient using Generalized Least Squares. Results: Application in several published meta-analyses of genetic association studies provides encouraging results, outperforming the commonly used method of comparing the results of first vs. subsequent studies. Conclusion: The particular method is intuitive, easily implemented and allows drawing conclusions based on formal statistical tests. A STATA command is available at http://bioinformatics.biol.uoa.gr/∼pbagos/metatrend/. © 2009 Elsevier Inc. All rights reserved.