Meta Analysis for Health Care Data

• Activity Number: 259
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Teaching of Statistics in the Health Sciences
• Author(s): Fanglong Dong*+
• Companies: University of Kansas SOM
• Keywords: meta analysis ; health care data ; arrow plot

Abstract

BACKGROUND: Meta-analyses of interventions in health care traditionally study dichotomous outcomes such as mortality. Increasingly, medical interventions are titrated to a surrogate outcome that is a continuous variable, such as a measure of cholesterol or glucose control. Meta-regression has correlated change in the clinical outcome based on change of the continuous surrogate by pooling outcomes in the intervention groups of trials, but this method ignores available data. METHODS: We have developed two dimensional meta-analytical plots of clinical outcome by surrogate outcome. For each trial an arrow leads from the results of the control group to the results of the intervention group. Dashed arrows indicate insignificant results. RESULTS: The change in the y-coordinate of each arrow is the relative outcome reduction. The slope of the arrow is the relation between the surrogate and clinical outcomes; a vertical or horizontal arrow indicates no relationship. This plot allows testing heterogeneity of slopes and modeling nonlinear relationships. CONCLUSION: We propose that two-dimensional arrow plots increase the information garnered from meta-analysis.