139-31: Problems Commonly Associated with Forest Plots Addressed Using High Resolution Graphics in SAS®

Forest plots are frequently used to display the results of meta-analyses and have been published in the medical literature since 1988 (Lewis & Clarke, 2001). A review of forest plots reveals that the format chosen can affect the accuracy with which the information is conveyed to the reader. One formatting convention is the use of symbol areas to reflect the relative contributions, or weights, of the studies included in the analysis. Typically a study with a larger weight is represented by a symbol with a proportionately larger area. The use of symbol areas to convey study weights is not optimal for a number of reasons. First, it is difficult for viewers to accurately decode symbol areas to enable a visual comparison of study weights. Second, as symbol areas increase in size they can obscure other information on the plot. Third, if a logarithmic scale is used, studies with equal weights should, but often do not, have symbols with the same physical dimensions. This issue becomes more apparent as the difference in effect sizes increases. The macro used to generate the figures in this paper provides SAS software users with an easy to use program to produce high-resolution forest plots. It will generate forest plots where symbol widths represent study weights, but it will also generate traditional forest plots where symbol areas represent study weights. Regardless of the symbol format or type of scale being used, symbols representing studies of equal weight will have the same dimension anywhere on the plot.