To the Editor.
—In their recent article on implications of randomized trials, Dr Mulrow and colleagues1 include a meta-analysis of 13 trials involving elderly hypertensive patients. The relevant data on several different endpoints are presented graphically as a plot of odds ratios (ORs) with 95% confidence intervals (CIs) for each individual trial and for all trials combined (their Figure 1).The use of visual plots of ORs and CIs has become increasingly common in reviews of the literature, for good reason, since the graphic approach is an efficient and effective means of conveying quantitative information. Unfortunately, in the article by Mulrow et al, the data were plotted on an arithmetic scale rather than appropriately on a logarithmic scale. Hence, they convey a misleading impression. On a log scale, the CI has the property of being symmetrical around the point estimate.2 A log scale corrects for the fact that although
Hatch MC, Godbold JH. Hypertension in the Elderly: Correction and Clarification. JAMA. 1995;274(1):25. doi:10.1001/jama.1995.03530010039024