To the Editor.
—One of the recommendations made by Dr Siegel and colleagues1 in their article on reporting cost-effectiveness analyses is that cost-effectiveness ratios should not appear for options that are less effective but more costly than an alternative (ie, are dominated).We agree with this approach when costs and effects are expressed as point estimates.2 However, the increasing frequency with which pharmacoeconomic evaluations are conducted alongside clinical trials makes it possible to express costs and effects as means and variances and to compare them by using confidence intervals (CIs). The use of CIs makes it possible to measure uncertainty related to variability in sample data.3 In this case, the CI of the cost-effectiveness ratio could range from a negative value to a positive one. This situation may occur, for instance, if sample size is not large enough to demonstrate statistically significant differences from a clinical perspective,
Sacristán JA, Obenchain RL. Reporting Cost-effectiveness Analyses With Confidence. JAMA. 1997;277(5):375. doi:10.1001/jama.1997.03540290027020