It is becoming increasingly more common for a randomized controlled trial of a new therapy to include a prospective economic evaluation. The advantage of such trial-based cost-effectiveness is that conventional principles of statistical inference can be used to quantify uncertainty in the estimate of the incremental cost-effectiveness ratio (ICER). Numerous articles in the recent literature have outlined and compared various approaches for determining confidence intervals for the ICER. In this paper we address the issue of power and sample size in trial-based cost-effectiveness analysis. Our approach is to determine the required sample size to ensure that the resulting confidence interval is narrow enough to distinguish between two regions in the cost-effectiveness plane: one in which the new therapy is considered to be cost-effective and one in which it is not. As a result, for a given sample size, the cost-effectiveness plane is divided into two regions, separated by an ellipse centred at the origin, such that the sample size is adequate only if the truth lies on or outside the ellipse.