The recent article by Glassman and colleagues [1] contains several errors and may mislead readers about the promises and limitations of cost-effectiveness analysis.

First, their Table 2 reflects a 2.5% risk for rupture with aneurysms 3.5 to 3.9 cm in diameter, but their text reports a Figure of 0.25%. Second, their Table 3, from which most of their conclusions are drawn, is inaccurate. A common mistake in calculating lives saved is not using a fixed reference point. In this case, a reasonable reference point is the number of patients who would die if no patients had surgery for aneurysms. Using their example, 2900 of 40 000 patients would die under those circumstances (3080 if the 2.5% risk for rupture is used for aneurysms <4 cm rather than the 0.25% we used). In either case, the number of lives saved for each threshold at each of the three hospital networks is calculated relative to this reference point. We believe that the correct figures are presented in the Table 1 on page 166.

Many of the authors' conclusions are not preserved with this new table. For example, the authors' view that the three hospital networks differ little in total number of lives saved at the 5-cm threshold for surgery is no longer supported. Similarly, although the authors' Table suggested that increasing numbers of lives are saved at the 6-cm threshold as one moves from network A to network C, our Table reveals the opposite. The authors' result could never be achieved given the different surgical mortality rates. In the new table, network A not only saves more lives at the 4-cm threshold, it saves each life at a lower cost than network B or C.

Finally, we agree that cost-effectiveness analyses often involve tradeoffs and that those tradeoffs can be hard to balance. But hard-to-balance tradeoffs are not an argument against using cost-effectiveness analyses to develop decision rules. An important purpose of cost-effectiveness analysis is to make these tradeoffs explicit. Sometimes, these analyses reveal that some strategies are clearly better than others. The miscalculations by Glassman and colleagues understate the persuasiveness of quantitative health policy analysis. After this analysis, for example, who would consider using a 4-cm threshold at network B or C? More lives could be saved, at a lower cost, by using a 6-cm threshold at network A. Similarly, a 5-cm threshold at network C is clearly worse than a 6-cm threshold at network A or B.

We believe that few could support using the 4-cm threshold at network A after viewing this analysis. Although this strategy carries an average cost of $526 316 per life saved, the relative extravagance of this strategy is most evident when one considers that the 40 additional lives saved at this threshold, compared with the 5-cm threshold at the same network, increase costs by $400 000 000. In other words, each additional life saved costs $10 000 000.

Although the authors created their example to illustrate the limitations of cost-effectiveness analysis, the correct figures ironically reveal how much these analyses contribute. The virtue of these analyses is that they often reveal which strategies are dominated, which are extravagant, and which are truly close calls.