Dr. Gandjour raises an interesting point: The exclusion of survivor costs (costs associated with a treatment because it extends the patient's life) could affect the calculated cost-effectiveness ratio. He suggests that we calculated unduly favorable cost-effectiveness ratios because we excluded such costs.

The purpose of cost-effectiveness analysis is to provide a metric of comparison of potential uses of limited resources (the "medical commons" [1]). There is no gold standard for the threshold of willingness to pay. To be useful, such analyses must use standard methods, as we have done (2).

In response to Dr. Gandjour's suggestion, we searched PubMed to identify all cost-effectiveness analyses that were published in 5 major medical journals (JAMA, Annals of Internal Medicine, The New England Journal of Medicine, Lancet, and British Medical Journal) from 1 January 2003 to 31 December 2005. Of 43 articles that we identified, only 2 included survivor costs; in both articles, these costs were partial and disease-unadjusted.

Nyman (3) found little consensus regarding the inclusion of survivor costs in cost-effectiveness analysis. Even those researchers who favor the inclusion of such costs disagree on the type of costs to be included. Consequently, a dependable method that enables the inclusion of survivor costs has not been developed. Furthermore, the inclusion of such costs would bias analyses toward nonintervention unless an adjustment of willingness-to-pay thresholds occurred.

When the calculated cost-effectiveness ratio is close to the acceptable threshold and the survivor's future cost of care is expected to offset future earnings, the addition of survivor costs could transform an otherwise cost-acceptable intervention into an unacceptably expensive one. Invasive candidiasis in patients in intensive care units, the focus of our analysis, has been associated with a high mortality rate and shortened life expectancy. Both of these factors were incorporated into the analysis, but impaired future productivity was not. If one includes survivor costs, then such productivity might also be included. Therefore, the inclusion of survivor costs would make our analysis less useful and generalizable without affecting our conclusion that selective empirical use of anti-Candida therapy in patients in the intensive care unit is a reasonable strategy.