OBJECTIVE Selective endothelial transplantation in the form of Descemet's stripping endothelial keratoplasty (DSEK) is rapidly replacing traditional full-thickness penetrating keratoplasty (PK) for endothelial disease. An incremental cost-effectiveness analysis was performed to determine whether the benefits of DSEK are worth the additional costs. METHODS Retrospective cohort study. METHODS Patients at the Singapore National Eye Center, a tertiary eye center in Singapore, with Fuchs' dystrophy or bullous keratopathy who underwent either PK or DSEK. METHODS Patients underwent either PK (n = 171) or DSEK (n = 93) from January 2001 through December 2007. Data were collected from inpatient and outpatient notes corresponding to the time immediately before the procedure to up to 3 years after. METHODS Improvements in best spectacle-corrected visual acuity were used to calculate the increase in quality-adjusted life years (QALYs) 3 years after the procedure. This was combined with hospital charges (a proxy for costs) to determine incremental cost-effectiveness ratios (ICERs) comparing PK with no intervention and DSEK with PK. RESULTS Three-year charges for DSEK and PK were $7476 and $7236, respectively. The regression-adjusted improvement in visual acuity for PK relative to no intervention was -0.613 logarithm of the minimum angle of resolution (logMAR) units (P<0.001), and for DSEK relative to PK, it was -0.199 logMAR units (P = 0.045). The regression-adjusted marginal gain in utility for PK relative to no intervention was 0.128 QALYs (P<0.001) and for DSEK relative to PK was 0.046 QALYs (P = 0.031). This resulted in ICERs of $56 409 per QALY for PK relative to no intervention and $5209 per QALY for the more expensive DSEK relative to PK. CONCLUSIONS If the goal is to maximize societal health gains given fixed resources, DSEK should be the preferred strategy. For a fixed budget, it is possible to achieve greater QALY gains by providing DSEK to as many patients as possible (and nothing to others), rather than providing PK.