Purpose: The purpose of this study was to compare methods of cost effectiveness analyses from a net benefit perspective using the economic data from the Clopidogrel for the Reduction of Events During Observation (CREDO) trial. Methods: Net benefit analysis was applied to the CREDO economic data for cost effective thresholds ranging from $0 to $100,000. Net benefit was estimated by bootstrap, ordinary least squares regression (OLS) and robust regression methods. Robust regression has been suggested as an alternative to OLS to account for outliers in economic data. Results: The results of the CREDO trial showed a 27% relative risk reduction in cardiovascular events for clopidogrel plus aspirin vs. aspirin alone over a one year follow-up period. An economic analysis from an incremental cost effective ratio (ICER) perspective indicated that clopidogrel was cost effective with an ICER of $4,353 per life year gained using Framingham estimates of survival, and $3,460 per life year gained using Saskatchewan estimates. The net benefit analysis indicated that clopidogrel was cost effective with net monetary benefit of $1,275 ($69-$2,481) at a cost effective threshold of $4,000 (OLS p = 0.0384) for Framingham estimates, and $1,432 ($225-$2,639, OLS p = 0.0201) for Saskatchewan estimates for the same threshold. The bootstrap estimate was similar: $1,282 ($61-$2,518). Robust regression estimates varied depending on the method, but were generally $200-$300 less than OLS and bootstrap estimates. Robust methods that weighted residuals gave estimates over the range of thresholds that differed by less than $400 ($923 - $1,317). Conclusions: Net benefit analysis of the CREDO economic data supported the conclusion that clopidogrel was cost effective. However, OLS regression assumptions of normally distributed residuals and homogeneous variance were clearly violated. Robust regression methods also indicated cost effectiveness of clopidogrel, but resulted in similar estimates of benefit across cost effectiveness thresholds rather than increasing benefit with increasing cost effective thresholds as would be expected. Unmet OLS regression assumptions along with differences in net benefit estimates depending on robust regression methods suggest additional investigation of the use of net benefit as an alternative to the cost effectiveness ratio is needed.