OBJECTIVES: Many authors have addressed the issue of handling uncertainty in stochastic cost-effectiveness analyses. They generally oppose parametric (Fieller's theorem, confidence box or ellipse...) and non parametric (bootstrap) methods. The parametric bootstrap is at the bridge of these two families of methods. Parametric bootstrap was used to assess confidence interval of the incremental cost-effectiveness ratio (ICER) of a drug used in the treatment of post-infarction patients and the estimated CI was compared to that provided by non parametric bootstrap. METHODS: Data were derived from a placebo-controlled clinical trial on 1749 patients included just after MI and followed over 4 years. Patient-level costs were computed to reflect the US Health care system. The parametric bootstrap was based on the assumption that the incremental costs and effects differentials follow a normal bivariate distribution. Five thousand re-samples were made from the normal bivariate density function. The non parametric bootstrap was based on 5000 re-samples from the original sample. Mean ICER estimates and percentiles CI are presented for both techniques. RESULTS: The mean ICER from the original sample was estimated to US$2574 per life-year saved. The parametric bootstrap provided an estimated ICER of US$2726/LYS and the 95% CI was [1164; 5210]. From the non parametric bootstrap, the ICER estimate was US$3,419/LYS and the 95% CI was [1660; 5254]. CONCLUSIONS: The parametric bootstrap provided a CI, which seemed to be in accordance with that of the non parametric bootstrap. It still needs to be compared to other parametric and non parametric methods, particularly in terms of coverage properties. It can be a useful tool, as it allows assumptions on the ICER joint distribution, without implying specific textbook formulae.