OBJECTIVES: Immunosuppresive therapy is required to prevent graft rejection. Older medicines such as tacrolimus are paradoxically toxic to the kidney, whereas newer therapies such as sirolimus (Rapamune) are not. The purpose of this study was to evaluate the relative cost-utility of sirolimus versus tacrolimus in the UK. METHODS: A stochastic simulation model was constructed using clinical trial and real-life data comparing the two treatments. Time duration was up to 20-years, 2003 prices, discounted at 6% for costs and 1.5% benefits and from an NHS perspective. Simulated events included patient and graft survival, haemodialysis, peritoneal dialysis, re-transplants and acute rejection. Costs were summed for events and various maintenance therapies. Utility was differentially accredited depending upon survival using the alternative renal replacement therapies. Outcome was predicted using post-transplant creatinine levels up to 3-years. Extensive statistical economic analysis and sensitivity analysis was undertaken. RESULTS: Extensive validation demonstrated that the simulation was very reliable. Over the 10-year horizon, sirolimus gained 0.58yrs (discounted) of functioning graft over tacrolimus, resulting in an incremental cost per year of functioning graft that was dominant (ICER was calculated at -£39,576). Over a 20-year time horizon cost effectiveness of sirolimus over tacrolimus further improved with an average discounted gain in years of a functioning graft of 1.5yrs, resulting in an incremental cost-utility that was dominant (ICUR -£46,695). The number of haemodialysis events was 48,243 on sirolimus versus 127,829 129, on tacrolimus and peritoneal dialysis events 40,872 versus 105,249, respectively. Sirolimus remained dominant over tacrolimus under all scenarios. This findings were robust using statistical economic analysis and sensitivity analysis. CONCLUSIONS: Sirolimus was far more cost-effective than tacrolimus and was economically ‘dominant’. The magnitude of this difference indicates that this finding is likely to be geographically generalisable.