OBJECTIVE: In multinational trials, hospital costing is estimated by assigning diagnosis- and country-specific unit cost estimates to hospitalizations. Using this methodology, the standard error for hospitalization costs is underestimated. Our objective is to expand methods for international hospital costing to increase the validity of hospitalization costs and to explore the effects of using different methodologies on statistical power. METHODS: In this analysis, 'standard' costing methodology assigns a fixed cost to each hospitalization. The 'expanded' methodology involves: (1) using DRG weights to adjust country-specific cost estimates for diagnoses for which cost data were not collected; (2) differentiating between costs that occur on the day of admission and all remaining hospital days; and (3) adjusting cost estimates by length of stay for each hospitalization. Using data from a subset of placebo-treated patients enrolled in a multinational study of a new treatment for heart failure, we compared estimates of total hospitalization costs using 'standard' and 'expanded' methodologies. RESULTS: Total hospitalization costs were estimated for 200 patients who had at least one hospitalization during the follow-up period (mean=1.42). Using 'standard' and 'expanded' methods, average total costs were $5,911 and $5,881, respectively. The corresponding standard errors were $269 and $417, an increase of 55% resulting in a loss of power. Assuming homoskedasticity, equal sample size, ?=0.05, 2-sided t-test, and a $500 difference in costs, one would have 25.7% power using 'standard' costing methods but only 13.5% when using 'expanded' methods. With a $1,000 difference, one would have 74.5% and 39.3% power, respectively. CONCLUSIONS: The results of a cost comparison between treatment strategies can be affected by the methodology used to calculate costs. Studies that consider the intensity of hospital care, or that consider the variability in resource utilization, require a larger sample size than studies that don't consider these issues to have equal power.