OBJECTIVES: To propose a sample size calculation method for randomized trial-based cost-effectiveness analyses (RTBCEA), coherent with recommendations to express uncertainty through cost-effectiveness probabilities and expected value of information. METHODS: Let’s estimate the sample size of a RTBCEA with 2 parallel-groups of equal size n, assuming equal variance of costs (s²_C) and effectiveness (s²_E) in each group. Our method is based on the incremental net monetary benefit (b~N(μ_b,σ²_b)). A realization of this random variable in the planned RTBCEA would yield a sampling distribution of mean μ_b and variance 2σ²_bi/n, where σ²_biis the variance of the individual net monetary benefit in each group. σ²_bi=λ²σ²_E+σ²_C–2λρσ_Eσ_C, where λ is the ceiling cost-effectiveness ratio, and ρ the correlation between cost and effectiveness. Using this sampling distribution of b, the remaining value of perfect information (or the expected value of perfect information depending on n (EVPI_(n))) is calculated, specifying the size of the target population (N) and a discount rate (r). The total cost of the planned RTBCEA is defined by a fixed cost (C_f) and a cost/patient (C_v). The optimal sample size of one arm in the planned RTBCEA is n when EVPI_(n)=C_f+2nC_v