OBJECTIVES: The gap between “efficacy” in clinical trials and ”effectiveness” in real practice due to patient selection, non-compliance and treatment patterns has been broadly recognized among health-care decision makers. However little attention has been paid for statistical errors accompanied in clinical trials, which may cause the gap. The purpose of this study is to develop a new analytical framework to estimate the degree of the gap caused by statistical errors. METHODS: The expected cost-effectiveness ratio (TP) in the real world was formulated in terms of a vector by using a decision analytic approach assuming an evidence-based treatment choice from two treatment options: a new treatment (TA) and a conventional treatment (TB). The value of TP varies depending on the parameters such as cost-effectiveness evidence of TA and TB, type ? and type ? errors, and a Bayesian prior probability that TA is better than TB. The prior probability is usually estimated by the following two steps: 1) Bayesian inference based on binomial modeling, and 2) Normal approximation of the binomial modeling with the log odds ratio. Then the derived prior probability was applied for our formula to estimate the gap, called “gap formula”, by using a hypothetical dataset. In addition, a Monte Carlo simulation was performed as probabilistic sensitivity analysis. RESULTS: Adaptive estimations of Bayes provided flexible evaluation for the gap formula between real world and clinical trials. The result of the Monte Carlo simulations quantitatively illustrated with distributions how much the TP could be deviated, mainly depending on the type ? error, from the ideal average cost-effectiveness ratio. CONCLUSION: The series of methods we developed can be applied in adaptive design for trial-based cost-effectiveness analyses such as sample-size calculation considering the influence of statistical errors.