Two-stage Fill-it-up design method incorporating historical control data in binary endpoint clinical trials

Leveraging historical control data to augment randomized control data in clinical trials has become an important strategy for improving the efficiency of statistical inference, particularly in contexts with limited sample availability. However, potential heterogeneity between historical and current datasets may introduce bias in treatment effect estimation and compromise inferential validity. This study proposes a two-stage "Fill-it-up" design for clinical trials with binary endpoints to enable the rigorous integration of historical control data under controlled statistical risk. Analytical procedures for sample size determination and practical implementation steps for both stages are provided. Simulation studies demonstrate that the family-wise error rate can be effectively controlled below the pre-specified significance level, while the average sample size is substantially reduced compared with a conventional single-stage design that excludes historical controls. The efficiency gains become more pronounced as between-group heterogeneity decreases. The proposed two-stage Fill-it-up design offers a frequentist framework for safely and efficiently incorporating historical control data into binary endpoint trials. Given that additional recruitment is required when equivalence is not established, its practical application is best suited for studies where there is strong prior confidence in the quality and comparability of historical data, such as in rare disease or pediatric settings. This design provides a pragmatic approach for enhancing the efficiency and ethical sustainability of modern clinical research.