Basket trials test a single therapeutic treatment on several patient populations under one master protocol. A desirable adaptive design feature is the ability to incorporate new baskets to an ongoing trial. Limited basket sample sizes can result in reduced power and precision of treatment effect estimates, which could be amplified in added baskets due to the shorter recruitment time. While various Bayesian information borrowing techniques have been introduced to tackle the issue of small sample sizes, the impact of including new baskets into the borrowing model has yet to be investigated. We explore approaches for adding baskets to an ongoing trial under information borrowing. Basket trials have pre-defined efficacy criteria to determine whether the treatment is effective for patients in each basket. The efficacy criteria are often calibrated a-priori in order to control the basket-wise type I error rate to a nominal level. Traditionally, this is done under a null scenario in which the treatment is ineffective in all baskets, however, we show that calibrating under this scenario alone will not guarantee error control under alternative scenarios. We propose a novel calibration approach that is more robust to false decision making. Simulation studies are conducted to assess the performance of the approaches for adding a basket, which is monitored through type I error rate control and power. The results display a substantial improvement in power for a new basket, however, this comes with potential inflation of error rates. We show that this can be reduced under the proposed calibration procedure.
