gms | German Medical Science

Introduction: Using a group sequential or adaptive design within a clinical trial, an investigator can perform planned interim analyses to draw conclusions from the data collected so far. However, data collection and analysis needs time, and new patients will enter the trial while the interim analysis is ongoing. Moreover, when using a binary endpoint, some patients may not yet have completed their follow-up at the date of the interim analysis and may thus not be analyzed. Such patients are referred to as “interim patients”. The number of interim patients essentially depends on duration of the interim analysis.

If the superiority of a treatment is shown in the interim analysis, the trial is stopped and the relevant office of regulatory affairs is informed. Although interim patients are not part of the interim analysis, the data collected on those patients have to be sent to the office of regulatory affairs as well and will thus be analyzed in some way. If the effect of treatment on interim patients turns out to contrast the superiority observed in the interim analysis, this may nevertheless lead to the withdrawal of the positive conclusion. Therefore, it appears sensible to consider interim patients already in the planning phase of a trial in order to avoid spending type I error in situations of contrasting effects of treatment on interim patients.

Methods: General strategies for including patients recruited during interim analyses of a trial were proposed by Faldum et al. [1]. Here, we explicitly exploit and discuss a two-stage adaptive design with interim patients for the exact binomial test. The exact binomial test might be the method of choice in a phase II setting when sample size is limited (so that potential normal approximations do not hold true) and when the primary endpoint is binary, e.g. tumor response, which is defined by whether a response to treatment is observed after a fixed time span or not. The success rate of the (small) population under the new treatment is then compared to that of a historic control.

A two-stage adaptive design algorithm with interim patients is explicitly described and performance of the algorithm is illustrated using a clinical example. In particular, the algorithm avoids rejection of the null hypothesis after the interim analysis in case of contrasting effects of treatment on interim patients.

Results: Standard group sequential or adaptive designs do not take repeated analysis with interim patients into account. Consequently, the resulting test is conservative. By means of the procedure considered here, this shortcoming is removed and the repeated analysis with interim patients is explicitly taken into account already in the planning phase of a trial. In particular, the power of the trial may this way be increased as compared to standard group sequential or adaptive designs.

Discussion: The techniques developed in [1] were applied to construct a two-stage adaptive design for the exact binomial test. The proposed design is easy to implement and sample size determination is easy to perform. In particular, the test decisions for interim patients as well as the test decisions for the second stage do not depend on the results of the first stage. Therefore, the proposed design is particularly easy to communicate and interpret in a medical background and is additionally more powerful than the corresponding standard adaptive designs without interim patients.