gms | German Medical Science

Introduction: Formal proof of efficacy requires that in a prospective clinical trial, superiority (or sometimes non-inferiority) towards a control is demonstrated. Often, one primary endpoint is specified, but various diseases (e.g., Alzheimer’s disease) are complex. Thus, treatment success needs to be based on the assessment of more than one primary endpoint. If co-primary endpoints are defined, all endpoints have to be significant as a requirement for study success. No adjustment of the type-1-error is needed here, but the sample size needs to be increased in most cases to maintain an acceptable power. Recently, the so-called “dual endpoint concept” has been proposed, where a significant result in at least one primary endpoint is sufficient to claim study success. The study-wise type-1-error is controlled at the pre-defined level by applying a Bonferroni-split of the type-1-error between endpoints. Obviously, such an approach is not in line with conclusive decision-making because study success could be declared when one endpoint shows significant superiority while a deterioration is observed for the other one. Consequently, this concept is not included in the EMA multiplicity guideline [1].

Methods: Based on Röhmel et al. [2], we propose an approach that dissolves the described contradictory situation. We investigated different additional conditions to the dual endpoint concept to exclude a pre-specified non-acceptable deterioration for all primary endpoints. Specifically, either the treatment effect estimate needs to be on the right side of zero, a positive trend is required for the primary endpoints, or non-inferiority has to be demonstrated. Additional, superiority to the control has to be shown for at least one of the primary endpoints. Our approach is realized by the performance of a hierarchical two-step procedure. In the first step, the respective additional condition to exclude harm has to be fulfilled for all primary endpoints. Secondly, superiority needs to be shown for at least one. A simulation study was performed to examine our approach in comparison to the dual endpoint concept regarding costs in terms of power, and, therefore sample size. The traditional co-primary concept was also included as a yardstick. Simulation scenarios included various treatment effects and correlations for three different sample sizes.

Results and conclusion: The simulation study shows that if planning assumptions are correct, additional conditions to the dual endpoint concept for claiming study success avoid post-hoc discussions and improve interpretation with limited impact on costs in terms of sample size. Moreover, our proposed approach has the advantage that the minimum requirements for endpoints can be modeled flexibly for a collection of practical needs. In summary, our work emphasizes that the more aspects are discussed and pre-specified in the planning stage, the better the certainty and interpretability of its results.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.