gms | German Medical Science

Background: Phase IIb actively controlled trial with event outcomes shall be designed aiming to obtain sufficient information for planning the subsequent phase III trial. Hence, the sample size of the Phase IIb trial shall be large enough to determine the treatment outcomes with sufficient precision.

Methods: A study is planned to compare two groups, denoted by X and Y with respect to specific event rates. The expected event rates are 25% in group Y and approx. 15% in group X for an observation period of 15 months. At the end of the study, EX events were observed in group X and EY events in group Y. The study will be considered positive if the probability of superiority of X (defined as 10 percent points increase in survival compared to Y) given that the observed event rate is at least a given number XX%.

Two methods have been used: a) a frequentist simulation using truncated exponentially distributed event times and b) a Bayesian derivation of the conditional probability of superiority based on Beta-priors, in order to determine a cost efficent sample size of the Phase IIb trial. In addition, it was planned to provide adequate timing and criteria for interim futility analyses.

Results: Various simulation scenarios were evaluated by the trial team. The key outcome was that almost any numerical advantage of the observed event rates (i.e. EX<EY) should lead to a progression to the Phase III trial, when the estimated probability P[HR<0.80] would be at least 50%, as this result would support the assumption of relevant superiority of X.

For example, using a total sample size of 100 subjects, the expected number of events would be at least 16 with a probability of 94% under the planning assumptions. The simulated probility of success in this case would be 80%. For the alternative case of an equal event rate of 25% in both groups, the simulated success rate was derived as 29%. The interim analyses could be planned at 6, 12, and 18 events, using the futility criteria of Freidlin [1].

Conclusion: On one hand, the frequentist approach for the sample size determination was easier to implement and to discuss with non-statistical team members. On the other hand, the criteria for the interim analyses were discussed favourable for the Bayesian approach.

Employee of medac.

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