Thanks Biobonic
Predetermined efficacy analysis can allow to stop the trial for efficacy, or stop the trial for futility as that has happened for Pfizer trial. Usually, predetermined, or interim, analysis are set at at least 50% Of the required events. With less events, it is very likely that any futility analysis will not detect with enough certainty that the trial will be futile in the end or reversely that the drug is efficient (lack of power). To stop early for efficacy, the effect size has to be much greater than anticipated.
In Cel SCI's case there is not such predetermined study planned, because the power of the study is set at design by the protocol. There have been many convincing and well documented explanations as to why this study could not be ended early for efficacy. However it could very well have been ended earlier for futility because of the conditional power analysis they are doing.
Conditional power also allowed to increase the sample size if required (eg compensate dropouts and enroll more), meaning that there won't be enough people/events in a timely manner. In CEl SCI's case, with 298 events needed, 928 recruited, I do not see really how more enrollment would have been required, either than to accelerate the number of events, because 298 can always be reached with 928.
The fact that there is not such predetermined stop points, does not mean that significance cannot be calculated early by the IDMC. In fact, significance has to be calculated when they do conduct the "conditional power and sample size". It is not only intended at determining if more enrollment is required, is it vitally aimed at simulating a "best case scenario" (for instance if all the events left fall into the control arm), and see if significance will be shown in the end. If significance will not be shown in the best case scenario, then it will not be shown whatever happens and the study should be shut down. On the other hand, they can simulate a "worse case scenario" and see what happens if all the remaining events fall into the test arm. If significance is reached still, that means that the effect size is likely above the 10% from the study design, that the study can continue and will be successful.
The in-between scenario, when significance is shown in the best case scenario and not in the worse case scenario happens throughout the IDMC meeting. That's the standard scenario, however when it comes to the very end, with 10 events left, or even five events left, there is little probably that the effect size is right at 10%. It should be either close to 0 or close to the phII observed 30%+
Therefore, with this in mind, thinking that the IDMC only had 1 or 2 events left last time that they met, for me it is mathematically highly unlikely that they have not seen significance in both scenarios : the best and the worse. If IDMC has done its job of course.
Fosco
