No abstract

No abstract

final result might have shown a significant difference in favour of placebo. Such a result also would have been as 'clear as mud' as the plausibility of a toxic effect of progesterone was zero. We would have concluded a type-1 error.We also applied a Bayesian interpretation to the interim results, comparing the original null and alternative hypotheses. The likelihood ratio of the interim result was 133 for the null hypothesis over the alternative. If we started with equipoise regarding these hypotheses (50% chance of progesterone being effective or not), the probability of progesterone having no effect would be >99% following the interim analysis.Under the circumstances, we decided that the scientific gain from continuing the trial was minimal and that stopping the trial was the best option.