negative comments., page-13

Yes Moreforme agreed – interim analyses are fairly common in P3 trials for the reason you outline – cost. It would be handy to see the report and read the argument. But on the face of it if the author is arguing that because the interim analysis is performed on a smaller sample size it will be less likely to find a statistically sig effect then this is very misleading.

In a carefully designed sequential P3 trial critical values are predetermined to indicate whether a tx is worse than, better than, or not able be shown is different from your control group. An analysis on your outcome is performed at say 100 patients – and the test statistic compared to the critical value adjusted for the N (in terms of CIs) in the sample at the time.

When you think about it the design makes sense – why keep going at great expense on trial that is never going work – or on the other hand why keep going if the effect is proven earlier than expected. So interim analysis in the context of a sequential trial is hardly a disaster in my book – and certainly doesn’t reduce your chances of finding a positive result – as is being claimed.