<
Can one estimate the final HR by assuming we track close to the combined study? If so, this seems to be a slam-dunk. (e.g. Final combined HR from previous study/1.28 * 1.25)>
I've been away all day and too many posts to catch up... Now, some assumptions:
1. First, the integrated data were adjusted somewhat to match with the given data from today PR, esp, the given HR of 1.25 and its 95% CI (Confidence Interval) [0.95,1.64].
2. Next, the company still did not to release the interim alpha or trigger. So we need to guess what that was in order to compute. From the fact that the interim occurred in May and DNDN have been saying that the event rate tracked the integrated data, the trigger was likely in the 240 range. Let's assume 243 as that is 75% of 304.
So, keep in mind those assumptions for the below:
1. As Mark Frohlich mentioned, the trial was originally powered via the final results of the integrated data with a target HR of 1.45. Since that final result was based on 164 events out of 225 enrolled patients, the corresponding number of events for IMPACT would be 373 (out of 512). However, extrapolating the above interim HR of 1.25 forward, the HR at 373 would have been about 1.40. So, the original target HR 1.45 was a bit high, especially since the original final trigger was only 360.
2. If the HR was around 1.40 at 373 events, the HR at the old final trigger of 360 would have been around 1.38. Continuing from there, the HR of the current final look at 304 would be around 1.30 with 95% CI [1.02,1.66].
3. The corresponding power for the final at 304 would likely be slightly higher than 60%. If they had stayed with the 360 trigger, the power would have been about 80%. So the SPA change did result in a significant power reduction given what we know of the data today. But anything above 50% is considered a good chance for this sort of estimation.
