Once, we need to do an Interim Analysis, I didn’t understand why we need to do it. Later I learned in Clinical Trial studies, our ultimate goal is trying to get approval for FDA submission at the end of studies. While exhausting and long progress, sometimes we perform Interim Analysis before the completed trial to access Safety, Futility, and Efficacy to decide if we should keep the study.
Today we will focus on the method to access the Interim analysis for Efficacy.
Interim analysis for Efficacy:
Perform the Primary Outcome prior to the end of the study to stop the trial prematurely if the treatment has demonstrated superiority.
Potential Problem for Multiple Comparision:
The P-value varies along the study time.
as we can see there are regions where P-value demonstrated the significance, whereas other parts it didn’t.
Possible Solution to adjust Type I error rules:
Reject Null when P-value is extremely small (<0.001) at each time point.
–Con: Hard to achieve, a little too conservative
–Pro: Guarantee the overall P-value <0.05.
P-value is the same at each interim analysis.
–Con: Number of interim analysis must be fixed, smaller aphla for
–Pro: Easy to use, Best chance to Reject Null early
In the early phase, adjust the alpha level to very small so Reject Null when p-value is extremely small. Later phase the alpha level becomes larger.
–Con: Hard to adjust
–Pro: Increasing chance to Reject Null as more data collected.
P.S.if the interim analysis is based on Safety, we won’t worry about Multiple Comparisons.
Futility is deflated alpha rate.
Thanks 77 for sharing the Harvard Catalyst class!