In a design where formal hypothesis testing is used and the interim analyses provide multiple opportunities to stop for efficacy, the overall trial significance level will be greater than nominal significance levels used at each stage. Any stopping rules should adjust for this, and correspondingly any confidence intervals or p-values presented must be calculated to adjust for the possibility of stopping earlier and for having reached the observed stage in the trial.<br>Conversely, a trial that only has the option to stop early for futility will conservatively preserve the overall significance level. Here the nominal confidence interval and p-value at the end of the trial can be used. Investigations should be made into the effect on the power of the trial, and only if the power is substantially reduced should adjustments be used.
đang được dịch, vui lòng đợi..
