A method for the calculation of confidence intervals following a sequential clinical trial is proposed which is more accurate than methods based solely on continuous monitoring assumptions, but is not as computationally laborious as previous methods suggested for group sequential trials. The calculation takes account of the excess of the sample path over the boundary at the final inspection of the trial, which may be substantial after group sequential monitoring, and is ignored when continuous monitoring is assumed. Accuracy of such confidence intervals after a triangular test is investigated by simulation of coverage probabilities, individual confidence limit probabilities and p-value curves.
- Clinical Trials as Topic/methods
- Confidence Intervals
- Mathematical Computing
- Models, Biological
- Monitoring, Physiologic