Evaluate individual patient level outcome data within live, blinded clinical trials which was previously inaccessible without breaking blinding.
Identify in real-time the point which sufficient data has been collected to predict a failed trial outcome at the agreed confidence level.
Protect trial integrity by going further than unblinded interim analyses to eradicate potential
risks of bias
Using 25,000 trial scenarios derived from real trial data on clinicaltrials.gov, our current computational model recommended stopping:
- 1/4 of failing trials before 33.4%
- 1/2 of failing trials before 49.7%
- 3/4 of failing trials before 68.4%
This model had a low type II error rate of 0.01 for predicting trial outcome.