zahrcast
zahrcast
Sia forecasts the results an ongoing trial will report at maturity, with credible intervals, from the data collected so far. It links each patient's disease burden to progression and death, and borrows from historical trials.
Interim decisions come a few months into follow-up, when most patients are still censored. The usual answer is to wait for more events. Enrollment and spending continue while you wait, and a pivotal trial costs a median of about $19m.1
The burden measurements already collected often carry information about the final result that an event count does not use.
The range of results the trial could still report, with the probability of each. The same output answers:
Most interim methods extrapolate a survival curve from the events observed so far. Sia models each patient's disease burden over time and derives progression and death from it. A patient with four assessments adds little to an event count, but a lot to an estimate of their trajectory.
Fit to each patient's measurements and events so far, borrowing from historical trials.
Simulate each patient forward from today, thousands of times. Some futures end in progression.
Each simulation is a completed trial. Read any endpoint off them, with its uncertainty.
Every forecast comes with its uncertainty, as a credible interval or a probability.
Sia estimates how much to borrow from historical trials. When they disagree, it borrows less.
Forecasts are scored against data that arrived after the cut-off they were made at.
Sia was fitted at early cut-offs of completed trials and checked against their final results.2