Bayesian Predictive Probability of Success (PPoS) Calculator
Estimates the predictive probability that a trial will meet its success criterion at the final analysis, given the data observed so far. PPoS supports interim Go/No-Go and futility decisions by averaging the chance of eventual success over the posterior for the unknown parameter.
How it works
Given interim data and a prior, the tool forms the posterior for the treatment effect, then integrates over that posterior to project the probability that the completed trial will cross its pre-specified success threshold. Efficacy and futility thresholds are configurable. Predictive probability differs from posterior probability (the chance the effect exists now) and from frequentist power (a pre-trial property under a fixed effect).
When to use it
- You are running an interim analysis and want a Go/No-Go or futility read based on the projected chance of final success.
- You want to combine observed interim data with a prior rather than rely on a fixed assumed effect.
- You are quantifying whether continuing the trial is likely to be worthwhile.
Assumptions & limitations
- Predictive probability is not frequentist power and does not by itself control the frequentist type I error rate; a monitoring rule based on it should have its operating characteristics evaluated by simulation.
- Results depend on the chosen prior and success criterion; a different prior or threshold can change the Go/No-Go conclusion, so sensitivity analysis is advisable.
- PPoS quantifies the projected chance of success for the planned analysis — it is not a guarantee about any single realized trial.
For the full methodology, derivation, and worked examples, read the complete guide.