Statistical Calculators › Bayesian Sample Size Methods › Posterior Probability of Success at an Interim Analysis
← All Calculators

Bayesian Sample Size Methods

Posterior Probability of Success at an Interim Analysis

Bayesian interim-analysis calculator for a binary endpoint. Computes the posterior probability that the response probability exceeds a prespecified null value, using a Beta prior and the observed interim number of responses. Runs entirely in your browser.

Interim Data & Bayesian Prior

Specify the interim number of responses, the response rate under the null, and the Beta prior for the response probability.
Beta(a, b) prior for p

Interim Result

The posterior probability is the Bayesian probability, after observing the interim data, that the response probability exceeds the specified null value.
Enter interim data and click Calculate Posterior Probability.

Methodology

This calculator uses the conjugate Beta-Binomial model for a binary endpoint. If the interim data contain x responses among n patients and the prior distribution is Beta(α, β), the posterior distribution is Beta(α + x, β + n − x).

Prior: p ∼ Beta(α, β)
Data: x | p ∼ Binomial(n, p)
Posterior: p | data ∼ Beta(α + x, β + n − x)

Posterior Probability of Success = P(p > p0 | data) = 1 − Ip0(α + x, β + n − x)

Here, Ip0(·) is the regularized incomplete beta function. The calculator evaluates the beta tail probability directly, without Monte Carlo simulation or an external statistical server.

Interim Success Rule

A trial is classified here as meeting the specified Bayesian interim success criterion when P(p > p0 | data) ≥ η, where η is the user-specified posterior probability threshold. If the posterior probability is below the threshold, the calculator reports that the specified success criterion has not been reached.

This is a posterior-probability approach: it evaluates the probability of the hypothesis given the data already observed at the interim analysis. It should not be confused with Bayesian predictive probability of success, which additionally integrates over possible future observations. The FDA's Bayesian guidance distinguishes these two interim approaches.

Worked Validation Example

A published illustration considers 12 responses among 20 patients, a uniform Beta(1,1) prior, and a null response probability of 0.50. The resulting posterior probability is approximately 0.81.

x = 12    n = 20
p0 = 0.50
Prior = Beta(1, 1)
Posterior = Beta(13, 9)
P(p > 0.50 | data) = 0.80834484 ≈ 0.81

The calculator reproduces this value to numerical precision. With a success threshold of 0.95, this particular interim result therefore does not meet the specified Bayesian success criterion.

References

Saville, B.R., Connor, J.T., Ayers, G.D., & Alvarez, J. (2014). The utility of Bayesian predictive probabilities for interim monitoring of clinical trials. Clinical Trials, 11(4), 485–494.

The paper gives the worked binary-endpoint illustration of 12 responses among 20 patients under a uniform prior, with P(p > 0.50 | data) = 0.81. Published article

U.S. Food and Drug Administration. (2010). Guidance for the Use of Bayesian Statistics in Medical Device Clinical Trials.

The FDA guidance describes posterior-probability approaches at interim analyses and distinguishes them from predictive-distribution approaches. FDA guidance

Statsols /. Bayesian Clinical Trial Design Software. Bayesian module as supporting Bayesian testing and predictive-power methods for clinical trial design and interim assessment.