bayesassurance: An R package for calculating sample size and Bayesian assurance

TL;DR

The paper introduces the bayesassurance R package, which calculates Bayesian assurance for clinical trial design using simulation-based methods, accommodating various assumptions and linking Bayesian and frequentist approaches.

Contribution

It presents a novel R package with a two-stage Bayesian framework for computing assurance, enhancing flexibility and addressing limitations of single-prior methods.

Findings

Demonstrates the package's application through detailed examples.

Shows how Bayesian assurance can overlap with frequentist power.

Highlights advantages of the two-stage Bayesian approach.

Abstract

We present a bayesassurance R package that computes the Bayesian assurance under various settings characterized by different assumptions and objectives. The package offers a constructive set of simulation-based functions suitable for addressing a wide range of clinical trial study design problems. We provide a detailed description of the underlying framework embedded within each of the power and assurance functions and demonstrate their usage through a series of worked-out examples. Through these examples, we hope to corroborate the advantages that come with using a two-stage generalized structure. We also illustrate scenarios where the Bayesian assurance and frequentist power overlap, allowing the user to address both Bayesian and classical inference problems provided that the parameters are properly defined. All assurance-related functions included in this R package rely on a two-stage Bayesian method that assigns two distinct priors to evaluate the unconditional probability of observing a positive outcome, which in turn addresses subtle limitations that take place when using the standard single-prior approach.