Bayesian statistical methods continue to gain in popularity with researchers thanks to their ability to integrate prior information, real world data and expert opinions into their estimates. This results in better decision making at key milestones in drug development.
What is Assurance (Bayesian Power)?
- Assurance is a calculation that integrates a number of parameters surrounding a trial or drug development that provides an indication of the "true probability of success of a trial". In frequentist terms, this is akin to providing a statistically significant result.
- Assurance can be calculated as the expected statistical power based on a prior distribution of the unknown parameters related to the effect size. This provides a "Bayesian Power" estimate which can be used as an alternative to the traditional power analysis used in most intervention research.
- Assurance may also be considered as a Bayesian formalization of a sensitivity analysis
How does Bayesian Assurance work?
- To calculate Bayesian Assurance, priors have to be elicited from expert opinions and/or existing data and then integrated into frameworks such as the Sheffield Elicitation Framework (SHELF).
- Through the elicitation process all relevant data is summarized, reviewed and implicitly weighted.
- Based on the experts responses and weighting, biostatisticians can develop a probability distribution that demonstrates both current knowledge and uncertainty of the current milestone move.
- Assurance (probability of success) is the power averaged over all plausible values by assigning prior to one or more parameters, providing a summary statistic for the effect of parameter uncertainty. This can highlight unforeseen problems and lead to better decision making or appropriate reviews.
The benefits of Bayesian Assurance
- Bayesian Assurance - The True Probability of Success
Using the Bayesian module in nQuery and a framework such as SHELF leads to a user friendly method to use Assurance (A) in decision making across many levels. Having a greater understanding of the true probability of success for a given scenario allows for better decisions regards finance and risk.
- Identify Study Design Threats & Opportunities
- Bayesian Techniques such as Prior Elicitation provides a systematic approach to reduce the analysis gap between completed studies and planned studies when there is a lack of reliable evidence or scientific consensus.
- Examining experts prior distribution may often reveal concerning aspects of the study design that was previously overlooked.
- The elicited prior distribution can be used to assess various study designs. These can ultimately be influential in the appropriate ‘go’ or ‘no go’ decision.
- The elicitation process highlights not only the rationale for believing in the likely effect of the drug, but the gaps in knowledge and/or sources of uncertainty
- Communicate Complex Findings to Non-Statisticians
The prior elicitation process is methodically and scientifically-driven. It enables a more vigorous review of data and decision making. However through the nQuery Bayesian module, researchers are presented a Bayesian Assurance calculation that can be communicated to non-statisticians, including internal governance boards, in an easy to understand way.
A more robust understanding of risks and insights means these can be formally acknowledged and plans made to mitigate these risks to reduce costs and thus reduce the likelihood of failure.
Being able to formally capture internal and expert opinion, integrate existing data and identify obstacles in the study protocol, teams can provide boards with formally appropriate estimates of the probability of trial success as well as robust plans for interim decision rules where appropriate, enabling better portfolio and company-wide decision making.
Using nQuery for your Bayesian clinical trial design
nQuery presents an easy-to-use formal method of discovering the true probability of success. Bayesian Assurance is a vital contextual tool in the biostatisticians planning toolbox. It places uncertainty at the heart of sample size determination.