Evidence synthesis from aggregate recurrent event data for clinical trial design and analysis

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Evidence Synthesis

Background:

• Historical clinical trial data is vital for designing, monitoring, and analyzing new studies.
• Synthesizing evidence from historical trials, often in aggregate form, is essential for informed decision-making.
• Recurrent event endpoints are common in many therapeutic areas but pose challenges for meta-analysis.

Purpose of the Study:

• To develop and validate methods for evidence synthesis of historical trials with recurrent event endpoints when only aggregate data is available.
• To enable meta-analysis using negative binomial regression models without individual patient data.
• To illustrate the application of these methods in designing and analyzing a new clinical trial.

Main Methods:

• Described approaches for back-calculating model parameter estimates and standard errors from summary statistics.
• Utilized approximate Bayesian computation and a quadratic approximation to the log-likelihood.
• Employed a Bayesian hierarchical meta-analysis model to estimate parameters.

Main Results:

• Simulations demonstrated that the proposed approach yields inference comparable to using individual patient data.
• The back-calculation method effectively synthesizes aggregate data from historical trials.
• The methodology was successfully illustrated using data from 11 historical trials for a severe eosinophilic asthma trial.

Conclusions:

• The developed methods provide a viable solution for meta-analysis of recurrent event data from historical trials with aggregate summary statistics.
• This approach enhances the utility of published literature for informing future clinical trial design and analysis.
• Accurate meta-analysis is achievable even without direct access to individual patient data.