Optimal timing for an accelerated interim futility analysis incorporating real world data

Area of Science:

• Clinical Trials
• Biostatistics
• Health Services Research

Background:

• Randomized controlled trials (RCTs) use interim monitoring for early stopping due to safety, efficacy, or futility.
• Futility monitoring aids resource reallocation, but conventional methods require substantial data, limiting early decisions.
• Early futility stopping doesn't inflate Type-I error, offering opportunities for external data integration to boost efficiency.

Purpose of the Study:

• To develop and evaluate a Bayesian approach for interim futility monitoring in RCTs.
• To leverage external real-world data (RWD) to improve the efficiency of futility monitoring.
• To assess the impact of data exchangeability on sample size reduction.

Main Methods:

• Proposed a Bayesian framework using Semi-Supervised MIXture Multi-source Exchangeability Models (SSMIX) for futility monitoring.
• Incorporated RWD and trial data, accounting for measured and unmeasured differences.
• Utilized predictive probabilities for futility assessment and investigated optimal timing relative to expected sample size under the null hypothesis.

Main Results:

• The Bayesian approach reduced the expected sample size by approximately 70 participants under the null when external and trial data were exchangeable.
• Even when exchangeability assumptions were not fully met, the method still achieved a 10-20 participant reduction in expected sample size under the null.
• The proposed futility monitoring maintained approximately 80% power across evaluated scenarios.

Conclusions:

• Integrating external RWD into interim futility monitoring offers a promising strategy to enhance clinical trial efficiency.
• This approach avoids Type-I error inflation, making it a valuable tool for optimizing resource allocation in clinical research.
• Developing regulatory-accepted methods that combine RWD and trial data is crucial for efficient clinical trial operations.