Validation of Predictive Analyses for Interim Decisions in Clinical Trials

Area of Science:

• Clinical Trials
• Biostatistics
• Translational Medicine

Background:

• Adaptive clinical trials dynamically adjust based on interim predictions.
• Poor selection of Prediction Analyses and Interim Decisions (PAID) plans risks patient safety and trial integrity.

Purpose of the Study:

• To present a method for evaluating and comparing candidate PAID plans for adaptive trials.
• To guide the incorporation of predictive data into critical interim decisions.

Main Methods:

• Leveraging completed trial datasets for validation of PAID plans.
• Utilizing interpretable validation metrics to assess candidate PAID strategies.
• Examining various PAID complexities, including biomarkers and external data, in a glioblastoma trial.

Main Results:

• Validation using historical data and electronic health records supports PAID plan selection.
• Ad hoc simulations without real-world data lead to suboptimal PAID evaluations and inaccurate trial operating characteristics.

Conclusions:

• Real-world data and completed trial analyses are essential for selecting optimal PAID plans.
• This approach enhances the reliability of predictive models and interim analysis rules in adaptive trials.