Unblinded by the Night: Predictive Power for Complex Bayesian Adaptive Trials When Sight Privileges Vary

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Adaptive Trial Design

Background:

• Clinical trials require robust bias reduction strategies, including blinding investigators and sponsors.
• Complex adaptive designs necessitate unblinded statisticians for interim analyses and oversight.
• Data and Safety Monitoring Boards (DSMBs) require timely, accurate information for safety and efficacy monitoring.

Purpose of the Study:

• To propose a novel method for predicting clinical trial outcomes using current data.
• To enable blinded decision-making without compromising trial integrity.
• To provide updated Bayesian predictive power for complex adaptive trial designs.

Main Methods:

• Utilized Bayesian predictive power as a trial prediction method.
• Applied the method to simulated data from the Hyperbaric Oxygen Brain Injury Treatment (HOBIT) trial.
• Demonstrated updated Bayesian predictive power calculations for blinded and unblinded groups.

Main Results:

• Bayesian predictive power effectively predicts trial outcomes and sample size distribution.
• The approach facilitates resource allocation and decision-making for different trial teams.
• Updated predictions offer valuable insights into future trial behavior.

Conclusions:

• The proposed approach enhances guidance during clinical trial conduction.
• Bayesian predictive power is applicable to various complex adaptive trial designs.
• This method supports informed decision-making for both blinded and unblinded stakeholders.