Predictive event modelling in multicenter clinical trials with waiting time to response

Area of Science:

• Statistics
• Clinical Trials
• Biostatistics

Background:

• Predictive modeling in clinical trials is essential for event forecasting.
• Existing methods may not adequately account for dynamic recruitment and complex event types.

Purpose of the Study:

• To develop an advanced statistical technique for predictive event modeling in ongoing multicenter clinical trials.
• To provide predictive means and bounds for the number of events over time, incorporating patient recruitment and risk factors.

Main Methods:

• Utilized an advanced Poisson-gamma model for patient recruitment, accounting for temporal and inter-center variations.
• Employed finite Markov chains in continuous time to model event occurrences (recurrence, death, loss-to-follow-up).
• Integrated Bayesian re-estimation for predictive recruitment rate adjustments based on interim trial data.

Main Results:

• Derived closed-form expressions for predictive mean and bounds of future events under Markovian assumptions.
• The technique effectively models various recruitment and event scenarios in ongoing oncology trials.
• Demonstrated efficient application through case studies of real-world oncology trials.

Conclusions:

• The developed statistical technique offers robust predictive capabilities for event counts in ongoing clinical trials.
• This method enhances trial monitoring and decision-making by providing reliable future event estimations.
• Applicable to diverse clinical trial designs, particularly in oncology, to manage recruitment and event prediction.