Sample size determination for decision-centered pragmatic trials

Area of Science:

• Clinical Trials Methodology
• Decision Analysis
• Biostatistics

Background:

• Pragmatic randomized trials (RCTs) are crucial for comparing real-world treatment options.
• Determining appropriate sample size in pragmatic RCTs is challenging, balancing ethical considerations and resource allocation.
• Existing methods for sample size determination in pragmatic RCTs are limited.

Purpose of the Study:

• To develop a decision-analytic (DA) method for sample size determination in pragmatic RCTs.
• To anchor sample size calculations to stakeholder-defined minimally important differences (MIDs).
• To enable selection of the superior treatment based on net clinical benefit.

Main Methods:

• Modeled a two-arm RCT using a decision tree incorporating treatment benefits and harms.
• Weighted outcomes by stakeholder values and preferences (relative value, RV).
• Set sample size to ensure expected loss from wrong decisions (γ·Δ) does not exceed acceptable regret (ARg).

Main Results:

• The DA approach required approximately half the participants compared to conventional frequentist designs (mean N=89 vs. 168).
• Sample size reductions in three of four recent pragmatic trials ranged from 47% to 72%.
• The framework identified potential issues in one trial, highlighting the need to align effect size, stakeholder preferences, and decision risk.

Conclusions:

• Designing pragmatic RCTs around decision quality offers transparency and ethical coherence.
• The DA method can substantially reduce sample size, potentially lowering accrual barriers.
• This approach accelerates the generation of actionable pragmatic evidence.