Individual allocation had an advantage over cluster randomization in statistical efficiency in some circumstances

Area of Science:

• Biostatistics
• Clinical Trial Design
• Epidemiology

Background:

• Contamination in control groups of randomized trials can reduce statistical power.
• Cluster randomized trials are often used to mitigate contamination but can be less sample-efficient.
• Complier Average Causal Effect (CACE) analysis offers a method to address contamination in individually randomized trials.

Purpose of the Study:

• To compare the statistical efficiency of individual randomized trials using CACE analysis versus cluster randomized trials.
• To evaluate the impact of contamination on sample size and statistical power in different trial designs.

Main Methods:

• Monte Carlo simulations were used to generate trial data with varying levels of control group contamination.
• Simulations assumed a false null hypothesis to assess the risk of Type II errors (failing to reject a false null).
• Statistical power and sample size requirements were evaluated for both individual and cluster randomized designs.

Main Results:

• Increasing contamination levels significantly decrease the power of studies to detect true treatment effects.
• Individual randomized trials using CACE analysis maintain a sample size advantage over cluster randomized trials when contamination is less than 30%.
• This advantage persists even with small cluster sizes and low intra-class correlation coefficients (ICC).

Conclusions:

• Individual randomization with CACE analysis can be statistically more efficient than cluster randomization under specific conditions.
• Precise measurement or estimation of contamination is key to leveraging the efficiency of individual allocation.
• The choice of trial design should consider the expected level of contamination and the need for statistical efficiency.