Influence of selection bias on the test decision. A simulation study.

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Medical Research Integrity

Background:

• Selection bias in clinical trials occurs when patients are selectively assigned to treatment groups.
• Even with allocation concealment in randomized trials, predictable assignments can introduce selection bias.

Purpose of the Study:

• To investigate the impact of selection bias on type I error rates in unmasked randomized trials using permuted block randomization.
• To incorporate practical assumptions like patient characteristic misclassification for a clinically relevant error estimate.
• To compare investigator biasing strategies and consider patient availability to establish an upper bound for type I error.

Main Methods:

• Simulations were conducted using SAS to evaluate selection bias effects under various conditions.
• Key factors examined included different block sizes, selection effects, biasing strategies, and patient classification success rates.
• The study simulated practical scenarios to assess the influence on type I error rates.

Main Results:

• Type I error rates frequently exceeded the 5% significance level, reaching up to 21%.
• While cautious biasing strategies and misclassification could reduce, they did not eliminate selection bias.
• The number of screened patients was approximately three times the required number for the trial.

Conclusions:

• Selection bias significantly influences test decisions in unmasked randomized trials with permuted block randomization and allocation concealment.
• The impact of selection bias should not be overlooked when designing and reporting clinical trials.
• Incorporating selection bias assessment is crucial for maintaining the integrity of clinical trial evaluations.