Randomizing a clinical trial in neuro-degenerative disease

Area of Science:

• Clinical Trials
• Biostatistics
• Neurology

Background:

• Parkinson's disease clinical trials require robust randomization methods.
• Existing methods may not fully account for patient covariate distributions.
• A sample of 144 Parkinson's patients provided response and prognostic factor data.

Purpose of the Study:

• To evaluate and compare different randomization rules for a sequential, two-treatment, two-site clinical trial.
• To develop a simulation model using empirical patient data for trial analysis.
• To recommend optimal allocation rules for Parkinson's disease trials.

Main Methods:

• Utilized a sample of 144 Parkinson's patients for covariate distribution modeling.
• Employed a two-stage algorithm for simulating covariate distributions.
• Performed simulations to compare six different allocation rules based on imbalance and bias.
• Generated empirical marginal distributions from a correlated multivariate normal distribution.

Main Results:

• Quantified the loss due to imbalance and potential bias for each allocation rule.
• Identified specific randomization rules that perform better under simulated conditions.
• Demonstrated the utility of using empirical patient data in simulation models.

Conclusions:

• The study recommends two distinct allocation rules, one for each trial site.
• Recommendations are contingent upon the target number of patients to be enrolled.
• The findings offer practical guidance for designing Parkinson's disease clinical trials.