A re-randomisation design for clinical trials.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Medical Research Methodology

Background:

• Clinical trial recruitment frequently faces challenges, leading to underpowered studies and suboptimal patient care.
• Many conditions necessitate repeated treatments or multiple attempts for success, posing recruitment hurdles for traditional trial designs.

Purpose of the Study:

• To introduce and discuss a re-randomisation design suitable for clinical scenarios requiring multiple treatments per patient.
• To outline the conditions under which this design can effectively address recruitment issues and improve trial efficiency.

Main Methods:

• Describes a re-randomisation design where each patient can be independently randomized on multiple occasions.
• Specifies conditions for unbiased estimates: completion of follow-up before re-randomisation, independent randomisations, and constant treatment effect.
• Emphasizes the need for analysis to account for within-patient correlation.

Main Results:

• The re-randomisation design yields asymptotically unbiased treatment effect estimates and correct type I error rates when conditions are met.
• This design typically offers higher statistical power compared to parallel group trials with equivalent numbers of observations.
• Successful implementation requires careful adherence to procedural and analytical requirements.

Conclusions:

• The re-randomisation design offers a viable solution to enhance clinical trial recruitment rates.
• It maintains unbiased treatment effect estimation and correct type I error rates when appropriately applied.
• This innovative design can significantly increase statistical power in various clinical research settings.