How many strata in an RCT? A flexible approach

Area of Science:

• Biostatistics
• Clinical Trial Design
• Epidemiological Research

Background:

• Prognostic factors are crucial in cancer trial design and analysis.
• Overstratification must be avoided by limiting the number of strata.
• Existing guidance can be improved with explicit principles and adaptability.

Purpose of the Study:

• To propose a new method for determining the number of strata in clinical trials.
• To provide clinicians with a practical tool for trial design.
• To balance the inclusion of prognostic factors with the risk of overstratification.

Main Methods:

• Utilizes a Poisson distribution to model observations per stratum.
• Calculates the number of strata based on sample size, minimum stratum size, and acceptable risk.
• Programmable in Excel for ease of use.

Main Results:

• For a 250-patient trial, <1% risk of <10 patients/stratum suggests no more than 13 strata.
• For survival analysis with 170 deaths, no more than 9 strata are prudent.
• This can often be achieved using just two prognostic variables.

Conclusions:

• The proposed method is flexible and principle-based.
• Applicable to both clinical trial design and epidemiological studies.
• A valuable tool for researchers and clinicians.