Sample size and power calculations for open cohort longitudinal cluster randomized trials

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Epidemiology

Background:

• Accurate sample size and power calculations are crucial for longitudinal cluster randomized trials.
• Traditional methods often assume closed cohorts or single measurements per participant.
• Open cohort designs, with variable participant measurements, are common but lack specific sample size formulas.

Purpose of the Study:

• To develop and present sample size and power formulas for longitudinal cluster randomized trials with open cohort sampling structures.
• To unify existing methods for closed cohort and repeated cross-sectional designs.
• To precisely guide the adjustment of participant-level autocorrelation for open cohorts.

Main Methods:

• Derivation of new sample size and power formulas accommodating open cohorts.
• Analysis of designs maintaining constant cluster size but variable participant measurements.
• Integration of participant-level autocorrelation adjustments as suggested by Feldman and McKinlay (1994).

Main Results:

• The presented formulas allow for open cohort sampling structures in sample size and power calculations.
• The degree of "openness" significantly impacts sample size and statistical power.
• The results unify and extend prior work on closed cohort and repeated cross-sectional designs.

Conclusions:

• The new formulas provide a precise method for sample size and power calculations in open cohort longitudinal cluster randomized trials.
• Understanding the impact of "openness" is vital for efficient trial design and resource allocation.
• These findings offer a unified approach applicable to various open cohort sampling schemes and error structures.