Power Analysis for Population-Based Longitudinal Studies Investigating Gene-Environment Interactions in Chronic...

Area of Science:

• Epidemiology
• Biostatistics
• Longitudinal Studies

Background:

• Conventional sample size calculations for population-based longitudinal studies often overestimate statistical power.
• Factors like measurement errors and unmeasured etiological determinants are frequently overlooked.
• Simulation-based approaches offer greater flexibility and accuracy by incorporating these complex factors.

Purpose of the Study:

• To investigate the statistical power of the Canadian Longitudinal Study on Aging (CLSA) to detect effects of environmental and genetic risk factors on chronic diseases.
• To explore design alternatives for enhancing statistical power in population-based longitudinal studies.
• To compare simulation-based sample size calculations with conventional methods.

Main Methods:

• An illness-death model was used for a simulation study.
• The study simulated the Canadian Longitudinal Study on Aging (CLSA) design with 30,000 participants.
• Statistical power was assessed based on varying prevalence of risk exposures, disease commonality, and measurement accuracy.

Main Results:

• Statistical power increased with higher prevalence of risk exposures, more common diseases, and accurate measurement.
• The CLSA's 3-year data collection frequency slightly reduced statistical power compared to continuous monitoring.
• Simulation-based calculations yielded higher minimum detectable hazard ratios than conventional methods, indicating greater accuracy.

Conclusions:

• Simulation-based sample size calculation is a superior method for complex longitudinal studies.
• Increasing measurement frequency and accuracy can enhance statistical power for a given sample size.
• The CLSA demonstrates sufficient power to detect moderate effects under specific conditions, but simulation provides a more realistic assessment.