From: Analysis of multicenter clinical trials with very low event rates
Number of centers | RE | MH | GEE | GEE-small sample correction |
---|---|---|---|---|
Small number of centers (e.g., 5) | Recommended | X Low type I error and power | X Inflated type I error | X Nominal type I error, but low power Convergence issues |
Moderate or large number of centers (e.g., 50 to 100) | Recommended (slightly biased odds ratio with a small total sample size (e.g., 200)) | X Low type I error and power Convergence issues when the sample size is small (e.g., 200) | △ Slightly inflated type I error in some scenarios | Recommended |