Conclusion: Neither approach was always optimal. However, if interest lies in detecting changes in overall outcome incidence (intercept), Calibration regression may be superior. If interest lies in detecting changes in other covariates (slope), Direct Approach may be superior.