Effect of Comparison Group on Inference about Effect Modification by Demographic Factors in Cohort Risk Regression