Assessing heterogeneity of treatment effect in a clinical trial with the proportional interactions model