Fig. 1

Illustration of the different strengths of interaction used in the simulation study. A hazard ratio larger than 1 indicates a higher risk for death under treatment T=1, and a hazard ratio below 1 a higher risk under treatment T=0. For the scenario with no biomarker–treatment interaction, the hazard ratio between the treatment groups is independent of the biomarker value. For the scenario with a quantitative biomarker–treatment interaction, the risk for an event is smaller under T=1 compared to T=0 for all (probable) values of B, but the difference between groups decreases with increasing values of B. For the scenario with a qualitative biomarker–treatment interaction, the risk for an event is lower for T=1 compared to T=0 for small values of B and vice versa for large values of B