In Oncology First-in-Human (FIH) single drug dose finding trials, the model-based approach Continuous Reassessed Method (CRM) is often used due to its efficiency and flexibility. A combination therapy is a therapy that uses more than one medication. In a combination dose finding trial, CRM Bayesian model allows an easy incorporation of the prior single drug dose finding trial information in the combination dose finding model, but the potential drug-drug interaction may also need to be considered in the model. The pre-information of interaction effect is usually quite limited prior to a phase I trial. Therefore, we aim to investigate whether the interaction effect can be ignored in the statistical model, under a variety of interaction and dose-response scenarios. We find that the interaction effect can be ignored in most of the realistic settings. But when the interaction effect is extremely high, a complicated model with interaction parameter works much better. In the presentation, we will also advise the run-in approach, which pre-specify the dose escalation path. Before observing any dose limiting toxicity (DLT), this rule-based run-in path is suggested to be used.