In the process of evaluating a new medical treatment, it's crucial to base conclusions on solid evidence. However, when direct experimentation is not feasible, what should researchers primarily depend on?