19 November 2002 Dear Editor, REPLY We appreciate the thoughtful letter from Dr Cole suggesting that the best way to treat patients is as participants in ran­domized controlled trials (RCT). Researchers in rapidly advancing fields, such as haematology−oncology plan new trials based on the results of prior trials, often before those prior trial results are published. By the time trial results are published, new questions have surfaced, so that new trials must be under way to refine or redefine therapies that, while proven, are almost never the final answer. The stated principle of evidence-based medicine is that patients should be treated according to the best available evidence from systematic studies in human subjects. High quality RCT provide, after completion, the highest level of evidence. Under many circumstances, use of therapies that have been shown to be effective is clearly the best approach. However, when the best therapy is still not good enough, and new therapies are under consideration, the patient who participates in an RCT has at least an even chance of getting the best therapy. Participants will get either the best currently known therapy (control patients) or a new therapy that may − or may not − be even better (intervention patients). Under ordinary circumstances outside of trials, even if the new therapy is available, the doctor making the choice of therapy in the absence of evidence might happen to choose wrongly between these two options. In this situation it is possible that all of his or her patients will get the worse therapy. Certainly, without the patients who participate in RCT, there would be no high quality evidence on which to base therapeutic decisions about any condition. In fact, if there is equipoise, that is if uncertainty exists regarding the best therapy, then the most ethical way to treat a patient is on a trial. In addition, there is good observational evidence that patients who join trials fare better than those who do not (Braunholtz DA, Edwards SJL, Lilford RJ. Are randomized clinical trials good for us (in the short term)? Evidence for a ‘trial effect’. J. Clin. Epidemiol. 2001; 54: 217−24). Unfortunately, most patients do not have access to RCT, even when uncertainty about treatments exists. Our patients would be best served by having all of their tests and treatments based on high quality evidence. When such evidence is not available, or is outdated because of rapid changes in a field, all patients should be enrolled in large effectiveness trials that address relevant clinical questions in the ‘real’ world of clinical practice.