ABSTRACT
Randomized clinical trials have played a fundamental role in modern medicine. For many treatments, the variation in prognosis among patients exceeds the size of the treatment effect and evaluation of the treatment is very error-prone without large randomized clinical trials. In oncology, developments in biotechnology and advances in understanding tumor biology have provided a better understanding of this variability. In many cases, tumors of a given primary site appear to be heterogeneous with regard to the mutations that cause them, their natural course, and their response to therapy. The heterogeneity in response to therapy is likely to be an important component of the observed small average treatment effects seen in many phase III randomized clinical trials. Small average treatment effects in common diseases for inexpensive and well-tolerated drugs like aspirin can be worthwhile and of public health importance. Small average treatment effects for very expensive drugs with serious adverse effects are much less acceptable for patients and for health care economics. Small average effects imply large NNT values, where NNT denotes the number of patients needed to treat in order to save one life or beneˆt one patient, on average. When the average treatment effect is small, it is likely that most patients do not beneˆt but are at risk for the adverse effects of the drug and may have missed the opportunity to receive a drug that might have helped them. Detecting small treatment effects requires large randomized clinical trials as the sample size is approximately inversely proportional to the square of the treatment effect to be detected.
