The introduction of pay for performance (P4P) in healthcare raises a problem for economists: is quality of care soluble in quantities? The quantified standards of quality used in P4P programs come from the results of Evidence based medicine (EBM) and from its main method of research: random controlled trials. We develop in this paper two major epistemological criticisms of this use of EBM. On the one hand, the implicit epistemology of statistical objects used by this method is problematic: the disease is close to the behavior of 2 the patient and the methodology used in randomized clinical trials is not neutral. On the other hand, it is also questionable to consider care service as product.