"Nevertheless, it is important to mention that according to an influential 2001 meta-analysis comparing placebo-treatment arms with no treatment, placebos make no clinical difference.[20] However, the methodological difficulties of combining diverse conditions in a single meta-analysis are serious, and the authors’ updated meta-analysis includes more nuanced conclusions.[21] Furthermore, the meta-analyses implicitly assume the biological reductionist definition of placebo, ignoring the psychosocial context of sickness and health. The subjects in those studies were entered into a research programme and received attention and care. Therefore, in the way that the studies of the meta-analysis were actually conducted, there was plenty of room for the placebo effect without the placebo, which would spuriously suggest that the placebo was no more effective than no treatment, though in fact both had the potential to produce a placebo effect."
from "The moral case for the clinical placebo"
http://jme.bmj.com/content/40/4/219.full