Learning from FAME: the need for sham controls in trials of stable coronary disease.