How large a role should expert opinion have in the use of evidence-based medicine?