The specificity of medical facts: the case of diabetology