Using the smallest worthwhile difference to interpret clinical evidence