The risk of bias: quality or validity?