Bias in uncontrolled therapeutic trials in rheumatology due to selection of populations with extreme characteristics.