Oral contraceptives and rheumatoid arthritis: Bias in observational studies of cause-effect relationships