Selection of subjects for hospital-based epidemiologic studies based on outward manifestations of disease

Selection of controls with the same outward manifestations of disease as th e case group has been proposed as a means of avoiding selection bias in hos pital-based case-control studies. The same strategy, however, can lead to s election bias in registry-based case-control studies that use control disea ses with similar manifestations whose diagnoses might have been associated with the exposure. Matching exposed and unexposed subjects by outward manif estation of disease can be used in cohort and cross-sectional studies aimin g at decreasing selection bias. This strategy in these study designs may le ad to over-matching, but this will not bias the relative-risk estimates. Ef ficiency considerations in applying this strategy require further investiga tion.