Participation (percentageinterviewed among those approached) was low in the three studies, particularly for controls.However, it is unlikely that participation bias would completely explain our findings becauseit is unlikely that genotype frequencies vary by willingness to participate (21).