Attrition bias occurs in research studies when participants drop out, leading to non-random loss of data that can skew results and undermine the validity of conclusions. This bias is particularly problematic in longitudinal studies where differential dropout rates between groups may significantly affect the study outcomes and interpretations.