Bias is less likely to invalidate a cohort study used to assess the risk of disease due to an exposure because: The presence or absence of exposure is determined prior to disease occurrence Disease status is always known at the start of the study Incident cases of the disease are distinguished from prevalent cases in the analyses Analyses are based on a comparison of the proportion of prevalent cases of disease in each group, rather than simply the absolute number of cases in each group The sample sizes in cohort studies are small .