Survivor bias

In this case-control study, cases are drawn from a hospital population. Out of all those with disease in the hospital (16 blue circles), 4 cases agree to be part of the study (shaded square). These 4 may represent a biased sample in that those with severe disease or those who die before being approached for participation are not represented. Out of all those with no disease in the general population, 4 controls agree to be part of the study. By comparing these 4 cases and 4 controls, any genes found to be selectively enriched (or depleted) in the cases may be linked to occurrence of the disease but may also be linked to severity or survival.
In a prospective study, blood is taken at baseline so that even if someone develops disease and dies before recruitment into a case-control study could happen they are still represented in the cohort study, ie, the cases are more representative.