Table 3.

Select characteristics of well-established designs for gene–environment interaction.

characteristic	study design
family-based	case–control	cohort
potential for population stratification bias	nil if appropriately analysed	varies; able to be minimized via good design, genomic control	varies but generally less than retrospective case–control study; able to be minimized via good design, genomic control
potential for recall bias	moderate to high	moderate to high	nil
potential for survivor bias	moderate to high	moderate to high	nil to moderate if DNA is not obtained on all cases and controls at base-line
ability to use plasma phenotypes in cases	no	no	yes
required sample sizes achievable?	common disease: yes	common disease: yes	common disease: yes with adequate follow up
rare disease: yes	rare disease: yes	rare disease: no, unless multiple studies are pooled