Step 1: Start from how risk of association is quantified in each design. A $2\times2$ table has cells $a,b,c,d$, and the odds ratio equals \[OR = \frac{a \times d}{b \times c}.\] The diagonal multiplication gives the name "cross product ratio."
Step 2: Case-control studies move backward from outcome to exposure, so true incidence is unavailable and the odds ratio becomes the natural measure of association.
Step 3: By elimination, cohort uses relative risk, cross sectional uses prevalence ratios, and an RCT yields relative risk or risk difference. Only the case-control design is defined by the cross product ratio. \[\boxed{\text{Case control}}\]