Which method is used to control confounding in the design phase of a cohort study?

• A. Randomization is the only method.
• B. Restriction, stratification, matching, and, if possible, randomization.
• C. Increasing sample size alone.
• D. Ignoring confounders.

Answer: (B)

Controlling confounding in the design phase means implementing study procedures that prevent or minimize differences in confounders between the exposed and unexposed groups before outcomes are measured. The best approach uses a combination of restriction, stratification, matching, and, when feasible, randomization.

Restriction limits participants to specific values of a potential confounder (for example, only non-smokers), which removes the confounding influence entirely within that restricted group. Matching pairs or groups subjects who share similar confounding characteristics with their comparison group, making the groups more comparable. Stratification plans for analyzing or sampling within strata of the confounder, so comparisons are made within more homogeneous subgroups. If exposure can be randomized, random assignment balances both known and unknown confounders across groups, effectively eliminating confounding bias by design.

Increasing sample size alone improves precision but does not remove bias from confounding. Ignoring confounders is unacceptable because it leaves bias in the estimated association.