Question of the Day
Sunday, November 8, 2026
A cross-sectional study finds that coffee consumption is associated with a higher risk of pancreatic cancer (odds ratio 2.1). When the analysis is stratified by cigarette smoking, the odds ratio for coffee within both smokers and nonsmokers falls to approximately 1.0. Smoking is strongly associated with both coffee consumption and pancreatic cancer.
Which of the following best explains the change in the odds ratio after stratification?
Answer: A. Confounding by smoking
Smoking meets all criteria for a confounder: it is associated with the exposure (coffee drinkers smoke more), is an independent risk factor for the outcome, and is not on the causal pathway between coffee and cancer. When the crude association (OR 2.1) collapses to about 1.0 uniformly across strata, the apparent effect was entirely due to confounding by smoking. Effect modification would instead produce different, non-uniform odds ratios across strata (e.g., strong in smokers, absent in nonsmokers), which is not seen here. Information bias and random error would not be resolved simply by stratifying on a third variable, and reverse causation would require the cancer to cause coffee drinking. Uniform collapse of a stratified estimate toward the null is the hallmark of confounding.
Just show me the answer
Answer: A. Confounding by smoking
Smoking meets all criteria for a confounder: it is associated with the exposure (coffee drinkers smoke more), is an independent risk factor for the outcome, and is not on the causal pathway between coffee and cancer. When the crude association (OR 2.1) collapses to about 1.0 uniformly across strata, the apparent effect was entirely due to confounding by smoking. Effect modification would instead produce different, non-uniform odds ratios across strata (e.g., strong in smokers, absent in nonsmokers), which is not seen here. Information bias and random error would not be resolved simply by stratifying on a third variable, and reverse causation would require the cancer to cause coffee drinking. Uniform collapse of a stratified estimate toward the null is the hallmark of confounding.
Thousands more board questions like this. Free, and you earn points for every one.
Start free on ResiLearn