A cohort study is conducted to investigate the association between regular alcohol consumption and the incidence of oral cancer. The investigators analyze the initial data and calculate a crude relative risk (RR) of 3.2 (95% CI, 2.4-4.1). However, the researchers suspect that concurrent tobacco smoking might be influencing the results. To further investigate this, they perform a stratified analysis based on smoking status. The RR for oral cancer among smokers is calculated as 1.2 (95% CI, 0.9-1.5), and the RR among non-smokers is calculated as 1.1 (95% CI, 0.8-1.4). Which of the following concepts is best demonstrated by the difference between the crude and stratified relative risks in this study?
- ConfoundingAnswer
- BEffect modification
- CSelection bias
- DRecall bias
- EObserver bias
Answer
Confounding
The scenario describes confounding. Confounding is present when a third variable is associated with the exposure, affects the outcome, and is not on the causal pathway. It is typically identified when the crude measure of association differs significantly from the stratum-specific measures, while the stratum-specific measures are similar to each other. In this case, the crude RR (3.2) is much higher than the stratified RRs (1.2 and 1.1), which are nearly identical, confirming that smoking was confounding the relationship between alcohol and oral cancer.
Step-by-Step Solution
Key Concept
Stratified analysis is used to differentiate between confounding (where stratum-specific relative risks are similar to each other but different from the crude relative risk) and effect modification (where stratum-specific relative risks differ significantly from each other).
Estimated Time:1m 0s