Effect modification (also termed statistical interaction) occurs when an external variable positively or negatively impacts the observed effect of a risk factor on disease factor (e.g., incidence, progression). When stratified analysis is performed based upon this external variable, there will be a significant difference in risk between the stratified groups. Unlike confounding, effect modification is a biological phenomenon in which the exposure has a different impact in different circumstances.

In this vignette, the risk of lung cancer is found to be 10-folds higher in individuals with chronic asbestos exposure. However, when the participants were stratified into two groups based on their smoking status (third variable), the results were modified. The arm with non-smokers showed an increase in the risk of lung cancer by only 3-folds. Most studies suggest that smoking increases the risk of lung cancer by about 20-folds. However, smokers with chronic asbestos exposure had about a 64-folds increased risk of lung cancer. In other words, the effects of smoking and asbestos exposure were more than multiplicative (3*20=60). This suggests synergism or interaction, meaning that the effect of smoking is somehow magnified in individuals who have also been exposed to asbestos. Multivariable methods can also be used to assess for effect modification.

Option Analysis

❌ Measurement bias:
Incorrect: Measurement bias occurs when information collected for use as a study variable (e.g., exposure, disease outcome) is inaccurate. However, this vignette does not describe any issues that could affect the classification process.

❌ Confounding:
Incorrect: Confounding is when a third variable that has not been factored into the study (the confounder) has an association with the dependent variable (outcome). It can suggest a causal association between the outcome and independent variable, when in fact, there is none. In this case, smoking synergized (modified) the effect of asbestos exposure, which is more suggestive of effect modification.

✅ Effect modification:
Correct: See Main Explanation.

❌ Observer bias:
Incorrect: Observer bias occurs when measurement of a variable or classification is influenced by the researcher’s knowledge of which group a participant is in. It is minimized by performing a blinded experiment. However, bias in the classification of participants is not discussed in the above vignette.

❌ Recall bias:
Incorrect: Recall bias results from inaccurate recollection of past exposure by individuals participating in the study. Recall bias typically occurs in retrospective studies (e.g. case-control study).

Study Design And Data Interpretation

Study Mode

Test Mode