A prospective cohort study was conducted to assess the effects of oral contraceptive (OC) use on the incidence of breast cancer. A crude analysis of the study suggests an association between OC use and breast cancer (relative risk [RR] = 1.40, 95% confidence interval [95%CI]: 1.01, 1.96). Further analysis shows that in women with a family history of breast cancer, OC use increases the risk of breast cancer (RR = 3.33, 95%CI: 2.51, 4.41). In women without a family history of breast cancer, a significantly smaller effect is observed (RR = 1.27, 95%CI: 1.01, 1.60). The results of this study are best explained by which of the following?
Show Explanatory Sources
Effect modification (interaction) results when an extraneous variable (modifier) changes the direction or strength of an association between a risk factor and a disease. A modifier is associated with the disease but not the risk factor. Effect modification can sometimes be confused with confounding, the bias that results when the exposure-disease relationship is obscured by the effect of an extraneous factor (confounder) associated with both exposure and disease.
Stratified analysis by the extraneous variable can help distinguish whether that variable is a confounder or an effect modifier, as follows:
Confounder (left panel): The measures of association (eg, relative risk [RR] of bladder cancer among alcohol users) calculated in each of the stratified groups (eg, smokers and nonsmokers) are similar to one another (eg, nonstatistically significant RR = 0.95 [95%CI: 0.51, 1.76] and RR = 1.03 [95%CI: 0.34, 3.13]), but they differ from the measure of association calculated in the crude analysis (eg, statistically significant RR = 1.81 [95%CI: 1.06, 3.10]); stratification can remove the effects of the confounder (eg, smoking).
Effect modifier (right panel): The measures of association calculated in each of the strata are significantly different (in strength or direction) from one another; stratification can make the effects of the modifier more apparent.
In this example, family history of breast cancer (BC) modifies the association between oral contraceptive (OC) use and BC. Crude analysis reveals a significant association (ie, RR = 1.40 [95%CI: 1.01, 1.96]) between OC use and BC. However, once the association is stratified by family history of BC, the strength of the association changes (ie, with BC family history, RR = 3.33 [95%CI: 2.51, 4.41]; with no BC family history, RR = 1.27 [95%CI: 1.01, 1.60]), reflecting a significantly higher risk for BC among those who take OCs if they also have a positive BC family history. In other words, positive BC family history acts as an effect modifier.
(Choice A) Given that the measures of association are significantly different from one another at the strata level (RR = 3.33 [95%CI: 2.51, 4.41] vs RR = 1.27 [95%CI: 1.01, 1.60]), BC family history is not a confounder.
(Choice C) Latency period is the time required for a given exposure to have a measurable effect on the outcome. This study provided no information on how long oral OCs must be used to have an effect on BC risk in susceptible patients.
(Choices D and E) Observer bias and selection bias are errors in study design involving misclassification of data and inappropriate selection of participants, respectively.
Educational objective:
Effect modification results when an external variable positively or negatively impacts the effect of a risk factor on the disease of interest. Stratified analysis helps determine whether a variable is a confounder or an effect modifier.