Examkin

Question

Difficulty: MediumBias, Confounding, and Validity

Investigators are analyzing data from a national health registry to assess the safety profile of a recently introduced calcitonin gene-related peptide (CGRP) antagonist used for acute migraine treatment. Specifically, they are evaluating the incidence of ischemic stroke over a 5-year period among 15,000 adult patients. Initial unadjusted analysis reveals that patients using the CGRP antagonist have a higher risk of ischemic stroke compared to those using standard NSAID therapy, yielding a relative risk (RR) of 1.7 (p=0.03p = 0.03). Suspecting that underlying patient characteristics might be influencing these results, the analysts separate the cohort into two groups: those with a documented history of migraine with aura and those without aura. Upon re-evaluating the data, the RR for ischemic stroke among patients with migraine with aura is 3.1 (p<0.01p < 0.01). Conversely, among patients without aura, the RR is 1.02 (p=0.88p = 0.88). Which of the following epidemiological phenomena best explains the function of the 'migraine with aura' variable in this analysis?

  1. A
    Confounding
  2. Effect modificationAnswer
  3. C
    Type I error
  4. D
    Selection bias
  5. E
    Observer bias

Answer

Effect modification
The correct answer is effect modification. This epidemiological phenomenon occurs when the association between an exposure (CGRP antagonist) and an outcome (ischemic stroke) is significantly altered by the presence of a third variable (migraine with aura). In this scenario, the risk of stroke is only elevated in patients who have migraines with aura (RR = 3.1), while those without aura experience no increased risk (RR = 1.02). Because the relative risks in the two strata are substantially different from one another, the third variable is acting as an effect modifier rather than a confounder.

Step-by-Step Solution

1
Analyze the initial crude data.
The unadjusted relative risk is 1.7, suggesting an overall increased risk of stroke with the new medication.
Establishing the baseline association is necessary before evaluating the impact of a third variable.
2
Evaluate the data after stratification by the third variable (migraine with aura).
The relative risk is 3.1 in the aura group and 1.02 in the non-aura group.
Stratification allows investigators to see if the third variable is masking a true effect or if the effect fundamentally differs between groups.
3
Compare the stratum-specific relative risks to each other and to the crude relative risk.
The stratum-specific relative risks (3.1 and 1.02) are significantly different from each other. Therefore, the presence of aura modifies the biological effect of the medication on stroke risk.
Significantly divergent relative risks across strata define effect modification, whereas similar relative risks across strata that differ from the crude relative risk define confounding.

Key Concept

Effect Modification vs. Confounding
Rate this question

Topics

Bias, Confounding, and Validity

More questions from this topic