A case-control study investigates an association between herbicide exposure and non-Hodgkin lymphoma. Controls are matched by age, gender, and race; exposure is determined by interview. The odds ratio is 3.2 (CI 1.4-5.4). Which factor is most likely to affect the validity of this study?

A case-control study investigates an association between herbicide exposure and non-Hodgkin lymphoma. Controls are matched by age, gender, and race; exposure is determined by interview. The odds ratio is 3.2 (CI 1.4-5.4). Which factor is most likely to affect the validity of this study?

• A. Confounding
• B. Biased measurement of exposure
• C. Random error
• D. Selection bias

Answer: (B)

Information bias from how exposure is measured drives the validity issue here. In this case-control study, exposure to herbicides is determined by interview. Cases (those with non-Hodgkin lymphoma) may be more likely to remember or report past herbicide exposure than controls, or may report it with greater certainty after disease experience. This differential misclassification of exposure biases the odds ratio upward, making the association appear stronger than it truly is. While matching helps reduce confounding by age, gender, and race, it doesn’t fix biases from data collection. Random error would cause imprecision and isn’t the main threat given the reported confidence interval; selection bias could be a concern if cases and controls differed in how they were chosen, but the interview-based exposure assessment is the clearer, more plausible source of systematic error here.

Information bias from how exposure is measured drives the validity issue here. In this case-control study, exposure to herbicides is determined by interview. Cases (those with non-Hodgkin lymphoma) may be more likely to remember or report past herbicide exposure than controls, or may report it with greater certainty after disease experience. This differential misclassification of exposure biases the odds ratio upward, making the association appear stronger than it truly is.

While matching helps reduce confounding by age, gender, and race, it doesn’t fix biases from data collection. Random error would cause imprecision and isn’t the main threat given the reported confidence interval; selection bias could be a concern if cases and controls differed in how they were chosen, but the interview-based exposure assessment is the clearer, more plausible source of systematic error here.