A study uses questionnaire-based exposure assessment to measure pesticide exposure and finds no association with cancer (OR = 1.0, 95% CI 0.9–1.1). The questionnaire is later shown to have substantial non-differential misclassification. What is the correct interpretation?
AThe null result is reliable — the wide sample size compensates for measurement error
BThe null result is uninterpretable as evidence of no effect — non-differential misclassification biases toward null, so a true association may have been attenuated away
CThe result confirms no association, because non-differential error makes findings more conservative
DDifferential misclassification must be responsible for the null finding
Non-differential misclassification (error unrelated to disease status) systematically biases risk estimates toward the null — it shrinks true associations toward an OR of 1.0. A null finding under these conditions cannot distinguish 'no true effect' from 'a true effect that crude measurement failed to detect.' This is one of the most important interpretive cautions in environmental epidemiology. Option C reverses the concern: 'more conservative' sounds safe, but the conservative direction here means suppressing real associations.
Question 2 Multiple Choice
Why are biomarkers often preferred over residential or occupational history as exposure measures in environmental epidemiology?
ABiomarkers are cheaper and less invasive than environmental monitoring
BBiomarkers integrate all exposure routes and reflect the internal dose — the amount that actually reached body tissues — rather than an assumed external concentration
CBiomarkers eliminate differential misclassification because they are objective measurements
DBiomarkers measure long-term cumulative exposure more accurately than any other method
The key advantage of biomarkers is that they measure internal dose — what was actually absorbed through all routes (ingestion, inhalation, dermal contact) combined. A residential history assumes people were exposed to what was measured in their environment, which may be false. Option D is wrong: many biomarkers reflect only recent exposure (e.g., urinary metabolites with short half-lives), making them poor for chronic exposures. Option C is also wrong — biomarkers can still be subject to differential misclassification if, for example, disease affects metabolism.
Question 3 True / False
Differential exposure misclassification typically biases risk estimates toward the null, making associations appear smaller than they truly are.
TTrue
FFalse
Answer: False
Only NON-differential misclassification consistently biases toward the null. Differential misclassification — where measurement error differs by disease status (e.g., cases recall exposures differently than controls) — can bias estimates in either direction: it can inflate associations, deflate them, or even reverse them. This unpredictability is precisely what makes differential misclassification more dangerous for causal inference than non-differential misclassification.
Question 4 True / False
A study using proxy exposure measures (e.g., occupation as a surrogate for chemical exposure) that finds no elevated risk cannot reliably distinguish a true null effect from an effect too small to survive the attenuation caused by exposure misclassification.
TTrue
FFalse
Answer: True
This is the fundamental interpretive challenge of non-differential misclassification. The bias toward null means observed null findings are consistent with either truly no effect or with a real effect that the crude proxy measure was too noisy to detect. Exposure validation substudies are designed precisely to address this: by comparing the proxy to a gold-standard measure in a subsample, researchers can estimate the degree of attenuation and correct for it.
Question 5 Short Answer
Explain why non-differential exposure misclassification biases risk estimates toward the null, and what methodological implication this has for interpreting null findings in environmental epidemiology.
Think about your answer, then reveal below.
Model answer: Non-differential misclassification means exposure measurement error is equally bad in cases and controls — it is unrelated to disease status. When you classify truly 'exposed' people as 'unexposed' (and vice versa) at random, you mix up the two groups. Cases contaminate the unexposed category and vice versa, blurring the difference in disease rates between exposure groups. The result is that observed relative risks are pushed toward 1.0 (no association). The implication is that null findings from studies with crude exposure proxies are ambiguous: they may reflect genuine absence of effect, or they may reflect real effects that measurement noise has erased. This is why exposure validation substudies and biomarker confirmation are critical for interpreting negative results.
The attenuation-toward-null phenomenon means that non-differential misclassification always works against finding associations — it is a conservative bias in the sense that it reduces type I errors (false positives) but increases type II errors (false negatives). Studies using better exposure measures systematically find stronger associations than those using cruder proxies, even when studying the same relationship.