Related Experiment Videos
Toward a clearer definition of confounding
1National Institute of Environmental Health Sciences, Research Triangle Park, NC 27709.
American Journal of Epidemiology
|January 1, 1993
Summary
Adjusting for factors caused by exposure can bias epidemiologic study results. This occurs even when these factors are not on the direct causal pathway between exposure and disease, leading to misleading findings.
Area of Science:
- Epidemiology
- Biostatistics
Background:
- Epidemiologic studies aim to estimate exposure effects on disease.
- Standard practice advises against adjusting for intermediate factors on the causal pathway.
- Factors influenced by exposure but not on the direct causal pathway are often adjusted for.
Purpose of the Study:
- To illustrate how adjusting for factors caused by exposure can introduce bias in etiologic studies.
- To demonstrate that bias can occur even when the adjusted factor is not an intermediate variable.
Main Methods:
- The study uses illustrative examples to demonstrate the phenomenon.
- It analyzes potential biases in prospective and retrospective etiologic studies.
Main Results:
- Adjusting for factors that are caused in part by the exposure and correlated with the outcome can lead to biased effect estimates.
- This bias is not limited to intermediate variables on the causal pathway.
Conclusions:
- Researchers must be cautious when deciding which variables to adjust for in epidemiologic analyses.
- Failure to account for the causal relationships between exposure, mediators, and confounders can yield misleading results.