Background: A potential source of bias in epidemiological studies comes from studying people at different stages of disease progression. This can result in biased selection of cases or in errors of measurement of exposures.
Methods: We use stage of disease at the time of diagnosis to evaluate how inclusion of people at different stages in the disease process can influence associations between environmental exposures and colon cancer. Data used were generated from a large case-control study of colon cancer.
Results: For most environmental exposures evaluated, including physical activity, body size, use of aspirin and of non-steroidal anti-inflammatory drugs, and dietary intake of folate and fibre, we did not observe differences in patterns of association by stage of disease at diagnosis. However, for total energy and red meat intake (men only), alcohol consumption, cigarette smoking, and family history of colorectal cancer among first degree relatives, patterns of associations were stronger when colon cancer was detected at an earlier stage of disease progression than when it was detected at a more advanced stage.
Conclusions: Most exposures did not differ by stage of disease, thus selectively excluding cases at different disease stages should not influence associations between these exposures and colon cancer. Associations for other factors, such as alcohol consumption and cigarette smoking, may be biased from asking cases with advanced disease to recall a non-disease-free time period. Associations with family history may also be biased if those with a family history of colorectal cancer are detected at an earlier stage and therefore more likely to participate in epidemiological studies.