Cross-sectional versus longitudinal investigations of the diet-cancer relation.

Within a prospective cohort study on diet and cancer, information was collected on cancer prevalence and baseline meat consumption. A nested case-control study on meat and cancer was conducted with 656 prevalent colorectal cases, 1,894 breast cancer cases, and 4,701 controls. When analyzed cross-sectionally, prevalence odds ratios for eating meat rarely versus regularly were 2.08 for female colorectal and 1.75 for breast cancer. In the longitudinal analysis, cases who started consuming meat rarely after diagnosis were excluded, resulting in odds ratios of 0.51 for female colorectal and 1.17 for breast cancer. These opposite findings highlight the problem of cross-sectional designs.