Can we really derive etiology from human-based studies?

This paper discusses the need for and limitations of human-based stud ies of diet and cancer. Four problems of such studies remain especially acute: assessment and monitoring of dietary practice; assessment of com pliance in diet change studies; validation of dietary questionnaires; and the use of intermediate biological end points. These limit human-based studies' utilization of recent advances in the cellular and molecular biology of cancer. Those researchers who emphasize human-based studies cannot proceed without acknowledgment of and interface with these advances, which promise more precise definition of both exposure and disease and may facilitate evaluation of associations of dietary practice and carcinogenesis in the context of human-based studies. To take advantage of these advances will require critical attention to the causal mechanisms by which intermediate biological end points link dietary exposures with cancer risk.