Epidemiological studies raise somewhat different ethical issues than those that usually confront IRBs in their review of biomedical research. Although epidemiologic research seldom risks direct harm to subjects, it may still wrong them if it invades their interests (such as privacy) or treats them merely as means. Review by an IRB is thus justified if it improves the benefit-risk ratio; even more important, in most circumstances, informed consent should be obtained in advance from subjects, to promote subjects' autonomy, to improve the quality of research, to regularize the relationship of investigators and potential subjects, and to protect subjects' privacy. Nonetheless, in certain circumstances-illustrated here by considering three hypothetical studies-alternatives to the usual consent rules are acceptable.