'Race' and 'ethnicity' are increasingly being used as variables in health research. However, studies have been mainly descriptive and have not been used to develop and evaluate strategies to improve health care. In part this reflects the poor analytical standards. The status of the concepts of 'race' and 'ethnicity' as research tools are rarely considered and there is poor consistency in terminology. This paper gives an overview of the research literature and raises questions about the validity of 'race' and 'ethnicity' as epidemiological variables. The tendency to collect routine ethnic data and include ethnic variables in an ad hoc and uncritical way in the United Kingdom and other countries may help transform minorities into mere statistical categories and produce data and findings which reinforce stereotypes. Multiculturalist ethnic health explanations also tend to displace more material explanations of health outcomes. It is concluded that more thought and care is needed before data are routinely categorized by ethnicity, or race or ethnicity are included as variables in research.