Epidemiological research involving people has inherent risks. The Council for International Organization of Medical Sciences provides guidance on the ethical principles of epidemiological research, including respect for people, non-maleficence, beneficience, and justice. In this article we discuss the challenges of applying each principle to population research on domestic violence, an issue in which poorly designed research could put women in violent relationships at substantial risk. Main concerns include ensuring safety of respondents in a context in which many live with their abuser, protecting confidentiality when breaches could provoke an attack, and ensuring the interview process is affirming and does not cause distress. The inherent risks entailed in research can only be justified if the interview is used to provide information on available services and is a source of immediate referral when necessary, if high-quality data are obtained, and if findings are used to raise awareness of, and improve services for, women who experience domestic violence.