Ethical duties cannot be outsourced
Leonard H. Glantz of the Boston University School of Public Health writes in Nature's Correspondence page (Nature 449, 139; 2007):
Your News Feature 'Trial and error', on the problems with research ethics committees designed to establish whether a proposed experiment is ethically sound (Nature 448, 530–532; 2007), presents avoidance of liability and the desire to retain power as the main reasons why institutions favour local control over centralized review. But institutions are ethically, not just legally, responsible for what happens to human subjects under their care.
Research is a suspect activity designed to advance knowledge, not benefit individuals. This does not denigrate its importance but rather reminds us why experiments involving humans are regulated differently from other kinds of research, and more heavily.
If a central institutional review board says it's fine to enrol patients into a project, this does not mean that the institution involved can ignore its obligation to protect the rights and welfare of human subjects in its facility. Any institution that outsources its ethical responsibilities towards subjects should not be allowed to conduct research on human beings.