since it is, by definition, an express deviation from those routines, taken to gather generalizable knowledge. Instead, research must first be judged by its guiding protocol. An acceptable research protocol should satisfy the criteria of ethically and scientifically sound research, including scientific merit (i.e., avaluable scientific question and valid research methods), risk minimization, and a favorable risk-benefit ratio in which risks to participants are justified by the anticipated benefits of the If the protocol is acceptable, investigators’ research duty of care in implementing it should include three elements. First, the investigator must enroll only eligible subjects. Enrolling ineligible people can endanger those individuals, compromise the scientific integrity of the study and thwart its hoped-for benefits to future populations. Second, the investigator should honor enrollees’ individual needs via whatever flexibility the protocol affords for individualizing care or, if necessary, by removing from the study any subject who is suffering undue harm. Some of the cases discussed above exhibit deficiencies of this sort. In Ken~ke?~ researchers arguably should not have kept a schizophrenic patient in the study when his depression, psychosis, and side-effects grew markedly worse. Similarly, in Burton u. Brooklyn Doctors Hospital,6o a premature infant was kept in a study evaluating the effects of high-dose oxygen on vision even after it had become clear the infant’s eyes were becoming damaged. Third, the research duty of care requires stopping an entire study if its harms to enrolled subjects have become unacceptable in light of anticipated benefits. Many higher-risk studies have Data and Safety Monitoring Boards (DSMBs) to assess research results at various intervals to determine whether a study must be stopped or modified.61 Those that do not should at least have a monitoring plan that can trigger review if accrued results reveal significant problems. As research cases are adjudicated according to these distinctive duties of care, another difference with medical malpractice will emerge. While the latter is fleshed out by expert witnesses from the medical profession, the specifics of research duties must come from different sources. One major source is regulatory?2 The Federal government’s Common Rule governs most human subjects research, and IRBs must ensure each project under their purview follows those rules.63 However, federal regulations do not currently govern all human research,”* and they only cover certain kinds of conduct, such as informed consent, not the entire range of conduct and care appropriate for research. Hence, courts must also include other sources for defining investigators’ duties to their subjects. The broader scientific community may play an important role, such as to define what constitutes sound scientific methodology. Additionally, the bioethics community may help to identify what kinds of risks are acceptable, and which may be too great to impose on human subjects. Courts may introduce their own criteria if, for instance, they find that an IRB has approved an unreasonably dangerous study. Even if only rarely, courts do sometimes supplant otherwise-applicable tort standards with their own j~dgments.6~ In sum, the substantive duty of care that physician-investigators owe their research subjects begins with familiar medical malpractice doctrines, because clinical research has elements of ordinary medical practice. However, conduct that is distinctive to research should be litigated under a research-focused standard of care. Breach of Fiduciary Duty Investigators Are Not Fiduciaries Research likewise diverges markedly from ordinary medicine in the applicability of fiduciary doctrines. To be sure, the concept of “fiduciary” is not entirely clear,”6 and the standard physician-patient relationship is not always deemed fiduciary in the most classic ~ense.~7 Courts have been reluctant, for instance, to permit breach of fiduciary duty as a separate cause of action against a physician.68 Nevertheless, courts have felt free to call the physician-patient relationship fiduciary or at least a relationship of trust and confidence.@ The physician has knowledge and skills whose application require discretion, and for which the patient has significant need.7O The physician voluntarily undertakes to act primarily for the benefit of the ~atient.7~ Classic fiduciaries loyalty duties apply, such as to refrain from exploiting the latter’s vulnerability for his own gain and to avoid conflicts of interest wherever p0ssible.7~ In sharp contrast, the very nature of research precludes a fiduciary relationship between investigators and subjects. By definition the researcher’s goal is not the betterment of any particular participant. It is the successful completion of the research, in hope of helping future patients. This is not a situation in which a standard fidelity obligation requires a transient compromise, such as a decision whether to violate patient confidentiality to warn a third party that the patient poses a danger. Rather, a completely different allegiance permeates the relationship. The investigator’s entire purpose, his number one loyalty, is already pegged on something other than the patient. It is to the protocol. To conclude that the investigator-subject relationship is not and cannot be fiduciary is sharply at odds with