Cluster randomized trials
In recent years, there has been a growing interest in generating dependable evidence about the effectiveness of health policies, programs and practices, using randomized designs. In some studies, randomization at the individual (patient) level may not be feasible because the intervention is designed to be implemented at the group level or because the hypothesized mechanism of action of the intervention operates at the group level. In health services implementation research  for example, the intervention may be administered to the health professional or may involve changes to the health care organization; and in trials of infectious disease interventions, a vaccine may be administered at the individual level but its effects observed among those in the wider community as a consequence of herd immunity. Randomization at the individual level may also be undesirable for methodological reasons such as the need to avoid contamination (for example, in trials of behavioural interventions) when individuals in close proximity are randomized to competing interventions. Finally, randomizing individuals may complicate the trial organization and implementation, for example, in developing nation settings where special equipment or personnel are required or permission from political authorities must be obtained to conduct the trial. The cluster randomized design  has thus become an increasingly important methodological tool in health and health services research: in cluster randomized trials (also known as group randomized or place randomized trials), intact social units or clusters of individuals, such as medical practices, communities, schools or villages are randomized to intervention or control conditions. The interventions may be delivered to the entire randomized group as a unit, or to individuals within each group, but all members of a group receive the same intervention; outcomes are then observed on individual cluster members (or subsamples of members) to evaluate the effect of the experimental intervention. Note that, although outcomes are observed on individuals, they may be aggregated at the cluster-level, for example, percentage of X-ray requests by physicians.
Ethical challenges in cluster randomized trials
The substantial methodological differences between cluster randomized trials and conventional randomized trials pose serious challenges to the current conceptual framework for research ethics. Contemporary research ethics is largely structured around the protection of the autonomy and welfare interests of individual research subjects. Ethical principles governing the conduct of clinical research are laid out in the Belmont Report . The ethical principle of respect for persons means that choices of autonomous individuals ought to be taken seriously and that persons who cannot responsibly choose for themselves are entitled to protection. This principle is the source of the moral rules requiring informed consent from research subjects and protection of confidential health information. The ethical principle of beneficence means that researchers have an obligation to protect subjects from avoidable harm and, where possible, to promote their welfare interests. It is the source of a variety of moral rules that guide the ethical analysis of study benefits and harms . The ethical principle of justice means that study subjects ought to be treated fairly. It grounds requirements that the vulnerable, such as children or incapable adults, not be included as a population of mere convenience. Recently, a novel ethical principle of respect for communities has been proposed . It requires that investigators have an obligation to respect communal values, protect and empower social institutions, and, where applicable, abide by the decisions of legitimate communal authorities.
The ethical implications of randomizing groups rather than individuals have as yet not been thoroughly explored [6–11]. This is illustrated by the ethical issues arising from three different cluster randomized trials: In the Community Intervention Trial for Smoking Cessation (COMMIT) , twenty-two cities in the USA and Canada were randomized to either a community-level antismoking intervention delivered through mass media, health care professionals and worksites, or to a no-intervention control. Baseline data were collected from a random sample of citizens using a telephone survey, and a sample of smokers identified and followed with annual telephone surveys to ascertain smoking cessation status. The investigators interacted with the community by forming a board of community representatives. Individual respondents were not aware that they were involved in a trial, although they provided verbal consent to complete the telephone surveys. Questions raised by this study include: Is consent required from the communities involved?  If so, from whom? Is the municipal government empowered to make these decisions?  What criteria ought community decision-makers use?  When the intervention is targeted at the entire community, it may be impossible to obtain individual informed consent from all citizens – is this ethically legitimate, or does an inability to consent to or opt out of a community-level intervention violate the rights of individual citizens? 
Analogous questions arise in quality improvement interventions where the intervention may be targeted at health professionals but affect patient care. For example, in a cluster randomized trial to improve adherence to guidelines for hypertension drug prescribing, physicians and nurse practitioners were randomized to receive either general guideline education or education plus patient-specific reminders about hypertension . Outcomes included overall compliance with guidelines and adequacy of patients' blood pressure control as ascertained by review of patients' medical records. The requirement for informed consent for both practitioners and individual subjects was waived by the Research Ethics Board (REB). Questions raised by this study include: Who are the research subjects: health professionals, patients, or both? Should consent be required from the participating health professionals  or is there a professional obligation for health professionals to participate in quality improvement research that obviates the need for their informed consent? If the study is using routinely-collected data with identifying information removed, is patient consent required? How are we to understand this study with respect to the analysis of benefits and harms? 
A third, very different study randomized villages in Nepal to provide nutritional supplements to women of child-bearing age . Villages were randomized to one of four study arms: vitamin A supplements, β-carotene supplements, both supplements, or placebo. The outcome of interest was mortality associated with pregnancy and childbirth. Community leaders agreed to randomization of communities, while individual women gave verbal consent to receive the supplements and provide data. A sample of women who became pregnant underwent further investigations, including blood sampling. Mortality and other variables were collected prospectively by study workers. Further information regarding fatalities was obtained from interviews with the families of any subjects who died. While this study shares many of the issues raised in the first example, additional questions include: Do investigators bear any special obligations to subjects because of the developing nation setting ? Is there an obligation to offer ancillary benefits to the control arm or to all study arms (e.g., additional health care unrelated to the study question) ?
Inadequacy of Current Research Ethics Guidelines
Given the uncertainty in the literature as to how to address the ethical problems presented by cluster randomized trials, it is not surprising that current research ethics guidelines do not address these issues. Relevant international and national guidelines include the World Medical Association Declaration of Helsinki , the Council of International Organizations of Medical Sciences (CIOMS) International Ethical Guidelines for Biomedical Research Involving Human Subjects , the International Conference of Harmonization Guideline for Good Clinical Practice , CIOMS 1991 International Guidelines for Ethical Review of Epidemiological Studies , Canadian Tri-Council Policy Statement , United Kingdom Medical Research Council Guidelines for Good Clinical Practice in Clinical Trials , and US federal regulations . These guiding documents were designed to protect the welfare and liberty interests of individual research subjects, but little guidance is provided with respect to community-based research, let alone cluster randomized trials. The sole exception is the UK Medical Research Council document (Cluster Randomized Trials: Methodological and Ethical Considerations) ; however, this document does not address the broad scope of ethical issues identified above, and its applicability to the regulatory environments of other countries is uncertain.
Researchers need direction on these ethical challenges in order to guide the ethically appropriate design and conduct of cluster randomized trials. REBs may be unfamiliar with this increasingly important study methodology. In the absence of formal guidelines for cluster randomized trials, REBs may fail to consider all of the relevant ethical issues generated by a study protocol, resulting in inadequate subject protection. Variable interpretation of ethical requirements for cluster randomized trials may lead to problems initiating multi-jurisdictional cluster randomized trials, and to unequal treatment of subjects in different jurisdictions. For example, Chaney et al.  reported on a series of health services implementation research studies involving primary care centres in the USA in which a combined total of more than 100 ethics review applications, amendments, and renewals were submitted at 17 research sites. Substantial variation in process and outcomes among sites was reported, with 35% of sites considering the research as exempt from review, 41% granting expedited review, and 18% requiring full review; approximately half of the REBs had ethical concerns necessitating changes to consent documents, and the number of days from submission to approval varied widely among sites, ranging from 3 to 82 days. Dziak et al.  reported on the ethics review process in a health services research study involving 15 primary care sites; REBs varied in the types of review required, and the number of days from submission to approval ranged from 5 to 172 days. According to the authors, variability in requirements for informed consent resulted in significantly different response rates among sites and affected sample generalizability.