As specified in the Common Rule, one way in which an investigator may intervene on human research subjects is by manipulating their environment. With respect to healthcare CRTs, Mann and Reyes have interpreted this to mean that an intervention designed to alter a healthcare professional's practice pattern entails a manipulation of the environment of all patients whose care may be affected . According to Mann and Reyes, every patient of a professional whose care may be influenced by a study intervention in a CRT meets the regulatory definition of a human research subject. This claim is understandable, given that much of the literature on the ethics of CRTs assumes that all cluster members (in this case, patients) will be subjects [2, 20, 25–28]. If correct, this view has important implications for the conduct of CRTs. If, in CRTs targeted at health professionals or health systems in general, all patients are considered human research subjects, the administrative burdens associated with protecting patients as human research subjects may threaten the feasibility of trials. We explore below whether the indirect effects of a research intervention at the group level (such as an educational intervention administered to health professionals) implies that all individuals within the group (including the patients of professionals participating) must be considered human research subjects.
1. Does environmental manipulation confer the status of human research subject on individuals?
The CIOMS International Ethical Guidelines for Biomedical Research Involving Human Subjects defines research as any study that manipulates an individual's social or physical environment . These guidelines, together with the CIOMS International Ethical Guidelines for Epidemiological Studies, address issues of informed consent and harm-benefit analysis in epidemiologic research . Therefore, without actually defining research subject, both CIOMS guidelines appear to acknowledge that individuals who may be directly affected by public health interventions that manipulate the environment, such as water fluoridation or pesticide use, are human research subjects [17, 20].
An unduly broad interpretation of environmental manipulation is, however, untenable. It seems implausible that everyone whose environment is manipulated in the context of a research project should be considered a human research subject. The term environment refers to "the surroundings or conditions in which a person, animal, or plant lives or operates" . Using this definition, "manipulations of the...subject's environment" would imply that entire populations might be involved in a research study. For example, studies in particle physics at the Large Hadron Collider meet the Common Rule definition of research, in that they are systematic investigations designed to develop generalizable knowledge and constitute "manipulations of...the environment" . Do the citizens of France and Switzerland thus become human research subjects? If experiments at the Large Hadron Collider are theoretically capable of producing microscopic black holes, does that imply that everyone on earth is a human research subject ? Clearly, a more restrictive interpretation of environmental manipulation is required.
It is helpful to consider what the US National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research meant when it included environmental manipulation in its definition of a human research subject. In a paper prepared for the National Commission in 1975 (included in the 1979 appendix to the Belmont Report), Robert Levine defined research as "...manipulation, observation, or other study of a human being -- or of anything related to that human being that might subsequently result in manipulation of that human being -- done with the intent of developing new knowledge and which differs in any way from customary medical (or other professional) practice" . Specific reference to manipulation of an individual's environment did not appear until the National Commission's 1978 Report and Recommendations: Institutional Review Boards which replaced Levine's definition of "anything related to that human being that might subsequently result in manipulation of that human being" with "manipulations of...the subject's environment" . This latter language was subsequently incorporated into the Common Rule. Exploring the reasons for this change in regulatory language helps us understand how to interpret environmental manipulation in the context of human subjects research.
Given the research ethics literature at the time, it seems likely that the National Commission was seeking to protect individuals who participated in studies evaluating the psychological effects of various environmental stimuli. Inclusion of environmental manipulation in the criteria for human research subjects seemed intended to cover research that deliberately manipulated subjects and jeopardized their welfare without physical contact. Examples of such research include studies examining the psychological and behavioural effects of habitation in simulated fallout shelters sponsored by the US Office of Civil and Defense Mobilization and studies evaluating the psychological effects of other environmental manipulations, such as sensory deprivation [32–34]. Environmental manipulations in the Civil Defense studies involved living in a confined space for prolonged periods of time, crowding, and variations in air quality, temperature and the availability of potable water. These manipulations of the environment caused human research subjects physical and psychological discomfort and jeopardized their welfare.
These studies, and the ethical concerns that they highlight, support a narrow reading of the environmental manipulation clause. We suggest that environmental manipulation must be designed to have a direct effect on individuals if they are to be considered human research subjects. We believe that this interpretation is consistent with the intent of the National Commission. We will therefore expand our criteria specifying human research subjects to include individuals who are deliberately affected via manipulation of their environment (see Table 1).
Our conclusion is also consistent with the language enshrined in the final report of the National Bioethics Advisory Commission. In its 2001 Report and Recommendations: Ethical and Policy Issues in Research Involving Human Participants, the Commission agreed that the term human research subject "connotes the fact that the individual is 'subjected' to an action by the investigator."  The Commission specifically recommended that "Research be considered to involve human participants when individuals 1) are exposed to manipulations, interventions, observations or other interactions with investigators or 2) are identifiable through research using biological materials, medical and or other records, or databases" .
2. Do indirect effects of CRT interventions on health professionals or health systems confer the status of human research subject on patients?
Mann and Reyes assert that educational or quality improvement interventions in a CRT designed to change physicians' practice patterns constitute manipulations of the environment of participating physicians' patients . As a result, they conclude that all patients of physicians whose care may be indirectly affected in such trials are human research subjects. We respectfully disagree.
First, for an individual to be properly considered a human research subject, the environmental manipulation must be designed to produce a direct effect on that individual. This is not the case in CRTs that intervene on health professionals. The interventions under evaluation in these studies are intended to change health professionals' behaviour by increasing their use of evidence-based strategies to improve care. Patients are not manipulated by these interventions; they are only indirectly affected by them.
Second, even if a change in professionals' practice patterns did constitute a deliberate manipulation of their patients' environment (a claim that we do not grant), that manipulation does not jeopardize patients' interests or mean that they ought to be considered human research subjects. Distinguishing between clinical practice and clinical research may help clarify matters, As Levine writes:
"If a physician proceeds in [the] interaction with a patient to bring what [the physician] considers to be the best available technique and technology to bear on the problems of that patient with the intent of doing the most possible good for that patient, this may be considered the pure practice of medicine. By contrast, if a physician interacts with an individual with the intent of developing new knowledge (not primarily for the benefit of that individual), this activity may be classified as research" .
In a CRT that intervenes on health professionals, the participating health professionals are research subjects, not investigators. They are still bound by professional responsibility to act in their patients' best interests. Thus, patient welfare is not put at risk in CRTs that evaluate interventions aimed at improving practice.
Some studies evaluate patient level effects as an outcome measure. The fact that a patient level effect may be measurable is relevant to patients only insofar as their private health information may be used, or they may be asked to submit to surveys or additional examinations to evaluate the outcome of the CRT. When this occurs, the patients of health professionals participating in an educational or quality improvement CRT ought to be considered human research subjects.
CRTs evaluating quality improvement initiatives or system-wide innovations aimed at healthcare organizations should be treated similarly to CRTs evaluating educational interventions aimed at providers. Patients need only be considered research subjects if they are directly intervened upon, if they interact with investigators, or if identifiable private information is used.
CRTs evaluating novel modes of healthcare delivery must be treated differently if the experimental intervention involves a departure from standard care . Such CRTs are conducted because the efficacy of the mode of delivery is uncertain. Randomization to clusters is undertaken for logistical reasons and to avoid treatment contamination. Novel modes of healthcare delivery are therefore best thought of as experimental interventions directed at patients. In such studies, the patients should be considered research subjects because they are directly intervened upon.
3. Implications for CRTs in fields other than healthcare
In some CRTs, particularly in public health, the purpose of the experimental intervention is to deliberately manipulate individuals via their environment. For example, in the COMMIT study, described above, multi-modal community interventions (environmental manipulations) were intended to produce behavioral change in smokers living in intervention communities [3, 4]. Another CRT evaluating interventions aimed at individuals via environmental manipulation compared rates of diarrheal illness in communities randomly assigned to water treatment with flocculant disinfectant or a control . In these studies, the purpose of the environmental manipulation is to intervene on individual community members. The residents of communities in these studies are therefore human research subjects and are entitled to regulatory protections. However, this does not mean that informed consent is required. Rather, many of these studies would meet criteria for a waiver of informed consent that are laid out in various national and international research ethics guidelines [11, 17, 18, 20]. These guidelines generally require that the research not be feasible if consent were required, and that the interventions in the study pose only minimal risk to subjects -- conditions that are often met by CRTs in public health. (When informed consent must be obtained in CRTs is the subject of another paper in this series).
CRTs in education are roughly analogous to CRTs in healthcare and the teacher-student relationship has many of the characteristics of a fiduciary relationship . If a CRT is used to evaluate the effect of a continuing education intervention for teachers, the indirect effect on students does not require treating them as human research subjects. However, CRTs of experimental curricular programs may be analogous to CRTs evaluating novel methods of health service delivery, and may require treating students as human research subjects.