Approved by SACHRP March 13, 2014
- Definition of a Cluster Randomized Trial
- Scientific validity
- Not research and Not research involving human subjects
- Overlap with QI projects as defined in OHRP FAQs
- Which institutions are engaged in research?
- Can CRTs meet the definition of exempt research under 45 CFR 46.101(b)(1) through (b)(6)?
- Who is a subject?
- Identifying the risks and benefits of the research
When is consent necessary for subjects?
- Waiver or alteration of consent
- Consent cannot be obtained until after randomization
- Voluntary participation
- Refusal to participate will involve no penalty or loss of benefits to which the subject is otherwise entitled and the subject may discontinue participation at any time without penalty or loss of benefits to which the subject is otherwise entitled.
- When can deception be used to help blinding?
- The role of gatekeepers.
Subparts B, C, and D
Cluster randomized trial (CRT) designs are frequently used in human subjects research. These trials bring up unique issues of regulatory application. The purpose of these recommendations is to address the application of US HHS and FDA regulations to cluster randomized trials.
Definition of a Cluster Randomized Trial
The central defining feature of a cluster randomized trial (CRT) is that randomization occurs on a group level rather than an individual level. In a traditional randomized clinical trial, subjects are randomized sequentially as each subject is identified and then enrolled in the study. In contrast, in a CRT the randomization occurs as a function of being a member of a group. In addition, there can be several layers of groupings as well, for instance by school district, school, and class, or by health care facility, medical provider, and each medical provider’s patients.
Examples of Cluster Randomized Trials
It is useful to distinguish between different kinds of CRTs based on the level at which the intervention is delivered. In a cluster-cluster trial, the intervention is delivered at the cluster-level. Usually, the intervention is "not divisible at the individual level" and is therefore necessarily delivered on a cluster-wide basis. In a professional-cluster trial, the intervention is delivered to health or other professionals working within clusters (providers, teachers, employers), while outcomes are then collected on the individual cluster members (patients, students, workers). An individual-cluster trial most closely resembles a standard randomized controlled trial in that the intervention is delivered directly to the individuals themselves. Usually, in an individual-cluster trial it would be possible to randomize by individual rather than cluster, but the cluster method is used for a methodological reason such as the prevention of exposure to trial aims among subjects. Another factor of the individual cluster design is that subjects have the ability to decline participation. Three examples follow to provide illustration.
The COMMIT study (Community Intervention Trial for Smoking Cessation) was designed to test the effectiveness of a comprehensive, community-oriented approach to influence citizens’ smoking behaviors. As the intervention is delivered at the community-level, this is an example of a cluster-cluster trial. Twenty-two communities with populations between 50,000 and 250,000 in the USA and Canada were randomly assigned to intervention or control. The intervention included activities focused on public education using mass media and organized community events, training of health care providers in cessation techniques, promotion of smoke-free policies in health care facilities and worksites, promotion of policies to restrict the sale of tobacco to youth, and development of smoking cessation resources and activities in each community. Population-based surveys, using random digit telephone dialing, were used to measure outcomes. Before randomization and at the end of the study, cross-sectional samples of approximately 2500 households per community were surveyed about their smoking behaviors. In addition, cohorts of approximately 550 heavy and 550 light-to-moderate smokers, willing to be contacted annually about their smoking status, were identified in each community. Main outcome measures were cross-sectional changes in the prevalence of smoking from pre- to post-intervention, and quit rates in the cohorts of smokers. Although the intervention significantly improved quit rates among light-to-moderate smokers, there was no significant effect on quit rates among heavy smokers or on the community prevalence of smoking.
Linder et al. used a cluster randomized trial to test a set of novel enhancements to electronic health records, designed to improve tobacco treatment and counseling in primary care. The enhancements included smoking status icons, tobacco treatment reminders, and facilitated ordering of medication and counseling referrals. Twenty-six primary care practices (521 clinicians) using electronic health records in Massachusetts were randomized to intervention or control. The enhancements were introduced to intervention practices with an introductory e-mail to clinicians, a practice visit by an investigator, and periodic e-mails to encourage use of the enhancements. Clinicians in control practices received no intervention. Because the intervention was targeted at health professionals, this is an example of a professional-cluster trial. Practices instead of physicians were randomized to facilitate the introduction of the intervention, reduce contamination, and potentially increase the effectiveness of the intervention through peer effects. The primary outcome was the proportion of documented smokers who made contact with a smoking cessation counselor. Secondary outcomes included documentation of smoking status in the electronic records and prescription of cessation medications. Over a 9 month period, data on 315,962 visits by 132,630 patients in the control and intervention practices were collected from the practice electronic records. The institutional review board granted a waiver of informed consent for included clinicians and patients. The intervention significantly increased contact with a cessation counselor as well as documentation of smoking status, but no difference was found in prescription of smoking cessation medications.
The ObaapaVitA trial was a double-blind, placebo-controlled trial to evaluate the effect of weekly, low-dose Vitamin A supplementation on pregnancy-related and all-cause female mortality in Ghana. As interventions were delivered to individual women, this is an example of an individual-cluster trial. A total of 1086 small clusters of compounds were randomized to either vitamin A or placebo capsules. Fieldworkers visited all compounds over a 1-2 month period to recruit women for the trial. All women of reproductive age who provided informed consent were enrolled in the trial (104,484 women in the treatment arm and 103,297 in placebo). Capsules were distributed during home visits undertaken every 4 weeks. Fieldworkers gathered data on pregnancies, births, and deaths. The study found no significant effect of Vitamin A supplementation on pregnancy-related or all-cause maternal mortality. Although individual randomization could in theory have been used, the use of cluster randomization considerably simplified the trial organization and fieldwork. The trial area was divided into small geographical clusters of compounds, designed to contain a maximum of 120 women each. Each fieldworker was responsible for an area of four contiguous clusters and expected to visit women in one cluster per week over a 4-week cycle. Randomization of clusters also minimized the possibility of women receiving the wrong capsules as fieldworkers only had one type of capsule in their possession during any week. Furthermore, cluster randomization allowed implementation of an extensive information, education, and communication campaign to promote adherence through radio messages, loudspeaker vans and drum beaters, messages delivered through churches and mosques, posters, and health workers.
Scientific rationale for use of CRT designs
Generally, the reasons for adopting a CRT almost always rest on practical (e.g., cluster-level intervention), logistical, or other considerations (see below). The CRT offers few scientific advantages over an individually randomized trial. The advantages that do exist need to be weighed against several disadvantages and limitations.
Advantages/Reasons for Use
The appropriate use of the CRT is driven by the nature of the intervention, the logistics of implementing the intervention, and the particular scientific question of interest.
When the trial is evaluating a cluster-level intervention (cluster-cluster trial), a CRT is the only design option. For example, a large-scale community health trial for the prevention of cardiovascular disease involving television, radio and billboards, cannot possibly be evaluated using individual randomization. Other examples of cluster-level interventions requiring a CRT include interventions that involve changing the environment, such as fluoridation of community water supplies, and innovative changes in health service delivery or administration, such as the provision of improved HIV testing services at designated centers.
Another common reason for choosing a CRT is to avoid contamination. This is a common justification in both professional-cluster and individual-cluster trials. Contamination occurs when individuals in the control arm are partially exposed to the intervention through interaction with individuals receiving the study intervention, thus biasing the results towards smaller effect sizes. Contamination may arise at both the health professional and individual levels. For example, in a trial of an educational intervention administered by a health provider, it would be difficult for a health provider to educate some patients and not others; further, at the individual level, patients attending the same clinic may discuss the educational intervention in the waiting room. The only way to avoid these risks is to randomize health providers, rather than patients. The risk of contamination is particularly great in the case of unblinded or behavioral interventions. For example, in a CRT for prevention of coronary heart disease, worksites may be randomized to minimize the likelihood of workers in different intervention groups sharing information about the trial. Increasing the sample size of an individually randomized trial to allow for contamination may sometimes be preferable to adopting cluster randomization, given the methodological challenges presented by this design.
Another reason for choosing a CRT occurs when indirect effects of a study intervention are of interest. For instance, in vaccine studies the overall effectiveness of a vaccine is a combination of individual immunity conferred by the vaccine and the reduced chance of encountering an infectious person (so-called "herd immunity"; see below).
Other common reasons for adopting the design in individual-cluster trials relate to logistical or administrative convenience. Cluster randomization may considerably simplify fieldwork (see example 3 above). CRTs may offer cost savings in some circumstances. For example, a trial that requires the use of expensive equipment or personnel (e.g., nurse specialists) would be less costly when implemented as a CRT, because the equipment or personnel need be provided to only half the centers as opposed to all centers if individuals within centers were randomized. In some trials, the outcome measure may be a rate defined at the level of the cluster with the data easily obtainable from routine administrative databases available for each cluster; individual randomization would require data directly from individuals with accompanying increases in cost and administrative requirements. Cluster randomization may help ensure that the intervention is fully or properly implemented. For example, in example 3 above, cluster randomization may have helped to prevent sharing or swapping of medications among community members (in the hope of getting some benefit should they be randomized to placebo). Cluster randomization may enhance compliance, promote publicity at the cluster level, or reinforce the effective use of a new technology within a cluster. Cluster randomization may be required for political reasons. For example, a design whereby only half of the members in a community receive an intervention may not be acceptable to decision makers or village elders, and may cause resentment among those being denied the intervention. Similar reasons may apply in professional-cluster trials: for example, it may not be acceptable to physicians to have only some of their patients offered a screening intervention. In these trials, the only feasible way to secure cooperation and successfully recruit participants is to use cluster randomization.
Whereas individually randomized trials provide information only about the direct effect of an intervention on the people who receive it, CRTs allow one to also study whether people benefit from an intervention provided to other members of the community (i.e., indirect effects of an intervention). Indirect effects are particularly important in studies of infectious diseases. For example, the effects of vaccines that are designed to block the transmission of a parasite that spreads malaria cannot be evaluated in an individually randomized trial. To examine the effect of such a vaccine on infection rates in the community, a CRT is required. Similar considerations apply in studies of HIV transmission where an intervention may be designed to reduce the "infectiousness" of HIV-infected individuals to their sexual partners. Such an effect could not be measured in an individually randomized trial.
The CRT design is statistically less efficient than an individually randomized design. For the same total number of subjects, CRTs with positive intracluster correlation always have less power than an individually randomized trial; the sample size calculation must take the intracluster correlation into account to ensure an adequately powered trial. This means that a larger sample size is required to yield the same power as an individually randomized trial. The loss of efficiency is a direct result of positive correlation among responses from individuals in the same cluster. CRTs require special methods to be used in sample size calculation as well as in statistical analysis as standard methods are usually invalid. CRTs are therefore more complex to design and analyze. Results from CRTs may also be more difficult to interpret. First, selection biases are a more serious concern in CRTs than in individually randomized trials, particularly when randomization of clusters is necessary prior to participant recruitment and allocation concealment is not possible. Second, imbalances between study arms are more likely in CRTs because the number of clusters randomized is often quite small. Given these methodological challenges associated with cluster randomization, individual randomization is always the method of choice, unless there are cogent reasons for adopting a CRT design.
Determining whether the activity is research involving human subjects
A threshold process for the ethical review of research is determining whether the activity does in fact meet the definition of research under the HHS regulations, and if it does meet the definition of research, then determining whether or not there are human subjects involved. The use of a CRT design does not determine in and of itself whether or not a given project meets the definition of research or meets the definition of research including human subjects. Certain activities that use cluster randomized design will not meet the definition of research, and some activities using cluster randomized design will not meet the definition of research involving human subjects.
Overlap with Quality Improvement (QI) projects as defined in OHRP FAQs
Often CRTs will meet the definition of “research” in 45 CFR 46, and the various definitions of “clinical investigation” in the FDA regulations 21 CFR Parts 50, 56, 312, and 812. However, CRTs may also meet the definition of a quality improvement project as defined in the OHRP FAQs on quality improvement projects. Thus, one of the threshold regulatory issues to consider with a given CRT is whether or not it is research or a clinical investigation under the regulatory definitions. If a CRT does not meet those definitions, then as a regulatory matter the project does not meet the requirements for IRB review and informed consent.
The OHRP FAQs provide two examples of QI activities that do not meet the definition of research. First, the HHS regulations for the protection of human subjects in research (45 CFR part 46) do not apply to quality improvement activities conducted by one or more institutions whose purposes are limited to: "(a) implementing a practice to improve the quality of patient care, and (b) collecting patient or provider data regarding the implementation of the practice for clinical, practical, or administrative purposes."
This type of QI activity could be conducted using a cluster randomized design. For instance, two hospitals could be randomized, with one hospital implementing a practice to improve the quality of patient care, while the other hospital does not implement the practice. Examples could include having nursing staff wash their hands once an hour, or having two additional nursing staff working on each shift. The fact that this was done using a cluster randomized design would not in and of itself cause this activity to be research under 45 CFR 46.
Similarly, this type of cluster randomized QI activity involving FDA regulated products may not meet the definition of a clinical investigation. For example, a hospital could use one type of approved air mattress for burn victims in one wing of a burn unit, and a different approved air mattress in the other wing, and then collect patient satisfaction information about each mattress based on noise level. As another example, the purpose of such a trial could be to provide data on cost effectiveness rather than to establish the safety or effectiveness of the mattresses. In both cases, this trial would not meet the FDA definition of clinical investigation of a device.
The OHRP FAQs also provide second example of QI activities that do not meet the definition of research. The HHS regulations for the protection of human subjects in research (45 CFR part 46) do not apply to quality improvement activities if their purposes are limited to: "(a) delivering healthcare, and (b) measuring and reporting provider performance data for clinical, practical, or administrative uses." A cluster randomized design could be used to deliver different healthcare methods on two floors of a hospital. For example, the floors could be randomized to using one brand of catheter on one floor and a different brand on the other floor for the purpose of addressing the observation that medical providers are reluctant to make use of a newer, less expensive model at the hospital because they are concerned that the catheter will not be as cost effective. As long as the data collected is used for clinical, practical, or administrative uses, the project would not qualify either as research under 45 CFR 46. As long as the purpose is not to collect safety or efficacy data about the devices, it also is not a clinical investigation under 21 CFR 56 or 812.
Public health projects
Public health authorities often will try various methods of public health interventions, varied across neighborhoods or other jurisdictional units, in an effort to determine the most effective or efficient interventions. In certain cases, these activities will not meet the definition of research because they are not intended to produce generalizable knowledge. For example, within one city jurisdiction a health department that provides school nursing services may determine to vary vaccination delivery practices among schools, providing required vaccinations directly and on-site in one set of schools, but in other schools requiring parents to seek vaccinations from public health clinics or private physicians. The public health authority then can compare vaccination rates among the schools, all in order to understand whether on-site vaccination services best achieve acceptable vaccination rates among school children. Similarly, delivery of STD and HIV screening services, and community promotion of those services, can be varied by clinic and neighborhood, in order to determine the most effective and efficient use of limited screening resources.
In these cases and others, the purpose of varying the intervention among service delivery sites and neighborhoods is not to derive generalizable knowledge, even though aggregated experiences, if accompanied by adequate data gathering, might give rise to publishable findings that are generalizable knowledge. Instead, the purpose of these interventions –which typically are discretionary public benefit interventions, not interventions dictated by patient “rights” to care and services – is to promote the most optimal allocation of limited public health resources. Randomization is done at a level far beyond the individual patient because (1) such a design is more efficient than individual randomization and moreover (2) the public health authority’s own success and failure is measured on an aggregate, not individual patient, level. Thus, such public health activities are most often regarded not as “research,” but as the delivery of an acceptable range of public health interventions, grouped and then measured by service delivery site or neighborhood. Recipients of public health services are generally not thought to be required to undergo a consent process regarding the variant of service that they are receiving. However, whether a public health intervention qualifies as research must be determined on a case by case basis. The fact that an activity is a public health intervention does not automatically exclude it from also meeting the regulatory definition of research.
Which institutions are engaged in research?
After it has been determined that a project is research and that it is research involving human subjects, the next threshold issue is to determine which institutions are engaged in the research. When an institution is engaged in research, then the institution is required to oversee the research in compliance with HHS regulations, including issues such as IRB review, informed consent, and registration with OHRP. In cluster randomized trials, it can be difficult to determine which institutions are engaged in research, particularly in studies such as the example involving a community smoking cessation program. An analysis must be performed for each institution involved in a CRT to determine if it meets the criteria of being engaged in research. The OHRP guidance document “Engagement of Institutions in Human Subject Research” provides criteria for determining if given institutions are engaged. It may be necessary to first determine which participants in the research are subjects, as discussed below, before being able to apply some of the criteria in the guidance.
Can CRTs meet the definition of exempt research under 45 CFR 46.101(b)(1) through (b)(6)?
Yes, CRTs can be exempt research under all of the exemption categories if they meet the exemption criteria. The CRT design does not in and of itself determine that the criteria are met or not.
Who is a subject?
An essential issue in the application of the regulations to CRTs is determining which individuals meet the definition of a human subject under 45 CFR 46 and the FDA regulations. The potential subjects include medical providers, their patients, teachers, their students, and individuals who are the targets of the cluster randomized research activities. Under 45 CFR 46, the definition of a human subject is:
(f) Human subject means a living individual about whom an investigator (whether professional or student) conducting research obtains
(1) Data through intervention or interaction with the individual, or
(2) Identifiable private information.
Intervention includes both physical procedures by which data are gathered (for example, venipuncture) and manipulations of the subject or the subject's environment that are performed for research purposes. Interaction includes communication or interpersonal contact between investigator and subject. Private information includes information about behavior that occurs in a context in which an individual can reasonably expect that no observation or recording is taking place, and information which has been provided for specific purposes by an individual and which the individual can reasonably expect will not be made public (for example, a medical record). Private information must be individually identifiable (i.e., the identity of the subject is or may readily be ascertained by the investigator or associated with the information) in order for obtaining the information to constitute research involving human subjects.
It is important to note that the definition includes a disjunctive “or” between the two sub-clauses, such that an individual becomes a subject if either the investigator obtains data through intervention or interaction with the individual, or if the investigator obtains identifiable private information about the individual even when there is not an intervention or interaction. Either condition suffices to make the individual a subject; both conditions do not have to be satisfied.
It is also important to note that for the purpose of defining intervention or interaction, the term “data” does not refer only to identifiable data. Collecting aggregate data from a population or sub-population exposed to a research intervention, such as general infection rates across hospitals or lead blood levels, meets the definition of obtaining data about a living individual. Even though you are only measuring data from a sub-population, all of the exposed individuals are subjects. This falls within the intent of the regulations.
In certain cases, such as professional cluster designs, a difficult issue is determining which individuals are subjects. To provide direction on how to make this determination, it is important to elucidate the meaning of the phrase “manipulations of the subject or the subject’s environment” in the definition of an “intervention.” The regulation states that “Intervention includes both physical procedures by which data are gathered (for example, venipuncture) and manipulations of the subject or the subject's environment that are performed for research purposes.” SACHRP recommends that a “but for” test may be helpful in interpreting this phrase. The “but for” test in its most simple form asks “but for the existence of X, would Y occur?” The “but for” test is used in law to establish causation for the purpose of determining liability, for instance by considering whether an injury would not have occurred but for the defendant's negligent act. For the purposes of determining whether an individual will be a subject in a professional cluster design study, the “but for” test can be articulated, “but for the existence of the research, would the environment still be manipulated in this same way if the research didn’t take place? When the individual’s environment has a reasonable possibility of being manipulated by the existence of the research, then the individual is a research subject. For the purposes of medical research, the test can be phrased, “but for the existence of the research study, would the subject’s medical care be manipulated in the same way if the research didn’t take place?” The guidance should instruct IRBs to consider a short chain of events that could lead to manipulation of the subjects’ care, rather than thinking of attenuated chains of unlikely events that could lead to the manipulation of the subject or the subject’s environment.
It is important to note that if the activity is not research, then the application of the “but for” test is moot. It only answers the question of whether the research manipulates the individuals’ environment, not other activities, such as QI projects, insurance policy changes, or changes in hospital staffing for budget reasons.
The FDA regulations (which apply to “clinical investigations”) provide a different definition of a human subject. “Human subject means an individual who is or becomes a participant in research, either as a recipient of the test articles or as a control. A subject may be either a healthy human or a patient.” (21 CFR 56.102(e)). For devices, human subject also includes an individual “on whose specimen an investigational device is used” (21 CFR 812.3(p)). The application of the FDA definition of a human subject to CRTs is not as complicated as the application of the HHS definition. However, it is important to note that under FDA regulations, persons in a control arm qualify as subjects.
To illustrate the process for determining whether providers, their patients, teachers, their students, and other individuals are research subjects, we provide three examples.
A. Antibiotic Ointment versus Antibacterial Soap in ICU: Consider a CRT in which ICUs are randomized to the use of either antibiotic ointment or antibacterial soap on the patients to prevent the spread of staph infection. The purpose is to study the safety and efficacy of the products.
Are the ICU patients subjects under 45 CFR 46 in this proposed study? Yes, the ICU patients are subjects under 45 CFR 46. The research involves interaction with the patients because they are exposed to different products used in the ICU to prevent the spread of staph infection.
Are the ICU patients subjects under FDA regulations? Yes, the ICU patients are subjects under the FDA regulations because the safety and efficacy of medical products are being tested in the research, and it qualifies as a clinical investigation that requires IRB review. It is likely it would not need an IND, because if fits under the IND exception at 21 CFR 312.2. However, that does not alleviate the need for IRB review and oversight. In addition, consent cannot be waived because this is an FDA regulated study. If a subject doesn’t agree to participation, then their data cannot be used for the study purposes. In some cases, such as this one, the patient may end up being exposed to the research intervention because it is the only intervention available at that site.
B. Strategy Training of Psychiatric Care Staff and Physicians for Reduction of Seclusion, Restraint, and Aggressive Behavior: Inpatient psychiatric acute care units for patients with serious mental illness are randomly assigned to one of three approaches to manage aggressive behavior on the unit: the first approach provides medical staff with classroom training emphasizing the value of reducing the use of seclusion and restraint use; the second arm provides medical staff with training focused on methods of early identification of aggressive behavior, de-escalation and intervention using alternatives to seclusion and restraint where possible; and the third arm involves no intervention in standard practice. The measured outcomes will be episodes of patient seclusion, application of restraint, and episodes of aggressive behavior by patients. The researchers who provide the intervention also interview the medical staff at the end of the study about their views on the effectiveness of the intervention, and record identifiable information about the medical staff’s application of the techniques being taught in the first and second arm. In the third arm, there are interviews with the medical staff to gather their views on standard techniques.
Are the medical staff subjects under HHS regulations? Yes, the medical staff are subjects under 45 CFR 46 because the researchers interact with the medical staff to obtain research data about the staff.
Are the patients subjects under the HHS regulations? Yes, the patients are subjects under 45 CFR 46 because their environment is manipulated by the research. If the research did not exist, the patients in the first and second arm would receive different care than if it did exist. In addition, if the researchers gather private identifiable information about the patients, then they would be research subjects on that criteria as well.
Are the medical staff subjects under the FDA regulations? No, the medical staff are not subjects under the FDA regulations because they are neither recipients of FDA regulated test articles nor controls.
Are the patients subjects under the FDA regulations? No, the patients are not subjects under the FDA regulations because they are neither recipients of FDA regulated test articles nor controls. However, if the use of FDA regulated drugs were part of the applied techniques, or if the safety or efficacy of an FDA regulated device were being studied, then the patients would be research subjects under the FDA regulations. In this case, however, the interventions did not include FDA regulated products.
C. Smoking Cessation Study: In the smoking cessation example of a CRT above, communities are randomized to be exposed or not to a smoking cessation campaign.
Are the members of the communities all subjects under 45 CFR 46? Yes, the community members are subjects under 45 CFR 46, because their environment is manipulated by the presence of the smoking cessation campaign. However, as discussed below, this research will qualify for a waiver of consent because the manipulation involves minimal risk.
There are research designs under which certain groups of individuals will not meet the definition of a subject. For instance, consider a study in which school counselors are randomized by school to use one of two different checklists of criteria to detect possible cases of child abuse. The effectiveness of the two different checklists will be tested only by interviews and surveys with the school counselors. There will be no effort to measure any rates of child abuse detection or follow up in the community, and no contact with the students in the schools. In this case, the students are not research subjects because no data is obtained about them through intervention or interaction, and no private identifiable information is collected, even though their environment has potentially been manipulated.
Each proposed CRT research project requires a careful analysis as to whether the various levels and clusters of participants are research subjects. Even if certain clusters of individuals are found not to meet the definition of research subjects, the IRB or institution may wish to consider whether there are issues of unacceptable risks, lack of informed consent, or other issues affecting that cluster population.
Identifying the risks and benefits of the research
The criteria for IRB approval require that IRBs determine that risks are minimized and that the risk/benefit ratio is appropriate. In addition, subjects must be informed of risks and benefits as part of the consent process. The risks and benefits to the subjects in each level and cluster must be considered (e.g., randomized medical providers and their patients).
One issue that arises is that there is not uniformity in the regulated community in designating which risks are in fact research risks. The regulations direct that, “in evaluating risks and benefits, the IRB should consider only those risks and benefits that may result from the research (as distinguished from risks and benefits of therapies subjects would receive even if not participating in the research).” This is a clear cut issue when an investigational new product is being tested, but it is not as clear when the research involves a registry, a phase IV study, arms in which standard of care is provided, or public health interventions. In many CRTs, there is not uniformity as to the risks of the research, particularly in those arms that involve standard of care interventions. Some IRBs will consider the risks of a standard of care arm as the risks of the research, and also consider the benefits of that arm as research benefits. Other IRBs will consider the risks of a standard of care arm as not being research risks, and the benefits as not being benefits of research.
In the example involving the randomization of ICU units to antibiotic ointment or antibacterial soap, the use of both approved products fall within standard of care. The risks include the fact that one product may cause more adverse events, such as skin irritation to the soap or allergic reaction to the antibiotic, or that one product may not be as effective in preventing the spread of staph infection and subjects may develop treatment-resistant staph infection.
For another example of the risks involved in CRTs, consider a community-based study to address the recognized problem of post-partum hemorrhage in rural Indonesia. One hundred villages will be randomized to either be provided with misoprostol, an inexpensive drug to treat post-partum hemorrhage, or have no access as a matter of local standard of care. In the active arm of the study, pregnant women will be asked to consent to participate in the research, and will receive tablets of misoprostol in a small baggy with directions on use presented in pictures. After they have their children, the women in the active arm will be interviewed to see if they had postpartum bleeding and used the misoprostol. In the control arm of the study, pregnant women will be identified by professional surveyors, but there is no intervention in their care and they will not be asked to provide consent. After they have their children, the women’s level of postpartum bleeding will be determined by professional surveyors. In the active arm, the women face the risks of misoprostol, particularly if they take it while still pregnant prior to the birth of the infant. In the control arm, the women face the risks of untreated postpartum bleeding as they would have if the research did not exist. This illustrates that there can be above minimal risk in CRTs.
Informed consent and waiver or alteration of consent in CRTs
The HHS and FDA regulations require that research subjects consent to their participation in research unless a waiver of consent is acceptable. The standards for waiver of consent under the two sets of regulations differ. Under HHS regulations, there are certain waivers of consent possible. Under the FDA regulations, distinct waivers of consent are not possible. The most common of these waivers is 45 CFR 46.116(d), whereby consent can be waived if four conditions are met:
(1) The research involves no more than minimal risk to the subjects;
(2) The waiver or alteration will not adversely affect the rights and welfare of the subjects;
(3) The research could not practicably be carried out without the waiver or alteration; and
(4) Whenever appropriate, the subjects will be provided with additional pertinent information after
In addition, 116(d) can be used to alter the consent such that certain information is not disclosed.
However, FDA has not adopted the 116(d) regulatory criteria for waiver of consent. Therefore, to determine if consent can be waived or altered, there first must be an analysis of which regulations apply, and then for each cohort of subjects, appropriate regulatory analysis must be applied in order to determine if consent may be waived under the applicable regulations.
The CRT is a design that requires careful justification. There are usually cogent reasons to adopt a CRT: in the case of cluster-cluster and professional-cluster trials, the reasons are usually obvious; in the case of individual-cluster trials, individual randomization is possible — at least in principle — but contamination, efficiency, or political factors argue for the use of a CRT. The use of a CRT does not change the general presumption that individual informed consent is required, unless conditions to justify a waiver of consent exist. According to the Ottawa Statement, "an inappropriate reason to adopt a CRT is the mistaken belief that the need to seek informed consent can be avoided by using cluster randomization".
When IRBs approve a waiver of consent for a CRT, the IRB, institution and investigator may wish to consider whether it would be appropriate to perform community outreach to provide knowledge to the affected population of existence of research. This does not substitute for informed consent from individuals, but may be respectful of autonomy in those cases where the IRB has made the finding that the research meets the regulatory criteria for waiver of approval.
Is it acceptable to obtain consent after randomization?
Under the HHS regulations, IRBs can consider whether not obtaining consent must be justified as an alteration of consent under 45 CFR 46.116(d). One issue that arises in CRTs is that clusters of potential subjects may be randomized on a group basis without being approached for consent prior to the randomization. In these cases, IRBs should consider whether subjects should be asked if they will allow the use of the data collected at the earliest possible opportunity where such consent is required. At that point, depending on the study design, subjects may also exercise their right not to participate in the research going forward. However, the right not to participate will depend on the study design. In some cases, it may not be possible to decide not to participate, because for instance the randomized intervention is the only treatment available. An example of this is the research described above in which subjects are randomized by ICU to be cleansed with either an antibiotic ointment or antibacterial soap. If a subject doesn’t agree to participation, then their data cannot be used for the study purposes. In some cases, such as this one, the patient may end up being exposed to the research intervention because it is the only intervention available at that site. This issue does not automatically cause the research to be unacceptable from a regulatory perspective. However, the IRB must consider whether not obtaining consent at the time of randomization exposes the subjects to an unacceptable level of risk or an unacceptable restriction on autonomy. The potential subjects should be given the opportunity to consent at the first opportunity.
Voluntary Participation; Opportunity to decline participation
Another issue that arises with CRTs is that when the intervention is administered at the level of the cluster, such as the community or the institution, the subjects often may not have an opportunity to decline participation after their group has been randomized because the entire cluster is affected. For instance, in the smoking cessation example above, subjects located in the communities randomized to either have the smoking cessation campaign or not have no choice as to whether to participate. They will either be exposed to the campaign or not.
A regulatory element of consent is that “Refusal to participate will involve no penalty or loss of benefits to which the subject is otherwise entitled, and the subject may discontinue participation at any time without penalty or loss of benefits to which the subject is otherwise entitled”
Another regulatory issue that can arise is that subjects must be informed that their refusal to participate in the research will not lead to any penalty or loss of benefits to which they are otherwise entitled, and that the same applies if they discontinue participation. In the SATURN study, school districts were randomized to having or not having after drug testing programs for after-school sports. One potential subject complained that if she did not participate in the research and be tested for illegal drugs then she could not participate in after-school sports. IRBs must always consider whether this situation applies in any CRT study.
When can incomplete disclosure through an alteration of consent be used in a CRT to avoid contamination?
45 CFR 46 also allows for alterations of consent, and this approach is commonly used to allow deception in certain types of research in order to strengthen the validity of the research. This same technique can be used in a CRT if the criteria at 45 CFR 46.116(d) for an alteration of consent are met. As one example, a study design may be such that if subjects are told about both arms of a study, then it could contaminate the results. In this case, if the 116(d) criteria are met then information about the other arm can be withheld.
In an FDA regulated study, it is also acceptable to withhold certain information in the initial consent process if the withholding is essential to ensure blinding. This would not be considered deception, just as blinding is not considered deception. However, it would usually be appropriate to disclose the withheld information to the subject at such point as it is no longer necessary for the blinding.
Subparts B, C, and D
If the CRT involves the protected subject populations that are addressed in 45 CFR 46 Subparts B, C, and D, then the IRB must apply those regulatory criteria.
For instance, if the research involves pregnant women, fetuses, neonates of uncertain viability, or non-viable neonates, then Subpart B must be applied. Once the subjects are identified and the risks assessed, then the application of Subpart B should be straightforward. As a useful standard for determining whether these populations are involved, it is acceptable to not consider these groups to be included unless the investigator or IRB has direct knowledge that any of the subjects are pregnant or otherwise meet the definitions, as long as the research is of minimal risk to them. If the research would be of more than minimal risk to these subjects, then the research should be modified to ensure those risks are acceptable or to exclude these populations. If the investigator or IRB learn that these populations have become enrolled, then the IRB should consider whether their continued participation is acceptable under the regulatory requirements.
If the research involves prisoners, then Subpart C must be applied. Once the subjects are identified and the risks assessed, then the application of Subpart C should be straightforward. As a useful standard for determining whether these populations are involved, it is acceptable to not consider these groups to be included unless the investigator or IRB has direct knowledge that any of the subjects are prisoners. If the investigator or IRB learn that these populations have become enrolled, then the IRB should consider whether their continued participation is acceptable under the regulatory requirements. [The earlier SACHRP recommendations on incidental inclusion of prisoners in research apply here as well.]
For subpart D, once it has been established who the subjects are and what the risk levels are, then the application of Subpart D will proceed as normal.
The Role of Gatekeepers
As with any research project, researchers performing CRTs must obtain the agreement of gatekeepers such as nursing home directors, school principals, and other officials to conduct research at a given organization. However, as a regulatory matter, that permission cannot substitute for the informed consent of the subjects in a CRT, or an IRB approved waiver of consent.
Recommended modifications to the regulations
SACHRP recommends that certain modifications to the regulations would allow the ethical conduct of CRTs and other research designs to proceed. First, an earlier SACHRP recommendation was that when applying the 45 CFR 46.116(d), IRBs should be able to use a component analysis approach to allow consideration of whether the proposed alteration of consent is of minimal risk, rather than whether the research as a whole is minimal risk. This earlier recommendation is available in the SACHRP January 2013 letter to the Secretary, attachment D, section d(1).
Second, we recommend that FDA adopt the provisions for waiver of consent that exist under 45 CFR 46.116(d). This would allow for certain minimal risk research under FDA regulation to be conducted with this waiver of consent when the criteria are met.
Third, we recommend that HHS consider altering the definition of a human subject in the HHS regulations. One factor making the HHS definition of a research subject difficult to use in the context of CRTs is that the current regulations appear to assume that the subjects who experience an intervention or interaction are the same subjects about whom data are collected from those interventions or interactions.. Yet in many CRT designs, individuals experience the research intervention by virtue of their being a member of the cluster, but data are not collected directly about all of the individuals within that cluster. For instance, in a CRT looking at the effectiveness of HIV prevention strategies with randomization on a community-wide basis, the study design might require collection of data from only 20 percent of the individuals affected by the intervention in order to prove the efficacy of the intervention, even though the entire cluster is affected by the intervention. For these reasons, and in order to protect persons in clusters who do undergo research interventions, SACHRP recommends that the HHS definition of a research subject be interpreted (or, if necessary, amended) so that these cluster member individuals from whom no data or identifiable private information is collected, are nevertheless afforded appropriate protection under the human subject protection regulations.
As one possible means to this end, we recommend consideration of the following change to the HHS definition of a human subject:
(f) Human subject means a living individual (1) who receives a research interaction or intervention whose effects on humans are studied, or (2) about whom an investigator obtains identifiable private information.
Research Intervention includes both physical procedures (for example, venipuncture) and manipulations of the subject or the subject's environment that are performed for research purposes Interaction includes communication or interpersonal contact between investigator and subject. Private information includes information about behavior that occurs in a context in which an individual can reasonably expect that no observation or recording is taking place, and information which has been provided for specific purposes by an individual and which the individual can reasonably expect will not be made public (for example, a medical record). Private information must be individually identifiable (i.e., the identity of the subject is or may readily be ascertained by the investigator or associated with the information) in order for obtaining the information to constitute research involving human subjects.
The purpose of this change is to remove the necessity to collect data to cause an individual to become a human subject. Rather, just an interaction or intervention for research purposes would be sufficient. It is important to note that this could end up including research staff as being subjects.
We offer these recommendations to help the Secretary provide advice on the application of US regulations to cluster randomized trials.