Additional Thoughts on a Community Randomized Trial of Multi-Component HIV Prevention Programs
As discussed in Chapter 3, while evidence shows that the package of services in multi-component prevention programs are effective in reducing drug-related HIV risks, questions remain about the specific contribution of individual elements to reductions in risk behavior and HIV incidence. This issue is important from a policy perspective because certain aspects of the multi-component programs can be resource intensive. Further research is needed to help identify the most effective and cost-effective combination of programs that are feasible to implement in high-risk countries.
One approach for resolving the remaining questions is a multinational, multi-site trial randomized at the community level (community randomized trial), to evaluate the effectiveness and cost-effectiveness of multi-component programs of increasing complexity. Such a trial could determine whether programs without specific elements are less or possibly more effective and cost-effective, thereby providing policymakers with the tools to make informed decisions about which components to include.
While multiple smaller studies might address some of the questions, such studies might not have sufficient power to demonstrate effectiveness or sufficient sample size to control for confounding effects. Furthermore, because of the difficulty of finding truly single component programs to evaluate in isolation, smaller studies would likely have difficulty disentangling the effects of multiple components in one program.
While not intended to be comprehensive, the following sections provide
an overview of how such a trial might be designed and the operational, logistical, and ethical issues that might arise.
TRIAL DESIGN FEATURES
In a community randomized trial, some communities receive certain added interventions while other communities awaiting such interventions serve as comparison sites (often called “controls”). A stepped-wedge design would be most appropriate for a community randomized trial of multi-component HIV prevention programs. This design involves sequential rollout of an intervention (whereby intervention components are added to a standard package) in participating communities over time. Areas that are yet to receive a specific intervention serve as controls for the intervention area(s). This design is particularly relevant where an intervention may do more good than harm (making a factorial design, in which certain participants do not receive the intervention, unethical). Such a design is also appropriate where, for logistical, practical, or financial reasons, a program cannot simultaneously deliver an intervention to all participants (Gambia Hepatitis Study Group, 1987).
A community randomized trial makes particular sense for injecting drug users because they may share the same drug-using network and compare their treatment experiences. Thus randomizing participants on an individual basis could create situations where the control group could not be insulated from the intervention group, potentially contaminating the control regimen and blunting the study’s ability to detect important differences.
The multi-national, multi-site design feature of the community randomized trial is appropriate for several reasons. First, the burden on individual research sites in implementing a large trial across many communities with sufficient HIV incidence to evaluate interventions is high. Spreading the sample across many sites reduces the burden for a particular research group.
Second, implementing the study at numerous sites allows investigators to evaluate whether the interventions work in different communities. While the interventions would need to be tailored somewhat to the local context, this has proven feasible in trials of similar complexity. The National Institutes of Mental Health (NIMH) Collaborative HIV/STD Prevention Trial is a good example of a multi-national, multi-site community randomized trial. Funded by the National Institutes of Health, the study relies on local public opinion leaders to promote HIV testing and counseling and encourages risk reduction in five developing countries: China, India, Peru, Russia, and Zimbabwe (RTI International, 2005).
Third, using multiple sites increases the statistical power of the study. Because participants within one community are likely to have similar out-
comes, the power of a community trial will be lower than a trial of the same size employing randomization at the individual-level (i.e., the effective sample size will be between the number of participants and the number of sites). Increasing the number of communities will provide more power than increasing the number of participants in a single community. Increasing the number of communities in a single trial does increase logistics and cost, however, and care must be taken to provide adequate resources to enable local partners to carry out the study.
It would be important to select trial sites where the anticipated effect size (percent reduction in the incidence of HIV) would justify the effort. The trial would provide an opportunity to collect data on both biological and behavioral outcomes. One of the limitations of previous evaluations is the reliance on self-reported changes in behavior as a proxy for changes in risk HIV transmission. This study, in contrast, would have the opportunity to measure actual HIV incidence. The trial would also represent an opportunity to evaluate the impact of multi-component prevention programs on HCV transmission. Investigators could consider the feasibility of addressing this primary endpoint during the design phase of the trial.
Secondary outcomes might include subjective and objective measures of risk behavior, including drug-related behavior (such as self-reports of needle sharing and needle disinfection) and sexual behavior (such as self-reports of condom use). Secondary outcomes might also include potential harm at the individual and community level (such as an increase in discarded needles or recruitment of new users). Mediating variables (e.g., self-efficacy in not sharing, or knowledge about infectious disease transmission through sharing of cotton or water) could also be collected. Mediating and moderating variables are important for scrutinizing causal pathways, which distinguish successful interventions from less successful ones. A formative evaluation component could also provide information on the best implementation strategies for the prevention program. Data on program costs and cost-effectiveness could also be collected to help inform decisions on allocating resources.
DESIGN AND IMPLEMENTATION CHALLENGES
A trial of this size and complexity would present major operational and logistical challenges. For example, before the actual trial began, investigators would have to collect baseline measures of biological and behavioral outcomes at all study sites. Investigators would also have to conduct follow-up assessments periodically to measure outcomes. A study of this size would take several years and require extensive cooperation and monitoring among an array of organizations and stakeholders. However, while the proposed rigorous study would present significant challenges, that real-
ity does not obviate the need for it. Experience with the NIMH Collaborative HIV/STD Prevention Trial shows that a trial of this complexity is feasible (Personal communication, D. Celentano, co-principle investigator on NIMH trial, Johns Hopkins University, August 11, 2006).
In approaching such a multi-national and multi-faceted study, critical ethical issues must be anticipated and essential safeguards put in place. Such a trial could be designed in a highly ethical manner consistent with local values. If undertaken, such a study would require extensive and closely coordinated planning and scrupulous attention to the ethical underpinnings. While the details of establishing such assurances and safeguards depend on the specific protocol(s), the basic principles for protecting human rights would apply. Below we outline some of these parameters.
The principles and strategies for ethical safeguards are well identified and can ensure the rights of research participants, as guaranteed under the Declaration of Helsinki and elaborated in the Belmont Accords. All such research must safeguard against exploitation of individuals and communities. Individuals should be afforded the benefits of enhanced treatments (the principle of beneficence), and receive only complementary pre-specified approaches or combinations of approaches whose comparative advantages remain uncertain (the principle of equipoise). Participation in research must be fully informed and freely given according to international and local norms for the particular methodologies being employed, and without repercussion or coercion (respect for dignity and autonomy). Community access to participation should be widely available (distributive justice). For example, extensive promotion of the study would allow as many communities as possible to be considered for participation.
The evaluations would be most informative if conducted in high-risk countries that are initiating programs to protect against HIV among IDUs, given that the evidence base is largely drawn from developed countries. However, the contributions of various components of a comprehensive program are also unknown in countries with more extensive existing HIV prevention programs for IDUs. Thus, communities in these countries might also participate in the trial, furthering the distributive justice of the study.
As noted, a stepped-wedge design would guard against ethical concerns about withholding effective therapies because all communities would eventually receive the interventions. A “no treatment” or “minimal treatment” control arm would be inappropriate. Rather, control communities should have a substantial prevention program equaling or exceeding that already available in the study sites. For example, in evaluating the effectiveness of a needle and syringe exchange component, investigators might provide a basic package of services such as voluntary HIV counseling and testing to promote behavioral change, education on needle disinfection, and referrals to health services and drug treatment in both the control and experimental
communities. Needle and syringe exchange could then be added in the experimental communities.
A multi-national, multi-sector steering committee could help ensure the highest levels of scientific rigor and ethical protection for the trial. Full scientific vetting of protocol(s), informed consent provisions, data quality assurances, periodic review of progress, and transparency of results are requisite for such research integrity. An essential feature of the proposed trial is community involvement in its design and implementation, as well as the development of a community partnership to address concerns prior to and throughout the trial. Plans for sustainability and dissemination of effective interventions are also critical.
TENSION FOR POLICYMAKERS
Policymakers presented with the suggestion for further research confront an inherent tension. On the one hand, there is evidence that multi-component programs are effective in reducing drug-related risk behavior. As noted in Chapter 3, lack of complete or definitive evidence should not impede implementation of programs vital for protecting public health. On the other hand, this report presents a compelling argument for collecting more evidence, derived from high-risk countries themselves. The policymaking process must be understood as providing ever-greater clarification. A randomized community trial would assure policymakers who are initiating or extending HIV prevention programs that they will learn important lessons, and that they can apply those lessons in a timely and effective way to minimize expenditures and maximize the benefits in their countries.
Gambia Hepatitis Study Group. 1987. The Gambia Hepatitis Intervention Study. Cancer Research. 47:5782–5787.
Research Triangle Institute (RTI) International. 2005. NIMH Collaborative HIV/STD Prevention Trial Project Overview. [Online]. Available: http://www.rti.org/u10/overview.cfm [accessed August 22, 2006].