Research with vulnerable populations: considering the Bucharest Early Intervention Project (part 2).

In an earlier post, I looked at a research study by Nelson et al. [1] on how the cognitive development of young abandoned children in Romania was affected by being raised in institutional versus foster care conditions. Specifically, I examined the explanation the researchers gave to argue that their work was not only scientifically sound but also ethical.
In this post, I examine the accompanying policy forum article, Millum and Emmanuel, “The Ethics of International Research with Abandoned Children” [2]. Millum and Emanuel are in the Department of Bioethics at the Clinical Center of the National Institutes of Health. As such, it’s not unreasonable to assume that they are not coming to their understanding of this research — and to the question of whether it rises to the appropriate ethical level — from the point of view that good science should trump all other interests.

To lay the groundwork for their consideration of the Bucharest Early Intervention Project (BEIP), Millum and Emanuel first explain four ethical issues that are key in international biomedical research:
1. Standard of care.
Should the intervention being investigated in the study be compared to the local “standard of care,” or to the standard of care in the U.S. (or whatever relatively wealthy nation the scientists doing the study are from)? This was a big issue in research conducted in the 1990s in sub-Sahara Africa on AZT’s effect on maternal transmission of HIV. Whether “AIDS Clinical Trials Group protocol 076” — where a short-course AZT intervention was compared to a placebo — was ethical was discussed in the New England Journal of Medicine by Harold Varmus and David Satcher [3], who defended the protocol, and by Marcia Angell [4], who criticized its ethics (or at least the direction they seemed to be going). The argument boiled down to two central issues: Would be possible to evaluate the efficacy of short-course AZT intervention if the control group received the U.S. standard of care rather than a placebo? And, given the researchers’ ethical obligations toward their research subjects, could they knowingly expose those in the control group to what amounted to no treatment (the placebo) when the Declaration of Helsinki requires that those in the control group be given the best current treatment (even if that best treatment is not a common local treatment).
How should scientists resolve this apparent conflict between what is required by ethics and what is required to get meaningful data? (Don’t forget, you can’t ethically justify any research with human subjects that does not promise to generate reliable data.)
Millum and Emanuel offer the following response to questions around standard of care:

[I]t is permissible to offer research participants in developing countries less-effective interventions than those used in developed countries if doing so (i) is scientifically necessary to answer an important question; (ii) does not deny anyone treatment they would otherwise receive; and (iii) is intended to develop interventions that will benefit the developing country. [5]

The desire for clean data — even on an important scientific question — is not by itself enough here to warrant choosing a local standard of care for your control group that would not be permitted elsewhere.
2. Informed consent.
Humans being relevantly ethically different from beakers and reagents, we recognize that they should have a say in whether to participate in scientific research — and that they should be given sufficient information to make the best choice possible as to whether participation or non-participation is in their own best interests.
In most international clinical trials, though, it is not sufficient to hand subjects your standard multi-page “informed consent” form and get them to sign it, even if you’ve translated it into their language. (It’s not clear that these forms work terribly well to secure informed consent in the U.S., either.) Some research subjects may not read very well, or at all, and they may have a very different understanding of health and disease than does the western biomedical community.
Does this mean that divides of language, literacy, and culture make it impossible to secure meaningful consent from research subjects in the developing world. Some people think so, but others would label this judgment paternalistic. Millum and Emmanuel point out, “Available data do not show informed consent is invalid [in international biomedical research].” [6]
3. Ancillary care.
It is quite likely that the health care situation for research participants in the developing world is not as good as is the standard situation in many developed nations. Millum and Emmanuel write:

Although researchers in developing countries often feel obliged to provide other treatments that their subjects desperately need [above and beyond what is required for safety or scientific validity during the research], the nature and extent of these obligations are not well defined. The most complete account justifies ancillary care obligations because research participants entrust aspects of their health to investigators through the procedures they undergo. [7]

In some sense, we can cast ancillary care as a benefit to the subjects in exchange for the risks they undergo by participating in the research. Some have worried whether in desperate conditions such ancillary care can act almost to coerce people (who really need the care) to participate in a research study. Coercion undercuts freely given consent.
4. Posttrial benefits.
What happens when the study is over and the scientists leave town? Do the scientists have any further responsibilities toward the population from which they drew their research subjects?
Arguably, they do. Otherwise, we could be looking at a situation in which an impoverished population takes on the risks of testing new treatments whose benefits are then only made available to wealthier populations. That uneven distribution of risks and benefits is recognized as unjust, and scientists are obligated to minimize such injustices.
Given these four ethical issues, what do Millum and Emmanuel have to say about the BEIP study?

The initial reaction to the BEIP may be that the extreme vulnerability of abandoned, institutionalized children renders any research on them unethical. Not only are they unable to give informed consent, there is no clear guardian acting in their best interests. This puts them at greater risk of being selected for reasons of convenience rather than scientific necessity. While these are valid concerns, familiar safeguards can protect such children. People who cannot consent can be protected by enrolling them only in minimal-risk research, whose risks do not exceed those of everyday life. None of the study’s assessments of functioning were likely to harm the children. Restricting the participation of vulnerable groups, such as prisoners and the institutionalized, to research that addresses important questions relevant to their situation protects against unfair subject selection. The BEIP study aimed to produce results that would primarily benefit abandoned, institutionalized children. [8]

Vulnerable groups like children — especially really young children like the ones in the BEIP — can’t give informed consent themselves. Consent can be rendered by proxy, but there have to be safeguards to ensure that the kids’ best interests are protected. (Note that it is entirely appropriate to be paternalistic toward babies.)
Nelson et al. made efforts to ensure that risks were minimized for the kids in the research. And, the knowledge generated in this research was expected to help the situation of other abandoned kids in Romanian institutions — which is to say, posttrial, the benefits of the research would come right back to the community to which the scientists came to do the research.
Of course, some benefits from this knowledge might also come to children in the U.S., none of whom participated in the BEIP.
In Millum and Emanuel’s paper, standard of care and clinical equipoise are the ethical issues with the BEIP that seem to require the closest scrutiny:

The appropriate standard of care for a clinical trial depends on the research question being answered. Although the importance of equipoise, i.e., uncertainty among experts, is disputed, ethics requires at least that the research address an important question and be scientifically valid. Trials that do not meet these conditions lack social value. If a research study will not generate socially useful knowledge, it wastes resources and exposes participants to risks and burdens for no good reason.
The BEIP study addresses an important question; the welfare of institutionalized children depends on choosing correctly between further institutional care or switching to foster care. Prior data focused on adoption and did not directly address this comparison, had selection biases, and lacked a definitive randomized trial. Thus, the study appears to fulfill equipoise. Moreover, although both institutional and foster care can sometimes result in maltreatment, study participation was unlikely to cause net harm to the children; no child was put at additional risk to obtain the results, which reduces the ethical reasons for worrying about equipoise. [9]

The study was set up to generate a good scientific answer to an important question which the existing research had not answered, and the kids being studied were not exposed to significant risk of harm beyond what they were already experiencing before the researchers began the study.
Well and good, but we might worry about how existing inequalities in living conditions internationally might not become the cover for scientists to “outsource” their research, running trials in countries whose prevailing standard of care are close enough to no intervention (by our wealthy, industrialized lights) that the trials look an awful lot like placebo-controlled double-blind studies that can’t easily be conducted at home in the First World.
This is where the accounting for who bears the risks and who reaps the benefits comes in. In the case of the BEIP, Millum and Emanuel are persuaded that the immediate benefits of knowledge produced in the study will accrue to abandoned Romanian children:

[T]he instigation of the study by the Romanian Secretary of State for Child Protection and official reaction to its results indicate that the study had a high likelihood of having an impact on these children’s lives; state policy was likely to adopt the BEIP conclusions. This impact is not guaranteed: social and economic circumstances or government policies might change, resources may not be made available for foster care, or the conclusions of the research may be disputed. Certainty about implementation cannot be required to ethically proceed with a study. Instead, researchers must judge the likelihood that their work will generate health benefits, and proceed on the basis of its expected benefit. The expected benefits to Romania’s abandoned children appear to provide ample justification for the BEIP. [10]

Maybe, after asking for the knowledge the BEIP produced, the Romanian child protection officials might decide (for reasons of cost, perhaps) that they could not act in the way this knowledge would seem to dictate. But this mere possibility wouldn’t be reason enough not to do the research (and indeed, Romanian child welfare policies actually changed in response to the results of the BEIP).
But, while we’re looking at just distributions of risks and benefits, there is one group of Romanian kids who clearly cannot benefit from the BEIP findings: the kids in the study who weren’t moved to foster care:

[T]he children who remained in institutional care cannot now receive the benefit of early foster care, which the trial showed to be superior for some developmental outcomes.
This consideration does not make the BEIP study unethical… However, it does indicate that researchers need to pay special attention to how results get implemented when the benefits cannot accrue to participants. For instance, trial designs that move participants into the arm that is doing better during the course of the trial can be employed. According to the researchers, limited funds foreclosed this option. Alternatively, as done in the BEIP, researchers can present valid scientific results as soon as possible to those parties who can act on them. These parties have responsibilities to the participants, too; because the children were involved in research for the benefit of Romanian society, the representatives of society should ensure that the children get the care they deserve. [11]

This last point is a really powerful one. Nelson et al. take pains in their paper to demonstrate that they recognize their obligations to their research subjects, and that they are acting on those obligations. From their paper:

Because government-sponsored foster care was limited to about one family when our study commenced, we created our own foster care program. After extensive advertising followed by screening, we recruited 56 foster families into the project. [12]

In other words, the foster care system in Romania came into existence as a direct result of this study. Further:

[W]e were aware from the outset of the policy implications of our work, and as the study progressed we made our results available to government officials and child protection professionals. Indeed, several years after our study began, the Romanian government passed a law that prohibits institutionalizing children less than 2 years old, unless the child is severely handicapped. [13]

At least in part because of the study, the presumption in Romania is no longer in favor of institutionalization. But as many duties toward these children as the researchers take on (which is entirely appropriate, given the relationship between researcher and research subject), it is not up to the scientists to address every unmet obligation the larger society has to these kids. These kids did not pop into a laboratory ex nihilo. They were born into a society. The other people in that society — the adults, the people with political and economic power — had duties to these kids from the very start, and those duties don’t go away simply because researchers, by entering into a particular relationship with the kids, take on a special set of obligations to them.
We can argue that the scientists have an obligation to do what they can for the children, owing to the special duties in the researcher-research subject relationship. But so also does the Romanian government. So also do the Romanian people. Arguably, so might people and governments in wealthier nations have an obligation to do what they can for these kids.
And what we can do may be limited by our resources and our authority. It is heartening that Romanian child protection policies and practices have changed as a result of the findings of the BEIP, but Nelson et al. did not have the legal authority to abolish institutionalization of abandoned babies. That was something only Romanian lawmakers could do.
I think Nelson et al. make a good case (and Millum and Emanuel seem to support this judgment) that they are serious in their intentions to run an ethical study, and are making every effort to look out for their research subjects and for the larger population from which they were drawn. But it’s hard not to feel some unease nonetheless — especially about the kids in the institutional care group in the study, whose cognitive development may have taken a permanent hit.
Was there no way to get all of these kids out of institutional care and into foster care? Did we really need a controlled study to tell us that foster care was better?
This may just be one of those places where some people (including policy and law makers, child protection workers, economic planners, and even parents) think that its vital to establish that the “knowledge” driving our decisions is really knowledge (which, to the scientists, means connected to evidence of a certain sort in a specific way), and other people (including policy and law makers, child protection workers, economic planners, and even parents) think it’s always better to err on the side of caution even if this leaves us with more uncertainty in term of what we think we know.
On a visceral level, I understand the impulse not to take unnecessary risks — especially when those risks will be borne by little kids. But I also believe that we do better at assessing (and avoiding) the risks when we have better knowledge.
I don’t think this is a tension that goes away. Rather, I think this is another good reason for scientists and the larger society to talk to each other (and to listen to each other), especially when it comes to the details of how science might be used to find answers to questions of pressing societal importance.
——-
[1] Charles A. Nelson, III, Charles H. Zeanah, Nathan A. Fox, Peter J. Marshall, Anna T. Smyke, and Donald Guthrie, “Cognitive Recovery in Socially Deprived Young Children: The Bucharest Early Intervention Project,” Science 21 December 2007: Vol. 318. no. 5858, pp. 1937 – 1940.
[2] Joseph Millum and Ezekiel J. Emanuel, “The Ethics of International Research with Abandoned Children,” Science 21 December 2007: Vol. 318. no. 5858, pp. 1874 – 1875.
[3] Harold Varmus and David Satcher, “Ethical Complexities of Conducting Research in Developing Countries,” New England Journal of Medicine October 1997: Vol. 337. pp. 1003 – 1005.
[4] Marcia Angell, “The Ethics of Clinical Research in the Third World,” New England Journal of Medicine September 1997: Vol. 337. pp. 847 – 849.
[5] Millum and Emanuel, p. 1874.
[6] Millum and Emanuel, p. 1874.
[7] Millum and Emanuel, p. 1874.
[8] Millum and Emanuel, pp. 1874-1875.
[9] Millum and Emanuel, p. 1875.
[10] Millum and Emanuel, p. 1875.
[11] Millum and Emanuel, p. 1875.
[12] Nelson et al., p. 1938.
[13] Nelson et al., p. 1939.
Millum, J., Emanuel, E.J. (2007). ETHICS: The Ethics of International Research with Abandoned Children. Science, 318(5858), 1874-1875. DOI: 10.1126/science.1153822