Randomized controlled trials (RCTs) are the gold standard for evidence, but they are very costly and time consuming. And because the populations enrolled in RCTs must be carefully specified and the settings in which the studies are conducted are themselves not typical, it is often difficult to extrapolate results from RCTs to ordinary practice situations.
So what can we do to get more information about the relative quality and cost of different health interventions?
Imagine the following situation. A and B are widely used treatments for a common condition. While clinicians and patients may have preferences for one or the other, by Freedman’s well-known criteria a state of clinical equipoise exists with regard to A and B – i.e.,- “there is no consensus within the expert clinical community about the comparative merits of [A versus B].”
In order to gain information about the relative effectiveness of A and B several health plans agree to participate in a Cluster Randomized Trial. Some health plans will favor A; the others will favor B. “Favor” means that unless clinicians have specific reasons for choosing the non-preferred agent they will use the preferred one. The health plan data bases will be used to follow patterns of side effects, medication changes and clinical outcomes.
This intervention does not require all individuals to receive the treatment to which their group is assigned, only that the fractions treated with A and B in the two groups differ by a substantial amount. Such separation allows estimation of population-level difference in outcomes, which can inform overall treatment strategies (“in general A and B yield similar outcomes,” or “in general people treated with A fare better”).
I recently joined with a group of colleagues to examine the ethics of this kind of drug trial. We (a) interviewed health plan patients, their physicians, and health plan administrators, (b) discussed CRTs with the Harvard Pilgrim Health Care Ethics Advisory Group, and (c) systematically reviewed the relevant ethics literature. We published our conclusions in a Hastings Center Report article - "Comparing Drug Effectiveness at Health Plans: The Ethics of Cluster Randomized Trials." We focused on two large questions - (a) is individual informed consent required for participation in the kind of trial we envisioned and (b) if it is not required, what takes its place?
We concluded that individual informed consent is not required for participation in CRTs of widely used standard treatments that are in a state of equipoise. If a physician believes that the non-preferred agent is indicated she can prescribe it. If a patient wants the non-preferred agent she can receive it. Practicing physicians are familiar with the state of “internal equipoise” in which they would be equally comfortable recommending one agent or another. A CRT of the kind we envisioned simply asks the physician to prescribe the preferred agent unless she has a specific reason for prescribing the alternative. There is nothing in this chain of steps that is inherently different from ordinary practice. With regard to the prescribing of A or B there is no “experiment” being done and nothing requiring informed consent beyond what is ordinarily required in clinical practice.
Not surprisingly, the U.K., which insures the entire population through its National Health Service, and which faces explicit trade-off questions on a daily basis, has paid more attention to the ethics of comparative effectiveness research than we have in the U.S. In 2002 the Medical Research Council published "Cluster Randomized Trials: Methodological and Ethical Considerations. "Here's the essence of what it had to say about informed consent:
An individual body, or mechanism that can represent the interests of the cluster is required – for convenience this will be labeled the “cluster representation mechanism” (CRM). The appropriate nature of the CRM will vary depending on the circumstances – both the nature of the cluster and the nature of the intervention. Thus, agreement to fluoridate the water supply might be obtained by plebiscite, while GPs might agree to the distribution of an information leaflet to people in their waiting rooms...The CRM must produce a formal document for the cluster that certifies and sets out its ability to [consent] (sufficient knowledge of the circumstances, beliefs, and values of members of the cluster, any delegated authority from/for the cluster, lack of conflicts of interest)...The CRM has essentially the same rights as a patient in an individually randomized trial --- including the absolute right to withdraw the cluster, without adverse impact on the cluster, if it decided that the study was not now in the interests of the community.We agreed with this analysis.
Our interviews with patients, however, showed that some were perplexed by the idea of clinical equipoise. These patients believed that their doctors knew what was best for them. The idea that large swaths of medical practice do not have an evidence base showing that alternative A is better than B was foreign to them.
Conducting comparative effectiveness CRTs of the kind we envision would involve explicit acknowledgement of the degree to which important areas of medical practice are not evidence based. While we did not see individual informed consent as required for conducting this kind of CRT, we recognized that participation in research is not a typical health care activity. Just as teaching hospitals are expected to make clear that they are replete with students and to explain why this is so and what it might mean to patients, we believe that health plans and medical organizations should do the same with regard to participation in research - even research that does not require individual informed consent.
Information about comparative effectiveness is a necessary condition for getting a better grip on our health care system, but it is not sufficient. The more difficult step will be having the guts to use the information we develop.
In addition, those who lead the effort to strengthen comparative effectiveness research will have to educate the public about why the research is so important. The U.S. public has a strong faith in medical science and a rapacious appetite for medical miracles. Transforming our delivery systems into true learning organizations will require open recognition of medical uncertainty in ways that are not often acknowledged by physicians or understood by patients. But using everyday medical practice as a source of systematic learning about treatment effectiveness would be a salutary change in the culture of medical care.