In 2006, the Centers for Disease Control and Prevention issued guidelines advising broad HIV screening in the United States, even among groups with little apparent risk. The motivations behind these guidelines were clear: More than a million Americans are living with HIV infection; one in five don’t know it. This is a serious problem for individual patients, who are often diagnosed late in the course of their HIV disease, and a lost opportunity for prevention. Patients receiving effective HIV treatment generally reduce their risk behaviors and are less infectious because they have reduced viral load.
Economic studies find that it’s cost-effective to screen for HIV in general medical settings, even when such screening identifies only a few new cases for every thousand people tested. The advent of cheap and accurate rapid test technologies makes broad HIV screening an especially promising public policy.
It’s still a big challenge, financially and organizationally. The most difficult challenge is to work HIV screening within the everyday workflow and patient interactions in busy settings that serve millions of people who show up everyday for something other than an HIV test.
The traditional practice of requiring risk-reduction counseling for uninfected persons aggravates these challenges. Done right, patient-centered risk-reduction counseling often requires a 30-minute session. A trained counselor leads the patient through a discussion of his specific HIV risks, helping him to develop an achievable risk-reduction action plan.
In most settings, well over 99 percent of HIV tests yield negative results. So it’s a real burden to provide such counseling to every HIV-negative patient. Some facilities that might otherwise offer HIV testing refuse to do if they are also expected to provide risk-reduction counseling. If such counseling appreciably reduces HIV infection, it would be worth the additional costs and burdens. Public health practitioners deeply disagree on precisely this point, with little definitive evidence available to decide this point.
Lisa Metsch recently led a major randomized trial called Project AWARE that explored these questions. (Lisa and I are frequent collaborators, though not on this study.) She is the Stephen Smith professor and chair of the Department of Sociomedical Sciences at Columbia University’s Mailman School of Public Health.
This week, her team published their results in the Journal of the American Medical Association. They examined the impact of risk-reduction counseling within a sample of 5,012 patients who all received HIV tests. 2,505 patients received risk-reduction counseling along with their HIV test. The other 2,507 received only brief information regarding HIV and the rapid testing process. The research team then examined the rates of new sexually transmitted infections (STIs) in the two groups, to examine the effectiveness of counseling in reducing behavioral risk.
Their bottom line was clear: Evidence-based counseling doubled the cost of HIV testing. As shown below, the counseling conferred no additional benefit in preventing sexually transmitted infections.
These results are disappointing, but they are also clarifying for public policy. Indeed, this is precisely the kind of negative finding of greatest use for health policy. The AWARE trial cost $12 million. That’s an absolute bargain when it calls into question a widespread, burdensome and costly practice that apparently fails to improve health outcomes.
I caught up with Lisa by telephone. The below transcript has been edited for length and clarity.
Harold Pollack: Your team conducted a randomized trial comparing HIV risk behaviors in two groups, one provided HIV testing accompanied by risk-reduction behavioral counseling, and another provided the same testing without such counseling. Why was this an important comparison to make?
Lisa Metsch: In 2006, the Centers for Disease Control and Prevention released a new guideline statement for adult and adolescent HIV testing in health-care settings. In an effort to make HIV screening more routine in medical care, CDC made some important departures from prior recommendations. Before 2006, there was a much greater emphasis on testing individuals with specific risk factors.
Our work focused on one major change in what CDC was recommending. Voluntary HIV testing has traditionally included risk-reduction counseling. CDC’s 2006 statement recommended that risk reduction counseling was not going to be required at the time of HIV testing. This change was highly controversial. Many people in the field protested.
By 2006, there was increasing evidence supporting the benefits and cost-effectiveness of HIV testing, which can tangibly increase life expectancy for people living with HIV. CDC argued that the counseling takes time. Care sites might not have the resources to provide it. So requiring it would detract from the goal of getting as many people tested as possible.
We knew that about one in five people living with HIV did not know their status. If CDC wanted to make HIV testing more routine, the risk reduction counseling was seen as a real barrier. Moreover, there wasn’t a large literature to establish the actual effectiveness of such counseling in HIV prevention. One prominent 1999 meta-analysis basically said that counseling was not effective for HIV-negative persons. A major study called Project RESPECT found more optimistic results. This was published in JAMA 15 years ago. The RESPECT counseling study was done before HAART therapy, before rapid testing. It was based on data collected in the mid-'90s, when HIV was a much more immediately life-threatening disease, not the generally-chronic disease it usually is today. I and many others felt that the field needed a contemporary study to inform this issue.
HP: Just to summarize, tell me if I get this right. In 2006, the Centers for Disease Control issued a guidance statement which tried to make HIV screening more universal than it’s traditionally been, reaching populations of lower-risk people in all sorts of settings. They did so in light of evidence that such measures were cost-effective, and because of the importance of finding people living with HIV who don't know their status.
When testing is done in such low-risk populations, maybe 99.5 percent of people will test HIV-negative. If one realistically wants to test millions of people, the testing procedure itself must minimize the costs and burdens of conducting it in the real-world workflow of a hospital emergency department, substance abuse treatment facility, or wherever else you are. So if counseling for HIV-negative people isn’t especially valuable, requiring it is counter-productive. This simply makes testing too burdensome in many places.
LM: That’s right. And the strongest supporting evidence for such counseling at the time of testing was largely based on studies conducted in the 1990s, when HIV was a very different disease.
For all the reasons you mentioned, there was a compelling argument to move away from counseling. But before we jumped to conclusions, we give counseling its day in court by running a rigorous trial. We worked with the NIDA Clinical Trials Network from the National Institute on Drug Abuse and performed a randomized trial in twelve drug treatment programs throughout the U.S.
To simplify a bit without going into all the technical details, we offered one treatment group testing and evidence-based risk reduction counseling. For this group, we used the Project RESPECT model previously supported by the CDC. We then offered the other treatment group HIV testing followed by information only, as supported by the new (2006) CDC recommendations. We found no effect of risk reduction counseling at six months follow-up. That study had some key limitations, though. Our primary outcomes were based on patients’ self-reported risk behaviors. We also felt that we were not working with a group of individuals likely to face high HIV risks, as was done in the earlier work from the 1990s.
Just as we completed that study, along came the American Recovery and Reinvestment Act of 2009, the stimulus. I and my colleague Grant Colfax, who was at the San Francisco Department of Health (now director of the Office of National AIDS Policy) received a $12.3 million grant for what became the AWARE study.
We conducted a randomized trial of an evidence-based risk-reduction counseling model. Unlike in our previous work, we were able to have biological markers as the primary outcome: new cases of sexually transmitted infection. We also had a large enough sample — roughly 5,000 people, 2,500 in each of the two treatment arms — to examine the potential impact of such counseling within important subgroups. For example, we could examine whether counseling was more effective among men than among women. We could compare men having sex with men with other groups, different race/ethnic groups, different age groups.
HP: Can you say a little bit about who these 5,000 people were, and why they were in the study?
LM: We went to nine sexually transmitted disease clinics, mostly in places that have high rates of HIV infection: Miami, Los Angeles, San Francisco, Seattle, Washington D.C., and others. Those under the age of 18 and those self-reporting as having HIV were not eligible for the study. We recruited 5,000 study subjects and enrolled people in eight months. As I mentioned, this was done with stimulus funds. So we had a mandate to move quickly. Men comprised two-thirds of our sample. About 28 percent of our sample were men having sex with men.
(Note to readers: As I’ve written before in this space, MSM account for about 30,000 of the roughly 50,000 new HIV infections that occur in the U.S. every year. Understanding the impact of risk-reduction counseling in this groups is therefore fundamental — HAP.)
HP: Let me ask an embarrassingly basic question. HIV is a rare disease. Even among 5,000 people, you would expect maybe 12 people to become HIV-infected over a six-month period. How can you establish whether or not counseling is effective if HIV is such a rare outcome?
LM: That’s right. And that’s why our outcome was new sexually transmitted infections of all kinds. We would have needed a prohibitively large sample to see any plausible differences in HIV infections themselves. So instead, we used STIs as a proxy, a “surrogate marker” for HIV. That’s certainly not a perfect approach. It has some limitations. Yet we used the same outcome that was used in the earlier studies I mentioned, which faced the same difficulty. If people are acquiring new sexually transmitted infections, they are continuing to have unprotected sex.
HP: I should mention to readers that about 12 percent of this sample contracted some sort of STI over the study period. (See the accompanying graphic.) In contrast, less than 0.5 percent contracted HIV. Even a $12 million study in a pretty high-risk population requires these surrogate measures to understand what’s happening with the highest risk behaviors.
Another embarrassingly simple question: Wouldn't it be cheaper to just ask people how many sex partners they have, whether they consistently use condoms, and so on?
LM: We did ask these questions, and the results are reported in our paper. But when you want to know whether counseling actually reduces disease incidence, the biological markers provided much stronger evidence. These were the markers that were used in earlier studies. Hopefully this kind of evidence will influence policy makers, public health practitioners and others.
HP: Reading the paper, counseling seems to have a protective effect if you look at some of the reported risk behaviors. Yet these protective effects don’t carry over, when one looks past self-reports to the number of participants who were actually found to contract gonorrhea or some other STI.
The pattern raises an obvious possibility. When you do a counseling intervention, you clearly teach people one thing: What you, the researcher, really hope to hear. When we do studies that ask people to self-report regarding various stigmatized or high-risk behaviors, I wonder how often we contaminate our own results. We telegraph to people what we want to hear. Then we come back six months later to ask the predictable questions.
LM: Right. That’s why the best studies collect both kinds of data. Our a priori primary study outcome was a composite biological measure of new sexually transmitted infections. This is less subject to such bias.
HP: If I am a policymaker and I am reading this, what should I draw from your study?
LM: Our data clearly showed that there was not a beneficial effect of providing counseling at the time of HIV testing. Such counseling was more expensive as well. In an era of limited resources, I conclude that there are better ways to spend our precious HIV prevention funds. If we implement a more focused approach, we can probably test more people every day in real-world settings, and use our prevention resources more efficiently.
I should specify that I'm talking about counseling for people who test HIV-negative. Post-test counseling for those who test HIV-positive remains essential. Such counseling must provide and ensure follow-up medical care, address psychosocial needs, and provide other supports.
HP: Two things struck me about your last comments. First, money is also only one issue in-play. When you add risk-reduction counseling, testing becomes a much bigger burden for managers and front-line professionals who operate social service, medical or public health interventions. HIV testing also becomes more of a burden for patients. And don't forget, these are not people who showed up because they want an HIV test. They presented for care for something else they care more about. If there is little public health benefit, it’s self-defeating to promote a more elaborate process that’s more difficult for everybody.
LM: I agree, as does an accompanying editorial to our JAMA piece.
HP: The other irony concerns the small amount of money at stake. Without counseling, HIV testing costs something like $28 per patient. Add the counseling, and this rises to about $56 for every person tested. It’s symptomatic of public health’s financial standing that $56 is too much in some settings. Given all the wasteful items in our $2.8 trillion health-care system, that $28 difference shouldn’t decide things, but it does.
I should also mention another take-away lesson from your study. Without a huge amount of evidence, public health researchers and practitioners hold strong views on both sides of the counseling debate. We must hold ourselves accountable to rigorous experimental trials to find out how effective risk-reduction counseling actually is. My guess is that you studied above-average interventions provided under well-supported, well-resourced conditions. You still found almost no clinical benefit. That’s disappointing and surprising, but that’s why the rigorous work needs to be done.