Will a New NIH-Funded Study Tell Us Whether Immediate AIDS Treatment Slows HIV Transmission?

August 4, 2009

Each year in Sub-Saharan Africa there are about 2-million new cases of HIV infection, most of whom would not need antiretroviral therapy (ART) under current guidelines for 8 to 12 more years. Since donors have not managed to place on treatment more than about half of those needing it each year, the 8 to 12 year lag between infection and need for treatment has been seen as a breathing space. But a radical new idea discussed at last week’s Conference on HIV Pathogenesis, Treatment and Prevention in Cape Town, South Africa would start the 2-million newly infected people on ART within days or weeks of their infection.

Would immediate treatment work as a prevention strategy? At the IAS meeting, Dr. Anthony Fauci, Director of the National Institute of Allergy and Infectious Disease, told the South African Times that “[w]e will be doing research to determine the feasibility of several assumptions underlying [the immediate treatment strategy].” Now the University of North Carolina’s Institute for Global Health and Infectious Disease has announced that researchers William Miller and Audrey Pettifor have been awarded a $3.5 million, four-year research grant to test some of the assumptions flagged in Dr. Fauci’s conference talk.

The study will recruit 30 newly infected people for each of three comparison groups

As Pettifor explains on a local radio broadcast (helpfully flagged by the Kaiser Foundation’s Global Health News), the study will find and follow 90 HIV-infected Malawians who are in a 12 week window immediately after infection. The subjects will be randomly allocated to three groups or “arms” of the trial. The “control” arm will receive one session of standard post-test counseling, which includes advice on how to protect their partners from HIV infection. Patients in a second arm will also receive a more intense HIV prevention intervention, which includes five sessions of risk-reduction counseling. The third group of patients will receive not only the post-test counseling and the additional five sessions of prevention counseling, but also will immediately begin antiretroviral treatment. The question is whether the group beginning treatment transmits fewer HIV infections than either of the other two groups.

90 subjects seems like an awfully small sample

Doesn’t it seem that 90 subjects is an awfully small sample, especially when split three ways? While it is risky to try to second-guess the study’s peer reviewers without seeing the proposal they funded, one can raise some basic questions about the statistical challenge confronting the researchers based on the little that has so far been revealed.

Assume that even with post-test counseling the rate of new infections among the partners of high-risk adults in Malawi is about one per-cent per year. Then in order to justify enthusiasm for universal continent-wide HIV testing followed by immediate ART, we would want the infection rate among the partners of those getting ART to be reduced by at least half, say to half a percent per year. So the researchers would need a sample size large enough to reliably detect a difference of one half of a percent. How large would such a sample have to be?

Let’s say that we would want to be 90 percent sure that immediate treatment constitutes more effective prevention than either of the two alternatives before investing HIV prevention resources in this strategy. In the language of statisticians, this means we want the study to have a “statistical power” of 90 percent. If you have a copy of Stata, you can calculate the sample sizes necessary to have a 90% chance of finding a half a percentage point difference by just typing in the following command:

sampsi 0.005 0.01

The answer is that the immediate treatment group and each of the other groups of patients must have 6,650 HIV-negative partners. Since the adult infection rate in Malawi is about 25%, the total number of partners for all three groups would have to be at least: (3 * 6,650)/( 1 – 0.25) = 26,600. That’s a lot of people whom the researchers must convince to be tested after sex with their patients. And those who are HIV negative must consent to testing regularly for up to a year after that.

The sample size requirements are even worse if the intensive counseling arm works as one might hope it would. Suppose the control arm partners have an undiminished rate of new infections of one percent, and the intensive counseling arm has a reduced infection rate of only a half of a percent. In this case, to prove the superior effectiveness of ART, the researchers will have to show that the clients of the “counseling-plus-ART” group have an even smaller rate of new infections, say of a quarter of a percentage point per year. The sample size calculation in Stata can be done with the command, sampsi .005 0.0025, and yields a requirement of 13,348 subjects in each of the second and third arm of the study. This consideration increases to 45,000 the number of sexual partners who would have to be tested and followed ( = (13,348*2 + 6,650)/(1 – 0.25) ). Divided by 90 patients, the authors seem to be assuming that each patient will have on average at least 494 unique partners per year who (a) have no other infected partners and (b) return for testing for a full year after enrollment. These last two requirements might double or even triple the number of partners to be studied per patient, to 1,000 or 1,400 for each of the 90 recruited patients in the study.

By following the patients’ partners over more than one year, the sample size requirements can be reduced, but attrition problems are likely to increase. From where I sit, it seems that the chances of success would have been improved by aiming for perhaps five times as many newly HIV-infected subjects in order to reduce the number of sexual partners per subject to the more easily manageable range of 200 to 300 per subject.

It will be interesting to learn how the researchers have in fact designed their study, what the logic was of their design choices – and how successful they are at recruiting and following the number of patients and partners that they will need. But even if their work and that of other researchers supports the feasibility and effectiveness of immediate ART for prevention on a small scale, the intervention must meet two other tests proposed by Dr. Fauci: (1) feasibility of scale-up and (2) cost-effectiveness relative to alternative ways to prevent HIV infection or prolong life.

Given the four-year expected duration of the study and the difficulties I have suggested above, it seems to me that immediate treatment is years away from inclusion among the most promising tools for HIV prevention.

You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.

7 Responses to “Will a New NIH-Funded Study Tell Us Whether Immediate AIDS Treatment Slows HIV Transmission?”

Kim Yi Dionne Says:
August 4th, 2009 at 6:01 pm

Just to clarify, the adult HIV prevalence rate in Malawi is actually 11.7%, according to the Sentinel Surveillance Results from 2007 presented at the 2008 National AIDS Commission Research Dissemination Conference. Of course, the resulting math would only reduce the sample sizes you predict by about 4,000 and 8,000 subjects, respectively, therefore requiring an assumption that each patient will have on average roughly 400 unique partners per year.

I can’t believe that Fauci et al would have overlooked the statistical power problem… and am wondering why the two treatments are necessary (why not just post-test counseling vs. post-test counseling + ART?).
Nandini Oomman Says:
August 5th, 2009 at 4:43 pm

This idea–immediate treatment as a prevention strategy–doesn’t make sense to me from a policy point of view (apart from the study design flaws that won’t lead to immediate answers anyway!. Where are the resources to put 2 million newly infected people every year (although the number should decrease if the intervention is working!)on ART within days or weeks of infection? And how on earth will one identify them? Getting people to test for HIV is a continuing challenge, so how do we propose finding these new infections and treating them within a time bound window? Also seems a bit nonsensical to me to think about another resource intensive treatment program when the resource pie for GH is already being cut in to slivers rather than good, chunky pieces to support programs that work. I guess we are looking for a magic bullet to prevent HIV transmission, but for now there is nothing real or feasible about this idea.
Kim Yi Dionne Says:
August 5th, 2009 at 7:51 pm

There was a lot of talk before that there just weren’t resources for scaling up testing and treatment in Africa — but we see that has changed. We are also seeing a lot of mathematical modeling suggesting that immediate treatment could be an effective prevention strategy. Given these, I actually very much like the idea of running a field experiment to see whether there will indeed be an impact on incidence. Unfortunately, this is a very expensive experiment and as we’ve already said, the design has some incredible flaws. So, based on the cost-effectiveness of running the experiment (as well as the cost-effectiveness of treating so many people), I would prefer money spent on some rigorous/experimental testing of behavior change interventions.
Elwyn Chomba Says:
August 7th, 2009 at 8:04 am

How long would the newly infected be on ARVs is it for life?It is time that research was focussed on suscitanable cost effective preventable stategies.How does one screen for new infections at poulation level?We are still grabbling to see whether the PMCT srategy has been effevtive especially in sub Saharan Region.
Audrey Pettifor Says:
August 7th, 2009 at 5:48 pm

Thank you for the post. As co-PI of the study we would like to clarify a few key points that were not covered in the press release. Importantly this study is a pilot. This grant was funded in response to an RFA for interventions that examining the potential of behavioral plus biomedical interventions (a number of studies have been funded by NIH to examine this idea). Our outcomes are not transmission, but rather an assessment of the potential effects of behavioral and treatment interventions–we will assess the feasibility of transmission events as an outcome. Our goal is to obtain sufficient information to be able to use mathematical modeling to assess whether a large trial would be warranted–investigators at Imperial College are leading this effort. The idea was to generate PILOT data for the mathematical modeling (thus the small sample size!).

Another important component of the grant is validating rapid tests to find people with HIV infection earlier- a major limitation right now is that NAAT testing has to be done, often with pooling specimens which is very time consuming and expensive. We are validating some newer rapid tests which may help identify people with AHI.

Understanding the potential role of acute infection in the spread of HIV infection and the potential effect of behavioral and biological interventions during this time period is key to determining whether larger scale efforts are warranted.
Mead Says:
August 11th, 2009 at 12:23 pm

Thanks, Audrey, for clarifying the purposes of this interesting research project. Practical strategies to implement prevention among those with acute HIV infection (AHI) would be a great contribution to slowing the epidemic. Readers might be curious about NAAT, which is “nucleic acid amplification testing”. The 2008 report from the “WHO Working Group on HIV Incidence Assays” discusses the difficulties with NAAT testing that Audrey alludes to. Readers can find the report here:
http://www.who.int/diagnostics_laboratory/links/hiviwg_geneva_01_08_report.pdf
Marco Gomes Says:
August 11th, 2009 at 8:12 pm

Not to say that health education and condom promotion should be forgotten, nor to deny that the social and economic determinants of the HIV epidemic (such as poverty, migrant labor, urbanization and gender inequality) should be addressed. But in the short to medium term, the results of the study should provide a message of hope for those dedicated to controlling this devastating epidemic.

The epidemic of infection with the human immunodeficiency virus (HIV) and the acquired immunodeficiency syndrome (AIDS) poses a major ethical question: How can we control the epidemic and the harm that it causes without unjustly discriminating against particular social groups and without unnecessarily infringing on the freedom of individuals? This question pertains to three spheres of public policy in the United States: public health, the delivery of health care, and research. In the public health sphere, vigorous educational efforts will be required, as will modified approaches to intravenous drug use, prostitution, and homosexual and bisexual sexual activity. Carefully targeted, voluntary testing and screening programs should be coupled with counseling and with guarantees of confidentiality and nondiscrimination where these are appropriate. Both health care workers and the health care system have a moral obligation to provide care to people with HIV infection, but heroic self-sacrifice should not be required provided that infection control precautions are observed. Patients with neurological involvement and terminally ill patients will benefit from statutes allowing recognition of advance directives about preferred modes of care or nontreatment. There is a moral imperative to perform intensive research directed toward the understanding, treatment, and prevention of HIV infection and AIDS. The research process will raise challenging ethical questions.