Friday, October 14, 2011

Did Gates Foundation's Avahan Project Really Avert 100,000 HIV Infections?

The Lancet's article 'Assessment of population-level effect of Avahan, an HIV-prevention initiative in India' makes interesting reading, not least because most HIV prevention 'initiatives' appear to be assessed in a cursory manner, as if all effort goes into finding good news and burying bad news. The researchers go to great lengths to present figures and analyses that would allow anyone to judge the merits of Avahan themselves.

The paper is not especially accessible to the layperson, nor is this the responsibility of the authors. Making it so accessible is a job for the scientific media. But they seem shy of doing any more than repeating the publicity mantra about the Gates Foundation's work averting 100,000 new HIV infections over a period of five years in a population of 300-330 million. It's "all thanks to a Bill & Melinda Gates Foundation...project, called Avahan", says the Times of India.

The BBC doesn't gush so much, but nor does it give the reader any reason to question the headline figure, suggesting that The Lancet study confirms that the investment (of $258m according to BBC and The Lancet, $338m according to Times of India) paid off. They cite The Lancet paper's claim that Avahan "had a beneficial effect" and that the findings support "investment in well-managed HIV prevention programmes", neither of which are terribly surprising.

But behind the hype, and even The Lancet article seems to accentuate the positive, what the researchers found was a rather small effect with a very large margin of error. While you would expect a set of interventions to have some effect, as opposed to none at all, this project involved a very large investment over a relatively long period of time.

There are two reservations I have about this study, the first being my overall reservation about HIV prevention interventions; HIV is not all, perhaps not even mainly, sexually transmitted. While this program also targeted risk groups who face non sexual risks, such as intravenous drug users, author Lalit Dandona tells the Times of India "HIV mainly spreads in India through sexual intercourse".

My second reservation is the contention that "interventions like safe-sex counselling by peers, treatment for sexually transmitted infections, distribution of free condoms and needle and syringe exchange programmes among the most-at-risk population prevented the virus from spreading among the general population" (my emphasis); I don't believe that HIV spreads, to any substantial extent, from 'high risk' populations to the general population.

Consider the high prevalence groups in African countries, such as Kenya. They are not 'high risk' populations in the sense HIV researchers use, they are not primarily sex workers, men who have sex with men, intravenous drug users or even truckers. The mystery UNAIDS doesn't like to acknowledge is how people who don't have 'high risk' behaviors can have such high HIV prevalence, considering how slowly HIV spreads through penile-vaginal sex.

High prevalence groups, such as members of the Luo tribe in Nyanza province, do not appear to 'spread' HIV throughout the country, not even to the neighboring people of the Kuria and Kisii tribes or those of the Luhya tribe in the province just North of Nyanza, where HIV prevalence is far lower. The population of Nairobi and other high population density areas do not 'spread' HIV to low population density areas.

While prevalence among Luhya overall is far lower than that among the Luo, prevalence among Luhya women is far higher than it is among men. As a high prevalence group, Luhya women don't even appear to 'spread' HIV among Luhya men very efficiently. There are 10 infected Luhya women for every 1.7 infected men. Many high HIV prevalence groups do not have identifiably high sexual risk behaviors.

Treatment of sexually transmitted infections, while important in its own right, has not been shown to have much impact on HIV transmission. Similar remarks apply to distribution of free condoms; condoms do protect against HIV transmission, but only sexual transmission. And needle exchange programs do reduce HIV transmission among intravenous drug users (though perhaps not as effectively as is sometimes claimed). But these 'high risk' groups have never been demonstrated to 'spread' HIV among the general population.

As the opening sentence of The Lancet paper says "The aim of Avahan...was to reduce HIV transmission in the general population through large-scale prevention interventions focused on high-risk groups." Sex workers, their clients and their clients' partners, intravenous drug users and their partners all together don't account for a large proportion of HIV transmissions in countries where Modes of Transmission Surveys have been carried out.

While much has been made by the media about men who have sex with men also having sex with women, it has not been shown that this plays a significant role in 'spreading' HIV from this group to the general population. The number of men who have sex with men also having sex with women is unlikely to be very high, so the number of women involved is also going to be low. This phenomenon is probably far more significant in the minds of journalists (and possibly researchers) than it is in the real world.

The research does not show any mechanism by which HIV prevalence somehow radiates out from 'high risk' groups to the general population through sexual behavior. Nor does it shed any light on how people who only face low sexual risk levels are infected at higher rates than could be expected. It would be disturbing if the several thousand dollars per infection 'averted' failed to avert any infections, but it is unlikely that the number is anywhere near 100,000.

With HIV, it appears that 'high risk' does not always result in high prevalence, and high prevalence has not always been explained by high risk levels. This is a serious anomaly and HIV research that fails to account for it will be flawed at best, totally invalid at worst.


No comments: