Almost on cue, this week's Nature has published some words that begin to make sense about the "epidemic of AIDS" in Africa. Coming on the heels of the JAMA/Lancet backsliding on HAART's efficacy and the usefullness of viral load measurements in deciding the appropriate time and type of clinical intervention, middle and low level AIDS, Inc. spin doctors, who have been kept extra busy the past weeks, will have even more spinning to do. It has already started, as is clear from the internal bandages in the article itself (reprinted below).
Michael Peterelis, who first brought our attention to the publication, has an excellent report of this item in which he places it in the context of previous public admissions of gross error in estimates.
Published online: 30 October 2006;
New methods suggest AIDS toll lower than estimated
By Erika Check
There is no doubt that the number of HIV infections worldwide is still on the rise: the toll is up to 40 million, according to UNAIDS. But are scientists using the right method to count the cases?
This year's UNAIDS estimates of the epidemic are lower than those released at the end of last year for 11 southern African countries and for China. And in August, Indian epidemiologists cast doubt on their country's high estimate of HIV cases. [...]
Based on sentinel surveillance, UNAIDS estimates that India has 5.7 million HIV-infected individuals—slightly more than the government estimate of 5.2 million—earning India the unhappy distinction of being the country with the most HIV infections.
But when researchers led by Lalit Dandona, an epidemiologist at the Administrative Staff College of India, ran a population-based survey, their data suggested that the real number could be as low as 3.5 million.
Dandona's group gave HIV tests to as many people as would consent to the study in one district of Andhra Pradesh state. Based on their estimate, only 47,000 people in the district are HIV-positive, less than half of the official estimate of 113,000 people. The team then calculated an adjustment factor for this bias from their data, then applied it to data from other states, arriving at their substantially lower number for the country.
"We were very surprised," says Dandona. "Before we released this to the public, we spent a lot of time making sure we were not underestimating."
Other epidemiologists say that they're not surprised the official estimates are off.
"Many of us in the field have suspected that the standard methods of estimation have resulted in overestimates," says Willi McFarland, director of HIV/AIDS Statistics and Epidemiology for the San Francisco Department of Public Health.
Sentinel surveys are great for tracking trends—to pinpoint sites where infections are rising, for instance—but they're not ideal for predicting absolute numbers, he says.
One explanation for India's overestimate may be that the poor are more likely to be infected than are the wealthy, and a higher proportion of the poor go to public clinics.
Officials are still mulling over the implications of the study. UNAIDS expert Peter Ghys says population-based surveys in other states indicate that the birth clinic HIV rates accurately reflect HIV infections in the general population.
India's National AIDS Control Organisation adds that it is reviewing the methods but is not likely to conduct large population-based surveys because they are too expensive.