Several prominent demographers and scientists have vigorously refuted Health Minister Manto Tshabalala-Msimang's claim that South Africa's HIV epidemic is declining and that the country "may be making some real progress in its response to the HIV epidemic".
Tshabalala-Msimang's statement was based on a national survey of HIV prevalence among pregnant women, which researchers are describing as deeply flawed.
The report, posted on the health department's website on 29 August, states that prevalence among pregnant women fell from 29.1 percent in 2006 to 28 percent in 2007.
Prof Rob Dorrington, a demographer at the University of Cape Town, and his colleague, Prof David Bourne, point out in a letter published in the South African Medical Journal (SAMJ) this week that the 2007 survey employed a methodology not only radically different from that used in 2006, but also "manifestly wrong".
The authors detected a problem when they noticed that changes in prevalence by age group did not tally with the change in overall prevalence, and that district figures were inconsistent with provincial estimates.
They deduced that in 2006 the results from district antenatal clinics were simply totalled to derive prevalence estimates for the country's nine provinces, but in 2007 the health department began weighting provincial figures according to age groups, based on general population estimates for age distribution.
Dorrington and Bourne describe the new methodology as "clearly problematic" because the age distribution of women attending antenatal clinics is very different from that of the female population as a whole.
"Since the prevalence of HIV also has a distinct age pattern, and prevalence is lower in the youngest and oldest age groups, using the population of all women to reweight the data will inevitably underestimate the prevalence of women attending public antenatal clinics," they wrote.
After recalculating the 2007 figures, using the same method applied to the 2006 data, the authors estimated HIV prevalence among pregnant women at 29.4 percent. Antenatal prevalence figures are used in combination with other surveys and mathematical models to determine HIV prevalence in the overall population, but the revised figure suggests that the number of South Africans living with HIV has probably not declined.
Dorrington and Bourne also recalculated the estimates for each of the provinces and found that the age weighting had given rise to "absurd results", particularly for the Western Cape, where prevalence dropped from 15.1 percent in 2006 to 12.6 percent in 2007, according to the survey.
In reality, only two districts in the province showed small declines and, after totalling data from all the districts, the authors estimated a prevalence rate of 15.3 percent.
The Treatment Action Campaign (TAC), a national AIDS lobby group, released a statement on Monday saying that the health department had "made a mistake or misled the public" by failing to state the change of methodology in the 2007 report.
Analysis of these kind of data appears to be becoming increasingly beyond the skills of the Department of Health
TAC spokesperson Lesley Odendal recommended that the health minister's claims of a drop in HIV prevalence "be treated with scepticism".
She also pointed out that even a slight increase in prevalence would not necessarily indicate that the HIV epidemic had worsened. "It has become extremely difficult to interpret the meaning of prevalence in recent years for two reasons: the HIV epidemic has matured into an AIDS epidemic [with more people dying], and antiretroviral treatment is helping people with HIV to live much longer."
In their letter to the SAMJ, Dorrington and Bourne note that the survey was "surprisingly silent" on the impact of life-prolonging antiretroviral (ARV) treatment on prevalence.
About 500,000 HIV-positive South Africans have benefited from the roll-out of ARV treatment, which could actually cause prevalence to increase, depending on mortality rates, but the 2007 survey revised the total number of people living with HIV down from 5.41 million in 2006 to 5.27 million in 2007.
Dorrington and Bourne conclude that "analysis of these data appears to be becoming increasingly beyond the skills of the Department of Health" and recommend that the government enlist the help of the broader scientific community to help interpret future figures on prevalence.
The Department of Health said it will release a statement responding to the criticisms.