Aids figures may be too optimistic

Figures for HIV infection collected from the annual antenatal survey are the most accurate forecast of Aids deaths.

DR MALEGAPURU MAKGOBA, president of the Medical Research Council, stresses the need for more statistics about the incidence of HIV/Aids, so that South African policymakers can have the same standard of information at their fingertips as the Centers for Disease Control (CDC) in Atlanta provides for America. What a pity then that the health department's expert epidemiology unit was closed down in 1996, as part of the overall restructuring of the department, and its vital work of tracking infectious disease outsourced instead.

For 20 years the epidemiology unit was run by CDC-trained Dr Horst Kustner, a former mission hospital doctor. It is due to his efforts that the country has a reliable picture of the growth pattern of the epidemic and can forecast the number of people who are likely to die of Aids. That picture is drawn from the annual survey of HIV infection among women attending public antenatal clinics, the project that Kustner launched in 1990.

Those who lament the "window of opportunity" that was missed in the early 1990s may be surprised that the health department established an Aids advisory group as far back as 1983 and an Aids control group a few years later. Kustner recalls, "For four years we were looking at a disease that affected only white homosexual males and in very small numbers," he says, "with the occasional case resulting from a blood transfusion or a drug addict using an infected needle. But in 1987 we found the first two or three cases among women. Optimists on the advisory group believed that it would just go away, but then we began finding cases among Malawians working on the mines and among prostitutes."

As soon as a blood test for HIV became available, Kustner argued long and hard for the introduction of an annual antenatal survey as the most reliable and cost-effective way of tracking the disease. Pregnant women are sexually active and more likely than other groups to be representative of the general population. Since blood samples are taken routinely from pregnant women to test for a number of infections and medical conditions, it is not too difficult to add one more. However, it was several years before he succeeded in convincing the department and the epidemiology unit was given the go-ahead to start the surveys.

Kustner explains that they decided to follow the British recommendation of unlinked, anonymous, confidential HIV-testing. Using this design there is no way of identifying whose blood is tested. The district laboratories are simply requested to test a certain number of blood samples over a time period and, after rendering them anonymous, to report the results. No individual can be told her test results or even know if she is included in the sample. Also in accordance with British practice, women were not asked whether they wished to take part or not. This, says Kustner, is important, because research shows that where testing for HIV is voluntary, the results may significantly underestimate the incidence of the disease.

The survey is still undertaken every October - now outsourced through the Medical Research Council - but Kustner points out that the original sampling design has been changed. It remains unlinked and anonymous but, following new ethical guidelines, the women are now asked whether they wish to take part in the survey or not. This means, he says, that the women tested are in effect volunteers. While they might indeed be randomly chosen, they are certainly no longer randomly included in the sample tested. This is the crucial difference. "Bias may have crept in. I wouldn't be at all surprised if the current drop in increase rates in some provinces is due to this," he says and recommends reverting immediately to the original methodology. The latest figures, from the tenth antenatal survey in 1999, are nevertheless the best available for assessing the progress of the disease (map 1). They show that in KwaZulu-Natal infection rates have reached a plateau at 33 per cent. Although some optimism about that levelling off was expressed in the press reports earlier this year, according to the specialists it is no surprise at all. Dr Brian Williams, the epidemiologist who runs the Carletonville Aids project and who has prepared the maps and graphs on this page, comments: "For a number of reasons the average cannot go much higher than 35 per cent, though specific groups may show higher levels."

The big question, says Williams, is will the other provinces inexorably follow the same path as KwaZulu-Natal or will they level off at lower rates, as happened in Thailand. He points out that in 1990 Thailand had the same HIV prevalence among pregnant women as South Africa (graph 1), but determined government action successfully turned the tide. The 1999 antenatal survey talks cautiously of the epidemic having entered a phase of stabilisation. But, says Williams, data from several more years will be necessary to confirm this trend.

For 4.2 million South Africans, of course, it is already too late. The numbers of Aids-related deaths a year that will follow from the present rates of HIV infection are shown in graph 2. Even in the lower prevalence provinces there are hotspots of infection in the densely populated metropolitan centres (map 2) and the mining districts. According to Williams previously unimaginable infection rates have been found among the general population of a West Rand mining town, where 60 per cent of 25-year-old women are HIV-positive.

The report of the 1999 antenatal survey recommends a successor programme that would include incidence testing to count new cases of infection. But what will the data be used for? Epidemiology is a practical discipline, says Kustner. "Practical people argue that it is not more and better data this country needs, but more appropriate responses. The data cannot save lives, only the responses to the data can do that. And it is not only the government that has to respond, but society at large, through drastic changes in behaviour."