When it comes to sub-Saharan Africa's devastating AIDS crisis, there is an understandable tendency to latch onto any scrap of good news.

Figures suggesting the epidemic is waning in some countries are being trumpeted by governments and international donor agencies as evidence that their prevention efforts are succeeding.

Kenya's National AIDS Control Council recently ascribed a small drop in the country's HIV infection rate to people absorbing the messages in awareness campaigns and changing their behaviour accordingly.

South Africa's health minister, Manto Tshabalala-Msimang, claimed that the first evidence of declining HIV prevalence in pregnant women - from 30.2 percent in 2005 to 29.1 percent in the latest survey - was mainly due to "our continued focus on prevention as the mainstay of our response to combat HIV".

But the real story behind increases and decreases in HIV prevalence is far less clear. "There's an awful lot of vested interests, but it's sufficiently murky that no one really knows what's going on," Prof John Hargrove, director of the Centre of Excellence in Epidemiological Modelling and Analysis (SACEMA) at the University of Stellenbosch, South Africa, told IRIN/PlusNews.

Twenty-five years is not long to get to grips with an epidemic that has evolved very differently in different parts of the world: in Europe, North America and Asia it has largely been confined to high-risk groups like injecting drug users, sex workers and men who have sex with men; in southern Africa it has spread rapidly via heterosexual networks.

Although theories abound, "nobody really knows why southern Africa is worst affected", said Dr Brian Williams, another epidemiologist at SACEMA. "And if we don't know that, it's very difficult to explain why prevalence is going up or down."

Lack of reliable data

Part of the problem was having adequate, reliable surveillance figures. In general, said Hargrove, the data had been "bitty" and mostly derived from urban populations.

The first generation of HIV-prevalence figures were obtained by testing pregnant women at antenatal clinics, but the age groups of the women, and the fact they were clearly having unprotected sex, meant the numbers tended to overestimate HIV infections in the general population.

Where possible, antenatal surveys are now combined with more representative data gathered in household surveys, but UNAIDS noted in its 2005 epidemic update that the high numbers of people who refused to be tested in household surveys, or were absent from home, could lead to underestimations of HIV prevalence.

While prevalence only tells us how many people are living with HIV and AIDS, incidence measures the number of new HIV infections occurring during a specific period. Incidence provides the most up-to-date and revealing snapshot of an epidemic, but the technology for determining recent infections is still quite new and prohibitively expensive for most African countries.

In the absence of such surveys, HIV prevalence in people aged 15 to 20 is often used as a proxy, because it is probable that most infections in this age group are recent.

The variety and unreliability of most surveillance methods causes epidemiologists like Hargrove and Williams to take any news of apparent declines in HIV prevalence with a large pinch of salt.

For years, Uganda has been held up as the poster child of successful prevention policies: from a peak adult HIV-infection rate of about 15 percent in the early 1990s, UNAIDS now estimates Uganda's prevalence at 6.7 percent.

Former President Yoweri Museveni swiftly responded to the emerging crisis as early as the late 1980s, and grassroots campaigns communicated basic prevention messages, such as abstinence from sex before marriage, being faithful to one's partner and the use of condoms. The ABC approach, as it has now been dubbed, combined with Museveni's leadership, have been widely credited with reducing risky sexual behaviour and lowering the prevalence rate.

But Williams pointed out that the evidence for Uganda's falling infection rate was "not really clear", and was based on a handful of antenatal surveys in the capital, Kampala. "We're desperate for a success story, so Uganda will be a success story regardless of the lack of evidence," he said.

Justin Parkhurst, of the London School of Hygiene and Tropical Medicine, also questioned the "so-called proof" of Uganda's success in reducing HIV infections in the British medical journal, The Lancet. He pointed out that the evidence supporting prevalence declines had been based on "selective pieces of information, which have been falsely presented as representative of the nation as a whole."

Parkhurst suggested that governments in low- and middle-income countries were under pressure to respond to donor fatigue by exaggerating the success of their AIDS programmes. "The standard of proof for policy recommendations seems to have been lowered, to provide the international community with the African success story it wants, or even needs," he concluded.

If Uganda's prevalence had indeed declined, there was still no sure way of determining why. Parkhurst cautioned against attributing the decline to "a few specific interventions introduced by the Ugandan government": not only were there numerous players in the AIDS fight besides the government, but "individuals can change their behaviour for reasons unrelated to intervention programmes".

Williams believed that while real behaviour changes, such as having fewer partners and higher condom use, might have taken place, they had less to do with the government's efforts and more with the widespread experience of watching friends and relatives die from AIDS-related illnesses.

Natural history of an epidemic

The dynamics of an epidemic can also bring about changes in HIV prevalence: in the early phases, HIV infections have tended to rise steeply and then level off as they reached a "saturation" point in the population; at a later stage, HIV prevalence might start declining, not necessarily because of widespread behaviour change, but because the number of people dying from AIDS-related illnesses has outpaced the number of new infections.

When the mortality rate of those infected reaches a balance with the incidence of new infections, prevalence will plateau - the stage South Africa is currently experiencing.

Paradoxically, the impact of a national antiretroviral (ARV) programme that keeps large numbers of HIV-infected people alive for longer might actually increase prevalence, or offset a lower rate of new infections.

This could explain why a country with a large ARV programme, like Botswana, has not seen significant declines in HIV prevalence, while Zimbabwe, with it's relatively small programme, has. But the real story is probably far more complex, and impossible to decipher at present, due to the lack of investment in research, monitoring and tracking national AIDS epidemics.

"Billions have been spent on virology, but we just haven't done enough basic public health research," said Williams. "Very few studies have been done trying to understand what's actually going on."

Ideally, such a study would need to monitor several thousand people over a period of at least five years, testing them regularly for HIV. According to Williams, such studies have not been done, and even in-depth evaluations of the impact of specific prevention programmes have been few and far between.

Social indicators

On the thorny question of whether prevention programmes have had a direct impact on HIV prevalence, Whiteside was as reluctant to give a definitive answer as the epidemiologists: "We can't say for sure, but equally we can't say they haven't," he said.

"There is a natural history [of an epidemic], and perhaps we've underestimated it ... The trouble is, we're looking at things that are going to take years to develop, and our monitoring and evaluation tends to be short-term."

In the absence of reliable long-term data, Whiteside believed the key to interpreting HIV/AIDS figures was "to understand what is going on in our societies more broadly".

He suggested that looking at social indicators such as the rates of rape and teenage pregnancies, or the numbers of children completing school, could provide indirect evidence of behaviour change, or lack of it.