In a statement issued at its headquarters in Geneva, UNAIDS warned against interpreting the survey as a sign that the HIV/AIDS crisis in Kenya had been overestimated, and a senior official said key data in the study seemed flawed.

"There's something wrong there," that official, UNAIDS Chief Scientific Advisor Catherine Hankins, said in an interview with AFP.

The study, the Kenya Demographic and Health Survey, is carried out every five years to plan health and social policies.

The latest survey was carried out in September last year and, for the first time, asked people whether they would be tested for HIV.

The tests were carried out and validated by officials from the US Centers for Disease Control and Prevention (CDC).

People from representative households across the country took part in the survey, but only around 70 percent of respondents -- 6,000 people -- agreed to take an HIV test. The remaining 30 percent refused.

The preliminary results suggest Kenya has an infection rate of 6.7 percent, compared with UNAIDS' far higher estimate of 9.4 percent.

The director of Kenya's National AIDS and Sexually Transmitted Diseases Programme (NASCOP), Kenneth Chebet, maintained the figures reflected an overall decline in the number of HIV-infected Kenyans, "from about 2.5 million people four years ago to an average of 1.4 million."

The UNAIDS statement said it welcomed household studies because they had the potential to add an important tool to the meagre data sources available for estimating the spread of HIV/AIDS in Africa.

The agency's main data source is the HIV status of women who attend antenatal clinics, whose blood is tested routinely for the virus.

But the statement stabbed a finger at the study's dramatic difference in gender infection rates -- 8.7 percent among women, but just 4.5 percent among men, which is almost unheard-of in AIDS research.

"Nowhere, in any country in the world, do you see a twofold difference between men and women in the population," Hankins said in the interview with

"When you look at the prevalence among men, that just hits you right in the face -- there's something wrong there, there's something wrong."

The likely reason for this big difference, she said, was the 30 percent of people who had refused to take a test.

Many of these could have been men who engaged in risky sexual behaviour but, fearful of the stigma, probably shunned taking the test, she suggested.

Hankins stressed that household surveys were welcomed by the agency, and noted that similar surveys from four other African countries had been taken into account when UNAIDS last December published an update of the global AIDS pandemic.

By factoring in those surveys and other sources, the agency downgraded its estimate, saying it believed around 40 million people had AIDS or the human immunodeficiency virus (HIV) by the end of 2003, compared with 42 million people a year earlier.

Two-thirds of the total live in sub-Saharan Africa, according to that estimate.

The debate is important, for the Kenyan study is likely to provide ammunition to those hardline conservatives who accuse UNAIDS and other agencies of AIDS scaremongering.

UNAIDS' heavy dependence on HIV rates from antenatal clinics in Africa has been a matter of debate among AIDS specialists.

The advantage of this data is that 100 percent of women get an HIV test and the method is simple and a good comparative tool for countries across the continent.

However, the clinics themselves are based mainly in urban areas rather than rural regions, and do not reflect HIV prevalence among gays, prostitutes and injecting drug users.

TERRA.WIRE