The natural assumption was that he must have picked up the disease from his mother in the womb, but her HIV test came back negative. So where did Sipho catch the virus? No one can be sure, but it is most likely that he was infected in hospital, perhaps by a needle that had not been sterilised after being used on an infected patient.

The World Health Organization thinks that tragedies like Sipho's are very much the exception. It estimates that unsafe injections during healthcare account for just 2.5 per cent of HIV cases in Africa, and that the vast majority of infections are via sex.

But some researchers believe the role of dirty needles has been greatly underestimated. If they are right, relatively simple measures could save millions of people worldwide.

This week, the group Physicians for Human Rights based in Washington DC sent an open letter to the WHO and UNAIDS. It calls for more resources to be spent on preventing infection by dirty needles. The letter says people should be educated about the dangers, and measures taken such as providing syringes that cannot be used more than once.

'Firestorm of protest'

But the WHO and UNAIDS have long resisted the suggestion that injections are an important driver of the epidemic. "It has been a huge struggle to make the case that this is a significant part of the epidemic," says Ernest Drucker, an AIDS expert at Yeshiva University in New York. "We've run into a firestorm of protest."

"The worry is that if too much attention is paid to unsafe injections it will take away from the message about sexual transmission," says James Whitworth at the London School of Hygiene and Tropical Medicine, who backs the WHO position. Another fear is that vaccination programmes will be undermined if injections are seen as risky.

While these concerns might be valid, critics argue the consequences of downplaying the role of dirty needles are far worse. The most vociferous of them is David Gisselquist, an independent researcher in Hershey, Pennsylvania, who has published a string of papers highlighting dirty needles as a major risk factor (New Scientist print edition, 1 March 2003).

Using the WHO's own estimate that 7.6 per cent of infections in 1988 were from dirty needles or blood transfusions, he says healthcare is to blame for 10 million infected people today. If needles cause closer to half of all infections, as Gisselquist believes, tackling the problem would have kept the epidemic confined to high-risk groups, he claims.

"In Asia, if we don't get that message out, the epidemic could really blow up," he warns. The WHO's own figures, based on observations in hospitals and clinics, suggest that up to 75 per cent of injections in parts of south-east Asia are carried out using unsterilised equipment, compared with just 20 per cent in sub-Saharan Africa.

Unpublished report

Gisselquist's work prompted the WHO to hold a meeting on unsafe injections in March 2003. He says data supporting his claims was presented, but it was not reflected in the meeting's conclusions. Instead, the press release proclaimed: "An expert group has reaffirmed that unsafe sexual practices are responsible for the vast majority of HIV infections in sub-Saharan Africa."

Six months before the meeting, UNAIDS drew up a report, which has been seen by New Scientist, that contradicts this position. Based on a review of 23 studies, it concludes that in sub-Saharan Africa, "contaminated injections may cause between 12 and 33 per cent of new HIV infections". That is far higher than the accepted 2.5 per cent figure.

That report has never been published, prompting Gisselquist to accuse the WHO of ignoring evidence that does not support its views. But according to Peter Ghys of UNAIDS in Geneva, the document was a preliminary draft that has since been incorporated into a much larger summary of the evidence. That study, due to be published early next year, will support the WHO estimate of about 2.5 per cent.

George Schmid, a senior researcher on HIV at the WHO in Geneva and author of the revised study, says the apparent change of view arises because a statistical technique used in the 2002 draft is inappropriate for HIV.

The reviewed studies calculate a "population attributable fraction", the proportion of infections in the population due to a specific risk factor. Schmid says this method works for non-infectious diseases, but not when infected people can affect the future course of the disease by infecting other people.

Second birthday
 

Gisselquist's critics also ask why hepatitis C, which is mainly spread by needles, does not mirror the pattern of HIV infection, and why HIV has spread in some countries with relatively good healthcare.

In response, Gisselquist claims hundreds of studies have reported significant numbers of children who, like Sipho, have contracted the disease despite having HIV-negative parents or parents with a different HIV strain.

A study of nearly 10,000 South Africans released in 2002, for instance, found that 5.6 per cent of children aged between 2 and 14 were infected. Most children infected by their mothers die before their second birthday, so the surprisingly high figure points to infection routes other than sex being important.

But Schmid says the results of all these studies are questionable. For instance, the instrument used to collect samples in the South African study was not approved by the FDA for use on children, he says. Schmid is now helping to design a follow-up study.