Doubts have been raised about the reliability of a trial suggesting success for a vaccine against HIV.
In the large-scale trial in Thailand, a combination of vaccines seemed to give volunteers a protective effect of 31%.
The US military and Thai government, who co-sponsored the trial, said the effect was not caused by random chance but was statistically significant.
But new data, being published at a conference in Paris on Tuesday, is believed to question that assertion.
It was the world's largest clinical trial of a HIV vaccine - involving 16,000 people in Thailand aged between 18 and 30.
Among the 8,000 volunteers who had been given the combination of vaccines, 51 had gone on to become infected with the virus.
Of the group given a placebo, there were 74 positive cases.
The numbers were small, but according to Seth Berkley of the International Aids Vaccine Initiative, the results were "exciting news and a significant scientific achievement".
He said: "Now we have got a vaccine candidate that appears to show a protective effect in humans, albeit partially."
Lack of detail
But there were concerns as well. Researchers were not able to indicate just how the vaccines were working.
And, as more data was given to scientists, the claims about statistical significance began to look increasingly shaky.
Gus Cairns, who works with UK HIV information charity NAM, said: "This particular study was in the awkward position of producing a result that was only just statistically significant.
"This means there was a one in 26 chance that the results could be due to pure chance - and that this may not reflect anything at all.
"That's difficult. And there is also subset analysis of this study that if you only look at the people who strictly adhered to the protocol - ie took all their vaccine doses - then it becomes not statistically significant."
The problem is that the initial figures given for the numbers infected included all those who got HIV once the trial started, including those who got it in the course of the six-month regime of injections.
But if these were excluded, as they would be in many trials, then the numbers change - and so do the claims of a protective effect.
Study power reduced
These details have been substantiated by the US military HIV research programme.
In an update to the study, they indicated that if you looked at the data in this way, it does not reach statistical significance and the study's power is reduced.
Researchers are expected to give a more detailed breakdown at the Global HIV vaccines meeting in Paris.
Many scientists have been upset that the initial information about the trial came out via a press conference, rather than via a peer reviewed journal.
But according to Dr Aland Bernstein, executive director of the Global HIV vaccine enterprise, the issue of how the information came out does not matter.
"At the end of the day what matters to me is the long run. If that work doesn't hold up that's fine, we'll hear no more of it.
"If it holds up, and we'll only know that over the weeks and months ahead, then it is an important contribution - and whether we heard those results on 24 September or 24 October it doesn't matter in the long run."