New York:
The full results from an AIDS vaccine trial in Thailand, released Tuesday, showed that the vaccine's protective effect might be even weaker than researchers first admitted.
However, the complicated six-shot, two-vaccine regimen may have briefly worked better in the first year after it was given, and also may have worked better in Thais at average risk from heterosexual sex, rather than those who used drugs or men who had sex with men. Those offshoot results could open avenues for future research, scientists said.
Meanwhile, the continuing debate over whether the vaccine's slight protective effect found in the study was real or just a fluke got only more complicated.
Last month, researchers from the U.S. military and the Thai government said their three-year study of about 16,400 Thai men and women suggested that the new vaccine, known as RV144, could protect 31 percent of those getting it.
Although no one would consider licensing such a weak vaccine, the announcement made headlines around the world because no other AIDS vaccine trial in 20 years had protected anyone.
But a controversy soon emerged. Of the roughly 8,200 people who got the vaccine, only 51 became infected, while among the roughly 8,200 who received a placebo, 74 became infected. The authors of the study conceded that the difference was just barely statistically significant.
Rival researchers with whom they shared the full trial data in private soon began grumbling that it could be analyzed in other ways that made the results meaningless.
The full release of data - which took place simultaneously at a medical conference in Paris and online in the New England Journal of Medicine - showed three different analyses.
The previously released one, known as the "modified intent-to-treat" analysis and showing the vaccine to be 31 percent effective, included everyone in the trial except seven people, who researchers later realized were infected before it began. Dr. Anthony S. Fauci, representing the National Institutes of Health, which oversaw the trial, described that analysis as the "gold standard."
A second, the "per protocol" analysis, included only the 12,450 subjects who got the entire vaccine series or the placebo and stayed in the trial to the end. It showed the vaccine to be only 26 percent effective, and there was also a 16 percent chance the results were due to chance. (Five percent is the usual limit in clinical trials.)
A third, the full "intent to treat," included the seven previously infected subjects and also showed the vaccine to be 26 percent effective. It had an 8 percent chance that the results were meaningless.
In an accompanying editorial, Raphael Dolin of Harvard Medical School described the vaccine as "modest" at best and "unlikely to be a public health control measure," though the results suggested further direction for research.
Different statisticians interpreted the results differently.
Philip B. Stark, a professor of statistics at the University of California, Berkeley, said he considered the full intent to treat "the most kosher analysis."
"Once you start modifying, you start introducing new opportunities for confounding the results," he added.
But Donald A. Berry, chairman of biostatistics at the M.D. Anderson Cancer Center at the University of Texas, said he accepted the modified intent to treat analysis the researchers preferred, because it was the endpoint they said early on they would use.
However, in the end, he said, all three results "are really the same, so it doesn't make much of a difference."
But his overall impression, said Berry - who had no connection to the study or any of its rivals - is that the vaccine does not work. So many trials of AIDS vaccine have been conducted he said, that it is likely that even one that consisted of injections of water would, just by chance, show a weak effectiveness like this.