AIDS Vaccine Trial Produces Disappointment and Confusion
On Saturday, 15 February, a team of data analysts from VaxGen checked into a hotel a short drive from their offices in Brisbane, California, using fictitious names. The scientists had come to "unblind" the results of the first-ever, real-world test of an AIDS vaccine, a 4-year study involving 5000 people. VaxGen was so concerned that no one outside this team--and in particular no investors--would learn anything about the results until the company formally released them that the scientists agreed to speak to their families and other outsiders during their weeklong stay only through an intermediary. "There were concerns that people would read something into the tone of our voices or the spring in our walks," says Phillip Berman, the company's chief scientist. Had outsiders spotted the scientists around midnight, they may well have detected that the results were bitterly disappointing.
Late that evening, Berman and his colleagues saw the first comparison of HIV infection rates in those who received placebo shots versus the two-thirds of trial participants who received the vaccine itself. "We saw absolutely no difference between the vaccine and placebo groups," says Berman. "Our first reaction was, 'There's a mistake.' " He adds: "Everyone was pretty depressed."
The next day, however, the mood brightened. When the scientists broke the data down into racial groups--which they say was part of the original design--the vaccine appeared to have worked in black, Asian, and mixed-raced participants. "The numbers were small, which concerned us," says Berman, who with Donald Francis, VaxGen's president, helped launch the company in 1995 from the ashes of an AIDS vaccine program at the biotech powerhouse Genentech. "But the result was highly statistically significant," adds Berman. Subsequent analyses of the immune responses in the vaccinated group further buttressed the scientists' confidence. "They were pretty incredible results," Berman reports.
Berman and his colleagues released the results in a Webcast on 24 February. Francis, a high-profile AIDS epidemiologist since the earliest days of the epidemic, opened the unusual briefing. "This is a very interesting time that we're in," said Francis. Indeed, VaxGen's analyses touched off a debate that promises to reverberate around the community for weeks, if not months, to come.
The trial, which took place in the United States, Puerto Rico, Canada, and the Netherlands, predominantly involved gay men at high risk of becoming infected. The researchers reported that 5.7% of the vaccinated group and 5.8% of those who received the placebo became infected with HIV, strongly indicating that the vaccine offered no benefit whatever. But VaxGen CEO Lance Gordon argued that the analyses of racial subgroups showed "clear evidence of vaccine efficacy." In particular, VaxGen showed data indicating that the vaccine--which contains a genetically engineered version of HIV's surface protein --protected 67% of the black, Asian, and mixed-race participants. An analysis of blacks alone revealed that the vaccine worked a startling 78% of the time (see table).
Black and white? Despite overall negative results, VaxGen's Donald Francis sees hope in the subgroup analyses.
CREDITS: (TOP TO BOTTOM) HELEN DAVIS/SYGMA/CORBIS; SOURCE:VAXGEN
Hours after VaxGen made the data public, AIDS researchers began pasting asterisks and red flags all over the results. The subgroup analyses "are very interesting and provocative," says Anthony Fauci, head of the U.S. National Institute of Allergy and Infectious Diseases. "I was quite surprised." But Fauci--whose institute, over the strenuous objections of Francis and Berman, all but abandoned this vaccine in 1994 when early human trials proved disappointing--stresses that, overall, "efficacy did not occur." And he adds: "I'd like to have hard-core statisticians really pore over the data and let us know what it means."
Steven Self, a biostatistician at the University of Washington, Seattle, who specializes in AIDS vaccines (and who once advised VaxGen about the trial's study design), has strong reservations about subset analyses in general. "Subset analyses are notoriously difficult to interpret, and they're doubly difficult when the overall result is nil, which is the case here," says Self.
Seth Berkley, head of the International AIDS Vaccine Initiative, a nonprofit organization that bankrolls development of products, is more blunt. He notes that VaxGen's subanalysis hinged on "just 13 infections" among black participants. "Pull out 13 people from a large study--no matter whether they're left-handed homosexuals or whatever --and I certainly worry more about the results." Four AIDS advocacy groups--the AIDS Vaccine Advocacy Coalition, Project Inform, the Treatment Action Group, and the Gay Men's Health Crisis--issued a joint statement calling VaxGen's analysis "misleading and premature." The groups urged the company to submit the results to an outside panel of experts.
VaxGen argues that the differences may be explained by the fact that vaccinated blacks had higher levels of anti-HIV antibodies than their white counterparts did, but the company has not yet released the data to back that assertion. John Moore, an HIV antibody expert at Cornell University who has long voiced strong criticisms of the vaccine, is skeptical. "I don't buy this," says Moore. "I don't think there's a biological mechanism that will explain this."
"It's not implausible that there would be differences in immune responses in different populations," says geneticist Stephen O'Brien of the U.S. National Cancer Institute. "But I wouldn't take this as proof." He notes that African Americans have on average a 20% admixture of Caucasian genes, which "makes genetic association studies with small numbers suspicious." HIV geneticist Richard Kaslow of the University of Alabama, Birmingham, says he's especially dubious about the huge differences: 3.8% protection in the overall study group, but 78% among blacks. "That this would select a marker in blacks that has a 20-fold difference in vaccine efficacy is not impossible, but it's hard for me to imagine," says Kaslow.
Berman readily concedes the need for further study. "You can't analyze a 4-year trial in a week," he says. VaxGen plans to discuss its results in detail next month at an AIDS meeting in Banff, Canada. The company also has an efficacy trial of its vaccine under way in Thailand, which should yield results by this fall.
Issue of 28 Feb 2003,
Copyright © 2003 by The American Association for the Advancement of Science. All rights reserved.