The first anti-AIDS vaginal gel to make it through late-stage testing failed to stop HIV infection in a study of 6,000 South African women, disappointed researchers announced Monday.

The study was marred by low use of the gel, which could have undermined results, they said. Women used it less than half the number of times they had sex, and only 10 percent said they used it every time as directed.

Scientists are still analyzing the results to see if this made a difference. They also plan more tests on a revamped gel containing an AIDS drug that they hope will work better.

The gel used in the current study did prove safe, however, and researchers called that a watershed event.

But for now, the effort is the latest disappointment in two decades of trying to develop a microbicide - a cream or gel women could use to lower their risk of getting HIV through sex. A female-controlled method is especially needed in poor countries where women often can't persuade men to use condoms.

A year ago, scientists stopped two late-stage tests of a different gel after early results suggested it might raise the risk of HIV infection instead of lowering it.

The new study tested Carraguard, a microbicide developed by the nonprofit, New York-based Population Council. It contains carrageenan, which comes from seaweed and is widely used in the food and cosmetics industries as a gel, stabilizer and thickening agent. Lab, animal and early human tests suggested it might prevent HIV and other sexually spread infections.

The latest study was done from March 2004 through March 2007 in Gugulethu, Isipingo and Soshanguve, all in South Africa.

More than 9,000 women, average age 31, volunteered for the study. About 27 percent tested positive for HIV and were disqualified. In all, 6,202 women were randomly given either Carraguard or a placebo gel. Neither the women nor the study staff knew who received what. All received safe-sex counseling and condoms.

Women participated from nine months to two years, with 4,244 completing the study. About 18 percent dropped out, often because they became pregnant and the gel is not known to be safe for use in pregnancy. Another 13 percent could not be found for follow up information.

At the end of the study, there were 134 new HIV infections in the Carraguard group and 151 in the fake gel group - a rate of 3.3 infections per 100 women each year in the microbicide group and 3.7 for the placebo group.

"The results are comparable," with no statistically significant difference, said Khatija Ahmed, a microbiologist who headed the study's Setshaba Research Centre site near Pretoria.

However, women in the study used the gels only 44 percent of the time, and some used it hardly at all. Researchers are still analyzing the numbers to see what that means. If nonuse was far greater in the microbicide group than the placebo group, "it could have had an impact on our final study results," said Barbara Friedland, the study's behavioral coordinator.

A plus: reported condom use doubled, from 33 percent at the start of the study to 64 percent during it. Other sexually spread infections declined.

The study was paid for by the Bill & Melinda Gates Foundation and the U.S. Agency for International Development, or USAID.

Jeff Spieler, an official at USAID, called the trial "groundbreaking work" in a statement. "We have always known that the path to developing a successful microbicide would be a long one."

The Population Council hopes to start tests this year of a revamped Carraguard containing an experimental AIDS drug, MIV-150. The group also has studies under way of a contraceptive version of the gel, Carraguard plus hormones.