The first AIDS vaccine tested in a large population of people at high risk for the disease has proven to be largely ineffective, according to data released early today by the vaccine's manufacturer.

AIDSVAX reduced the rate of infection 3.8 percent in people receiving the vaccine, compared with those who received placebo injections, said VaxGen, based in Brisbane, Calif.. Vaccines generally need to be at least 70 percent effective in reducing infection from human immunodeficiency virus (HIV) to gain approval for widespread use. VaxGen officials had said that a real-world efficacy of 30 percent for AIDSVAX might be enough to make the product useful in some populations.

The vaccine appeared to be effective, however, in a subgroup of recipients, notably African Americans. Among them, 2 percent who received the vaccine became infected with HIV, compared with 8.1 percent who were given the placebo -- a statistically significant difference. When Asians and mixed-race volunteers were added to the group of blacks -- totaling, in all, about 500 of the 5,000 volunteers -- the protective effect was nearly as strong.

"It appears that blacks, Asians and the other non-white volunteers were able to induce a higher level of antibody than others. There appears to be a correlation between that and protection. We need to continue to do more analysis," VaxGen spokesman James Key said of that finding.

Company officials were not available to comment on what their next step will be. They scheduled two telephone conferences with reporters and investors today.

The vaccine was tested in 5,417 volunteers at high risk for HIV infection in the United States, Canada, Puerto Rico and the Netherlands. The company is testing a similar vaccine in Thailand, but the results of that study will not be known until later this year.

More than a dozen vaccines have been tested in small numbers of people for their ability to stimulate immunity or, in some cases, slow the progression of HIV infection. The AIDSVAX trial, however, was the first one to test a vaccine's ability to protect against infection in a large and diverse population of volunteers.

The vaccine consists of a protein, called gp120, that is on the outer surface of the AIDS virus and is one of many viral structures that stimulate production by the immune system of disease-fighting antibodies. In this case, the gp120 was made by recombinant DNA technology, not by extraction from the virus itself.

About twice as many people were randomly assigned to be given the vaccine as to receive a placebo, or inactive, immunization. The vaccination schedule consisted of three initial shots spaced three months apart, and then booster shots every six months. Although slightly more than 5,400 people were enrolled in the study, only 5,009 received at least three shots. The results announced today were limited to them.

Over three years, 5.7 percent of the people receiving AIDSVAX became infected with HIV, compared with 5.8 percent of people receiving the placebo shots.

It is unclear what significance the finding of variable efficacy among different racial groups might mean in terms of the vaccine's possible usefulness. Although some racial and ethnic groups are at higher risk for certain infections -- and therefore are more urgent targets for a vaccine's use -- licensing vaccines for specific racial groups is without precedent.

The volunteers in this trial consisted of 5,108 gay or bisexual men, and 309 women at high risk of HIV infection because they are the sexual partners of those men, or of intravenous drug users. All of the volunteers were given frequent counseling to practice safer sex, and not to count on protection from the vaccine.