AZT, the only drug approved for treating the AIDS virus directly, appears significantly less effective in blacks and Hispanics who get the drug early in their disease than in whites started on AZT at the same early stage, according to a preliminary study announced yesterday.

The researchers who conducted the study, sponsored by the Department of Veterans Affairs and presented at a Food and Drug Administration advisory committee meeting yesterday, stressed that their data were "fragile" and that more studies were needed to determine whether their results were real or a statistical fluke.

When given late in the disease, AZT appeared equally effective in all patients at prolonging survival.

Other studies of AZT presented at the day-long meeting also failed to confirm the differences between whites and minorities seen in the VA trial, which led members of the committee to recommend against changing federal guidelines for treatment of HIV disease on the basis of the study.

"It is difficult to know where chance has confused things here," said David Feigal, an epidemiologist at the University of California at San Diego and a member of the committee. "The small number {of people} in this trial is a problem. . . . We're going to need confirmation."

But many researchers and AIDS activists present at yesterday's meeting, appeared shaken by the results, which contradict the conventional wisdom that AZT is most effective for all HIV patients when it is given in the earliest stages of the infection. The study also suggests that researchers may have paid insufficient attention to how differences of race, ethnicity or socio-economic status affect AIDS treatment.

The VA researchers said they checked whether other factors could explain their findings. For example, they found no significant differences among the groups in diligence of taking their medicine or in ratios of intravenous drug users to homosexuals.

Researchers generally dismissed genetic differences as a factor and instead suggested that patients from minority groups, who are often more likely to be poor, are beset by additional diseases and have less access to routine health care.

"These are very disquieting results," said Anne Gershon, a pediatrician at the Columbia University College of Physicians and Surgeons in New York. "We don't know where we stand."

"Last year when we stormed the National Institutes of Health, our motto was '10 years, one billion dollars, one drug.' " said Mark Harrington, an AIDS activist with ACT-UP. "If I was a person of color looking at this study I would say '10 years, one billion dollars, no drug.' "

The VA study, which ran for four years, involved 338 HIV-infected people divided into two groups. The first group was given AZT on or around the time their T-cell blood counts dropped to 500 per cubic millimeter. T-cells are the immune system cells targeted and killed by HIV. When the count drops below 100, death is typically imminent. This is the point at which federal health officials have for over a year recommended that AZT be administered to HIV-infected people.

The second group was given a placebo until their T-cells dropped below 200, when they too were given AZT. This fits the guideline originally promulgated for AZT therapy.

The study's overall results, lumping whites and minorities together, were consistent with previous studies of AZT. Those given AZT earlier remained healthier longer than those given the drug when their T-cell levels dropped below 200. Length of survival for the two groups was about the same.

The surprising results emerged only when the groups were divided along racial and ethnic lines. Among the 118 blacks and Hispanics, the time it took to develop AIDS symptoms was the same, regardless of when AZT therapy was begun. More surprisingly, those who received the drug early, died far more quickly than those who received it later.

The VA researchers conceded their data need to be confirmed. Other researchers warned that the study was not designed to show racial differences in response to AZT and that retrospective analyses of this sort are notoriously unreliable.

Following the VA study, three other previously completed AZT studies were re-analyzed to see if they also showed a racial difference in response to AZT. None of the three showed any differences as striking as in the VA study, although because of differences in the design of the trials, some researchers said that reliable comparisons could not be made.