The decline in AIDS deaths in the United States that began three years ago has started to level off, and there are disturbing signs that new HIV infections may be rising among young gay men, researchers reported yesterday.

While the benefits of powerful antiviral drugs introduced in recent years remain dramatic by any measure, there also is evidence that excessive confidence in them may be prompting some people at risk for HIV infection to practice unsafe sex, researchers said.

The number of AIDS deaths nationwide dropped 42 percent in 1997, but fell only 20 percent in 1998, the researchers reported at the National HIV Prevention Conference in Atlanta, a meeting sponsored by the federal Centers for Disease Control and Prevention. Most of 1998's reduction occurred in the first three months, with the number of AIDS deaths per month holding relatively steady for the remainder of the year.

A similar trend was seen in the annual decline in new cases of AIDS, which is the advanced stage of infection with HIV, the human immunodeficiency virus. They fell 18 percent in 1997, but only 11 percent in 1998. More recently, epidemiologists in San Francisco determined that one-fifth of all people testing positive for HIV acquired the infection in the previous six months, a sign that spread of the disease continues apace.

The new statistics, following several years of relatively good news about AIDS, offer sobering evidence that the epidemic is far from being under control, experts said.

"The data tell us that this is still an unstable epidemic," said Helene D. Gayle, director of the CDC's division of HIV prevention. "No matter what new drugs we have, prevention is ultimately going to be our best weapon."

The slowing of the decline in AIDS mortality is not unexpected. Life-saving or life-prolonging therapies always have their most dramatic effects, in epidemiological terms, soon after they are widely adopted by previously untreated populations. When many deaths are prevented in one year, the pool of untreated people shrinks, reducing the number of lives that potentially can be saved with the treatment in subsequent years.

In the case of antiviral treatment for HIV infection, this began in 1996, the first year that multiple "protease inhibitor" drugs became available in the United States. Today, 85 percent of people with HIV who qualify for the combination therapies are taking them.

Nationally, AIDS deaths have dropped from about 50,000 a year in 1995 to 17,000 a year now. In some places, the decline has been breathtaking. In the county that includes Seattle, there were, on average, 435 deaths from AIDS each year from 1993 to 1995. Last year, there were 78. In Seattle, the annual death rate fell from 59 per 100 people with AIDS in 1987 to 4 deaths per 100 people with AIDS in 1998.

Factors other than the expected effects of better treatment also can contribute to the flattening of the downward trend in AIDS mortality now being reported. These include the appearance of drug-resistant infections and the inability of some patients to take the medicines, either because of side effects or inconvenience. How much those factors may explain the trend is unknown.

The trend in AIDS mortality, however, doesn't necessarily reflect the trend in new HIV infections. That's because of the long lag (on average, about a decade) between the time infection occurs and when the virus has done sufficient damage to the immune system to cause the problems that fit the official definition of AIDS. The new therapies are prolonging the silent period of infection -- indefinitely, it appears, in some cases -- making recent AIDS mortality an even less reliable indicator of HIV "incidence," or rate of new infection.

Recently, however, researchers have used paired AIDS blood tests, one very sensitive and the other relatively insensitive, to distinguish long-standing infections from recent ones. (The new ones give a positive result only on the very sensitive test.) This has provided an unusually good lens with which to study the epidemic among various groups of people at risk.

At the San Francisco Department of Public Health's anonymous HIV counseling and testing sites, 1.1 percent of men reporting homosexual activity were newly infected in the second half of 1997, said Willi McFarland, a physician and epidemiologist. In the first half of this year, however, 2.8 percent of such men were newly infected, as detected by the paired blood tests.

Although in statistical terms the difference between those two percentages is not significant -- the 2.8 percent could be an upward blip that occurred by chance -- other data suggest a trend may be starting.

The rate of rectal gonorrhea in San Francisco -- a powerful predictor of a man's risk for acquiring HIV -- went from 20 cases per 100,000 men in 1994 to 40 cases per 100,000 in 1997. At the same time, the percentage of gay men in the city who report using condoms all the time has fallen from 70 to 60. The percentage of men reporting multiple anal-sex partners has risen during that period.

"At best, we can conclude that incidence of HIV is not going down in San Francisco in the last four years," McFarland said. "In light of the data on . . . gonorrhea and high-risk sexual behavior, the recent upward trend . . . among testers becomes more worrisome."

At the anonymous testing sites in San Francisco, gay male users of injected drugs were the group with the highest percentage of new infections in 1998 and 1999. HIV incidence, however, was zero or near zero for other risk groups. Last year, the sites found only one woman newly infected with the virus, and in four years have found no recent infections in heterosexual men, heterosexual men who inject drugs or women who inject drugs.

In a study of 96,000 people seen at clinics for sexually transmitted diseases in six cities (Baltimore, Miami, New Orleans, Houston, Denver and Los Angeles) between 1991 and 1997, about 8 percent of gay men became infected each year. Black gay men had the highest rate (11 percent a year), followed by Hispanics (7.5 percent) and whites (6.5 percent).

That study, unlike the one in San Francisco, found no upward trend in new infections. However, the six-city study ended in 1997, so it is not known whether HIV incidence has risen in the last year or so, said Hillard Weinstock, the CDC physician who led it.

In a study of about 400 gay men in West Hollywood, Calif., all of whom had heard of protease inhibitors, infected people who were "optimistic" about the drugs' ability to prevent infection used condoms 66 percent of the time, compared with 80 percent condom use among those who were "pessimistic" about the drugs. Among the uninfected, the optimists used condoms 74 percent of the time, and the pessimists 85 percent of the time, reported Sheila Murphy of the University of Southern California.

Although the rates of new infection, and the overall prevalence of HIV, varied from city to city, nearly all recent studies have found black gay men at highest risk. The CDC is especially targeting African Americans for its prevention programs, with its budget for prevention efforts in the black community now at $138 million a year, compared with $11 million a year in the late 1980s.

AIDS Trends

The decline in AIDS deaths has leveled off and new infections among gay men may be rising.

Men account for 70 percent of new infections. Estimates of new infections among men . . .

. . . by race/ethnicity

Black 50%

White 30%

Hispanic 20%

. . . by transmission risk

Homosexual 60%

Heterosexual 15%

Intravenous drug user 25%

SOURCE: 1999 National HIV Prevention Conference