November 10, 2001

UNCERTAIN ESTIMATES OF NEW HIV INFECTIONS

The federal Centers for Disease Control and Prevention routinely reports that new HIV infections have leveled off.

For instance, the June 1 issue of the Morbidity and Mortality Weekly Report, marking the 20th anniversary of the emergence of the epidemic, said ``new infections in the United States, which had risen rapidly to a peak of 150,000 per year in the mid-1980s, declined to an estimated 40,000 per year since 1992.''

The weekly report attributes the decline to better prevention efforts and wider social awareness.

Isn't it more than a little odd that the number has scarcely changed in almost a decade when so much else about HIV and AIDS has changed in that time?

Most dramatically, the death rate from AIDS has plunged as new drug regimens allow many HIV-positive Americans to live for years longer before developing AIDS, if indeed they ever do. The number of AIDS cases among men who have sex with men has dropped quite sharply, the number of cases among users of injected drugs has also dropped though less steeply and the number of cases acquired through heterosexual contact has stayed fairly level, so those cases are a larger percentage of the total than they were 10 years ago.

David Murray wondered about that too, especially since he knew that blood-test sites in 36 states had reported that positive test results had dropped from 58,000 in 1991 to 30,000 in 1998 and 21,000 in 2000. That's not the national total, obviously, because it's not all the states and it isn't the only way to get tested, but shouldn't the national total show a similar decline?

Murray is the director of the Statistical Assessment Service, which publishes a little newsletter, Vital Stats, that casts a skeptical eye on reports of dubious statistical merit, especially those that are swallowed whole by the media (on the Web at stats.org).

So he wondered in print where the CDC numbers come from, and in the October issue he explains. The CDC doesn't count new infections, it estimates them.

Dr. John Karon of the CDC outlined for Murray how the estimate is arrived at. First, there are 60 million American men aged 18 to 44, and the CDC assumes that roughly 2 percent of them are likely to be gay and to engage in risky sexual behavior.

Of the 1.2 million men at risk, 400,000 were already infected by 1992. The CDC further assumes that 2 percent of the remaining 800,000 become HIV-positive each year, accounting for 16,000 new cases.

But those cases, the CDC estimates, are only 60 percent of the total among men, because some are infected through shared needles or heterosexual contacts. So the total among men is close to 27,000. If 30 percent of new cases are among women -- another estimate -- then the overall total is 38,000, commonly rounded up to 40,000.

One can't say that the result is wrong; rather, there are so many assumptions built in that no one could have any idea whether it's right. Nor does it take into account shifts in the pattern of infection or in diagnosis and treatment.

It could be too high; the final number depends crucially on the starting-point figure of how many gay men were already infected in 1992, a number that has been retroactively scaled back from what it was believed to be at the time.

But it could equally well be too low. A different CDC study, also included in the June 1 weekly report, used a new method of antibody testing that can distinguish between recent infections and long-standing ones. Though the authors cautioned that the figures were preliminary, they found that the rate of new infections among young gay men tested in six cities was 4.4 percent overall, comparable to rates last seen in the early 1990s. Lead researcher Linda Valleroy called the numbers ``explosive.''

In the United States AIDS is a matter of huge medical -- and political -- concern. If even here it is essentially impossible to find out how rapidly people are becoming infected, then how accurate, Murray wonders, can the worldwide estimates possibly be, given that many of them are from countries with rudimentary medical infrastructure? He's concerned that in many places many deaths are being reported as AIDS-related when no diagnosis has actually been made. That's a situation ripe for political exploitation.

There's no reason to believe the CDC's steady-state estimate of new infections is politically motivated, except the general one that politics influences just about everything having to do with AIDS. But that's reason enough to wonder.

(764 words)