When flu flies the coop: a pandemic threatens.

When a nasty strain of influenza first jumped from poultry to people in Hong Kong in 1997, government officials there ordered the slaughter and cremation of more than a million domestic birds. That action squelched the human outbreak, but the virus didn't go away. Six years later, that flu, known as avian influenza A H5N1, again began felling people and large numbers of birds, and the trend continues. This time, it's not confined to one country but is spreading across Asia.

So far, this virus has rarely if ever passed directly from one person to another, as the annual human influenzas do. But each new host, regardless of its species, is like a lottery ticket for the virus, giving it yet another opportunity to evolve the characteristics that would enable it to spread person to person. Many scientists say that it's only a matter of time before that happens.

The consequences of this influenza spreading among people could be disastrous on a global scale. The H5N1 virus is currently more deadly to people than were the viruses that caused past pandemics, or global outbreaks, of influenza that killed millions of people.

The worst flu pandemic on record, in 1918 and 1919, killed at least 20 million worldwide. That flu strain was lethal in about 2.5 percent of cases, giving it a fatality rate far higher than that of annual, run-of-the-mill flu viruses, says virologist Robert G. Webster of St. Jude Children's Research Hospital in Memphis.

Those typical influenzas are lethal to people over 65 years old in only about 0.1 percent of cases, and they kill less than 0.0001 percent of infected children, who constitute another relatively vulnerable group.

Since reemerging in people in late 2003, the H5N1 flu has killed about half of the 100-some people it has infected, and it has been lethal to people in all stages of life. "If it does go human-to-human and maintains anything like [a fatality rate of] 50 percent, it would be devastating," says Webster.

According to an estimate published in June by the nonprofit organization Trust for America's Health, even a pandemic strain with a fatality rate of about 20 percent could kill half-a-million people in the United States and send 2 million more to the hospital.

As governments around the world scramble to arm themselves with flu-fighting medicines and to develop public health protocols, some scientists are taking stock of the evolving threat. In addition to tracking the spread of the virus, they're studying different outbreak scenarios to determine the best strategy for responding when a pandemic strain finally strikes.

If an outbreak began today, it would catch the world unprepared.

FOWL DEEDS Among birds, the H5N1 virus is spreading at an alarming pace. Early this spring, it appeared to be confined mainly to poultry in southern China and Southeast Asia, where hundreds of millions of domestic birds have died or been culled. But during May and June, an outbreak around Qinghai Lake in central China killed more than 1,000 wild waterfowl.

In a report in the July 14 Nature, Webster and eight colleagues in China describe that outbreak and the genetic traits of the viral strain responsible for it. Influenza strains show slight genetic differences that researchers use to determine the pedigree of a virus. Such analysis suggests that the Qinghai Lake strain originated in a single bird that picked it up, the team says, "most probably from poultry in southern China." The infected birds at Qinghai Lake included bar-headed geese, gulls, and other migratory species, so the virus maybe as mobile as they are. This was the first report of transmission of the virus among migratory birds.

In another line of research, George F. Gao of the Chinese Academy of Sciences in Beijing and his colleagues tested the Qinghai strain's virulence by experimentally infecting eight chickens and eight mice. All the birds died within 20 hours, and the rodents survived no longer than 4 days, the scientists report in the Aug. 19 Science. Those findings indicate that the virus retained its lethal nature during its migration.

Gao's team notes that migratory birds that spend the summer at Qinghai Lake tend to overwinter not only in Southeast Asia but also in India and Tibet, where the virus hadn't been identified when they first published their report online on July 6. Carried by migrating waterfowl, H5N1 may well cross the Himalayas this fall and infect wild and domestic flocks in the Asian subcontinent, the scientists say.

Since the Qinghai outbreak, the virus has been identified or suspected in birds in Kazakhstan, Mongolia, Siberia, and Tibet. The virus has turned up almost as far west as Asia's boundary with Europe.

"It's probably going to spread across Russia to Europe," says Webster.

Russian officials have indicated that they also consider Azerbaijan, Georgia, Iran, Iraq, and Ukraine likely to acquire the virus from birds migrating out of Russia. United Nations representatives have also labeled some Mediterranean and central European countries at risk.

Aside from playing a role in the geographic spread of H5N1, infections in waterfowl have revealed some novel and troubling aspects of the virus.

First, avian-influenza viruses, including the H5N1 virus collected in 1997, generally aren't harmful to wild bird species. Avian-flu viruses, says Webster, have "lived with the aquatic birds for probably millions of years in perfect harmony. It's only when they come into domestic poultry that they kill."

But in late 2002, the H5N1 virus became highly pathogenic to waterfowl. An apparently new strain killed wild ducks, geese, swans, and flamingos in one outbreak in Hong Kong. "That is a very, very unusual feature," says Webster.

Second, he says, the virus has since evolved in a worrisome direction. In laboratory experiments in mallard ducks, it rapidly shifted from being potentially fatal to causing only asymptomatic infections. Nevertheless, it remained highly virulent to domestic chickens and, presumably, to people. A resilient wild waterfowl, such as the mallard, could therefore become a permanent biological reservoir for a strain of avian flu with pandemic-causing potential.

In the study, Webster and other St. Jude researchers, along with collaborators in China, Indonesia, Thailand, and Vietnam, tested 11 virus samples that had been isolated in 2003 or 2004 from infected birds and people. They also included three strains from 1997 through 2001.

At specialized facilities designed to handle highly hazardous biological materials, the scientists exposed healthy mallards to H5N1 virus and then housed the infected birds with other, unexposed ducks. The viruses spread readily among the animals. The viruses isolated after 2002 killed a quarter of the ducks. The older viruses weren't lethal to those animals.

The researchers then tested whether the lethal strains' virulence had shifted during the first experiment. They isolated H5N1 virus from four ducks that had survived an infection with an H5N1 sample that had killed at least one other bird in the same cage. When exposed to any one of the four new samples, fresh ducks showed no sign of illness, suggesting that the viruses had evolved in the course of a single infection to be relatively harmless to other members of the species.

However, when researchers tested two of the same new samples on chickens, all of the birds died. That indicates that the viruses' lethality to other species was unmitigated, Webster and his colleagues say in the July 26 Proceedings of the National Academy of Sciences.

Taken together, Webster says, the study findings hint that large numbers of wild ducks and other waterfowl may now be carrying dangerous strains of H5N1 without showing symptoms. Those birds could be flying below the radar of public health programs that investigate possible avian-flu outbreaks only when they receive reports of dead or dying birds.

This so-called passive surveillance, as opposed to active testing of apparently healthy birds, is the most that may be feasible in many regions of Asia where financial resources are scarce.

If the virus already is widespread but largely concealed in some wild birds, control measures such as culling overtly sick flocks will have little effectiveness.

PRACTICE PANDEMICS Draconian measures to contain and eliminate infected birds will become even less valuable if H5N1 changes into a strain that is easily communicable among people. At that stage, deflecting a pandemic might depend on how, when, and where governments deploy key medical tools and public health policies, such as travel lock downs and quarantines.

Nearly all the confirmed human infections since 2003 appear to have resulted from contact with infected birds. But a few people already appear to have been infected with H5N1 by relatives with whom they've had household contact.

In the Jan. 27 New England Journal of Medicine, researchers in Thailand, along with two U.S. colleagues, reported a pair of probable human-to-human infections, both originating with the same infected person in a Thai home in 2004. Interpersonal transmission may also have caused two cases in Indonesia in July.

In past pandemics, a bird flu moved into people after the virus swapped genes with another flu virus, for instance, the agent causing a swine flu. But that won't necessarily be the case this time.

Any cluster of human cases with a strain that doesn't look like a standard winter-flu strain would be sufficient to indicate an imminent threat warranting a coordinated global response, according to biostatistician Ira M. Longini Jr. of Emory University in Atlanta.

Two research teams recently published studies analyzing how that response might unfold. Both groups used mathematical models to consider the relative importance of viral characteristics, antiviral drugs, and measures for preventing transmission in the early days of an outbreak, while the virus is still near its source.

"Neither one of the models shows that we can contain it at the source," stresses Emory biostatistician M. Elizabeth Halloran, who coauthored one of the studies with Longini and other researchers. "Both show that, with really concerted effort, there is a possibility that we could contain it."

"That should not lull anybody," Halloran says.

In the hypothetical scenario that Halloran and Longini considered, an outbreak begins in rural Southeast Asia. The researchers used data from a recent Thai census and other sources to estimate such factors as how densely populated the strain's birthplace might be and how many times each day typical residents interact closely enough that they might spread the infection. By using different values for the outbreak strain's ease of transmission, the investigators plotted several possible courses of the pandemic.

They also added flu-fighting countermeasures to the model, including the distribution of antiviral drugs such as oseltamivir (Tamiflu) and the use of a partially effective vaccine. They assumed that it would take at least a week from onset of the first human illness for health officials to recognize the outbreak and respond.

The World Health Organization (WHO) and some national governments have stockpiled oseltamivir, which could be used to prevent infections or to treat people already sick with flu.

"A mobile stockpile of Tamiflu that could be moved anywhere on the planet rapidly is the way to go," Longini says. On the basis of their study's results, he and his colleagues say in the Aug. 12 Science, "the current WHO stockpile of 120,000 courses could possibly be sufficient to contain a pandemic if the stockpile were deployed at the source of the emerging strain within two to three weeks of detection."

Successful containment would depend on many factors, however, and the model could be overly optimistic, the researchers acknowledge.

The other modeling analysis, conducted by Neil M. Ferguson of Imperial College London and his colleagues, predicts in an upcoming Nature that "a stockpile of 3 million courses of antiviral drugs should be sufficient." That study treated Thailand as the epicenter of the epidemic.

This analysis assumed that an effective vaccine wouldn't be widely available by the time a pandemic begins. Anthony S. Fauci, director of the National Institute of Allergy and Infectious Diseases, in Bethesda, Md. announced on Aug. 6 that government scientists had found an experimental vaccine against H5N1 that seems effective. The investigators observed a strong immune response in people who received two shots of vaccine, but its efficacy in preventing disease isn't known because no volunteer was exposed to the actual pathogen.

Secretary of Health and Human Services Michael O. Leavitt said in July that the United States will stockpile enough vaccine for 20 million people and antiviral medicine for 20 million more.

But, according to Fauci, the government has purchased only enough H5N1 vaccine, from French manufacturer Sanofi-Pasteur, to inoculate 450,000 people at the dose used in the recent test. Acquiring more vaccine will take months, because few facilities are capable of mass-producing it.

Furthermore, the pandemic strain will almost certainly not correspond precisely to the strain that the Sanofi-Pasteur vaccine targets. That will limit the vaccine's effectiveness.

The existing vaccine could still shield some inoculated people and, among those who get infected despite the shot, the treatment could reduce the severity of the disease and the rate at which they spread flu to others. "Even if it didn't completely protect people, it would still [make other] measures more effective," says Halloran.

If the transmissibility of the virus is sufficiently high, however, or if the preventive pharmaceuticals can't be deployed in time, a local outbreak could expand like a wildfire. In that ease, Halloran's and Ferguson's models indicate, social procedures such as canceling schools and dismissing office workers could be of particular importance.

But those actions are socially and economically disruptive, and they offer no guarantee of choking the pandemic, Halloran says. If people decide not to leave home to get food, she says, even pizza deliverers could become vehicles for spreading influenza.

In an interconnected world, finding protective isolation isn't as simple as staying far from the crowd.