WARNING SIGNS: A 13-year-old boy infected with the H5N1 avian flu virus goes into hospital quarantine north of Bangkok in October 2004. Among 71 known cases in Vietnam, Cambodia and Thailand, one instance of transmission between people has already been identified. Image: EPA
In anticipating the next pandemic, flu specialists think the H5N1 avian flu strain, which has jumped from birds to dozens of people in Asia, will inevitably adapt to spreading from person to person. The first local outbreaks could then quickly fan out across the globe. If the disease follows the pattern of previous pandemics, a third of the world's population could be infected and perhaps 1 percent of those people might die. That is, unless the inevitable could be delayed long enough for countries to ramp up production of vaccines and antiviral drugs.
A bold idea circulating among flu experts offers a way to create that delay. The concept is simple: detect the first clusters quickly and then slow or squelch the emerging virus by blanketing the outbreak area with antivirals. In the past, "no one even considered this thought of containment on the agenda," says Emory University biostatistician Ira M. Longini, Jr. "Now we have a control tool, and we know a lot more about how these things emerge."
Longini is one of several researchers using computer models to test the strategy. At a conference in February, he described some of his findings for possible scenarios in a hypothetical Southeast Asian rural community of about 500,000 people. Density, demographics, travel habits, household sizes, work sites and schools are all based on Thai government data, but Longini thinks they can also be extrapolated to neighboring countries.
By simulating each person's susceptibility and daily contacts, Longini's model projects how the adapted flu strain might spread. In epidemiology, an all-important variable is the disease's "reproductive number"--the average number of new infections that one infected person will cause. This figure, abbreviated R0 (R-"naught"), is typically low for flu--the 1918 pandemic strain's R0 was around 2, according to Longini. Flu moves notoriously fast, however, because its incubation period is brief: within a day of infection a person may begin transmitting the virus, unwittingly, because symptoms appear only on the second day.
By plugging these parameters into the model and running each of several scenarios 100 times, Longini produces probabilities for different outcomes. A single villager could initiate a chain of human transmissions of a virus with an R0 of 1.4, for example. If health officials detected the outbreak 14 days later and then began targeting victims and their contacts for treatment and prophylaxis with the antiviral drug Tamiflu, the outbreak would be contained 98 percent of the time. Just 2 percent of the time more than 500 people would become infected, but the outbreak would rarely escape the region. In all scenarios, odds of containment became better still if everyone in the geographic area received prophylactic Tamiflu when outbreaks were first detected. "Draconian quarantine" also helped, Longini adds.
The higher the R0, however, the lower the likelihood of containing the virus. When the R0 is set at 2.4, for example, the outbreak quickly grows uncontrollably large in 75 percent of the simulations--with the exception of scenarios in which the population has been vaccinated in advance, even if the vaccine is not a perfect match to the adapted H5N1 strain. "With prevaccination, you can contain even a large R0 with antivirals," Longini explains. "It basically buys you time; it effectively lowers the reproductive number."
Such modeling enables officials "to get a handle on how much antiviral would actually be required to control an outbreak," says Nancy J. Cox, chief of the Influenza Branch at the U.S. Centers for Disease Control and Prevention. In reality, successful containment would depend on variables that cannot be predicted, Cox cautions, such as whether an infected person carried the new flu to a large city, where contacts in crowded public spaces would be harder to trace.