In This Issue >>

Disease Forecasting

Disease Forecasting by Jim Duffy, Page 4

Model Epidemics

"Viral chatter" is a phrase Don Burke uses often in talking about emerging zoonotic diseases. It describes the way a virus on the verge of sparking an epidemic might behave in an almost hyperactive manner, generating one surge after another in reports of human infections. Burke intends the phrase to echo its war-on-terrorism counterpart, "intelligence chatter."

"It's the same idea," he says. "You have these multiple introductions, and taken all together they lend credence to this idea that something is about to happen. But we can't pinpoint exactly the whens and wheres of what that might be."

In recent years, Southeast Asia has been a hotbed of viral chatter from the avian flu strain H5N1 (the name describes its surface array of hemagglutin and neuraminidase proteins). First isolated from birds in 1961, H5N1 didn't cause a public health scare until 1997 when it bounced out of the poultry population in Hong Kong and infected 18 people, killing six, before fading.

By early 2004, however, H5N1 had again exploded, killing an estimated 100 million birds in eight countries. Human infections and deaths have been documented recently in Thailand, Vietnam, Cambodia and Indonesia.

Uncommonly virulent, the virus has stirred fears of an outbreak on the order of the major influenza pandemics of the 20th century. In 1918, the flu killed as many as 50 million people worldwide, including 500,000 in the United States. Less lethal pandemics in 1957 to 1958 and 1968 to 1969 still managed to cause 70,000 and 34,000 U.S. deaths, respectively, according to the CDC.

Scott Dowell, MD, MPH '90, is familiar with H5N1 chatter. As former director of the Bangkok-based International Emerging Infections Program (IEIP), a joint effort of the CDC and Thailand's Ministry of Public Health, he's been on the front lines, confirming human infections and tracking reported transmissions.

"What we're doing on the ground isn't really about any new advanced techniques," Dowell says. "At this point, we're falling back on traditional approaches."

Because it's so easy for physicians to miss a new, virulent strain amid outbreaks of regular flu, it's critical to identify clinical features clearly enough to ensure timely recognition of cases. So far, Dowell says, this flu has a tendency to infect young people who've been in contact with sick poultry. Patients show signs of pneumonia and lymphopenia, then progress quickly into acute respiratory distress syndrome.

Dowell and IEIP also collaborated with Thai authorities on preparations for a potential epidemic. While this work incorporates tried-and-true public health strategies as well, it's also been strongly influenced by research undertaken by Burke and Derek Cummings, a research associate in International Health. Their work has "had a big impact on the way we've been thinking," says Dowell, who moved back to the CDC in Atlanta in July to become a senior advisor to the director at the Coordinating Center for Infectious Diseases.

Working with an international team, Burke and Cummings recently published a paper in the journal Nature in which they projected via a computer model the various courses H5N1 might take in Thailand on a path from viral chatter to global threat. They then tested a range of scenarios to see whether an epidemic could be held in check so effectively that it would never escape Thailand.

Cummings, who has studied the 1918 flu in some detail, wasn't optimistic at the outset. "Given what's happened in past pandemics, there wasn't a lot of hope," he confesses.

Computer models can predict various scenarios of how an epidemic may unfold, yielding clues for how best to respond. But one situation never varies: when infection numbers get too high, the cause is lost.

In search of reliable data sets, Cummings scoured Thai government offices, academic institutions and international organizations. The model encodes each one of 85 million individuals in Thailand. It distributes them in line with real-world population densities and in accord with the size, age and gender distribution of households. It also includes measures of the distance each travels to work or school and their number of co-workers or classmates.

The modeling team first ran a simulated epidemic with no public health protections in place. That flu followed a classic epidemic curve and reached into every corner of the country within five months, on its way to becoming a global threat.

Next, they tested an overly optimistic scenario, assuming that this new flu would be relatively easy to identify and would spread at a rate in accord with "run-of-the-mill" flu. In addition, the public health response would be fast, effective and would focus on administering prophylactic doses of antiviral medicines to broadly defined contact rings of infected individuals. Under these conditions, the epidemic was contained in Thailand nearly 100 percent of the time.

Then Cummings, Burke and their collaborators devised and tested a range of scenarios with the sorts of surprises and complications inevitable in the real world. It was here that they braced for the worst outcomes, but that's not what the model delivered. As long as the new strain didn't quickly gain a supercharged level of transmissibility (in the 20th century, only the 1918 flu seems to have pulled this trick) and as long as the public health response was reasonably efficient, the containment strategy showed as much likelihood of success as failure under most scenarios.

A key and somewhat counterintuitive finding concerned antiviral medicines. Successful outcomes in any scenario emerge only if overall infection numbers are kept quite low. Once the numbers get out of hand, the cause is lost. In a sense, however, this is good news. It means that the antiviral stockpile needed to fight a transmissible H5N1 doesn't need to be as large—or as prohibitively expensive—as previously feared.

"You're looking at something in the high hundreds of thousands of antiviral courses," Cummings says. "It could go up to 1 million, maybe 2 million"—figures much lower than the many millions of doses that had originally been anticipated.

The finding opened eyes at CDC. Dowell says that new funding anticipated from the U.S. Congress will enable IEIP to begin training and outfitting as many as 100 rapid-response influenza teams and to start building an antiviral stockpile.

"The implications of this new way of thinking are enormous," he says. "There have been debates here about whether this is really feasible. One of my Thai colleagues said, 'You know, it may or may not work. But there is little lost in trying. And there is such a big potential loss if we don't try.' I think he's got it right."

So does Cummings.

"Without a doubt, the best hope we have is that a human-to-human transmissible agent never develops," he says. "But this research shows that there's a second hope. At least at the early stages, an agent is probably not going to be all that well adapted to human transmission. That gives us a chance. It's still a long shot. But it's worth pursuing."

<< Previous Page

Support JHSPH

The Johns Hopkins Bloomberg School of Public Health strives every day to keep millions of people around the world safe from injury or illness.

Invest in
Public Health >>