Disease modeling's dynamic duo

When constructing a building, a solid foundation is critical. One's experience and skill in building the substructure ensures a base capable of supporting growth. The same holds true for assembling a new research program: bring in the best and brightest, and the science will soar.

The Hutchinson Center has already benefited from the addition of Drs. M. Elizabeth "Betz" Halloran and Ira Longini, longtime biostatistical collaborators at Emory University who joined the Public Health Sciences Division in January. In their first few months here, they got a scientific computing cluster up and running to enable their work in disease modeling, published findings in Science, and regularly advised government officials on flu-pandemic preparedness. The pair are key recruits in a new vaccine-research center planned as a joint venture with the University of Washington, where both teach in the Department of Biostatistics.

"It was a real coup to be able to bring Betz and Ira to the Center," said Dr. Steve Self, head of the Biostatistics and Biomathematics Program. "They are one of the few research teams in the world that integrate sophisticated mathematical modeling of infectious-disease dynamics with strong foundations in the design and analysis of field studies."

Halloran said she and Longini are excited to utilize the strengths already present in the PHS Division's Statistical Center for HIV/AIDS Research and Prevention (SCHARP). "Scientists at the Center have huge capacity in the biostatistical area, so there's a lot we can build on," she said. "There are great people here, and hopefully, we can get some of them working together in a more cohesive manner. That's part of our role, as well as recruiting new people."

"Our goal is to have the biggest computational epidemiology group in the United States, a really vibrant center. People here are not afraid to think big."

Longini and Halloran's work has long been on the Center's radar. Their assessment of HIV vaccines in the early 1990s caught the attention of Center scientists years before the HIV Vaccine Trials Network was established here. "Betz and I and our little group at Emory pretty much pioneered that whole area of design and analysis of HIV-vaccine trials," said Longini, who's worked with Halloran for 16 years. "We felt these vaccines probably wouldn't completely protect against infection, so researchers really needed to think about trials, designs and analysis to measure how well vaccines reduced disease progression and transmission to others." Both were very involved in the analysis of the early HIV-vaccine trials in Thailand and the United States.

The pair had the field of disease modeling — using mathematical and statistical methods to study infectious disease — practically to themselves for the past two decades. However, in the wake of the Sept. 11 attacks, they suddenly found their skill set in hot demand.

The federal government contracted them and several other researchers to create simulations of a smallpox outbreak caused by bioterrorists. Their findings, which showed careful use of the smallpox vaccine could contain the disease, were published in Science. "There was interest in our work so that the United States would be better prepared and have more capacity," Halloran said.

The growing concern over a possible influenza pandemic caused by the H5N1 "bird flu" strain has brought an even greater spotlight to Halloran and Longini's skills. Since 2004, they have been charter members of a new, interdisciplinary network of disease modelers who help the government understand and prepare for infectious-disease outbreaks. The network, called MIDAS (Models of Infectious Disease Agent Study), receives about $7 million annually from the National Institute of General Medical Sciences, part of the National Institutes of Health.

The initial MIDAS models, released in 2005, focused on an outbreak of pandemic influenza in Southeast Asia, where the first bird flu-related cases appeared in people. The researchers tested the effect of different intervention strategies, including vaccinating people before an outbreak with a somewhat effective vaccine, distributing antiviral medications, closing schools, and quarantining the neighborhoods of infected individuals. They found that a combination of these measures, if implemented early, might contain the outbreak at the source.

The researchers published findings this spring about what might happen if pandemic flu reached other parts of the world, specifically the United States. They developed new computational models (customized programs) to reflect U.S. demographics and transportation patterns. The study results suggest that the rapid production and distribution of vaccines, even those poorly matched to the outbreak strain, could help contain the spread of the disease, particularly if children were given the vaccine early. They also found that travel restrictions alone would not effectively control the spread of the illness.

Longini is glad science is guiding government plans for disease preparedness. "We're pretty much at the cutting edge of driving national policy," he said. "We talk directly to the White House, the Department of Homeland Security and the Department of Health and Human Services, and they're absolutely listening."

"We're interested in the science, not policy," Halloran said. "But we can share our research with the policymakers and help them structure their thinking. For example, if I told you a vaccine is 90 percent efficacious, does that mean that there's a 90 percent chance that I'm 100 percent protected and I'll never get this disease, or does it mean that if I get exposed, my probability of getting the disease is decreased by 90 percent? Those are all things to think about that are quantitatively based."

Predicting the potential spread of an infectious disease requires much more than simply connecting cities on a map. Halloran and Longini champion use of stochastic models, which take into account more real-world unpredictability, as well as many factors about the disease and the affected population. Such models track how a virtual person might behave in a simulated community. Each individual has a chance of catching or spreading an infection through encounters with others at home, work, school and elsewhere.

Disease-modeling simulators

In constructing these models — which Longini compares to creating a recipe —: the researchers start with assumptions about how people interact and how infectious agents spread. Onto this, they add known or estimated information about actual communities and the infectious agent that might emerge there. They can also introduce and evaluate the effectiveness of different interventions, such as vaccination or quarantine. The scientists can modify the models to simulate different situations, such as a more urban community or a more contagious virus.

Because the models are very complex, researchers use high-performance computers — like the computing cluster Halloran and Longini recently had installed — to generate the simulations. The models may run for weeks at a time, producing millions of different possible outcomes. But no single set of results or single model can predict exactly what will happen. As a result, scientists often ask different models the same questions. When different models yield similar results, researchers have more confidence in the predictions.

The research partners plan to create disease-modeling simulators for illnesses, such as HIV and malaria, as well as sexually transmitted diseases, including gonorrhea and chlamydia. "What we're learning about the way disease is propagated through social-contact networks should also help us with acute infectious diseases," Longini said.

Halloran and Longini and their computers are currently supported through the Center's general fund, which is sustained in part by private donations.

Both researchers took less-than-direct paths to their scientific specialty. Interested in infectious diseases since he was a boy, Longini learned about the field of using mathematical, statistical thinking to understand and control infectious diseases while in graduate school. He taught at a university in Columbia and at the University of Michigan. He also conducted research field studies in Latin America and Bangladesh before settling at Emory.

As a child, Halloran planned to be a physicist. "I grew up in the Cold War, thinking the atom bomb was going to drop in my back yard," she said. She studied math and physics in college and then became interested in medicine. She moved to Germany and became a medical doctor, with thoughts of moving to Africa to practice tropical medicine. Returning to the United States, she discovered mathematical modeling in infectious disease while pursuing public-health studies at Harvard.

"I was finishing up my schooling in my late 30s, and I had a lot of degrees but no real job. I was familiar with Ira's methods for analyzing infectious-disease data and somebody said, 'We're looking for someone for Ira to talk to,' and so I went."

Their productive partnership has kept them working together for almost two decades. Today the research partners can practically finish each other's sentences, and their mutual respect runs deep. "We trust each other and share a sense of humor," Halloran said.

"It's hard to find a really good colleague who thinks alike, but also differently. Betz is that person," Longini said. "To discover another person so passionately interested in using mathematical statistical ideas to mitigate and control infectious diseases is a rare find."

Stretched thin by the many demands for their expertise, they look forward to growing their niche at the Hutchinson Center and UW. "We need more people," Halloran said. "We call ourselves a mom and pop shop, but that's why we came here. The number of scientific projects and ideas that we have decreases our ability to actually solve them, so our goal is to get more people involved. It's a great opportunity."