The Iceberg Effect

Ron Brookmeyer sounds the depths of epidemiological icebergs for a living.

When a new disease outbreak emerges, one of the biggest concerns for public health officials and the public is the question of whether the patients diagnosed with the disease represent the majority of those already infected. Are those patients it? Or are they just the tip of a huge iceberg of infection that already reaches deep down below the level of clinical detection, waiting to emerge at some point to endanger the lives of many?

The key to the iceberg effect is the disease's incubation period, the time between infection and the first symptoms of illness. Incubation periods can vary widely by disease. For example, the median incubation period for HIV is about 10 years, while anthrax has a median incubation period of 11 days.

Brookmeyer, PhD, professor, Biostatistics, belongs to a pioneering group of scientists who first developed statistical models that allow researchers to apply the incubation period and other factors to a population in order to sketch the true shape and size of an emergent epidemic. 

"Here's a simplified example: The median incubation period is the length of time by which half of the people infected will develop symptoms. If we established that a disease had an 11-day median incubation period, and we knew that there had already been 10 cases at the 11-day point after the exposure, we would then know that we needed to look for approximately 10 more cases," Brookmeyer explains.

He and his colleagues first applied the techniques during the emergence of the AIDS epidemic in the 1980s. By combining the data available on HIV's incubation period with reports of AIDS cases and other factors monitored by physicians, they were able to produce a clearer picture of the true spread of AIDS.

As concern grew over the possibility of a bioterror attack, Brookmeyer joined efforts at the School to help prepare for such an event. His first effort in this area actually took place more than a year before Sept.11, when D. A. Henderson, then-director of the School's Center for Civilian Biodefense Strategies, asked him to take a look at the data available on an outbreak of anthrax in 1979.

That outbreak took place in Sverdlovsk in the former Soviet Union and killed 70 people. At first linked to contaminated meat, the outbreak was later traced to an accidental release of anthrax spores from a nearby Soviet biowarfare laboratory.

Anthrax spores are essentially inert, but after entering the body they germinate into their disease-causing form after a period of time that can vary by days or weeks. The result of Brookmeyer's study was a mathematical curve that shows over time what percentage of a group of people exposed to a sufficiently large dose of anthrax will develop the inhalation form of the disease.

Initially, Brookmeyer says, the work was lukewarmly received. "I gave a couple of talks about the results, and I saw a lot of people getting glazed looks that said, 'What is anthrax again?'" he recalls.

After the anthrax attacks last fall, though, many more people started paying attention. Brookmeyer and Natalie Blades, a graduate student at the School, resolved to use the findings on Sverdlovsk to gain new insight into the U.S. anthrax outbreaks. Their goal was to determine how many cases of inhalation anthrax were prevented by the use of prophylactic antibiotics among the thousands of people most at risk of exposure.

They analyzed eight cases of inhalation anthrax in three clusters: in Florida at a supermarket tabloid, among New Jersey postal workers, and among District of Columbia postal workers. Three additional cases of inhalation anthrax could not be studied because investigators were never able to determine the sources of exposure that led to those cases.

For the eight cases, they had information on exposure times (with the exception of the Florida cluster, where an exposure date had to be estimated because a letter was never found), the time when each case of anthrax began, and the time when prophylactic antibiotics were started.

Based on this data and the anthrax incubation period, they estimated antibiotics had prevented nine cases of potentially fatal inhalation anthrax. Their results were published in the March 8 issue of Science.

Brookmeyer thinks statistical modeling techniques can help scientists prepare for and respond to bioterror attacks in several ways, including assessing potential responses in advance in simulations of attacks, predicting the size of an outbreak based on early data, and weighing the potential hazards of prophylactic treatment against the risks of infection.

He says it might even be possible to use statistical modeling to help officials zero in on the tip of the iceberg: an as-yet undetermined source of an emergent outbreak.

"If we can help pinpoint the exposure date, then we can help identify other people who may have been around the cases at the same time and were therefore exposed and in need of treatment," he explains, adding, "Of course, helping find an exposure date may also help find the perpetrator."

How well known are the incubation periods for other potential bioterror agents like smallpox and plague? Thomas Inglesby, deputy director of the Center for Civilian Biodefense Strategies, says incubation periods for both pathogens have been previously established.

"For smallpox, the range of the expected incubation period is about seven to 17 days, and for plague it's one to six days," Inglesby says. "It's not clear, though, how weaponization will affect those figures. Adding some form of carrier to a smallpox powder, as was done in the anthrax attacks of 2001, or finding a way to deliver a particularly large dose to victims might alter the incubation period."

Brookmeyer is running simulations to see how scientists can better account for sources of uncertainty like dosage, ages of those exposed, and the suppressed immune systems of HIV patients, transplant recipients, and others.