May 12, 2006
Revisiting a 1979 anthrax release may spawn a better bioterror response
By Dawn Levy
In 1979, anthrax was accidentally released in the city of Sverdlovsk (pop. 1,200,000) in the former Soviet Union, infecting about 80 to 100 people and killing at least 70. Russian officials claimed at the time that tainted meat sold on the black market was responsible; American officials argued that a nearby biological weapons facility released the killer spores. In the early 1990s, Harvard researchers visited the city to piece together the epidemiology of the outbreak. Their investigation, published in Science magazine in 1994, concluded that the Soviet cover story was false.
Now, physicist Dean A. Wilkening, director of the science program at Stanford's Center for International Security and Cooperation (CISAC), has revisited this Cold War tragedy and used its real-world data to improve our ability to model the medical effects of inhalational anthrax. This, in turn, allows him to model more accurately hypothetical scenarios such as the release of a kilogram of aerosolized anthrax in Washington, D.C., today.
The models researchers have used in such thought experiments "predict very different outcomes," says Wilkening, whose work to better understand the human effects of inhalational anthrax was supported by grants from the John D. and Catherine T. MacArthur Foundation and the Carnegie Corporation. Using real-world data from the Sverdlovsk outbreak and from limited nonhuman primate experiments, he was able to eliminate two of four theoretical models currently used in "what if?" scenarios that inform bioterrorism policies ranging from how much medicine we should have on hand in the Strategic National Stockpile to how rigorous post-attack decontamination efforts need to be. He reports his findings in the May 1 issue of Proceedings of the National Academy of Sciences.
"To date, researchers haven't paid enough attention to which model they use," Wilkening says. "Different models can give predictions that vary by a factor of 10 or more, so it matters which model one uses for predicting the human effects of inhalational anthrax." Wilkening aims to anchor models on the best available data and provide realistic models that the bioterrorism community can employ in policy studies.
The Sverdlovsk outbreak is "a sort of natural experiment," he says. "It's a tragic incident, but it also is a very valuable source of scientific data that one can use to distinguish between the four models currently in use." The upshot of his analysis is that two of the models currently in use are not accurate for predicting the human response to inhalational anthrax.
Insufficient data is available to resolve which of the remaining two models he examined is most accurate. That answer will have to await further data from costly nonhuman primate experiments, should they ever be performed (none are planned). "We have to use both [models] right now, or use them as bounding cases," he advises.
Wilkening explored four policy issues that illustrate the consequences of choosing different models: 1) calculating how many anthrax-exposed people would become infected and how many would die; 2) assessing if decontamination would be needed; 3) determining how soon exposed people would show symptoms and how soon doctors would recognize those symptoms as anthrax; and 4) calculating how soon exposed people need to receive antibiotics to avoid contracting the disease.
"To figure out what happens in a bioterrorist event, you need to know two basic properties about the pathogen you're dealing with," Wilkening says. One is the dose-response curve, which determines the likelihood of becoming infected at different exposure levelsthe higher the dose of anthrax you get, the higher the probability that you will become infected. The dose at which 50 percent of an exposed population becomes infected, called the ID50, is around 10,000 spores. The other basic property is the incubation-period distribution, or the time the pathogen takes to grow in the body before symptoms first appear.
Wilkening's study brought dose-dependence to a debate over how long the incubation period is for inhalational anthrax. Published data from vaccine efficacy tests in which nonhuman primates were challenged with high doses of anthraxup to a million sporesindicate an incubation period of one to five days. Data from Sverdlovsk, which exposed people to low doses probably on the order of 1 to 10 spores, indicate a longer incubation period, about 10 days. Whereas previous authors have debated whether nonhuman primate experiments or the Sverdlovsk data should be used to determine the incubation period for inhalational anthrax in humans, Wilkening demonstrates that both estimates are correct, with the difference between them being due to the dose dependence of the incubation period and the very different doses received in each case.
"If you are exposed to a higher dose, there is a much higher chance that an anthrax spore will germinate quickly, thus leading to a shorter incubation period," he says. "Sverdlovsk was a low-dose exposure event and, consequently, one would expect anthrax spore germination to take a longer time, thus leading to a longer incubation period."
Truth and consequences
Russian officials confiscated the medical records of the Sverdlovsk victims and have so far refused to release details of what happened on April 2, 1979. "It would be nice to know exactly what happened, because that would allow us to model the event more accurately," Wilkening says.
Nevertheless, based on weather and other data from the day of the event, scientists think that around 2 p.m. spores, or dormant cells that revive under the right conditions, were released from a military facility, and the Bacillus anthracis spores spread up to 5 kilometers downwind. People breathed in the spores, which geminated and incubated in the body for between four to 40 days before people began to feel ill or show signs of illness such as sore throat, coughing, pains, aches and runny nosethe same symptoms as fluthat indicated they had entered what doctors call the prodromal phase. Within four days, people passed the point of no return, called the fulminant phase, in which toxins from the bacteria had built up to such an extent that people went into shock and died.
It's impossible to save those who've entered the fulminant phase and difficult to save those who've entered the prodromal phase. But if people can start treatment after exposure but before symptoms appear, there's a good chance that they will survivea conclusion Wilkening draws from work by colleagues at Stanford's Center for Health Policy. Treatment primarily consists of antibiotics such as ciprofloxacin, doxycycline or penicillin. While a vaccine to prevent anthrax exists, it is not yet available for the general public but would be made available to people exposed to anthrax, according to the Centers for Disease Control and Prevention website.
In his study, Wilkening ruled out two of the four models because they either did not fit the Sverdlovsk data or the nonhuman primate data, or both. "There are two models that people have used that should no longer be used to predict fatalities, models B and C." (The four models used in his analysis are labeled A-D for convenience.)
Using the two remaining models A and D, he predicted that a hypothetical attack releasing 1 kilogram of anthrax spores in Washington, D.C., would infect between 4,000 and 50,000 people, most of whom would die if not treated quickly with antibiotics. The difference of a factor of 10, Wilkening points out, is "an uncertainty with which we must live for the time being until better data can resolve which of the models A or D is more accurate."
Regarding decontamination efforts, the higher the probability of becoming infected at low exposure levels, the greater the need for effective decontamination, especially for indoor environments. Spores "by nature are hardy," Wilkening says. In the soil, out of the way of sunlight, they can last for a decade. "Residual contamination can be a very serious problem in the wake of an attack," Wilkening says. "Unfortunately, both models A and D predict that residual surface contamination from anthrax spores will be a problem. Consequently, we need to come up with effective indoor decontamination strategies."
Analysts such as Professor Lawrence Wein of the Graduate School of Business are considering the issue. Last year, he assessed decontamination and concluded cleaning buildings to make them safe to reoccupy was a billion-dollar proposition.
In addition, the four models make very different predictions about when symptoms would occur. The day after exposure, they predict between 10 and 1,000 people feeling sick, with more people getting sick in the viable versus discredited models.
"In terms of detecting the outbreak rapidly, this is a good thing because it says that doctors could recognize it [sooner]," Wilkening says.
In terms of treating people before they reach the prodromal phase, however, this is a bad thing because people become sick quicker. Wilkening's analysis may help policymakers reassess how fast antibiotics need to reach people. His best model says administering antibiotics by day three saves 90 percent of exposed people. "Today we cannot meet the three-day requirement," he warns.
A photo of Wilkening is available on the web at http://newsphotos.stanford.edu.