The Overstated Risks of Censored Bird-Flu Papers -- and Overstated Benefits

The controversy about two censored scientific papers boils down to one question: Do the risks of publishing them outweigh the benefits? A third paper published today argues the risks should be discounted—but it may provide evidence to diminish the benefits as well.

The first two papers, among the most famous nonpublished articles ever, were held back in December because the U.S. government feared they contained the recipe for a deadly pandemic strain of bird flu. They showed that the avian influenza virus, which does not ordinarily pass from person to person, could be engineered to do so. Since the virus, known as H5N1, has an official fatality rate of nearly 60 percent when it does infect human beings, that seemed to the National Institutes of Health like a risky bit of knowledge, and officials put the kibosh on publication.

That infuriated not just the lead authors of each paper, Ron Fouchier of Erasmus Medical College, in the Netherlands, and Yoshihiro Kawaoka of the University of Wisconsin at Madison, but a wide swath of the scientific community.

The brief new paper, published online today in Science, asserts that a fatality rate of nearly 60 percent is a wild overestimate. Peter Palese, the lead author and chair of the microbiology department at the Mount Sinai School of Medicine, in New York City, has fresh numbers to back this up. He points out that the current death rate is based on a few hundred severe, hospitalized cases, the only cases that meet stringent World Health Organization criteria for linking infection to death.

What this fails to account for, he says, are the vast numbers of milder cases in the general population.

He and two Mount Sinai colleagues, Taia T. Wang and Michael K. Parides, looked at 20 studies of more than 12,500 people in several Asian countries, where bird flu is prevalent. They found evidence that 1 to 2 percent of those people had been infected, as shown by antibodies to the virus in their blood, among other things. But the people didn’t die.

“If one assumes a 1-2% infection rate in exposed populations,” the scientists write, “this would likely translate into millions of people who have been infected, worldwide.” Add those to the total case number, rather than using only severe cases, and the death rate plummets and bird flu appears to be a much milder disease.

Mr. Palese has previously argued that a lower death rate changes the whole risk-benefit equation: Lower risks mean the potential benefits—new medications, more accurate warnings of a breakout disease—should take precedence.

But not so fast. Paul Keim, a professor of biology at Northern Arizona University, says Mr. Palese’s conclusions have another effect: They reduce the value of the benefits he touts. Mr. Keim is the acting chair of the National Science Advisory Board on Biosecurity, the federal panel that reviewed the infamous papers and suggested they be redacted.

“It would seem to me that this new definition of infection affects both sides of the equation,” Mr. Keim wrote in an e-mail. “If true, one can argue that a human-transmissible H5N1 virus is less dangerous, but then again it also argues that the beneficial aspects of the research (e.g., vaccine development) are less important too.”

A mild disease that doesn’t harm many people, in other words, doesn’t require an urgent remedy. And so the risks—a terrorist using information in the papers to modify the virus, or the virus accidentally getting out of a lab—may still outpoint the value of publication.

The “new definition of infection” is also coming in for sharp criticism. Michael T. Osterholm, director of the Center for Infectious Disease Research and Policy at the University of Minnesota (and like Mr. Keim a member of the federal advisory board), said that Mr. Palese lumped together different kinds of bird flu to get his high prevalence / low death ratio. “It’s like putting your head in the refrigerator, your feet in the oven, and saying on average the temperature is just right,” he said.

The big error, he said, was including about 2000 cases from a 1997 flu outbreak in Hong Kong together with flu outbreaks after 2004. “They are completely different, with completely different genotypes,” he said. And if Hong Kong was excluded, the overall infection rate drops from the high levels suggested by Mr. Palese down to less than half a percent. That’s what previous studies have shown, Mr. Osterholm said, and it means a higher fatality rate.

Mr. Osterholm also published a paper on bird flu risks today, in the journal mBio, and in it he attacked the assertion that mild cases of infection have been missed. Studies of groups likely to be exposed or infected, like poultry workers, have not detected mild illnesses caused by H5N1, he said.

But the big point, Mr. Osterholm concludes, is that even low death rates are still frighteningly high. “If you took the fatality rate of the 1918 flu pandemic and reduced it by a factor of 20, you’d still have a terrible disease,” he said. So instead of arguing about rates, he said, researchers should be figuring out ways to safely share information about a disease that could do a lot of harm.