Scientists led by virologist Peter Palese of Mount Sinai School of Medicine in New York argue in an analysis published in the online edition of the journal Science that the WHO, a U.N. agency, is calculating the death rate using an estimate of human bird flu cases that is simply too low.

Palese and his colleagues did not offer a specific death rate for people infected by bird flu. But based on figures cited in their analysis, the rate appears to be under 1 percent.

The WHO stood by its calculations and some experts criticized the Palese team's findings, saying they were based on misleading data. As of Thursday, the WHO counts 586 cases of people infected by bird flu. Of those, 346 died, for a fatality rate of 59 percent.

Palese declined requests for an interview, and asked his co-authors not to speak to reporters, according to the Mount Sinai press office.

The important scientific journals Science and Nature are holding off on publishing papers on two experiments that created mutant, more contagious forms of the H5N1 bird flu virus. The delay comes at the request of a U.S. biosecurity panel for fear the research could fall into the wrong hands and be used to create a pandemic that might kill tens of millions of people.

Researchers in the United States and the Netherlands have agreed to a temporary halt to their work. Scientists and public health officials meeting at the WHO last week agreed that the moratorium should remain in place until they can fully assess the risks posed by the research.

Science and Nature have announced their intention to eventually publish the papers in full. The new study could support arguments that fears about the research are overblown.

"There has been a great scare among the public whipped up among the press in the last few months. That needs to be dealt with," Science editor-in-chief Bruce Alberts said last week.

A spokeswoman for the journal said the controversy over bird flu research did not play a role in the decision to publish the new paper. "All Science papers are evaluated on their own merits," spokeswoman Kathleen Wren said. "The question that this paper addresses, namely the prevalence and fatality rate of the virus, is an important one in itself for public health."

Some scientists said there was little coincidence in the timing of the study's publication. They noted that Palese published similar findings last week in the journal Proceedings of the National Academy of Sciences and that it is unusual for Science to publish a paper when key data have appeared elsewhere.

"The editors of Science and Nature are the most powerful people in science," said an influenza epidemiologist who asked not to be named for fear of retribution. "This is the editors of Science saying H5N1's fatality rate isn't 50 percent, so we don't need to worry about a (possible) lab release."

The new findings published in Science also contradict fresh assessments by Michael Osterholm of the University of Minnesota, a member of the National Science Advisory Board for Biosecurity (NSABB). Osterholm cites evidence that the death rate from bird flu is at least as high as the WHO reports.

According to the WHO, the bird flu human death rate ranges from about 30 percent in Egypt to more than 80 percent in Indonesia and Cambodia, WHO spokesman Gregory Hartl said. Hartl said the WHO "is still comfortable estimating a fatality rate between 30 percent and 60 percent" despite the Palese analysis.

"CASE FATALITY RATE"

In the new paper, Palese's team revisited the calculation of the "case fatality rate" of H5N1, which equals the number of H5N1 deaths divided by the number of known human infections.

Palese's team says the WHO's estimate of cases is too low. To count as an H5N1 infection, the WHO requires that a patient have a fever above 100.3 degrees Fahrenheit and that an approved lab confirm the presence of H5N1 by molecular diagnosis.

That does not account for rural patients unable to get to a hospital, infected people who do not show symptoms and other factors, the researchers said. As a result, many more than 586 people may have been infected with H5N1 and survived, they said.

To support that idea, they cited earlier studies assessing the prevalence of H5N1 infection. Based on blood samples from 14,000 people, these studies estimated infection rates of 0.2 percent to 5.6 percent, the researchers reported.

Palese's team settled on 1 percent to 2 percent. If even 1 percent of people in rural areas where H5N1 is rife have been infected, then "millions of people ... have been infected worldwide," he and his colleagues wrote in Science.

In that case, the fatality rate of H5N1 is at least 10 times less than the estimate by the WHO, the researchers said. More precisely, if millions of people have been infected but only 346 died, the death rate would far less than 1 percent, based on figures presented by Palese's team.

A case fatality rate of a few percent would still make H5N1 far deadlier than the most pandemics. Seasonal flu kills fewer than 1 percent of those it infects. The 1918-19 Spanish influenza pandemic that claimed 20 million to 40 million lives killed around 2 percent of those infected.

Dr. Andrew Hayward, a flu expert at University College London, calls the new analysis an "important study."

The people "most likely to be tested for H5N1 are severe cases that end up in hospital," Hayward said. "People who are ill enough to get tested for H5N1 have a very high risk of dying, but this work suggests that many more people in the community become infected" and do not die.

"SIGNIFICANT PROBLEMS"

Other experts expressed doubts about the new study.

"There are significant problems with the analysis," said Arturo Casadevall, chairman of the department of microbiology and immunology at the Albert Einstein College of Medicine in New York and a member of the NSABB. Particularly worrisome, he said by email, is "the inclusion of several studies in the meta-analysis that may not have been appropriate."

One of those studies counted infections from the 1997 Hong Kong bird flu outbreak. That strain of H5N1 is genetically different from later forms and should not have been included in an analysis of fatality rates, Osterholm said.

"The inclusion of the Hong Kong study, which was on a different virus, makes this paper seriously flawed methodologically," Osterholm said in an interview.

In a far different analysis to be published Friday in mBio, the journal of the American Society for Microbiology, Osterholm and a colleague conclude that H5N1 kills up to 80 percent of people it infects.

"The opposite of what Palese suggests actually occurs," Osterholm said. That is, mild and asymptomatic H5N1 infections are counted through scientific surveys of blood samples, but some deaths are missed "because H5N1 infection was not considered, testing was not done, or no testing was available," he said.

If the WHO is wrong on the death rate by a factor of, say, 20? "Even if H5N1 had a 20-fold lower mortality," said Osterholm, "it would still kill more people than the 1918 pandemic."

(Editing by Michele Gershberg and Will Dunham)

• A Permanent Economic Emergency