A simple math problem lies at the heart of a heated debate over whether scientists should be allowed to publish provocative research into the transmissibility of H5N1 flu. Assuming the avian virus could spread easily among people, just how deadly would an H5N1 pandemic be for humans?

Flu scientists tend to shy away from that question, suggesting that it is not possible to predict how lethal the virus would still be after undergoing the necessary changes to adapt to human physiology. But inevitably, people look for clues to what appears to be the best predictor of the virus's future path—its current behavior. And that appears downright terrifying: as many as 59 percent of people known to have contracted the virus have died from the infection.

More specifically, of the 584 people who have tested positive with what the World Health Organization (WHO) confirms is H5N1, 345 have died. (These numbers are current as of February 8, 2012.)

But what if H5N1 isn't as deadly as the official numbers suggest?

Indeed, two researchers have charged into the already fraught H5N1 publication controversy insisting the numbers are wrong, that the true mortality rate is likely to be much, much lower and that bad policy is being driven by the inflated figures.

Peter Palese, a noted influenza virologist at Mount Sinai School of Medicine in New York City, and Vincent Racaniello, a professor of microbiology at Columbia University Medical Center, also in New York City, are among a vocal group of scientists who vehemently oppose any decision to suppress the details of research conducted by Yoshihiro Kawaoka of the University of Wisconsin–Madison and Ron Fouchier of Erasmus Medical Center in the Netherlands.

Fouchier and Kawaoka had—at the request of the National Institutes of Health—figured out whether the H5N1 virus could become more transmissible in non-avian species. Their efforts reportedly revealed that just a few mutations were all that was needed to create a bird flu virus that is easily transmitted between ferrets. In addition, Fouchier said that his strain remained just as deadly to ferrets as it had been to birds, although Kawaoka later declared that his lab strain was not lethal.

Palese suggested in a perspective article co-authored by Taia Wang and published ahead of print on January 25, 2012 in Proceedings of the National Academy of Sciences that the case fatality rate of H5N1 human was almost certainly "orders of magnitude" too high.

Starting with the current 59 percent rate, if you start pushing the decimal point left, 59 becomes 5.9, which becomes 0.59 or even 0.059. Each adjustment of the decimal corresponds to an order of magnitude. (For comparison's sake, the mortality rate of current seasonal flu is less than 0.1 percent whereas researchers estimate that the mortality rate of the killer 1918 flu pandemic was around 2 percent.)

Racaniello, who did his thesis research under Palese, suggested on his popular Virology Blog in early January that the estimates of H5N1's killing potential were vastly overrated. Citing a recently published study that found what might be H5N1 antibodies in the blood of some villagers in Thailand, he mused that if 9 percent of rural Asians had antibodies to the virus, the perception of how dangerous H5N1 is would change dramatically.

In the flu world, few people would argue that Palese and Racaniello are wrong that the case/fatality rate is too high. It might be difficult, though, to find many who agreed with their conclusion on what that means about the virus.

It is widely accepted that the cases that come to light and get tallied by WHO are only an unknown portion of the total human infections that have occurred. Official case counts are certainly missing some infections—but not enough to morph H5N1 into a benign virus, a number of flu scientists agree in interviews for Scientific American.

"I think all these numbers are flexible, and Peter is undoubtedly right it's not 60 percent. But I don't know what it is. And I don't think he does either," says Robert Krug, chairman of Genetics and Microbiology at the University of Texas at Austin, where his work focuses on the molecular mechanisms at play during influenza infection.

"It's dangerous. How dangerous? I have no idea…. I'm sure it's less than 60 percent but it's still too high for the world to tolerate a (human-to-human) transmissible H5N1 virus," says Krug, who believes both papers should be published in full.

The problem with the case/fatality rate, as Palese pointed out in PNAS, is that human infections with what is still a bird virus generally only come to the attention of medical authorities when someone gets really sick. In fact, in order to count as a case by WHO's definition, a person must have a high fever, known exposure to the virus, and needs to test positive for H5N1. A specimen for a test would generally only be taken at a hospital and that facility would have to have access to a laboratory. If H5N1 is causing mild cases, they are unlikely to come to light under that definition. Is a person living in a remote Cambodian village who feels lousy for a couple of days going to seek that kind of medical care? If there are H5N1 cases like that, the fact they are being missed artificially lowers the denominator.

"If the only cases you know about are the ones who are going to die, then you might believe that the case/fatality rate is very high because you lack surveillance of less symptomatic cases," says John J. Treanor, chief of the Infectious Diseases Division at the University of Rochester Medical Center in New York State.

But what of the numerator, or the number of deaths? For the case/fatality rate to plummet, the numerator must be a smaller fraction of the total cases. But it is clear the numerator is off as well, notes Tim Uyeki, an influenza epidemiologist at the U.S. Centers for Disease Control who has spent a lot of time in the field studying human H5N1 cases and outbreaks.

Uyeki points as an example to the first report in the scientific literature of presumed person-to-person spread of H5N1. It was a cluster of three infections that started with an 11-year-old girl who fell ill in September 2004. She lived with an aunt while her mother worked in a distant city. Both the aunt and the mother, who came home to care for the girl, got sick; the mother and daughter died.

All three clearly had H5N1—a throat swab confirmed it in the aunt and virus was found in tissue from the mother. But the hospital had thought the girl had dengue fever. By the time they realized these were H5N1 cases, the girl had died and her body was cremated. Officially that cluster went down on the books as two cases, not three. There are other cases that were designated as probable infections but which never made the official count, Uyeki says.

Given the limitations of the system for finding human cases, researchers have been conducting what are known as sero-surveys—drawing blood samples from groups of people who were likely exposed to the virus to see if they have antibodies specific to it. That would be a sign that they had been infected and survived. More than 20 such studies have been completed since 1997, when the first known cases of H5N1 infection in humans cropped up. Groups that have been tested included workers who culled infected chickens, health care workers who cared for H5N1 patients, people who worked in live animal markets and people who lived in villages where cases have occurred. The studies have been done in China, Indonesia, Nigeria, Cambodia, Thailand and elsewhere, important because different subfamilies of H5N1 viruses circulate in different parts of the world and some—hypothetically—may cause more severe disease than others.

Most of the sero-surveys have been small; few have contained more than 500 people. Whereas one study—among poultry market workers in Hong Kong in 1997—found around 10 per cent had H5N1 antibodies, most reported either no positives or low rates of people with antibodies. Some were under 1 percent, two were in the 3 to 4 percent range.

The study Racaniello drew on to argue H5N1 infection was more prevalent (and thus less lethal) than official numbers suggest looked for evidence of antibodies in 800 Thai adults living in villages where outbreaks of H5N1 had occurred in birds and where at least one human infection had been reported. The researchers found 5.6 percent had elevated antibodies to one H5N1 virus and 3.5 percent to another.

Not everyone agrees, however, that this particular study can be used to support Racaniello's argument. The threshold used in the Thai research as evidence of antibodies is substantially lower than most studies use. With a cutoff that low, says Malik Peiris, chair of the Department of Microbiology at the University of Hong Kong, one cannot be sure whether what is being detected is antibody-specific to H5N1, or antibodies to other flu viruses that happen to cross-react with the H5N1 test. Having low levels of antibodies that react to—and might even protect against—H5N1 does not prove that the person was infected with H5N1, Peiris says.

The senior author of the Thai study, Gregory Gray, chair of the Department of Environmental and Global Health at the University of Florida, says his group used the low threshold because they know antibody levels wane over time. They were looking for "subtle evidence" of infections that might have occurred years previous. But Gray says the results should not be overinterpreted. "It is a stretch to say this is population-based and also a stretch to say these all represented H5N1 infections," he says.

Although Krug, Treanor, Uyeki and Peiris all agree the official 59 percent H5N1 case fatality rate is not the true number, none takes much comfort from the fact. Krug is agitated that the controversy over the studies is drawing attention away from their key message—this virus can adapt to spread in mammals, which may include humans. And Treanor scoffs at the idea that concern over H5N1 is overblown. "If H5 is not dangerous, why are we even bothering to study it at all?" he asks. "I think it is without a doubt the case that it is not as dangerous as it looks from the cases that we have. But it is still without a doubt an extremely dangerous virus—particularly if it gained the ability to spread from person to person."

As for how far off the case/fatality rate is, there is no way of knowing. Uyeki, who has studied the issue at length, gives his estimate: "Are we missing some [cases]? Yeah, probably we're missing some. But are we missing hundreds of thousands? No, I don't think so. Are we missing tens of thousands? Probably not. Are we missing hundreds? Possibly. It's really hard to know."