News & AnalysisBird Flu Controversy

Dead Reckoning the Lethality of Bird Flu

+ See all authors and affiliations

Science  17 Feb 2012:
Vol. 335, Issue 6070, pp. 786
DOI: 10.1126/science.335.6070.786

“Peter, you know there's science and there are facts, and you know you can't have your own facts.”

Those fighting words came from Michael Osterholm on 2 February at the New York Academy of Sciences, where he and Peter Palese, both prominent influenza researchers, debated just how deadly the avian virus known as H5N1 is to humans. That question bears directly on the current fracas over H5N1 variants created by two labs that spread easily among ferrets, a model for human transmission (see main text). Palese, a virologist at Mount Sinai Medical Center in New York City, thinks official figures overstate the lethality of H5N1 to humans, exaggerating the risk of the new experiments. Osterholm, an epidemiologist at the University of Minnesota, Twin Cities, asserts that Palese cherry-picks studies to discount the threat. “You misrepresented the data, and that's propaganda,” Osterholm charged.

Bird's-eye view.

Poultry vets (purple) had significantly higher antibodies to avian H1 strains when several dilutions were combined.


The fact is, the facts aren't clear.

Without question, H5N1 kills many of those it sickens but does not spread readily among people. As of 8 February, the World Health Organization (WHO) says, H5N1 had killed 59% of the 584 confirmed cases in humans since 2003. If H5N1 artificially or naturally acquired the ability to transmit easily among mammals, jumped into humans, and remained highly pathogenic, it could trigger what Osterholm called the “worst pandemic” ever seen. Osterholm is a member of the U.S. National Science Advisory Board for Biosecurity (NSABB), which made the contentious recommendation that journals redact key details before publishing the studies to prevent aiding bioterrorists.

Palese says making the details freely available poses little risk and will advance the field. In a 25 January online perspective he co-authored in the Proceedings of the National Academy of Sciences (PNAS), Palese notes that WHO ignores subclinical cases of H5N1, which one recently published study suggested occurred in 45 of 800 people (5.6%) tested in rural Thailand. “Even if only a low percentage of the rural population is asymptomatically/subclinically infected, the case fatality rate that is offered by the WHO—and that is driving this controversy—is likely orders of magnitude too high,” Palese's PNAS perspective argued.

Osterholm noted that several studies contradict the Thai report, published in the 15 October 2011 issue of Clinical Infectious Diseases, which he called “by far the worst one of all.” Epidemiologist Gregory Gray of the University of Florida, Gainesville, who led that study, counters that it's one of the largest and most carefully done serology surveys yet published. “Is it perfect? No,” Gray says. “But it's the best we can do.”

Osterholm focused on the antibody assay used by Gray's group to assess exposure to H5N1. The researchers performed a “serial dilution” that mixes antibody-containing sera with inert liquid at various concentrations and then tests whether they can prevent infections in culture. More dilute solutions can stop virus only if the starting sera contain higher titers of H5N1 antibodies, so “neutralizing” titers of 1:80 are more compelling than 1:10. WHO only ascribes symptomatic cases to H5N1 if they have titers of 1:80; combining published studies of asymptomatic infection with that cutoff, prevalence is 0.48%, Osterholm says. Gray's study used a 1:10 cutoff and found an 11-fold higher prevalence.

Gray says the 1:80 cutoff makes sense in the clinic, when people have acute infections, but not in a retrospective epidemiology study. He notes that he and his colleagues analyzed villagers 2 years after H5N1 was last detected in Thailand, at which point antibody levels likely would have waned. And 1:10 was simply the minimum requirement used to define a person as positive: Titers went as high as 1:40. Their final analysis combined the various titers using a statistical technique called proportional odds modeling, validated in an earlier study of avian viruses (see graph).

Osterholm told the New York audience, “I could find a number of you in this room positive using that study design.” Not so, Gray says. “We've done thousands of these, and very few people have 1:10,” he says. “I'm sure there are some false positives, but it's not an egregious thing.” To more precisely determine subclinical infection rates, Gray says the field needs prospective studies—which he has under way—that capture antibodies from people during and immediately after exposure.

Osterholm says NSABB didn't debate H5N1's actual lethality. “If this virus was 20-fold lower in mortality, it would still be a very, very catastrophic pandemic,” he says.

But what if H5N1 is 1000-fold less lethal than WHO estimates and Palese has a valid point? Osterholm agrees that's possible but contends scientists must err on the side of caution. As he said at the debate, “We can't afford to be wrong.”

Related Content


Navigate This Article