Debora MacKenzie, reporter
How deadly is bird flu? In public, everyone - including reporters like me - repeats the World Health Organization's numbers: currently 587 "known" cases, of whom 346 have died. That makes a case-fatality rate of 59 per cent.
That's about twice the estimated case-fatality rate of the Black Death. That's why everyone is scared H5N1 will become easy for humans to catch and transmit - and why there has been such an outcry over recent research in which H5N1 became transmissible by aerosol in ferrets. Now that death rate is being questioned.
Peter Palese and colleagues at the Mount Sinai School of Medicine in New York think H5N1 may not be much more deadly than normal flu. Other epidemiologists question that conclusion. Such uncertainty about the H5N1 death rate is not new: I've been hearing about it in the bar at flu meetings for years.
The case fatality rate is a fraction: deaths divided by cases. The more cases there are - the bigger the denominator - the smaller the rate. The WHO only accepts a case as official if it's serious enough to go to hospital and get tested. If there are a lot of mild cases, they're being missed - and the death rate is lower, maybe a lot lower.
For that reason, ever since H5N1 started its global spread in 2004, researchers have surveyed people who were exposed to the virus to see if they have antibodies to it - evidence they were infected. Palese reviewed those studies, and concluded that, on average, between 1 and 2 per cent of exposed people get infected but not seriously ill. It's unknown how many people worldwide have been exposed, but Palese says it could be millions - making the denominator on that fraction very large, and the death rate maybe no worse than ordinary flu.
In response, Michael Osterholm and Nick Kelley at the University of Minnesota in Minneapolis also reviewed the existing studies - but excluded those involving the somewhat different H5N1 that hit humans in 1997, and a problematic Thai study. Only three of 13 studies found clear evidence of infection in exposed people, and those found it in only 0.5 per cent - only 26 people.
Meanwhile, they say, following exposed people to see if they got sick has not turned up any cases. And we are probably missing serious cases, too. They figure the death rate for H5N1 is unlikely to be massively lower than 59 per cent - but stress that even if it is 20-fold lower, that would still be worse than the 1918 flu pandemic, which killed tens of millions of mainly young people.
What neither group mentions is that like H5N1, the swine flu pandemic of 2009 was an animal flu that hadn't adapted our immune systems yet - and it behaved very oddly. It was mild or even symptom-free in many, but really bad in some ?- of whom quite a few died. Maybe that's happening with bird flu.
The real bottom line, I think, goes to the Minnesota group. We need to work out how to study the pandemic potential of this virus and share the results safely, they say - and also, how to make sure it doesn't escape from a lab while we're doing it.
Yes, we do. And we need to face up to the fact - as the Minnesota team also points out - that if it does go pandemic, our current vaccines and drugs will be little help. That's what needs our attention. Now.