Thursday, February 23, 2012

Science: Peter Palese On The CFR of H5N1



# 6169



One of the livelier `sideshows’ in the vigorous debate over the safety and wisdom of H5N1 lab research has been the argument over just how deadly the H5N1 virus really is.


Although the virus first appeared 15 years ago in Hong Kong that question remains difficult (perhaps, impossible) to answer.


The media has reported, ad nauseum, that the virus kills 6 out of 10 people it infects. Which would give it a CFR (Case Fatality Ratio) about 60%.


But then it gets complicated.


We literally have no good read on how many people the virus has actually infected.  If `mild’ or asymptomatic infections occur, then they are unlikely to be noticed.


A large part of the reason is that surveillance is less than robust in many of the regions where the virus is endemic. Testing – when it is done (and often it isn’t) – has long suffered from sensitivity problems. 


Beyond that, many people who get sick (and even die) in places like Cambodia, Vietnam, or China may do so without ever seeking medical care.


Intuitively we can say that these cases almost certainly exist, but we have no way to accurately measure them. In truth, even the CDC can’t tell you exactly how many people are sickened or die from the flu (or any other cause) each year.


Most scientists would grant that what we see with the H5N1 virus is probably the tip of the iceberg. It is the size of what lies beneath that visible tip that is heavily disputed.


Today we’ve a paper, authored by Professor Peter Palese  et. al., that argues that we are likely missing a great many uncounted H5N1 infections, and that the virus is far less lethal than has been assumed in the past.


This is similar to the argument that Vincent Racaniello  used last month in his blog Should we fear avian H5N1 influenza?


The reason behind this debate – beyond the obvious need to quantify an important aspect of this virus – is the debate over whether this virus is too dangerous to work with in a BSL-3 laboratory environment. 


Relegating it to the highest containment labs (BSL-4), it has been argued, would severely restrict the number of scientists (and countries) who could work on this pathogen



One way to get an idea of the uncounted number of people infected by a virus in a population is to conduct regional seroprevalence studies.


You essentially check antibody levels against a specific virus in a representative group of the population.


Unfortunately, you don’t end up with a clear cut Infected/Not Infected reading from these tests. You get an antibody titer level, and that requires a subjective decision as to what level constitutes proof of a `previous infection’.


Set the bar too high, and you rule out possible cases whose antibodies have declined over time (or who were exposed to an antigenically different H5N1 strain).


Make it too low, and you may count people who were exposed to a non-H5 virus or who received such a low viral load as to not develop illness or immunity. 


Over the years we’ve seen a number of seroprevalence studies on the H5N1 virus, and the results (and methods) have varied considerably.


Palese’s paper appears today in the journal Science (the embargo lifts at 2pm EST), and here he makes his case:


Seroevidence for H5N1 Influenza Infections in Humans: Meta-analysis

Taia T. Wang,  Michael K. Parides,  Peter Palese


The prevalence of avian H5N1 influenza A infections in humans has not been definitively determined. Cases of H5N1 infection in humans confirmed by the World Health Organization (WHO) are fewer than 600 in number with an overall case fatality rate of >50%.

We hypothesize that the stringent criteria for confirmation of a human case of H5N1 by WHO does not account for a majority of infections, but rather, the select few hospitalized cases that are more likely to be severe and result in poor clinical outcome.

Meta-analysis shows that 1-2% of more than 12,500 study participants from 20 studies had seroevidence for prior H5N1 infection.


While stating that we don’t have definitive numbers, Palese writes that if one assumes a 1-2% infection rate among exposed populations, there would likely be millions of people who have been infected by the H5N1 virus.


Palese grants that deaths from the virus may also be undercounted, and calls for better studies (something that I think everyone, regardless of where they stand on this issue, would agree with).


If we assume that Professor Palese is correct in his assumptions, then the H5N1 would seem to be not much more dangerous than some strains of seasonal flu.


A comforting thought.


But one that seems at odds with the argument proffered by many scientists calling for aggressive research into the H5N1 virus because of its serious `pandemic potential’.


Of course not everyone views the available data in quite the same way as does Professor Palese.


We’ll be getting a counter-argument either later today or tomorrow from Michael Osterholm and Nick Kelley of CIDRAP, who have their own analysis coming out in mBio this week (see press release).


Stay tuned.