Thursday, August 04, 2005

On the mirage of stopping bird flu

It pains me to say some work of people I know and respect is a load of crap, but it has to be said in this instance. It is widely reported (e.g., BBC) that a bird flu pandemic is stoppable "if governments work together." Whether a pandemic with H5N1 (as opposed to another influenza A subtype) happens or not is no longer within the control of any government (if it ever was). I (and others, notably Henry Niman) have said this before, but it must be said again: once this virus gains the capability of being transmitted from person to person like other influenza subtypes that circulate in human populations, there will be no way to prevent its global spread. The contrary idea, as reported in this week's scientific journals Nature and Science, is so heavily qualified it might as well have never been made in the first place.

Indeed the requirements for the claim to be valid essentially make this a "trans-scientific proposition," i.e., one that can be phrased in scientific terms but not practically carried out, like weighing the moon on a balance.

Here are the conditions required to "stop" an outbreak that begins in Thailand, according to two research teams, one in England, the other in the US:
Firstly, the virus would have to be identified while confined to about 30 people, they told Nature.

In addition, antiviral drugs would have to be distributed rapidly to the 20,000 individuals nearest those infected.

They estimate an international stockpile of three million courses of the treatment would be enough to contain an outbreak.

But it would mean having to deploy the drug anywhere in the world at short notice.
Another team from Emory University in Atlanta, the US, led by Dr Ira Longini, simulated an outbreak in a population of 500,000 in rural Thailand, where people mixed in a variety of settings, including households, schools, workplaces and a hospital.

Provided targeted use of antiviral drugs was adopted within 21 days it would be possible to contain an outbreak, they found, as long as each infected person was not likely to infect more than an average of 1.6 people.

If it was more infective than this, household quarantines would also be necessary, they said.
These expressions of hope are based on some rather shaky foundations, computer models with many assumptions built in (among them an R0 of 1.6 in the Longini model, likely an underestimate). I am not knocking computer models. I construct and use them myself. But they shouldn't be used in circumstances where the margin of error is as slight as in this case and they shouldn't be interpreted to provide reliable assurance things would evolve in the way predicted. They may be useful for understanding the qualitative dynamics of disease spread, but in this case we are too close to the sensitive points (the bifurcation points in the parameter regimes) to have any confidence in the results.

Even if the models were valid (a big "if"), just a glance at the initial conditions above would show the futility of relying on this strategy. Early detection, a relatively small cluster, rapid distribution of an antiviral stockpile that doesn't exist, all in a three week period in a rural area with little public health infrastructure and porous borders. And that would be the minimal requirement.

So I hate to disagree with Elizabeth Halloran of Emory, an infectious disease epidemiologist of note and a genuine expert. In her view, as reported by the BBC,
"Our findings indicate that we have reason to be somewhat hopeful.

"If - or, more likely, when - an outbreak occurs in humans, there is a chance of containing it and preventing a pandemic."
There is no reason at all to be hopeful. If the pandemic never materializes it will be from dumb luck or for biological reasons we as yet don't understand. The task at hand, as I have noted in many other posts, is to prepare to manage the consequences, not hold out false hope that anything we do can avoid them altogether.