Wednesday, July 1

CDC hoodoo and swine flu lethality

"One contributing factor for death in our patients may have been delayed admission and delayed initiation of oseltamivir [Tamiflu]."

-- From Pneumonia and Respiratory Failure from Swine-Origin Influenza A (H1N1) in Mexico: an original article published June 29, 2009 in The New England Journal of Medicine.

The article is written by a team of physicians at Mexico City's National Institute of Respiratory Diseases (INER); it reports on the clinical and epidemiological characteristics of the first 18 persons with pneumonia and laboratory-confirmed S-OIV infection ("swine flu") hospitalized at INER.

The authors use "may" to qualify their observation about Tamiflu partly because their study only involves 18 patients. And because their methodology was simply to retroactively review medical charts and radiological and laboratory findings on the 18 patients.

However, Tamiflu's life-saving role in fighting swine flu is by now established through anecdotal accounts from physicians and public health officials in the USA and around the world.

The use of antibiotics has also played a life-saving role in curing secondary infections arising from swine flu infection, as the Mexican study also highlights. And yet the CDC continues to avoid publicly acknowledging the implications of the low death count. So at the risk of being tedious I'm going to repeat myself:

From the May 20, 2009 Pundita post titled Stop misleading the public about the true lethality of H1N1/swine flu virus. Memo to CDC, WHO, New York City Department of Health:
Officials at the above-named organizations have deployed a verbal sleight-of-hand in their communications with the public that is as dangerous as the swine flu virus itself:

The officials have announced that the death rate from swine flu infections is low. They have leapfrogged from this fact to the inference that the virus is not particularly lethal. Would that be with or without the administration of Tamiflu to affected patients?

Now let's stop clowning around. Officials do not know for certain how lethal swine flu is. But it doesn't take a lot of data collection to figure out that if a teenaged healthy patient presents with symptoms of a 103 degree and rising fever, and who is unable to move, and who "looks as if she was run over by a truck," as the father of one swine flu sufferer described his daughter's condition, she might not be long for this world without rapid medical intervention. The student in question started a quick recovery after two doses of Tamiflu.

From this, and from many aspects of the data so far, it is very likely that Tamiflu, and not a wimpy version of a swine flu virus, explains the low death rate so far from H1N1 infections.

That also explains the higher death count in Mexico. Money says that many of the Mexico deaths occurred before doctors in Mexico realized they were dealing with more than an ordinary flu virus and thought to administer Tamiflu or Relenza.[...]
Another verbal trick is the claim that the number of deaths in the United States from swine flu is very low in comparison to the 36,000 deaths in the USA from the annual influenzas.

One would think that after hearing health officials invoke the same number, year in and year out, some quick-minded reporter would ask, 'Why is it always 36,000 Americans who die each year from flu? Is there ever a year when far more than that number die from flu, or far less?'

As I've mentioned in earlier posts such a reporter exists. Her name is Kelly O'Meara. At the time of her investigation she was (and I assume still is) anti-vaccine -- or at least anti-flu vaccine. But because she suspected that the CDC was inflating the number of annual deaths from influenza, she was like a terrier with the magic number.

A CDC spokesman, Curtis Allen, admitted to O'Meara that the 36,000 statistic was not a "real" number; it was a computer-generated guess. A wild guess.

Allen told her, "There are a couple problems with determining the number of deaths related to the flu because most people don't die from the influenza...We don't know exactly how many people get the flu each year because it's not a reportable disease and most physicians don't do the test [nasal swab] to indicate whether it's influenza."

However, using data she got from the CDC, O'Meara discovered that "The greatest number of actual influenza deaths recorded since 1979 were 3,006 in 1981." (See this post for citations.)

That doesn't necessarily mean the number of deaths from influenza for those years was that low; it means that the U.S. health system is not set up to accurately record the number of annual deaths from influenza.

It also means the CDC is unable to distinguish between deaths from influenza and deaths from underlying medical conditions in people who've died while suffering from the flu.

What all that means is it's useless to attempt to determine the true lethality of the 2009 swine flu by comparing the number of deaths from the disease to the CDC's magic number of 36,000.

The same holds true for trying to determine the lethality of swine flu by comparing deaths from the disease to annual worldwide deaths from seasonal flu because that annual number is also a computer-generated wild guess.

It is the same for using the number of deaths from the 1918-20 pandemic or any other pandemic in the attempt to determine the lethality of the 2009 swine flu. The number -- the range of numbers -- is a wild guess.

In short, mathematical models that assess the lethality of the 2009 swine flu by using death rates for previous pandemics and seasonal flu deaths are hoodoo.

Yet it's been extremely useful hoodoo, hasn't it? Useful, that is, for governments and the public health systems they fund. One is hard pressed to find a news report or official statement about swine flu deaths that does not conjure the magic numbers; this, to put the number of swine flu deaths in 'context.'

So it's also been a dangerous hoodoo; it's led the public to believe that swine flu is no more lethal than a garden-variety influenza.

And with that belief has come a casual attitude about protections against swine flu and a very negative attitude about epidemic-fighting measures that are inconvenient or a hardship, such as quarantine and self-quarantine.

The hoodoo is losing some of its power now that the death count from swine flu is starting to rise, which brings me to a passage in a June 19 Wall Street Journal report:
[...] Those who dismiss the H1N1 bug, also known as swine flu, as a hoax like to point out the death rate, so far, has been extremely low. Seasonal influenza generally kills about 36,000 Americans a year, while the H1N1 virus has only resulted in 44 confirmed deaths, as of June 12.

But those numbers are misleading, say health experts. They point out that the official death count includes only victims whose diagnoses were confirmed by a special laboratory test. The actual number of infections is believed to be far higher, and should continue to climb as the virus gains traction.[...]
I'm happy to learn that at least some "health experts" are pointing out that the magic 36,000 number is misleading, even though they don't seem to be anxious to reveal the magic trick. Yet I had to read the passage twice before its implications sank in:

I'd fumed because U.S. health departments had quickly ceased testing for all but the most serious swine flu cases, which meant they couldn't provide a good estimate of the number of swine flu cases. But like an idiot I'd assumed that the death count was accurate.

Duh, Pundita, they wouldn't automatically test for swine flu in every person who'd died in the United States since March.

So. We don't even have an accurate statistic for the number swine flu deaths, do we? In that event it would be a help if the CDC, WHO, and all others who publish a swine flu death tally would add "Estimated" alongside the numbers. It'd be a help to the public, that is.


Barnaby Dawson said...

Hmmmm. I agree with most of what you say. But describing the CDC annual seasonal flu deaths estimate as a wild guess is not reasonable in my opinion. Its true the estimates vary massively from year to year and estimates from differing methods often differ by many thousands. Still the order of magnitude here (and 1st significant figure perhaps) is what counts here and estimates from differing methods agree that the annual deaths are in the tens of thousands (and agree reasonably well on the first significant figure).

You're completely right though about comparing estimates from mathematical models (for seasonal flu) to the number of cases (for swine flu) you happen to be lucky and detect.

Indeed I would say that describing the figures as an "estimate" would still be misleading. "Lower bound" would be more accurate. After all the difference in (poorly) detected figures and (ok) estimated figures for seasonal flu is about a factor of 20.

One other thing. Your comments on the Mexican strain might be correct but there has now been a study which suggests otherwise. Several strains of swine flu have been given to ferrets (a good animal model with respect to flu) and most strains had a low lethality but the mexican strain killed 50%. Obviously they're ferrets not people but it does show that the differences between the mexican strain and others may be signficant.

Great post Pundita. Very few observers seem to have picked up this estimation issue.

Pundita said...

Barnaby - Thank you for your comments. Re your comment that a "Mexican strain" of swine flu killed 50% of ferrets in a study -- could you send me your source and (if you're referring to one of the two studies recently published in Science) the exact sentences that refer to a Mexican strain? Thank you.

I will reply at greater length to your observations in a future Pundita post.