…That’s what I would be saying, from actual sources instead of the mainstream news, if I believed that a sample size of 30 was likely to give statistically significant results.
Anne Schuchat, M.D., interim deputy director for the CDC’s (Centers for Disease Control) Science and Public Health Program, is on it. AAFP quotes her as saying:
“We were surprised by the frequency of obesity among the severe cases that we’ve been tracking. I do think it’s an important result. The question of whether people with obesity need to be treated differently in terms of anti-viral treatment or seasonal flu vaccinations is one we’re looking into.”
Isn’t that a nice little ambiguity? Did she mean that she thinks obesity is protective, or a risk factor? In what way does she think should the medical profession be looking to treat obese people differently?
Forward to the CDC Telebriefing on Investigation of Human Cases of H1N1 Flu:
“If there truly is an increased risk of severe complications on obese patients, it would be important to take steps to attend to that. One unfortunate statistic right now is that the U.S. is experiencing an epidemic of obesity. We had much higher rates of obesity in the U.S. than we had 10 or 20 years ago. Both in children and adults. So it?s hard for us to say at this point to say whether the number of patients with reported obesity is significantly higher than we would expect. It was a bit surprising on first glimpse. A lot of theorys people are entertaining now. Some people with morbid obesity have a syndrome where they have fairly severe respiratory compromised just based on the extra weight they?re carrying on their chest.”
OK, so Schuchat thinks The Fatties get the flu bad, real bad, because fat lungs don’t work properly, because they’re surrounded by evil suffocating chest fat.
Let’s look at the figures, shall we? According to CDC data, the population of California has an obesity prevalence of 22.6%. So what’s the frequency of obesity among severe, hospitalised H1N1 flu cases in California? 30%? 50%?
Schuchat was commenting on this Morbidity and Mortality Weekly Report (MMWR), Hospitalized Patients with Novel Influenza A (H1N1) Virus Infection — California, April–May, 2009. This MMWR looked at the characteristics of 30 people hospitalised for severe H1N1 flu infection.
How many of the severely ill were obese? 4. FOUR.
That’s 13%. That’s 41% LOWER than expected, if people were sickening randomly with severe swine flu.
So, I ask, CDC: What the fuck?
And, science media, how about reading sources yourself? I found this with a two-minute Google search. But each and every one of you has decided that Teh Fat Causes Swine Flu. This is primary-school mathematics FAIL, folks.
~~~
Update 19 June 2009: CBC.ca reports one more datapoint, in “Risk factors for severe swine flu may explain disease’s enigmas: experts”
Four of the 23 New Yorkers who died were obese.
That’s 17%. Obesity prevalence in New York is 25%. The journalist goes on:
Still, Harper says it’s too soon to say whether that’s a risk factor in and of itself; the U.S. national obesity rate is about 25 per cent, or roughly six out of every 23 people. It may be that some of the things that go hand-in-hand with obesity — like early heart disease or diabetes — are the real risk factors.
*headdesk*
Categories: health
Soggy here, innit? 
The Wrong Conversation? No, Doctors Don’t Need to Mention Weight Loss More 
Pamela Denoon Lecture 2016: Nelly Thomas 
Here’s 200 things to keep me busy for the rest of this year 
Hoyden About Town 
Ooh, nice catch. Demonstrates how habitual the assumption that obesity is the big bad has become amongst health practitioners! Which of course leads to the moral condemnation of those who are obese as responsible for their own ill health (but never their own good health), unlike the slender members of the population (who obviously are clearly just unlucky if they fall ill). How sad that they can’t wander around saying ‘you brought this [swine flu]* on yourself, evil fatty!’ anymore. (Not that that’ll stop them.)
Yeah, I know, a tiny sample size and all, but still! [grumps at moralism of medicine].
* insert almost any health problem here.
It sure won’t stop them. Because now the headlines have read “Obesity Increases Swine Flu Risk”, and the CDC has said “Obesity Increases Swine Flu Risk”, and that is all the vast majority of people, including healthcare workers, will read. How many doctors do you think critically read MMWR figures instead of pre-digested soundbites? Have you seen ANY media outlet debunk this crap?
Maths is fun – but probably takes longer than just quoting what everyone else said because it’s OBVIOUSLY true! Of course, there’s always the problem of overweight and obese people being less likely to seek care – because we know we’ll just be told “eat less and exercise more”, not matter what’s actually wrong.
But won’t somebody think of the media outlets? If they don’t blame fat people, who will they blame? /sarcasm
“If there truly is an increased risk of severe complications on obese patients, it would be important to take steps to attend to that.”
I don’t have a problem with this statement. If obesity, for some medical reason not something you just made up, makes people with swine flu sicker then it’s something that needs to be addressed. The fact that there is absolutely no evidence supplied to show that this is actually the case is the problem. It’s all one big IF. Sure, you claim an obesity epidemic etc etc, but if that doesn’t actually make any difference to who gets sick and how badly, then why are you making a fuss about it CDC?
There’s also the fact that we never seem to hear “If there truly is an increased risk of severe complications on slim patients, it would be important to take steps to attend to that.” I can’t get over the fact that supposedly reputable public health agencies have looked at this data and just completely got it backwards. I feel a bit like I’m missing something, but I’ve been over it a number of times, and can’t see what.
But as you pointed out 30 isn’t a proper size sample, so we actually don’t know from those figures whether obesity is a risk factor or not. Anne Schuchat maybe commenting about anecdotal evidence she’s seen, or she maybe talking rubbish.Who knows?
It’s a two point thing, isn’t it?
1. 30 isn’t a sufficiently large sample
2. even if it was, obese people are underrepresented, not overrepresented, in the figures as they stand
Fine: I’m not sure where you’re getting that level of uncertainty from. The Telebriefing transcript I linked, though poorly transcribed, makes it clear that she was talking about the California series of 30, and calling obesity “an underlying medical condition that would put them at higher risk for influenza” in that context.
As far as I can see, the CDC has presented no other data on H1N1 and obesity. If anyone finds some, please let us know here in comments.
The level of uncertainty comes from the insufficient sample numbers, which you pointed out in the post. So, maybe it comes from some other evidence she hasn’t presented. Or maybe there’s no evidence whatsoever.
In the UK the government has been falling over themselves to point out that everyone hospitalised for swine flu over here so far has “an underlying medical condition” to minimize panic, though I don’t know whether that is meant to include obesity (I wonder if assumptions are being made based on the higher death rate in Mexico and the weight profile of people there). The woman who just died in Glasgow was definitely overweight … but that might have something to do with giving birth a fortnight ago…
Even if you bought the argument that teh fat is a) a medical condition and b) increases your susceptibility to H1 N1 what exactly would they expect to do about it? To be considered obese you’d probably be, say, three stone overweight (19.5kg) or more. We already know that it’s not healthy to lose more than half a kilo per week so it would take 39 WEEKS to eliminate the risk, and even then most people would be instructed to lose the weight by restricting their diet which will have a negative knock-on effect on their immune system.
Pass me the cheesecake, I’m not scared.
But that’s because slim=normal, of course! So you don’t have to take steps to attend to an increased risk of severe complications on slim patients, cos it’s not an increased risk, it’s a normal risk, dontchaknow. So you just do whatever you normally do.
/snark
*sigh*
You know, I thought you were going to say that based on a sample of 30, they were going to deny vaccine, or restrict vaccine, to fat people. I can’t even begin to imagine how anyone could conclude an increased risk from that data.
I guess fat people are at higher risk has become the null hypothesis. Both invalid, and unacceptable.
A little statistics lesson here. Just because something has a small sample size doesn’t mean that the results can’t be statistically significant. If you have a small sample, it just means that your estimated effect has to be many more standard deviations away from the mean in order to believe that the effect isn’t just random. A sample of 30 is reasonably large and for that sample size, an effect only needs to be 2 standard deviations away from the mean in order to be considered statistically significant. You may disagree with the premise of the study, but based on what has been reported, there’s nothing wrong with their statistical analysis.
I also think you may have misinterpreted the intent of looking at obesity. The CDC is trying to identify other health conditions that are comorbid or make someone more prone to swine flu. A number of the people who had swine flu had other medical conditions. I think that they are using obesity as a bad proxy for other conditions like diabetes or heart disease. On the other hand, weight and height are probably one of the few data points that they could get for all of the patients in their sample.
NAWFrance:
What statistical analysis? Can you point to it?
Again, the prevalence of obesity in the hospitalised sample was LOWER than would be expected by chance alone. I don’t know whether it’s statistically significantly lower, because no analysis has been published. If you can do a back-of-the-envelope on this particular raw data set (preferably with an explanation of why the particular test you choose is appropriate for this particular data set), we’re all ears!
I don’t “disagree with the premise of the study” (which appears to be nothing more than a case-series); I argue that the conclusion drawn by the CDC spokesperson is the exact opposite of what the data is suggesting.
What leads you to that conclusion? They have data right there on diabetes, cardiac disease, chronic lung disease, immunosuppression, seizure disorders, and pregnancy. I’m not sure where you’re getting them using obesity as a proxy for co-morbidities.
If they had used obesity as a proxy for, say, diabetes (and diabetes data was absent), then looked at the data in good faith: what they’d be finding is that there is no evidence that diabetes raises the risk of severe H1N1 flu, that very preliminary data suggested that it might be protective, and that this issue deserved further investigation.
A little statistics lesson here. Just because something has a small sample size doesn’t mean that the results can’t be statistically significant.
And to add to what Lauredhel has already pointed out, statistically significance is irrelevant if the sample isn’t representative. There are a raft of reasons why any given group may be over- or under-represented in such a study (and I use that word advisedly), and without a scrap of analysis, it is impossible to say anything much about representation. Otherwise, as Lauredhel said, we’d be shouting that obesity was protective for severe swine flu.
My guess is there’s underlying data they haven’t published that they’re bringing to that statement in the AAP article.
The figure that isn’t given in either the MMWR or CDC briefing is the chronic condition breakdown of the people who are infected with the virus, in addition to those who are hospitalised. Without those figures I don’t think it’s possible to make a call on their maths.
Their call on obesity as a significant risk factor would be most properly based on the hospitalisation rate of infected patients.
With respect, Lauredhel, I think you’ve made an assumption, in quoting the background rate of obesity, that the population of infected patients reflects the same demographics as the general population. That’s not necessarily the case and without those figures (which I’d guess the CDC probably has) it’s not really possible to draw a conclusion on the reliability of their statement about obesity significantly increasing risk for complications in H1N1.
Their call on obesity as a significant risk factor would be most properly based on the hospitalisation rate of infected patients.
Do you think so? I guess it depends on what you want to measure. Surely if obese people have lower infection rates in the first place, that impacts on the risk factor of obesity?
And I don’t think Lauredhel has made any assumptions at all – at least she hasn’t drawn any conclusions, so I can’t see how she has made assumptions. She has just questioned the conclusions drawn in the article. And no-one has any business publishing conclusions without publishing data.
BTW, the back of the envelope calculation with an assumption of a SD of 10 (with no idea about variance, it is a reasonable guess for a number than can range from 0 to 100) and 29 degrees of freedom, the result is actually significant. Less that 0.01, possibly more – the t statistic comes out at 5.2ish, with 0.05 significance at 2.1ish. As I said though, with no concept of the representativeness of the sample, that is pretty meaningless.
Surely if obese people have lower infection rates in the first place, that impacts on the risk factor of obesity?
Sure, but what the CDC is talking about is what to do about it in public policy.
So, for taking extremes to illustrate my point, if you have (from the 330 or so confirmed cases) 300 obese patients and only four hospitalised, you may be able to argue that obesity is an increased risk factor for contracting H1N1, but not for hospitalisation.
If, on the other hand, you have only four from 330 confirmed cases and all four are hospitalised you have completely different set of circumstances. Both of those would impact on the recommended treatment and vaccination regimes – but presumably in different ways.
And no-one has any business publishing conclusions without publishing data.
I don’t think that’s the case, particularly in developing pandemic situations – the good epidemiological studies are going to be months and months away, and peer reviewed ones are even further (though, to be fair, I haven’t had more than a cursory look for data on infection demographics, and there may well be published info our there).
The CDC and WHO are working with what they have and giving the best advice they can, given a developing situation. I also think my opening comment – My guess is there’s underlying data they haven’t published that they’re bringing to that statement in the AAP article is a fair one.
But how do you know that obesity is the causal factor? What if the patient’s obesity is linked to a chronic lung condition. Why wouldn’t you identify the lung condition as the cause rather than the obesity. She hasn’t said whether these patients were only obese, with no other illnesses. I’d be pretty surprised if that were the case.
The data for the case series are all there in the MMWR I’ve linked. Pulling out the four:
– 7 yo with asthma and obesity, hospitalised 4 days with asthma exacerbation
– 30 yo with diabetes (doesn’t say which type) and obesity, hospitalised 1 day for viral syndrome & vomiting
– 40 yo with asthma, hypertension, and obesity, still hospitalised 18D at the time of MMWR release, pneumonia and respiratory failure
– 41 yo with autoimmune hepatitis/biliary cirrhosis s/p liver transplant, HTN, obesity, hospitalised 6 days with viral syndrome
Crankynick: I’m big on “show me the data”. As you’ve read before on this blog, mainstream media and very often authors and organisations themselves are not infrequently not very good at interpreting data, when that data relates to existing societal prejudices. Since I started doing this, I’ve frankly been quite surprised at just how bad they are; I started out with far more faith in the process than I have now.
I read all soundbites with a critical eye (and mix my metaphors with a gleeful wave). I don’t posit more supportive data that they’re just not telling me about without a good reason to posit those things. And I think this is mainly because my underlying assumption isn’t “they’re correct” any more.
And to expand a little with a gratuitous kid story: one day our six year old was at his dad’s lab, and the boss, a scientist (bio sciences), heard the Lad say “Well if I can’t see it, I don’t believe it!”. The boss has reserved him a job for the future, and has been quoting him ever since.
I think he gets it from somewhere.
And I think this is mainly because my underlying assumption isn’t “they’re correct” any more.
Sure, and I’m not taking issue with the underlying point – that’s certainly what happens, and that underlying data should be questioned.
As a (partly) science media journalist myself I’m not going to take issue with criticisms of my profession on that front – with exceptions for a few outstanding journos, the criticisms of the way that science is dealt with in the media aren’t unfair. I’ve been guilty of a few horrendous fuck-ups myself.
I probably would take issue with you concluding comment, though – And, science media, how about reading sources yourself? I found this with a two-minute Google search, particularly in this case.
I’d argue that the journo who wrote that report should very probably have gone back and asked those questions, and incorporated the responses into their article (if they had got responses, which is by no means certain).
From my perspective, I think the journo would have been making a serious mistake if he had gone and done a similar set of figures themselves, as you did, and incorporated them into the piece. I certainly don’t have the skill set to conduct even the most basic epidemiological research, and you risk getting the shit kicked out of you by the professionals if you try.
What I (mostly occasionally sometimes) have developed is a skill set to catch obvious data holes and question them – constructing something meaningful on the back of that is much more difficult, though.
Doing the in depth analysis of that kind of thing is one of the strengths of blogs, btw – I think sites like Bad Science and Orac (and a raft of others), should be on the morning reading list of every science/health journo everywhere. I’ve certainly learned more from them than I did at Uni, or from most other sources.
And that was a little OT towards the end, so my apologies.
(it’s on-topic enough for me!) Definitely. Yetm look at what the MSM and blogs have had to say about this (not exhaustive or carefully representative, just yoinked straight from Google results):
And there are a whole lot more that say things like “Swine flu hits sick people the hardest”, including obesity uncritically under the category of “sicknesses” – which is a whole nother, but connected, issue. While I think blogs have potential along these lines, in this particular issue, I think they’re (we’re) failing.
One exception: “Obesity and H1N1” at Pandemic Flu Watch.
And here’s my favourite Headless Fattie from that search!
I’ve seen three commenters (of many), and no other bloggers/journalists, raising questions about the data. (If anyone finds any, please point them out in-thread.) Most of the comments I’ve seen are sneerings along the lines of “Well, we all KNOW fat is unhealthy! This is just more proof!” and “Another reason why the overweight and obese should pay more for their health insurance!”
No argument with the above – that’s one of the major problems with press conference reporting, sadly: the wire service journos who do the pressers don’t have time or (necessarily) the ability to spot check claims or find the holes at the time, and when it goes out with a header like that it’s almost inevitably picked up unquestioned and re-run.
It’s at that point that journos should be questioning the data, but too often don’t.
Just going back to my comment earlier on, though, about public policy decisions: these two WSJ articles give an insight into the decisions that are going to need to be made on the back of these risk factors.
A background obesity prevalence of 22.6% in California would mean that, if obesity were a priority target for vaccination, 8.2 million vac courses would be needed to cover that indication alone – that’s around $US80-120 million, assuming a per dose cost of $US10-15.
These are decisions involving big money for public health authorities, which means I tend to assume there’s some background data there to justify it (and yeah, feel free to describe me as hopelessly naive for continuing to assume that public health decisions would actually be based on solid data – my response to that would be that I’d assume the figures would be done, even if they don’t pay attention to them).
CrankyNick wrote:
Except Lauredhel didn’t assume anything. She pointed out the fact that the data they presented shows less-frequent severe illness among obese people than among slim people. Nick, on the other hand, reaches his conclusions by assuming all over the place.
CrankyNick, eat your words.
“I’d argue that the journo who wrote that report should very probably have gone back and asked those questions, and incorporated the responses into their article (if they had got responses, which is by no means certain).”
Journalists? Ask questions? That’s so last century. All journalists do these days is cut and paste media releases, surely?
OK, let me rephrase the comment from above:
I don’t know that they’re basing their recommendation on risk in obesity on epidemiological data collected which specifically includes assessments of obesity, chronic medical conditions or specific other categories that they’re claiming as risk factors.
I do know that the CDC collects data on a regional basis that includes demographic data: “Varying segments of the population are affected by influenza and may require targeted interventions. These groups are determined through influenza surveillance. Here.
I also know that the EU collects demographic data and information on underlying conditions of patients diagnosed with H1N1 – here, and that they say that preliminary analysis of the initial few hundred cases reported at European level shows that the epidemiological pattern in the EU+3 countries does not differ from what was documented in the Americas – I think it’s reasonable to then say that similar data is collected in the US.
(It’s probably worth noting that the (June 11) EU data doesn’t include obesity in its list of the (24 from 292? unclear) cases of patients with underlying conditions, though. No idea whether that means it isn’t being collected, or just isn’t being published)
In support the first comment I made – that you can’t infer, as I think Lauredhel did initially, that the population severely affected by the disease necessarily reflects the same demographics as the broader population – the EU report notes this: …Therefore, the demographic characteristics of cases documented in the EU so far do not reflect the overall population at risk of infection, but rather the population contributing to seeding events (travellers) and amplification of transmission (school children and teenagers) in the early stage of the spread of a new influenza virus strain.
I don’t have any particular view on whether obesity is an risk factor for H1N1, btw – I was noting that there doesn’t appear to be enough published data to judge either way from the outside.
But I think it’s also possible the data does exist, is being used to inform the precautionary statements (and preparations) of public health authorities and just has not yet been made publicly available,
crankynick:
I suspect we’re probably picking at side-nits here, but I’ve been back over my post a couple of times, and I’m still not sure what you mean by this. Talking about the risk of “severe swine flu”, or chance of being among “severe, hospitalised H1N1 flu cases”, or the odds of sickening “with severe swine flu”, says nothing at all about the pool of milder H1N1 cases or the total pool of H1N1 cases. The two groups I’m comparing, just as I believe the CDC is, are (1) the population, and (2) people hospitalised with H1N1 flu.
At no point have I talked about the total pool of infected people and their various risks of going on to develop more severe complications. That’s not been part of this conversation.
Hope that clears things up?
One of the reasons I haven’t tried to address that is the fact that we don’t have any good data on the total number of people infected, and we won’t have unless someone does fairly large-scale random testing of a representative sample of the population over time. (Which they might, but that will take a while.) That particular denominator can only be delineated by testing everyone, not just sick people who (a) present to the doctor and (b) get swabs taken (and (c) get accurate results that are then reported, etc.). That missing denominator is, as Mystery Rays has perspicaciously observed, going to be an issue anywhere where there is asymmetrical or poor access to healthcare and/or a pool of relatively mildly affected folk.
At no point have I talked about the total pool of infected people and their various risks of going on to develop more severe complications
I guess the point that I was trying to make is that I think that you probably should have been – as should the CDC, if they were going to publicise risk assessments including obesity as a risk factor.
I think you have to know what proportion of the total population of people who were infected with H1N1 influenza were also obese before you can draw any conclusions about the risk factors – both for contraction and for progression to serious illness. If you don’t have that middle number, neither set of analyses mean very much.
I’m not sure we’re in substantive disagreement, though.
Update just posted:
CBC.ca reports one more datapoint, in “Risk factors for severe swine flu may explain disease’s enigmas: experts”
That’s 17%. Obesity prevalence in New York is 25%. The journalist goes on:
*headdesk*
Note that in these two datapoints, there is of course no statistical adjustment for other factors that are potential confounders, because they’re simple case series instead of controlled studies.
Obesity is more prevalent among those who are poor and those who are already ill. I would expect to see significantly higher levels of complicated illness or death in obese people purely because of these confounders, even if obesity was not an independent risk factor at all.
Yet, on the only two datapoints disclosed so far, complicated illness and death is lower among obese folks (and by oddly similar amounts).
Still waiting on that data they’re withholding and using to inform the Fatty McSwineFlu (thanks lilacsigil) panic. *whistles*