Jeff Schwimmer and his University of California-San Diego colleagues seem to think so. In a press-release that is being cut-and-pasted across across the internet, Schwimmer tells of a study he conducted which appeared to show that kids who caught a cold from the adenovirus 36 (AD36) were fatter than kids who avoided that virus.
To understand this study, you must become savvy to the lingo pediatricians use. Foremost is their use of the word children. Say that word to most civilians and they conjure up images of pre-teens; but to a pediatrician it implies anybody under 19, and sometimes even under 22. There is nothing wrong with this, of course, but if you were unaware of the distinction while examining pediatric study results, you could easily fool yourself.
Schwimmer et alia “examined 124 children, ages 8 to 18, for the presence of antibodies specific to” AD36. They also measured these kids’ weight, height, ethnicity (two-thirds were Hispanic; one quarter white; a tenth black), sex (just over half were boys), and so forth. But why look for previous infections of the AD36 virus? According to the press release:
While an association between AD-36 and obesity in both animals and human adults has been previously described, the particulars remain poorly understood. For example, it is not known how often or under what circumstances AD-36 infects, why the virus affects people differently and whether weight gain is the result of an active infection or a lasting change in a person’s metabolism. In cell cultures, Schwimmer said, the virus infects pre-adipocytes or immature fat cells, prompting them to develop more quickly and proliferate in greater numbers than normal.
In other words, some suspect that, through an inexactly specified and largely unknown mechanism, the virus makes people pack it on. Of course, the virus also might have nothing to do with weight gain.
Half the kids (67) in the study were “obese” as specified through the problematic body mass index (BMI; which over-classifies actual obesity). Fifteen-percent (19) of the kids were found to have been previously exposed to AD36, and 78% percent of those exposed were obese. That is, 15 out of the 124 total kids were both exposed to the virus and were said to be too fat.
I cannot be more specific about the numbers because although the press release promised the paper would in the 20 September on-line version of Pediatrics, I could not locate it.
The main finding is that kids “who were AD-36-positive weighed almost 50 pounds more, on average, than children who were AD-36-negative.” Findings of this kind usually imply that the distribution of weight in the two groups is identical, except that the mean of the exposed group is higher. This is a strong assumption and is usually only met approximately, if at all. But never mind: the study had a statistician, and we can assume he made the proper adjustments for deviations from this assumption (such as for ethnicity).
However, the one measure that cries out for adjusting is age. Just think: exposed kids are on average 50 pounds heavier—50 pounds! That’s an enormous difference! Why, it’s the same difference in weight as between the average 18-year-old and average 9-year-old. And, say, this study had 18- and 9-year-old kids, didn’t it?
Now, what if it was the older kids who predominated in the group exposed to AD36? Then it would falsely appear that exposure leads to weight gain. Importantly, older kids have obviously had more time to be exposed to the virus. The small number of kids and their wide range of ages in this study could easily lead to a false signal.
What’s more plausible: kids who are exposed to a virus plump up, or that sedentary fat kids catch colds easier than their lither compatriots? That is, Schwimmer might have it backwards: obesity might cause colds!
There is also some indication that Schwimmer had a desire that the study show that AD36 causes obesity. From the press release:
What Schwimmer does hope the findings will do is get people to “move away from assigning blame, and broaden the way we think about obesity. Currently, there’s a somewhat simplistic belief that obesity is just a person’s own fault, or in the case of children, the fault of the family. But, that’s an overly simplistic view, and it’s not helpful,” he said.
Rephrasing: it’s not the kids’ fault that they’re fat. Sitting on their keisters for endless hours in front of a glowing screen is less plausible as a cause of obesity than is a virus working through mysterious means. Further, it is not helpful to yell that kids should turn off the damn computer/TV/video game, stop shoving so much cake down their pie holes, and get outside.
Thanks to reader and contributer Ari Schwartz for suggesting this topic.
Update Thanks to reader and soon-to-be contributer Katie who has tracked down the paper (it was not where we had supposed).
The data indicated that for the AD36 Negative group, the ages were 8-11 yrs (18%), 12-15 yrs (65%), and 15-18 yrs (17%). For the AD36 Positive group, the ages were 8-11 yrs (5%), 12-15 yrs (32%), and 15-18 yrs (63%). This is as we surmised above: more older kids are in the AD36 group. The weight average for the AD36 Negative group was 69 kg (+/- 24 kg SD); for the AD36 Positive group is was 93 kg (+/- 24 kg SD).
Thus, one sure finding is that older kids are heavier: in fact, they were about 24 kg heavier, which translated to about 50 lbs. It is at least good to see that the press release got this figure correct.
Another finding is that older kids are more likely to have been previously infected by AD36, also as we surmised (in two years, you have have plenty of colds).
There was no word in the paper that the authors adjusted for age. Indeed, there were no limitations of any kind mentioned.
When I went to school 15% of 67 was 10.05.
Did something change? Transcription error somewhere?
Bob,
Typo. Should have read “Fifteen-percent of the kids“, not obsese. Thanks.
You are both wrong. See the Junkfood Science website by Sandy Szwarc. Will the myth of overeating and under exercising as a cause of obesity never end?
Mr. JH must have had a bad cold, hmmm, around age 40.
I blame all my temporary weight gains on chocolate bars, cookies, cakes, pies, etc. And I credit my pants and skirts for helping me lose weight.
Briggs,
You are a hopeless romantic. If AD36 exposure occurred in the older kids I wouldn’t be surprised at the following headline.
Well, maybe not but I do wonder at times.
How many families adhere to that old adage, feed a cold, starve a fever?
Do these genes make my butt look fat?
Sorry, I couldn’t help myself (except to another slice of pie).
Can it really be that poor of a study? And where were the editors and reviewers of Pediatrics?
Another fallacy.
does not tend to make kids “obese”. The proper term is “morbidly obese”. [Thumbs excepted, of course]. Stop being so politically correct!
Bernie,
It’s hard to credit. But if they adjusted for age in the weight finding, then the raw signal must be enormous. I have still be unable to find the original paper.
This statement is found in the news release.
“An estimated 17 percent of American children and adolescents are obese. Obese children have a 70 to 80 percent chance of becoming obese adults.”
67/124 children in the study were found to be obese. 54%
I suspect a selection problem exists.
Although it is a little known fact, retrospective studies like this one have virtually no inferential value. Indeed, no causal interpretation can be drawn from the statistical analysis of observational data. The authors might (or might not) have reported odds ratios as a way of summarizing the data, which would have been appropriate. But cause-and-effect is simply NOT a logical conclusion that can be obtained from restrospective observational studies. Ever.
That’s not just a rule of statistics. It is a Law of Logical Inference, and hence of the Scientific Method.
The authors are not doing science. They are doing something other than science. They purport to be scientists, and took your money with the promise that they would do science, but they did something else.
I hesitate to use the word “fraud” because it will just make everybody mad at me. Therefore I request that you all come up with another word that we can use to describe what the authors did, a word I can use without fear of angry recriminations.
Mike D.,
Depends upon what you mean by “observational data”. Experimental data is observational data is it not? The real problem with this study is the inference of causality when (apparently) only two variables were used and the absence of consideration of possible confounding factors. Knowing that A and B are correlated while lacking any other information makes causality impossible to determine. When and how the necessary data are obtained would not necessarily be relevant to their utility in assigning cause.
Hard to see how any experimental data could be obtained ethically for this study. Does that mean no attempt should be made to answer the question? What if data were found that mimic those of an experiment? Would it be OK to use it? Would inference based upon it be valid?
Mike D,
Rather than ‘fraud’, you can call this “the narrative fallacy.” Or a post-hoc rationalization.
Fraud isn’t the right word because fraud suggests intent. These guys didn’t intend to lie– they just don’t know any better, and the publisher of whatever journal published what sounds like an exciting story.
Also, I’d like to note that although I’m a very avid video gamer, I also exercise regularly and while I was never one for team sports myself, I’ve been a distance runner and hiker for a long time.
Well, I was a distance runner. Then I ruined my left foot. Now I’m a moderate-length biker.
Not all video gamers are illiterate fatties, dear Boomers. Many of us are fit and educated. Even worse, I can attribute some of my cognitive skill to games.
All,
See the important update above. The paper has been tracked down. I’m not sure how rare this is, but my guesses turned out to be correct.
The article is behind a paywall. It is not accessible at the Boston Public Library on line until 3 years after it is published.
Twenty-four is one of my favorite numbers (I am not lying), but it’s no reason for the number to show up more than once or twice. It could happen, but I am inclined to think something is wrong. Also, it seems inconceivable to me that the researchers and referees would overlook the obvious fact that hat older kids are heavier.
Where is the article? Yeah, I have to read it for myself to believe it.
Conceive it, JH, conceive it.
DAV says: Experimental data is observational data is it not? …
Yes, all data are “observed”. Or else they are phony. But the accepted usage in science is to separate “observational” studies from “experimental” studies. The latter are where confounding factors are controlled by the reseachers. That does not mean that all observational studies are worthless; just worthless at cause-and-effect inference. Nor does it mean that all experimental studies are good at cause-and-effect; they have limitations, too.
You point out correctly that experimenting with humans is unethical and probably impossible. In order for the study to be experimental, some of the subjects would have had to be infected with the virus deliberately (along with controls of all the other confounding factors). Such an experiment would be theoretically possible on lab rats, but not on people.
Does that mean the question is uninteresting? No, but it does mean it is extraordinarily difficult to derive any scientific inference about it. Scientists face this kind of issue all the time. Some questions are intractable (in fact, most of them). Crafting tractable questions of interest is a very important aspect of the scientific method. So is making proper inferential statements.
Science is not religion or philosophy. There is no question that cannot be considered under the rubrics of religion and philosophy. Science is much more limited in that respect.
Re “post hoc rationalization”. Maybe. Better than “junk pseudo-science” because it still gives the researchers the semblance of belonging to the cult of “science”.
The people who pass out the research dollars need to catch this stuff before they fund it. There ought to be some scientific method controls at the money spigot. Instead, the spigot is open to members of the cult and their respective institutions — who also have zero scientific methods controls. In fact they celebrate “post hoc rationalizations and pat themselves on the back for making them.
Where do they award the Nobel prize for Medicine? The authors of this article should be put on a no-fly list specifically for that airport!
An aside:
For all intents and purposes, climate science is entirely observational. Hence any cause-and-effect inference in climatology is hugely suspect and technically illogical. But you already knew that…
Jerrry,
Now, now. This study is no different than (tens of? hundreds of?) thousands of others where the authors wanted something to be true more than they wanted to find reasons it couldn’t be. You’ll note also that we have not disproven that AB36 causes fat children.
Instead of fraud, I suggest science porn (as was used I guess in climate porn). The scientific basis of the claim is unimportant, only the pleasure of the mediatic sound bite. It’s a pseudo scientific claim that appeal to our basis instincts.
I kinda favor scienciness to describe the product of these scientists’ efforts.
Mike D. “Fluff”?
Mr. Briggs,
There are so called Thirty-Six Stratagems, mostly attributed to Sun-Tzu. Sorry, conceiving without further investigation is not one of them.
Anyway, I see that you’ve rounded the numbers.
It also reports the results for BMI, which is a more reliable indicator of obesity. Let’ note that the paper concludes that there is an association of obesity with the presence of neutralizing antibody to AD36 in children. It doesn’t claim a cause-and-effect relationship.
JH,
BMI is not that great a measure; but anyway, I see that you agree with my results. True, they did not claim “Causality!” with 100% certainty, but they sure did lay the insinuation on thick.
I rounded the first 24 from 23.6 and the second 24 from 24.2; the third 24 (difference in weights) was from 23.8. Fair? And, since we only have an n = 127, I’m sure you’ll agree that to express the uncertainty in the standard deviation to three decimal places is not warranted. I did the rounding the authors should have done.
Ah, now mister statistician, something they never taught me in school.
How accurate is the sample standard deviation? i.e. If I had a set of data from a known distribution, how close should I expect the sample standard deviation to be to the actual SD of my distribution? I would guess that it is within about 10% (but not more precise than that) if I have a hundred elements to my data.
I remember learning about the Student’s T to measure the probability that your sample mean was in a neighborhood of the population mean. Can I use the Chi Square distribution can be use in a similar way with variances.
Mike D: “For all intents and purposes, climate science is entirely observational.”
And Monte Carlo climate models are also useless. Trying to take obervations of a chaotic system and fit it to a model with many parameterized simulations built into it is asking for trouble.
I remember once how I was testing a matrix solution with random numbers as inputs. The matrix solution was working correctly but every once in a while I would get a diagonal symmetric result with a diagonal of whole numbers where you would have thought you would get random fractions. Turns out random number generators built on computer math can sometimes fuze themselves into the equation and come back out in unexpected ways.
One must be very careful when back testing models with randomly generated data, especially if they are based on linear algebra and you are using a random generator that produces patterns.
Matt:
Is there a link to the actual study? Will Katie provide a follow up post? I ask because it appears as if this is a grossly misleading interpretation of the data. Is it really as bad as it appears?
Truth. I do push-ups during loading screens and sit-ups during cut-scenes. It’s only a sedentary activity if you’re lazy.
Pingback: We are exposed to more viruses as we age. In other words, A VIRUS IS MAKIN MAH BABBY FAT! : erv
http://www.aap.org/digitaleditions/pediatrics/
You can sign up for free preview if you don’t have university library or other access.
What’s more plausible: kids who are exposed to a virus plump up, or that sedentary fat kids catch colds easier than their lither compatriots? That is, Schwimmer might have it backwards: obesity might cause colds!
My only problem with this statement is that whether or not they “catch” a cold (i.e. show symptoms) and whether or not they are exposed to it are two different beasts. Measuring for the presence of antibodies (as done in the paper) would show who was exposed, not who “caught” it.
One could even go so far as to argue that the less sedentary kids would be expected to be in more sports which would increase interaction with other children and increase the odds of being exposed earlier as compared to the kid sitting at home alone in his room playing Halo.
Overall good article. I certainly hope they corrected for age, though. If they didn’t I think it would make my science cry :*(
Poodle and others,
As I mentioned in the update, no correction for age appears in the paper. Readers interested in a criticism of the paper’s virology should click through the link to “We are exposed…” just above.
The crying you hear is my science…
Pingback: Instapundit » Blog Archive » COULD AN ADENOVIRUS INFECTION lead to obesity? And what about a vaccine? UPDATE: Some study cri…
This is hardly the first study on adenovirus-36 and obesity. They’ve already tested the virus on experimental animals in controlled settings, and found that infection results in obesity. See, for instance,
http://www.ncbi.nlm.nih.gov/pubmed/17135605
from four years ago. To quote from the abstract, “Despite similar food intakes, infected rats attained significantly greater body weight and fat pad weight by 30 weeks post-inoculation. Epididymal-inguinal, retroperitoneal, and visceral fat pad weights of the infected group were greater by 60%, 46%, and 86%, respectively (p < 0.00001)."
If you put the search terms "adenovirus 36 obesity" into the PubMed search box on that page, you'll get links to most of the other work that's been done.
Can we squelch this myth about climate science that Mike D started in the thread? The CO2-warming causal relationship comes from basic physical chemistry and thermodynamics and has been perfectly well known since the 1890s.
Norman Yarvin,
Another place to search for papers is in Schwimmer’s bibliography. Yes, and Schwimmer’s will be the study that is cited by the next group writing a paper on AD36 and obesity. After all, it’s been peer reviewed. Thus it is guaranteed to be certain, right?
Apparently, they also found this in chickens, years and years back, which is a bit easier to duplicate. Of course, fat chickens are probably not a bad thing, but fat sick chickens are. (Not good eatin’.)
This really isn’t surprising, when you think about it. Some illnesses make it hard to put on weight afterward, or hard to get color back in your cheeks. So one that makes it hard to get rid of fat normally isn’t all that wild.
However, I agree that we need better written, better conducted studies to prove stuff like this, and it’s weird that the US and the Australian/UK studies each seem to be unmentioned in stories about the other. It’s all Shwimmer and no Nikhil Dhurandhar or Atkinson, or vice versa. (And since Dhurandhar apparently discovered this weirdness and since he also lives in the US where he could be called by reporters, that’s a heck of an omission.)
Shrug. Don’t suppose science and medical reporters do much more than rewrite press releases, most of the time these days. Probably not interested enough to run a search, much less talk to a live body.
I think there is a golden opportunity for publication here – just use their methods (ages 8-18, no age in the model).
I hypothesize that doing better on an algebra test will be associated with greater weight, and can give interviews saying that math may be bad for your health, as we suspected. We can go on to test “ever broken a leg”, “better at archery”, “plays hockey”, “reads more”, and “has eaten raw tuna”.
Seriously, that they did not even mention the age problem in discussion is a sign that no reviewer or editor complained, but that seems impossible.
Norman Yarvin
re: other papers in PubMed:
I came to the same conclusion by using AD-36 + obesity + 2010 in Google Scholar.
There seems to be a lot of background on the subject. I totally agree that at face value this paper adds little or nothing.
FWIW:
Another paper, also behind a paywall, is:
International Journal of Pediatric Obesity
April 2010, Vol. 5, No. 2 : Pages 157-160
Human adenovirus-36 antibody status is associated with obesity in children
Richard L. Atkinson, Insil Lee, Hye-Jung Shin, Jia He
(doi: 10.3109/17477160903111789)
http://informahealthcare.com/doi/abs/10.3109/17477160903111789
If age 14.8 ± 1.9 corresponds to Mean ± Coeff. of Variation, what does 2 figures for the range represent?