Junkfood Science: Oh what a tangled web we weave — Sir Walter Scott (1771-1832)

July 26, 2007

Oh what a tangled web we weave — Sir Walter Scott (1771-1832)

I wasn’t even going to write on this “study” because it has elicited a plethora of thinly veiled hate speech in the blogosphere and media, and was such junk science, I was certain no one would take it seriously.

I was wrong.

Within hours of the press releases, a massive, well-orchestrated marketing campaign was off and running. By dinnertime yesterday, Google noted 300 nearly identical articles had been published about it and there were 500 by this morning. Television and radio reporters have been gushing over it, with MSNBC reporting that having a fat friend can make you fat and be downright dangerous for your health. It isn’t genes, environmental influences or hanging around together, either, they reported, but the obesity epidemic could be spreading via email and instant messaging. [The power of the internet as a new energy modality?] ABC news told viewers this morning that having fat friends was bad for you and that fat friends could be spreading the attitude that being fat was okay and making it acceptable to have unhealthy behaviors. The groupthink instantly made it acceptable to publicly make fat people out to be pariahs and no one noticed the hurt it was causing.

Let’s not beat around the bush. This study was used to justify and promote the social shunning and discrimination of fat people.

Not one health or medical writer, even at the most prestigious consumer or medical publications, has critically reported on this study. Not one has pointed out its unorthodox methods, its findings that conflict with known science and known biological mechanisms, or the flawed and contradictory findings within the study itself.

Gina Kolata at the New York Times wrote one of the earliest reports:

Obesity Can Spread, Study Says

Obesity can spread from person to person, much like a virus, researchers are reporting today. When one person gained weight, their close friends tended to gain weight, too. Their study, published in the New England Journal of Medicine, involved a detailed analysis of a large social network of 12,067 people who had been closely followed for 32 years, from 1971 until 2003... people were most likely to become obese when a friend became obese. That increased a person’s chances of becoming obese by 57 percent.

There was no effect when a neighbor gained or lost weight, however, and family members had less influence than friends. Proximity did not seem to matter: the influence of the friend remained even if the friend was hundreds of miles away. And the greatest influence of all was between mutual close friends. There, if one became obese, the odds of the other becoming obese were nearly tripled. The same effect seemed to occur for weight loss, the investigators say. But since most people were gaining, not losing, over the 32 years of the study, the result was an obesity epidemic.

Dr. Nicholas Christakis, a physician and professor of medical sociology at Harvard Medical School and a principal investigator in the new study, says one explanation is that friends affect each others’ perception of fatness. [The co-author is James H. Fowler, a political science professor at the University of California, San Deigo.] When a close friend becomes obese, obesity may not look so bad. “You change your idea of what is an acceptable body type by looking at the people around you,” Dr. Christakis said.... The investigators say their findings can help explain why Americans have become fatter in recent years — each person who became obese was likely to drag some friends with them.

She went on to report that their analysis was unique and could explain why people have gotten fatter. Genetics can determine ranges of weights around 30 pounds, she claimed, but “that leaves a large role for the environment.”

We are to believe, it seems, the media images that we’ve all gained gargantuan amounts of weight, rather than the average 7 - 10 pounds actually evidenced over recent decades among our increasingly diverse population, as reported by Dr. Jeffrey Friedman, head of the Laboratory of Molecular Genetics at Rockefeller University in New York.

We are also to overlook the genetics of obesity that has been recognized by scientists for more than half a century. “The heritability of obesity is equivalent to that of height and greater than that of almost every other condition that has been studied,” said Dr. Friedman in a 2004 review of the genetic science in Nature Medicine. The simplistic notion that weight can be controlled by diet and exercise and proper behavior “is at odds with substantial scientific evidence illuminating a precise and powerful biologic system that maintains body weight within a relatively narrow range,” he wrote. Already, many of the genes which regulate body weight have been identified, and these “genes balance calorie intake and energy expenditure with considerable precision,” with us having little long-term control over things.

Instead, Ms Kolata went on to interpret the study, opining:

If the new research is correct, it may mean that something in the environment seeded what many call an obesity epidemic, leading a few people to gain weight. Then social networks let the obesity spread rapidly.

It also may mean that the way to avoid becoming fat is to avoid having fat friends.

That is not the message they meant to convey, say the study investigators... Friends are good for your overall health, [Christakis] explains. So why not make friends with a thin person, he suggests, and let the thin person’s behavior influence you and your obese friend?

Ms Kolata lent the study additional credibility in readers’ minds by reporting glowing expert opinions.

Their research has taken obesity specialists and social scientists aback. But many say the finding is pathbreaking, and can shed new light on how and why people have gotten so fat so fast. “It is an extraordinarily subtle and sophisticated way of getting a handle on aspects of the environment that are not normally considered,” said Dr. Rudolph Leibel, an obesity researcher at Columbia University.

Dr. Richard Suzman, who directs the office of behavioral and social research programs at the National Institute on Aging, called it “one of the most exciting studies to come out of medical sociology in decades.” The institute financed the study.

Medical writers at publications such as MedPageToday were equally credulous, giving physicians the action point: “Explain to patients who ask that this observational study found that the likelihood of a person becoming obese is heavily influenced by obesity in their friends, siblings, and spouse.” Medical professional readers were told: “Surprisingly, the researchers found, the greatest effect was not among those sharing the same genes or the same household, but among friends, even those living apart.”

No need to go on, as you’ve heard all of this, too. But what you haven’t heard was that this paper wasn’t actually a study, researching people using recognized, proven sound medical research methodology.

It was computer animation and, in essence, created a virtual reality.

The fact it was published in the New England Journal of Medicine, a “peer-reviewed” medical journal, as scientific research is proof only that fancy degrees, prestigious reputations and affiliations have become meaningless.

How’d they do that?

The cohort these authors used was the 5,124 children of the original Framingham Heart Study participants. And in actuality, only 45% of them proved to be connected. The researchers dredged through “handwritten administrative tracking sheets that had been used since 1971” in the Framingham Heart Study. The information they used from the tracking sheets was the contact information that had been gathered on the participants at each of the seven interviews between 1971 and 2003 to facilitate future follow-up, asking them to name a close friend or person who might know where they are in a few years should they move, get married, etc. [Someone you know who keeps a good address book doesn't mean they're your best friend. This information was never meant to be used to research social relationships.] The tracking sheets also noted their relatives.

Using this contact information, the authors then created three levels of friendships based on if the cohort had listed someone or the person had listed them, and if they had mutually listed each other. They put the 38,611 different people connections — friendships, family and neighbors — into a computer and created a computer model the likes medical research has never seen. It’s one for Dr. John Brignell’s computer modeling record book.:)

As they described:

We graphed the network with the use of the Kamada–Kawai algorithm in Pajek software. We generated videos of the network by means of the Social Network Image Animator (known as SoNIA). We examined whether our data conformed to theoretical network models such as the small-world, scale-free, and hierarchical types.

In other words, they made up a computer animation. SoNIA is a Java-based package for visualizing dynamic or longitudinal data, currently under development by two young developers at Stanford, Dan McFarland and Skye Bender-deMoll. Pajek is a free, non-commercial use software program for drawing large networks of relationships.

These authors chose to use the Kamada-Kawai drawing algorithm, but as Kim Holmberg, MSc, PhD, of the Department of Information Studies at Åbo Akademi University in Oslo, Norway, illustrated at Webometrics, different drawing algorithms will give completely different results using the same data. For instance, compare the network of municipal websites in the region of Varsinais-Suomi as drawn using Normal interlinking, using Fruchterman Reingold 2D, and using BibExcel, with frequencies, to Kamada-Kawai.

Here is the social network map [courtesy of James Fowler, UC San Diego] these authors drew “of 2,200 people, the largest group of connected individuals in the Framingham Heart Study, in the year 2000. Each circle represents one person, and the size of each circle is proportional to that person's body-mass index. Yellow circles indicate people who are considered medically obese and green circles indicate people who are not obese. Lines indicate family and friendship ties:”

This “new science of networking” appears to be the up and coming computer research technique and some actually believe it should be a new medical specialty. An editorial in the New England Journal of Medicine, written by Albert-László Barabási, Ph.D., a computer scientist at the University of Notre Dame, Notre Dame, Indiana, said:

The role of links and connections does not stop here. In the past few years, we learned that network effects increasingly affect all aspects of biologic and medical research, from disease mechanisms to drug discovery. It is only a matter of time until these advances will start to affect medical practice as well, marking the emergence of a new field that may be aptly called network medicine.

But there’s more to this study. For their statistical analysis to come up with links, these authors created models and ran countless simulations, with so many assumptions and complex selections and elimination of variables, that no one could hope to unravel it:

The use of a time-lagged dependent variable (lagged to the previous examination) eliminated serial correlation in the errors (evaluated with a Lagrange multiplier test) and also substantially controlled for the ego’s genetic endowment and any intrinsic, stable predisposition to obesity. The use of a lagged independent variable for an alter’s weight status controlled for homophily....A significant coefficient for this variable would suggest either that an alter’s weight affected an ego’s weight or that an ego and an alter experienced contemporaneous events affecting both their weights. We estimated these models in varied ego–alter pair types....

We calculated 95% confidence intervals by simulating the first difference in the alter’s contemporaneous obesity (changing from 0 to 1), using 1000 randomly drawn sets of estimates from the coefficient covariance matrix and assuming mean values for all other variables. All tests were two-tailed. The sensitivity of the results was assessed with multiple additional analyses.

In other words, any pretense that these statistical machinations in any way resemble reality is a myth.

What did they find? None of the odds ratios their computer model came up with were tenable. But they didn't simply admit the null findings. Instead, they reported that obesity was associated less with genetic, familial ties; less with geographical proximity, as in immediate neighbors or even friends hanging out together socially; less with even being married and living, eating and sleeping together; than in simply being friends with a fat person. [But among the fine print: the weight gain of a fat friend wasn’t “contagious” if the friends were the opposite sex or among two females; the finding was only statistically significant among men.]

They made no efforts to give any physiological explanations for these implausible findings or how long-distance relationships might be more associated with obesity than genetics. Nor, did they have any data on the closeness of the friendships or how often people were in contact with their supposedly fattening friends.

Forgetting that their study was a data dredge looking for correlations, which is unable to ever demonstrate causation, Christalkis said it showed the social effect was “a direct, causal relationship.” In their University of California, San Diego, press release, he said: “What appears to be happening is that a person becoming obese most likely causes a change of norms about what counts as an appropriate body size. People come to think that it is okay to be bigger since those around them are bigger, and this sensibility spreads.”

Embellishing its significance

In the study conclusions, the authors said that their findings and “the observation that geographic distance does not modify the effect,” ruled out environmental factors or behavioral imitation effects, as well as genetics or childhood experiences, to explain the obesity epidemic. They attributed the rise in obesity to a change in the “general perception of the social norms regarding the acceptability of obesity.”

Then, in the very next paragraph, they contradicted that, proposing:

Yet the relevance of social influence also suggests that it may be possible to harness this same force to slow the spread of obesity. Network phenomena might be exploited to spread positive health behaviors, in part because people’s perceptions of their own risk of illness may depend on the people around them... The observation that people are embedded in social networks suggests that both bad and good behaviors might spread over a range of social ties. This highlights the necessity of approaching obesity not only as a clinical problem but also as a public health problem.

The press release went on to advance the “profound policy implications of the study,” far beyond any evidence the paper could support. Community-wide weight loss programs that “harness the force of networks” to compel social changes are already being hailed as the solution to the obesity epidemic.

“The findings lend support to treating people in groups or even whole communities,” reported the Washington Post. “If these close social environments can promote a disease, they can also promote solutions to disease,” said William H. Dietz of the Centers for Disease Control and Prevention. “These same social networks might be used to turn a disease like obesity around.” According to the Post, the results of this study support forming relationships with people who have healthful lifestyles.

This is not the first time Christakis has proposed far-reaching public policy implications for social networks and advocated societal-wide interventions for the collective good. In an editorial in a 2004 issue of the British Medical Journal, he said the effects of social networks require a “rethinking of the relative value of healthcare interventions or of the conduct of clinical trials... Doctors, trialists, patients, or policy makers might see reason to take them into account when choosing treatment or evaluating benefit.” He suggested:

When the cost-benefit assessment is made by policy makers with a collective viewpoint, all the downstream costs and benefits of health care accruing to a group might be relevant, and the argument in favour of accounting for collateral effects might be even more compelling than that perceived by individual doctors or patients. From a societal perspective the assessment of the cost effectiveness of medical interventions might change substantially if the benefits of an intervention are seen as including the collateral positive effects and the costs as including the collateral negative effects... For example, preventing a death from heart attack, which is clearly desirable from the individual's perspective, may mean that we have to forego the motivation that would otherwise have accrued to others to whom the patient is connected to improve their own health habits. Another provocative implication is that policy makers might value socially connected individuals — such as married people — more when it comes to health care since benefits might be multiplicative in such people...

Even if someone isn’t fat and won’t derive direct benefit from weight loss in their neighbor, he said, the greater good it will have on influencing others gives it socially important value. “Exercise or smoking cessation in one person may prompt numerous others to behave similarly. Conversely, there may be epidemics of disorders such as obesity, alcoholism, suicide, or depression that might spread in a peer to peer fashion.”

Whenever we come across a study with questionable science, that receives massive marketing and media attention far beyond its merit, and is being used to support sweeping public policy change, it’s worthwhile to ask why. But, in this case, Dr. Christakis declared “no potential conflicts of interest relevant to this paper.”

He is, however, on the Executive Committee of the Robert Wood Johnson Scholars in Health Policy program at Harvard and was elected to the Institute of Medicine last year. The Scholars website says it is “the most sought-after interdisciplinary post-doctoral fellowship program in the social sciences. Its purpose is to foster the development of a new generation of creative thinkers in health policy research.” Each year, it enables twelve Ph.D.s up to five years “to undertake two-year fellowships without any of the usual obligations of teaching and university administration” and is a program of the Robert Wood Johnson Foundation. The fact that RWJF is the largest foundation in the world funding societal policies against obesity, smoking and alcohol, we are to believe, does not constitute a potential conflict of interest.

This study illustrates the difference between politics and good science. The reporting and responses from media and medical professionals have illustrated the difference between stereotype versus knowledge, understanding and compassion. There is absolutely no credible science to support stigma against any group. You cannot “catch” fat from associating with a fat person anymore than you can catch “black” from a black person.

What the science knows about obesity “should be sufficient to end the opprobrium of the obese,” said Dr. Friedman. “To end the stigma of obesity, the scientific community must communicate more effectively a growing body of compelling evidence indicating that morbid obesity is the result of differences in biology and not a personal choice.”

The public trusts medical and journalism professionals to give them reliable information to help them. Over the past 24 hours, that trust has proven to be undeserved.

© 2007 Sandy Szwarc

Bookmark and Share