This sounds familiar

I just finished reading two books on obesity: The Obesity Myth by Paul Campos and Rethinking Thin by Gina Kolata, both via Megan McArdle. I strongly recommend both books. Campos’ is more advocacy and moral indignation - he’s a professor of law - which makes it a livelier read. (For my money, you can skip Chapters 15, 17, and 18.) Kolata’s is more strictly factual - she’s a science writer for the New York Times - which makes it more compelling. If you only read one, read Kolata.

There’s lot of interesting stuff in both books and I may revisit them in a future post. For now, though, I want to talk about something in Kolata’s book that sounded awfully familiar.

One thread of Kolata’s book traces the history of worry about overweight and the diets that were prescribed to overcome it from around 1800 to the present. In Chapter 8, The Fat Wars, she has brought us up to the present. The part of the story I’m interested in begins in 2001 with the surgeon general’s report entitled “A Call to Action to Prevent and Decrease Overweight and Obesity”. Kolata quotes one political science who calls this “crisis and obsession” and continues:

We’d gotten all the cultural mileage out of blaming tobacco for our sins. The surgeon general’s report gave us a new sin, and it resonated, had a sense of truth.

According to Sander Gilman of Emory University:

The operative term is moral panic. There are moments when certain things become the focus of the society because they are believed to be ... a danger to the society, and it is believed that if you focus on it you will be able to avoid it or cure it.

Kolata goes on to describe studies done to determine “How many American’s die each year simply because they are so fat?” She cites two studies published in the Journal of the American Medical Association, one in 1999 and one in 2004, both by researchers from the Centers for Disease Control. The earlier paper claimed that between 280,000 and 350,000 “Americans die each year because they weigh too much”. The later paper claimed “obesity and overweight were causing an extra 400,000 deaths a year”.

After the first paper was published, however, Katherine Flegal, a statistician at the National Center for Health Statistics, and David Williamson, an epidemiologist at the CDC, took at look at the work and discovered problems with the data. The paper was not a new study but rather used data from existing studies. Some of those studies relied partially on self-reporting of weight and at least some were based on a population that was not a representative sample of the United States population. Furthermore some (or perhaps all) of the original studies:

also tended to remove people from their populations, excluding large numbers ... because the researchers thought they were sick. Such practices have become increasingly common in studies of obesity, although they are rarely seen elsewhere in epidemiology because there is no agreed-upon way to decide whom to remove, or when, and such a practice inevitably alters the results.

As if the data problems weren’t enough, Flegal and Williamson also discovered an error in how the data was manipulated. When trying to determine causality, researchers must attempt to control for other factors that can affect the outcome; in this case, other factors are ones that might affect mortality rates, such as age, sex, and smoking. The researcher on the 1999 paper assumed the data he was using had not already controlled for those factors so he controlled for them. However, the studies had already controlled for those factors so he overcontrolled, so to speak. Flegal and Williamson estimated the data manipulation problem would overstate deaths from obesity by 15 to 20 percent. The two wrote up a paper detailing the problems and attempted to get it published. It took four years and four rejections without review before the American Journal of Public Health accepted it. When the second paper attributing 400,000 deaths a year to excess weight appeared, Flegal and Williamson determined it used “the same problematic methods”.

After documenting the errors in the 1999 paper, Flegal and Williamson began to look for a way to do the “how many die because they’re fat” study correctly. They enlisted the aid of two statisticians from the National Cancer Institute to help. Although new to obesity research:

Barry Graubard and Mitchell Gail ... had developed highly regarded and sophisticated methods for estimating cancer risk ... The statistical challenge was the same as with cancer, and their experience with national cancer center data could carry over...

The foursome used data from the Federal Health and Nutritional Examination Survey which they considered both trustworthy - the subjects’ weight and height were measured as part of the data collection process - and representative of the United States population. Their results were very, very different from those in the 1999 and 2004 papers. What Flegal and her colleagues found was a U-shaped mortality curve. Those at the tails - the very thin and the very obese - had an increased risk of death; the normal weight and the overweight did not. In fact, the overweight people had a slightly decreased risk of death compared to those classified as normal weight.

Less than a year after the 2004 paper estimating 400,000 deaths from fat every year, Flegal, Williamson, Graubard, and Gail published their paper, also in the Journal of the American Medical Association. The response was swift. Those with no involvement in obesity research looked at the methodology and praised the paper. On the other hand, “[r]esearchers and organizations that had made a career out of sounding alarms about obesity and overweight” attacked in full force. The Harvard School of Public Health held a seminar to refute Flegal’s paper. They claimed it contradicted The Nurses’ Health Study which was true but that study suffered from the problems of being non-representational and of throwing out up to 90% of the participants for one reason or another. The seminar claimed Flegal’s group had failed to exclude people who should have been excluded. Not true, say the foursome, they analyzed their data with and without smokers and with and without the chronically ill: the results were always the same. In a further attempt to bury the paper, the critics claimed the foursome did not have the credentials to do the work they’d done because they weren’t physicians. Flegal and Williamson have Ph.D.'s in nutrition; Gail actually has an MD - from Harvard, no less - plus a Ph.D. in statistics; Graubard has a Ph.D. in statistics:

“I’ve published papers with most of our critics. and our credentials didn’t seem to be an issue then,” Williamson says.

Then Harvard issued a press release announcing that public opinion had not been swayed by Flegal’s paper:

Most Americans have not changed their minds about the seriousness of the obesity problem and do not believe that scientific experts are overestimating the health risks of obesity.

It went on to, in essence, congratulate the 41% of the American people who continued to believe being seriously overweight kills as many people as smoking and the almost 50% who continued to believe that even someone who was moderately overweight was likely to die too soon. In other words, for failing to give any credence to the ridiculous new study by Flegal, et. al.

Furthermore, researchers in the obesity field wrote editorials in and letters to the editors of their journals urging everyone to join in pushing back against Flegal’s paper. The editor-in-chief of Obesity Management directed his words to doctors. He told them that although few of their patients were likely to have read the Flegal article they had probably heard about it on the news and had gotten the message not to worry too much about being overweight. He continued:

I do not think this is the message you want them to have. It is difficult enough to get many people interested in addressing their excessive weight. Second, when we need more resources not just to study the problem but to begin to do something about it for the public, this provides ammunition for those who oppose this allocation of resources.

In summing up her experience, Flegal says:

Everyone thinks they already know the answer. Anything that doesn’t fit, they have to explain it away or ignore it. All these people who just know weight loss is good for you. It’s just taken for granted regardless of the evidence. [snip]

I expected people would get perturbed, but I really didn’t expect the way they did it. All those erroneous so-called fact sheets. And these misinterpretations and making up things we’d said. [snip]

It seems like some researchers say to themselves, “If the data don’t come out as strongly as we want, then let’s just work on these data until they come out the way we want them to.”

If all this sounds as painfully familiar to you as it does to me, then you are probably aware of the history of another moral panic: Apocalyptic Anthropogenic Global Warming. The Flegal paper sounds like obesity’s hockey stick controversy. The only researchers considered to have acceptable credentials are those who agree that obesity (or AAGW) is a crisis; that is, those who reinforce the moral panic. The refusal on the part of the obesity establishment to even consider Flegal’s findings echoes Al Gore telling Congress that the science of AAGW is settled. And the warning from Obesity Management about the dangerous impact of Flegal’s paper - and especially the concern that it “could provide ammunition” to those who don’t want to fund obesity research and obesity solutions - is within hailing distance of Paul Krugman insisting that those who don’t get with the AAGW program are committing “treason against the planet”.

In many ways, this is all normal at least in the obesity field. (AAGW, not so much since the “establishment” wisdom is so very new.) Thomas Kuhn’s The Structure of Scientific Revolutions explains the slow process by which scientific consensus is destroyed and reformed: evidence is presented that contradicts the existing paradigm; the evidence is ignored, explained away, or dismissed; more evidence is presented; and the cycle repeats until the weight of contradictory evidence brings down the existing view of the world and erects a new one in its place. Part of what is unusual about the obesity issue is that the struggle is more visible than usual, the data is more accessible, and it is much easier to compare what the science says to what we experience as individuals than is the case with more obscure fields . How many of us can explain the shift from (speaking very simplistically) gravity to relativity as the paradigm for understanding the rotation of the planets? And, equally important, how many of us must consider that shift in our day to day lives?

And that, of course, is a way in which the obesity paradigm does resemble the AAGW paradigm: the obesity paradigm is already having dramatic effects on our lives both personally and politically and is likely to have even more dramatic effects in the future. The interlocking elements of the obesity paradigm - being overweight causes early death; being overweight causes more illness and thus incurs more medical costs; intervening in children’s lives can reduce their weight; losing a few pounds can ward off Type 2 diabetes; and most of all, anyone who really tries really hard can lose a large amount of weight and keep it off - are informing our discussions of health care and health care costs; are deforming our funding of school programs; and are partially driving our massive spending on weight loss. (Think about how many of the uninsured we could insure for that money.). Yet reading Kolata and Campos reveals that all of these elements are of questionable accuracy and one - that any one can lose weight and keep it off - has been shown to be false for over a hundred years and has been repeatedly proven to be false in study after study after study in the 20th century.

In addition, the existing obesity paradigm almost certainly damages the health of the overweight. Most directly, there are concerns about the health effects of weight cycling, that is repeatedly losing and regaining weight. Beyond that, because it insists that being overweight makes it impossible to be in good health, the paradigm prescribes exercise solely because it will promote weight loss. When the pounds don’t come off, the overweight quit exercising, thus losing the benefits of exercise which are as helpful to the overweight as to those with more slender forms. Similarly, the existing obesity paradigm emphasizes eating for thinness rather than health which leads not only to calorie counts below those considered minimally necessary for obtaining vital nutrients but also to the adoption of all sorts of unbalanced diets. Even when a diet plan consists of a balanced menu, the emphasis is on the weight-reducing aspect rather than the healthy-eating aspect. Just as with exercise, when the healthy eating plan does not produce sustainable weight loss, it will be abandoned - often in favor of one of the unbalanced plans. Once again the only concern is with weight loss and so the overweight are never told that eating a more healthy diet can improve their health even if no weight loss occurs.

Even worse, the existing obesity paradigm may increase the number of overweight people through an insistence on diets that not only don’t work but could actually exacerbate the problem. This seems obvious to anyone who has ever dieted or been around someone who has dieted. The almost inevitable outcome is that any weight lost is regained and then even more weight is added. This outcome is implied by the studies of what happens to normal weight people who are placed on calorie restricted diets by researchers. These “dieters” obsess about food, dream about food, do anything to obtain food on the sly. When once again allowed to eat as much as they want, some of them consume food in huge quantities, eating 5,000-6,000 calories at a sitting, 8,000-10,000 calories in a day. They are driven to eat just as they are driven to breathe. Similarly when scientists examine “metabolic changes ... psychiatric conditions ... body temperature and pulse” in fat people who have lost weight they find that:

Fat people who lose large amounts of weight may look like someone who was never fat, but they are very different. In fact, by every measurement, they seemed like people who were starving.

Small wonder, then, that the usual reaction to a diet is to eventually eat everything that doesn’t eat you first. And even smaller wonder that so many “successful” dieters end up a short time later even heavier than they were before they started dieting. What is wonderful is that faced with long-term success rates estimated at between 1% and 5%, the obesity establishment continues to prescribe the same treatments - and the overweight continue to follow them, convinced that a 95-99% failure rate somehow indicates that they personally lack willpower. Even more wonderful are the suggestions that somehow “getting” people to lose weight can and should be one of the paths to reduced health care spending.

It is disastrous enough for the overweight individuals involved to go through the weight loss process over and over and over again when we have known for decades that failure is virtually assured. It would be even more catastrophic if we as a nation fooled ourselves into thinking that our health care spending could be slashed if only people would lose weight.

Reality based, my foot.

*****

Reading:

Megan McArdle has been doing very interesting writing on obesity, especially on how intransigent it is and how uncertain it is that the government can do anything about it. This is the most recent post I’ve seen by her on the subject:

A Rambling Response on Obesity

You can start here and work backwards - and possibly forwards if she keeps writing on the subject.


In the health care cost debate, the focus is on higher medical costs for the overweight rather than on higher mortality. This makes sense: if the overweight have the decency to die quickly they will require less health care. I’ve taken a quick look at two studies in this area.

One of them was cited in the Flegal paper and asserts worse health and therefore higher medical costs for the overweight. I cannot tell from the author’s description whether the study sample was self-selected; if so, I would have qualms about that. The other problem is Campos’ point: we do not know if obesity is a problem in and of itself or simply a marker for lack of exercise and unhealthy eating. That is, if a study finds the obese have more health problems, is that because they are obese or is it because they eat poorly and do not exercise?

The other study was cited by TigerHawk and asserts higher medical costs; it provides a dollar figure. I barely skimmed this one but found a problem on page 9:

This analysis has several limitations. One is the reliance on self-reported height and weight.

Thank goodness nobody ever lies about their height.

This study also shares the Campos problem (emphasis mine):

Finally, the regression-based approach allows for quantifying the spending attributable to obesity by payer and point of service, but it does not directly allow for apportioning spending across specific diseases or the underlying behavior that causes excess weight (that is, poor diet and inactivity). This should be an area of future research.

I disagree with the author’s claim that we know these factors cause excess weight but I agree they should be controlled for if we are to make any claims about the health costs of overweight. Imagine if we discovered that walking for a half-hour every day made people healthier no matter what their weight? As hard as it is to get people to exercise, I can pretty much guarantee it would be easier to persuade people to walk 30 minutes a day than to get them to lose 30 pounds and keep it off.

I would also very much like to see spending apportioned across specific diseases. What exactly are we spending extra money on when it comes to the overweight?