More on Flegal and obesity

Ampersand responds to my earlier post on Flegal's JAMA study. Read Amp's essay first and then check out my replies.

Amp thinks I'm not being fair to the Center for Consumer Freedom, the group behind the ad onslaught against the Centers for Disease Control. (Cf. US Ad Blitz Dismisses Obesity as Hype [Reuters]).

The Center for Consumer Freedom is launching an all out attack on the CDC. The lobby group isn't content to congratulate the Centers for publishing new and better research. It wants to punish the CDC and its scientists for their previous findings. The CCF's main strategy is to impugn the motives of the CDC. They imply that the agency knowingly disseminated bad research to scare the public.

The CCF claims that the earlier study was a flawed piece of propaganda, but they don’t offer any evidence to support that allegation. The press release simply reports that some scientists were critical of the study's methodology. The fact that someone, somewhere criticized some scientific methodology isn't a case. We can't decide whether the study was done well unless we hear their arguments and those of the study's defenders.

Maybe some of our epidemiologically savvy readers would like to weigh in on the merits of the earlier CDC study.

The general consensus seems to be that the Flegal study is better than its predecessor. But that’s how science works. It's a learning process. Findings are published, the scientific community reacts, and new research tries to overcome the shortcomings of earlier work.

Amp writes:

“There’s nothing wrong with skepticism, of course. But I can’t help but notice that, when it comes to the harms of obesity, suddenly Majikthise’s skepticism vanishes, and alternative ideas aren’t even mentioned:
…There is no doubt that obesity increases the risk of death and ill-health. […] this study suggests that if obesity rates increase, excess deaths will increase as well.
In fact, Majikthise sounds more certain about the obesity/death association than the JAMA authors themselves do.”

No one doubts that there are some excess deaths attributable to obesity, least of all the authors of this study. If obesity becomes more common, then there will be more excess deaths attributable to it. That’s just arithmetic.

Amp wonders why the authors of the study felt compelled to point out that their research didn't address the impact of overweight on other non-lethal health problems. I think their disclaimer is an important caveat. The CCF spin is that overweight is harmless or beneficial. In fact, we’re getting better at compensating for some of the chronic diseases it causes or exacerbates. The authors argue that overweight and obesity are less dangerous today because of better treatments for high blood pressure, high cholesterol, and diabetes. This is good news, but it's not license for complacency. Preventing chronic disease is as important as reducing excess mortality.

I may have given the misleading impression that I support reducing diets as a pillar of public health policy. That wasn't my intent. In the long run the most effective public health interventions will be measures that benefit people of all body types: PhysEd in public schools, community bike trails, tax credits for worksites with exercise facilities, better prenatal care (to ensure healthy weight gains during pregnancy and prevent gestational diabetes), increased support for breast feeding (because breast fed babies are less likely to become obese), zoning and tax incentives to lure fresh food stores to the inner city, an end to the corn syrup subsidy, and so on.

Update: What does this study mean for parents? Maternal and Child Health lays it on the line.

Revere the epidemiologist at Effect Measure gives Flegal's methodology a first-glance thumbs up and offers a backgrounder on the shady Center for Consumer Freedom.

Comments

[Martin]

I'll try to argue why I think the two studies are not necessarily contradictory (caveated that I only have access to the abstracts).

The Flegal study looked at the relative risk of death due to obesity, but did not look at the actual number of deaths due to “poor diet and physical inactivity,” which was what Mokdad looked at when he came up with his 400,000 deaths estimate. Choosing relative risk also means that Flegal is somewhat limited in what she can look at because she must choose a baseline mortality (overweight), which becomes a relative risk of 1. The choice of a baseline mortality is a distinguishing feature between the Flegal and Mokdad studies because Mokdad did not assume that there was a baseline mortality due to poor diet and physical inactivity.

In defense of Mokdad, I would say the he is shooting at a moving target as the study is designed to look at the proportion of deaths that can be explained by poor diet and physical inactivity. This is a moving target because to determine the proportion for, for example, coronary heart disease is a non-trivial exercise since our understanding of the etiology of this disease (and others) is constantly changing. In addition, there is no indication in the abstract that Mokdad and his team evaluated each of the studies included in his study for quality. This is a flaw and is a legitimate reason to doubt Mokdad's article, however, Flegal and Mokdad are still not clearly at odds with one another.

[Hedley Lamarr]

These studies flow back and forth. Recall a few years ago there were alternating studies every few months on the advantages/dangers of drinking coffee. Next it was red wine. Now it is too much weight. Just wait a while and all will swing back the other way.

[Dan McEnroe]

>...better treatments for high blood pressure, high cholesterol, and diabetes.

Most of which are mind-numbingly expensive, and since these conditions impact seniors in greater numbers their treatments will put greater strain on already overburdened Medicare AND Medicaid. It's all connected - don't these people get it?

I watched diabetes kill my dad one piece at a time and now it's going to work on my mom. If these chuckleheads want to rush headlong into that fate using the kind words of David Brooks as their justification, well, my initial impulse is to say, go on and I hope you enjoy your prosthesis. But then I remember that my family's going to end up footing the bill for their stupidity.

[slapshot57]

not an epidemiologist but I have read more scientific journal articles than most and I was not too impressed with this study. First and most glaringly it bases a person's health on BMI, which as anyone who's a gym nut knows is a horrible measure of fitness. My BMI at 6'2" 190 puts me in the slightly overweight category, but my body fat percentage is really low. I'd have a lot more faith in this study if they look at people's fat percentages as a measure of weight instead, because that is what leads to health problems.

Additionally, the BMI for underweight is reidiculous. I was 155 pounds in high school, looked anorexic and yet still fell into the normal weight. Of course if I was 145 that would be ridiculously unhealthy.

Anyways, bad study, doesn't really tell us anything about fitness vs. mortality, yadda yadda yadda people will be using this to justify not going to the gym. Good job JAMA!

[galnoir]

Who (or what) is CCF? From their Web site:

The Center for Consumer Freedom is a nonprofit coalition of restaurants, food companies, and consumers working together to promote personal responsibility and protect consumer choices … Consumer freedom is the right of adults and parents to choose what they eat, drink, and how they enjoy themselves. Defending enjoyment is what we're all about! … The Center for Consumer Freedom is supported by restaurants, food companies and more than 1,000 concerned individuals.

They also oppose the "Green Fringe"—that is, the movement toward healthy, sustainable food production. (Yay, CAFOs! Boo, organic! Mad cow disease? Pshaw!)

In other words … it's all about the bottom line for McDonald's, the Outback, etc. These people don't care about your health. They just want to (1) encourage everyone to buy their products; and (2) maintain the right to produce those products as cheaply as possible. If they can convince you that obesity is not a health problem, then they figure you'll be mollified into continuing to hork back the fries and gallon-sized soft drinks. Because it's all about freedom! and choice!

I think there's been debate back and forth between Majikthise, Amp, etc., about the relative dangers of obesity—but I don't think there's much debate at all that no matter what you weigh, you're better off consuming more fruits, veggies, and whole grains and less processed crap (and being as active as possible). That's not what CCF wants you to believe. They are no friends of the overweight or obese. They are no friends of anyone who gives a damn about being healthy.

[Primer]

Precision vs. Accuracy

My 0.02 dollar. There is no question Flegal’s study is more statistically precise than the previous studies she cites. It remains to be seen, however, if it is at all accurate.

Let X = weight status (underweight, normal, overweight, obese)
Let Y = mortality rate
Let Z = prescriptions for public health

Both Flegal and previous study researchers attempted to determine the relationship between X and Y. Additionally, they hope to show that this relationship has implications for Z. Neither Flegal nor the previous study’s researchers accurately defined X. Additionally, Flegal admits that concentrating only on X’s relation to Y might miss the broader point in respect to Z.

[Lindsay Beyerstein]

I would make the opposite argument. Flegal's estimates are less but probably more. One of the strengths of the paper is that it includes the error of estimate for all the figures reported. For whatever reason, a lot of the earlier work on this topic didn't report margins of error. (I haven't been reproducing those. But the error bars around the 112,000 and the 86,000 are huge.) But since their methdology is better, it's more likely that their estimates are in the ballpark.

Flegal et al. don't really make a lot of suggestions about public health policy.

[Mary]

Some thin people will NEVER give up their claims to moral superiority, it doesn't matter what the research says. It's just so nice to have people to look down on, isn't it? Read "Faces At The Bottom of the Well" by Derrick Bell. He was talking about race but he has your number, Lindsay.

[Alon Levy]

Mary, did you actually read the post? Research says being obese is a health risk.

[Mary]

Being alive is a health risk. Pregnancy and childbirth are HUGE health risks; as is driving a car, living in an urban area, or next to a factory or electric plant.

So now why know why the blog is called "magic thighs"?

[Virginia]

"Research says being obese is a health risk."

The biggest problem here is that so few of those who comment about this stuff online actually know anything about epidemiology! Here is the situation for those who don't have any epi education: To make a true causal inference (prove A causes B), you must run an experiment. You must randomly assign people to groups of "obese" or "not obese" to ensure that there are no other differences between the groups other than their body size. No differences in smoking, exercise, diet, genetics, zip. The only difference between the groups can be body size. Then you check their health outcomes. Of course, we cannot do this, so we have to use less-than-perfect research designs. We have to take people who are already at a body weight and try to draw conclusions. If research is done properly, it will control for factors such as weight-cycling, stress, nutrition, and physical activity (and not through self-report which is notoriously inaccurate.) Few studies have done this. Those that have found that obese fit people live just as long as thin fit people. THIN UNFIT PEOPLE are at the highest risk of premature death, including death from heart disease. The problem is that there are more Obese Unfit people than Thin Unfit people in most studies, so basic comparison studies reach unsound conclusions. Want proof that weight itself does not affect health status? Look at the 2004 study on liposuction. Liposuction reduced people's BMI and removed them from the "obese" category but did not decrease their heart disease and diabetes risks! "But studies show that moderate weight loss leads to better health," you say? Show me one study that showed that WEIGHT LOSS, not INCREASE PHYSICAL ACTIVITY and BETTER NUTRITION, led to better health. You can't find one? I'm not surprised. The fact of the matter is that an "overweight" person who leads a healthy life will live far longer than a thin person who leads the average American life. Until more studies properly control for lifestyle factors, stress, and weight-cycling, we must be suspicious about their conclusion that extra weight itself is the culprit.

[Virginia]

For those who are still confused, try this scenario:

If eating too much popcorn and getting too little sunshine caused SOME people to lose their hair and also led to an increased risk of cancer, would we conclude that baldness causes cancer? Would we begin assuming ALL bald people must have had too much popcorn or stayed inside too much, and that all these people will end up with cancer? Would we begin assuming that those with hair are healthy? Would we rush to address the "baldness epidemic" and say that "baldness costs the country billions of dollars a year in health care costs" because it is linked to cancer? Would we develop a reality show called "Rapunzel" to see who could grow their hair the fastest, all the while insisting that this is in the interest of public health and cancer prevention?

We do the public a great disservice when when implicate weight instead of the true culprits - nutrition, exercise, weight-cycling, and stress. We neglect the hundreds of thin people in this world who binge on thickburgers and honestly believe they are not at risk because we have lied and said the WEIGHT is the culprit (when it is only a visual correlate for SOME people). We also neglect the hundreds of "overweight" people who eat right and exercise, because they are led to believe their WEIGHT is the culprit, so they will go to extremes to get rid of that weight, probably yo-yo, and find themselves in poorer health than before. Remember, research shows that these people were in the lowest risk category if they already eat right and exercise and simply naturally have the extra pounds. Of course, when those people die early, their doctors will assume it is their weight that did it instead of implicating their history of weight cycling!

[revere]

I'm not sure where Virginia studied epidemiology, but the problem of causal inference is considerably more difficult than she lets on (I am a chronic disease epidemiologist of many years experience). The implication that only experiments provide valid evidence for causal inference is incorrect. Causal necessity is a slippery subject, granted, but simplificaitons like that are not helpful. Randomization does not allow you to say there is no difference between the groups in factors other than the one you are studying. On the contrary, there usually are differences which is why you use statistical methods to evaluate the probability that the differences in outcome could have been caused by chance alone. The randomization allows you to use those statistical methods for that purpose, that is all. The notion that non-experimental (i.e., observational) studies like epidemiology are somehow less "scientific" would also imply that geology, astronomy, cosmology and paleontology were lesser sciences because they are observational, too.

There are many difficulties in doing observational studies. BMI is not the ideal indicator for "obesity" but it is measureable and relatively reliable. The Flegal study did adjust for potential confounding (a failing of the earlier Moktad study) and used population based and more recent data assessed longitudinally (the NHANES I, II, III cycles). BMI does not just consider weight but weight and height combined. That is still imperfect. But the data show that BMIs above 30 are associated with considerable excess mortality, which seems to bottom out around 25 which is the upper limit of the current recommended BMI from CDC. This suggests that there is an optimal window around 25 that might be recommended regarding mortality as an outcome. However the Flegal study did not consider other outcomes and optimal body weight/height/fat/distribution probably affects (and may be different) for other outcomes. Over at Effect Measure I will have a short post tomorrow (Sunday May 1) about obesity and influenza.

What is quite clear is that "obesity" is a very political subject. CDC fastened on it partly for political reasons (it was safe from their Bush masters) and they over reached, providing the perfect opening for right wing scum like the astroturf Center for Consumer Freedom (see my post here).

[Alon Levy]

Just to be sure, revere, you're measuring BMI in kilograms and meters, right?

[Mary]

Per Body & Soul, researchers at GlaxoSmithKline have been working on a promising vaccine that could protect women against a virus that causes the majority of cases of cervical cancer. The virus is sexually transmitted, so girls have to get the vaccine before they become sexually active. Yet there are people who'd rather have women get cancer than get vaacinated:

"[R]eligious groups are gearing up to oppose vaccination, despite a survey showing 80 per cent of parents favour vaccinating their daughters. "Abstinence is the best way to prevent HPV," says Bridget Maher of the Family Research Council, a leading Christian lobby group that has made much of the fact that, because it can spread by skin contact, condoms are not as effective against HPV as they are against other viruses such as HIV.

"Giving the HPV vaccine to young women could be potentially harmful, because they may see it as a licence to engage in premarital sex," Maher claims, though it is arguable how many young women have even heard of the virus."

Sound familar? Lindsay and many of the folks on this thread seem to think that studies showing extra weight may not be a big health risk are making the same essential argument that the wingnuts are: People are too stupid to be trusted with decisionmaking. Vaccinate them and they'll have promiscuous sex. Concede that a few "extra" pounds aren't a big deal and they'll gain 100. Too bad we aren't all as smart and thin as Lindsay.

[Alon Levy]

I fail to see any similarity between Lindsay's argument about obesity and the religious right's about vaccination.

[Virginia]

Revere,
I completely admit that I oversimplified the issue of causal inference, recognizing that many do not have even the most basic understanding of the issues involved in making such assumptions. As my original background was in social science, I certainly do not discount quasi-experimental methods for determining causation when experimental methods are not possible, and I know the value of controlling for potential confounders even when random assignment is used.

The problem with these studies is that they found body size (however you measure it) is related to mortality. Ice cream sales happen to be related to crime rates. When one goes up, the other goes up. This does not prove that ice cream causes crime or that crime causes people to eat ice cream. There is a third variable for which one must control -- Season/Weather. When the weather is warm, both go up. The same problem has happened with obesity research. People have found a correlation. Body size (however measured) goes up, mortality goes up. They have decided that one caused the other. They have failed to consider the confound of cardiorespiratory fitness. Those studies that have controlled for cardiorespiratory fitness have consistantly determined that body mass is not a cause of mortality. Fit obese people (by whatever standard of measurement) live just as long as fit thin people. Because there are more unfit obese people in the world than unfit thin people, we see a correlation between body size and mortality, but this correlation does not prove that body size CAUSED mortality. Lack of fitness causes early mortality whether one is thin or overweight.

Of course measurement and classifications are important in research, but the issue I am talking about is the basic research fallacy of deriving causal conclusions from correlational data without looking at potential confounds. The lack of understanding of these issues has led the media to implicate body size. This means that FIT obese people are led to believe they are at high risk for early mortality, even though the research clearly states they are not at any risk whatsoever. It also means that UNFIT lean people are led to believe they are not at high risk.

[revere]

Virginia: I'm afraid I cannot agree with your clarification. Assigning causal status to an association is an act of judgment, not an empirical outcome. The question is not one of correlation or of confounding, but of interpretation.

Your ice-cream example is faulty because it involves a cross-level inference (group exposure and group outcome attributed to individual exposure and outcome). This kind of ecological infernece is known to be open to serious bias from confounding and/or effect modifiers. The Flegal study was an individual level study, hence of a different category.

The connection between BMI and mortality is not just one of a single observed association but part of a rather large amount of cross referenced data from the molecular up to other observational (and occasionally experimental/intervention) studies. The data strongly validate BMI as a surrogate for some factor affecting mortality, perhaps the increased burden of carrying around extra weight or a marker for something else, perhaps a physiological factor related to cardiovascular mortality. This does not mean there aren't other causal factors involved that may increase or mitigate that risk for an individual. But that doesn't invalidate the legitimate (and strong and robust) findings that BMIs over 30 (not grossly obese by any means) carry considerable increased risk of dying. Where BMI as a marker for those factors that affect mortality sits may be open to debate, but I doubt that a reasonable interpretation of the body of evidence would say it is an adventitious or spurious association as you imply.

[Virginia]

Revere,
Can you explain the outcomes of studies that show that obese people (BMI over 30 and high body fat percentage measured by hydrostatic testing) who have high levels of cardiorespiratory fitness (measured, if I remember correctly, by treadmill tests) have the same relative risk of death for cardiovascular disease as lean and average sized people with the same cardiorespiratory fitness levels? How can body composition be a risk factor if this is what the research shows? Also, a basic epidemiological principle for establishing causation is that cessation of the proposed risk factor leads to cessation of the outcome. The 2004 study on the effects of massive liposuction took formerly obese people and put them in the non-obese category. The study did not find a decreased risk of heart disease or diabetes. Cessation of the exposure did not lead to cessation of the outcome. Many studies claim that "moderate weight loss leads to decreased risk." Unlike the lipo study, how many of those studies produced this moderate weight loss without changes in fitness and nutrition? Did they really show weight loss led to decreased risk, or did they show increased fitness and nutrition led to decreased risk? As I am completing a research paper on these issues, I would appreciate having somebody point me to studies that properly controlled for cardiorespiratory fitness and nutrition and still found increase risk of mortality based on body mass, and I would like to know what those authors concluded should be recommended to a person who has large body mass but already exercises and eats right. Obviously, liposuction won't decrease their risk, and according to the studies I currently have in my literature review, such people are already in the lowest risk category. If you have recommendations for other studies to include, please send them my way. I'd rather see them now before I finish the paper.

[Virginia]

Ooops... I didn't mean to say "a basic epidemiological principle for establishing causation is that cessation of the proposed risk factor leads to cessation of the outcome." Clearly, sometimes a risk factor can have lasting effects on the outcome even after it has ceased. I meant to say that this is one of the POSSIBLE ways to establish a causal relationship.

[Andrew Spark]

Obesity has been shown to have a substantial negative effect on longevity, reducing the length of life of people who are severely obese by an estimated 5 to 20 years.