by Marion Nestle
Jan 9 2013

Let’s Ask Marion: Can It Really Be Healthier To Be Overweight?

Every now and then, Kerry Trueman challenges me with a difficult question about some current topic.  Our most recent exchange, published yesterday in the Huffington Post, is about the so-called “obesity paradox” (defined below).

Trueman: There’s a brouhaha in the blogosphere over Paul Campos’ NY Times op-ed in which he claims that our current definition of what constitutes a ‘healthy weight’ is dead wrong. Campos cites a new analysis from The Journal of the American Medical Association claiming that overweight or obese individuals have a lower mortality rate than people whose weight is ‘normal.’

He speculates that our obsession with obesity is a misguided and manufactured controversy foisted on us by the multibillion dollar weight loss industry and Big Pharma. Marion, you’ve been called a lot of things over the years, but have you ever been accused of being in cahoots with Jenny Craig and Eli Lilly?

Nestle. I love it when people invoke conspiracy theories to deny that obesity raises disease risks.

Paul Campos is a lawyer. He views matters of diet and health from a legal perspective. From that viewpoint, if a statistical analysis shows little effect of obesity on mortality except among very obese people, then nobody need be concerned about weight gain except at the extreme.

If only the science of diet and health were that simple. Scientists, alas, must struggle with a number of vexing questions about such studies:

  • Does a finding of statistical significance necessarily imply clinical or biological significance?
  • Do statistical findings based on populations necessarily count for individuals?
  • Do statistical associations provide guidelines for behavior?
  • Are the methods used in statistical studies adequate to draw conclusions about behavior?

We are talking here about a huge meta-analysis of 97 studies of obesity and mortality carried out by Katherine Flegal and her colleagues at the National Center for Health Statistics.

When I read papers by excellent statisticians published in prestigious, peer-reviewed journals, I start by taking the results at face value. Then I ask critical questions about what the results might mean.

I found the figures in the paper difficult to follow so I’ve summarized the results below:


Normal 18.5 – 25 1.00
Overweight 25 – 30 0.94 (95% CI, 0.91-0.96)
Obesity, Grade 1 30 – 35 0.95 (95% CI, 0.88-1.01)
Obesity, Grades 2 and 3 >35 1.29 (95% CI, 1.18-1.41)

My interpretation: Compared to people with BMI’s in the normal range, those with BMI’s considered overweight or somewhat obese display no increased risk of mortality. Indeed, those in the obesity grade 1 category may have a slightly reduced risk. The study only finds an increased risk of mortality—by 29%—when the BMI exceeds 35.

My first reaction? This is not news.

Since this is a study of previous studies, we’ve seen results like this before. Flegal herself published a similar analysis in JAMA in 2005. In that paper, she presented the results in a way that is easier to visualize:




These earlier results show what is called a “J-shaped” curve, meaning that the risk of mortality increases at BMI’s below as well as above the normal range.

The new study is consistent with these earlier—and equally controversial—results.

But the earlier results point to some of the difficulties in interpretation.

  • What’s going on at the lower end of these curves?
  • Are mortality results skewed by people who are ill and weigh less?
  • Is BMI an adequate indicator of health status?

At the moment, there is no way to answer these questions at a level of precision that might satisfy legal thinkers.

What we do know is that obesity above the normal range sometimes—never always—raises the risk for chronic diseases like type 2 diabetes, heart disease, and others. Risk is about probability. Risk never implies legal certainty.

It seems clear that for some people—perhaps many—having a chronic disease does not cause a decrease in lifespan. Since 1970, people throughout the world have gained slightly more than ten years of life expectancy overall, but are now spending many more years living with injury, illnesses, and disabilities—conditions caused by cigarette smoking, excessive alcohol consumption, and poor diet.

With respect to overweight, this is sometimes called the “obesity paradox,” understood as the survival benefits of modest overweight and obesity for people who already have a specific medical condition.  For them, many studies—not all—show that survival is modestly better at higher weights.

The Flegal study deals only with the statistical significance of the mortality survival component of the paradox. It does not deal with issues related to the effects of obesity on quality of life.

As a result, some of my colleagues have made harsh comments about the study, calling it “a pile of rubbish.”

I wouldn’t go that far. The study is what it is—a statistical meta-analysis. It’s up to scientists and clinicians to figure out whether statistics like these have meaning in the real world.

In the real world, it doesn’t take much overweight to induce type-2 diabetes in susceptible individuals, and many people—not all—with type-2 diabetes can completely eliminate symptoms by losing a few pounds.

Campos may be willing to let his weight go to where it will and take his chances that statistics are on his side. That’s his choice.

As for me, I’d prefer to avoid weight-related illnesses for as long as I can. I’m hedging my bets and continuing to watch my weight.

  • Michael

    Thanks Marion. Walter Willett and his colleague JoAnn Manson at the Harvard School of Public Health have been hammering away at two methodological errors in nearly all studies about the BMI-life expectancy connection (including this meta-analysis) ever since a study they did in 1995 in the New England Journal of Medicine that first proved it (and they’ve replicated it since). To wit: MOST studies that explore this relationship fail to take account of two very major things that themselves cause you to both have a lower body weight and to die early, instead of the low body weight itself causing you to die early. (This is similar to how a hot day causes both an increase in ice cream sales and an increase in drownings in swimming pools, resulting in an artificial correlation between ice cream sales and drownings).

    One is smoking. Nicotine suppresses the appetite, with the result that smokers eat less, with the result that their body weights tend to be lower than those of nonsmokers. Some people actively use cigs to help themselves lose weight; others crave sweets during quitting, in part because the have suddenly lost the appetite-suppressing effects of smoking; and most just go through their lives not realizing that they’d be hungrier and likely fatter if they weren’t spending half of their waking hours with a carcinogenic appetite suppression device jammed between their lips.

    Smokers, of course, also die earlier. So if you don’t take special care to adjust for the effect of smoking on BMI, you wind up making it look like slimmer people die sooner BECAUSE they’re slimmer.

    Manson and Willett have found that even the normal way to statistically adjust for the effects of smoking are insufficient to fully take account of the effects of smoking, so that the only way to FULLY separate the effects from smoking on BMI and mortality is to analyze smokers completely separately from nonsmokers). Most studies fail to do this properly; Flegal et al didn’t make ANY adjustment for smoking in this study.

    The second thing is that people who are have chronic disease start *losing* weight many years before it kills them. This is often quite obvious in people who have had cancer, but that’s only the fastest and most obvious kind of case. People start to slowly lose weight many years before they die, because their body is falling apart: the Baltimore Longitudinal Study of Aging found that it begins as much as 9 years before death. So if you don’t (a) put to onie side deaths that happen within a few years of recruiting a person into the study, or (b) take multiple measures of a person’s body weight and watch for a sudden weight loss, you’ll again get an artifactual result that heavier people are less likely to die, for a reason that is the reverse of what is really going on.

    Here is the original study where Manson and Willett showed this:

    Here is a subsequent study in men:
    (Look at the Table, and compare what it looks like in the “Biased subgroup” using the “Confounded Model” vs “Men with >2 years of follow-up” in the “Optimal Model”. The former shows a man of BMI <20.0 ( very slim!) to be 1.68 times as likely to die as a normal-weight man; the other shows him to be ~12% less likely to die than such a man. And the former shows negligible increased risk for being quite overweight (BMI 27.5–29.9), while the latter shows him to be about a third more likely to die than a normal-weight man).

    And here is a recent, very large “meta-analysis” pooling data from many studies, including 1.46 million white adults (similar to what Flegal did, but separating out smokers and with an analysis of people with no evidence of pre-existing heart disease):

    Here is the key figure comparing the 2 analyses:

    In their Willett’s and Manson’s own studies (as opposed to this pooled analysis of many researchers’ studies), they were able to control for more variables, but this shows the general gist: being overweight looks far riskier, and being very slim looks far more dangerous, in the “all subjects” group than the nonsmokers who didn’t already have evidence of cancer or heart disease at baseline.

    Others have proven this too. But, the media get excited every time one of these inadequately-adjusted studies says you’re better off overweight. Sigh .

    There are other issues too, related to the simple fact that BMI is a very crude measure of a person’s physique: at the same BMI, you can be almost all muscle or almost all fat. MOST people with low BMIs are slim, and MOST people with high BMIs are fat, but not all — and, as they age, MOST people lose muscle and gain fat, yet remain at roughly the same weight. They’re actually very fat, in terms of how much deadly visceral adipose tissue they’re carrying around, and increasingly frail as they lose muscle and bone, but their BMI barely budges, and if they weren’t fat in youth they won’t register as fat in old age either. BMI alone will classify many people as ‘normal weight’ who are actually “Thin on the Outside, Fat Inside” (TOFI), which similarly makes being “thin” look risky and in the process downplays the risk (by comparison to “normal-weight” (but fat!) people) of being obviously, visibly overweight.

  • Alexandra

    Isn’t it true that people who are slightly over-weight will last longer in a hospital setting? Perhaps that partially accounts for the slight difference in mortality?

    Is there a difference, ethnically, in terms of BMI and how accurately it reflects health? I am a 6 foot tall woman of Dutch and Swedish heritage. I haven’t had a normal BMI since I was pre-pubescent, but no one I know would describe me as overweight and I am very healthy.

    Very interesting in any case, thank-you! I am fascinated by stats and how they do or do not actually help us in the “real” world.

  • Michael

    (One detail in my comment was phrased poorly. Flegal’s meta-analysis excluded studies that didn’t *adjust*, in the conventional way, for smoking. But it didn’t take the further step of looking at data for nonsmokers as an entirely separate category, which is key, as Manson and Willett have shown and emphasized).

  • Beenie

    Can we please get off the weight-loss bandwagon and start talking about practicing healthy habits instead? For me, daily exercise and mindful, healthy eating has NEVER resulted in any significant weight loss, but it has improved my health a great deal. When a person focuses too much on weight-loss, believing that thinness is the ultimate path to wellness and success, it becomes an obsession, and destructive behaviors that lead to ill health (i.e. the thousands of people in this country with eating disorders and the health problems associated with yo-yo dieting) are often the result.

  • Without a controlled randomized trial it is too difficult to come to a conclusion. That in itself should tell us that the treat is small in overweight and slightly obese people.

    It is also possible that extreme obesity is a result of an underlying problem that would cause earlier mortality in an environment were scarcity of food prevents obesity.

    -Nortin Hadler, MD, believes what is now called diabetes 2 is largely a construct invented by the pharmaceutical industry. Read this 2009 interview. He is the author of Worried Sick. A prescription for health in an overtreated America, which has a section on diabetes 2.

  • Campos’ polemical tone is unhelpful. It’s better, I think, not to respond in kind, with irrelevant potshots at conspiracy theorists, lawyers, and “the media.”

    Underneath his rhetoric, I see Campos making two linked claims:
    (a) The correlation between BMI and mortality is not as simply and clear as is commonly thought.
    (b) The causal role of overweight and obesity in mortality is not as well-understood as is commonly thought.

    Maybe he’s saying something stronger — that there’s no correlation, or that excess adiposity doesn’t cause excess mortality, for example; the polemical tone is unhelpful precisely because it’s difficult to tell how strong of a claim he’s actually making. But we can start by assessing the evidence and arguments for these two weaker claims.

    Now, I’m by no means an expert, but I am scientifically literate, and it seems to me there’s something to these evidence and arguments. Indeed, I don’t think there’s all that much disagreement between you and Campos. You both agree that a purely observational study isn’t powerful enough to sort out the causal relationships. You both agree (I think) that the correlation follows a “J-shaped” curve that doesn’t show much difference between normal weight and overweight. So you seem to agree on (a).

    Furthermore, you both agree that there are numerous other influential factors: weight loss during terminal illness, innate susceptibility to diseases like type-2 diabetes, socioeconomic status, dieting and weight cycling, social discrimination, cigarette smoking (as Michael points out), and so on. So then you seem to agree on (b) as well.

  • My problem with BMI is this: we are using a measurable physical characteristic to measure behavior, since behavior is notoriously hard to measure.

    By this I mean that the BMI, when used over a large population, is a good indicator of whether people are eating right and exercising. People who do those two things tend to be thinner in relation to their height – but not all people.

    There are some studies that show that you can have a high BMI and be just as healthy as your skinny counterparts, provided you are eating right (even if you take in extra calories) and, in particular, exercising. See this one in Puerto Rico:

    This is why I dislike use of BMI as an indicator of anything for an individual. It’s a statistical tool, but shouldn’t be used as a measure of individual health.

  • Stephanie

    Could someone link papers for me that specifically reference FAT as being the CAUSE of a disease? Not habits, not diet, not age, not amounts of exercise, but FAT, and FAT alone?

    I’m not convinced that smokers should be excluded or analyzed separately. If the question is about the effect of BMI on longevity – you can’t just exclude an unhealthy habit that may fall predominantly in one group. Doing so simply skews the data in to support the Flat Earth mantra of ‘fat bad, thin good’ that is so well established.

    To also suggest that all of the dead thin people MUST have been fat before and started losing weight because of their fat diseases is wishful thinking. Of course that will be true of some, but really, is it so impossible to believe that life, death, and illness could be more complicated than just fat or thin??

    Thin people get heart disease, thin people get cancer, thin people even get Type 2 Diabetes. You cannot just wish for a magic bullet of thinness to keep you healthy.

  • Pingback: For The Love Of Food | Healthy Eating Tips - Upgrade Your Healthstyle | Summer Tomato()

  • Anthro


    I don’t think overeating “mindfully” is any better than doing so “mindlessly”. If the result is too much excess flab, most people’s health will suffer–regardless of spiritual status.
    @other comments in general

    It’s interesting how people want to find evidence to support their wishful thinking, i.e., being overweight is okay, when (especially in the case of obesity) it is a threat to health for populations and usually for individuals as well.

    Thank you, Marion, for going through this carefully and with a science-based eye for those of us who are interested but don’t have the statistical, epidemiological, and other science-cred to analyze these things. I am science-literate, but I know when to rely on experts and I choose them carefully. I do not include “brave maverick doctors” who write books for the popular literature genre to be among them. I much prefer science-credentialed professors of nutrition and public health. 🙂

  • Good article, I think that overweight people are more exposed when it comes to diseases, or food related diseases, not to mention if combined with smoking, alcohol or stress.

  • Michael


    Could someone link papers for me that specifically reference FAT as being the CAUSE of a disease? Not habits, not diet, not age, not amounts of exercise, but FAT, and FAT alone?

    First, no one is saying that fat is THE cause of disease — just that it’s a cause, just like smoking, air pollution, saturated fat, various genetic mutations and variants, and aflatoxin.

    Even granted that: To some extent, I think that what you’re asking sets a bar that is logistically almost impossible to jump with the scientific method, without some very unethical and expensive human experiments, precisely because habits, diet, and amount of exercise are, of course, the main thing that determine body fat. But scientists make their best efforts to disentangle them all, and the evidence is pretty clear that fat per se (and especially visceral fat (surrounding your internal organs in the gut area) and even more so ectopic fat (deposited in the liver and in muscle tissue in response to chronic energy imbalance) badly disrupt normal metabolism and lead to chronic disease and death.

    I suppose the most direct evidence that fat tissue per se is a health problem is a series of studies in rats, in which scientists removed the visceral-type fat (which includes, in these animals, ‘epididymal’ fat) from one group of rats and compared them to others where they didn’t. This improved insulin sensitivity, reversed age- and diet-induced insulin resistance, and extended their lives — even in animals that were not especially overweight (but more dramatically in those that were).

    You can’t do these kinds of experiments in humans, because it’s too difficult and dangerous to remove visceral and ectopic fat from humans; liposuction only targets the subcutaneous fat, and has no such benefits. But a close second is bariatric surgery. Despite the obvious riskiness of surgery itself, when obese people have their energy intakes substantially reduced, they immediately begin enjoying healthier metabolism, their risk of diabetes goes down, pre-existing diabetes is often reversed, and on average they live substantially longer. This is especially notable in the early days after surgery, when they live on what I think we can agree is a pretty poor diet (Ensure-type sugared protein drinks with vitamin supplements), and can’t exercise because of the need to recover from the surgery itself, yet they still enjoy dramatic metabolic improvements almost from day one.

    I’m not convinced that smokers should be excluded or analyzed separately. If the question is about the effect of BMI on longevity – you can’t just exclude an unhealthy habit that may fall predominantly in one group.

    But this is exactly what you’re calling for in your opening sentence: excluding unhealthy habits and looking at the fat alone. If smoking kills (and it does) and tends to make people thinner (and it does), then looking at smokers to see if thinness is healthy is bound to be misleading. And, as I noted, the data shows this clearly when you actually compare the results of running the data both ways.

    really, is it so impossible to believe that life, death, and illness could be more complicated than just fat or thin??

    Again: no one is arguing that life and health are entirely the result of fat (or even visceral and ectopic fat). Even if you’re thin, you can still make your health worse if you smoke, don’t exercise, live next to an old coal power plant, and eat a lot of sugar and trans-fatty acids while eschewing vegetables. But if you are carrying around a lot of visceral fat, and you don’t smoke, exercise regularly, live in an area with clean air, and eat plenty of vegetables and litle processed food (and, to be fair, the way that most people who are more than a little overweight got that way and stay that way is lack of exercise and/or an unhealthy diet) — still, if that’s you, you can still increase your odds of living a long and healthy life if you lose the toxic fat tissue.

  • Pingback: diet review workout – Saturday Stories: Big Food, Good Science vs. Bad Science, & Overweight Health()

  • Elaine Magee

    Maybe some of us are hedging our bets that ultimately it is best to eat and exercise for the health of it and let the pounds (and BMI) fall where they may. Our anti-obesity obsession isn’t helping the body image nor the eating disorder risk of our population.

    I’ve written 25 books and have never used the BMI tables. How can I when it groups men and women together and all ethnic groups. I choose health as my motivation and not pounds. And I strive as a nutrition educator to not promote obsession and restriction but to promote a healthy relationship with food, a total (mind, body, spirit) view of health and to foster health body image particularly in women. Call me crazy.