A Second Look at the Link Between Obesity and Mortality

A controversial study published earlier this year in the Journal of the American Medical Association shows that overweight people have significantly lower mortality risk than normal weight individuals, and slightly obese people have the same mortality risk as normal weight individuals.

This meta-analysis, headed by statistician Katherine Flegal, Ph.D., at the National Center for Health Statistics, looked at almost 100 studies that included 3 million people and over 270,000 deaths. They concluded that while overweight and slightly obese appears protective against early mortality, those with a body mass index (BMI) over 35 have a clear increase in risk of early death. The conclusions of this meta-analysis are consistent with other observations of lower mortality among overweight and moderately obese patients.

Many public health practitioners are concerned with the ways these findings are being presented to the public. Virginia Hughes in Nature explains “some public-health experts fear…that people could take that message as a general endorsement of weight gain.” Health practitioners are understandably in disagreement how best to translate these findings into policy, bringing up the utility of BMI in assessing risk in the first place.

Walter Willett, chair of the nutrition department at the Harvard School of Public Health, told National Public Radio that “this study is really a pile of rubbish, and no one should waste their time reading it.” He argues that weight and BMI remain only one measure of health risk, and that practitioners need to look at the individual’s habits and lifestyle taken as a whole.

In the same JAMA issue, authors Steven Heymsfield and William Cefalu, both physicians, also argue that “not all patients classified as being overweight or having grade 1 obesity, particularly those with chronic diseases, can be assumed to require weight loss treatment. Establishing BMI is only the first step toward a more comprehensive risk evaluation.”

BMI is a comprehensive tool that measures height-to-weight ratio and was originally devised in 1832 by Adolphe Quetelet. Insurance companies, who performed the first large-scale studies on the correlation between health and death, found that overweight individuals tended to die earlier than those of normal weight, and used BMI as their metric.

Subsequent studies have shown as an individual’s BMI goes up, so does the risk of developing type 2 diabetes, hypertension and heart disease. In 1972, Ancel Keys published the “Indices of Relative Weight and Obesity,” and BMI began to be used as the best predictor for health by public health agencies and practitioners. In 1998, the NIH categorized BMI arbitrarily to the easy-to-round numbers that we are familiar with today.

While the advantage of using BMI is that it is cheap and easy, the downside is that it fails to distinguish between lean and fatty body mass. Many studies depend on self-reported height and weight, yet people tend to overestimate their height, and underestimate their weight, giving them a lower BMI than they actually have. So while BMI is publicity friendly and something we are all generally familiar with, it is an imprecise measurement of health.

Taking a waist circumference may be a better way than BMI to assess health risk (waist-to-hip ratios can also be used, although waste circumference is easier to measure and interpret, making it a better choice). Abdominal obesity, or having an “apple shaped” body, is associated with a higher risk of death from heart disease even at normal BMI’s. In fact, multiple studies show that waist circumference is a more effective predictor than BMI of risk of death from heart disease and type 2 diabetes.

Since the publication of the JAMA article, the American Medical Association has released a resolution defining obesity as a disease. In this declaration, the AMA discusses the merits of using BMI along with waist circumference as a metric for diagnoses (H-440.866). It remains unclear, however, how many practitioners are even familiar with what a waist circumference is or how to measure it correctly.

Epidemiologists are hesitant to jump to the conclusion that being overweight or slightly obese may be “good” for you. There may be explanations for the protective effect that do not have to do with weight at all. For example, as Flegal and her co-authors point out, heavier individuals may present with illness earlier and seek out care earlier, leading to better outcomes. They also may have a greater likelihood of receiving optimal medical care by conscientious practitioners who have learned to be hyper-vigilant about connections between weight and health.

There are also explanations that fat does have a protective metabolic effect in older individuals, which may bestow a benefit of more metabolic reserves. Researchers also argue that the weight-death relationship found in Flegal’s study may be confounded by smoking and sickness, where smokers and those who are chronically ill tend to be thinner but die earlier. These issues are addressed in the JAMA article, but some researchers are unsatisfied to the extent of analysis.

Although death was the main metric in this study, it is important to remember that morbidities such as diabetes and cardiovascular disease continue to have strong evidence to support a correlation with overweight. This makes clarifying messages to the public even more challenging. In an editorial in Nature, the controversy surrounding this argument boils down to how to effectively communicate these studies to the public.

The editors maintain that “the political mantra on public-health advice is clear: don’t send mixed messages.” How best to present these studies to an American public where 40% of men and 30% of women are overweight and costs associated with obesity are $150 billion dollars annually, remains a major public health challenge.

Lua Wilkinson, MA, RD, CDN is a PhD student in Nutrition at Cornell University, where she studies digital social networks in the promotion of maternal and child nutrition globally. Her work has appeared in The Atlantic, Tea Leaf Nation, The China Beat, Savage Minds, and Asia Healthcare Blog.

24 replies »

  1. Interesting discussion – One thing that is being missed completely is that there is a definite answer that already exists that solves all these problems – It is Health at Every Size – No need for measurements of any kind – weight, BMI, waist, bio-electric, etc. etc. Many larger people are perfectly healthy and we have no way to “fix” any of the above mentioned parameters in a permanent way with any but a tiny minority of folks – and weight loss treatments all carry a high risk of iatrogenic consequences both psychological and physiological – The answer – Health at Every Size is common sense, evidence based and quite simple – though not necessarily easy – weight neutral – for example – if you have a fat person with type II diabetes – treat them the same way you would a thin person with type II diabetes – research for any of this available on request – Jon

  2. I would question why even use BMI? Why not use HbA1c and dysmorphia and… If BMI is the initial screen, how many non-obese people are left untreated due their “healthy weight”. If everyone were screened annually for the Metabolic Syndrome (which ironically does not use weight as criteria), then all would have a better opportunity of catching abnormal conditions early, right?

  3. It would be interesting to re-evaluate “obesity” data with different BMI/obesity guidelines. I recommend 40 obesity. I think there would be fewer “paradoxical” observations if the guidelines/criteria better matched the data. I think the problem is that NHANES I data helped set initial guidelines yet NHANES II & III do not paint same picture.

    However, this would decrease obesity prevalence and, thus, interest in obesity prevention, right? Only about 30-35% would be overweight + obesity; and only 10-15% obese. A far cry from the 70% value that is used now.

    Thanks Lua

  4. Maybe this is usefull– Obesity can be viewed as a syndrome or a collection of signs and symptoms. Some are behavioral, some are genetic and some are metabolic. The treatment is similarly complex. Let’s call it a ‘lifestyle change’ which includes a mixture of medication, exercise, stress management (even prayer) etc. So, this condition may best be defined like conditions in the DSM-V. For example, a BMI over 25 AND dysmorphia AND an A1c over 7 AND … It may be best treated in the same way. Therapy and diet and exercise and stress management and medication.

    The bottom line issue then is that western medicine is not set up to define or treat a complex and chronic condition that crosses disciplines. Perhaps healthcare is too fragmented to deal with obesity without a paradigm shift.

  5. Maybe related to this – I would like more definitive information on the role of insulin in obesity, type 2 DM, metabolic syndrome, perhaps hypertension. It seems that we conflate type 1 and type 2 too much – assuming the problem in type 2 is blood glucose levels due to too little insulin when it may be too much insulin causing insulin insensitivity. I’m curious if we’ve done enough to understand this, when insulin is a powerful appetite stimulant and we don’t recognize the benefit of managing insulin/glucose metabolism in an entirely different way. I haven’t heard enough on this – it seems too easy to say to obese people – stop eating, exercise; if you can’t do that it’s your own fault and we can wash our hands of you.

  6. Why don’t we deal with weight issues in terms of treating disease processes? E.g. – if we have studies that indicate that folks with hypertension and DM – those dx processes dissipate with weight loss, it seems to me that that could be a recommendation along with exercise and fruits and veggies – i.e, don’t “treat” weight, per se, but as a factor to be modified only inso far as it contributes to other uncontroversial actual disease processes ….

  7. The AMA defining obesity as a disease was a big mistake, since the definition of obesity is based on BMI and as the authors suggests, BMI is a poor general indicator of obesity. If you can’t define a disease, don’t call it a disease.

  8. Susan, I could not agree with you more. In fact, my colleagues and I at Cornell often have this same discussion: how do we create useful public health messages that do NOT further stigmatize obesity? Do we even have any?

  9. Susan, I wish you had put your last name so I could write to you individually and tell you that this comment should be tweeted out to the entire wellness industry, which is basically built on BMIs.

    Just like Edward Everett, the keynote speaker at Gettysburg said to Abraham Lincoln, “I wish I had captured in two hours what you said in 2 minutes,” you just obliterated years of work by the wellness ignorati with one 200-word comment

  10. It seems to me that we have the completely wrong emphasis here. We don’t need a CLEAR message. We need a USEFUL message that is also an ACCURATE message.

    We seem to be determined to panic over a health statistic we understand poorly and have even fewer resources to treat. Even if BMI really is the gold standard in health risk assessments and the problem really is weight rather than something else correlated with weight, we still don’t have any terribly good obesity treatments.

    To make an extreme connection, it seems rather like making sure you say clearly, “Cancer is strongly correlated with risk of death. We recommend that people with cancer stop having cancer.”

    It seems to me that the best advice would be to calm the heck down about obesity and continue to research until there is a message that is fit for public consumption. Meanwhile, there are plenty of healthy behaviors with strong research behind them from which we can draw a clear and useful message.

    As a bonus, we can try to stop the social damage that this entire panic has already done to obese people.

  11. First, Lua, thank you for this thoughtful posting.

    Second, the lesson I learn here is that reducing the human condition to a number, particular one administered by HR people, is a dumb idea. Perhaps there should be an exception for morbid obesity but you don’t need a test to determine that.

  12. Do you think the AMA should think about increasing the BMI standard for overweight to something that matches closer to what this meta-analysis shows as higher risk for mortality? Even when individuals with “lower” BMI’s maintain a higher risk for morbidities such as diabetes, heart disease and hypertension?

  13. I think part of the issue IS just this: who is defining overweight, obesity, unhealthy? In my research, it appears that these are defined arbitrarily by numerous organizations. Part of the difficulty in this meta-analysis was that everyone (including Walter Willet) has their own way of defining BMI, and the gold standard isn’t particularly “golden” in being able to define risk.

  14. BMI IS inadequate, and study after study show that waist circumference is a better measurement than BMI. We can discuss the danger or unimportance of mixed messages, but the bottom line is: why are we using BMI when waist circumference is better? And why didn’t the AMA do a better job of addressing this in their statement on obesity-as-disease?

  15. Health in complicated, and it’s a bit silly to think we can honestly have a single twitter-like blurb, as much as those savvy in social media might like it. A BMI of 27 means very different things in an active 25 year old, an inactive 25 year old with type two diabetes, an inactive 45 year old with hypertension, and an active 90 year old. Now add in ethnicity (certain ethnicities are affected more or less by weight gain when it comes to diseases like diabetes), other dietary habits, environmental exposures, etc. You just can’t simultaneously generalize and be accurate for even the 97.5% of the population that the dietary guidelines seek to cover adequately. Meta-analysis are lovely, until it comes to a specific patient with their own quirks.

  16. Simple solution – just redefine “normal weight” as 25-29.9 …
    As pointed out, the original descriptions of “normal”, “overweight”, etc were arbitrary – so just match up the descriptions with the “optimal” BMI these studies indicate …

  17. One more nail in the wellness industry coffin, where BMI is the Gold Standard in health risk assessments, along with getting PSA tests for prostate cancer, of course

    I play on an ultimate frisbee team where two of my best and fastest teammates have BMIs at or close to 30 — they are the poster children for why this wellness industry obsession with BMI is nonsense

  18. Clearly BMI is inadequate as an obesity measurement. Just look at the difference in body composition between males and females of the same BMI, and also how body composition changes with age. If someone weighs the same at age 70 as at age 20, they do not have the same bodies! Work needs to be done to determine what people should be told. People can handle the truth and won’t just gain weight because now they’ve been told it’s okay. I recommend: don’t worry about mixed messages; make sure the message is accurate, and people can dive deeper if they so choose. They are now spending $$$ and much effort to lose weight – they deserve to understand why with as much specificity as professionals are able to.