UK United Kingdom

Monday’s medical myth: overweight people live longer

We’ve long known that too much excess weight increases your likelihood of dying prematurely. Or does it? A large review of the evidence published in the Journal of the American Medical Association (JAMA…

There’s no need to temper our efforts to address the obesity epidemic just yet. Image from

We’ve long known that too much excess weight increases your likelihood of dying prematurely. Or does it?

A large review of the evidence published in the Journal of the American Medical Association (JAMA) in January found that overweight people live longer than leaner people.

But don’t reach for the pie and chips just yet.

There is no denying that the high and rising prevalence of obesity and overweight is a major health problem. With 70% of men and 56% of women either overweight or obese, Australia is among the worst-affected countries in the world.

We commonly measure body size with body mass index, or BMI: a person’s weight in kilograms divided by the square of their height in metres. It is a good average measure of the amount of fat a person has in their body, which is useful for classifying people in population studies.

The World Health Organization (WHO) defines a “normal” BMI range as 18.5 to 24.9 and this corresponds to a weight of 49 to 65kg for Australian women of average height (1.62 m) and 58 to 77kg for men of average height (1.76 m).

Recent estimates have shown for the first time that a high BMI is now responsible for the greatest burden of disease in Australia and New Zealand, ahead of smoking and high blood pressure.

But the challenges represented by the obesity epidemic have raised the question of what the ideal BMI is for life expectancy.

Two large and well-conducted studies in the Lancet and New England Journal of Medicine examined this issue by gathering data on more than two million people who had their BMI calculated and were followed for their risk of dying over a defined time period.

The large numbers of participants and detailed individual data in these studies means the researchers were able to look at how small differences in BMI relate to the risk of death, accounting for a range of other factors known to influence this relationship, including illnesses that could potentially affect BMI.

Despite the diversity of populations covered by the studies, and the differences in methods, their findings are remarkably consistent: people with a BMI at the upper end of the WHO “normal” level (22.5 to 24.9) have the lowest death rates.

You can see this visually represented by the J-shaped curve in the graph below: as BMI goes up in increments of 2.5 above and below the 22.5-to-24.9 category, so do death rates.

Hypothetical data showing the risk of death by fine BMI categories

Why do people with a BMI at the upper end of the “normal” range have the lowest risk of death?

People with BMIs above this optimal level have an increased risk of dying, especially from heart disease, most likely due to increases in blood pressure, cholesterol levels and diabetes caused by excess body fat. They also have an increased risk of dying of cancer.

People with BMI levels below the optimal level also have increased death rates, particularly from respiratory diseases (such as chronic bronchitis) and cancer. The increased risk of death in the people with lower BMIs may also be because chronic illness has caused them to lose weight.

So with this evidence in mind, how does the JAMA paper reach such a different conclusion?

It comes down to the way the data from each of the studies have been collated and presented.

The JAMA review used broad classifications for underweight, normal weight, overweight, and obese and very obese, rather than the 2.5 increments of BMI. People with a BMI of 18.5 to 24.9 were included in the “normal weight” category; we can see from the graph above that this broad category includes people with the lowest risk of death, combined with people with a higher risk of death.

The “normal weight” category was then used as the comparison group for the studies, and has an average risk of death that is higher than the risk in the broad “overweight” category.

This skews the optimal weight finding and changes the shape of the curve, from J-shaped to tick-shaped.

Hypothetical data showing the risk of death by broad BMI categories

The result? One that suggests being overweight makes you live longer.

There are varying reasons why the researchers might have used these broad groups, including the fact that many studies are too small to be able to present statistically reliable results according to finer gradations in BMI.

The bottom line is that too much fat is bad for your health and increases your risk of dying prematurely. We still have a long way to go in our fight against obesity and, if anything, we need to redouble, not reduce, our efforts.

Articles also by These Authors

Sign in to Favourite

Join the conversation

15 Comments sorted by

  1. Colin MacGillivray

    Architect, retired, Sarawak

    Excellent to read this article, it answers my own misgivings about BMI numbers. The first graph shows BMI 19 and 27 have the same death risk. So the recommended BMI range (normal weight) should be a little bit higher, say up to 27. As a 1.8m tall, 80kg, BMI 24.7, 66 year old, no-one thinks I'm overweight.
    Probably the authorities think it's better to aim a bit low at 25 but it's dishonest.

  2. Gary Cassidy

    Monash University

    RE "The increased risk of death in the people with lower BMIs may also be because chronic illness has caused them to lose weight."

    Has there been any work to separate this out from the results? I know of a few people who have lost considerable weight before dying.

    1. Sheri Mills

      Secondary School Home Economics Coordinator

      In reply to Gary Cassidy

      Yes, agree Gary. Although I am delighted to find that my BMI of 23 is ideal in terms of statistical risk, I think that if the data excluded people who had lost weight due to terminal illness (and whose subsequent deaths have skewed the results), we would probably find that a BMI of 19/20/21 would give very similar results.

  3. Citizen SG


    The results for normal or underweight patients may be reflective of disease in an older cohort. For example, for a given population of 60+ year old people, a low to normal weight might actually reflect the result of COPD, cancer etc and the lack of lean body mass can be predicitve of less resilience to disease.

    I wonder if the study would show that a normal body weight young person is less at risk of disease than an overweight young person: especially at less risk of developing chronic disease later in life.

  4. Emily Banks

    Scientific Director of the 45 and Up Study at Sax Institute

    Thanks for all the thoughtful comments. There has been quite a bit of work to try to exclude people with weight loss due to chronic illness from studies looking at the relationship of body mass index (BMI) to death rates. The Prospective Studies Collaboration went so far as to ignore deaths occurring in the first five years after baseline and the New England Journal of Medicine paper by Berrington de Gonzalez reports findings separately excluding smokers and people reporting certain illnesses, such as cancer, at the time when their BMI was measured.

    The general theory is that because some of these diseases, especially chronic obstructive airways disease, have such a long period of illness prior to death, it is probably not possible to completely exclude the effects of illness on BMI. What we have is a best estimate from studies that have dealt with this problem, at least in part. This means that the likely optimal BMI may be even lower than the 22.5 to 24.9 range.

    1. Lennert Veerman
      Lennert Veerman is a Friend of The Conversation.

      Senior Research Fellow, School of Population Health at University of Queensland

      In reply to Emily Banks

      Thanks for your informative article.

      This issue of 'reverse causality' remains a challenge. I agree that the optimal BMI is probably lower than the 22.5 to 24.9 range. (Disclosure: my own BMI is around 18.5.)

      I won't be telling you anything you don't know, but I particularly like one very clever study by Davey Smith et al, reported in the BMJ in 2009 ( The authors eliminated reverse causality bias by using the BMI of people's children, instead of their…

      Read more
  5. Chris Borthwick


    Yes, plainly 'normal' should be from 22-27, shifting the green column sideways.

    My real query, though, is why the rise in obesity hasn't affected the all-causes deathrate at all. Australian life expectancy has continued to rise by three months a year, just as it has since 1890, with absolutely no bends or bumps (well, there was a glitch for a few years in the sixties, but that's long gone).

    1. Gary Cassidy

      Monash University

      In reply to Chris Borthwick

      Hi Chris,

      RE "Australian life expectancy has continued to rise by three months a year, just as it has since 1890"

      That's in interesting fact. Where did you get this from?

  6. Ariel Atkinson

    Research Assistant, University of North Carolina at Chapel Hill

    This article would greatly benefit from some statistics. What are the confidence intervals for each fine BMI? What differences are truly significant? I guess reading the original articles is required, and what I'm saying is I wouldn't get too hung up on the small incremental differences in BMI. It's not clear if there's statistically significant difference in risk among all the people in the normal category and the lower end of the overweight category.

  7. Abbie Noiraude

    logged in via Facebook

    Since the BMI was 'devised' between 1830 and 1850 has anyone ever recently tried to unravel, or even revisit its authority? It was 'devised' by a polymath, not a medical practitioner or a medical scientist.
    Why are we still sticking to something that was devised so long ago? We no longer believe in many theories ( eg a fat baby is a healthy baby) so why are we persisting with something that may indeed be a false assumption?

    Having such a narrow spectrum for what is 'normal' or what is healthy may be a misnomer. "Health" is far more than just a includes diet ( someone with a 'normal' BMI may be a smoker or a poor nutritional eater), genetics, exercise, stress, mental health, etc.
    Of course someone morbidly obese is not healthy, nor is an anorexia sufferer.
    Time for a revisit of this 'devised' number?

    1. Eric Anderson

      Eternal Student

      In reply to Abbie Noiraude

      Yes, BMI is a very flawed number. For a summary of some common criticisms (without, alas, citations) see:

      I've seen a variety of alternatives suggested for various circumstances (waist circumfrence, waist-hip ratios, etc.), but nothing seems to yet gained widespread acceptance.

      I attribute the continued use of such an obviously flawed metric to (a) prejudice and (b) path dependency. Standards are hard to change.

  8. John Doyle


    It's obvious that use of the BMI index is popular because it's so simple to discuss and use. Doesn't really matter in those terms if it's nonsense logic, it still says something of use.
    Might it be possible to massage it to get a more personal result?
    For example one to incorporate fat to muscle percentage. That would avoid fit athletes being classed as obese, for example. Even a multiplier of hip to waist difference would help.