New study claims that “healthy obesity” is a myth

Recently, researchers from Mount Sinai Hospital in Toronto attempted to determine whether healthy obesity really exists. This report recently splashed into the news after an editorial published in the Annals of Internal Medicine declared that “healthy obesity is a myth“.

Consistent with this statement, the original report by Kramer and colleagues found that metabolically “healthy” obese people have a 24% higher rate of all-cause mortality relative to metabolically healthy people of normal weight. In reality, this 24% boost is little more than rounding error. In the same study, people with poor metabolic health had a 165-214% increase in mortality, regardless of body mass, Yes, even the skinny ones.

Study overview

Kramer and colleagues conducted a meta-analysis of studies that looked at the relationship between BMI (normal weight, overweight, obese), mortality (all-cause or cardiovascular deaths) and metabolic health. Metabolic health was determined in these studies based upon key determinants of health like waist circumference, fasting triglyceride (e.g. fat) levels, fasting glucose levels, blood pressure, high-density lipoprotein (HDL or “good”) cholesterol levels and the use of anti-hypertensive or glucose-lowering medications. Some studies also looked at inflammatory markers and insulin resistance (which is associated with diabetes).

In their analysis, the researchers included 12 prospective studies with a total of 67,127 participants that studied the relationship between BMI, mortality and metabolic health. About half of these studies had a follow-up for 10 years or longer duration, which represents a key difference between this study and past reports that have generally considered durations of less than 10 years.

I have reproduced one key observation that didn’t make it into the headlines in Figure 1 (below). As is apparent, there is a strong relationship between body mass index and metabolic health, with the ratio of healthy to unhealthy individuals shifting from about 7:1 in the normal weight category, less than 2:1 in the overweight class and less than 1:1 among obese individuals.

Figure 1. Proportion of individuals in the overall sample by BMI category (normal weight, overweight, obese) as a function of metabolic health. The proportion of metabolically healthy individuals decreases from about 7:1 in the normal weight group to less than 1:1 in the obese group.

Figure 1. Proportion of individuals in the overall sample by BMI category (normal weight, overweight, obese) as a function of metabolic health. The proportion of metabolically healthy individuals decreases from about 7:1 in the normal weight group to less than 1:1 in the obese group. More than 40% of obese people are classified as “healthy”.

In other words, people with higher body mass tend to have poorer metabolic health. Still, a large proportion of these obese peopleĀ (>40%) were still classified as healthy.

Additionally, the researchers divided participants into a series of groups: metabolically healthy people who were i) normal weight, ii) overweight or iii) obese and metabolically unhealthy people who were iv) normal weight, v) overweight or vi) obese.

When they compared these groups, a number of interesting findings emerged.

First, people who were metabolically healthy, regardless of body weight, had similar mortality rates, slightly favoring the normal weight group (by between 10% and 24%). However, this small difference was only significant when the researchers compared “healthy” obese with “healthy” normal weight individuals in studies that had a 10-year follow-up, with a relative risk of 1.24x higher in the obese group (24% difference). When studies with shorter follow-ups were included, the relative risk was similar but not significant.

In other words, there is a small, but significant effect of body mass on mortality in metabolically healthy individuals.

Second, when metabolically unhealthy individuals were compared to metabolically healthy people of normal weight, there was a massive effect of poor metabolic health, regardless of BMI. The relative risk of mortality was between 2.65-3.14 for all groups of metabolically unhealthy individuals (obese, overweight, normal weight). This represents a whopping 165-214% increase in the incidence of mortality among these groups, relative to the healthy normal weight individuals.

In short, having good metabolic health puts you in the clear. Being obese at the same time doesn’t have much of an effect.

The key factors that clearly predicted higher mortality in all groups of unhealthy individuals were elevated systolic blood pressure, decreased HDL levels, modestly elevated LDL levels and elevated fasting triglyceride and glucose levels. Healthy obese and healthy normal weight individuals differed more modestly on a number of factors, including waist circumference, insulin resistance, cholesterol levels and blood pressure.


What conclusions can we draw from this study?

1) It looks as though being obese leads to poor metabolic health, and poor metabolic health produces a major increase in mortality. Moreover, people with normal weight were 3-4 times more likely to be metabolically healthy than obese individuals. That’s number one, and should not be downplayed.

2) It is clear that for those people who are metabolically healthy, body mass has a small effect on mortality, which is tiny compared to the effect of poor metabolic health. Based upon this, I cannot agree with the conclusion that “healthy obese” individuals do not exist. Clearly they do, and in large numbers. In fact, the 1.24x increased risk in this healthy obese group was compared to healthy normal weight individuals. Relative to all normal weight individuals, it is unlikely that there is any significant difference in mortality.

While the authors of this paper propose that metabolically healthy obese individuals are slowly developing disease risk, this study does not directly test this hypothesis. It is true that people in the “healthy obese” category have higher levels of some metabolic markers of poor health than normal weight individuals, but clearly these markers have not caused large increases in mortality over the 10-year span studied here. In order to show that healthy obese individuals are transitioning to unhealthy obesity (rather than being a completely distinct group of people), a long-term follow-up comparing metabolic markers of health across the obesity cycle would be necessary.

Overall, I suspect that the safest bet is to avoid obesity in general. This should cut the risk of poor metabolic health dramatically and lead to a reduced chance of mortality, consistent with most health advice. However, there doesn’t seem to be much evidence to completely ignore people who are both healthy and obese. Based upon this study, there are likely lots of these “healthy obese” individuals, and further, they only appear to have slightly elevated levels of mortality. This doesn’t imply that everyone should assume that metabolic health is simply a stroke of luck: while there are probably a number of genetic factors involved, I suspect that lifestyle (diet, exercise) also plays a pretty big role in this metabolic advantage. I will have to look into this at some future date.

ResearchBlogging.orgKramer CK, Zinman B, & Retnakaran R (2013). Are Metabolically Healthy Overweight and Obesity Benign Conditions?: A Systematic Review and Meta-analysis. Annals of internal medicine, 159 (11), 758-69 PMID: 24297192


3 thoughts on “New study claims that “healthy obesity” is a myth

  1. Carolyn

    The Kramer study is really flawed. It doesn’t adjust for anything, not even for age and sex, which are obvious confounding factors. A much better meta-analysis on the same topic (with most of the same studies) was published in October: Combined effect of obesity and cardio-metabolic abnormality on the risk of cardiovascular disease: A meta-analysis of prospective cohort studies Jingyao Fan, Yiqing Song, Yu Chen, Rutai Hui, Weili Zhang International journal of cardiology 12 October 2013 (volume 168 issue 5 Pages 4761-4768 DOI: 10.1016/j.ijcard.2013.07.230)

    1. zstansfi Post author

      Thanks for letting me know – I assumed that the studies included in the meta-analysis already accounted for confounds. I will try to take a look at the details of the two papers and write-up a further discussion soon.

      Ideally, it additional to age and sex it would also be good to see illness, alcohol use and smoking-status included in a regression model.


Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>