Skip to main content

BMI, Obesity & mortality: three grand challenges

BMI, Obesity & mortality: three grand challenges

BMI (kg/m2) is one of the most widely used anthropometric measures and is virtually the sole criterion for judging obesity, its extent and its links to disease and mortality. When something is that widely used, there is a tendency to forget about any shortcomings in its use. In this blog I look at three aspects of BMI to remind one and all that in the BMI-obesity-mortality triangle, all is not rosy.

Challenge 1. How good is obesity at predicting body fatness.

It is important to recognise that the definition of obesity is based on a correlate of body fatness but is itself not a direct measure of human fatness. One study[1]has examined the relationship between % body fat measured using the technique of bioelectrical impedance and BMI in a large (13,601) sample of US adults. Using a statistical method of evaluating true and false positives, BMI was classed as “good” for men aged less than 60 years and as “excellent” for women of that age. For those above 60, the true sensitivity was classified as “fair” for men and again “excellent” for women. The authors note: “In our results, BMI showed an unacceptable low sensitivity for detecting body fatness, with more than half of obese subjects (by body fat measurement) being labeled as normal or overweight by BMI”.   Shorter people tend to be so because of shorter legs meaning that their trunk, the heaviest part of their body, contributes disproportionately to BMI compared to taller people. Indeed data exists to show persons with short legs can have BMI values that are 5 units above what they would be for persons of average height[2].

Challenge 2. Is the BMI-mortality the curve constant for age?

Conventional wisdom states that the relationship between BMI and mortality is U-shaped. That is, there is a minimal risk where BMI ranges from 20 to 25. On either side of this optimal range of BMI, obesity rates rise. They rise slowly in the initial deviation from this range and then they rise rapidly at BMI values below 18.0 or above 30. The U-shaped nature of the BMI-mortality link is often explained by the fact that at the lower end od BMI there are many smokers who are generally thinner that the average and also have a higher risk of disease. In 1983, the Royal College of Physicians (RCP) in London issued a report on obesity[3]. The report showed that the U-Shaped nature of the curve linking BMI and mortality is identical among “never smokers” and “smokers” (20 cigarettes per day). So the sharp rose in mortality at the lower extremes of BMI is not explained by smoking but may be explained by a higher percentage individuals who are light in weight as a result of some diagnosed or non-diagnosed disease.

The U-shaped mature of the BMI-mortality link is held to be universally true for both sexes and for all ages. But that is completely incorrect. The 1983 RCP report  looked at the link between BMI and mortality across ages. The rule that a BMI range for the lowest level of mortality was 20-25 held true up to the age of 50. For the next decade, mortality rose steadily as BMI fell below 22 and it rose again steadily above a BMI of 30. In between there was no rise in mortality with changing BMI. For those aged 60 to 69 years, the same steady rise in mortality below a BMI of about 22 was again seen but there did not appear to be any rise in obesity at any BMI above that value. When averaged across all ages, the conventional wisdom applied but above the age of 50, this was not the case. Two years later, researchers in the US confirmed that finding and they identified 23 other reports in the literature, which supported the notion that BMI should be looked at differently in different age groups[4].

Challenge 3. Is a higher BMI always bad ~ the obesity paradox.

Research conducted using US data gathered across three national surveys where weight and height were directly measured showed that the lowest risk of mortality was found to be in the BMI range of 25-30[5]. Now that was a direct challenge to conventional wisdom. This was true across all age groups and both in the total population and among non-smokers. The research also showed that the risk of mortality with obesity, fell with age. Indeed, among 70 year olds being obese increased the risk of mortality by 3% and being severely overweight, the in crease was 17%. Comparable figures for those in the age range 20-59 were 20% and 83% respectively. The obesity paradox had been born. Now it is a flourishing area of research. In 2013, the same team took a look at the global literature and ended up examining data from 97 studies, involving 2.88 million people with 270,000 deaths[6]. The same pattern was found. The nadir in the BMI-mortality link was found in what is deemed to be overweight with a BMI of 25-30. The “obesity paradox” holds true if we focus specifically on the disease most associated with obesity, namely diabetes. One study drew on the US National Health Interview surveys from 1997 to 2006 studied 74,710 subjects and divided them into those with or without diabetes[7]. Each of these was divided into quintiles (fifths) of BMI. For those in the lowest quintile of BMI, mortality was about 4.5 times higher among those with diabetes compared to those without diabetes. However, at the top fifth of BMI this figure fell to 1.7. In other words, as people with diabetes got fatter, they lived longer.  When looked at subjects who never smoked, the protective pattern of increasing adiposity remained. Many similar studies are appearing, almost all of which confirm the existence of an obesity paradox. So,

We can bury our hand in the sand and pretend that these challenges don’t exist or we can get together and try to design studies, which will help unravel these anomalies.

[1] Romero-Corral A et al (2008) Int J Obes 32, 959-966
[2] Garn S et al (1986) Am J Clin Nutr 44, 996-7
[3]Royal College of Physicians (1983) J Roy Coll Phys Lond 17, 3-58
[4]Anres R et al (1985) Ann Inter Med 103, 1030-3
[5]Flegal KM et al (2005) JAMA 293, 1861-1867
[6]Flegal KM et al (2013) JAMA 309, 71-82
[7]Jackson CL et al (2013) J Gen Intern Med 29, 25-33


Popular posts from this blog