BMI: The Badly Misused Index

BOTTOM LINE

Body Mass Index (BMI) – a statistical tool to screen for the macro-nutritional status of populations – to see whether populations are generally underweight, adequately nourished or overweight. It’s a measure of your weight (in kgs) relative to your height (in meters) squared, but an inadequate measure of your body fat percentage (BF%) and skeletal muscle mass.

HOW SHOULD BMI BE USED?

BMI is cheap, quick and reasonably useful for tracking patterns and prognoses of certain chronic cardiometabolic diseases and nutrition trends. Researchers can compare large groups of people without needing expensive scans or detailed body composition assessments.

But in a clinical setting, when assessing an individual, it really is no longer considered sufficient, on its own, to diagnose obesity or determine an individual’s health risk. Current recommendations are to interpret BMI alongside other anthropometric and clinical measures, such as:

• waist circumference
• waist-to-hip ratio
• waist-to-height ratio 
• hip circumference
• performance metrics of function and fitness
• physiological measures such as blood pressure, cholesterol levels, and physical examination

Together, these measures account for differences in sex, ethnicity, visceral adipose tissue and skeletal muscle mass, which are much more useful when assessing an individual’s metabolic health and allowing treatment to be tailored more appropriately.

THE BACK STORY

In 1832, the Quetelet Index, a statistical measure outlined by Adolphe Quetelet, a Belgian statistician, mathematician and astronomer, was designed to quantify the characteristics of the “normal” man. 

Then in 1950s, insurance entered the picture: Louis I. Dublin, statistician and vice-president of Metropolitan Life Insurance Company, developed tables of “normal” weights to classify their obese policyholders.

In 1972, Ancel Keys, a physiologist, christened the measure as ‘BMI’. Keys was actually just looking for a simple, accessible measure to use in population research. He did not intend the BMI to be used for medical purposes.

BMI wasn’t designed to diagnose individuals. It was designed to describe populations.

So, we have 2 statisticians and a physiologist merrily fiddling with a metric for assessing “averages in a population”, but not one of them was trying to determine if an individual was healthy.

Even for a lay statistician (yours truly, who was forced to study stats for 5 years 🙄), there are some rather niggly statistical problems with Keys’ 1972 BMI study. These include the facts that:

• The different study cohorts were not age matched.
• No women were included in the studies.
• The populations were not properly representative of different ethnicities and backgrounds

As if those statistical niggles weren’t enough, Keys himself agreed that BMI is not the best indicator of BF%. Recent research reinforced his observation, showing that BMI can simultaneously overestimate and underestimate body fat.

If you have read any of my previous blogs, you will already have heard me bang on about the fact that, when it comes to body fat, where you carry it matters, almost more than how much you carry. Your visceral fat, together with your skeletal muscle mass, are really important metabolic health variables, yet BMI tells us remarkably little about either.

You can have that tall, skinny individual who has to jump around in the shower just to get wet, yet carries very little muscle and earns a distinction on the BMI. Meanwhile, a short, lean elite athlete can score a mere C, simply because muscle pushes him/her into the “overweight” category.  And then there’s the poor eight-month-pregnant mum-to-be… she gets a very red F.

And yet, despite these fairly glaring shortcomings, laypeople, doctors, insurance companies, the health industry smorgasbord of coaches, trainers, dieticians, etc seem happy to overlook them and continue using BMI to inform health decisions about individuals.

We are happily led to believe that the same “normal” BMI range applies equally to a 62-year-old Asian woman and a 30-year-old Caucasian man. Never mind that women, older adults and most Asian populations weren’t properly represented in the original study. Yet if both have a BMI of 23, they earn exactly the same report card.

That’s quite a leap of faith.

Then body composition comes along and makes an even bigger mess of things. A person can have a “normal” BMI, but carry excess visceral fat and therefore have an elevated metabolic risk. Conversely, improving body composition through resistance training may increase muscle while reducing fat, causing BMI to stay the same—or even increase—despite genuine improvements in health.

The trouble is, BMI hasn’t the faintest idea what your weight is actually made of.

At one point, a BMI-defined “obese” group included men whose BF% ranged from “athletic” to “obese”.

As if confusing muscle with fat wasn’t enough, BMI doesn’t seem terribly interested in fitness either. Someone with a “normal” BMI who can’t actually run for the bus may be at greater risk of all-cause mortality than someone in the “overweight” BMI category who runs marathons.

In a remarkable study of more than 40 000 people, researchers compared BMI classifications with actual markers of metabolic health (e.g. cholesterol and blood glucose). Their findings suggested that more than 70 million American adults could be put in the wrong box—classified as metabolically healthy or unhealthy—based on BMI.

BMI is a very accessible metric for anyone with good maths or a calculator. It can also shape how people see themselves—particularly adolescents. Two teenagers may have very similar BMIs yet vastly different body compositions: one may be “skinny-fat”, carrying little muscle and more body fat, while the other is shorter, stronger and more muscular. Ironically, BMI may reward the former and penalise the latter. That can whisper to physical health, but shout at mental wellbeing.

In 2015, Nutall made a rather telling observation: at that time, if the average BMI of Western populations sat around the prescribed World Health Organization (WHO) cut-off for “overweight”, then more than half the population would inevitably end up being labelled overweight or obese.

Fast-forward two decades, and that’s pretty much where we’ve landed. Today, around 54–60% of adults in OECD countries (such as the UK, USA, Canada and Australia) are classified as “overweight” or “obese”, leaving only 40–46% in the “normal” BMI range. So, what the h&*# is going on? Are we witnessing an obesity epidemic, a genuine change in body composition, or have the BMI categories themselves become part of the story? Nuttall (2015) even questioned whether terms such as “overweight” and “pre-obese” unfairly stigmatise what has, statistically speaking, been a common body size for decades.

The debate is very much alive.

Have the BMI categories kept pace with the evidence? Have they been updated to better reflect women, people outside of the 18-to-60-years old cohort of Keys’ original study, and different ethnic populations?

Not really.

To be fair, there have been some attempts:

• In 1993 the WHO developed uniform categories of the index. Prior to this, it seems they were a bit stricter than the National Institutes of Health (NIH) in their BMI criterion for “overweight” 😉
• In 1997 the International Obesity Task Force expanded the number of obesity classes.
• In 2002 the guideline for the category of “overweight” for Chinese adults was revised.
• In 2015 a similar revision happened for Indian adults
• Any adjustments for age have not been widely adopted, despite evidence that BMI predicts health differently across the lifespan. Nor have the categories been updated to reflect the fact that Western European populations have become taller and heavier over the centuries.

The bottom line is that BMI’s ability to predict future health varies between different groups of people. The number itself hasn’t changed—but the context around it has. As with so much in health, its interpretation requires our old friend… nuance!

That doesn’t mean BMI should be abandoned. It has survived for nearly two centuries because it remains cheap, quick and reasonably effective for identifying broad population trends. That explains much of its extraordinary staying power.

The problem isn’t the calculation itself. It’s what we’ve come to expect of it. When assessing an individual, BMI should be interpreted alongside other anthropometric, physiological and body composition measures—not as a diagnosis in its own right.

Alternatively, if BMI is to continue being used as a predictor of individual health risk, its categories and cut-offs need to evolve to better reflect today’s populations.

BMI isn’t a bad index. It may simply be a Badly Misused Index.

BMI gives us a number. Good clinical practice interprets the person.

References plus supporting statements