Select Page
Glenn DiCostanzo blog post

Most of us are overweight. That is what we are told, and as evidence, the News has images of out-of-shape people eating dessert on the sidewalk. At first glance it looks true, we are overweight. The data support the claim. But with 75% of adult Americans included, this statistic likely includes you too. Have you wondered what weight you are over? Looking closer at the data, it’s more forthright to state that 75% of adults are ‘categorized as’ overweight, and the fine print should add that it’s not uncommon to be categorized as overweight without being out-of-shape or unhealthy.

BMI (Body Mass Index) is implicated in the statistic. Worldwide, height and weight measurements are used to calculate BMI using a formula developed in the 1830s. BMI values enable simple comparisons between people of varying heights and weights. In the US, the National Institutions of Health and the CDC (Center for Disease Control) categorize BMI values to screen for health related problems. Many other countries have categories as well.

BMI receives scorn and criticism: It’s old, it’s a rule-of-thumb, it’s ethnically biased, it doesn’t differentiate lean muscle mass from fat, it doesn’t consider body shape or proportion,… The criticism may be justified, but BMI is just a number. The discontent is likely due to the weight-based categories which are independent of the BMI calculation.

Even before considering the accuracy of these categories, it seems reasonable to ask: why bother categorizing people with suggestive labels? Why attempt to simplify a small range of directly comparable numbers? Why call a BMI of 26 overweight when we could call it 26? If it is necessary to have categories to screen for health issues related to excess fat, why not measure fat directly?

Considering accuracy, since the BMI calculation does not differentiate lean muscle mass from fat, many fit athletes are classified as overweight. If a few fit persons were misclassified as overweight, we might conclude the category is conservative in screening for individuals with health problems, but this is not the case. A 2016 study* published in the International Journal of Obesity found that close to half of the individuals categorized as overweight, were metabolically healthy while over 30% of the respondents categorized as normal-weight were unhealthy.

***

In the graph above, the overweight label is applied to all BMI values of 25 and over; the majority of these persons are obese having a BMI of 30 and over. The 75% statistic comes from summing the overweight and obese categories. The table shows the sample percentages for the categories of Normal, Overweight, and Obese.

In the graph, adult weight categories are not applied to children and adolescents. The CDC uses age-based categories with a percentile approach where the overweight label starts at the 85th percentile. This is a good alternative method, but the percentiles are based on old data, not a child’s contemporary peers.

***

Dropping the weight categories isn’t meant to deny the existence of a global obesity epidemic which is often the story behind the headlines and images. Problem solving with respect to public health and the global obesity epidemic could benefit from precise language. A problem well defined is a problem half solved.* Our weight, physical fitness, and health are common topics of discussion in many social settings. We are in this together, but the weight categories and labels dilute the information provided by the BMI value. These labels do more to support sensational headlines and marketers than the persons labeled.  Eliminating the categories will assist problem solving by removing the inaccurate labels, promote data literacy by allowing the numbers to tell the story, and benefit social discourse by reducing the apprehension associated with being labeled. 

Glenn DiCostanzo
December 2021

Figure Caption: The data shown in the graphs and table are from the 2017-2020 NHANES. Python makes it easy to tag, count, group, and graph respondents based on weight categories and age using the pd.cut and groupby methods. The libraries used in this analysis include pandas, numpy, matplotlib, and seaborn.  The graph’s graphics were added in Adobe Illustrator.  The graph below provides and alternate perspective on the data shared in the table.

 

*AJ Tomiyama, 2016, “Misclassification of cardiometabolic health when using body mass index categories in NHANES 2005–2012,” International Journal of Obesity. 

 *”A problem well stated is a problem half solved.” Charles Kettering

Glenn DiCostanzo blog post
Share This