Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Misclassification of cardiometabolic health when using body mass index categories in NHANES 2005–2012


The United States Equal Employment Opportunity Commission has proposed rules allowing employers to penalize employees up to 30% of health insurance costs if they fail to meet ‘health’ criteria, such as reaching a specified body mass index (BMI). Our objective was to examine cardiometabolic health misclassifications given standard BMI categories. Participants (N=40 420) were individuals aged 18+ in the nationally representative 2005–2012 National Health and Nutrition Examination Survey. Using the blood pressure, triglyceride, cholesterol, glucose, insulin resistance and C-reactive protein data, population frequencies/percentages of metabolically healthy versus unhealthy individuals were stratified by BMI. Nearly half of overweight individuals, 29% of obese individuals and even 16% of obesity type 2/3 individuals were metabolically healthy. Moreover, over 30% of normal weight individuals were cardiometabolically unhealthy. There was no significant race-by-BMI interaction, but there was a significant gender-by-BMI interaction, F(4,64)=3.812, P=0.008. Using BMI categories as the main indicator of health, an estimated 74 936 678 US adults are misclassified as cardiometabolically unhealthy or cardiometabolically healthy. Policymakers should consider the unintended consequences of relying solely on BMI, and researchers should seek to improve diagnostic tools related to weight and cardiometabolic health.


The recently proposed rules by the Equal Employment Opportunity Commission (EEOC) allow employers to penalize an employee up to 30% of the cost of their health insurance if they fail to meet specific ‘health’ criteria, such as reaching a specified (lower) body mass index (BMI). Such a policy is based on the fact that overweight and obesity are associated with poor health conditions including diabetes, cardiovascular disease, hypertension and some cancers.1, 2 This kind of policy carries with it the major assumption that higher BMI individuals must uniformly face poor health. Yet, the relationship between BMI and health is complex, and focusing on between-BMI category variation in morbidity and mortality obscures substantial within-category variability in cardiometabolic health.3 Here, we test this assumption using the most recent nationally representative data available. We document the prevalence and demographic distribution of cardiometabolic health, highlighting the considerable number of individuals whose health status is misclassified when BMI categories are used as a proxy for actual health.

Misclassifying individual health on the basis of high BMI has numerous potential consequences. Not only do these types of punitive policies exacerbate the well-established economic consequences of being heavy,4 but they are also perceived as stigmatizing by heavier individuals,5 which can have a host of negative mental and physical health consequences.6 Furthermore, individuals with an overweight or obese BMI are often instructed by their physicians to lose weight. If these individuals are otherwise healthy, however, intentional weight loss may actually increase risk for mortality.7 The assumption underlying a policy like the EEOC’s also has potential consequences for lower BMI individuals. If these individuals are classified as healthy solely based on their BMI, they may not engage in proper preventive care or diagnoses may be delayed.

Given these potential consequences of misclassification, the goal of this study is to quantify the extent to which individual cardiometabolic health is mischaracterized when using established BMI categories. To do so, we draw on the most recently available nationally representative prevalence data on cardiometabolic health drawn from the 2005–2012 National Health and Nutrition Examination Survey (NHANES). We use the stringent definition of metabolic health described in Wildman et al.,3 which relies on the greatest number of criteria across multiple systems (blood pressure, triglycerides, cholesterol, glucose, insulin resistance and C-reactive protein), and was used with earlier data from NHANES.


NHANES provides data on the health and nutritional status of adults in the US through interviews and physical examinations. Sampling for NHANES is representative of the noninstitutionalized civilian US population and consists of ~5000 persons each year. The present analyses are based on NHANES participants 18 and older from 2005–2012, who completed the interview, examination and/or lab components of NHANES. N for analyses ranged from 12 351 (homeostasis model assessment) to 39 303 (demographics).

Age, gender, race/ethnicity and pregnancy were self-reported.8 For the present analyses, race was categorized as follows: Mexican American, Non-Hispanic white, Non-Hispanic black and other. Use of antihypertensive, lipid-lowering and antidiabetic medications was also self-reported.8

Height was measured to the nearest 0.1 cm by a stadiometer, and weight was measured in kilograms by a digital weight scale with participants wearing a standard examination gown.8 BMI values were calculated from measured height and weight values using the standard equation: weight(kg)/height(m)2. Waist circumference was measured to the nearest 0.1 cm at the end of participants’ normal expiration.8

Because a standardized definition of metabolic health has yet to be established, the present analyses used the definition outlined by Wildman et al.,3 which uses the greatest number of criteria among existing definition options. This defines metabolic health as 0–1 of the following metabolic abnormalities: (1) systolic/diastolic blood pressure 130/85 mm Hg or antihypertensive medication use; (2) fasting triglyceride level 150 mg dl−1 (1.69 mmol l−1); (3) high density lipoprotein-cholesterol level <40 mg dl−1 (1.04 mmol l−1) in men or <50 mg dl−1 (1.29 mmol l−1) in women or lipid-lowering medication use; (4) fasting glucose level 100 mg dl−1 (5.55 mmol l−1) or antidiabetic medication use; (5) homeostasis model assessment-IR >5.13 and (6) high-sensitivity C-reactive protein level >0.1 mg l−1 (0.95 nmol l−1).

Three consecutive blood pressure readings were averaged.8 Due to high prevalence of extremely low implausible diastolic blood pressure measurements, all diastolic blood pressure <35 mm Hg, were excluded in the present analyses. Triglycerides were determined by timed-endpoint.8 Glucose was determined by oxygen rate. High density lipoprotein-cholesterol was measured enzymatically through traditional precipitation methods.8 Insulin was measured using two-site enzyme immunoassay. CRP was quantified by latex-enhanced nephelometry.8

Statistical analyses

Data from the 2005–2006, 2007–2008, 2009–2010 and 2011–2012 data collection cycles were appended and the sampling weights modified as directed in NHANES documentation. All analyses were done on the non-pregnant subpopulation of the data. Female respondents who had a positive lab pregnancy test or self-reported as pregnant were excluded. Listwise deletion of missing data was done for all analyses.

Means/percentages were calculated for the overall population as well as five BMI categories: underweight, normal weight, overweight, obesity and obesity types 2/3 (combined due to low n). Logistic regressions controlling for age (top-coded at 80 years) were conducted using healthy versus not healthy as the outcome variable. Gender and BMI category were used as predictors.

Analyses were conducted using SAS 9.4 (SAS Institute, Cary, NC, USA) and SUDAAN 11.0.1 (Research Triangle Institute, Research Triangle Park, NC, USA). The sampling weight was adjusted for the multiple years following the method suggested in NHANES documentation.9 This revised sampling weight, clustering and stratification were incorporated into all analyses as recommended in NHANES documentation.10


Table 1 displays descriptive statistics. Table 2 presents population frequencies and percentages of healthy versus unhealthy metabolic status, stratified by BMI category. Although the relative percentage of healthy versus unhealthy individuals decreased in obesity, as expected, fully 19 761 047 obese US adults were classified as metabolically healthy. Supplementary Figure 1 displays the age-adjusted predicted population frequencies and percentages of healthy versus unhealthy stratified by BMI, further stratified by gender and race, respectively.

Table 1 Descriptive statistics of study sample, stratified by metabolic health status and BMI
Table 2 Estimated population frequency of metabolic status (%), stratified by BMI category, of non-pregnant adults

No significant race-by-BMI interaction emerged, F(12,64)=1.62, P=0.11). There was a significant gender-by-BMI interaction, F(4,64)=3.81, P=0.008, further qualified by examining specific meaningful combinations of gender and BMI. Pairwise comparisons within BMI-by-gender groups using Sidak correction for multiple comparisons indicated normal weight females had greater odds of being metabolically healthy than normal weight men (OR=1.41, P<0.001), as did women with type 2/3 obesity compared with men with type 2/3 obesity (OR=2.05, P=0.034). However, obese women were no more likely to be metabolically healthy than obese men (OR=1.13, P=0.909).


Overweight and obesity have long been considered uniformly detrimental to health, and recently proposed rules by the EEOC would codify this into policy. Yet focusing on BMI ignores overweight and obese individuals who are cardiometabolically healthy—nearly half of overweight individuals, ~29% of obese individuals, and ~16% of obesity type 2 and 3 individuals. For these individuals, having a healthcare provider prescribe weight loss could be a misuse of time, patient effort and resources. Focusing on BMI as a proxy for health may also contribute to and exacerbate weight stigmatization, an issue that is particularly concerning given healthcare providers evince high levels of anti-fat bias.11, 12 Moreover, this focus ignores the many individuals whose BMI is considered ‘normal’ yet are cardiometabolically unhealthy—30% of this population. When healthcare providers deem these individuals as ‘healthy’ merely because they are not overweight or obese, critical diagnoses could be delayed or missed altogether. Overall, we found that using BMI as the main indicator of cardiometabolic health misclassifies an estimated 74 936 678 individuals.

These results clearly indicate that health policies such as those proposed by the EEOC should not rely on BMI. Not only are such policies discriminatory, but they run the risk of overlooking more effective approaches. A recent component analysis suggests that the most effective health interventions are those that emphasize health behaviors, foster improved self-concept (for example, a sense of self-efficacy) and provide practical skills (for example, stress management); targeting weight and weight loss was found to be unnecessary to improve health.13 We recognize, however, that BMI may be seen as a quick, convenient and inexpensive marker of health in the clinical setting. Yet excessive focus on weight is likely to have detrimental consequences for the health and wellbeing of heavier individuals14 and thus should not be the principal outcome in health promotion efforts.15

Although obtaining blood markers is more time intensive, invasive, and costly, doing so can foster more accurate diagnosis and improved patient care. If lab markers are absolutely unobtainable, potential solutions are to instead use markers that researchers argue are a more accurate marker of health than BMI, such as physical activity and cardiorespiratory fitness,16, 17, 18 waist circumference19 or body fat percentage,16 or their combination. Regardless of the ultimate solution, the need for improved diagnostic tools related to cardiometabolic health is clear.

We contend that blood pressure, triglyceride, cholesterol, glucose, insulin resistance and C-reactive protein data are more accurate measures of health than BMI. However, this multi-system definition of cardiometabolic health should be confirmed using mortality data from longitudinal studies.20

In sum, a large proportion of US adults are misclassified as cardiometabolically unhealthy according to BMI categories, indicating that the EEOC and other entities should not rely on BMI when formulating health policy. Moreover, a clinical focus guided by weight and BMI may be misdirected. Future research should study overweight and obese individuals who are cardiometabolically healthy to understand how individuals can be healthy, no matter their BMI.


  1. 1

    Catenacci VA, Hill JO, Wyatt HR . The obesity epidemic. Clin Chest Med 2009; 30: 415–444.

    Article  Google Scholar 

  2. 2

    Guh DP, Zhang W, Bansback N, Amarsi Z, Birmingham CL, Anis AH . The incidence of co-morbidities related to obesity and overweight: A systematic review and meta-analysis. BMC Public Health 2009; 9: 88.

    Article  Google Scholar 

  3. 3

    Wildman RP, Muntner P, Reynolds K, McGinn AP, Rajpathak S, Wylie-Rosett J et al. The obese without cardiometabolic risk factor clustering and the normal weight with cardiometabolic risk factor clustering: Prevalence and correlates of 2 phenotypes among the US population (NHANES 1999-2004). Arch Intern Med 2008; 168: 1617–1624.

    Article  Google Scholar 

  4. 4

    Cawley J . The impact of obesity on wages. J Hum Resour 2004; 39: 451–474.

    Article  Google Scholar 

  5. 5

    Tannenbaum D, Valasek CJ, Knowles ED, Ditto PH . Incentivizing wellness in the workplace: Sticks (not carrots) send stigmatizing signals. Psychol Sci 2013; 24: 1512–1522.

    Article  Google Scholar 

  6. 6

    Hunger JM, Major B, Blodorn A, Miller C . Weighed down by stigma: How weight-based social identity threat influences weight gain and health. Soc Psychol Personal Compass 2015; 9: 255–268.

    Article  Google Scholar 

  7. 7

    Harrington M, Gibson S, Cottrell RC . A review and meta-analysis of the effect of weight loss on all-cause mortality risk. Nutr Res Rev 2009; 22: 93–108.

    Article  Google Scholar 

  8. 8

    National Health and Nutrition Examination Survey: Questionnaires, datasets, and related documentation. Available at:

  9. 9

    Task 2: When and how to construct weights when combining survey cycles. Available at:

  10. 10

    National Health and Nutrition Examination Survey: Analytic guidelines, 1999–2010. Available at:

  11. 11

    Tomiyama AJ, Finch LE, Belsky ACI, Buss J, Finley C, Schwartz MB et al. Weight bias in 2001 versus 2013: Contradictory attitudes among obesity researchers and health professionals. Obesity (Silver Spring) 2015; 23: 46–53.

    Article  Google Scholar 

  12. 12

    Phelan SM, Burgess DJ, Yeazel MW, Hellerstedt WL, Griffin JM, van Ryn M . Impact of weight bias and stigma on quality of care and outcomes for patients with obesity. Obes Rev 2015; 16: 319–326.

    CAS  Article  Google Scholar 

  13. 13

    Friedemann Smith C, Heneghan C, Ward A . Moving focus from weight to health. What Are the components used in interventions to improve cardiovascular health in children? PLoS One 2015; 10: e0135115.

    Article  Google Scholar 

  14. 14

    Hunger JM, Tomiyama AJ . A call to shift the public health focus away from weight. Am J Public Health 2015; 105: e3–e3.

    Article  Google Scholar 

  15. 15

    Ross R, Blair S, de Lannoy L, Després J-P, Lavie CJ . Changing the endpoints for determining effective obesity management. Prog Cardiovasc Dis 2015; 57: 330–336.

    Article  Google Scholar 

  16. 16

    Ortega FB, Lee D-C, Katzmarzyk PT, Ruiz JR, Sui X, Church TS et al. The intriguing metabolically healthy but obese phenotype: Cardiovascular prognosis and role of fitness. Eur Heart J 2013; 34: 389–397.

    Article  Google Scholar 

  17. 17

    Barry VW, Baruth M, Beets MW, Durstine JL, Liu J, Blair SN . Fitness vs. fatness on all-cause mortality: A meta-analysis. Prog Cardiovasc Dis 2014; 56: 382–390.

    Article  Google Scholar 

  18. 18

    Lavie CJ, De Schutter A, Milani RV . Healthy obese versus unhealthy lean: The obesity paradox. Nat Rev Endocrinol 2015; 11: 55–62.

    Article  Google Scholar 

  19. 19

    Plourde G, Karelis AD . Current issues in the identification and treatment of metabolically healthy but obese individuals. Nutr Metab Cardiovasc Dis 2014; 24: 455–459.

    CAS  Article  Google Scholar 

  20. 20

    Must A, Phillips SM, Naumova EN . Occurrence and timing of childhood overweight and mortality: Findings from the Third Harvard Growth Study. J Pediatr 2012; 160: 743–750.

    Article  Google Scholar 

Download references


AJT was supported by the Hellman Fellows fund. The Hellman Fellows Fund had no further role in the design of the study, in the collection, analysis and interpretation of the data, in writing the report and in deciding to submit the paper for publication. AJT and JMH conceived the study. AJT obtained funding. CW analyzed the data. AJT, JMH, JNC and CW interpreted the analyses and drafted the manuscript.

Author information



Corresponding author

Correspondence to A J Tomiyama.

Ethics declarations

Competing interests

The authors declare no conflict of interest.

Additional information

Supplementary Information accompanies this paper on International Journal of Obesity website

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Tomiyama, A., Hunger, J., Nguyen-Cuu, J. et al. Misclassification of cardiometabolic health when using body mass index categories in NHANES 2005–2012. Int J Obes 40, 883–886 (2016).

Download citation

Further reading


Quick links