HomeCirculationVol. 132, No. 5Identification of Obesity and Cardiovascular Risk in Ethnically and Racially Diverse Populations Free AccessResearch ArticlePDF/EPUBAboutView PDFView EPUBSections ToolsAdd to favoritesDownload citationsTrack citationsPermissions ShareShare onFacebookTwitterLinked InMendeleyReddit Jump toFree AccessResearch ArticlePDF/EPUBIdentification of Obesity and Cardiovascular Risk in Ethnically and Racially Diverse PopulationsA Scientific Statement From the American Heart Association Goutham Rao, MD, FAHA, Tiffany M. Powell-Wiley, MD, MPH, FAHA, Irma Ancheta, PhD, FAHA, Kristen Hairston, MD, Katherine Kirley, MD, MS, Scott A. Lear, PhD, Kari E. North, PhD, Latha Palaniappan, MD, MS, FAHA and Milagros C. Rosal, PhDon behalf of the American Heart Association Obesity Committee of the Council on Lifestyle and Cardiometabolic Health Goutham RaoGoutham Rao Search for more papers by this author , Tiffany M. Powell-WileyTiffany M. Powell-Wiley Search for more papers by this author , Irma AnchetaIrma Ancheta Search for more papers by this author , Kristen HairstonKristen Hairston Search for more papers by this author , Katherine KirleyKatherine Kirley Search for more papers by this author , Scott A. LearScott A. Lear Search for more papers by this author , Kari E. NorthKari E. North Search for more papers by this author , Latha PalaniappanLatha Palaniappan Search for more papers by this author and Milagros C. RosalMilagros C. Rosal Search for more papers by this author and on behalf of the American Heart Association Obesity Committee of the Council on Lifestyle and Cardiometabolic Health Originally published6 Jul 2015https://doi.org/10.1161/CIR.0000000000000223Circulation. 2015;132:457–472is corrected byCorrectionOther version(s) of this articleYou are viewing the most recent version of this article. Previous versions: January 1, 2015: Previous Version 1 Obesity, defined as excess fat (adipose) tissue accumulation that may impair health,1 is a highly prevalent and serious public health problem. Roughly 35.7% of American adults are obese.2 High rates of obesity are not limited to the United States or even to other highly developed countries. The prevalence of obesity in Mexico, for example, is comparable to that in the United States.3 Not surprisingly, rates of obesity-related illnesses including cardiovascular disease (CVD) are rising quickly worldwide. More than 25 million American adults have been diagnosed with diabetes mellitus.4 India is projected to have >100 million diabetic people by the year 2030.5 CVD is the number 1 cause of death worldwide.6 These grim statistics highlight the need for accurate identification of overweight and obese adults who are at high risk for obesity-related illnesses. Accurate identification of such people allows healthcare professionals, policymakers, and others to target prevention and treatment programs to those at the highest risk of morbidity and mortality. Unfortunately, the tools and measures currently available to identify obesity and associated risks are either impractical, inaccurate, or both. For example, the body mass index (BMI) is easy to calculate, and established cutoffs that define overweight (25 kg/m2) and obesity (30 kg/m2) are readily available and well known. The application of such cutoffs to a diverse population, as will be discussed, however, leads to misclassification of a large number of people. Many people with a normal BMI have high levels of adiposity and also are at high risk for obesity-related illness. Others with a high BMI have relatively normal levels of adiposity and are metabolically healthy. The problem of misclassification is especially important for racial and ethnic minorities, who make up nearly 40% of the American population today and will make up more than half of the population by the year 2050.7,8The purpose of this scientific statement is to describe the limitations of current simple measures, such as the use of BMI with standard thresholds to identify overweight and obesity, as applied to racial and ethnic minorities. The statement also addresses potential alternatives, as well as the diagnosis of obesity based on multiple risk factors, including race and ethnicity. On the basis of our extensive, systematic review of current literature and in collaboration with the American Heart Association’s (AHA) Obesity Committee of the Council on Lifestyle and Cardiometabolic Health, we have also made some key recommendations for clinical practice, research, and public health for improved identification of obesity and cardiovascular risks in a racially and ethnically diverse population.This statement is focused on racial and ethnic groups within the United States, although many of our conclusions are either drawn from or applicable to related populations around the world. We use the standard US Census Bureau classification of Hispanic origin and race.9 Although our statement is intended to encompass the full racial and ethnic diversity of the United States, we acknowledge that specific groups make up a large or growing segment of the population and also suffer disproportionately from obesity and CVD. Groups that are the focus of the discussion that follows include the following: (1) Blacks, non-Hispanic blacks, and those who identify as such and have primary or partial ancestry in sub-Saharan Africa. This group represents 42 million Americans (13.6% of the population).7 (2) Hispanic or Latinos (eg, people of Puerto Rican, Mexican, or Dominican origin), a group of diverse racial and ethnic origins that includes >50 million Americans (16.3% of the total population). Among Hispanics, Mexican-Americans, a group that includes those who identify as having primary or partial ancestry in the Republic of Mexico, represent 31.8 million Americans (10.0% of the American population). This group grew by 11.2 million people between 2000 and 2010 and now represents 63% of the US Hispanic population.10 (3) Asian Americans, the largest composite group of immigrants to the United States; this group, together with descendants, includes 17.3 million people (5.6% of the American population).11 This group includes East Asian Americans (people of part or whole Chinese, Korean, Japanese, Filipino, Vietnamese or other Southeast Asian ancestry) and South Asian Americans (people of part or whole Indian, Nepali, Pakistani, Bangladeshi, or Sri Lankan origin).This statement is intended to provide practical guidance for clinicians and therefore emphasizes simple anthropometric measures, including BMI, waist circumference (WC), and waist-hip ratio (WHR), alone or in combination with other measures of overall cardiovascular risk. However, we do briefly also describe the accuracy of currently less practical methods such as imaging and related methods for identification of adiposity in racial and ethnic minorities. The accuracy of anthropometric measures has been assessed on the basis of comparison to more reliable measures of adiposity (such as dual-energy X-ray absorptiometry [DEXA] scanning), as well as on the ability of such measures to distinguish between people with and without cardiovascular risks (eg, diabetes mellitus and other components of the metabolic syndrome). In some studies, the ability of anthropometric measures to predict cardiovascular or all-cause mortality has been estimated. For our purposes, we consider all 3 of these standards for evaluation of the value of anthropometric measures to be valid and important.We searched the PubMed database for English-language articles published between 1998 and 2013. As noted, our focus was racial and ethnic minority populations living in the United States. We limited our search to these populations and closely related populations from other parts of the world (eg, people of South Asian background living in India or the United Kingdom). Our search was limited to adults only. We used a variety of search terms in various combinations, including the following: obesity, adiposity, measurement, diagnosis, accuracy, race, ethnicity, blacks, Hispanics, Asians, Native Americans, south Asians, Indians, Chinese, BMI, WC, WHR, cardiovascular risk, and Framingham risk score. We considered both review articles and cross-sectional studies. The abstracts of all retrieved citations were reviewed for relevance. Relevant articles were shared and discussed by all authors. We also reviewed the reference lists of all reviewed articles to identify additional relevant studies. Members of the writing group who knew of relevant consensus guidelines, scientific statements, and articles not indexed in PubMed shared and reviewed these with other group members.Nature of the ProblemApplying Common Anthropometric Measures With Uniform Standards to Minority PopulationsBody Mass IndexBMI was developed as a measure of body fatness (adiposity) nearly 200 years ago by Adolphe Quetelet.12 BMI is the most widely used measure to identify overweight and obesity in research studies and is recommended in clinical settings as well. Cutoff points of 25 and 30 kg/m2 define overweight and obesity, respectively, according to both the National Institutes of Health (NIH) and World Health Organization (WHO).1,13 These cutoffs have been recommended for use in all populations and for both men and women. BMI is easy to calculate, and having uniform cutoffs in different populations greatly simplifies identification of overweight and obesity. As a diagnostic tool, however, using the BMI with these cutoffs is problematic for several reasons. As pointed out by Kuczmarski and Flegal,14 the origin of the current recommended cutoffs is not entirely straightforward. BMI cutoff points have evolved considerably over the past 30 years. At times, cutoffs were based on a reference standard, which means that cutoffs separated population segments based on the overall distribution of BMI. The recommended BMI cutoffs of 27.8 kg/m2 for men and 27.3 kg/m2 for women recommended by the NIH in 1985, for example, represented the 85th percentile for adults aged 20 to 29 years in the population.15 The current BMI cutoffs of 25 and 30 kg/m2 were based loosely on a criterion standard, which means that cutoffs were selected to reflect morbidity and mortality. Recent evidence, however, calls this rationale into question. There is some evidence, for example, that a BMI between 25 and 30 kg/m2 is associated with lower mortality than normal weight (BMI of 20–25 kg/m2) and that a BMI of 30 to 35 kg/m2 is not associated with increased mortality compared with normal weight.16 There is also evidence that current BMI cutoffs do not accurately predict the presence of cardiovascular risk factors or the incidence of heart disease in diverse populations.17,18Visceral adipose tissue (VAT) may better predict cardiovascular risk than BMI. VAT likely serves as the site of dysfunctional, hypertrophic adipocytes that promote inflammation, oxidative stress, and production of cytokines and adipokines.19,20 Consistent with the evidence of its metabolic activity, VAT has also been shown to predict CVD burden and incident cardiovascular events independent of other cardiometabolic risk factors.21–24 Additionally, VAT appears to improve cardiovascular risk prediction over BMI.18 Unfortunately, data describing the relationship between VAT and CVD in racially and ethnically diverse populations are extremely limited.23As a criterion standard for identifying people at high risk for cardiovascular morbidity, as well as cardiovascular and all-cause mortality, BMI with current cutoffs alone is of limited usefulness. An ideal tool for identifying overweight and obesity would be able to distinguish not only between people at lower and higher risk of cardiovascular events, cardiovascular morbidity, and cardiovascular and all-cause mortality but also between those with low and high levels of adiposity in general. Clearly, CVD and related mortality represent extremely serious consequences of obesity, but excess body fat has a much broader impact, for example, on the musculoskeletal system, sleep, the incidence of cancer, and a wide range of psychosocial issues. Can BMI with current cutoffs at least be used to identify people with excess adiposity? Adiposity can be accurately measured by a number of different techniques, including DEXA (“Nonanthropometric Imaging and Other Measures”). Unfortunately, even as a simple measure of adiposity, the current BMI cutoffs are inaccurate across the general population. The WHO reference standards for percent body fat (%BF) used to define obesity are 25% for men and 35% for women.25 Romero-Corral et al26 found that a BMI of 30 kg/m2 in a diverse sample of US adults had a high specificity (95% for men, 99% for women) to detect obesity defined by %BF but very poor sensitivity (36% for men, 49% for women). The inaccuracy of BMI in predicting adiposity across the general population has been confirmed by others.27The accuracy of BMI with standard, uniform thresholds as a measure of adiposity or cardiovascular risk is especially problematic in ethnic and racial minorities. Several studies have documented the poor sensitivity of BMI in Asian populations and the need for lower BMI thresholds. In a study of Japanese Americans, for example, in which elevated cardiovascular risk was defined as the presence of ≥2 risk factors (among them, high-density lipoprotein cholesterol [HDL-C] <40 mg/dL for men or <50 mg/dL for women; triglycerides >150 mg/dL; blood pressure >130/85 mm Hg or taking antihypertensive medication; or fasting plasma glucose >100 mg/dL, 2-hour oral glucose tolerance test >140 mg/dL, or use of glucose-lowering medications), the optimal BMI for identifying people at elevated risk was estimated at just 23.3 kg/m2 for women and 25.3 kg/m2 for men.28 BMI is also a poor measure of adiposity among Asians and Asian Americans.29 Just 7% of Asian American men and women have a BMI ≥30 kg/m2.30 According to one estimate, 25-year-old women of Chinese origin, for example, with a %BF of 35% (WHO threshold for obesity among women) have a BMI on the average of just 24.3 kg/m2, far below the BMI standard for obesity of 30 kg/m2.29The ability of BMI to predict cardiovascular risk and the associated optimal BMI cutoff values varies considerably in other racial and ethnic groups and between men and women. For example, the BMI cutoff with the maximum sensitivity and specificity for predicting diabetes mellitus among Mexican American men has been estimated at >27.8 kg/m2 compared with 29.5 kg/m2 among non-Hispanic white men. By contrast, the optimal cutoff among Mexican American women has been estimated at 30.4 kg/m2, higher than the cutoff of 27.7 kg/m2 among white women.31 Other estimates from Mexico itself are available that substantiate the need for lower cutoffs for predicting diabetes mellitus among men of Mexican origin compared with non-Hispanic white men. Data from the 2000 Encuesta National de Salud (ENSA), or Mexican National Health Survey, which enrolled 11 730 men and 26 647 women, were used to calculate optimal BMI cutoffs for predicting type 2 diabetes mellitus and hypertension in Mexican men and women. Optimal BMI cutoffs for predicting type 2 diabetes mellitus were 26.3 to 27.4 kg/m2 in men and 27.7 to 28.9 kg/m2 in women. Optimal cutoffs for predicting hypertension were 26.2 to 27.0 kg/m2 in men and 27.7 to 28.5 kg/m2 in women.32Another study of optimal cutoffs for predicting ≥1 of 3 risk factors (glucose >125 mg/dL or use of diabetes mellitus medication, blood pressure >140 mm Hg or use of antihypertensive medication, and low-density lipoprotein cholesterol >160 mg/dL, HDL-C <35 mg/dL for men/<45 mg/dL for women, or use of cholesterol-lowering medication) estimated comparable cutoffs of 23.9 kg/m2 for Mexican American men and 23.6 kg/m2 for white men.17 Optimal cutoffs differed significantly between Mexican American women (23.5 kg/m2) and white women (21.7 kg/m2). The optimal BMI cutoff value predictive of having >1 of 3 CVD risk factors for black men was estimated at 23.0 kg/m2, comparable to that among white men (23.6 kg/m2). In the same study, the optimal BMI cutoff among black women was 23.0 kg/m2, significantly higher than the cutoff of 21.7 kg/m2 among white women.17As among Asians, BMI with standard thresholds provides a poor estimate of adiposity among other racial and ethnic groups. Analysis of data and derivation of estimates from the National Health and Nutrition Examination Surveys (NHANES) by Heo et al33 revealed that non-Hispanic white men aged 18 to 84 years with a BMI of 30 kg/m2 have a %BF in the range of 29.8% to 32.3% depending on age, well above the WHO threshold for obesity. Non-Hispanic white men with a BMI of 25 kg/m2 have between 24.9% and 28.0% body fat depending on age, which suggests that a BMI threshold of 25 kg/m2 may be more appropriate for defining obesity in this group. BMI cutoffs of 25 and 30 kg/m2 also underestimate body fat among Mexican American men.33 Mexican-American men with a BMI of 25 kg/m2 have between 25.6% and 27.5% body fat; Mexican-American men with a BMI of 30 kg/m2 have between 30.1% and 31.9% body fat.Non-Hispanic black men have significantly lower %BF at BMI values of 25 and 30 kg/m2 than their white or Mexican-American counterparts. Non-Hispanic black men with a BMI of 25 kg/m2 have %BF in the range of 22.6% to 25.6% depending on age, and those with a BMI of 30 kg/m2 have %BF in the range of 27.5% to 30.0%.33 Among women with a BMI of 30 kg/m2, both non-Hispanic whites and Mexican-Americans have body fat percentages well above the WHO standard of 35%, with a range of 41.8% to 44.0%.34 As in the case of non-Hispanic black men, standard BMI cutoffs among non-Hispanic black women more closely correlate with the 35% body fat standard than for women of other backgrounds. Non-Hispanic black women have a %BF in the range of 35% to 37.7% at a BMI of 25 kg/m2 and 39.9% to 42.3% at a BMI of 30 kg/m2.34 On the basis of these estimates, Rahman and Berenson34 report that the sensitivity of BMI ≥30 kg/m2 for %BF >35% is greater among non-Hispanic black women (75% sensitive) than among non-Hispanic white (47.8% sensitive) and Hispanic (53.9% sensitive) women.The differences in body fat percentages described above may be associated with significant differences among different racial and ethnic groups in both skeletal and muscle mass. On average, non-Hispanic black men and women have greater skeletal and muscle mass than their non-Hispanic white counterparts, who in turn have greater skeletal and muscle mass than men and women of Asian origin.35–37 Skeletal and muscle mass in Hispanic men and women approximates that in non-Hispanic white men and women.37In conclusion, BMI with standard cutoffs for overweight and obesity does not reliably predict cardiovascular risk, cardiovascular morbidity, cardiovascular and all cause-mortality, or adiposity among men or women of different ethnic and racial backgrounds in the US population. The accuracy of BMI cutoffs in predicting adiposity is somewhat better among non-Hispanic blacks than among other groups.Waist CircumferenceElevated WC and WHR are associated with increased cardiovascular risk and premature death independent of BMI.38,39 Like BMI, both measures are correlated with overall adiposity, but they are more appropriately used as measures of abdominal obesity, which correlates with the amount of VAT, which in turn is strongly associated with cardiovascular risk.40,41 The mechanisms through which VAT leads to adverse metabolic effects are not well understood. Possible mechanisms include more labile release of fatty acids from VAT than from other adipose tissue and possibly a higher release of inflammatory molecules from VAT.42 Of the 2 measures, WC is easier to measure in a clinical setting because it requires only 1 measurement and has been more thoroughly studied. To assess obesity, the 2013 Guideline for the Management of Overweight and Obesity in Adults from the AHA, American College of Cardiology, and the Obesity Society recommends measurement of WC at least annually in all adults using cut points from either the NIH/National Heart, Lung, and Blood Institute (NHLBI) or WHO/International Diabetes Federation.43 Unfortunately, roughly half of primary care physicians seldom or never measure WC.44,45 Furthermore, errors in measurement and inconsistency of technique are common.46,47 The NIH/NHLBI cut points that distinguish between people at higher and lower cardiovascular risk are >102 cm (40”) for men and >88 cm (35”) for women.48 On the basis of a large random sample of people from the Netherlands, the WHO recommends cut points of >94 cm for men and 80 cm for women to identify those with increased risk of metabolic complications and >102 cm for men and >88 cm for women to identify those with substantially increased risk of metabolic complications.49 The International Diabetes Federation has established different cut points for populations of European (>94 cm for men and >80 cm for women) and Asian (>90 cm for men and >80 cm for women) origin.50There is considerable evidence that use of the NIH/NHLBI cut points in racial and ethnic minorities greatly underestimates the presence of metabolic abnormalities and cardiovascular risk. This is especially true in both South Asian and East Asian populations. For example, based on a cross-sectional sample of 15 239 Chinese adults aged 35 to 74 years, Wildman et al51 calculated WC cutoffs associated with the presence of ≥2 cardiovascular risk factors (among hypertension, dyslipidemia, and diabetes mellitus). A cutoff of 100 cm for men (slightly below the 102-cm NHLBI cutoff) had a sensitivity of just 8.5%, which means that >90% of men with ≥2 risk factors would be missed. Cutoffs of 85 and 90 cm in women had sensitivities of just 45.7% and 26.1%, respectively. Even the International Diabetes Federation’s more conservative cutoffs for Asian populations of 90 cm for men and 80 cm for women had sensitivities of just 32.8% and 61.6%, respectively. These results are consistent with other studies reviewed by Lear et al52 that indicated significant cardiovascular risk among both East and South Asian populations at much lower values of WC than among Europeans. The poor performance of the NIH/NHLBI cutoffs also applies to other populations in the United States. Zhu et al17 calculated WC cutoffs for the presence of ≥1 cardiovascular abnormalities (among hypertension, diabetes mellitus, or dyslipidemia, defined by standard criteria) based on data from the third NHANES. A cutoff of 101 cm had a sensitivity of just 37.4% (specificity 85.8%) among non-Hispanic black men and 40.6% (specificity 84.3%) among Mexican American men. The sensitivity was only slightly better (47.7%) among non-Hispanic white men. A cutoff of 89 cm had much better sensitivities of 74.0%, 81.4%, and 85.9% among non-Hispanic black, Mexican American, and non-Hispanic white men, respectively, although at the expense of specificity. Cutoffs of 83 cm and 94 cm had sensitivities of 84.2% and 57.5%, 83.2% and 57.9%, and 78.5% and 49.2% among non-Hispanic black, Mexican-American, and non-Hispanic white women, respectively. Additional data from the ENSA survey in Mexico estimated optimal cutoffs for predicting diabetes mellitus as 93 to 98 cm in men and 94 to 99 cm in women. Optimal cutoffs for predicting hypertension in the same study were 92 to 96 cm in men and 93 to 96 cm in women.28 The WHO, NIH, and International Diabetes Federation all recommend higher WC cutoffs for men and women. It is curious, therefore, that as is the case for BMI, optimal WC cutoffs for women of Mexican origin are higher than those for men of Mexican origin.Waist-Hip RatioThe WHO recommends cutoffs for WHR of ≥0.90 for men and ≥0.85 for women.23 Corresponding cutoffs from the US Department of Agriculture and US Department of Health and Human Services are ≥0.95 for men and ≥0.80 for women.53 WHR cutoffs have been thoroughly studied among Asian populations. A limited number of studies in other populations have recommended widely varying WHR cutoffs. On the basis of their review of available evidence, Lear et al52 recommend cutoffs of 0.90 and 0.80 for Asian men and women, respectively, the WHO cutoff for men and the US Department of Agriculture/US Department of Health and Human Services cutoff for women.Waist-Height RatioWaist-height ratio (WHtR) was proposed as a measure of central adiposity and a marker for future cardiovascular risk nearly 20 years ago.54 On the basis of a meta-analysis of 10 studies that included roughly 88 000 people, Lee et al55 concluded that adiposity indices that included a measure of central adiposity (WC, WHR, or WHtR) were superior to BMI in predicting cardiovascular risk, although no single measure that incorporated WC was superior to the others. Furthermore, combining BMI with a measure of central obesity did not improve the discriminatory value.55 A systematic review of 65 studies by Browning et al56 concluded that WHtR is superior to BMI for predicting CVD and diabetes mellitus. Unfortunately, subsequent studies from a number of countries including India, Iran, China, and Korea and from diverse populations within the United Kingdom, among others, have shown inconsistent results, which leaves the value of WHtR compared with other waist-based measures and BMI in question.57–60In summary, the universal NIH/NHLBI cutoffs for WC perform poorly as predictors of cardiovascular risk/metabolic abnormalities, especially in Asian populations. Cutoffs recommended by the WHO and the International Diabetes Federation perform slightly better. WHR has been less thoroughly studied, but based on available evidence, cutoffs recommended by the WHO and the US Department of Agriculture and US Department of Health and Human Services more accurately identify cardiovascular risk in diverse populations. Given, however, that WC is infrequently measured, it is unlikely that WHR, which requires 2 measurements and calculation of their ratio, is as practical a tool for most clinical settings. The value of the WHtR compared with BMI and other measures that incorporate WC is uncertain.What Are the Consequences of Misclassification?Missing Obesity in Large Numbers of PeopleReliance on current, standard, or universal anthropometric standards to diagnose obesity (BMI threshold of 30 kg/m2) may lead to missing excess adiposity in large numbers of people. This problem is especially relevant for populations of East Asian and South Asian origin. On the basis of the National Health Interview Survey, Oza-Frank et al61 reported that using the current BMI thresholds of overweight at 25 kg/m2 and obesity at 30 kg/m2, the proportions of Asian Americans who met the criteria for overweight and obese were just 29% and 7%, respectively, significantly lower than for the US population as a whole. As discussed above, however, Asians and Asian Americans have substantially higher levels of adiposity at the same BMI as non-Hispanic whites, and they also develop cardiovascular risks at lower levels of BMI.62 For these reasons, a WHO Expert Consultation recommended lower BMI thresholds for overweight and obesity among Asians. A comparison of these lower thresholds to the current US standards provides an estimate of the proportion of overweight and obesity that is missed if the thresholds of 25 and 30 kg/m2 are applied to different Asian groups61 (Table 1).Table 1. Proportion of Overweight/Obesity According to Different BMI ThresholdsAsian IndianChineseFilipinoOther AsianAsian (All)% Overweight (WHO Asian BMI standard: 23.0–27.4 kg/m2)46.738.246.540.743.0% Overweight (general BMI standard: 25.0–29.9 kg/m2)34.120.634.525.928.8Missed % overweight with use of general BMI standard12.617.61214.814.3% Obese (WHO Asian BMI standard: >27.5 kg/m2)16.68.820.814.515.2% Obese (general BMI standard: >30.0 kg/m2)6.74.210.27.37.1Missed % obese with use of general BMI standard9.94.610.67.28.1BMI indicates body mass index; and WHO, World Health Organization.Modified from Oza-Frank et al61 by permission of the American Diabetes Association. Copyright © 2009, the American Diabetes Association.Overdiagnosis of Obesity in Some GroupsPeople who are muscular but do not have excess adiposity have higher BMI, which allows a number of people to be inappropriately diagnosed as obese. Ode et al63 demonstrated in their study of athletes and nonathletes that the specificity of elevated BMI to diagnose excess adiposity in male varsity athletes, who have higher muscle mass, was only 27%.Using NHANES III data, Burkhauser and Cawley64 compared BMI with total body fat and %BF measured through bioelectrical impedance. On average, black females had 3.6 kg more fat-free mass (such as muscle, bone, and fluid) than white females, and black males had 1.3 kg more fat-free mass than white males (both differences were statistically significant). Black women also had on average 3.2 kg more total body fat, but their additional fat-free mass offset that, so their %BF was only 0.8% greater than that of white females, a nonsignificant difference. Black men were found not only to have more fat-free mass on average but also to have on average 2.3 kg less total body fat than white men. As a result, their average %BF was 2.9% lower than that of white men, a statistically significant difference. When %BF, instead of BMI, was used to define obesity, the race/ethnic gap in obesity between black and white women decreased significantly (with black women still significantly more likely to be obese). Using BMI criteria, 34% of black women were classified as obese, compared with 26% when %BF criteria were used. Similarly, for black males, the 19% classified as obese as defined by BMI cutoffs dropped to 13% when %BF was used (Table 2).Table 2. Overdiagnosis of Obesity in Bl