Association of Total Body Fat Percentiles, Obesity and Cardiometabolic Health Biomarkers in Adults in the United States Authors: Grant Herrington, Luyu Xie, Zhenglin Yuan, John Clark, Anna C Rivara, M Sunil Mathew, Sarah Khan, Deepali K Ernest, Jaime P Almandoz, Sarah E Messiah Published: 2026 ABSTRACT ABSTRACTObjectiveTo estimate total body fat percentage (TBFP) percentiles and their association with obesity and cardiometabolic risk biomarkers in the United States adult population.MethodsCross‐sectional data for adults (20–59 years) from the 2011–2018 National Health and Nutrition Examination Survey (NHANES) were analyzed (N = 14,347). TBFP was assessed using dual‐energy X‐ray absorptiometry. TBFP percentiles were calculated using age‐adjusted, weighted quantile regression. Receiver operating characteristic (ROC) analyses assessed the ability of TBFP to predict obesity (body mass index [BMI] ≥ 30 kg/m2), elevated HbA1c (≥ 5.7%), and alanine aminotransferase (ALT) thresholds (≥ 20 and ≥ 40 U/L); and the ability of BMI to predict elevated TBFP.ResultsTBFP percentiles were similar across groups, but non‐Hispanic Black men had significantly lower odds of TBFP ≥ 90th percentile compared with non‐Hispanic White men. TBFP demonstrated accuracy in predicting BMI‐defined obesity (AUC 0.83–0.91), and BMI demonstrated accuracy in predicting elevated TBFP (AUC 0.80–0.90). TBFP had limited ability to predict elevated HbA1c (AUC 0.52–0.65) and ALT (AUC 0.51–0.63).ConclusionsTBFP percentiles are associated with BMI‐defined obesity but are insufficient to predict cardiometabolic risk independently. These findings support the continued use of BMI as a screening tool and suggest that TBFP may complement but not replace existing measures.Keywords: anthropometric, body mass index, measurement, risk, total percent body fat Introduction Obesity, defined as excess adiposity, is the most common chronic disease in the United States (US) and is associated with over 200 complications and increased mortality [1, 2]. Historically, obesity has been classified using body mass index (BMI), but there have been calls to reclassify obesity based on levels of adiposity and its impact on health [3, 4, 5]. At a population level, elevated BMI is associated with multiple chronic diseases [4]. However, obesity‐related complications, such as hypertension and type 2 diabetes (T2D), occur at different BMI thresholds across racial and ethnic groups, leading to proposals for race‐ and ethnicity‐specific BMI cutoff points for screening [6, 7, 8, 9]. Although BMI is strongly correlated with adiposity, it does not directly measure body composition, including total body fat percentage (TBFP) or fat distribution, key factors in obesity‐related complications [4, 10]. This limitation may contribute to its weaker predictive value for cardiovascular risk [11, 12, 13, 14] compared to other anthropometric measures, such as waist‐to‐hip ratio (WHR) and total body fat percentage (TBF%), which exhibit stronger associations [15, 16]. Body composition measurement quantifies adipose tissue and its distribution, distinguishing fat mass from lean mass to inform obesity‐related risk. Common techniques include dual‐energy X‐ray absorptiometry (DXA) [17], bioelectrical impedance analysis (BIA) [18], air displacement plethysmography [19], computed tomography (CT) [20], and magnetic resonance imaging (MRI) [21]. Studies have shown that beyond total fat mass, the distribution and type of adipose tissue play a critical role in cardiometabolic health, with visceral adipose tissue and intrahepatic triglyceride content demonstrating stronger associations with cardiometabolic risk than subcutaneous fat [22, 23]. Although obesity has traditionally been defined as a BMI ≥ 30 kg/m2, limitations of BMI have led to the consideration of alternative methods such as TBFP. Current TBFP cutoffs to define groups at risk are not well defined but are initially proposed as ≥ 25% in men and ≥ 30–35% in women, with more recent thresholds approaching ≥ 30% in men and ≥ 42% in women [24, 25]. Several large epidemiologic studies have examined the relationship between directly measured adiposity and cardiometabolic risk factors [26, 27, 28, 29, 30, 31, 32, 33, 34]. Analyses from nationally representative datasets such as NHANES as well as large cohort studies have demonstrated that total body fat percentage is associated with adverse cardiometabolic profiles, including dysglycemia, liver enzyme abnormalities, and dyslipidemia. However, these studies have also shown that the predictive performance of directly measured adiposity is often comparable to that of BMI rather than being clearly superior. As a result, uncertainty remains regarding the extent to which more precise measures of body fat improve risk stratification beyond BMI in population‐based settings. In the pediatric and adolescent populations, percentiles of anthropometric measures such as BMI, waist circumference (WC), waist‐to‐hip‐ratio (WHR) and TBFP have been used to predict cardiometabolic risk, characterize racial and ethnic group‐specific cut‐offs, and monitor changes in adiposity over time [27, 35, 36]. In adults, data on TBFP percentiles are limited and lack sufficient representation or power to evaluate differences by race or ethnicity [37]. Given the limited data on adult TBFP percentile estimates in diverse populations and the increasing use of body composition in assessing obesity‐related cardiometabolic risk, the objective of this study was to [1] develop nationally representative TBFP percentiles stratified by race, ethnicity and sex, and [2] evaluate the association and predictive performance of TBFP in relation to cardiometabolic risk biomarkers, while directly comparing its performance with BMI within the same population. We hypothesized that TBFP, as a more direct measure of adiposity, would demonstrate differential associations with cardiometabolic risk biomarkers across sex and racial/ethnic groups, and that its predictive performance would differ from that of BMI [38, 39]. Materials and Methods Study Design A cross‐sectional design utilizing nationally representative data from 2011 to 2018 National Health and Nutrition Examination Surveys (NHANES) was conducted. This study followed the STROBE reporting guidelines [39]. The Institutional Review Board at the University of Texas Southwestern Medical Center ruled this study to be exempt from review because of the use of publicly available, de‐identified data for analysis. The United States National Center for Health Statistics Ethics Review Board approved the study (protocol #2021‐05) [40]. Data Source NHANES is a cross‐sectional survey conducted in 2‐year cycles at the National Center for Health Statistics to monitor the health and nutrition status of the US population [41]. NHANES is a national survey designed to collect data regarding the health of US non‐institutionalized adults and children using a complex multistage probability sampling methodology. For each survey cycle, a complex 3‐step weighting method is created to represent the United States noninstitutional civilian population. The purpose of weighting the NHANES sample data is to permit the analysis of estimates that would have been obtained if the entire sampling frame had been surveyed. Weighting takes into account several features of the surveys: the specific probabilities of selection for the individual domains that were oversampled (e.g., Mexican Americans and non‐Hispanic Blacks were oversampled in the 2001–2002 surveys) as well as nonresponse and differences between the sample and the total population. When analyzing combined 4‐year data sets, the correct sampling weights must be used to produce unbiased estimates. Participant Recruitment and Data Collection Procedures NHANES employs a multistage, probability‐based sampling strategy to ensure a nationally representative sample. This process involved the random selection of primary and secondary sampling units, followed by the recruitment of households and individuals for in‐depth health assessments and interviews. Trained interviewers conducted face‐to‐face household interviews using the Computer‐Assisted Personal Interview system to collect data on demographics, medical history, and prescription medication use. Physical examinations, including anthropometric assessments and biospecimen collection, were performed by trained health professionals at the Mobile Examination Center (MEC). Biomarker data were obtained from participants who completed the NHANES examination component at MECs, where standardized laboratory protocols were used. Written informed consent was obtained from all participants aged ≥ 18 years. Participant Eligibility Criteria Adults aged 20 years and older were included in the analysis if they had available data on total body fat percentage (TBFP), BMI, glycated hemoglobin A1c (HbA1c), and alanine aminotransferase (ALT). NHANES measures HbA1c in whole blood samples using high‐performance liquid chromatography (HPLC), while ALT is measured in serum or plasma using a kinetic rate method. Data for this study were obtained from four NHANES cycles (2011–2012, 2013–2014, 2015–2016, and 2017–2018), which were combined in accordance with NHANES analytic guidelines to increase sample size and improve the precision and generalizability of estimates. From the combined NHANES cycles, a total of 39,156 participants were initially eligible. We conducted a complete‐case analysis, excluding participants with missing data on total body fat percentage (TBFP), body mass index (BMI), glycated hemoglobin (HbA1c), or alanine aminotransferase (ALT). Because NHANES is a repeated cross‐sectional survey, exclusions reflect missing data within survey cycles rather than loss to follow‐up. Across these cycles, 39,156 participants were available, of whom 14,347 met the eligibility criteria for inclusion in the analytic sample after exclusions. Sample sizes contributed by each survey cycle are provided in Supporting Information S1: Table S1. Measures Anthropometric Measurements Total Body Fat Percentage Dual‐energy X‐ray absorptiometry (DXA), also known as a DEXA scan, measures total body fat, bone density, and lean tissue mass, providing a detailed assessment of body composition. DEXA scans were administered to eligible survey participants aged 8–59 in the NHANES 2011–2018 cycles. Only adults aged 20–59 years old were included in this analysis. One of this study's primary objective was to estimate TBFP percentile estimates for the US Mexican American, other Hispanic, non‐Hispanic Black, non‐Hispanic White, non‐Hispanic Asian, and other/multi‐racial adult populations. Race and ethnicity, the moderator(s) of interest, were self‐reported by NHANES participants. Body Mass Index NHANES body measurements data were collected in the Mobile Examination Center by trained health technicians. Generally, one technician completed the measurements and data recording. Occasionally, one technician conducted the measurements, and another technician assisted in data recording. Standing height was measured without shoes using a portal stadiometer. Weight was measured using a portal digital weight scale. Body Mass Index (BMI) was then calculated by dividing weight in kilograms by height in meters squared (kg/m2). Obesity was defined as a BMI of ≥ 30.0 [42]. Cardiometabolic Disease Biomarkers Glycated Hemoglobin Anthropometric measures of excess adiposity correlate strongly with obesity‐related complications, such as type 2 diabetes (T2DM) and metabolic dysfunction‐associated steatotic liver disease (MASLD) [27, 36]. Glycated hemoglobin (HbA1c), a marker of glycemic status that reflects average blood glucose over approximately 90 days, is a standard measure for diagnosing dysglycemia, with values of 5.7%–6.4% indicating prediabetes and ≥ 6.5% indicative of diabetes. While HbA1c is commonly used in epidemiologic studies, it does not capture all forms of dysglycemia, such as impaired fasting glucose or impaired glucose tolerance. However, given its clinical relevance and availability in NHANES data, we evaluated the accuracy of TBFP in predicting an HbA1c ≥ 5.7%, representing the threshold for prediabetes [43]. Alanine Aminotransferase MASLD is defined as hepatic steatosis accompanied by at least one cardiometabolic disease risk factor and no other identifiable cause [38]. Alanine aminotransferase (ALT) was used as a biomarker of liver injury rather than as a diagnostic measure of metabolic dysfunction–associated steatotic liver disease (MASLD). American Gastroenterological Association guidelines recommend ALT as part of a non‐invasive testing to risk‐stratify and stage MASLD, with recent guidelines suggesting lower cutoffs (> 33 U/L in males, > 25 U/L in females) compared to the traditional threshold (> 40 U/L) [44, 45, 46]. Given the ongoing debate on optimal ALT cutoffs, we evaluated the accuracy of TBFP in predicting ALT thresholds of 20 U/L and 40 U/L. Statistical Analysis Age‐adjusted weighted quantile regression analyses were used to estimate the 10th, 25th, 50th, 75th, 90th, 95th, and 99th percentiles of the distribution of TBFP for men (unweighted n = 6848) and women (unweighted n = 7499) aged 20–59 years by self‐identified race and ethnicity group. Receiver‐operating characteristic (ROC) analyses were used to assess the ability of TBFP to predict HgA1c ≥ 5.7% versus < 5.7% in individuals without type 2 diabetes and the ability of TBFP to predict ALT ≥ 20 versus < 20 U/L and ALT ≥ 40 versus < 40 U/L. Survey weights were adjusted according to the NHANES analytic guidelines to account for the combined survey cycles. All participants included in the analysis met eligibility criteria for both body composition assessment and cardiometabolic biomarker measurement. To evaluate the ability of TBFP to discriminate individuals with cardiometabolic risk, we employed receiver operating characteristic (ROC) analyses. ROC curves are commonly used to assess the performance of continuous measures in classifying binary outcomes and provide a summary measure of discrimination through the area under the curve (AUC), enabling direct comparison with established metrics such as BMI. The area AUC was used as a summary measure of classification performance, with higher values indicating better discrimination. Specifically, ROC analyses were also used to evaluate the ability of TBFP to predict obesity defined by BMI (BMI ≥ 30 vs. < 30 kg/m2) by race/ethnicity among men and women and the ability of BMI to predict obesity by TBFP (TBFP ≥ 25% vs. < 25% for men; ≥ 30% vs. < 30% for women) by race/ethnicity among men and women. We used cutoffs of TBFP of 25% for men and 30% for women as these were more conservative estimates which are commonly used clinically. Briefly, in ROC analyses, area under the curve (AUC) can range between 0.5–1, with higher values indicating better discrimination between cutoffs. In the literature, different labels have been used to determine “poor” versus “good” and while they are arbitrary, they serve the purpose of comparison [41]. In this paper, we use a cutoff of ≥ 0.8 to indicate a “good” prediction capacity. In addition to ROC analyses, we estimated optimal TBFP cut‐points for each cardiometabolic outcome using the Youden Index to identify thresholds that maximized sensitivity and specificity. All analyses were performed using SAS version 9.4 (SAS Institute Inc.), and a two‐sided p < 0.05 was considered statistically significant. Results The final sample consisted of 6848 men and 7499 women after applying inclusion and exclusion criteria. Age‐adjusted weighted quantile regression analyses produced TBFP percentiles (10th, 25th, 50th, 75th, 90th, 95th, and 99th) by sex and race and ethnicity (Table 1). Overall, among men, 6.3% (95% CI: 5.4%–7.3%) had a TBFP at or above the 90th percentile, with prevalence ranging from 4.7% in non‐Hispanic Black men (95% CI: 3.5%–5.9%) to 7.3% in non‐Hispanic Asian men (95% CI: 5.6%–9.0%) and 7.2% in other/multi‐racial men (95% CI: 2.9%–11.6%). Among women, 8.7% (95% CI: 7.6%–9.7%) had TBFP at or above the 90th percentile, with prevalence ranging from 5.8% in other/multi‐racial women (95% CI: 2.1%–9.6%) to 9.1% in non‐Hispanic White women (95% CI: 7.7%–10.5%). TABLE 1The estimated prevalence and 95% confidence intervals of total body fat percentage percentiles by sex and race and ethnicity.Race and ethnicityPrevalence of total body fat percentage percentiles≥ 10th≥ 25th≥ 50th≥ 75th≥ 90th≥ 95th≥ 99thMenOverall67.9 (65.7–70.1)55.9 (53.9–58.0)36.4 (34.6–38.2)17.1 (15.7–18.5)6.3 (5.4–7.3)3.0 (52.4–3.6)0.6 (0.4–0.9)Mexican American70.0 (66.1–73.9)56.1 (51.9–60.2)35.5 (31.5–39.5)14.8 (11.8–17.8)5.2 (3.4–7.1)2.7 (1.3–4.0)0.8 (0.1–1.6)Other Hispanic71.4 (67.3–75.4)57.2 (52.5–61.9)36.8 (32.2–41.4)15.9 (12.0–19.7)6.4 (4.1–8.7)2.6 (1.2–3.9)0.9 (0–1.9)Non‐Hispanic Black61.9 (59.3–64.6)48.9 (46.4–51.3)30.1 (27.7–32.6)12.7 (10.9–14.5)4.7 (3.5–5.9)2.4 (1.5–3.2)0.5 0.1–0.9)Non‐Hispanic White68.0 (65.2–70.9)56.7 (53.9–59.5)37.2 (34.7–39.7)18.2 (16.2–20.2)6.6 (5.2–8.0)3.2 (2.4–4.1)0.6 (0.3–0.9)Non‐Hispanic Asian75.3 (71.5–79.1)64.4 (60.2–68.6)45.0 (41.2–48.9)21.8 (19.2–24.5)7.3 (5.6–9.0)2.7 (1.8–3.7)0.5 (0–1.0)Other/multi‐racial59.7 (50.3–69.2)50.6 (41.2–60.0)31.5 (23.4–39.5)15.2 (9.6–20.7)7.2 (2.9–11.6)3.7 (0.4–7.0)1.1 (0–2.9)WomenOverall66.4 (64.0–68.8)56.0 (53.7–58.4)38.4 (36.2–40.5)20.2 (18.6–21.9)8.7 (7.6–9.7)4.5 (3.8–5.2)1.0 (0.7–1.3)Mexican American69.8 (66.5–73.1)59.2 (55.3–63.0)37.7 (34.1–41.2)20.2 (17.1–23.3)8.4 (6.3–10.6)4.4 (2.6–6.2)0.8 (0.3–1.4)Other Hispanic67.0 (62.5–71.6)56.4 (51.6–61.2)38.0 (31.6–41.4)19.8 (16.2–23.4)8.4 (5.8–11.0)4.8 (3.1–6.5)1.3 (0.4–2.1)Non‐Hispanic Black59.7 (56.2–63.1)53.9 (50.6–57.3)40.0 (37.2–42.9)22.1 (19.8–24.3)8.6 (7.4–9.8)4.4 (3.5–5.3)1.0 (0.5–1.5)Non‐Hispanic White67.7 (64.7–70.8)56.7 (54.1–62.5)39.1 (36.1–41.9)20.4 (18.1–22.7)9.1 (7.7–10.5)4.6 (3.6–5.7)1.0 (0.6–1.5)Non‐Hispanic Asian61.0 (57.6–64.5)48.2 (54.1–64.4)29.7 (26.4–33.1)16.1 (13.0–19.1)7.3 (5.6–8.9)4.3 (3.1–5.4)0.9 (0.3–1.5)Other/multi‐racial67.9 (58.7–77.1)57.0 (48.3–65.8)37.1 (30.3–43.9)18.7 (12.4–25.1)5.8 (2.1–9.6)2.8 (0.3–5.3)— a a Sample size too small to calculate the weighted prevalence and 95% CI. The estimated prevalence and 95% confidence intervals of total body fat percentage percentiles by sex and race and ethnicity. Sample size too small to calculate the weighted prevalence and 95% CI. Logistic regression analysis indicated that non‐Hispanic Black men had significantly lower odds (OR 0.67; 95% CI: 0.48–0.94) of having TBFP above the 90th percentile compared with non‐Hispanic White men after adjusting for age. No statistically significant differences were found for other racial and ethnic groups compared with non‐Hispanic White men or women (Table 2). TABLE 2Odds of total body fat percentage ≥ 90th versus < 90th percentile by sex, race and ethnicity.SexRace/ethnicityModel 1 a Model 2 b OR (95% CI) p‐valueOR (95% CI) p‐valueMenMexican American0.73 (0.46–1.14)0.1650.73 (0.46–1.15)0.166Other Hispanic0.90 (0.58–1.41)0.6530.91 (0.59–1.41)0.664Non‐Hispanic Black0.67 (0.48–0.94)0.0210.67 (0.48–0.94)0.022Non‐Hispanic White1 (ref)—1 (ref)—Non‐Hispanic Asian1.07 (0.76–1.49)0.7051.07 (0.77–1.49)0.692Other/multi‐racial1.09 (0.57–2.09)0.7891.10 (0.57–2.12)0.775WomenMexican American0.93 (0.68–1.28)0.6560.93 (0.68–1.27)0.639Other Hispanic0.93 (0.66–1.31)0.6790.94 (0.66–1.33)0.705Non‐Hispanic Black0.95 (0.78–1.16)0.6000.95 (0.78–1.16)0.600Non‐Hispanic White1 (ref)—1 (ref)—Non‐Hispanic Asian0.79 (0.59–1.07)0.1210.79 (0.58–1.07)0.126Other/multi‐racial0.63 (0.31–1.29)0.2030.63 (0.31–1.30)0.205 a Adjusted for age and race. b Adjusted for age, race, and survey cycle. Odds of total body fat percentage ≥ 90th versus < 90th percentile by sex, race and ethnicity. Adjusted for age and race. Adjusted for age, race, and survey cycle. TBFP demonstrated excellent predictive accuracy for identifying obesity defined by BMI ≥ 30 kg/m2, with area under the curve (AUC) values ranging from 0.827 to 0.904 for men and 0.856 to 0.914 for women across race and ethnic groups (Figure 1). Notably, non‐Hispanic Black men (AUC 0.904) and non‐Hispanic White women (AUC 0.914) demonstrated the highest predictive accuracy for obesity. Accuracy of total body fat percentage to predict obesity (BMI ≥ 30 vs. < 30 kg/m2) by race/ethnicity among (a) men and (b) women. Panel. Conversely, when assessing BMI's ability to predict obesity defined by commonly used TBFP thresholds (≥ 25% in men and ≥ 30% in women), BMI also demonstrated strong predictive accuracy (AUC values ranging from 0.817 to 0.893 in men and from 0.804 to 0.901 in women) (Figure 2). Non‐Hispanic Black men (AUC 0.893) and non‐Hispanic Black women (AUC 0.901) exhibited particularly strong predictive associations. Accuracy of body mass index to predict obesity (TBFP ≥ 25% vs. < 25% for men; ≥ 30% vs. < 30% for women) by race/ethnicity among (a) men and (b) women. However, TBFP showed limited predictive capacity for cardiometabolic disease risk biomarkers. For glycated hemoglobin (HbA1c ≥ 5.7%) in individuals without known diabetes, the AUC ranged from 0.524 to 0.648, indicating limited discriminative ability across racial and ethnic groups (Figure 3). Similarly, TBFP had modest accuracy for predicting elevated alanine aminotransferase (ALT), with AUC values ranging from 0.553 to 0.628 for ALT ≥ 20 U/L and from 0.511 to 0.595 for ALT ≥ 40 U/L. (Figure 4) Accuracy of total body fat percentage to predict HbA1c ≥ 5.7% versus < 5.7% in individuals without type 2 diabetes. Accuracy of total body fat percentage to predict (a) ALT ≥ 20 versus < 20 U/L and (b) ALT ≥ 40 versus < 40 U/L. To contextualize the performance of TBFP, we additionally examined BMI in relation to the same cardiometabolic risk factors. BMI demonstrated comparable or slightly stronger discrimination for these outcomes than TBFP across most racial/ethnic groups. Consistent with prior literature, BMI showed high accuracy in identifying individuals with elevated adiposity defined by TBFP thresholds (Figure 2), whereas TBFP showed more limited discrimination for HbA1c and ALT outcomes (Figures 3 and 4). Discussion In this nationally representative analysis from NHANES (2011–2018), we observed variability in TBFP percentiles by sex, race, and ethnicity, and found that TBFP demonstrated strong predictive accuracy for identifying BMI‐defined obesity (≥ 30 kg/m2) across all groups, while BMI similarly showed excellent accuracy in predicting elevated TBFP (≥ 25% in men, ≥ 30% in women). These findings reinforce the well‐established concordance between BMI‐defined obesity and excess adiposity, as previously reported in large epidemiologic studies, while extending prior work by providing nationally representative, race/ethnicity‐ and sex‐specific TBFP percentiles and evaluating their relationship with cardiometabolic biomarkers within a unified analytic framework. Although TBFP represents a more direct measure of adiposity, it did not consistently outperform BMI in predicting cardiometabolic risk markers in this study, with only modest discrimination observed for HbA1c and ALT outcomes. Consistent with prior literature, BMI remained a robust and practical predictor of cardiometabolic risk at the population level, likely reflecting its strong correlation with overall body mass and its established clinical thresholds. The modest predictive performance observed for TBFP may, in part, reflect methodological considerations, including the use of individual biomarkers such as HbA1c and ALT, each with recognized limitations, which may not fully capture the complexity of cardiometabolic dysfunction. As discussed, HbA1c was used as a marker of glycemic status but does not encompass the full spectrum of dysglycemia, including impaired fasting glucose and impaired glucose tolerance [37], while ALT was used as a biomarker of liver injury and does not provide a definitive diagnosis of metabolic dysfunction–associated steatotic liver disease (MASLD), which requires more comprehensive clinical assessment [47, 48]. In addition, TBFP reflects total adiposity but does not account for fat distribution, particularly visceral and ectopic fat depots, which are more strongly associated with cardiometabolic risk and adverse outcomes [24, 49, 50, 51]. As a result, the limited discriminatory ability observed in this study likely reflects a combination of biomarker limitations, the multifactorial nature of cardiometabolic disease, and the inability of global adiposity measures to capture biologically relevant fat distribution, rather than the absence of a relationship between adiposity and risk. These findings are consistent with prior studies demonstrating that more direct measures of adiposity do not consistently improve prediction of cardiometabolic outcomes beyond BMI in population‐based analyses [15, 16]. The recent Lancet Diabetes & Endocrinology Commission has emphasized the need to move beyond BMI alone in defining obesity by incorporating measures of adiposity and metabolic health [5] While our findings are discussed in the context of these recommendations, it is important to note that this study did not operationalize the full diagnostic framework proposed by the Commission, which incorporates multiple domains including hyperglycemia, glucose intolerance, and dyslipidemia. Rather, HbA1c was used as a marker of glycemic status, and future studies should incorporate composite measures of metabolic dysfunction to better align with evolving clinical frameworks for defining obesity‐related health risk. In this context, our findings support the continued use of BMI as a practical screening tool while highlighting the potential complementary role of TBFP in refining risk assessment, particularly across diverse populations where differences in body composition are not fully captured by BMI alone [38, 39]. Although BMI is an indirect measure of adiposity, it has been extensively validated across populations and remains strongly associated with cardiometabolic outcomes, whereas TBFP, despite its physiologic specificity, may be influenced by measurement variability and does not fully capture fat distribution, which is a key determinant of metabolic risk. Taken together, these findings suggest that TBFP should be considered complementary to BMI rather than a replacement, particularly in efforts to improve obesity classification and risk stratification. Total body fat percentage cutoffs of ≥ 25% in men and ≥ 35% in women have historically been used to define excess adiposity and were originally proposed to align with BMI thresholds (≥ 30 kg/m2) based on population‐level correlations between BMI and body fat percentage derived from DXA data across diverse groups [52, 53]. More recent work has suggested higher thresholds associated with metabolic risk (≥ 30% in men and ≥ 42% in women) [25], highlighting the influence of age, sex, and race/ethnicity on TBFP and the need for population‐specific interpretation. Although the development of TBFP percentiles is inherently descriptive, these findings provide important clinical and public health contexts by offering nationally representative reference values that allow interpretation of individual adiposity relative to population distributions, particularly in diverse groups with known differences in body composition. However, deriving clinically meaningful TBFP thresholds linked to cardiometabolic outcomes remains challenging, as the relationship between TBFP and metabolic risk is continuous and heterogeneous across populations. Accordingly, our findings should be interpreted as descriptive and comparative rather than prescriptive, and future research should focus on establishing standardized TBFP thresholds that incorporate population heterogeneity and are linked to clinically relevant endpoints. In addition, our results highlight the importance of considering alternative and complementary measures of adiposity that capture fat distribution, including waist circumference and waist‐to‐hip ratio, as well as imaging‐derived measures of visceral adiposity, which have demonstrated stronger associations with cardiometabolic risk and mortality than overall adiposity alone [54, 55, 56, 57]. Practical, low‐cost anthropometric measures that reflect central adiposity should be further developed and validated across diverse populations to support more precise and equitable obesity‐related risk assessment in clinical practice. Study Limitations and Strengths; Future Directions This study has several limitations that should be considered. First, the cross‐sectional design of NHANES precludes causal inference, and our analyses are intended to be predictive rather than etiologic. Second, we used a complete‐case analysis approach, excluding participants with missing data on key variables (TBFP, BMI, and cardiometabolic biomarkers). As a result, selection bias may be present if missingness was not completely at random. In addition, although NHANES employs rigorous, standardized data collection methods, biomarker data are obtained from participants who complete the examination component, which may introduce additional selection considerations. Third, the use of individual biomarkers such as HbA1c and ALT, while clinically relevant and widely available, does not fully capture the multidimensional nature of cardiometabolic disease. Finally, TBFP reflects total adiposity but does not account for fat distribution, particularly visceral adiposity, which may be more strongly associated with cardiometabolic risk. Despite these limitations, this study has several important strengths. We leveraged nationally representative NHANES data across multiple survey cycles, allowing for robust estimation of TBFP percentiles across diverse sex and racial/ethnic groups. The use of standardized DXA‐based body composition measures enhances the accuracy of adiposity assessment. Additionally, the large sample size and ability to directly compare TBFP and BMI within the same population provide a comprehensive evaluation of their relative and complementary roles in cardiometabolic risk assessment. Future research should focus on developing and validating race‐ and ethnicity‐specific thresholds for total body fat percentage (TBFP) to enhance the diagnostic accuracy and clinical relevance of obesity assessments. While TBFP is useful for estimating overall adiposity, we acknowledge its limited ability to assess cardiometabolic risk in isolation. To address this gap, future studies should investigate the relationship between easily performed, low‐cost assessments of regional adipose distribution, such as waist circumference and waist‐to‐hip ratio, and cardiometabolic dysfunction, events, and other outcomes. These investigations should account for variation across age, sex, and racial and ethnic groups to inform more personalized and equitable risk assessments. Broader inclusion of cardiometabolic markers, including lipid profiles, inflammatory biomarkers, and cardiovascular outcomes, is also warranted to expand the clinical utility of both BMI and adiposity measures. Ultimately, integrating feasible, cost‐effective tools into routine clinical workflows is essential to support timely, patient‐centered obesity care across diverse populations. Conclusion In conclusion, this study provides nationally representative TBFP percentiles for U.S. adults aged 20–59 years. While percentiles were largely consistent across groups, non‐Hispanic Black men had significantly lower odds of TBFP ≥ 90th percentile. Notably, TBFP percentiles had limited ability to predict cardiometabolic risk markers, which highlights the need for additional, accessible measures of fat distribution and demographically tailored cutoffs to improve clinical risk assessment. Funding This work was funded by the National Institutes of Health, National Institute on Minority Health and Health Disparities (Grants R01MD011686, R01MD011686‐05S2 and R01MD011686‐05S3). Conflicts of Interest Jaime P. Almandoz reported receiving advisory/consulting fees and/or other support from Boehringer Ingelheim, Eli Lilly and Company, Nestlé, Novo Nordisk A/S, and Wave Life Sciences. The other authors declared no conflicts of interest. Drs. Messiah and Xie and Mr. Mathew received funding from NIH/NIMHD and Dr. Messiah received additional funding from NICHD. Dr. Rivara received funding from NIDDK. Supporting information Table S1 Acknowledgments S.E.M. conceptualized the study, guided all statistical analyses, and contributed to writing the first draft of the paper. L.X. and Z.Y. performed the statistical analyses and generated the tables and figures. S.M. procured all the data. G.H., J.C., A.C.R., S.K., D.E., and J.A. assisted with writing the manuscript. J.A. provided medical expertise in interpretation of the findings. All authors approved the final manuscript for publication. Herrington G., Xie L., Yuan Z., et al., “Association of Total Body Fat Percentiles, Obesity and Cardiometabolic Health Biomarkers in Adults in the United States,” Obesity Science & Practice (2026): e70158, 10.1002/osp4.70158. Grant Herrington, and Luyu Xie are co‐first authors. References Associated Data Supplementary Materials Table S1