Differential associations of distinct anthropometric indices with heart failure in adults with MASLD: A cross-sectional study of NHANES 2017 to 2020

Abstract

Metabolic dysfunction-associated steatotic liver disease (MASLD) is closely linked to cardiovascular morbidity, yet it remains unclear which anthropometric indices most effectively capture the burden of prevalent heart failure (HF) in this population. This cross-sectional study utilized data from the National Health and Nutrition Examination Survey conducted from 2017 to March 2020 (pre-pandemic). MASLD was defined according to the 2023 Delphi consensus as hepatic steatosis plus ≥ 1 cardiometabolic risk factor. HF was defined according to self-reported physician diagnosis of congestive HF. Weighted multivariable logistic regression and restricted cubic splines were used to evaluate the associations of body mass index (BMI), waist circumference (WC), 10 × waist-to-height ratio (10 × WHtR), and body roundness index (BRI) with prevalent HF. Among 2753 participants with MASLD, 96 had HF. In the fully adjusted model, BMI remained independently associated with HF (odds ratio, 1.05; 95% confidence interval, 1.01–1.09). Conversely, the associations for central adiposity indices (WC, 10 × WHtR, and BRI) were attenuated and became nonsignificant after adjusting for metabolic comorbidities. WC tertile categories showed higher odds of HF for the second and third tertiles compared with the lowest tertile, although the overall trend was not statistically significant. Subgroup analyses suggested potential heterogeneity by race/ethnicity and diabetes status, with stronger associations observed among non-Hispanic Asian participants and those without diabetes. Restricted cubic splines analyses indicated linear dose-response relationships for all indices. In adults with MASLD, general obesity (BMI) was independently associated with HF, whereas associations for central adiposity indices were no longer statistically significant after adjustment for metabolic comorbidities. Notably, in subgroup analyses, central obesity markers such as 10 × WHtR and BRI showed stronger associations with HF in non-Hispanic Asians and participants without diabetes, suggesting potential heterogeneity in these associations across clinical subgroups.

1. Introduction

Metabolic dysfunction-associated steatotic liver disease (MASLD), recently renamed from nonalcoholic fatty liver disease to better reflect its underlying pathophysiology, has emerged as a major global public health challenge, affecting approximately 38% of the adult population worldwide.^[1–3] This nomenclature shift, formalized by a multi-society Delphi consensus in 2023, emphasizes the coexistence of hepatic steatosis with cardiometabolic risk factors, such as obesity, dysglycemia, and hypertension.^[4] Although MASLD can progress to advanced fibrosis and hepatocellular carcinoma, cardiovascular disease (CVD) rather than liver-related complications remains the leading cause of mortality in this population in most epidemiologic studies.^[5,6]

Among the spectrum of cardiovascular complications, heart failure (HF) is a significant but often underappreciated comorbidity. Emerging evidence supports organ crosstalk along a liver–heart axis, whereby systemic inflammation, insulin resistance, and hemodynamic perturbations may contribute to myocardial remodeling and dysfunction in steatotic liver disease.^[7–10] Several meta-analyses have demonstrated that individuals with steatotic liver disease face a substantially higher risk of developing HF independent of shared risk factors such as diabetes and hypertension.^[11,12] Given the increasing prevalence of both conditions, identifying high-risk individuals within the MASLD population is crucial for early risk stratification and intervention.

Obesity is a pivotal driver of MASLD and HF. However, the anthropometric indices that are most informative for characterizing HF burden in MASLD remain debatable. Body mass index (BMI), the traditional metric for defining obesity, has notable limitations: it fails to distinguish between lean mass and adipose tissue and does not capture body fat distribution.^[13] This is particularly relevant in MASLD, where visceral adiposity, rather than generalized obesity, is the primary driver of lipotoxicity and metabolic inflammation.^[14] Consequently, indices reflecting central obesity, such as waist circumference (WC) and waist-to-height ratio (WHtR), may better capture cardiometabolic risk than BMI.^[15] Furthermore, the body roundness index (BRI), a geometric index derived from height and WC to quantify body shape and improve the prediction of body fat and visceral adiposity, has been linked to incident CVD and mortality in population studies.^[16–18]

Despite these advances, few studies have systematically compared the performance of these diverse anthropometric indices, specifically within the newly defined MASLD population.^[19,20] It remains unclear whether novel markers of central adiposity, such as BRI or WHtR, show stronger associations with prevalent HF than traditional BMI in patients who already exhibit metabolic dysfunction. To address this knowledge gap, we utilized data from the 2017 to 2020 National Health and Nutrition Examination Survey (NHANES) 2017 to 2020. This study aimed to investigate the associations of BMI, WC, WHtR, and BRI with the prevalence of HF among Unites States (U.S.) adults with MASLD and to evaluate potential variations across different demographic and clinical subgroups.

2. Materials and methods

2.1. Study population

The study population was derived from the NHANES 2017 to March 2020 pre-pandemic cycle. Of the 15,560 participants with the required components available, individuals aged < 20 years (N = 6328) and pregnant participants (N = 87) were excluded. Participants without a valid vibration-controlled transient elastography (VCTE) examination or with missing controlled attenuation parameter (CAP) data were excluded (N = 1750). Hepatic steatosis was defined as CAP ≥ 274 dB/m; thus, participants with CAP < 274 dB/m were excluded (N = 4178). Participants with missing anthropometric measurements (WC or height; N = 113) were excluded. To minimize confounding from alternative liver disease etiologies, participants with viral hepatitis (hepatitis B surface antigen positivity or hepatitis C virus RNA positivity; N = 34) and heavy alcohol consumption (men > 30 g/day; women > 20 g/day; N = 189) were removed.^[21,22] Significant alcohol consumption was calculated from frequency and quantity questionnaires, assuming 14 g of ethanol per standard drink. This threshold aligns with the lower boundary distinguishing MASLD from metabolic dysfunction and alcohol-associated liver disease in the consensus nomenclature.

MASLD was defined as hepatic steatosis plus at least 1 cardiometabolic risk factor; those with steatosis but without any cardiometabolic risk factors were excluded (N = 11). Finally, individuals with missing data on covariates and outcome variables (education, alcohol-related variables, hypertension, diabetes, smoking, and HF status; N = 117) were excluded, resulting in a final analytic sample of 2753 participants. The participant selection process is illustrated in Figure 1.

Figure 1.

Participant selection for the MASLD cohort and analytic sample (NHANES 2017–March 2020 pre-pandemic). CAP = controlled attenuation parameter, MASLD = metabolic dysfunction-associated steatotic liver disease, n = number of participants, NHANES = National Health and Nutrition Examination Survey, VCTE = vibration-controlled transient elastography.

The NHANES protocol was approved by the National Center for Health Statistics Research Ethics Review Board, and written informed consent was obtained from all participants. Because the present study used publicly available de-identified NHANES data, no additional institutional review board approval was required.

2.2. Definition of MASLD

MASLD was defined in accordance with the 2023 multi-society Delphi consensus statement, requiring the presence of hepatic steatosis in conjunction with at least 1 cardiometabolic risk factor following the exclusion of other specific liver disease etiologies.^[2] Hepatic steatosis was objectively quantified using VCTE (FibroScan 502 V2 Touch, Echosens) conducted in the NHANES mobile examination center (MEC) by certified technicians following standardized protocols.^[21] We used the median CAP to assess liver fat, defining steatosis as a CAP score of ≥ 274 dB/m. While optimal CAP cutoffs vary across validation studies, this prespecified threshold was selected to harmonize with recent NHANES VCTE-based epidemiologic studies applying the new MASLD nomenclature.^[23–25] To ensure diagnostic accuracy, the analysis was restricted to participants with a complete elastography examination, characterized by a fasting time of ≥ 3 h, ≥10 valid stiffness measures, and an interquartile range/median ratio of < 30% for liver stiffness.^[21]

Upon establishing the presence of steatosis, MASLD was diagnosed if the participant exhibited at least one of the 5 cardiometabolic risk criteria derived from the consensus framework: overweight/obesity, defined as a BMI ≥ 25 kg/m² (or ≥ 23 kg/m² for Asian individuals),^[26] WC > 94 cm for men and > 80 cm for women; dysglycemia, indicated by fasting plasma glucose ≥ 100 mg/dL, hemoglobin A1c ≥ 5.7%, or use of antidiabetic medication; hypertension, defined as mean blood pressure ≥ 130/85 mm Hg or use of antihypertensive medication; hypertriglyceridemia, defined as plasma triglycerides ≥ 150 mg/dL or lipid-lowering treatment; or low HDL cholesterol (< 40 mg/dL for men, < 50 mg/dL for women) or lipid-lowering treatment.^[2] Because NHANES fasting laboratory measures (glucose and triglycerides) were collected only in a subsample, these values were treated as missing for participants not meeting fasting requirements, and metabolic risk was captured via hemoglobin A1c, non-fasting lipid proxies, and medication history to maximize cohort inclusion without compromising validity.^[2,23]

2.3. Exposures and outcomes

Anthropometric assessments, including measured weight, standing height, and WC, were performed by trained health technologists at the NHANES MEC using standardized procedures and calibrated equipment.^[27] BMI was calculated as the weight in kilograms divided by the height in meters squared (kg/m²). WC was measured at a standardized anatomic site, marked just above the uppermost lateral border of the right ilium at the midaxillary line, and recorded to the nearest 0.1 cm at the end of normal expiration. The primary exposures were BMI, 10 × WHtR, and BRI. WHtR was calculated as WC divided by height using consistent units (e.g., cm/cm) and rescaled for regression analyses as 10 × WHtR to express effect estimates per 0.1-unit increase in WHtR for more clinically interpretable scaling. BRI was computed from WC and height (m) using the geometric model proposed by Thomas et al, which conceptualizes body shape as an ellipse to better approximate central/visceral adiposity^[17]:

$${\displaystyle BRI=364.2-365.5\times \sqrt{(1-\frac{{(WC/2\pi )}^2}{{\left(0.5\times Height\right)}^2})}}$$

Higher values across these indices were interpreted to reflect greater general or central adiposity. Participants with missing data required to compute specific anthropometric indices were excluded from the analyses.

The outcome was HF, ascertained from the NHANES Medical Conditions Questionnaire administered during the standardized in-home interview using the computer-assisted personal interview system. Participants were classified as having HF if they answered “Yes” to item MCQ160b (“Has a doctor or other health professional ever told you that you had congestive HF?”) Those answering “No” were classified as not having HF.^[28] Responses of “Refused,” “Don’t know,” or missing were treated as missing outcome data and excluded to preserve the phenotype validity.

2.4. Covariates

To account for potential confounding factors, sociodemographic, lifestyle, and clinical characteristics derived from the standardized NHANES questionnaires were included as covariates. Sociodemographic variables included age (continuous), sex (male or female), and race/ethnicity, which were classified into 6 categories: Mexican American, other Hispanic, non-Hispanic White, non-Hispanic Black, non-Hispanic Asian, and other/multiracial. Educational attainment was stratified into 5 levels: < 9th grade, 9th to 11th grade (including 12th grade with no diploma), high school graduate/general educational development or equivalent, some college or AA degree, and college graduate or above.

Regarding lifestyle and health characteristics, smoking history was ascertained using the question, “Have you smoked at least 100 cigarettes in your entire life?” classifying participants as ever-smokers or never-smokers.^[29] Alcohol consumption history was evaluated based on the question, “Ever had a drink of any kind of alcohol?,” categorizing participants as ever drinkers or lifetime abstainers. Comorbidities were assessed using self-reported physician diagnoses. Hypertension was defined as an affirmative response to having been told by a doctor or other health professional that high blood pressure was present. Similarly, diabetes mellitus was defined by the question “Has a doctor or other health professional ever told you that you had diabetes?”; participants reporting a “borderline” diagnosis were grouped with those answering “no,” distinguishing them from those with a definitive diagnosis.

2.5. Study size and power considerations

The analytic sample size was determined by the availability of eligible participants in the NHANES 2017 to March 2020 pre-pandemic dataset with complete data on anthropometric indices, HF status, covariates, and survey design variables. Because this was a secondary analysis of an existing population-based survey, a prospective a priori sample size calculation was not applicable. Therefore, we performed a survey design-based Monte Carlo minimum detectable effect-size analysis for the 4 primary continuous anthropometric exposures. Using the complete-case sample for the fully adjusted model (n = 2753; HF events = 96), the observed NHANES complex survey design and covariate distribution were preserved. Binary HF outcomes were simulated under a range of candidate odds ratios (OR) per 1-standard deviation increase in each anthropometric index. For each candidate OR, the fully adjusted survey-weighted logistic regression model was refitted in 500 simulated datasets, and empirical power was calculated as the proportion of simulations with a 2-sided P value < .05. The minimum detectable OR was defined as the smallest OR achieving at least 80% empirical power. The estimated minimum detectable ORs were 1.64 for BMI, 1.75 for WC, 1.75 for 10 × WHtR, and 1.63 for BRI, each expressed per 1-standard deviation increase (Table S1). Given the limited number of HF events, particularly within some subgroup strata, subgroup and interaction analyses were considered exploratory.

2.6. Statistical analyses

All statistical analyses were performed using the R software (version 4.3.2; R Foundation for Statistical Computing). To account for the complex multistage probability sampling design of the NHANES, appropriate sample weights (MEC exam weights), clustering, and stratification were applied in the baseline characteristics comparisons and multivariable logistic regression models using the survey package.^[30] Continuous variables were expressed as weighted means with 95% confidence intervals (CIs), whereas categorical variables were presented as weighted proportions with 95% CIs. Baseline characteristics were stratified by HF status, and differences between groups were assessed using a survey-weighted Student t test for continuous variables and the Rao–Scott χ² test for categorical variables.

To investigate the associations between anthropometric indices (BMI, WC, WHtR, and BRI) and the prevalence of HF, weighted multivariable logistic regression models were constructed to estimate the OR and 95% CIs. Anthropometric indices were analyzed as continuous variables and categorized into tertiles, with the lowest tertile serving as the reference group. Tests for linear trends were performed by entering tertile categories as continuous ordinal variables in the models. Three sequential models were developed to control for potential confounders: model 1 was unadjusted; model 2 was adjusted for age, sex, race/ethnicity, and educational attainment; and model 3 was fully adjusted for age, sex, race/ethnicity, educational attainment, smoking status, alcohol consumption, hypertension, and diabetes mellitus.

Furthermore, restricted cubic splines with 3 knots (placed at the 10th, 50th, and 90th percentiles, consistent with Harrell recommended defaults) were utilized using the regression modeling strategies package to visualize the dose-response relationship and test for potential nonlinearity.^[31] Owing to the incompatibility of restricted cubic splines functions with complex survey design objects, these specific analyses were performed using weighted logistic regression models with normalized weights (calculated as the individual weight divided by the mean weight) to preserve the representativeness of point estimates.^[32] nonlinearity was formally evaluated using the analysis of variance method.

Subsequently, subgroup analyses were conducted to assess the consistency of associations across subgroups defined by age, sex, race/ethnicity, and comorbidities using a full complex survey design. Potential effect modification was evaluated by testing the significance of the multiplicative interaction term between exposure and stratification variables. Statistical significance was defined as a 2-sided P value of < .05.

3. Results

3.1. Baseline characteristics stratified by HF status

The final analytic sample comprised 2753 adults with MASLD, including 96 participants identified as having HF and 2657 without HF. Baseline characteristics stratified by HF status are presented in Table 1. Participants with HF were markedly older than those without HF (mean age 63.81 vs 50.71 years; P < .001). Socioeconomic and lifestyle profiles also differed; education level distribution varied significantly between groups (P = .010), and smoking was more prevalent among participants with HF (61.9 vs 41.0%; P = .004). In contrast, no statistically significant differences were observed in sex (P = .472), race/ethnicity (P = .110), or alcohol consumption (P = .877).

Table 1.

Baseline characteristics of the analytic MASLD sample stratified by self-reported HF status.

Characteristic	Overall (N = 2753)	No heart failure (N = 2657)	Heart failure (N = 96)	P value
Age, yrs	51.06 (49.71, 52.41)	50.71 (49.42, 52.01)	63.81 (61.15, 66.48)	< .001
Sex				.472
Male	44.5% (41.5, 47.5)	44.6% (41.6, 47.6)	39.8% (27.1, 52.5)
Female	55.5% (52.5, 58.5)	55.4% (52.4, 58.4)	60.2% (47.5, 72.9)
Race/ethnicity				.110
Mexican American	11.8% (8.6, 14.9)	12.0% (8.8, 15.2)	3.0% (0.2, 5.7)
Other Hispanic	7.2% (5.6, 8.9)	7.3% (5.7, 8.8)	6.0% (0.0, 13.3)
Non-Hispanic White	62.6% (57.4, 67.7)	62.3% (57.3, 67.3)	72.9% (56.6, 89.2)
Non-Hispanic Black	8.6% (6.3, 11.0)	8.5% (6.2, 10.8)	12.7% (4.0, 21.4)
Non-Hispanic Asian	5.2% (3.5, 6.9)	5.3% (3.6, 7.0)	1.7% (0.0, 3.5)
Other Race	4.6% (3.3, 5.9)	4.6% (3.3, 5.9)	3.7% (0.0, 8.6)
Education				.010
< 9th Grade	3.7% (2.8, 4.6)	3.6% (2.7, 4.5)	8.1% (2.2, 13.9)
9–11th Grade	6.6% (5.7, 7.4)	6.5% (5.6, 7.4)	8.3% (3.4, 13.2)
High School	29.8% (26.8, 32.7)	29.6% (26.4, 32.8)	35.8% (21.1, 50.4)
College	31.5% (29.0, 34.0)	31.2% (28.6, 33.8)	42.3% (29.5, 55.1)
College Graduate or above	28.4% (23.8, 33.1)	29.1% (24.3, 33.9)	5.6% (0.0, 11.8)
Smoking				.004
No	58.4% (55.0, 61.8)	59.0% (55.5, 62.4)	38.1% (26.3, 49.8)
Yes	41.6% (38.2, 45.0)	41.0% (37.6, 44.5)	61.9% (50.2, 73.7)
Alcohol				.877
No	7.9% (6.3, 9.6)	7.9% (6.4, 9.4)	8.8% (0.0, 20.6)
Yes	92.1% (90.4, 93.7)	92.1% (90.6, 93.6)	91.2% (79.4, 100.0)
Hypertension				< .001
No	55.7% (51.7, 59.8)	56.9% (52.9, 60.8)	14.0% (3.6, 24.4)
Yes	44.3% (40.2, 48.3)	43.1% (39.2, 47.1)	86.0% (75.6, 96.4)
Diabetes				< .001
No	81.3% (78.9, 83.7)	82.1% (79.8, 84.3)	53.4% (43.4, 63.5)
Yes	18.7% (16.3, 21.1)	17.9% (15.7, 20.2)	46.6% (36.5, 56.6)
BMI, kg/m2	33.65 (33.11, 34.18)	33.59 (33.05, 34.14)	35.53 (33.57, 37.48)	.124
WC, cm	110.98 (109.78, 112.18)	110.8 (109.55, 112.05)	117.64 (114.4, 120.89)	< .001
10 × WHtR	6.59 (6.51, 6.67)	6.58 (6.5, 6.66)	6.98 (6.75, 7.21)	.001
BRI	6.98 (6.79, 7.18)	6.96 (6.76, 7.16)	8.01 (7.34, 8.68)	.001

Data are presented as weighted means (95% confidence intervals) for continuous variables and weighted percentages (95% confidence intervals) for categorical variables. P values were calculated using survey-weighted Student’s t tests for continuous variables and Rao–Scott χ² tests for categorical variables.

BMI = body mass index, BRI = body roundness index, CI = confidence interval, HF = heart failure, MASLD = metabolic dysfunction–associated steatotic liver disease, N = number of participants, WC = waist circumference, WHtR = waist-to-height ratio.

Cardiometabolic comorbidities were substantially more common in the HF group. The prevalence of hypertension and diabetes was notably higher among participants with HF (86.0 vs 43.1%; P < .001) (46.6 vs 17.9%; P < .001). With respect to anthropometric indices, overall adiposity as measured by BMI did not differ significantly between the groups (35.53 vs 33.59 kg/m²; P = .124). However, measures capturing central obesity and body shape were consistently elevated in participants with HF, including WC (117.64 vs 110.80 cm; P < .001), 10 × WHtR (6.98 vs 6.58; P = .001), and BRI (8.01 vs 6.96; P = .001), suggesting a more adverse pattern of fat distribution despite comparable BMI.

3.2. Association between anthropometric indices and HF

Weighted multivariable logistic regression models were applied to examine the associations between general and central adiposity indices and the prevalence of HF among adults with MASLD (Table 2). In the crude model, BMI, WC, 10 × WHtR, and BRI were all positively associated with prevalent HF. Specifically, each 1 kg/m² increase in BMI was associated with a 4% higher odds of prevalent HF (OR, 1.04; 95% CI, 1.01–1.07; P = .029), each 1 cm increase in WC was associated with a 3% higher odds (OR, 1.03; 95% CI, 1.01–1.04; P < .001), each 1-unit increase in 10 × WHtR, corresponding to a 0.1-unit increase in WHtR, was associated with a 57% higher odds (OR, 1.57; 95% CI, 1.24–1.98; P < .001), and each 1-unit increase in BRI was associated with a 17% higher odds (OR, 1.17; 95% CI, 1.08–1.27; P = .001). Similar positive associations were observed in Model I after adjustment for age, sex, race/ethnicity, and education, including BMI (OR, 1.07; 95% CI, 1.03–1.11; P = .004), WC (OR, 1.03; 95% CI, 1.01–1.05; P = .004), 10 × WHtR (OR, 1.70; 95% CI, 1.23–2.35; P = .007), and BRI (OR, 1.21; 95% CI, 1.08–1.35; P = .007).

Table 2.

Survey-weighted associations of anthropometric indices with prevalent self-reported HF among adults with MASLD.

Anthropometric index	Crude Model		Model Ⅰ		Model Ⅱ
	OR (95% CI)	P value	OR (95% CI)	P value	OR (95% CI)	P value
BMI, kg/m2	1.04 (1.01, 1.07)	.029	1.07 (1.03, 1.11)	.004	1.05 (1.01, 1.09)	.038
BMI (tertile)
T1 (n = 922)	Reference		Reference		Reference
T2 (n = 913)	2.49 (1.05, 5.92)	.050	2.92 (1.21, 7.03)	.034	2.38 (1.07, 5.31)	.066
T3 (n = 918)	1.80 (0.96, 3.39)	.080	2.59 (1.30, 5.14)	.019	1.91 (0.98, 3.74)	.096
P for trend	.104		.026		.140
WC, cm	1.03 (1.01, 1.04)	< .001	1.03 (1.01, 1.05)	.004	1.02 (1.00, 1.04)	.074
WC (tertile)
T1 (n = 918)	Reference		Reference		Reference
T2 (n = 921)	4.51 (1.79, 11.41)	.004	4.23 (1.60, 11.22)	.013	3.49 (1.31, 9.33)	.037
T3 (n = 914)	2.26 (1.70, 3.00)	< .001	4.26 (1.97, 9.20)	.003	2.91 (1.25, 6.78)	.039
P for trend	< .001		.010		.109
10 × WHtR	1.57 (1.24, 1.98)	< .001	1.70 (1.23, 2.35)	.007	1.46 (1.00, 2.15)	.085
10 × WHtR (tertile)
T1 (n = 918)	Reference		Reference		Reference
T2 (n = 917)	2.12 (0.85, 5.31)	.122	1.81 (0.75, 4.37)	.209	1.40 (0.58, 3.39)	.472
T3 (n = 918)	2.40 (1.17, 4.94)	.026	2.16 (1.02, 4.57)	.068	1.44 (0.63, 3.28)	.411
P for trend	.025		.081		.469
BRI	1.17 (1.08, 1.27)	.001	1.21 (1.08, 1.35)	.007	1.15 (1.00, 1.32)	.079
BRI (tertile)
T1 (n = 918)	Reference		Reference		Reference
T2 (n = 917)	2.12 (0.85, 5.31)	.122	1.81 (0.75, 4.37)	.209	1.40 (0.58, 3.39)	.472
T3 (n = 918)	2.40 (1.17, 4.94)	.026	2.16 (1.02, 4.57)	.068	1.44 (0.63, 3.28)	.411
P for trend	.025		.081		.435

Data are presented as survey-weighted ORs and 95% CIs from logistic regression models. Continuous analyses were performed per 1 kg/m² increase in BMI, per 1 cm increase in WC, per 1-unit increase in 10 × WHtR, and per 1-unit increase in BRI. One unit increase in 10 × WHtR corresponds to a 0.1-unit increase in WHtR. Tertile analyses used the lowest tertile as the reference group. Sample sizes for tertile groups are presented as unweighted numbers. The crude model was unadjusted. Model I was adjusted for age, sex, race/ethnicity, and education. Model II was further adjusted for smoking status, alcohol consumption, diabetes, and hypertension.

BMI = body mass index, BRI = body roundness index, CI = confidence interval, HF = heart failure, MASLD = metabolic dysfunction–associated steatotic liver disease, n = number of participants, OR = odds ratio, WC = waist circumference, WHtR = waist-to-height ratio.

After further adjustment for smoking status, alcohol consumption, diabetes, and hypertension (Model II), the associations differed by indicator. BMI remained independently associated with HF (OR per 1 kg/m² increase, 1.05; 95% CI, 1.01–1.09; P = .038). In contrast, the continuous associations for central adiposity markers were attenuated and did not reach statistical significance (WC, P = .074; 10 × WHtR, P = .085; BRI, P = .079).

When modeled in tertiles, WC showed the clearest risk differentiation. Compared with the lowest tertile, both the second (OR, 3.49; 95% CI, 1.31–9.33; P = .037) and third tertiles (OR, 2.91; 95% CI, 1.25–6.78; P = .039) were associated with higher odds of HF in Model II, although the overall trend across tertiles was not statistically significant (P for trend = .109). BMI tertile analyses yielded elevated point estimates for higher categories but did not demonstrate a monotonic gradient after full adjustment (P for trend = .140). Tertile-based associations for 10 × WHtR and BRI were not statistically significant in Model II (both P for trend > .40).

Restricted cubic spline analyses (Fig. 2) further characterized the dose-response patterns. The adjusted spline curves were broadly consistent with a linear increase in the odds of HF across the observed ranges of BMI, WC, 10 × WHtR, and BRI, with no evidence supporting nonlinearity (all P for nonlinearity > .15). Notably, uncertainty increased at higher exposure levels for the central obesity indices, as reflected by the widening CIs in the upper tails.

Figure 2.

Dose-response relationships between anthropometric indices and prevalent HF in adults With MASLD. RCS analyses were performed to visualize the associations of (A) BMI, (B)WC, (C) 10 × WHtR, and (D) BRI with the odds of HF. BMI = body mass index, BRI = body roundness index, CI = confidence interval, HF = heart failure, MASLD = metabolic dysfunction-associated steatotic liver disease, RCS = restricted cubic spline, WC = waist circumference, WHtR = waist-to-height ratio.

3.3. Subgroup analyses

Subgroup analyses were conducted to assess the robustness of the associations between anthropometric indices and HF across major demographic and clinical strata, including age, sex, race/ethnicity, education, smoking status, alcohol consumption, hypertension, and diabetes (Figs. 3–6).

Figure 3.

Subgroup analysis for the association between BMI and HF. Data are adjusted for age, sex, race/ethnicity, education level, smoking status, alcohol consumption, diabetes, and hypertension. BMI = body mass index, CI = confidence interval, HF = heart failure, N = number of participants, OR = odds ratio.

Figure 6.

Subgroup analysis for the association between BRI and HF. Data are adjusted for age, sex, race/ethnicity, education level, smoking status, alcohol consumption, diabetes, and hypertension. BRI = body roundness index, CI = confidence interval, HF = heart failure, N = number of participants, OR = odds ratio.

Figure 4.

Subgroup analysis for the association between WC and HF. Data are adjusted for age, sex, race/ethnicity, education level, smoking status, alcohol consumption, diabetes, and hypertension. CI = confidence interval, HF = heart failure, N = number of participants, OR = odds ratio, WC = waist circumference.

Figure 5.

Subgroup analysis for the association between 10 × WHtR and HF. Data are adjusted for age, sex, race/ethnicity, education level, smoking status, alcohol consumption, diabetes, and hypertension. CI = confidence interval, HF = heart failure, N = number of participants, OR = odds ratio, WHtR = waist-to-height ratio.

Overall, the association between BMI and HF was broadly consistent across most subgroups, with no evidence of effect modification by age, sex, education, smoking, alcohol intake, or hypertension (all P for interaction > .05). In contrast, significant heterogeneity was observed for race/ethnicity (P for interaction = .009) and diabetes status (P for interaction = .004). Stronger associations were evident among non-Hispanic Asian (OR, 1.29; 95% CI, 1.12–1.48) and other/multiracial participants (OR, 1.26; 95% CI, 1.11–1.42), whereas estimates were comparatively attenuated in other racial/ethnic groups. The BMI–HF association was present among participants without diabetes (OR, 1.06; 95% CI, 1.02–1.11) but was not observed among those with diabetes.

For WC, associations were generally stable across strata; however, effect modification was detected for race/ethnicity (P for interaction = .019) and education level (P for interaction = .042). The association was stronger among non-Hispanic Asian participants and among individuals with < 9th grade education. Diabetes status also modified the WC–HF association (P for interaction = .024), with a significant association observed among participants without diabetes (OR, 1.03; 95% CI, 1.02–1.05) but not among those with diabetes.

The subgroup findings for 10 × WHtR and BRI were broadly concordant with those for BMI and WC. Both indices demonstrated significant effect modification by race/ethnicity (both P for interaction = .002) and diabetes status (both P for interaction ≤ .006). Associations were most pronounced among non-Hispanic Asian participants (10 × WHtR OR, 10.08; BRI OR, 2.23) and were largely confined to participants without diabetes. No significant interactions were observed for age, sex, hypertension, or smoking for either index, supporting the overall consistency of the associations across these strata.

4. Discussion

To our knowledge, this is the first study to systematically evaluate and compare the associations of traditional (BMI, WC) and novel (WHtR, BRI) anthropometric indices with HF prevalence, specifically within a nationally representative cohort of U.S. adults with MASLD. Our findings indicate that while all adiposity indices were associated with HF in unadjusted models, their independent associations diverged after accounting for established cardiometabolic risk factors. Notably, BMI remained independently associated with prevalent HF even after full adjustment, corresponding to a 5% higher odds per 1 kg/m² increment in BMI. In contrast, the fully adjusted estimates for central adiposity markers were attenuated and no longer statistically significant, although their point estimates still suggested positive associations, corresponding to a 2% higher odds per 1 cm increase in WC, a 46% higher odds per 1-unit increase in 10 × WHtR (equivalent to a 0.1-unit increase in WHtR), and a 15% higher odds per 1-unit increase in BRI. Furthermore, we identified significant effect modification by race/ethnicity and diabetes status, suggesting subgroup heterogeneity with stronger associations among non-Hispanic Asians and participants without diabetes.

Obesity is an established risk factor for incident HF in the general population, with risk increasing in a graded dose-response manner with greater adiposity; however, among patients with established HF, higher BMI has been associated with better survival: an “obesity paradox.”^[33] In the context of MASLD, the “liver–heart (cardiohepatic) axis” adds a layer of complexity. Our observation that participants with HF had a significantly higher prevalence of metabolic comorbidities aligns with established evidence identifying MASLD as a systemic condition associated with proinflammatory cytokines, oxidative stress, and endothelial dysfunction, which may contribute to adverse cardiac remodeling and cardiovascular risk.^[34] Interestingly, in our fully adjusted models, BMI was significantly associated with prevalent HF, whereas indicators of central adiposity (WC, 10 × WHtR, and BRI) were attenuated and were no longer statistically significant as continuous variables after full adjustment. One possible explanation for this discrepancy is the close interrelationship between central adiposity and cardiometabolic comorbidities. Central and visceral adiposity are strongly associated with insulin resistance, type 2 diabetes, and hypertension.^[35] Therefore, the attenuation of the associations for central adiposity indices after adjustment for diabetes and hypertension may reflect shared metabolic burden, statistical collinearity, or adjustment for variables closely related to both adiposity and HF, rather than evidence of a formally tested mediating pathway. Given the cross-sectional design, we did not perform formal mediation analysis, and the observed attenuation should not be interpreted as evidence of causal mediation. In contrast, BMI, as a measure of total body mass, captures not only adipose tissue but also the hemodynamic burden of excess weight, which may affect myocardial structure beyond metabolic signaling.^[36]

A critical finding of our study was the substantial heterogeneity observed across racial/ethnic subgroups. We found that the associations between anthropometric indices, particularly 10 × WHtR and BRI, and HF were the most pronounced among non-Hispanic Asians. This aligns with the “lean MASLD” phenotype described in Asian populations, where metabolic dysregulation and visceral fat accumulation occur at lower BMI thresholds than in Western populations.^[37] Our data suggest that for Asian patients with MASLD, general obesity metrics, such as BMI, may underestimate cardiovascular risk. The exceedingly high ORs for 10 × WHtR in this subgroup underscore the clinical utility of measuring central adiposity to capture “occult” obesity. This supports the growing consensus that ethnicity-specific anthropometric cutoffs are essential for accurate risk stratification in precision medicine.^[38]

Furthermore, the interaction between diabetes status and adiposity indices warrants further attention. We observed that obesity indices were significantly associated with HF in participants without diabetes, whereas the associations were attenuated and not statistically significant in those with established diabetes. This phenomenon may represent a “saturation effect,” where the metabolic burden associated with diabetes is substantial, and the incremental association of additional adiposity may be less apparent.^[39,40] Alternatively, this pattern could reflect heterogeneity in HF pathophysiology, where diabetes-related myocardial dysfunction may involve pathways distinct from those involved in obesity-related remodeling. For clinicians, this implies that anthropometric monitoring is paramount in MASLD patients before the onset of diabetes to prevent the transition to overt CVD.

BRI is a relatively novel geometric metric derived from height and WC to quantify body shape and approximate visceral adiposity. While recent studies have proposed BRI as an informative indicator of all-cause mortality,^[16] our study suggests that in the specific context of the MASLD-HF relationship, BRI did not show a stronger association than traditional metrics, such as WC, in the fully adjusted model for the overall population. This may be due to the substantial overlap between MASLD diagnosis itself and visceral adiposity; since the MASLD definition already selects for patients with metabolic dysfunction, the discriminatory power of visceral fat markers might be reduced compared to a general population cohort. Nevertheless, the dose-response curves (Fig. 2) exhibited a consistent linear trend, suggesting no apparent threshold within the observed range of adiposity in this population.

Our study has several strengths, including the use of rigorous NHANES data, the application of the new 2023 consensus MASLD definition, and the objective quantification of liver steatosis via VCTE, which avoids reliance on liver enzyme surrogates. However, several limitations should be acknowledged. First, the cross-sectional design precludes causal inference; we cannot determine whether adiposity preceded HF or whether HF-related fluid retention inflated anthropometric measures, although body shape indices (10 × WHtR, BRI) may be less sensitive to short-term fluid shifts than weight alone. Second, the diagnosis of HF was based on self-reported physician diagnosis rather than echocardiographic or adjudicated clinical assessment, which may have introduced outcome misclassification. Third, several key comorbidities, including hypertension and diabetes, were also based on self-reported physician diagnosis, which may have resulted in covariate misclassification. Fourth, although we adjusted for major sociodemographic, lifestyle, and clinical factors, residual confounding due to unmeasured or incompletely measured variables may remain.

5. Conclusions

In conclusion, among U.S. adults with MASLD, higher BMI was independently associated with higher odds of prevalent self-reported HF. In contrast, the associations of central adiposity indices, including WC, 10 × WHtR, and BRI, were attenuated and no longer statistically significant after adjustment for key cardiometabolic comorbidities. Although subgroup analyses suggested potential heterogeneity by race/ethnicity and diabetes status, particularly for central adiposity measures, these findings should be interpreted as exploratory and require confirmation in prospective studies.

Acknowledgments

The authors thank all NHANES participants and staff for their valuable efforts and contributions.

Author contributions

Conceptualization: Yun Liu, Bing Chen.

Data curation: Yun Liu, Xuemei Liu.

Formal analysis: Yun Liu, Xuemei Liu.

Investigation: Xuemei Liu.

Methodology: Yun Liu, Xuemei Liu, Bing Chen.

Project administration: Bing Chen.

Software: Yun Liu.

Supervision: Bing Chen.

Visualization: Yun Liu, Xuemei Liu.

Writing – original draft: Yun Liu, Xuemei Liu.

Writing – review & editing: Yun Liu, Xuemei Liu, Bing Chen.