The association between body roundness index and lumbar bone mineral density in U.S. adults: Result of a nationwide survey

Abstract

A growing body of evidence has established a close association between obesity and bone metabolism. The body roundness index (BRI), a novel anthropometric parameter, has demonstrated superior predictive capability for estimating both visceral fat percentage and total body fat percentage when compared with traditional measures such as Body Mass Index (BMI) and waist circumference. However, there is a paucity of research exploring the relationship between BRI and bone mineral density (BMD). Therefore, this investigation was designed to assess the association between BRI and lumbar spine BMD among adults in the United States. Data were obtained from the National Health and Nutrition Examination Survey (NHANES) cycles spanning 2011 to 2018, including participants aged 20 years and older. To evaluate the linear association between BRI and lumbar BMD, weighted multivariable regression models were applied. Additionally, weighted smooth curve fitting along with a 2-piecewise linear regression approach was utilized to detect potential nonlinear relationships. Stratified subgroup analyses were carried out based on age, sex, race/ethnicity, and BMI categories. The final analysis included a total of 10,996 adult participants. In the fully adjusted regression model, BRI exhibited a significant inverse association with lumbar BMD (β = −0.048; 95% CI: −0.059 to −0.037). This negative correlation persisted across most subgroups (all interaction P-value > .05), with the exception of BMI-defined strata. A nonlinear association between BRI and lumbar BMD was identified, with an inflection point observed at a BRI value of 7.63. Comparable nonlinear patterns were detected in subpopulations stratified by age (<40 years and ≥ 40 years), gender (male), and race/ethnicity (non-Hispanic White and other race). This study highlights a significant inverse and nonlinear relationship between the BRI and lumbar spine BMD in U.S. adults. These findings may offer novel insights for the development of targeted strategies in osteoporosis prevention and intervention.

1. Introduction

Osteoporosis (OP), a systemic skeletal disorder, is defined by the deterioration of bone microarchitecture and a decline in bone mineral density (BMD).^[1,2] Projections suggest that by 2030, more than 70 million individuals in the United States will be affected by bone loss or OP.^[3] The economic burden associated with OP is anticipated to rise sharply, from $57 billion in 2018 to over $95 billion annually by 2040 in the U.S.^[4] According to Johnell et al, osteoporosis accounts for more than 8.9 million fractures globally each year.^[5] Given these statistics, OP has emerged as a significant public health concern, underscoring the urgent need for the development of an objective, efficient, and user-friendly approach for early detection and prevention.

Obesity is defined by an abnormal or excessive accumulation of body fat that adversely impacts health and is strongly associated with various chronic conditions.^[6] Its global prevalence has seen a marked rise, reaching an estimated morbidity rate of approximately 30%.^[7,8] A substantial body of evidence supports a significant correlation between obesity and bone health.^[9,10] Utilizing anthropometric indicators to explore this association is of vital importance. Although body mass index (BMI) and waist circumference (WC) are commonly used as conventional measures of obesity, these indicators are limited in that they fail to distinguish between fat mass and lean muscle mass.^[11–13] Thomas et al proposed the body roundness index (BRI) as a novel obesity measure in 2013 by taking WC and height into account.^[14] Compared to BMI and WC, the BRI provides superior predictions of %visceral adipose tissue and %body fat. The BRI is highly associated with various diseases, including metabolic syndrome, diabetes, and cardiovascular diseases.^[15–20] Moreover, the BRI demonstrates better predictive ability than both BMI and WC in diabetes, colorectal cancer, nonalcoholic fatty liver disease, and left ventricular hypertrophy.^[21–25]

To the best of our knowledge, no prior research has examined the association between the BRI and BMD. NHANES is a large, nationally representative dataset with DXA measurements, which acts as a valuable resource for assessing this issue. Therefore, the objective of the present study was to investigate the correlation between BRI and lumbar spine BMD in U.S. adults aged 20 years and older, utilizing data from the 2011 to 2018 cycles of the NHANES.

2. Methods

2.1. Data source and study population

The NHANES, administered by the National Center for Health Statistics (NCHS), represents the largest cross-sectional, population-based survey globally. It evaluates the nutritional status and general health of the non-institutionalized U.S. population through a complex, multistage, stratified sampling design. NHANES data are collected and disseminated in biennial cycles.

For this study, data were drawn from 4 NHANES cycles spanning 2011 to 2018. Among the initial 39,156 participants, exclusions were made for individuals lacking lumbar BMD measurements (n = 20,384), those missing WC and height data (n = 174), participants diagnosed with cancer (n = 414), and individuals under 20 years of age (n = 7188). After these exclusions, a total of 10,996 participants were included in the final analysis (Fig. 1). The NHANES protocol received approval from the NCHS Ethics Review Board, and informed consent was obtained from all participants.

Figure 1.

Flowchart of participants selection. BMD = bone mineral density, BRI = body roundness index, NHANES = National Health and Nutrition Examination Survey.

2.2. BRI evaluation

BRI acted as the independent variable in this study, and the formula below was used to determine BRI^[14]:

$${\displaystyle \begin{array}{l}{\displaystyle BRI=364.2-365.5\times \sqrt{1-{\left[\frac{\frac{WC\left(m\right)}{2\pi }}{0.5\times Height\left(m\right)}\right]}^2}}\end{array}}$$

All anthropometric measurements were conducted by trained personnel within the mobile examination centers. WC was measured using a flexible tape positioned at the intersection of 2 anatomical landmarks: 1 drawn vertically along the right midaxillary line and the other placed horizontally just above the superior lateral margin of the right iliac crest. Height was measured with participants standing upright in a standardized posture.

2.3. Outcome variable

Lumbar BMD, serving as the dependent variable in this study, was measured using dual-energy X-ray absorptiometry (DXA) with Hologic Discovery model A densitometers (Hologic, Inc., Bedford), operated in conjunction with Apex version 3.2 software. All scans were performed by certified radiologic technologists to ensure accuracy and consistency.

2.4. Covariates

Covariates included: demographics data [age, sex, race, education, ratio of family income to poverty (PIR)], examination data (BMI), questionnaire data (smoking status, alcohol use, moderate and vigorous activities, diabetes, hypertension), laboratory data [serum glucose, glycated hemoglobin (HbA1c), alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), total protein, albumin, serum creatinine, serum uric acid, blood urea nitrogen (BUN), phosphorus, total calcium, triglyceride, total cholesterol, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C)], dietary data (protein intake, calcium intake).

BMI was calculated by dividing weight in kilograms by the square of height in meters. Participants who had smoked a minimum of 100 cigarettes over their lifetime were classified as smokers. Alcohol use status included 5 categories: never (<12 drinks in lifetime), former (≥12 drinks in lifetime or any past year and without alcohol intake last year), mild (females having ≥ 1 drinks/d, males having ≥ 2 drinks/d), moderate (females having ≥ 2 drinks/d, males having ≥ 3 drinks/d, or binge drinking ≥ 2 d/mo), heavy (females having ≥ 3 drinks/d, males having ≥ 4 drinks/d, or binge drinking ≥ 5 d/mo).^[26] Diabetes and hypertension status were determined based on participants’ self-reported history of a physician-diagnosed condition. Additional details of all covariates are accessible at www.cdc.gov/nchs/nhanes/.

2.5. Statistical analysis

All statistical analyses were performed using R software (version 4.3) and EmpowerStats (version 4.1). To ensure national representativeness, appropriate sampling weights were applied in accordance with the analytical guidelines provided by the NCHS.^[27,28] Continuous variables were presented as mean ± standard error (SE), while categorical variables were reported as counts and corresponding percentages [n (%)]. The weighted chi-square test and 1-way analysis of variance were utilized to compare population characteristics across quartiles of the BRI. The association between BRI and lumbar BMD was assessed using weighted multivariable linear regression models. For trend analysis, BRI was categorized into quartiles. Three models were developed: Model 1 included no covariate adjustments; Model 2 was adjusted for age, sex, and race/ethnicity; and Model 3 incorporated all covariates except for height and WC. Further analyses included stratified subgroup evaluations and interaction tests within the fully adjusted model. To explore potential nonlinear relationships between BRI and lumbar BMD, weighted smooth curve fitting and a 2-piecewise linear regression approach were applied. A P-value < .05 was considered to indicate statistical significance.

3. Results

3.1. Characteristics of study participants

The characteristics of participants are presented according to the BRI quartiles (Table 1). In general, this study enrolled 10,996 subjects, with a mean (SE) age of 39.01 (0.23) years and 52.87% of male participants. The mean (SE) values of BRI and lumbar BMD among all individuals were 5.16 (0.05) and 1.04 (0.00) g/cm², respectively. Compared with subjects in the bottom BRI quartile, those in the top quartile were more likely to be older, female, non-Hispanic Black, and Mexican American, demonstrated higher values of BMI, WC, smoking rate, serum glucose, HbA1c, ALT, AST, ALP, uric acid, triglyceride, total cholesterol, and LDL-C, and showed greater prevalences of diabetes and hypertension. In contrast, they exhibited lower education level, PIR, height, alcohol use level, physical activity, total protein, albumin, creatinine, BUN, phosphorus, total calcium, HDL-C, protein intake, calcium intake, and lumbar BMD.

Table 1.

Weighted characteristics of the study population based on body roundness index quartile.

Variables	Total N = 10,996	Body roundness index				P-value
		Q1 (1.049–3.541) N = 2749	Q2 (3.542–4.784) N = 2750	Q3 (4.785–6.360) N = 2749	Q4 (6.361–19.101) N = 2748
Age (yr)	39.011 (0.229)	33.625 (0.388)	39.554 (0.372)	41.512 (0.304)	41.549 (0.302)	<.0001
Sex, n (%)
Male	5696 (52.865)	1558 (53.598)	1585 (58.531)	1512 (56.974)	1041 (41.682)	<.0001
Female	5300 (47.135)	1191 (46.402)	1165 (41.469)	1237 (43.026)	1707 (58.318)
Race/ethnicity, n (%)
Non-Hispanic White	3681 (60.116)	998 (63.719)	913 (61.561)	854 (57.105)	916 (57.845)	<.0001
Non-Hispanic Black	2489 (12.197)	652 (13.032)	521 (9.810)	567 (10.930)	749 (15.211)
Mexican American	1604 (10.408)	168 (4.989)	340 (8.967)	550 (14.022)	546 (13.979)
Other race	3222 (17.279)	931 (18.259)	976 (19.662)	778 (17.942)	537 (12.965)
Education, n (%)
Less than high school	2042 (13.472)	368 (9.763)	476 (12.707)	602 (15.525)	596 (16.114)	<.0001
High school	2417 (21.989)	542 (18.882)	579 (20.136)	624 (23.991)	672 (25.221)
More than high school	6535 (64.528)	1838 (71.340)	1695 (67.157)	1522 (60.455)	1480 (58.665)
Unknown	2 (0.011)	1 (0.014)	0 (0.000)	1 (0.030)	0 (0.000)
PIR	2.937 (0.050)	2.941 (0.075)	3.151 (0.070)	2.949 (0.057)	2.687 (0.058)	<.0001
Height (cm)	169.635 (0.166)	171.190 (0.195)	170.444 (0.278)	169.633 (0.304)	167.109 (0.263)	<.0001
BMI (kg/m2)	28.962 (0.142)	22.112 (0.060)	26.224 (0.068)	30.169 (0.065)	37.954 (0.157)	<.0001
Waist circumference (cm)	98.140 (0.355)	79.964 (0.168)	92.173 (0.159)	102.271 (0.185)	119.636 (0.334)	<.0001
Smoking status, n (%)
Yes	4267 (40.574)	1021 (37.235)	1040 (39.562)	1091 (41.883)	1115 (43.862)	.011
No	6724 (59.400)	1726 (62.724)	1710 (60.438)	1656 (58.075)	1632 (56.120)
Unknow	5 (0.025)	2 (0.042)	0 (0.000)	2 (0.041)	1 (0.018)
Alcohol use, n (%)
Never	1338 (9.234)	284 (7.716)	329 (8.585)	349 (9.848)	376 (10.914)	<.0001
Former	892 (7.367)	168 (5.234)	196 (5.835)	227 (8.198)	301 (10.432)
Mild	3312 (31.499)	915 (33.527)	907 (35.283)	793 (30.222)	697 (26.566)
Moderate	1863 (18.815)	482 (20.041)	466 (19.588)	447 (17.148)	468 (18.402)
Heavy	2619 (25.516)	676 (26.731)	596 (22.992)	677 (26.364)	670 (26.083)
Unknow	972 (7.569)	224 (6.750)	256 (7.717)	256 (8.219)	236 (7.603)
Moderate activities, n (%)
Yes	7182 (69.896)	1877 (74.019)	1801 (71.949)	1803 (69.140)	1701 (64.071)	<.0001
No	3814 (30.104)	872 (25.981)	949 (28.051)	946 (30.860)	1047 (35.929)
Vigorous activities, n (%)
Yes	5264 (51.219)	1590 (62.069)	1408 (56.651)	1264 (48.512)	1002 (36.611)	<.0001
No	5732 (48.781)	1159 (37.931)	1342 (43.349)	1485 (51.488)	1746 (63.389)
Serum glucose (mmol/L)	5.394 (0.026)	4.919 (0.028)	5.186 (0.028)	5.506 (0.046)	6.005 (0.057)	<.0001
HbA1c (%)	5.519 (0.014)	5.255 (0.014)	5.383 (0.015)	5.588 (0.024)	5.873 (0.025)	<.0001
ALT (u/L)	26.304 (0.245)	20.640 (0.429)	25.656 (0.491)	28.582 (0.444)	30.585 (0.623)	<.0001
AST (u/L)	25.245 (0.195)	24.163 (0.396)	25.206 (0.437)	25.438 (0.431)	26.221 (0.500)	.043
ALP (u/L)	67.110 (0.424)	60.461 (0.518)	65.226 (0.599)	68.948 (0.533)	74.248 (0.776)	<.0001
Total protein (g/L)	71.444 (0.110)	71.712 (0.180)	71.433 (0.153)	71.480 (0.118)	71.137 (0.117)	.007
Albumin (g/L)	43.195 (0.074)	44.402 (0.112)	43.697 (0.087)	43.088 (0.081)	41.486 (0.106)	<.0001
Creatinine (umol/L)	75.898 (0.320)	76.604 (0.509)	77.102 (0.577)	77.145 (0.616)	72.523 (0.520)	<.0001
Uric acid (umol/L)	319.455 (1.292)	291.277 (1.867)	314.090 (2.224)	330.859 (2.344)	343.030 (1.991)	<.0001
BUN (mmol/L)	4.579 (0.028)	4.468 (0.048)	4.686 (0.045)	4.690 (0.046)	4.463 (0.037)	<.0001
Phosphorus (mmol/L)	1.198 (0.003)	1.231 (0.004)	1.193 (0.005)	1.185 (0.005)	1.184 (0.005)	<.0001
Total calcium (mmol/L)	2.343 (0.002)	2.357 (0.003)	2.347 (0.003)	2.340 (0.003)	2.326 (0.003)	<.0001
Triglyceride (mmol/L)	1.360 (0.027)	0.900 (0.021)	1.345 (0.062)	1.620 (0.052)	1.589 (0.057)	<.0001
Total cholesterol (mmol/L)	4.941 (0.018)	4.593 (0.026)	5.004 (0.037)	5.156 (0.028)	5.010 (0.030)	<.0001
Direct HDL-C (mmol/L)	1.360 (0.008)	1.560 (0.013)	1.397 (0.012)	1.266 (0.009)	1.209 (0.009)	<.0001
LDL-C (mmol/L)	2.951 (0.016)	2.627 (0.030)	2.990 (0.034)	3.151 (0.031)	3.048 (0.031)	<.0001
Diabetes, n (%)
Yes	783 (5.546)	47 (1.422)	123 (3.328)	209 (5.597)	404 (12.278)	<.0001
No	10,008 (92.746)	2687 (98.119)	2587 (95.016)	2484 (92.355)	2250 (84.983)
Borderline	198 (1.663)	15 (0.459)	38 (1.630)	54 (2.011)	91 (2.617)
Unknown	7 (0.045)	0 (0.000)	2 (0.026)	2 (0.037)	3 (0.122)
Hypertension, n (%)
Yes	2560 (22.048)	254 (8.465)	522 (18.708)	725 (24.922)	1059 (37.117)	<.0001
No	8426 (77.892)	2493 (91.500)	2227 (81.278)	2020 (74.946)	1686 (62.821)
Unknown	10 (0.060)	2 (0.034)	1 (0.014)	4 (0.132)	3 (0.063)
Protein intake (g)	87.619 (0.620)	89.087 (1.262)	88.336 (1.063)	88.176 (1.146)	84.731 (0.968)	.01
Calcium intake (mg)	1013.434 (9.416)	1055.829 (16.613)	1010.301 (14.212)	1013.194 (17.109)	972.557 (16.275)	.001
Lumbar BMD (g/cm2)	1.040 (0.002)	1.064 (0.004)	1.044 (0.004)	1.029 (0.004)	1.021 (0.004)	<.0001

Mean (SE) for continuous variables: the P-value was calculated by the weighted 1-way analysis of variance. n (%) for categorical variables: the P-value was calculated by the weighted chi-square test.

ALP = alkaline phosphatase, ALT = alanine aminotransferase, AST = aspartate aminotransferase, BMD = bone mineral density, BMI = body mass index, BUN = blood urea nitrogen, HbA1c = glycated hemoglobin, HDL-C = high-density lipoprotein cholesterol, LDL-C = low-density lipoprotein cholesterol, PIR = ratio of family income to poverty, Q = quartile.

3.2. Linear association between BRI and lumbar BMD

Table 2 shows the correlation between BRI and lumbar BMD in 3 linear regression models. When no confounding factor was adjusted, BRI was negatively linked with lumbar BMD in model 1 (β = −0.005, 95% CI: −0.006, −0.003). After partially and fully adjusting for covariates, this negative association still persisted in model 2 (β = −0.004, 95% CI: −0.005, −0.002) and model 3 (β = −0.048, 95% CI: −0.059, −0.037). Compared with the lowest BRI quartile, the lumbar BMD in the highest quartile dropped by 0.143 g/cm² (β = −0.143, 95% CI: −0.184, −0.102 and P for trend < .0001) after transforming BRI from a continuous variable to a classified variable (quartile).

Table 2.

Association between body roundness index and lumbar bone mineral density among adults.

Exposure	Model 1		Model 2		Model 3
	β (95% CI)	P-value	β (95% CI)	P-value	β (95% CI)	P-value
BRI (continuous)	−0.005 (−0.006, −0.003)	<.0001	−0.004 (−0.005, −0.002)	<.0001	−0.048 (−0.059, −0.037)	<.0001
BRI (quartile)
Q1 (1.049–3.541)	Reference		Reference		Reference
Q2 (3.542–4.784)	−0.020 (−0.031, −0.009)	<.001	−0.009 (−0.020, 0.002)	.099	−0.025 (−0.042, −0.008)	.007
Q3 (4.785–6.360)	−0.035 (−0.046, −0.023)	<.0001	−0.022 (−0.034, −0.010)	<.001	−0.076 (−0.100, −0.052)	<.0001
Q4 (6.361–19.101)	−0.043 (−0.054, −0.032)	<.0001	−0.036 (−0.047, −0.024)	<.0001	−0.143 (−0.184, −0.102)	<.0001
P for trend	<.0001		<.0001		<.0001

Model 1: no covariate was adjusted; Model 2: age, gender, and race were adjusted; Model 3: age, gender, race, PIR, education, BMI, smoking status, alcohol use, moderate activities, vigorous activities, serum glucose, HbA1c, ALT, AST, ALP, total protein, albumin, creatinine, uric acid, BUN, phosphorus, total calcium, total cholesterol, triglyceride, HDL-C, LDL-C, hypertension, diabetes, protein intake, and calcium intake were adjusted.

ALP = alkaline phosphatase, ALT = alanine aminotransferase, AST = aspartate aminotransferase, BMI = body mass index, BRI = body roundness index, BUN = blood urea nitrogen, HbA1c = glycated hemoglobin, HDL-C = high-density lipoprotein cholesterol, LDL-C = low-density lipoprotein cholesterol, PIR = ratio of family income to poverty, Q = quartile.

Stratified analysis was performed to assess the stability of the correlation between BRI and lumbar BMD across several subgroups, including age, gender, race, and BMI (Table 3). The negative association remained consistent in distinct age levels, genders, and races (all P for interaction > .05). When stratified by BMI levels, the relationship between BRI and lumbar BMD was more prominent in participants with BMI 25 to 30 kg/m² (β = −0.053, 95% CI: −0.068, −0.037) than those with BMI < 25 kg/m² (β = −0.006, 95% CI: −0.022, 0.010) and BMI ≥ 30 kg/m² (β = −0.002, 95% CI: −0.008, 0.003) (P for interaction < .0001).

Table 3.

Subgroup analysis of the association between body roundness index and lumbar bone mineral density.

Subgroup	β (95% CI)	P-value	P for interaction
Age (yr)
<40	−0.049 (−0.061, −0.037)	<.0001	.858
≥40	−0.048 (−0.062, −0.033)	<.0001
Sex
Male	−0.072 (−0.086, −0.058)	<.0001	.971
Female	−0.033 (−0.045, −0.021)	<.0001
Race/ethnicity
Non-Hispanic White	−0.050 (−0.065, −0.034)	<.0001	.153
Non-Hispanic Black	−0.056 (−0.073, −0.040)	<.0001
Mexican American	−0.031 (−0.043, −0.019)	<.0001
Other race	−0.040 (−0.055, −0.025)	<.0001
BMI (kg/m2)
<25	−0.006 (−0.022, 0.010)	.454	<.0001
≥25, <30	−0.053 (−0.068, −0.037)	<.0001
≥30	−0.002 (−0.008, 0.003)	.377

Age, gender, race, PIR, education, BMI, smoking status, alcohol use, moderate activities, vigorous activities, serum glucose, HbA1c, ALT, AST, ALP, total protein, albumin, creatinine, uric acid, BUN, phosphorus, total calcium, total cholesterol, triglyceride, HDL-C, LDL-C, hypertension, diabetes, protein intake, and calcium intake were adjusted, but the model was not adjusted for the stratification variables themselves.

ALP = alkaline phosphatase, ALT = alanine aminotransferase, AST = aspartate aminotransferase, BMI = body mass index, BUN = blood urea nitrogen, HbA1c = glycated hemoglobin, HDL-C = high-density lipoprotein cholesterol, LDL-C = low-density lipoprotein cholesterol, PIR = ratio of family income to poverty.

3.3. Nonlinear association between BRI and lumbar BMD

Smooth curve fitting revealed the nonlinear correlation between BRI and lumbar BMD with the inflection point of 7.63 (Fig. 2; Table 4). For subjects with BRI < 7.63, every 1-unit increment in BRI was related to a 0.061 g/cm² decrease in lumbar BMD (95% CI: −0.068, −0.054); meanwhile, for BRI > 7.63, every 1-unit growth in BRI was linked with a 0.031 g/cm² decrease in lumbar BMD (95% CI: −0.039, −0.024). Furthermore, similar nonlinear relationship also existed in participants aged < 40 years (inflection point: 7.28) and ≥40 years (inflection point: 8.50), males (inflection point: 8.53), non-Hispanic White (inflection point: 8.72), and other race (inflection point: 7.40) (Fig. 3A–C; Table 4). Interestingly, in participants with BMI ≥ 30 kg/m², the correlation between BRI and lumbar BMD was a U-shaped curve, and the inflection point was 7.62 (Fig. 3D; Table 4). It showed a negative association (β = −0.028, 95% CI: −0.037, −0.020) and a positive relationship (β = 0.012, 95% CI: 0.006–0.018) for individuals with BRI level < 7.62 and >7.62, respectively.

Figure 2.

The nonlinear association between body roundness index and lumbar bone mineral density. (A) Each black hollow point exhibits 1 participant. (B) Solid red line illustrates the fitted smooth curve among variables. Two blue bands illustrate the 95% confidence interval of the fit. Age, gender, race, PIR, education, BMI, smoking status, alcohol use, moderate activities, vigorous activities, serum glucose, HbA1c, ALT, AST, ALP, total protein, albumin, creatinine, uric acid, BUN, phosphorus, total calcium, total cholesterol, triglyceride, HDL-C, LDL-C, hypertension, diabetes, protein intake, and calcium intake were adjusted. ALP = alkaline phosphatase, ALT = alanine aminotransferase, AST = aspartate aminotransferase, BMI = body mass index, BUN = blood urea nitrogen, HbA1c = glycated hemoglobin, HDL-C = high-density lipoprotein cholesterol, LDL-C = low-density lipoprotein cholesterol, PIR = ratio of family income to poverty.

Table 4.

Threshold effect analysis of body roundness index on lumbar bone mineral density using the 2-piecewise linear regression model.

Outcome	Inflection point (K)	BRI < K	BRI > K	Log likelihood ratio
		Adjusted β (95% CI) P-value	Adjusted β (95% CI) P-value
Total	7.63	−0.061 (−0.068, −0.054) < .0001	−0.031 (−0.039, −0.024) < .0001	<.001
Subgroup analysis
Age < 40 yr old	7.28	−0.063 (−0.072, −0.054) < .0001	−0.034 (−0.044, −0.025) < .0001	<.001
Age ≥ 40 yr old	8.50	−0.059 (−0.069, −0.050) < .0001	−0.026 (−0.037, −0.014) < .0001	<.001
Male	8.53	−0.084 (−0.094, −0.073) < .0001	−0.033 (−0.048, −0.018) < .0001	<.001
Non-Hispanic White	8.72	−0.060 (−0.071, −0.050) < .0001	−0.025 (−0.039, −0.012) .0003	<.001
Other race	7.40	−0.050 (−0.063, −0.036) < .0001	−0.024 (−0.039, −0.009) .0019	<.001
BMI ≥ 30 kg/m2	7.62	−0.028 (−0.037, −0.020) < .0001	0.012 (0.006–0.018) < .0001	<.001

Age, gender, race, PIR, education, BMI, smoking status, alcohol use, moderate activities, vigorous activities, serum glucose, HbA1c, ALT, AST, ALP, total protein, albumin, creatinine, uric acid, BUN, phosphorus, total calcium, total cholesterol, triglyceride, HDL-C, LDL-C, hypertension, diabetes, protein intake, and calcium intake were adjusted, but the model was not adjusted for the stratification variables themselves.

ALP = alkaline phosphatase, ALT = alanine aminotransferase, AST = aspartate aminotransferase, BMI = body mass index, BRI = body roundness index, BUN = blood urea nitrogen, HbA1c = glycated hemoglobin, HDL-C = high-density lipoprotein cholesterol, LDL-C = low-density lipoprotein cholesterol, PIR = ratio of family income to poverty.

Figure 3.

Subgroup analysis for the association between body roundness index and lumbar bone mineral density stratified by (A) age, (B) gender, (C) race/ethnicity, (D) BMI. Age, gender, race, PIR, education, BMI, smoking status, alcohol use, moderate activities, vigorous activities, serum glucose, HbA1c, ALT, AST, ALP, total protein, albumin, creatinine, uric acid, BUN, phosphorus, total calcium, total cholesterol, triglyceride, HDL-C, LDL-C, hypertension, diabetes, protein intake, and calcium intake were adjusted, except the stratification variables. ALP = alkaline phosphatase, ALT = alanine aminotransferase, AST = aspartate aminotransferase, BMI = body mass index, BUN = blood urea nitrogen, HbA1c = glycated hemoglobin, HDL-C = high-density lipoprotein cholesterol, LDL-C = low-density lipoprotein cholesterol, PIR = ratio of family income to poverty.

4. Discussion

To the best of our knowledge, this study is the first to investigate the association between BRI and BMD in an adult population. In this large-scale cross-sectional analysis involving 10,996 US adults, a negative relationship between BRI and lumbar BMD was identified. This inverse association remained consistent across stratified analyses, with the exception of the BMI subgroup. Additionally, a nonlinear relationship was observed, with an inflection point occurring at a BRI value of 7.63.

Wung et al found that elevated BRI was significantly associated with higher BMD T-scores among 101 hemodialysis patients diagnosed with metabolic syndrome.^[29] Similarly, a lower BRI level was reported to be linked with decreased calcaneal ultrasound T-scores in a Taiwanese population.^[30] These findings differ from our results, which may be attributable to variations in sample size, study population, anatomical sites examined, measurement techniques, and selected bone health indicators.

Smooth curve fitting detected an inflection point of 7.63, which indicated that A decrease in BRI is associated with a reduction in lumbar BMD, and the lumbar BMD decreases more sharply at lower BRI values, whereas the effect is attenuated beyond the threshold of 7.63.

Globally, BMI and WC are widely recognized anthropometric measures for assessing obesity.^[31–34] Previous literature has consistently demonstrated a positive association between both BMI and WC with BMD. A population-based study involving 6143 U.S. adolescents identified a positive correlation between BMI and total BMD, with a saturation point observed at 21.5 kg/m².^[35] In a Mendelian randomization study including 336,107 participants, Song et al utilized BMI-associated single nucleotide polymorphisms and reported a causal relationship between higher BMI and increased BMD in both the heel calcaneus and lumbar spine.^[36] Furthermore, a meta-analysis concluded that obesity, as defined by BMI, was positively associated with BMD at the lumbar spine and femoral neck.^[37] Consistently, Alay et al analyzed data from 452 postmenopausal women in Turkey and observed significant positive associations between BMI, WC, femoral neck BMD, and L1 to L4 lumbar spine BMD.^[38] Likewise, a cross-sectional analysis of 2903 older adults aged 50 years and above from the 2017 to 2020 NHANES cycles revealed positive relationships between BMI, WC, and femoral neck BMD, with a BMI saturation threshold identified at 24.3 kg/m².^[39]

However, when obesity is evaluated using BMI and WC, a phenomenon known as the “obesity paradox” has been reported by numerous studies in recent years.^[40] This paradox refers to a seemingly counterintuitive observation that obesity may confer protective effects and be associated with improved outcomes in certain medical conditions.^[41,42] Some researchers argue that the obesity paradox may be an artifact arising from the limitations of BMI, particularly its inability to differentiate between fat mass and lean muscle mass.^[43–45] As a result, the validity of BMI and WC as accurate indicators of obesity has been questioned by scholars.^[46–49] To further elucidate the relationship between adiposity and bone health, Jiao et al employed DXA to assess body composition and reported an inverse association between total body fat percentage and BMD among 11,615 U.S. adults aged 18 years and older.^[50] Consistent with these findings, our results demonstrate that higher BRI is significantly associated with lower lumbar BMD, and this negative relationship persists across the majority of subgroup analyses. These results differ from prior studies that assessed obesity using BMI and WC, suggesting that BRI may serve as a more reliable and informative metric for evaluating obesity status.

The underlying mechanisms responsible for the adverse association between obesity and bone health remain incompletely understood. Several plausible biological pathways have been proposed. First, adiponectin, which facilitates the differentiation of bone marrow-derived mesenchymal stem cells (BMSCs) into osteoblasts through upregulation of CXCL1 and CXCL8, tends to be reduced in individuals with obesity.^[51,52] Second, leptin – a hormone predominantly secreted by white adipose tissue – exerts complex, dual effects on bone metabolism. On one hand, in vitro studies have demonstrated that leptin promotes stromal cell differentiation into osteoblasts while inhibiting osteoclastogenesis.^[53,54] Additionally, leptin-deficient (knockout) rat models display reduced femoral BMD and diminished bone volume.^[55] On the other hand, leptin negatively interacts with hypothalamic neuron-derived serotonin, thereby impairing bone formation.^[56] Overall, leptin is considered to exert a predominantly negative influence on bone health.^[6,57] Astudillo et al reported elevated leptin levels in individuals with obesity.^[53] Third, obesity has been associated with an increased prevalence of secondary hyperparathyroidism, characterized by elevated parathyroid hormone (PTH) levels, which contributes to decreased BMD.^[58] Fourth, obesity induces an increase in marrow adiposity and modulates adipocyte metabolism. Given that both adipocytes and osteoblasts originate from BMSCs,^[59,60] excess adiposity shifts the differentiation pathway toward adipogenesis at the expense of osteoblastogenesis.^[61] The accumulation of adipocytes within bone marrow alters the bone microenvironment, disrupts osteocyte function, reduces bone turnover, and may ultimately predispose to early onset of OP.^[62] Finally, obesity is closely linked with heightened inflammatory susceptibility.^[63] Within the bone marrow niche, increased adipocyte presence suppresses osteoblast differentiation, limits osteoprotegerin production, promotes osteoclastogenesis, and enhances the release of inflammatory and immunomodulatory cytokines, thereby stimulating osteoclast activity and bone resorption.^[64,65]

Admittedly, this study has several limitations. First, the cross-sectional design precludes any inference of causal relationships between BRI and lumbar BMD in adults. Second, the possibility remains that residual confounding factors were not fully accounted for, which may have influenced the observed associations. Third, the findings may not be generalizable to individuals with cancer, as this subgroup was excluded from the analysis. Despite these limitations, the use of a large, nationally representative sample of the U.S. population strengthens the reliability of our results and enables comprehensive subgroup analyses. Nevertheless, future longitudinal studies are warranted to validate and further elucidate the observed associations.

5. Conclusion

BRI was identified as a negative predictor of bone health among U.S. adults. A nonlinear association between BRI and lumbar BMD was observed, suggesting a complex relationship between adiposity and bone metabolism. These findings underscore the potential importance of maintaining an optimal BRI level as part of strategies aimed at preserving bone health and preventing osteoporosis.

Author contributions

Conceptualization: Ziyi Zhao, Jian Wang.

Methodology: Hongxiang Ji, Caifeng Wu.

Software: Zhengdan Wang.

Validation: Wenyu Liu, Xiaoheng Ding.

Visualization: Shengquan Ren, Chunlei Liu.

Writing – original draft: Xiaoheng Ding.

Writing – review & editing: Xiaoheng Ding.