Bone material strength index: differences by race/ethnicity in Caribbean Hispanic, non-Hispanic White, and non-Hispanic Black older US adults

Abstract

Racial differences in fracture risk are not fully explained by differences in areal BMD (aBMD) as measured by DXA. Other aspects of bone quality, such as bone material properties, are not captured by DXA, but may contribute to fracture risk. Few studies have investigated racial/ethnic differences in bone material properties, and none have included Hispanic Americans. We assessed differences in bone material strength index (BMSi) measured by impact micro indentation by race/ethnicity. Caribbean Hispanic (CH), non-Hispanic Black (NHB), and non-Hispanic White (NHW) participants from the Washington Heights Inwood Community Aging Project (N = 148), a population-based study of aging in New York City, were included. While there were racial differences in aBMD, BMSi did not differ by race/ethnicity (CH vs NHB vs NHW: 79.3 ± 6.3 vs 78.3 ± 15 vs 77.7 ± 5.2) before (p = .22) or after (p = .47) adjustment for age, sex, BMI, fracture status, diabetes, alcohol consumption, smoking, and osteoporosis treatment. There was no significant interaction between race and diabetes status in relation to BMSi. Stratifying by sex, results were similar. In summary, BMSi was similar among CH, NHB, and NHW participants, suggesting that differences in bone material strength, as captured by BMSi, do not account for established racial/ethnic differences in fracture risk in these groups.

Graphical Abstract

Graphical Abstract.

Introduction

There are nearly 60 million Hispanic Americans in the United States, and this population is rapidly growing.¹ Despite their size, Hispanic men and women remain underrepresented in studies of skeletal health. In the United States, racial and ethnic groups show differences in areal BMD (aBMD) but fracture risk does not uniformly parallel aBMD differences. On average, non-Hispanic Black (NHB) individuals have higher aBMD and lower fracture risk than non-Hispanic White (NHW) individuals. In contrast, East Asians have lower aBMD yet also experience a lower risk of hip fracture. Data for Hispanic individuals are more variable, with studies reporting higher, lower, or similar aBMD and fracture risk compared with NHW.²

Similarly, people with type-2 diabetes (T2DM) have similar or higher aBMD than those without T2DM yet they have 50%-80% increased risk of fracture.^3–5 These discrepancies suggest that factors beyond aBMD contribute significantly to fracture risk. Other aspects of bone quality, including bone size and geometry, microarchitecture, and material properties, may play important roles.⁶ Studies using HR-pQCT have provided some insight into skeletal microstructural factors associated with fracture that may also influence risk of fracture across race/ethnicity.^7–9 More recently, impact microindentation (IMI) has allowed for in vivo assessment of bone characteristics thought to reflect material properties. Several cross-sectional studies have shown bone material strength index (BMSi) is associated with fracture.^10–12 Most BMSi data have been acquired in White cohorts. To our knowledge, only 2 prior studies have compared racial/ethnic differences in BMSi. These were both convenience samples potentially subject to selection bias. In a study by Duarte Sosa, 42 Norwegian women had significantly lower BMSi compared to 46 Spanish women.¹³ In another study Dawson-Hughes et al. found (39 NHB and 149 NHW) a significant race-by-glycemia interaction that revealed NHB participants with diabetes had lower BMSi than their NHW counterparts.¹⁴ Black participants without diabetes or with prediabetes didn’t differ from White participants. No studies in the US have assessed Hispanic Americans.

Whether BMSi differs in Hispanic Americans compared to NHW and NHB and may contribute to racial/ethnic differences in fracture is unknown. In this population-based study, we aimed to assess differences in BMSi by race/ethnicity, in a skeletally well-characterized cohort of Caribbean Hispanic (CH), NHB, and NHW individuals with DXA and HR-pQCT data. The IMI analysis was prespecified in the study design. We also assessed the interaction between race and diabetes status given prior study findings.¹⁴ We hypothesized BMSi would vary across racial/ethnic groups with NHB individuals having the highest BMSi in congruence with the lower risk of fracture compared to NHW men and women reported in other studies.

Materials and methods

Design

This was a cross-sectional analysis of BMSi by race/ethnicity in a multi-ethnic cohort of elderly men and women who were participating in a population-based longitudinal cohort study of aging, the Washington Heights Inwood Community Aging Project (WHICAP). The Columbia University Irving Medical Center Institutional Review Board approved this study. All participants provided written informed consent.

Study population

Washington Heights Inwood Community Aging Project is a NIH-funded population-based prospective study of older, city-dwelling ambulatory residents (age ≥ 65) living in New York City. The design and recruitment are published.^15,16 Briefly, random sampling to recruit medicare beneficiaries age 65 and older from 3 zip codes in northern Manhattan was used. Returning WHICAP participants were invited for a one-time assessment with DXA and HR-pQCT (N = 950). Participants with bone imaging enrolling in or following up after the Osteoprobe became available at our center were screened for the microindentation substudy. To ensure participant safety, eligibility was limited to participants <90 yr of age and excluded those with lower leg conditions (eg, local skin disease, edema, and infection), bleeding risk, or local anesthetic allergy. Among 313 participants screened, 150 completed the procedure; 101 were excluded based on eligibility criteria, one was unable to complete the procedure due to excess fat tissue, and 61 declined. All Hispanic participants self-identified as CH descent, and 2 Asian participants were excluded from race/ethnicity–stratified analyses due to small sample size, resulting in an N = 148. Compared with the overall WHICAP cohort, participants who underwent IMI were slightly younger by design (73.9 ± 5.2 vs 76.5 ± 6.2 yr, p < .00001) compared to those who did not, but there were no differences in sex (female 568/71.0% vs 107/71.3%, p = 1.0), fracture history (343/43.3% vs 70/46.7%, p = .47), current osteoporosis treatment (66/8.3% vs 11/7.3%, p = .56), or diabetes (N, 235/29.4% vs N, 40/26.7%; p = .56).

Dual-energy X-ray absorptiometry

BMD was measured using a Hologic Discovery instrument (Hologic Inc.) at the LS (L1-L4), FN, TH, and the 1/3-radius (1/3-R). T-scores were determined using the manufacturer’s NHW reference norms. All skeletal sites were scanned unless hardware or other artifacts were present such metal implants, rods, plates or internal fixation devices that precluded measurement at the region of interest. In vivo precision is 1.28% at LS, 1.36% at TH, and 0.70% at 1/3-R.¹⁷

HR-pQCT/μFEA

Standard HR-pQCT scans of the non-dominant forearm and tibia were performed using an XtremeCT II scanner (Scanco Medical; 68 kVp, 900 μA, 43 s integration, 60.7 μm voxel size). Scans were acquired and analyzed by a single operator; those with motion scores >3 (scale 1-5) were excluded (7 radius scans).¹⁸ Imaging was conducted at 4% of ulnar and 7% of tibial length. Given the study focus, selected HR-pQCT measures previously reported in the larger WHICAP cohort are presented, including vBMD (total, trabecular, and cortical), trabecular microstructure (Tb.N, Tb.Th, and Tb.Sp), and cortical thickness (Ct.Th); in vivo short-term reproducibility (CV) is 0%-5%.^19,20 Whole bone stiffness was estimated via micro-finite element analysis (μFEA) using voxel-based conversion, applying 1% uniaxial strain with a homogeneous Young’s modulus of 10 GPa and Poisson’s ratio of 0.3 (Scanco’s FE software v1.13).

Impact microindentation

Bone material strength index was assessed by IMI at the midshaft of the tibia as previously described¹⁹ using the handheld microindenter (OsteoProbe, Active Life Scientific Inc.) SN 261 and SN 236 devices which automatically exclude invalid measurements. The calibration and performance data between these 2 devices are the same. A single experienced operator conducted all the measurements. To determine the indentation site, we measured the midpoint from the medial border of the tibial plateau to the distal edge of the medial malleolus on the anterior surface of the tibia. A skin disinfectant was applied to the skin and a local anesthetic was applied to the skin and the periosteum before inserting the Osteoprobe through the skin, perpendicular to the anterior tibial site, until reaching the surface of the bone. Ten valid indentations were performed at the measurement site, each 2 mm apart, followed by 10 indentations, each 2 mm apart, in a reference polymethylmethacrylate calibration block. Bone material strength index was computed by the software. Bone material strength index healthy reference intervals are 71.0-97.9 for men and 59.8-95.2 for women.²¹

Biochemistries

Glycosylated hemoglobin (HbA1c) was evaluated by boronate affinity chromatography with the Primus CLC 385 (Primus).

Questionnaire/clinical evaluation

Research personnel used questionnaires to collect demographic information as well as past medical history, medications, calcium/vitamin D intake, physical activity, alcohol consumption, smoking, and historical fractures) by interview.^22,23 Fractures were defined as clinical fractures of any bone other than digits, skull or face. Participants’ weight was measured with a balance beam scale and their height was measured using a wall-mounted stadiometer. A participant was categorized as having diabetes if they had been diagnosed by a physician or if they had an HbA1c ≥6.5%. No participants that underwent IMI were taking glucocorticoids.

Statistics

All statistical analyses were conducted using R (version 4.4.1, R Foundation for Statistical Computing). Sex, race/ethnicity, and fracture status was self-reported. Descriptive statistics were expressed as either means with SDs or absolute (n) and relative (%) frequencies. The primary analysis used one-way ANOVA and Fisher’s exact test to assess for between-group differences in continuous and categorical variables as appropriate. Post-hoc analysis utilized Tukey’s honest significant difference and pairwise-comparisons for significant differences in continuous and categorical variables, respectively. ANCOVA was used to adjust for between-group differences in covariates and those known to affect bone quality. The model included age, BMI, sex, history of fractures, diabetic status, alcohol consumption, smoking, and osteoporosis treatment. Race-diabetes interaction was assessed using a multiple linear regression model of BMSi that included the covariates and interaction term. The outcome variable was modeled as a continuous variable under the assumption of normally distributed residuals with constant variance. Standard diagnostic procedures (residual plots, normality of residuals, Cook’s distance, and leverage statistics) were performed to assess model assumptions. Visual inspection of the model’s diagnostic plots including residual vs fitted, Q-Q, and scale-location indicated that the assumptions of linearity, homoscedasticity, normality of residuals, and independence were adequately met. A 2-tailed p-value of <.05 was considered statistically significant. Model estimates were used to inform a detectable-difference-based power analysis, using the residual SD to calculate the minimum detectable difference between groups (2-sided α = .05; 80% power).

Results

As shown in Table 1, CH individuals had significantly higher BMI and were more likely to be current smokers compared to NHW individuals. Non-Hispanic White were approximately twice as likely to have a prior history of fracture compared to CH and NHB and NHB had a significantly higher alcohol consumption than CH. As shown in Figure 1, BMSi did not differ by race/ethnicity before (p = .22) or after adjustment for covariates (79.5 ± 7.5 vs 78.4 ± 6.9 vs 77 ± 6.8, respectively, p = .47).

Table 1.

Differences in anthropometric, lifestyle, and clinical factors by race/ethnicity.

	CH (n = 96)	NHB (n = 31)	NHW (n = 21)	p-value
Age (yr)	73.4 ± 6.1	74.8 ± 5.1	75 ± 4.8	.33
Percent female (n [%])	72 [75]	22 [71]	11 [52.3]	.18
Height (inches)	62.2 ± 3.4	65.1 ± 3.9	64.6 ± 3.5	.00010 a,b
Weight (pounds)	165.3 ± 29.2	171.9 ± 34.1	157.4 ± 32.5	.25
BMI (kg/m 2 )	30.1 ± 5.2	28.6 ± 5.2	26.4 ± 4.3	.0095 b
Calcium intake (mg/day)	1144.6 ± 576	1138.4 ± 600.6	1295.6 ± 658.6	.55
Vitamin D intake (IU/day)	2801/2 ± 7462.8	1284.1 ± 1387.5	1710 ± 1248	.43
Diabetic (n [%])	31 [32.3]	7 [22.6]	2 [9.5]	.084
History of prior fracture (%)	40 [41.7]	13 [41.9]	16 [76.2]	.016 b
Current osteoporosis tx (%)	10 [10.4]	0	1 [4.8]	.15
Past osteoporosis tx (%)	17 [17.7]	9.7	4 [19.1]	.61
Physical activity score	101.7 ± 57.9	97.1 ± 42.7	98.1 ± 46.3	.90
Current smokers (%)	4 [4.2]	1 [3.2]	0	.051 b
Alcohol intake (drinks/wk)	1 ± 3.3	4.2 ± 10.1	3.8 ± 6.4	.011 a

Data are presented as mean ± SD or percentage (%). Crude p-values reflect comparisons across the 3 groups using one-way ANOVA or Fisher’s exact tests, as appropriate. Post-hoc tests utilized Tukey’s honest significant difference and pairwise comparisons. Statistical significance was defined as p < .05. Superscript letters indicate significant pairwise differences between groups as indicated on the table. Data are presented as mean ± SD or %, p < .05 was considered significant.

^a p < .05 for CH vs NHB.

^b p < .05 for CH vs NHW.

^c p < .05 for NHB vs NHW.

Abbreviations: CH, Caribbean Hispanic; NHB, non-Hispanic Black; NHW, non-Hispanic White.

Figure 1.

Adjusted means of BMSi by race/ethnicity group. Bars represent the mean values (±SD) derived from an ANCOVA model adjusted for age, sex, BMI, diabetes status, fracture history, alcohol consumption, smoking, and osteoporosis treatment. Race categories include Caribbean Hispanic (CH), non-Hispanic Black (NHB), and non-Hispanic White (NHW).

Stratifying by sex, there were no differences in BMSi before (males p = .11; females p-value = .31) or after adjustment for covariates (males: 80.3 ± 10.3 vs 82.7 ± 17 vs 76.6 ± 16.2, p = .13; females: 79.3 ± 7.9 vs 76.9 ± 14.8 vs 78 ± 20.8 p = .30, for CH vs NHB vs NHW, respectively). Omitting, individuals on osteoporosis treatment, BMSi did not differ by race/ethnicity before (p = .64) or after adjusting for covariates (79.3 ± 8.8 vs 78.5 ± 14.3 vs 77.2 ± 17.7, p = .64 for CH vs NHB vs NHW, respectively). There was no significant interaction between race and diabetes status on BMSi (β = −2.40, p = .62). Bone material strength index did not differ by fracture status after covariate adjustment (80.8 ± 9.3 vs 80.1 ± 11.8; p = .37), nor across racial groups among participants with prior fracture (adjusted means: 80.8 ± 10, 82.3 ± 9.1, and 78.3 ± 9.3 for CH, NHB, and NHW, respectively; p = .30).

As expected, there were racial/ethnic differences in aBMD by DXA. Before adjustment, areal BMD was lower at the 1/3 radius in NHW compared to CH and NHB (p = .014; Table 2). After adjustment for covariates, aBMD at the spine approached significance (p = .084) and differences in 1/3 radius BMD remained significant. There were no between-group differences in aBMD at the FN and TH before or after adjusting for covariates.

Table 2.

Differences in DXA, HR-pQCT, and IMI by race/ethnicity.

	CH (n = 96)	NHB (n = 31)	NHW (n = 21)	p-value	Adjusted p-value
LS T-score	−0.6 ± 1.6	0.2 ± 3.8	−0.5 ± 1.6	.094	.084
FN T-score	−1.2 ± 0.9	−0.9 ± 2	−1.5 ± 1	.20	.14
TH T-score a	−0.8 ± 0.9	−0.4 ± 2	−1 ± 1.2	.24	.16
1/3 R T-score a	−1.3 ± 1.2	−0.4 ± 3	−1.5 ± 1.6	.014 b , c	.0086 b , c
Radius_Tot.vBMD a	255.9 ± 60.6	258.0 ± 52.6	239.8 ± 66.8	.5	.42
Radius_Tb.vBMD e	137.6 ± 37.5	136.8 ± 39.5	146.8 ± 42.8	.59	.55
Radius_Tb.N e	1.3 ± 0.2	1.3 ± 0.2	1.4 ± 0.2	.03 b	.021 b
Radius_Tb.Th e	0.23 ± 0.02	0.23 ± 0.01	0.22 ± 0.02	.075	.057
Radius_Tb.Sp e	0.77 ± 0.19	0.77 ± 0.17	0.68 ± 0.13	.12	.096
Radius_Ct.vBMD e	817.7 ± 75.5	827.3 ± 61.2	767.5 ± 84.1	.011 b,c	.0039 c
Radius_Ct.Th e	0.85 ± 0.19	0.90 ± 0.22	0.76 ± 0.23	.042  c	.0087 c
Tibia_Tot.vBMD e	254.5 ± 54.6	262.0 ± 46.3	263.4 ± 51.6	.67	.60
Tibia_Tb.vBMD e	149.9 ± 36.1	147.8 ± 38.8	178.7 ± 37.4	.0041 b,c	.003 b,c
Tibia_Tb.N e	1.2 ± 0.2	1.2 ± 0.2	1.4 ± 0.2	.0000033 b,c	.0000033 b,c
Tibia_Tb.Th e	0.26 ± 0.03	0.25 ± 0.02	0.25 ± 0.02	.12	.13
Tibia_Tb.Sp e	0.87 ± 0.22	0.82 ± 0.18	0.67 ± 0.09	.00051 b,c	.00041 b,c
Tibia_Ct.vBMD e	821.1 ± 80.6	811.8 ± 95.7	762.0 ± 75.3	.015 b	.0026 b,c
Tibia_Ct.Th e	1.27 ± 0.28	1.45 ± 0.37	1.24 ± 0.32	.018 a,c	.0029 a,c
BMSi	79.3 ± 6.3	78.3 ± 6.4	77.7 ± 5.2	.22	.47
Radius stiffness (kN/mm) e	53.5 ± 17.5	66.7 ± 35.1	51.9 ± 24.1	.18	.048 d
Tibia stiffness (kN/mm) e	165.9 ± 40.6	178.5 ± 96.7	179.1 ± 59.6	.18	.14

Differences in as BMD by DXA, microstructure by HR-pQCT, and stiffness. Data are presented as mean ± SD or percentage (%). Crude p-values reflect comparisons across the 2 groups using unpaired-samples t-tests or Fisher’s exact tests, as appropriate. Adjusted p-values for DXA, HR-pQCT, and IMI outcomes are derived from ANCOVA analysis controlling for age, sex, BMI, history of fracture, diabetes status, alcohol consumption, and smoking. Statistical significance was defined as p < .05. Superscript letters indicate significant pairwise differences between groups as indicated on the table. Data are presented as mean ± SD or %, p < .05 was considered significant. Adjusted p-values shown for DXA, HR-pQCT, and IMI are adjusted for age, BMI, sex, history of fractures, diabetic status, alcohol consumption, osteoporosis tx, and smoking.

^aMissing scans due to hardware/motion score >3.

^b p < .05 for CH vs NHB.

^c p < 0.05 for NHB vs NHW.

^d p < .05 for CH vs NHW.

^eNo pairwise comparisons significant.

Total hip: 1 scan missing for NHW.

Radius: 5 scans missing for CH, 2 for NHB, and 1 for NHW.

Tibia: 1 scan missing for NHB.

Abbreviations: 1/3 R, distal 1/3 of the radius; CH, Caribbean Hispanic; Ct.Th, cortical thickness; NHB, non-Hispanic Black; NHW, non-Hispanic White; Tb.N, trabecular number; Tb.Sp, trabecular separation; Tb.Th, trabecular thickness; Tot.vBMD, Tb.vBMD, Ct.vBMD, total, trabecular, and cortical volumetric BMD.

There were racial differences in skeletal microstructure and mechanical competence (Table 2). At the radius, Tb.N was higher in NHW compared to CH before and after adjustment. Ct.vBMD was higher in CH and NHB compared to NHW before adjustment but differences between CH and NHW were attenuated after adjustment. Ct.Th was higher in NHB compared to NHW before and after adjustment. Radial stiffness was 2.9% and 22.2% lower in NHW compared to CH and NHB, respectively (53.5 ± 17.5 vs 66.7 ± 35.1 vs 51.9 ± 24.1 kN; adjusted p-value = .048).

At the tibia, trabecular vBMD and Tb.N were higher in NHW than in NHB and CH, whereas cortical vBMD and Tb.Sp were higher in CH and NHB than in NHW, before and after adjustment. Ct.Th was higher in CH and NHB compared to NHW before and after adjustment. There were no differences in tibial stiffness before or after adjusting for covariates.

Discussion

To our knowledge, this is the first community-based study to assess how race/ethnicity affects BMSi and the first study to include Hispanic Americans. Despite expected racial/ethnic differences in some aBMD and HR-pQCT/μ-FEA parameters, we did not find BMSi differed by race/ethnicity in NHW, NHB, and CH individuals.^20,24–28 Our results are important because prior studies of BMSi have included very few non-white individuals. The current findings suggest material properties may not differ by race/ethnicity in these particular race/ethnic groups. Structural, densitometric, and geometric factors rather than intrinsic material properties may be the primary drivers of observed differences in skeletal strength across these groups. However, because BMSi is assessed at the tibia, a predominantly cortical site, our findings may primarily reflect cortical material properties and may not capture trabecular material heterogeneity that could contribute to fracture risk at other skeletal sites.

One prior study compared BMSi by race across glycemic status.¹⁴ This was a convenience sample that included 35 Black and 149 white participants, respectively. The study suggested Black participants with diabetes had lower BMSi compared to white participants with diabetes, but this was not the case in non-diabetic or pre-diabetic patients. In contrast, we found no interaction between DM and race/ethnicity on BMSi. This difference could be due to a difference in sampling method or the younger average age of participants in the prior study. The former study did not compare all white to all Black participants.

Another study compared BMSi in 42 Norwegian and 46 Spanish women who were recruited from a clinic.¹³ The mean BMSi value in Norwegian women was significantly lower when compared to Spanish women. Results of this study can’t be directly compared to ours. Caribbean Hispanic populations share Iberian ancestry with individuals from Spain but are genetically distinct due to admixture with African and Indigenous Caribbean ancestries. Our results extend those of prior studies by including Hispanic Americans.

Bone material strength index did not differ by prevalent fracture status or across racial groups in participants with prior fracture. Prior studies, many in postmenopausal White women, have reported inconsistent associations between BMSi and fragility fractures.^12,28–31 Bone material strength index’s clinical utility in the elderly remains uncertain. Moreover, the specific aspects of bone material properties reflected by BMSi remain incompletely defined.

Strengths and limitations

This study also had some limitations. First, the number of participants in the NHB and NHW groups was relatively small, but our study size was comparable to prior studies assessing racial differences in black and white individuals and almost twice as large as that comparing Norwegian and Spanish patients. With the current sample sizes, we had 80% power with a 2-tailed α of .05 to detect a minimum BMSi between-group mean difference of ~3 to 4 BMSi units which is higher than our observed difference of 1-2. We may have failed to detect differences below this, but such differences are not likely to be clinically significant. Most studies suggest BMSi differences of at least 3-5 in patients with and without fracture, consistent with our detectable difference.^{10,21,32–34} The null findings are unlikely to reflect limited power and instead support biologically comparable bone material strength across groups. As this study was performed exclusively in older adults, we cannot exclude the possibility of racial differences in BMSi in younger individuals.

This study also has notable strengths, including limiting selection bias by enrolling participants from a population-based cohort, which enhances generalizability. Additionally, this is the first study to include Hispanic Americans when comparing IMI across race/ethnicity. Further, we have detailed skeletal information on this cohort, including aBMD, HR-pQCT, and FEA along with extensive covariate information.

In summary, BMSi was similar among CH, NHB, and NHW participants, suggesting that differences in bone material strength, as captured by BMSi, do not account for established racial/ethnic differences in fracture risk in these racial/ethnic groups.

Author contributions

Luis Ortiz Vasquez (Formal analysis, Data curation, Writing—original draft, Writing—review & editing), Carmen Germosen (Data curation, Project administration), Isabella Rosillo (Data curation, Methodology), Sanchita Agarwal (Data curation, Formal analysis, Methodology, Writing—review & editing), Ragyie Rawal (Data curation, Formal analysis, Writing—review & editing), Ivelisse Colon (Data curation, Project administration), Mildense del Orbe (Data curation, Project administration), Mariana Bucovsky (Data curation, Project administration, Supervision), Mishaela Rubin (Data curation, Methodology, Writing—review & editing), and Marcella Walker (Conceptualization, Data curation, Funding acquisition, Project administration, Supervision, Writing—original draft, Writing—review & editing)

Funding

This work was supported by the National Institute of Health (NIH) grants R01AR071986, PO1AG07232, R01AG037212, RF1AG054023, and T35DK093430 and an investigator-initiated research grant from Amgen Inc.

Conflicts of interest

The authors have no conflict of interest.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.