Association of Trabecular Bone Score (TBS) with Incident Clinical and Radiographic Vertebral Fractures Adjusted for Lumbar Spine BMD in Older Men: A Prospective Cohort Study

Abstract

The association of Trabecular Bone Score (TBS) with incident clinical and radiographic vertebral fractures in older men is uncertain. TBS was estimated from baseline spine DXA scans for 5,831 older men (mean age 73.7 years) enrolled in the Osteoporotic Fractures in Men (MrOS) study. Cox proportional hazard models were used to determine the association of TBS (per 1 SD decrease) with incident clinical vertebral fractures. Logistic regression was used to determine the association between TBS (per 1 SD decrease) and incident radiographic vertebral fracture among the subset of 4,309 men with baseline and follow-up lateral spine radiographs (mean 4.6 years later). We also examined whether any associations varied by body mass index (BMI) category. TBS was associated with a 1.41 (95% C.I. 1.23 to 1.63) fold higher aged-adjusted odds of incident radiographic fracture, and this relationship did not vary by BMI (p-value=0.22 for interaction term). This association was no longer significant with further adjustment for lumbar spine BMD (OR 1.11, 95% C.I. 0.94 to 1.30). In contrast, the age-adjusted association of TBS with incident clinical vertebral fracture was stronger in men with lower BMI (≤ median value of 26.8 kg/m²; HR 2.28, 95% C.I. 1.82 to 2.87) than in men with higher BMI (> median; HR 1.60, 95% C.I. 1.31 to 1.94; p-value=0.0002 for interaction term). With further adjustment for lumbar spine BMD, the association of TBS with incident clinical vertebral fracture was substantially attenuated in both groups (HR 1.30 [95% C.I. 0.99 to 1.72] among men with lower BMI and 1.11 [95% C.I. 0.87 to 1.41] among men with higher BMI). In conclusion, TBS is not associated with incident clinical or radiographic vertebral fracture after consideration of age and lumbar spine BMD, with the possible exception of incident clinical vertebral fracture among men with lower BMI.

Introduction

Trabecular Bone Score (TBS) is associated with incident hip and major osteoporotic fractures in older men, ⁽¹⁾ but its association with incident vertebral fracture is uncertain. Among men in the Manitoba Bone Density Cohort, TBS was not associated with incident clinical vertebral fracture ascertained using administrative claims data, but the accuracy of claims data for incident vertebral fracture may be suboptimal.^(2,3) Additionally, among older men no prior study has examined the association of TBS with incident vertebral fractures identified on the basis of radiographs alone.

The association of TBS with trabecular volumetric spine BMD (assessed with computed tomography) is weaker with increasing body mass index (BMI),⁽⁴⁾ and TBS is not valid for individuals with a BMI > 37 kg/m². Hence, it is conceivable that the association of TBS with incident vertebral fractures varies by BMI, but no study to date has explicitly examined this question. It is also unknown whether assessment of TBS is associated with incident vertebral fracture in men after consideration of lumbar spine BMD and prevalent vertebral fracture status.

Our aim was to determine the associations of TBS with incident radiographic and clinical vertebral fractures in a cohort of community-dwelling older men.

Materials and Methods

From 2000 to 2002, 5994 community-dwelling men ≥65 years old were enrolled into the prospective MrOS study in six regions of the United States, described in previous publications.^(5,6)

Measurement of TBS and BMD

BMD was measured at the femoral neck and lumbar spine with QDR-4500 densitometers (Hologic, Bedford, MA, USA), at the baseline MrOS visit. Consistent with ISCD recommendations, deformed vertebral levels were excluded. TBS was scored on PA spine DXA scans using TBS iNsight (Med-Imaps SASU, Merignac, France, v2.1) for the lumbar vertebrae L1 to L4 that had not been excluded in the spine BMD measurement. We cross-calibrated densitometers across study centers with both a standard spine phantom⁽⁷⁾ and a TBS phantom to ensure consistency and quality of bone mass measurement.

Ascertainment of Prevalent Radiographic Vertebral Fracture

All technically adequate lateral lumbar and thoracic spine X-rays were scanned to digital format. Trained technicians used a validated triage process to identify men (n=2618) with unequivocally normal X-rays.⁽⁸⁾ An expert reader evaluated baseline radiographs of the remaining 3213 men for prevalent radiographic vertebral fracture using the Genant semi-quantitative (SQ) criteria.⁽⁹⁾

Measurement of other covariates

We measured participants’ height with a Harpenden stadiometer, weight with a balance beam or electronic scale, and calculated BMI as weight (kg) divided by height squared (m²). Participants self-reported status regarding fractures experienced after age 50, parental hip fracture, tobacco use and use of systemic glucocorticoids. Self-reported physical activity was assessed with the Physical Activity Scale for elderly (PASE) questionnaire.

Ascertainment of Incident Clinical Vertebral Fractures

Participants were queried every 4 months by questionnaire whether or not they had a fracture (>99% of follow-up contacts were completed in active surviving participants). Self-reported fractures were confirmed by comparison of the community clinical imaging study (spine radiographs, CT, and/or MRI) to the baseline MrOS lateral spine radiographs by the MrOS Coordinating Center study radiologist. Confirmed incident clinical vertebral fractures were those with an increase of SQ grade ≥1 compared to the baseline spine radiograph.

Ascertainment of Incident Radiographic Vertebral Fracture

We identified incident radiographic vertebral fractures by comparing baseline and follow-up spine x-rays using a previously described protocol.⁽⁹⁾ All follow-up films underwent a validated triage ⁽⁸⁾ to differentiate those x-rays that had no possible fractured vertebra from those that needed further review. A physician reader evaluated follow-up films from participants with one or more possible vertebral deformities using the SQ method, with the revision that for grade 1 incident fractures, endplate depression or cortical discontinuity was also required to distinguish fracture from non-fracture deformities. Incident radiographic vertebral fractures were defined as those with a change in SQ grade of ≥1 from baseline to follow-up radiograph.

Statistical Analysis

The associations of participant baseline characteristics with TBS category (normal [≥ 1.35], partially degraded [>1.20 and <1.35], or degraded [≤ 1.20]) were tested with ANOVA (or non-parametric equivalent) for continuous variables, and with chi-square statistic for categorical variables. The associations of TBS and other predictor variables with incident clinical vertebral fractures and incident radiographic vertebral fractures were estimated using, respectively, Cox proportional hazards and logistic regression models. TBS was expressed as a continuous variable and the HR (OR) was estimated per 1 SD decrease in TBS. Initial models were adjusted for age and study enrollment site. An interaction term between TBS and BMI was then added, and if that interaction term was significant (p-value<0.05), models were run in subsets of men at or below and in men above the median BMI value of 26.8 kg/m². Models were then adjusted first for lumbar spine BMD, then for prevalent radiographic vertebral fracture, and finally for self-reported prior fractures, parental history of hip fracture, BMI, current smoking, glucocorticoid use, and PASE score. Post-regression diagnostics were performed to assure that Cox models did not violate the proportional hazards assumption, and that logistic models were well calibrated (using the Hosmer-Lemeshow test).

We performed secondary analyses to examine if the association of TBS with incident vertebral fracture might also vary by BMD, running regression models adjusted for age, spine BMD, study site, and an interaction term between spine BMD and TBS.

Results

For analyses with incident clinical vertebral fracture as the outcome, the study population comprised 5,831 men (97.3% of all MrOS enrollees) with valid TBS scores, baseline lateral lumbar and thoracic spine radiographs interpretable for prevalent vertebral fracture, and a BMI < 37 kg/m². A total of 202 men (3.5%) had one or more confirmed incident clinical vertebral fractures over a mean follow-up time of 11.5 (SD=4.5) years. Among the subset of 4,309 men who also had interpretable follow-up lateral spine radiographs (taken a mean 4.6 [SD=0.4] years later), 196 men (4.6%) had one or more incident radiographic vertebral fractures. Men with lower TBS were older; had lower lumbar spine BMD and a higher prevalence of radiographic vertebral fracture; were less active and more likely to report prior clinical fracture, use of glucocorticoid therapy, and current smoking (Table 1).

Table 1.

Baseline Characteristics according to TBS Level

Characteristics	TBS Category (N=5831*) [Range: 0.66–1.73]			P-Value†
	Degraded (N= 2106) [≤1.200]	Partially Degraded (N = 2819) (>1.200 to <1.350)	Normal (N = 906) (≥ 1.350)
Age, years (SD)	73.8(5.8)	73.8(6.0)	73.1(5.6)	0.01
Lumbar Spine BMD g/cm2, mean (SD)	1.00(0.17)	1.08(0.18)	1.19(0.18)	<0.0001
Lumbar Spine BMD Category^ Osteoporosis Osteopenia Normal	138 (6.6%) 632 (30.0%) 1336 (63.4%)	61 (2.2% 511 (18.1%) 2247 (79.7%)	3 (0.3%) 57 (6.3%) 844 (93.4%)	<0.0001
Prevalent Vertebral Fracture (Grade 2 or 3), number (%)	229(10.9%)	167(5.9%)	40(4.4%)	<0.0001
Parental history of hip fracture, number (%)	285(13.5%)	358(12.7%)	102(11.3%)	0.23
Prior fracture after age 50, number (%)	417(19.8%)	474(16.8%)	110(12.1%)	<0.0001
On Glucocorticoid therapy, number (%)	68(3.2%)	44(1.6%)	12(1.3%)	<0.0001
Body Mass Index (BMI) kg/m2 mean (SD)	28.8(3.6)	26.4(3.0)	25.7(2.9)	<0.0001
Current Smoking status, number (%)	99(4.7%)	83(2.9%)	19(2.1%)	0.0002
PASE Score, mean (SD)	140.9(67.3)	150.2(68.5)	149.8(66.5)	<0.0001

^*N=5831 Men with valid TBS scores, baseline lateral spine radiographs interpretable for vertebral fracture and BMI < 37 kg/m²

^{^}Osteoporosis: lumbar spine BMD T-score ≤ −2.5; Osteopenia: T-score <−1.0 and >−2.5; Normal: T-score ≥ −1.0

^†p-value for differences across TBS category performed with a Fisher exact test for lumbar spine BMD, one-way analysis of variance for other continuous variables, and by chi-square for categorical variables

Each 1 SD decrease in TBS was associated with an age-adjusted 1.41 fold higher odds (95% C.I. 1.23 to 1.63) of incident radiographic vertebral fracture, and this did not vary by BMI category (p-value=0.22 for BMI-TBS interaction term). However, this association was not significant after further adjustment for lumbar spine BMD (OR 1.11, 95% C.I. 0.94 to 1.30). Prevalent radiographic vertebral fracture strongly predicted incident radiographic vertebral fracture after adjustment for age, TBS, and lumbar spine BMD (table 2).

Table 2.

Age- and Study Site-Adjusted Associations^* of TBS with Incident Vertebral Fracture

	BMI Category	Predictor	Model 1	Model 2	Model 3
Incident Radiographic Vertebral Fracture^			OR (95% C.I.)	OR (95% C.I.)	OR (95% C.I.)
	N/A	TBS†	1.41 (1.23,1.63)	1.11 (0.94,1.30)	1.04 (0.88,1.23)
		Lumbar Spine BMD†		1.95 (1.63,2.34)	1.80 (1.50,2.16)
		Prevalent Radiographic Vertebral Fracture			4.42 (3.10,6.29)
Incident Clinical Vertebral Fracture‡			HR (95% C.I.)	HR (95% C.I.)	HR (95% C.I.
	N/A	TBS†	1.62 (1.42, 1.84)	1.19 (1.02,1.38)	1.15 (0.98,1.34)
		Lumbar Spine BMD†		2.50 (2.09,3.00)	2.36 (1.98,2.83)
		Prevalent Radiographic Vertebral Fracture			2.29 (1.61,3.26)
	≤ 26.8 kg/m2	TBS†	2.28 (1.82,2.87)	1.30 (0.99,1.72)	1.24 (0.94,1.65)
		Lumbar Spine BMD†		2.53 (1.95,3.27)	2.42 (1.87,3.12)
		Prevalent Radiographic Vertebral Fracture			2.33 (1.49,3.66)
	>26.8 kg/m2	TBS†	1.60 (1.31,1.94)	1.11 (0.87,1.41)	1.08 (0.85,1.37)
		Lumbar Spine BMD†		2.46 (1.80,3.36)	2.33 (1.70,3.15)
		Prevalent. Radiographic Vertebral Fracture			2.22 (1.25,3.93)

^*Hazard and Odds Ratios that are significant at p<0.05 level are in bold italics

^†per standard deviation (SD) decrease of predictor variable

^{^}196 men had an incident radiographic vertebral fracture (mean follow-up time [SD] 4.6 [0.4]) years)

^‡202 men had an incident clinical vertebral fracture (mean follow-up time [SD] 11.5 [4.5] years)

Each 1 SD decrease in TBS was associated with an age-adjusted increase of incident clinical vertebral fracture (HR 1.62, 95% C.I. 1.42 to 1.84), and this association did vary by BMI (p-value = 0.002 for interaction). The age-adjusted association of TBS was higher among men with BMI ≤26.8 kg/m² (HR 2.28, 95% C.I. 1.82 to 2.87) than among men with BMI >26.8 kg/m² (HR 1.60, 95% C.I. 1.31 to 1.94, Table 2). After adjustment for lumbar spine BMD, the association of TBS with incident clinical vertebral fracture was substantially attenuated in both men with BMI≤26.8 kg/m² (HR 1.30, 95% C.I. 0.99 to 1.72) and men with BMI>26.8 kg/m² (HR 1.11, 95% C.I. 0.87 to 1.41). In contrast, the association of prevalent radiographic vertebral fracture with incident clinical vertebral fracture did not vary by BMI category (Table 2).

All of these associations were not altered by additional adjustment for prior clinical fracture, current use of glucocorticoid medication, current smoking status, body mass index, and PASE score (data not shown). The age- and spine BMD-adjusted associations of TBS with incident radiographic vertebral fracture and clinical vertebral fracture among all men, men with BMI below the median, and men with BMI above the median did not vary by spine BMD (p-values for interaction between TBS and spine BMD, respectively, 0.99, 0.67, 0.54, and 0.33).

Discussion

In this population-based cohort of U.S. older men, TBS was associated with age-adjusted increase in risk of incident clinical and radiographic vertebral fractures. However, these associations were substantially attenuated after adjustment for lumbar spine BMD.

We observed that the association between TBS and clinical vertebral fracture varied by BMI, being attenuated among those with higher BMI. This is consistent with the hypothesis that a larger thickness of soft tissue in front of and behind the lumbar spine may introduce more TBS measurement error,⁽¹⁰⁾ and supported by our previous finding that the association of TBS with volumetric trabecular BMD measured with lumbar spine QCT becomes weaker as BMI increases.⁽⁴⁾

In general, our results from MrOS are consistent with findings from the Manitoba Bone Density database as TBS in the latter study was not associated with incident clinical vertebral fractures in men after adjustment for lumbar spine BMD. In contrast, TBS has been shown to be associated with incident clinical vertebral fractures among women in the Manitoba cohort⁽¹¹⁾ and with incident radiographic vertebral fractures among women in the OPUS⁽¹²⁾ and JPOS⁽¹³⁾ studies, adjusted for both age and lumbar spine BMD. The reason for these observed differences between the sexes is unknown. Although men on average have increased abdominal girth compared to women, TBS is associated with incident hip and major osteoporotic fractures in Manitoba older men adjusted for age and lumbar spine BMD.⁽¹¹⁾ Moreover, in the FRAX cohorts TBS is equally predictive of major osteoporotic and hip fractures in men and women.⁽¹⁾

It remains unclear precisely what aspects of bone architecture and quality TBS actually represents, or why it predicts fractures.⁽¹⁴⁾ TBS is postulated to be a surrogate marker of trabecular microarchitecture,^(15,16) but is not associated with vertebral strength estimated with finite element modeling after adjustment for lumbar spine areal BMD.⁽¹⁷⁾ TBS only explains a small proportion of the variation of trabecular microarchitectural parameters (trabecular number, spacing and thickness, and structure model index [a measure of the plate vs rod-like configuration of the trabeculae]) measured on HRpQCT scans of the tibia or radius,^(18,19) and a moderate proportion of the variance of these parameters measured on iliac crest bone biopsies.⁽²⁰⁾

Our study has limitations. We did not have statistical power to examine the associations of TBS with incident vertebral fractures within the traditional categories of BMI (normal 18.5 to 24.9 kg/m²; overweight 25.0 to 29.9 kg/m²; and obese ≥ 30 kg/m²). Although 10% of the MrOS population is non-white, our results are not generalizable to non-Caucasian men. We did not test whether the associations of TBS with vertebral fracture varies by measures of abdominal girth (such as waist circumference or abdominal wall tissue thickness). It is possible that the association of TBS with incident vertebral fracture varies by these measures to a greater degree than by BMI; further studies are required to test these hypotheses.

In conclusion, TBS is not associated with incident vertebral fracture in older men once age and lumbar spine BMD are known, with the possible exception of clinical vertebral fracture in men with lower BMI. In contrast, lumbar spine BMD and prevalent vertebral fracture are strongly associated with incident radiographic and clinical vertebral fractures in older men. Delineation of the differences in vertebral macro and microarchitecture between men and women, and their associations with lumbar spine BMD might help explain the discrepant associations of TBS with incident vertebral fractures in men compared to women.