Over the past two decades, the significant burden of osteoporosis has been increasingly recognized, in terms of patient morbidity and mortality, healthcare utilization, and societal costs.1–5 Osteoporosis is a skeletal disease characterized by low bone mass and deterioration in bone microarchitecture that results in enhanced bone fragility and increased susceptibility to fractures.6 Bone mineral density (BMD) measured by dual X-ray absorptiometry (DXA) is the current reference standard for the diagnosis of osteoporosis.6–9 Fracture risk continuously increases with decreasing BMD, which is one of its strongest predictors.9–11 However, the greatest limitation of BMD measurement is that a considerable degree of overlap exists in BMD values between individuals with and those without subsequent fractures.12, 13 This observation is witness to the fact that BMD is not the only structural determinant of bone strength. Trabecular bone microarchitecture, for example, also appears to be a significant determinant of bone strength, complementary to bone density.14, 15
The trabecular bone score (TBS) is a novel gray-level texture measurement that is based on the analysis of two-dimensional (2D) projection images that can be extracted from DXA images. The TBS is capable of differentiating between two three-dimensional (3D) microarchitectures of cancellous bone that exhibit the same bone density but different trabecular characteristics.16, 17 TBS values, as evaluated from 2D projection DXA images, demonstrate highly significant correlations with the 3D characteristics of trabecular bone microarchitecture, independent of BMD.16–18 A low TBS value has been found to be associated with degraded bone microarchitecture, low connectivity, high trabecular spacing, and a reduced number of trabeculae; conversely, a high TBS value has been found to be associated with good bone microarchitecture, including high connectivity, low trabecular spacing, and an augmented number of trabeculae.17, 18 Hence, high values of TBS reflect strong, fracture-resistant microarchitecture, whereas low TBS reflects weak, fracture-prone microarchitecture.16, 17, 19, 20 Cancellous bone microarchitectural texture assessed by TBS was a predictor of osteoporotic fracture risk independent of spine and hip BMD in a retrospectively analyzed prospective cohort of 29,407 women, age 50 years or more, living in the Province of Manitoba, Canada. Furthermore, in this study, combining the TBS with BMD was superior to either TBS or BMD alone, with regard to fracture risk prediction.21
Based on the positive results in several fracture endpoint trials, aminobisphosphonates have become the mainstay of antiresorptive treatment aimed at reducing the risk of vertebral and nonvertebral, including hip, fractures in postmenopausal women with osteoporosis.22–24 As such, yearly infusions of 5 mg of zoledronate (ZOL), a third-generation aminobisphosphonate, showed pronounced antifracture efficacy in the HORIZON fracture endpoint trial, producing significant reductions in new vertebral, nonvertebral, and hip fractures over a 3-year period in postmenopausal women with osteoporosis.24 This fracture risk reduction was consistent with improvements in BMD observed at the lumbar spine (LS) and the hip, but the order of magnitude of fracture risk reduction exceeded that expected from BMD changes alone.24, 25 Micro–computed tomography (µCT) analyses of iliac crest bone biopsies performed in a subset of patients included in the HORIZON trial showed that patients treated with ZOL for 3 years had significantly higher trabecular bone volume and number of trabeculae, significantly lower trabecular separation, and numerically higher connectivity density, suggesting better preservation of trabecular structure with ZOL than with placebo (PLB).26 These findings were consistent with observations performed in ovariectomized rats, which showed that decreases in trabecular architectural parameters were dose-dependently attenuated by ZOL and accompanied by dose-related improvements in bone strength parameters.27
The effects of ZOL on the TBS, a noninvasive method for the in situ determination of intravertebral cancellous bone microarchitectural texture, are unknown. The aim of the present study was to compare the effects of yearly intravenous ZOL and PLB during 3 years, both on LS BMD and site-matched bone microarchitectural texture assessed by TBS.
Patients and Methods
The study was performed by retrospectively assessing DXA images to calculate TBS in the full subset of 107 postmenopausal women with osteoporosis recruited at the Department of the University Hospital of Berne, Switzerland, who were included in the HORIZON trial, a randomized, double-blind, placebo-controlled study comparing the effects of once-yearly intravenous ZOL and PLB over 3 years. All patients gave additional written informed consent to a local BMD substudy, which had been approved by the Cantonal Ethical Committee of Berne, Switzerland, as a local amendment to the protocol. TBS analysis was predefined and approved by the institutional board of the University Hospital of Berne, Switzerland.
The HORIZON trial was a (1:1) randomized, double-blind, placebo-controlled clinical trial that compared the effects of once-yearly intravenous ZOL versus PLB in postmenopausal women with osteoporosis. In the original study, a total of 7765 patients were treated after random assignment to a once-yearly infusion over 15 minutes of either ZOL 5 mg (n = 3889) or PLB (n = 3876) for a total length of follow-up of 3 years. All patients received adequate calcium and vitamin D3 supplementation. Major outcomes of the HORIZON trial were fracture incidence, change in BMD, and drug safety and tolerability. Full details on the HORIZON trial design and results have been published.24
Patients included in the present study were the subset of all patients randomized and included in the HORIZON trial at the participating center of the Department of Osteoporosis of the University Hospital of Berne, Switzerland, corresponding to 107 postmenopausal women with osteoporosis treated with either ZOL (n = 54) or PLB (n = 53) over 3 years.
Measurement of TBS
DXA images (Hologic QDR 4500A; Hologic Inc., Bedford, MA, USA) of the anteroposterior LS (L1–L4)—collected at baseline, and at follow-up evaluations 6, 12, 24, and 36 months after the initiation of intravenous ZOL or PLB treatment—were sent to the Bone Disease Unit of the University Hospital of Lausanne, Switzerland, for blinded calculation of TBS values using TBS iNsight Software (version 1.9; Med-Imaps, Bordeaux, France). TBS was evaluated in the same regions of measurement as those used for the LS BMD, with TBS calculated as the mean value of the individual measurements for vertebrae L1–L4 from the original DXA scans, excluding any fractures or degenerative changes, in accordance with International Society for Clinical Densitometry (ISCD) rules for individual vertebrae exclusion (more than 1 SD from immediately adjacent vertebrae).28 As a result, in the PLB group, 30 patients with all vertebrae, 15 patients with three vertebrae, and 7 patients with two vertebrae were included, with 1 patient being excluded because only one vertebra was assessable. In the ZOL group, 23 patients with all vertebrae, 26 patients with three vertebrae, and 4 patients with two vertebrae were included, with 1 patient excluded because only one vertebra was assessable. The majority of the exclusions were related to degenerative changes in L4. All analyses were performed on the same vertebrae for both BMD and TBS parameters so that the results can be directly compared. Previously reported precision for TBS ranged between 1.48% (in-house data) and 2.1%.21 At the Department of Osteoporosis of the University Hospital of Berne, the coefficient of variation for LS BMD measurements assessed in accordance with ISCD recommendations (15 outpatients representative of daily routine with triplicate measurements after repositioning) is 0.90% with a corresponding coefficient of variation of 1.12% for TBS. Thus, the least significant change (LSC) is 2.49% for LS BMD and 3.10% for TBS.
After checking for normality distribution, descriptive analysis included means and percentages with SDs. Change in BMD and TBS from baseline were calculated for each subject as the absolute change divided by the baseline value and converted into a percentage. Bivariate intergroup comparisons were performed using Student's t test and Pearson's χ2 analysis for continuous variables. Data were analyzed on an intention-to-treat basis. The threshold for statistical significance was set at p < 0.05; and all inferential tests were two-tailed. All statistical analyses were performed using Stata (Version 12; StataCorp LP, College Station, TX, USA).
As shown in Table 1, baseline characteristics (mean ± SD) were similar between groups in terms of age, 76.8 ± 5.0 years; BMI, 24.5 ± 3.6 kg/m2; TBS, 1.178 ± 0.1; but not for LS T-score (ZOL–2.9 ± 1.5 versus PLB–2.1 ± 1.5; p = 0.01). These baseline characteristics were generally consistent with the mean values reported in the full population of the HORIZON trial.24
Table 1. Baseline Characteristics
ZOL (n = 54)
PLB (n = 53)
Values are mean ± SD.
ZOL = zoledronate; PLB = placebo; ns = nonsignificant; BMI = body mass index; TBS = trabecular bone score.
76.6 ± 4.8
77.0 ± 5.2
157.8 ± 6.3
156.7 ± 4.8
61.3 ± 9.4
60.0 ± 9.0
24.6 ± 3.3
24.5 ± 3.8
Lumbar spine TBS
1.176 ± 0.11
1.181 ± 0.09
Lumbar spine T-score
–2.9 ± 1.5
–2.1 ± 1.5
Because of exclusion, no show, or inability to retrieve the DXA scans, the number of subjects varied by visit, as shown in Fig. 1.
As shown in Table 2 and Fig. 1, changes in LS BMD were significantly greater with ZOL than with PLB at all time points,29 reaching +2.59% versus +0.04% and +9.58% versus +1.38% at months 6 and 36, respectively (p < 0.0001 for all). The change in TBS was of significantly larger magnitude with ZOL compared to PLB at month 24 (+1.11% versus–0.45%; p = 0.049) and month 36 (+1.41% versus–0.49%, p = 0.031). Compared to baseline, TBS increased significantly with ZOL at months 24 and 36 (+1.11%; p < 0.05, and +1.41%; p < 0.04, respectively), but was not significantly different from baseline with PLB at any time point.
Table 2. Comparisons Between Lumbar Spine BMD and TBS Changes in ZOL-Treated Versus PLB-Treated Women Through 36 Months of Follow-Up
Change from baseline BMD (%)
p (PLB versus ZOL)
Change from baseline TBS
Values are mean ± SEM.
BMD = bone mineral density; TBS = trabecular bone score; ZOL = zoledronate; PLB = placebo; ns = nonsignificant.
Significantly different from baseline p < 0.05.
Significantly different from previous visit p < 0.05.
TBS showed a slightly greater sensitivity than BMD (26% versus 19%; McNemar test p < 0.6) with regard to the proportion of patients below LSC in the PLB group. As expected, BMD was more sensitive than TBS with regard to the proportion of patients achieving an increase above LSC in treated patients (96% versus 35%; p < 0.001).
Over 3 years, LS BMD and TBS were only weakly correlated (r = 0.20); and there were no significant correlations between the change in BMD and TBS from baseline at any follow-up time point.
In the present prospectively defined retrospective analysis of the full subset of women with postmenopausal osteoporosis randomized to once-yearly infusions of either ZOL or PLB in the HORIZON study center of the Department of Osteoporosis of the University Hospital of Berne, Switzerland, ZOL induced a significantly larger sustained increase in LS BMD over 36 months and allowed for preservation of bone microarchitectural texture assessed by TBS at month 36. Furthermore, BMD and TBS were only weakly correlated.
TBS is a novel tool for fracture risk assessment in postmenopausal women.21 Its primary value beyond BMD is that it may specifically assess some dimensions of vertebral bone microarchitectural texture, which is recognized to be an important determinant of bone strength independent of BMD.12, 14, 30, 31 Several studies have demonstrated that TBS was strongly correlated with 3D microarchitecture parameters (such as bone volume/trabecular volume [BV/TV], connectivity density, trabecular number [TbN], trabecular thickness [TbTh], trabecular spacing [TbSp], structure model index [SMI]) as assessed by using high-resolution CT (93 µm)17, 18 and µCT (35 µm).32 Correlations remained significant even at low resolution.33 Using TBS accurately differentiated between two 3D samples in which the amount of bone was identical, but characteristics differed with regard to TbN, TbTh, or TbSp.18 More recently, Roux and colleagues32 compared the TBS with microarchitectural parameters measured using µCT with a 35-µm isotropic resolution on lumbar vertebrae (L3) freshly harvested from 16 human donors. The TBS was significantly correlated to the BV/TV and the SMI (which reflects the rod-like versus plate-like nature of the structure) with correlations coefficients of r = 0.518 (p = 0.040) and–0.597 (p = 0.015), respectively. They concluded that TBS was significantly correlated to the most relevant microarchitectural parameters used to predict fracture risk (ie, BV/TV and SMI), might improve bone assessment in association with BMD using standard DXA, and could enhance fracture risk assessment without requiring µCT acquisitions in elderly patients. In summary, an elevated TBS reflects strong, fracture-resistant microarchitecture; a low TBS reflects weak, fracture-prone microarchitecture. TBS is therefore not a physical measurement but rather an index of the trabecular pattern of the measured bone strongly related to bone strength.
In the current study, we sought to establish whether TBS was useful for monitoring the effectiveness of antiosteoporosis therapy beyond BMD within the context of a randomized clinical trial. In the subset of the patient population intended for BMD analysis in the initial HORIZON trial, the between-group difference in LS BMD over 36 months measured by DXA was 7.0% (p < 0.01),34 which is consistent with the between-group difference observed in the present analysis (8.20%, p < 0.0001). Although BMD increased by 9.58% versus baseline with ZOL over 3 years, TBS increased by only 1.41%, which corresponds to a change ratio of approximately 7. This finding is consistent with recent observations made in a retrospective analysis of the cohort included in the large provincial database in Manitoba, Canada, in which most of the women treated against osteoporosis used a bisphosphonate and in which the change in BMD over 3 years approximated 2% per year, versus 0.2% per year for TBS (Krieg MA, Goertzen AL, Aubry-Rozier A, Hans D, Leslie WD. Effects of antiresorptive agents on bone micro-architecture assessed by trabecular bone score (TBS) in older women. unpublished work). Furthermore, in the present study, changes in BMD and in TBS were not correlated, supporting the findings of published research, in which the two parameters were found to measure different characteristics of bone (Krieg and colleagues, unpublished work).21
Once-yearly intravenous ZOL was shown to be significantly superior to PLB with regard to the reduction of vertebral and nonvertebral fracture risk, including at the hip.24, 35 However, the order of magnitude of fracture risk reduction exceeded that expected from BMD changes alone.25, 34 The analysis of iliac crest bone biopsies performed in a subset of HORIZON trial patients by µCT identified significantly better preservation of trabecular structure with ZOL than with PLB26 and microspectroscopy analysis of these biopsies showed an increased mineral/matrix ratio with a larger proportion of younger bone tissue with ZOL compared with PLB.36 These findings suggest that ZOL may exert its fracture risk reduction effects by increasing bone mass and by preserving bone quality in terms of bone microarchitecture and bone material properties. These findings are consistent with the improvements in TBS versus baseline observed with ZOL but not with PLB over 3 years.
This study can be considered a first step toward establishing a possible role for TBS in the monitoring of osteoporosis therapy at the individual patient level, which would suppose known precision and LSC.37 In the present study, the LSC was 3.1% for TBS, leading to 35% of treated patients being above LSC. Precision reported in this study is better than the generally reported precision for LS TBS reported in the literature, which ranges from 1.5% to 2.1%.21 This difference may be explained by the stringent training received by all technicians as well as by the specific study procedures required by the HORIZON trial. Although 35% of treated patients showed a significant TBS increase consistent with a microarchitectural texture improvement, 65% remained below LSC such that the overall effects of ZOL on bone trabeculae can be considered as a preservation of vertebral microarchitectural texture in the majority of patients (Krieg and colleagues, unpublished work).38
In several studies,16, 19–21 TBS has been found to: (1) be lower in postmenopausal women with a past osteoporotic fracture relative to age- and BMD-matched women without fracture; (2) as it decreases, incrementally increase the odds ratio for spine fracture when combined with LS BMD; (3) to be lower in women with fractures, irrespective of whether their BMD meets the criteria for osteoporosis or osteopenia; (4) prospectively predict fracture as well as LS BMD; and (5) to capture roughly one-third of fractures misclassified using the BMD-dependent WHO definition of osteoporosis alone. With this study being considered as a first step, further studies with other bone active substances, such as teriparatide, are needed to elucidate whether, and if so to what extent, the combination of BMD and TBS assessments may contribute to improve monitoring of osteoporosis treatment at the individual patient level.
The present study has some limitations due to the retrospective nature of the analysis. However, the analysis was predefined and performed by intention to treat by persons blinded to treatment allocation and patient characteristics. The analysis was applied to the complete subset of a large population of postmenopausal women with osteoporosis included in a randomized, placebo-controlled trial by a single center. Finally, baseline characteristics of patients included in the present analysis were comparable to those of the full patient population included in the HORIZON trial. Although the subset of 107 patients was large enough for comparing the effects of ZOL versus PLB on TBS, it was not large enough for establishing the predictive value of TBS alone or in combination with BMD with regard to fracture risk.
In a prospectively defined retrospective analysis of a subset of postmenopausal women with osteoporosis included in the HORIZON fracture endpoint trial by a single center, once-yearly intravenous ZOL therapy significantly increased LS BMD relative to PLB over 3 years and TBS as of 2 years. Change in BMD and TBS were not correlated, confirming that TBS reflects bone properties other than BMD, more specifically bone microarchitectural texture. Further research is warranted to delineate the value of TBS as an index of treatment-related antifracture effects, as well as to explore the effects of other classes of bone active substances on TBS.
TBS iNsight software is a product of Med-Imaps. DH is co-owner of the TBS patent and has corresponding ownership shares. All of the other authors state that they have no conflicts of interest.
We are grateful to Philippe Kress, MD, for reviewing and commenting on our manuscript.
Authors' roles: Study design: KL and DH. Study conduct: AP, HB, CS, SG, and RP. Data analysis: DH, RP, HB, AP, and SG. Data interpretation: KL, DH, and AP. Drafting manuscript: SG, AP, OL, CS, and KL. Revising manuscript content: KL and DH. Approving final version of manuscript: KL and DH. DH and KL take responsibility for the integrity of the data analysis.