Introduction

Bone is a dynamic tissue, and the formation and resorption are continuous processes during life [1]. Osteoporosis and fractures are important complications of aging in men. According to the National Osteoporosis Foundation, up to 25% of men over the age of 50 years will experience a fracture due to osteoporosis [2]. In 2050, the incidence of hip fracture in men is expected to increase by 310% worldwide [3]. Moreover, men suffering from any major fracture have a higher mortality rate than women [4]. Indeed, the lifetime risk of experiencing an osteoporotic fracture in Caucasian men is 13% and is greater than the lifetime risk of developing another frequent cause of morbimortality in men: prostate cancer (11.3%) [5, 6].

Currently, measurement of the areal bone mineral density (aBMD) in the lumbar spine and femoral neck using dual-energy X-ray absorptiometry (DXA) is most commonly used to diagnose osteoporosis and assess bone quantity and changes in bone mass across the life span [7]. DXA measurements are an indirect surrogate for bone strength and have some limitations on distinguishing patients at low and high fracture risk [7, 8]. For instance, it can neither discriminate between cortical and trabecular bones nor evaluate bone microarchitecture [9].

Trabecular bone score (TBS) is a measurement that can be applied to DXA images. It is a gray-level textural index of bone architecture derived from lumbar spine DXA images and is an independent indicator of fracture risk [10]. Additionally, TBS is an important determinant of bone strength [11]. The association of TBS with incident clinical fractures in men has been consistently estimated in two previous cohorts [11, 12].

High-resolution peripheral quantitative computed tomography (HR-pQCT) is a technology that provides three-dimensional images allowing assessment of volumetric bone density and bone microarchitecture components in the examined area [8, 13]. After the acquisition of standard and automated segmentation methods, dedicated finite element (FE) analysis provides a direct subject-specific estimation of bone strength [14,15,16,17,18,19].

Prior studies on HR-pQCT have not concomitantly evaluated bone features using different technologies, which is essential for determining the contribution of each one to fracture risk prediction in adult men [8, 16, 19,20,21,22,23,24]. Furthermore, no study has evaluated bone by HR-pQCT in men from Latin America. As the availability of HR-pQCT increases with a growing number of health professionals having access to the technology, there is an unmet need to establish reference data that can be used as a basis for assessing bone health in different populations.

The primary objective of the present cross-sectional study was to (a) establish age-related reference values for volumetric bone density and morphometry by HR-pQCT and FE parameters of distal radius and tibia, in a healthy male population, and (b) correlate bone stiffness of distal radius and tibia with other HR-pQCT parameters, areal BMD and TBS.

Methods

Participants

A total of 340 healthy men aged 20 to 92 years were enrolled in the study. This population included workers from the University of São Paulo School of Medicine, Hospital das Clinicas of the University of São Paulo, or family members of the employees. At least 40 subjects per decade of age were recruited. Exclusion criteria were bone-associated metabolic disease (rickets, primary hyperparathyroidism, osteomalacia, and Paget’s disease), prior non-traumatic fracture, chronic disease (diabetes mellitus, renal or liver failure, hyperthyroidism, hypothyroidism, and malabsorption), use of any medication that interferes with bone metabolism (bisphosphonates, teriparatide, glucocorticoids, anticonvulsants, anticoagulants), current smoking (tobacco use), and alcohol intake ≥ 3 U/day.

This study was approved by the Local Ethics in Research Committee of the São Paulo University School of Medicine, and all participants gave written informed consent.

Data collection

Demographic and anthropometric characteristics, including race, age, height, weight, and body mass index (BMI) were recorded. Race was defined based on self-reporting of the second generation of ancestors, an approach previously used for the Brazilian population [25, 26]. Individuals who reported having four White grandparents were classified as White. Individuals with black African and White ancestry (mixed race) were classified as non-White. When racial information on an individual’s grandparents was not available, the race of an individual was determined by the race of his parents. The descendants of other races were not included. All the participants underwent a standardized interviewer-administered questionnaire that ascertained clinical characteristics, including smoking (tobacco use) status, alcohol consumption, physical activity (including work activities), previous non-vertebral fracture, hip fracture in first-degree relatives, comorbidities, and current medication use [26,27,28].

Dual-energy X-ray absorptiometry (DXA)

Areal bone mineral density (aBMD, g/cm2) measurements of the lumbar spine (L1–L4) and proximal femur (femoral neck and total femur) were obtained by dual-energy X-ray absorptiometry (DXA) equipment (Hologic QDR 4500A; Hologic Inc. Bedford, MA, USA, Discovery model). All DXA measurements were performed by the same experienced technologist.

Precision error for DXA measurements was determined based on standard ISCD protocols [29]. We calculated the least significant change with 95% confidence to be 2.3% for the lumbar spine, 3.8% for the femoral neck, and 2.6% for the total femur.

Trabecular bone score (TBS) was calculated by the software TBS iNsight® version 2.1 (Med-Imaps, Bordeaux, France). It was analyzed as the mean value of the measurements for vertebrae L1–L4 of DXA images at exactly the same ROI as the spine BMD measurements. TBS values were classified according to the following criteria [10]: TBS ≥ 1.310: normal; 1.200 < TBS > 1.310: partially degraded microarchitecture; TBS ≤ 1.200: degraded microarchitecture. In accordance with manufacturer recommendations, TBS was only assessed in patients with a BMI above 15 kg/m2 and lower than 37 kg/m2 [30].

High-resolution peripheral quantitative computed tomography (HR-pQCT)

The non-dominant radius and left tibia were immobilized on a carbon fiber shell and scanned on the first-generation HR-pQCT scanner (XtremeCT I, Scanco Medical AG, Brüttisellen, Switzerland) using a standard scanning protocol (60 kVp, 1000 μA). The measurements included 110 slices, corresponding to a 9.02 mm scan area (voxel size of 82 μm), positioned 9.5 mm and 22.5 mm proximal to the reference line for the distal radius and tibia, respectively.

All examinations were conducted by a trained biomedical scientist who also carefully examined each scan for motion artifacts. In the case of significant motion artifacts identified during scan acquisition, a second examination was performed. All HR-pQCT images were scored based on a motion scale ranging from 0 (no movement) to 4 (severe blurring of the periosteal surface and discontinuities in the cortical layer) [31, 32]. For the current study, scans with a score of 4 were excluded. Quality control was daily monitored using a phantom calibration provided by the manufacturer.

The standard morphologic analysis and advanced cortical (automated segmentation) method of the scanner were used for the analysis and have been described in detail elsewhere [20, 33, 34]. The outcome variables used in our analyses were density parameters, including total volumetric bone density (Tt.vBMD; mg HA/cm3), trabecular volumetric bone density (Tb.vBMD; mg HA/cm3), and cortical volumetric bone density (Ct.vBMD; mg HA/cm3); structure parameters, including number (Tb.N; mm-1), thickness (Tb.Th; mm), and separation (Tb.Sp; mm) of trabeculae, cortical thickness (Ct.Th; mm), and cortical porosity (Ct.Po; %).

In our laboratory, HR-pQCT measurements in vivo had variation coefficients ranging from 1.49 to 7.59% at the distal radius and 0.78 to 6.35% at the distal tibia for morphometry parameters, and 0.93 to 1.41% at the distal radius and 0.25 to 1.16% at the distal tibia for densities [24, 33].

Finite element analysis (FEA)

To estimate bone strength from HR-pQCT measurements, linear finite element (FE) models of the distal radius and tibia were created directly from the HR-pQCT images using software-specific finite elements (Finite Element software v. 1.13, Scanco Medical AG, Switzerland, January 2009). The standard analysis of FEA comprises a virtual resistance test, which means that the computer analyzes the behavior of the bone tissue when it is submitted to a compressive force along its major axis. Two mechanical properties of the bone are estimated: Young’s modulus, a measure of the ability of a material to return to its original shape after removal of a stress force, thus indicating the tissue’s elasticity; and the Poisson effect, which is the tendency of a material to become thinner when it is stretched at a given axis. In other words, when a material is pulled, it increases its size in the axis of traction, and decreases its size in the other two axes. In response to the tensile force applied, the elasticity of the material will tend to bring it to its original shape. This trend can be understood as a force that will shrink the material in the direction of its stretching and will increase it in the other directions. The Poisson ratio is a ratio between the first and second forces [35].

The HR-pQCT images were filtered using a Laplace-Hamming filter and axial boundary conditions were assigned for compression tests with 1% strain (Young’s modulus of 6829 MPa, Poisson’s ratio of 0.3) [36].

Models of the distal radius and tibia were generated directly from the HR-pQCT images (Image Processing Language and FE Extension (IPLFE), Scanco Medical AG, Switzerland). The biomechanical properties studied were stiffness (S, kN/m) and failure load (F.Load, N).

Statistical analysis

Data were expressed as mean and standard deviation (SD), median and interquartile range (IQR), or number and percentage. To calculate the reference curves, multiple linear regression models were developed to predict the measurement of volumetric bone density, morphometry, and bone strength by FEA for distal radius and tibia as functions of age, weight, and height. Thus, the results of the microarchitecture parameters were illustrated using dispersion diagrams with their respective adjusted lines, using the mean of the height and weight measurements to estimate the models, as well as the respective 95% normality intervals. To evaluate the correlation between bone stiffness and other parameters obtained by HR-pQCT, Pearson’s correlations were calculated.

We observed that the relationship with age, in some situations, did not show a linear relationship. Higher degree polynomials were tested and the quadratic relationship was sufficient in most non-linear models as a function of age. In this way, based on statistical analysis, a quadratic relationship was considered sufficient to adjust the models.

Statistical analyses were performed using IBM-SPSS software, version 20.0. P-values <0.05 were statistically significant.

Results

Reference curves according to the aging were developed including 340 men, of which 73% were Caucasian. Demographic, anthropometric, and clinical data are presented in Table 1. HR-pQCT measurements that showed movement artifacts were redone shortly after acquisition; thus, no participant was excluded after the analysis was completed. Representative 3D images of the distal radius and tibia for each age group are shown in Fig. 1.

Table 1 Demographic and anthropometric characteristics, clinical data, and parameters of areal bone mineral density and trabecular bone score obtained by DXA of study population categorized by age
Full size table
Fig. 1
figure 1

3D images of the left distal radius (A) and left distal tibia (B) for each age group

Full size image

Forty-seven (14 %) individuals reported previous fractures, all of them occurring after trauma. In the sites of fractures were radius, ulna, wrist, ankle, tibia, and fibula. No man had had experienced earlier vertebral fracture or fractures due to bone fragility. No comorbidity clinically relevant to the bone was reported by the participants.

DXA and TBS measurements are described in Table 1. All subjects had a BMI in the range of 18–35 kg/m2 [median: 27 kg/m2 (IQR: 25–29)], and no TBS adjustment was necessary. When stratified by TBS categories, 03 individuals aged 30–39 years, 05 aged 40–49 years, 07 aged 50–59 years, 08 aged 60–69 years, 28 aged 70–79 years, and 22 aged ≥ 80 years presented TBS <1.310.

Volumetric bone density, structure, cortical porosity, and bone strength estimated by FEA obtained by HR-pQCT as a function of age, weight, and height are described in Table 2 for the distal radius and tibia.

Table 2 Reference values of bone geometry, volumetric bone density, structure, cortical porosity and bone strength estimated by finite element obtained by HR-pQCT at distal radius and tibia of 340 healthy men from 20 to 92 years old, expressed as median and interquartile interval
Full size table

Multiple linear regression models were developed to predict the volumetric bone density, morphometry parameters, and estimated bone strength by FEA at the distal radius and tibia. These curves are illustrated using scatter plots (Figs. 2 and 3), which showed a linear relationship with age only for thickness measurements (trabecular and cortical) of distal radius and tibia (Fig. 2D, E), and bone strength at distal radius (Fig. 3A, B).

Fig. 2
figure 2figure 2figure 2

Reference curves depending on the age of a volumetric bone density, structure, and cortical porosity parameters, being (A) trabecular volumetric bone density (Tb.vBMD), (B) cortical volumetric bone density (Ct.vBMD), (C) number of trabeculae (Tb.N), (D) trabecular thickness (Tb.Th), (E) cortical thickness (Ct.Th), (F) cortical porosity (Ct.Po) obtained by HR-pQCT in the radius region (left) and tibia (right) in healthy men distributed by age groups. The solid line represents the mean from the regression model, and dashed lines represent the confidence interval of 95% prevision

Full size image
Fig. 3
figure 3

Reference curves depending on the age of the bone strength parameters estimated by finite element, being (A) stiffness (S) and (B) failure load (F.Load) obtained by HR-pQCT in the radius region (left) and tibia (right) in healthy men distributed by age groups. The solid line represents the mean from the regression model, and dashed lines represent the confidence interval of 95% prevision

Full size image

Table 3 shows that, at the distal region of the radius and tibia, all density and morphometry parameters had a statistically significant association with age but not with weight and height in men (p < 0.05). Pearson’s correlations (r) between bone stiffness of peripheral regions with other variables measured by HR-pQCT at distal radius and tibia, areal BMD measures of the L1–L4 and left hip (femoral neck and total hip), and TBS are shown in Table 4.

Table 3 Multiple linear regression analysis to explain the bone microarchitecture parameters obtained by HR-pQCT at distal radius and tibia of 340 healthy men according to the age, weight, and height
Full size table
Table 4 Pearson’s correlation between bone stiffness parameter estimated by finite element analysis with bone parameters obtained by distal radius and tibia HR-pQCT scan, and areal BMD and TBS parameters obtained by DXA in 340 healthy men
Full size table

At the distal radius site, except for some trabecular parameters (Tb.N and Tb.Sp) and for cortical porosity, the other variables showed a statistically significant correlation with stiffness (p <0.05), being positive (r> 0) or negative (r <0) correlations. At the distal tibial site, all parameters were statistically correlated with stiffness (p <0.05).

Discussion

This study provides HR-pQCT reference data for a selected population of healthy men in Brazil and demonstrates that age, weight, and height are associated to changes in bone mass and microstructural properties which behave differently in the cortical and trabecular compartments of both distal radius and tibia.

An interesting feature of the current study is the weight and height adjustments of the bone parameters in addition to age adjustments. Speculatively, significant changes occurring in weight and height with aging may lead to an increased ratio of body fat which may affect bone metabolism and bone loss. Also, the gender restriction (only male population) is clinically useful given that risk factors for osteoporosis and fractures are different in men and women [5, 7, 20].

Concerning to the clinical applicability of HR-pQCT, the implementation of age- and sex-specific reference data in different populations is limited. In the current study, we developed reference values for volumetric bone density and morphometry parameters from HR-pQCT, and bone strength estimated by finite element, at the distal sites of the radius and tibia, categorized by decades, for a healthy male sample representing a main part of the Brazilian miscegenated population.

Different from DXA, to determine T-scores for the HR-pQCT data is difficult because of the great variety of existing parameters, which show distinct behaviors throughout life [8]. In addition, there are differences in the aging bone changes between the regions of radius and tibia [16, 24, 37].

In this HR-pQCT study, almost half of the statistical models had a quadratic relationship with age, and all parameters had a statistically significant association with age in men. Density and trabecular microarchitecture changes in the advanced age might be related to local trabecularization of the cortical bone [38]. The cortical porosity increased with age, mainly at the tibia site. According to Burt et al. [16], in adulthood, cortical porosity increases as bone resorption exceeds formation over the years. It suggests that this variable could be clinically relevant for the stratification of fracture risk and the classification of diseases such as osteoporosis [16].

In FEA, bone stiffness at the tibia had a greater correlation with other HR-pQCT and DXA parameters than at the radius. These results may be related to differences in weight-bearing capacities between the two sites [39].

Furthermore, bone stiffness was correlated to the cortical bone compartment in both analyzed regions. According to the literature, cortical vBMD influences bone capacity to absorb energy during the bone mechanical overload process and is likely to be an important determinant of fracture risk [40]. Moreover, consistent with previous reports, bone stiffness was significantly correlated with areal BMD, especially at the femoral neck [41].

Bone stiffness was also correlated with trabecular volumetric bone density and trabecular structural parameters by HR-pQCT and TBS. Of note, the trabecular compartment is crucial to bone strength. As demonstrated by Burrows et al. [42], the number of trabeculae is significantly higher for men than women in all age groups, suggesting that a larger trabecular structure would increase bone stiffness and consequently improve fracture resistance. In fact, according to a recent study from our group, men had more trabeculae than women in all age groups, as well bone strength parameters estimated by finite element were higher for men compared to women [43].

Thus, a weak to moderate correlation between stiffness and density/structural parameters was detected in both compartments and at both peripheral sites, suggesting that bone strength is dependent on both trabecular and cortical bone. Although the magnitude of the correlation between FEA and most bone microstructural parameters provided by HR-pQCT is low, it helps to determine the contribution of each of these variables to fracture risk prediction, given that FEA provides an estimate of bone strength. Our data are consistent with the literature, which shows that both the cortical and trabecular bones contribute to the predicted failure load [44].

Although approximately 70% of our cohort was considered Caucasian, Brazil has very large with substantial ethnic admixture and distinct lifestyle habits, configuring a miscegenated population [25]. It is different from the Caucasian population of North America and Europe, for example. It is known that there are ethnic differences in the skeletal parameters [45, 46]. Thus, reference data reported in this study may be more adequate for miscegenated populations.

Potential limitations of this study include degenerative changes in the lumbar spine such as osteophytes and aortic calcification that may have overestimated areal BMD (DXA). Besides this, the sample consisted of a healthy population that included individuals with osteopenia and osteoporosis, resulting in a broad spectrum of areal BMD. Moreover, it was not possible to separate participants into racial groups because of the larger number of Caucasian individuals than other races. The oldest participant in our study was 92 years old. Due to the limited number of participants aged 90 years or older in our cohort, there was no statistical ability to generate reference results over this decade.

In conclusion, the current study expands the HR-pQCT worldwide database. As this is a specific selected sample, the results may not be directly extrapolable to other male populations. Anyway, the study brings an adequate methodology for the construction of reference data in other populations. We hope it can stimulate longitudinal research on HR-pQCT.