Introduction

Patients with chronic kidney diseases (CKD) stages 3a–5d have higher risk for fractures than the general population, worldwide [1,2,3,4,5]. In addition, incident hip fractures are associated with substantial worsening of morbidity and mortality in dialyzed patients [6,7,8]. However, there is little consensus on the methods for evaluating bone status in CKD patients. To date, no large clinical trials specifically targeting CKD patients for testing the efficacy of treatments currently used for osteoporosis have been performed. Thus, we need to develop and validate diagnosis and treatment strategies for better management of bone fragility in this population. The current evaluation of fracture risk is based on a combination of various clinical factors plus quantitative imaging based on X-Ray attenuation by bone. As it passes through bone, the X-ray beam loses part of its energy, due to photon absorption (removal from the beam) and scattering (change of direction). The amount of attenuation depends on the intensity of the incident X-ray beam and the physical properties of bone (including the amount of minerals and the size of the bone). This allows to image bone and, eventually, after calibration, to deduce quantitative parameters such as bone mineral density (BMD).

Determinants of bone strength

Fractures occur when the load in a region of a bone exceeds the ultimate strength of that bone. The risk of fracture increases with falls and trauma in combination with a decrease in bone strength (illustrated in Fig. 1) [9]. Bone strength results from bone loss (i.e. reduced bone quantity) and/or reduced bone quality [10]. Bone quantity (i.e. bone mass) ensues, in adults, from the bone remodeling balance, which is determined by a number of elements including genomic, hormonal, nutritional and mechanical factors. Bone loss occurs when bone is more resorbed than formed. The level of bone remodeling (turnover) is defined by the ‘activation frequency’ which represents the birth rate of bone remodeling units. Bone quality depends on both bone material and structural properties [11]. Structural parameters correspond to bone micro- and macro-architecture (i.e. geometry and topology), while material properties depend on the quality of bone mineral and matrix. Microarchitecture (e.g. cortical porosity, trabecular number/thickness and spatial distribution) impacts bone strength at both cortical and trabecular levels. Bone mass and some of the structural properties can be evaluated using X-ray based tools. In addition, macro- and micro-architecture parameters, combined or not with mineral density data, are used for modeling biomechanical properties of bone. The spatial distribution of cortical bone is computed via several parameters, which allow to predict cortical bone strength. For instance, the cross-sectional moment of inertia reflects the distribution of the bone mass around a neutral or central axis of the bone and the buckling ratio is the ratio of the outer radius to the cortical thickness [12]. Other engineering methods such as finite element analysis (FEA) allow to test virtual mechanical stress on bones to evaluate bone strength.

Fig. 1
figure 1

Schematic illustration of determinants of fracture

Full size image

Deterioration of bone strength in CKD

CKD affects both structural and material properties of bone [13]. Bone primary and secondary mineralization are substantially impaired by renal insufficiency which leads to a decrease in material properties. Indeed, primary mineralization slows down or even ceases when osteomalacia or uremic mixed lesions occur while secondary mineralization (which takes place after primary mineralization, once the new bone osteons have been formed) can be increased in adynamic osteopathy or decreased in osteitis fibrosa.

Evaluation of bone fragility

The main characteristics of the quantitative X-ray-based devices are summarized in Table 1.

Table 1 Main characteristics of X-ray based tools used for assessment of bone status
Full size table

Standard X-rays and morphological evaluation of bone for fracture detection

Standard skeletal X-rays are useful morphological tools to detect fractures or fissures. This evaluation is guided by clinical symptoms and thorough physical examination. Indeed, while some fractures such as hip fractures are obvious on standard X-ray, others such as tarsal, metatarsal or rib fissures can remain difficult to assess and necessitate further evaluations, including technetium bone scans coupled to computerized tomography or magnetic resonance imaging (MRI) (Fig. 2). As for vertebral fractures, it is well known that they are underdiagnosed [14] and that spine fractures observed on lateral chest X-rays are often overlooked or not reported [15]. Osteopenia (i.e. reduction in bone density on radiographs) can be found on plain X-rays; however, this sign is observed when more than 30–40% of bone has been lost. In CKD, thin cortices or endocortical resorption can be detected, indicating more or less the severity of renal osteodystrophy. It is worth noting that standard X-ray is not sufficient to confirm or exclude the diagnosis of osteoporosis.

High-resolution X-ray images can be used to analyze bone texture, a surrogate of trabecular microarchitecture, via fractal analysis methods. The combination of this technique with BMD measurements has been shown to better predict fractures in osteoporotic women than BMD alone [16]. However, this technique is not widely available and there are no data on CKD patients.

Dual energy X-ray absorptiometry

Bone mineral density

DXA measures areal bone mineral density (g/cm2) at the hip (neck and total regions), the spine (L1/L2 to L5) and the radius (ultradistal, distal and proximal regions). The results are expressed as a T-score, i.e. the number of standard deviations above or below the mean BMD measured in sex-matched 25-year-old healthy subjects (i.e. at the end of peak bone mass acquisition). Densitometric osteoporosis is defined by a T-score lower than −2.5. Spine BMD may be overestimated due to vertebral osteoarthritis or aorta calcifications. Areal BMD is influenced by body size, which is a major limitation for its use in pediatric CKD populations. Epidemiological studies have clearly identified low BMD as a fair predictor of osteoporotic fractures. However, in a non-uremic population, up 40–50% of patients with osteoporotic vertebral fractures have BMD values higher than −2.5 [17], indicating that factors other than BMD must be taken into account when evaluating the fracture risk [18]. For instance, a prevalent fracture is unequivocally a risk for future fracture. Indeed, patients with fractures are at higher risk for future osteoporotic fractures of the spine, wrist and hip compared to patients with similar BMD with no fractures [19] and prior vertebral fractures are a better predictor of future fracture than low BMD alone [20].

Vertebral fracture assessment (VFA)

Two-thirds of patients with vertebral fractures (VF) are asymptomatic; therefore, these fractures are often ignored while it is essential, as seen above, to identify previous vertebral fractures. Lateral imaging of the spine using fan-beam methodology (which eliminates parallax errors in viewing the vertebral body, compared to routine spine X-ray) can be performed on the same device as DXA BMD measurement. These images are then analyzed by the VFA software which allows to identify prevalent vertebral fractures. However, the automatic fracture detection provided by the VFA software needs to be checked by a trained physician [21].

Trabecular bone score (TBS)

TBS is derived from an algorithm that analyzes the spatial organization of pixel intensity, which in turn corresponds to the differences in the X-ray absorption power of an osteoporotic bone versus a normal trabecular pattern [22]. TBS is not a direct measurement of bone microarchitecture but it is related to bone microarchitecture parameters.

Neck geometry and hip structural analysis

Due to a better understanding of bone biomechanics and of the role of geometry in bone strength, a number of techniques has been derived from DXA hip measurements such as hip structural analysis (HSA). The major limitations of these approaches are mainly due to the bi-dimensional nature of DXA measurements. HSA parameters correlate with BMD. Whether they provide additional information independent of BMD and improve fracture prediction is still a matter of controversy [23].

FRAX for evaluating fracture risk at the individual level

We have seen above that BMD measurement is a fair but not perfect predictor of fractures in the general population, in which 1/3 of fractures occur in patients with BMD higher than −2.5 [24]. The FRAX tool combines BMD measurement at the femoral neck and clinical risk factors for fracture such as weight, height, prevalent fracture, family history of hip fracture, and steroid use, but it does not take falls into account. This software, a World Health Organization (WHO) initiative, calculates the 10-year probability of hip fracture and major osteoporotic fractures at the individual level. Major osteoporotic fractures include hip, vertebral, wrist and humeral fractures.

DXA in CKD

At the radius, both cortical and trabecular measurements may be affected by the fistula in CKD-5d patients. Bone mineral content is not influenced by hemodialysis sessions [25]. BMD measured by DXA integrates bone quality and quantity properties but BMD does not provide any information on the underlying renal osteodystrophy. BMD measurement by DXA was not recommended by the 2009 Kidney Disease Improving Global Outcomes (KDIGO) guidelines, but only cross-sectional studies comparing BMD in CKD patients with and without a prevalent fracture were available at this time [26]. In 2015, a meta-analysis gathering data from 13 studies including 1782 patients at stages CKD 3–5d showed that BMD was significantly lower at the femoral neck, lumbar spine, the 1/3 and ultradistal radius in patients with fractures compared to those without, regardless of dialysis status [27]. These data, together with longitudinal studies, led the KDIGO working group to revise the guidelines [28] and recommend, in 2017 [29], BMD measurements in CKD patients.

In the Health, Aging and Body Composition Study, 2754 participants (70–79 years of age at enrolment), including 587 CKD patients (21%) were measured at baseline and followed for 11 years [30]. The CKD population consisted of 83% CKD stage 3a and 13% CKD 3b. The adjusted fracture hazard ratio (HR) for each standard deviation (SD) of lower femoral neck BMD was 2.69 [95% confidence interval (CI) 0.99–3.77] in CKD and 2.15 (1.80–2.57) in the non-uremic population. For hip fractures only, the adjusted femoral neck BMD hazard ratio was 5.82 (3.27–10.35) in CKD and 3.08 (2.29–4.14) in the non-uremic subjects. Thus, lower femoral neck BMD was associated with greater fracture risk regardless of renal function.

A second prospective cohort of 131 predialysis patients (mean age 62 years) was measured with DXA BMD at the total hip, lumbar spine, and ultradistal and 1/3 radius at baseline and after 2 years [31]. Most of the patients had type 2 diabetes. They included, at baseline, 34% CKD stage 2, 40% CKD stage 4 and 26% CKD stage 5. Low BMD at all sites and a greater annualized % decrease in BMD predicted fracture. Multivariate models showed that the odds ratio (OR) of fracture was 1.75 (1.30–2.20) for each SD of lower total hip BMD. Interestingly, in this study, HRpQCT was not better at predicting fracture than BMD measured by DXA.

Finally, a Japanese study measured 485 hemodialyzed (HD) patients (mean age 60 years) at baseline and 40 months later. Forty-six fractures occurred during the follow-up period. The adjusted fracture HR was 0.65 (0.47–0.90) for each SD of higher femoral neck BMD (p = 0.009). Receiver operating characteristic (ROC) analyses stratified according to parathyroid hormone (PTH) below or above the median value of 204 pg/ml (21.6 pmol/l), showed that the area under the curve (AUC) for femoral neck BMD was 0.72 in the lower stratum and 0.51 in the higher stratum [32].

Thus, while BMD measurement by DXA predicts fractures in CKD patients with bone and mineral disorders, a number of issues remain to be discussed. The community of nephrologists must be aware that BMD measurement by DXA has high specificity and moderate sensitivity. In addition, the FRAX software does not take into account CKD as an independent risk for fracture. Thus, it is likely that the FRAX score calculation underestimates fracture risk in this population. The follow-up rate needs to be outlined, knowing that the International Society for Clinical Densitometry recommends calculating “the least significant change” (LSC), i.e. the least amount of BMD change between two measurements that can be considered significant at the individual level, for a 95% confidence level (=precision error × 2.77). Precision characterizes the reproducibility of the measurement and depends on the operator, the DXA device and the population. Thus, it should be checked whether the LSC generally used in non-uremic populations (0.030 g/cm2 at the hip) is similar in CKD patients. The type of CKD patients who will benefit from DXA measurement remain to be defined, knowing that the current drugs for osteoporosis treatment are not recommended in CKD 4–5 patients. Finally, the reimbursement of this test in the CKD population needs to be addressed by Health Authorities in each country.

As for the techniques derived from DXA measurements such as TBS, scant data are available and it is difficult to draw conclusions about their utility in routine practice. In addition, aorta calcifications may alter TBS calculation in the CKD population. In a cohort of 1426 participants (aged ≥40 years, mean age 67 years) including 199 patients with glomerular filtration rate (GFR) below 60 ml/min/1.73 m2 (72.4% CKD stage 3a, 25.1% CKD stage 3b, and 2.5% CKD stage 4) lower lumbar spine TBS was independently associated with a higher fracture risk in adults with reduced kidney function [33]. In a smaller population of 53 CKD-5d patients, analyzed at the time of kidney transplantation, spinal X-ray detected prevalent asymptomatic fractures in 26% of patients. TBS was 8% lower in CKD patients than in controls; however, TBS was similar in CKD patients with or without fractures [34].

Quantitative computed tomography

Central quantitative computerized tomography

This technique, which uses standard computerized tomography, provides volumetric BMD (g/cm3), after calibration with a phantom, as well as macro-geometry parameters at the level of the hip and the spine, the bone sites prone to osteoporotic fractures. It is not influenced by osteoarthritis and its measurements of bone geometry are true 3D parameters, while DXA-derived evaluation is an extrapolation from 2D parameters [35]. Its drawbacks are the higher radiation doses and greater costs for routine diagnosis.

Peripheral quantitative computed tomography

In addition to volumetric BMD, macro- and micro-architecture parameters can be evaluated with an accuracy that depends on the spatial resolution of the device (Table 1). Interestingly, while DXA use is limited in children with CKD due to the confounding effect of smaller body size and opposing PTH effects on the trabecular and cortical envelopes, peripheral quantitative computed tomography does not suffer from these limitations [36]. Peripheral as well as high resolution QCT are not used in routine practice due to the lack of device availability.

Peripheral QCT (pQCT)

pQCT analyses the trabecular and cortical compartments separately at the tibia and radius (resolution 400–500 µm). Recently, clinical cone beam computed tomography improved this technique and its resolution (220 µm). It allows to measure large portions of distal bones with fair spatial resolution and limited irradiation [37]. In a cross-sectional study of 52 CKD 5d patients, including 27 patients with fractures, pQCT analysis showed that a decrease in cortical density, area and thickness was associated with fractures—with the OR varying from 3 to 16—while DXA BMD was not [38].

High-resolution peripheral QCT (HR-pQCT)

HR-pQCT measures volumetric BMD at the distal tibia and radius. Acquisition time is 15 min per site. This device also provides information on trabecular and cortical microarchitecture parameters such as trabecular thickness, number and distribution and cortical porosity (Fig. 3a–c). The XtremeCT II (Scanco®, Brüttisellen, Switzerland) has an 82 µm spatial resolution. Longitudinal follow-up is possible due to specific software. This technique allows to compare two images of the same bone slice and enables to describe where formation and resorption took place during the observation period [39].

HR-pQCT analysis helps us to better understand the mechanisms of bone loss in CKD [40]. Nickolas at al, studied 53 CKD patients (including ten hemodialysis patients) with HR-pQCT and DXA at baseline and after 1.5 years [41]. They found a significant decrease in DXA BMD at the total hip and ultradistal radius. Cortical area, density, and thickness at the distal radius were reduced significantly while cortical porosity increased. Most interestingly, time-averaged levels of PTH and bone turnover markers predicted cortical deterioration. Thus, for the first time, the relationship between the severity of secondary hyperparathyroidism and cortical bone loss was evidenced, thanks to these longitudinal data. The same team demonstrated the persistent bone loss at the peripheral skeleton despite corticosteroid withdrawal in kidney transplant patients [42]. Thus, cortical porosity and cortical thickness are critically affected by CKD [43]. Cortical architecture depends mostly on modeling levels at the periosteal surface and remodeling at the endosteal surface and within the cortex. The endocortical (inner) third of cortices is the most active surface, where trabecular bone is formed at the expense of the cortex, a process highly deleterious for bone strength (Fig. 2a, b). Recent studies confirmed that cortical porosity is highly heterogeneous and demonstrated that thorough analysis of this heterogeneity, using HR-pQCT, would improve our knowledge of how cortical bone can deteriorate rather quickly [44] in the osteoporotic process associated with CKD [43]. Finally, HR-pCQT has been used to analyze the complex relationships between bone structure and vascular calcifications in the general population [45] and in CKD [46].

Fig. 2
figure 2

a, b MRI T1-weighted imaging of the foot in a CKD-5D 75-year-old patient referred for “subacute arthritis of the foot”. Arrows fracture localized at the proximal metaphysis of the first metatarsal bone (a) and anterior part of the calcaneus (b). c Standard X ray: The calcaneus fracture is barely visible

Full size image

Very high-resolution QCT, nano-CT and synchrotron radiation CT

It is possible, for research purposes, to analyze ex vivo bone biopsies at higher resolutions (10 µm to 10 nm) using nano quantitative CT or synchrotron radiation CT [47] (Fig. 3d, e). Synchrotron radiation provides a high-energy monochromatic X-ray beam, which yields high-quality images. These approaches allow to examine micro and nano structures of bone including collagen and mineral properties such as the degree of mineralization of bone (DMB), a strong determinant of bone strength [48] as well as osteocytes lacunae and their canaliculi network [49] (Fig. 3).

Fig. 3
figure 3

a–c HRpQCT images of tibia (a, c) and radius (b) in a CKD patient a, b with increased cortical porosity at the radius level compared to healthy control (c). d, e Synchrotron radiation computerized tomography images of iliac crest bone biopsies from a CKD-5d (d) and non-uremic osteoporotic patient (e). Note in a and d the increase in cortical porosity (arrow) and in d the deep erosion lacunae (dotted arrow), the thin cortices (arrow heads) and the increase in cortical porosity at the endocortical surface (arrows)

Full size image

Magnetic resonance imaging (MRI)

MRI can distinguish microarchitecture deterioration in patients with various metabolic bone diseases, compared to controls, as shown by a number of cross-sectional studies with small sample sizes (summarized in [50]). In the trabecular compartment, MRI images the marrow content since bone signal is hypointense. The voxel sizes range from 130 to 250 µm with a slice thickness of 400–1500 µm (for a scan time of 10–15 min). Bone can be analyzed at any site, including spine and hip, and there is no patient irradiation. Using high to ultrahigh field scanners (3-7T) and specific sequences may improve image and quantification accuracy. Fifty CKD patients were analyzed with 1.5T MRI early and 6 months after kidney transplantation. All patients received glucocorticoids. While vertebral BMD decreased by 3% during follow-up, trabecular microarchitecture parameters did not change significantly. In contrast, FEA analyses of bone strength such as cortical and trabecular stiffness and failure strength were significantly reduced overtime [51]. Recently, ultrashort echo time MRI has made it possible to analyze cortical bone. This technique is based on the measurement of concentrations of bone water (BW) at two levels (bound to collagen and within the porosities). Techawiboonwong et al. reported that CKD patients with renal osteodystrophy had higher BW than premenopausal and postmenopausal controls (by 135 and 43%, respectively) while no difference in volumetric BMD between CKD patients and controls was observed. Taken together, these preliminary data suggest that MRI could yield additional information beyond BMD, to better assess bone fragility [52].

Conclusions

Fractures have become an increasing concern in the CKD population. HRpQCT remains the research tool that helps us to better understand the mechanisms of increased bone fragility. DXA BMD is available worldwide and is a fair predictor of fractures with a good specificity but insufficient sensitivity. Now, DXA measurements are recommended for assessment of bone status in the CKD population—however a number of questions remains regarding the modalities of these measurements and the therapeutic strategies to implement when high risk for fracture is detected in patients with late-stage CKD.