The studies on TBS in maintenance dialysis (15–26) are listed in descending chronological order in Table 1.

Studies that compared TBS in dialysis and control subjects (15, 21, 22, 25, 26) universally found a statistically significant lower score, hence a degraded microarchitecture, in those on dialysis. Most were case-control studies that controlled for age and sex (21), age, sex, and body mass index (BMI) (15, 25), or age, sex, and lumbar spine BMD (22) for a total of 254 cases. Most cases were hemodialysis patients (n = 236) while only 18 patients were on peritoneal dialysis and 2 were transplant recipients The largest and smallest TBS differences between dialysis patients and controls were encountered in the study of Yavropoulou et al. (25) (1.11 ± 0.16 vs. 1.3 ± 0.13) and Dusceac et al. (22) (0.07 [0.03–0.1] lower in HD) respectively. Only one study (22) reported the TBS T- and Z-scores and found them to be is ~0.8 SD lower in hemodialysis patients.

As in the general population, TBS was associated with numerous demographics or bone-related factors like age, sex, BMI, or BMD. As a result, accounting for as many as possible of these factors is essential when comparing TBS across different groups. Unfortunately, none of the above-mentioned case-control studies matched non-dialysis patients for more than three factors. However, the homogeneity of the results from different cohorts and methodological approaches suggest that TBS is, indeed, significantly lower in dialysis patients.

Studies that compared TBS with international normative found a TBS lower than 1.31 (28) in 34.2% (21), 42% (23) and 35% (24) of cases. Luckman et al. (17), in 47 CKD patients of whom 23 were on dialysis, found a TBS lower than 1.37 in 53% of cases and Brunerova et al. (18) found a TBS lower than 1.23 in 47.5% of cases. We have to note that the normative used in the above-mentioned studies are derived from the general population and there is no consensus on TBS normal/abnormal values in the end-stage renal disease.

Most studies found significant correlations between TBS and demographic, bone- or CKD-related factors. As expected, most studies that reported on the association between TBS and age found a significant negative correlation (16–18, 20, 21). Also, most studies found a positive correlation between TBS and BMD (16–18, 20–22, 24), similar to the general population (29). It is interesting to note that TBS correlated not only to lumbar spine BMD (r between 0.18 and 0.5) (16–18, 20–22, 24) but also with femoral neck (r between 0.25 and 0.4) (17, 18, 20–22) or 1/3 radius BMD (r between 0.17 and 0.38) (20, 22, 24). Dialysis vintage inversely correlated with TBS in one study (21) but not in others (22, 24, 26).

As increased bone turnover is present in a significant number of dialysis patients several studies tried to find a correlation between bone turnover markers and TBS. However, the results are heterogeneous with studies reporting inverse (20, 24), direct (16), or no correlations (18, 20, 21, 26). Specifically, Dusceac et al. (20) found a negative correlation between TBS and serum parathyroid hormone (PTH) or C-terminal cross laps of type 1 collage (β-CTx), Aleksova et al. (24) found a negative correlation between TBS and PTH, procollagen type 1-N Propeptide (P1NP) and alkaline phosphatase (ALP) while Yun et al. (16) reported a positive correlation with PTH. Studies reporting an inverse correlation between TBS and bone turnover markers found correlation coefficients around 0.2.

The most important mechanisms behind low TBS in maintenance dialysis are the same as those in the general population, namely advancing age and low BMD. All studies found robust correlations between TBS and age or BMD and these correlations remained significant in multivariate regression models that included numerous other demographic or clinical factors.

Two interesting studies (17, 23) correlated TBS with structural bone parameters and microarchitecture assessed by high-resolution peripheral computed tomography (HR-pQCT) and histomorphometry from transiliac crest biopsy. Although both studies included also patients with stage 3–4 CKD, most subjects were on permanent dialysis (66 and 48%, respectively) so the conclusions can be extrapolated to this review. Both studies found direct correlations between TBS and parameters of trabecular bone-like trabecular bone volume (BV/TV), trabecular thickness (TbTh), or trabecular width (TbWi), and inverse correlations with trabecular space or trabecular porosity. Moreover, they also found significant associations with the cortical thickness (CtTh), cortical area, or cortical density. Unexpectedly, HR-pQCT findings of increased cortical bone with higher TBS were paralleled by a lower cortical width at histomorphometry. It is worth mentioning that all these findings were heterogeneous between radius HR-pQCT, tibia HR-pQCT, and iliac crest histomorphometry. As expected, both studies reported significant correlations with areal or volumetric BMD.

In a recent study, Yavropoulou et al. (15) evaluated the serum concentrations of bone metabolism-related microRNAs in 30 patients on permanent hemodialysis and found that some of these small molecules were significantly downregulated compared to healthy controls. MicroRNA-23a-3p, which targets the runt-related transcription factor, was specifically inversely correlated with TBS (rho = −0.503) and remained significant after adjusting for BMD.

Also, in a recent study by Yun et al. (16), TBS was prospectively inversely associated with new-onset cardiovascular events (coronary artery disease, stroke, or peripheral arterial occlusive disease), suggesting that profoundly altered mineral metabolism leads to deranged microarchitecture and vascular calcifications.

Further findings showed that three studies (16, 20, 24) tried to find a correlation between the etiology of CKD and TBS or BMD. Aleksova et al. (24) found significantly lower TBS and LS or femoral neck BMD in those undergoing simultaneous pancreatic kidney transplantation (those with type 1 diabetes mellitus). Similar to the general population, Dusceac et al. (20) found a significantly higher LS BMD in patients with diabetic kidney disease vs. other etiologies but no differences in TBS. Lastly, Yun et al. (16) found no difference in TBS in dialysis patients with or without diabetes.

The majority of studies also reported data on fracture and fracture risk (16, 19–21, 23–26). Most of the data come from cross-sectional studies that compared TBS between patients with and without prevalent fractures (21, 23–26). Four of these studies reported similar TBS between dialysis patients with and without fracture (21, 23, 25, 26) while Aleksova et al. (24) found significantly lower TBS in patients with non-vertebral fractures. Although four out of five studies reported negative results (non-significant differences), the TBS was lower, albeit non-significant in patients with prevalent fracture suggesting that lack of statistically significant results might be due to underpowering. Also, in a cross-sectional study, Dusceac et al. (20) found that the risk of fracture calculated by FRAX increased after adjusting for TBS in hemodialysis patients and was significantly higher than in age, sex and BMD matched controls.

Two prospective studies measured TBS at baseline and followed the patients for 20 (16) or 24 (19) months. The results were heterogeneous, with one study (16) showing lower TBS in those with incident fractures (1 vertebral, 1 hip, 2 lower extremities, and 3 upper extremities) while the other (19) found no significant differences (1 vertebral, 2 hip, 2 forearms, 1 humerus, and 1 rib).

The trabecular bone score measures trabecular bone and microarchitecture at the lumbar spine level and was not developed for other skeletal sites. However, hip fracture is the most prevalent type of fracture in maintenance dialysis with rates 1.7 to 98 times higher than in the general population, depending on the age and sex (1). There are concerns that TBS, captured at the lumbar spine level, might not reflect hip microarchitecture and hence, hip fracture risk. Currently, there is no study in maintenance dialysis associating low TBS with a higher incidence or prevalence of hip fracture. However, available studies found robust correlations between lumbar spine measured TBS and hip (17, 18, 20–22) or 1/3 radius (20, 22, 24) measured areal BMD. Also, studies found significant associations between TBS and various parameters of trabecular microarchitecture measured by HR-pQCT at the ultradistal radius or tibia or by histomorphometry at the iliac crest. Taken together, these findings suggest that lumbar spine measured TBS has the ability to capture the widespread skeletal alterations characteristic of CKD-MBD. Moreover, in the general population, TBS was shown to predict hip fracture independently of BMD and age (12).

The trabecular bone score is measured at the lumbar spine level, a region commonly affected by osteoarthritis and, more importantly, by vascular calcification so prevalent in maintenance dialysis. These led to speculations that TBS might not reflect the actual bone microarchitecture in chronic dialysis. Available data show that, at least in the general population, TBS is not affected by spinal osteoarthritis (30). The only study that tackled the relation between aortic calcifications and TBS in patients on permanent dialysis (27) showed an inverse correlation between these parameters, a finding interpreted as TBS and bone microarchitecture are an expression of the deranged mineral metabolism of CKD. In the same study lumbar (27) spine BMD was not associated with aortic calcifications. It is worth noting that in the study of Dusceac et al. (20) the correlations lines between TBS and lumbar spine BMD in patients and BMD-matched controls run parallel, suggesting a lack of significant derangements of aortic calcifications on TBS.

Also, there are some other putative factors like image noise, BMI, or water content of soft tissue that could limit TBS use. Image noise tends to lower TBS but the clinical impact is very low (31). Similarly, although BMI or fluid retention (in relation to dialysis timing) might impact TBS, the clinical relevance is minimal (32, 33). Moreover, the vast majority of dialysis patients included in the cited studies did not have extremely low or high BMI.