Dr Gordon serves as a consultant for and holds stock in Orthometrix Inc. All other authors state that they have no conflicts of interest.
Cortical PQCT Measures Are Associated With Fractures in Dialysis Patients
Article first published online: 17 JAN 2006
Copyright © 2006 ASBMR
Journal of Bone and Mineral Research
Volume 21, Issue 4, pages 543–548, April 2006
How to Cite
Jamal, S. A., Gilbert, J., Gordon, C. and Bauer, D. C. (2006), Cortical PQCT Measures Are Associated With Fractures in Dialysis Patients. J Bone Miner Res, 21: 543–548. doi: 10.1359/jbmr.060105
- Issue published online: 4 DEC 2009
- Article first published online: 17 JAN 2006
- Manuscript Accepted: 6 JAN 2006
- Manuscript Revised: 27 OCT 2005
- Manuscript Received: 14 SEP 2005
- dialysis patients;
To determine if pQCT could identify HD patients with fractures, we conducted a cross-sectional study in 52 men and women on HD. We found that cortical, but not trabecular, pQCT measures were associated with fractures.
Introduction: Fractures are common in hemodialysis (HD) patients, yet DXA is inconsistently associated with fractures. One explanation for this lack of association may be that HD patients have a selective decrease in cortical density not identified by standard DXA.
Materials and Methods: We used pQCT to examine cross-sectional associations between cortical and trabecular measures and fractures in 36 men and 16 women, ≥50 years of age, on HD for at least 1 year. We confirmed low-trauma nonspine fractures since starting HD. Prevalent vertebral fractures were identified by morphometry of lateral spine X-rays. pQCT measurements of the nondominant radius included trabecular density, cortical density, total area, cortical area, and cortical thickness. We also obtained DXA measurements of the hip and lumbar spine. We used logistic regression models, adjusted for age, weight, and sex, to examine the association between fracture (vertebral and/or self-reported nonspine) and each pQCT measure. Results are reported as ORs per SD decrease in the independent variable.
Results: The mean age was 65.8 ± 9.0 (SD) years, the mean weight was 72.3 ± 15.6 kg, most (32 of 52) subjects were white, and there were 32 fractures in 27 subjects (prevalent vertebral fracture or low-trauma fracture) since starting dialysis. A decrease in cortical density was associated with fractures (OR = 16.67; 95% CI: 2.94–83.33), as was a decrease in cortical area (OR = 3.04; 95% CI: 1.28–7.25) and a decrease in cortical thickness (OR = 3.26; 95% CI: 1.36–7.87). Fractures were not associated with pQCT trabecular density (OR = 1.18; 95% CI: 0.6–2.33), total area (OR = 1.1; 95% CI: 0.59–1.7), or DXA measurements of the hip and spine.
Conclusions: Cortical parameters of the radius were associated with fractures in HD patients. If confirmed in prospective studies, these findings may explain the lack of association between fracture and standard DXA measurements and raise the possibility that pQCT could be used to identify HD patients at high risk of fracture.
Fifty percent of men and women on long-term hemodialysis (HD) will sustain a fracture.(1–7) DXA is inconsistently associated with fractures in this group.(2,8–17) For example, we have previously reported that DXA of the hip and spine, while low in the majority of 104 HD patients, was not associated with fractures.(8)
One reason for the lack of association between DXA and fracture may be the presence of underlying metabolic bone disease, one of the most common being osteitis fibrosa from secondary hyperparathyroidism.(18–22) Hyperparathyroid bone disease has preferential effects on cortical components of bone, and measures such as DXA, which do not distinguish between trabecular and cortical components, may not detect cortical abnormalities in HD patients.(23,24)
In contrast to DXA, pQCT discriminates between trabecular and cortical components, allows assessment of volumetric density, and can be used to derive indices of bone strength. pQCT studies in HD patients consistently show a selective decrease in cortical measures caused by elevated levels of PTH.(25–27) However, the associations between pQCT measures and fractures are not known.
We assessed the relative contributions of pQCT (trabecular and cortical) and DXA measurements to fracture risk among HD patients. We hypothesized that cortical pQCT measures would be associated with fractures but not trabecular pQCT or BMD by DXA.
MATERIALS AND METHODS
We recruited subjects over 5 months from two hemodialysis centers (n = 500 HD patients) in Toronto, Canada. Potential participants were identified by chart review. We included men and women, ≥50 years of age, on HD for at least 1 year (n = 118). We excluded subjects with prior renal transplants, premenopausal women, and women currently taking hormone replacement therapy (HRT). Radiographs, BMD measurements, and pQCT measurements were obtained at a single site. All participants gave written informed consent. The study was approved by the appropriate institutional review boards.
A study investigator (SAJ) and research coordinator reviewed dialysis unit medical records for demographic data, dialysis frequency (3, 4, or ≥5 days/week), etiology of renal failure, medications that may influence cortical and/or trabecular components of bone (nonthiazide diuretics, thyroid hormone, anticonvulsants, prior steroid use, and erythropoietin), the dialysate calcium concentration, and use of calcium salts and vitamin D sterols. We recorded values, obtained within 30 days of study entry, for serum calcium, phosphate, intact PTH, alkaline phosphatase, and thyroid-stimulating hormone (TSH). We also recorded the most recent (within 1 week of study entry) dialysis weight and measured height at the time of DXA measurement using a Harpenden Stadiometer (Holtain, Crymych, UK).
Subjects were interviewed by a research coordinator about age, race, current alcohol consumption, current caffeine consumption, current smoking, physical activity, and family history of osteoporosis (fractures in a first-degree relative). We inquired about fractures, falls in the past 12 months, the presence of diabetes, and prior parathyroidectomy. The questionnaire has been used in several populations, including HD patients.(8)
We inquired about fractures since starting dialysis and classified fractures as low-trauma or traumatic according to the World Health Organization definition.(28) In our study, we included only low-trauma fractures defined as fractures that occur from minimal trauma, such as a fall from a standing height or less, in our analyses. We confirmed all self-reported low-trauma fractures by review of radiographs or radiology reports.
Prevalent vertebral fractures
We obtained antero-posterior and lateral radiographs of the thoracic and lumbar spine and identified fractures by quantitative morphometry. A prevalent fracture was defined as a decrease of at least 20% and 4 mm or greater in any vertical dimension compared with the vertebra above or below one or more thoracic and lumbar vertebrae.(29) Films were interpreted by a single experienced radiologist who was blinded to the clinical history and results of BMD testing.
pQCT measurements were taken of the nondominant forearm using the Stratec XCT 2000 scanner (Stratec Medizintechnik, Pforzheim, Germany). Measurements were obtained at two sites along the radius: 4% of the ulna length and 20% of the ulna length. Each scan was acquired with a 0.25-mm voxel and at a slice thickness of 2.5 mm. The precise position of the two cross-sectional measurements were determined in a scout view using the medial endplate of the radius as an anatomic marker and automatically set by the software at 4% or 20% of ulna length. Image processing and calculation of the various bone indices were performed using the manufacturer software package (version 5.5D). For all distal 4% scans, an iterative contour-detection algorithm was used to separate the radius from the soft tissue background. Once separated from the soft tissue background, total BMD and total bone area were calculated. Trabecular bone in the radius was identified after concentrically peeling off 55% of the bone pixels. Trabecular BMD was measured as the mean density within the remaining 45% of the total area. Cortical bone properties were assessed at the proximal site using the default threshold value of 710 mg/cm3. The following parameters were assessed at the proximal site: volumetric cortical density (mg/cm3), total area (mm2; cross-sectional area of the bone after the soft tissue has been peeled off), cortical area (mm2; the area that is assigned to be purely cortical), and cortical thickness (mm; the difference between the outer and inner radius in the circular ring model).
Geometric measures related to bone strength were also derived from the proximal scan. Specifically, the mean thickness of the cortical shell was calculated using the circular ring model, which has been shown to accurately determine cortical thicknesses in the proximal radius.(30,31) The ability of a bone to resist bending or torsion was assessed using the axial and polar moments of inertia. Both moments were derived from the proximal images using the commercial pQCT software that determines each parameter by examining the spatial distribution of the voxels containing bone relative to the center of mass of the bone.
BMD by DXA
BMD at the lumbar spine (L1–L4) and hip (total hip and femoral neck) was assessed by DXA using a DPX-L absorptiometer (Lunar Corp, Madison, WI, USA). A single experienced technologist, blinded to clinical history and the results of spinal radiographs, performed all DXA measurements (the intraclass correlation coefficient for DXA measurements was 0.98).
We used bivariate analyses (t-test, χ2, or Fisher's exact test, as appropriate) to examine differences in demographic and lifestyle factors, medical history, cause of renal failure, and duration of dialysis by fracture. We considered three classifications for fracture: (1) prevalent vertebral fractures by spinal radiographs, (2) a history of a nonspine fractures since starting dialysis, and (3) either of the first two classifications.
We used logistic regression to examine the relationships between fractures and radiographic measurements: BMD at the lumbar spine (L1–L4), femoral neck, and total hip and pQCT measures (trabecular density, cortical density, cortical area, total area, cortical thickness, bending strength, and twisting strength). We considered each predictor as a continuous variable and adjusted our analyses for age, weight, and sex. Results are expressed as the OR of fracture per SD decrease in the predictor.
To examine the discriminative ability of these tests, we constructed receiver operating characteristic (ROC) curves for each predictor variable. To compare BMD and pQCT measures directly, we compared the area under the ROC curves (AUC) for each pQCT and BMD measurement using a χ2 test.(32,33) Analyses were performed with STATA Version 7.0 (STATA Corp., College Station, TX, USA). Statistical tests were considered significant at a two-tailed level of 0.05 and were not adjusted for multiple comparisons.
Characteristics of study subjects
We identified 118 eligible subjects of whom 66 declined to participate (17 because of difficulty obtaining transportation and 49 because they were “too busy”), leaving 52 subjects. Of the 66 who declined, 61 (92%) allowed us to conduct a chart review to compare demographic characteristics of nonparticipants and participants. Subjects who declined to participate were more often women (35 of 61, 57%) than those who agreed (15 of 52, 29%; p < 0.001), but there were no other significant differences.
Most subjects were physically active men who took calcium and vitamin D supplements. The mean age was 65.8 ± 9.0 years, the median duration of dialysis was 40.2 months (interquartile range, 23.4–61.2 months), and roughly one in three participants had renal failure from diabetes. Mean levels of calcium, phosphate, and alkaline phosphatase were within the normal range. The mean serum PTH, 33.8 pM (range, 12.8–57.2 pM), was above the normal range (1.3–5.4 pM) in all subjects. There were no differences by sex or between subjects with and without fractures (Table 1).
There were 32 fractures in 27 subjects; 12 subjects had prevalent spine fractures; 20 self-reported low-trauma fractures (8 wrist, 7 ankle, 2 rib, 1 hip, 1 humerus, 1 tibia) since starting dialysis and 5 subjects had both a prevalent spine fracture and reported a low-trauma fracture.
Association between pQCT measures, DXA, and fractures
After adjusting for age, weight, and sex, a decrease in cortical density was associated with increased odds of fracture (OR = 16.67; 95% CI: 2.94–83.33), as was a decrease in cortical area (OR = 3.04; 95% CI: 1.28–7.25) and a decrease in cortical thickness (OR= 3.26; 95% CI: 1.36–7.87). Decreased bending strength and decreased torsional strength, derived from proximal pQCT images, were also associated with an increased risk of fracture (OR for bending strength = 5.99; 95% CI: 1.79–20.0; OR for twisting strength = 5.35; 95% CI: 1.72–16.67; Table 2). Results were similar when we considered prevalent vertebral fractures and self-reported fractures separately (data not shown). There was no association between fractures and either trabecular BMD or total area assessed by pQCT (Table 2).
Serum PTH was inversely correlated with cortical density (r = 0.62). More modest inverse correlations were found with serum PTH and cortical area (r = 0.4) and cortical thickness (r = 0.32). Serum PTH was not associated with fractures (OR = 1.01; 95% CI: 0.99–1.02), and adjusting for PTH did not substantially change the associations between any of the pQCT measurements and fracture (data not shown).
As previously reported,(34) we did not find any statistically significant association between fractures (prevalent vertebral and/or self-reported) and decreased BMD at the lumbar spine (OR = 0.53; 95% CI: 0.27–1.06), femoral neck (OR = 0.69; 95% CI: 0.36–1.30) or total hip (OR = 1.10; 95% CI 0.85–2.08; Table 2). Results were similar when we considered prevalent vertebral fractures and self-reported fractures separately (data not shown).
ROC curves confirmed that BMD testing at the lumbar spine (AUC: 0.63; 95% CI: 0.48–0.78), femoral neck (AUC: 0.42; 95% CI: 0.23–0.57), and total hip (AUC: 0.56; 95% CI: 0.30–0.63) did not discriminate among subjects with and without fractures. Similarly, neither trabecular density (AUC: 0.52; 95% CI: 0.34–0.69) nor total area by pQCT (AUC: 0.52; 95% CI: 0.34–0.62) was able to discriminate between subjects by fracture. In contrast, cortical density (AUC: 0.89; 95% CI: 0.90–0.99), cortical thickness (AUC: 0.78; 95% CI: 0.63–0.93), and cortical area (AUC: 73; 95% CI: 0.58–0.89) were able to discriminate between subjects with and without fractures, and the AUC for each for each of these cortical pQCT measures was statistically significantly greater than for any of the BMD measurements (Figs. 1A and 1B).
We found that cortical density, thickness, and area, assessed by pQCT, were strongly associated with fractures among men and women on HD. Similarly, bending strength and torsional strength, which incorporate cortical measures, were associated with fracture. In contrast, trabecular assessments by pQCT and DXA measurements of hip and spine BMD were not associated with fractures.
There are two other published studies of pQCT in HD patients, both of which report a decrease in cortical density, area, and thickness, but no change in trabecular density, compared with healthy controls.(25,26) Our results are consistent with these findings. No previous studies report on the association between pQCT parameters and fractures. The strong association we found between cortical parameters and fracture needs confirmation in prospective studies. Ultimately this research may result in a noninvasive method to identify HD patients at risk for fracture.
The preferential decrease in cortical parameters may explain the inconsistent findings from studies that report on the associations between DXA and fracture in HD patients. In our study, we did not find an association between axial BMD and fractures, but we measured only trabecular sites. Most,(8–14) but not all(2,15–17) other studies, also failed to show an association between BMD at trabecular sites and fractures. Indeed, some data suggest that PTH has an anabolic effect on trabecular bone.(24) However, studies that examined cortical-rich sites showed a consistent association between low BMD and fractures.(9–11,35,36)
The association we observed between cortical parameters and fractures may be mediated by hyperparathyroid bone disease, one of the most common types of metabolic bone disease in HD patients. Histomorphometric studies show that hyperparathyroidism preferentially affects cortical bone, resulting in subperiosteal resorption (reflected by a decrease in cortical area), intracortical porosity (reflected by a decrease in cortical density), and endocortical reabsorption (reflected by a decrease in cortical thickness).(24) In keeping with this hypothesis, we found serum PTH was correlated with cortical measures, but surprisingly, we did not find an association between serum PTH and fractures. This is in contrast to other population-based studies that report a modest association between high levels of PTH (>600 pM) and fracture (caused by hyperparathyroid bone disease)(37) and studies that report an association between low levels of PTH (<3.0 pM) and fracture (presumably caused by adynamic bone disease).(2,18,38) None of the subjects in our study had a serum PTH >600 pM, and serum PTH was <3.0 pM in only two subjects. As such, we were unable to examine the association between fractures and extreme values of PTH. Another potential explanation for why we failed to show an association between high levels of PTH and fractures may have been because we measured intact PTH only on one occasion, after subjects had been on dialysis for at least 1 year. PTH levels were probably higher before starting dialysis, and this is when we may have observed a stronger association between serum PTH and fractures. As well, we had a small number of subjects with relatively few fractures and may have not been able to show an association because of insufficient power. Alternatively, it may be that there are factors, in addition to PTH, that influence cortical bone and fracture risk in HD patients.
Our study has some limitations. Our data were cross-sectional and, as such, we cannot comment on the ability of pQCT or BMD testing by DXA to prospectively identify HD patients who will fracture. We did not measure forearm DXA and cannot comment on the ability of this site, which reflects predominantly cortical bone, to discriminate between subjects with and without fracture. We did not perform bone biopsies, and as such, were unable to directly determine the relationship between serum PTH, cortical parameters, and osteitis fibrosa. Finally, the small number of patients with only a few fractures limits the generalizability of our positive findings and raises the possibility that our negative findings may be caused by lack of statistical power.
In summary, we found that cortical parameters assessed by pQCT were strongly associated with fractures in HD patients. This may be partially explained by underlying hyperparathyroid bone disease. Further studies are needed to determine the relationship between PTH and fracture and to determine the mechanisms by which cortical parameters are decreased.
This work was funded by the Dean's Fund and the Connaught New Staff Matching Grant, both from the University of Toronto.
- 211983 Renal osteodystrophy: Classification, cause and treatment. FrameB, PottsJTJ (eds.) Clinical Disorders of Bone and Mineral Metabolism. Excerpta Medica, Amsterdam, The Netherlands, 254–258., , , ,
- 28World Health Organization 1994 Assessment of Osteoporotic Fracture Risk and Its Role in Screening for Postmenopausal Osteoporosis, World Health Organization, Geneva, Switzerland.
- 342004 Assessing fracture risk in patients with renal failure. J Bone Miner Res 19: S471., , ,
- 362004 Hip arthropathy in long-term (20 years or more) hemodialysis patients. Dial Transplant 33: 486–491., ,