SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. References

Despite the excellent performance in clinical practice and research, the dual-energy X-ray absorptiometry is restricted by the inherent planar nature of the measurement and the inability to discriminate between trabecular and cortical components of bone. Recently, a new peripheral tomographic scanner (Norland/Stratec XCT 3000) was introduced for versatile measurements of human long bone characteristics in vivo, including trabecular and cortical density (TrD and CoD, respectively), respective cross-sectional areas (TrA and CoA), bone strength index (BSI), and bone mineral content (BMC). We evaluated the technical performance of the scanner using different phantoms and determined the in vivo precision of the above-noted applications by measuring twice several sites of upper and lower limbs of 19 and 36 volunteers aged 23–60 years. The bone scans were performed, with intermediate positioning of the subject, at two different anatomic sites of the forearm, three sites of the upper arm, three sites of the shank, and two sites of the thigh, with the respective skeletal sites representing different bone compositions and sizes. According to phantom measurements, the XCT 3000 appeared to be a highly linear, stable, and precise (coefficient of variation [CV] about 0.2%) system in vitro. The soft tissue thickness, however, had a linear effect on density values and a nonlinear effect on BMC, whereas the effect on cross-sectional area was marginal. The in vivo root mean square CV (CVrms) values for the long bone ends ranged from 0.9% (distal tibia) to 2.7% (distal femur) for TrD, from 1.8% (distal femur) to 7.6% (distal radius) for TrA, from 2.0% (distal tibia) to 6.8% (proximal tibia) for CoD, from 1.8% (distal femur) to 4.9% (proximal tibia) for CoA, and from 4.2% (distal tibia) to 7.7% (distal radius) for BSI. The corresponding CVrms values for the long bone shafts ranged from 0.5% (midshaft of humerus) to 1.4% (midshaft of fibula) for CoD, from 1.7% (midshaft of tibia) to 4.6% (proximal shaft of humerus) for CoA, and from 2.5% (midshaft of tibia) to 7.5% (proximal shaft of humerus) for BSI. There was no interoperator effect on precision. This study provided, for the first time, independent precision data for the new XCT 3000 peripheral quantitative computed tomography (pQCT) scanner in various applications of human long bones (radius, ulna, humerus, tibia, fibula, and femur) and gave practical guidelines and procedures on how to employ this versatile method in clinical and research applications. The technical performance of the tested system was excellent and it allowed, with a low radiation dose, precise in vivo evaluation of trabecular and cortical density, cross-sectional area, and BMC of selected skeletal sites. The potential effect of the soft tissue thickness on density and mineral content values need to be recognized. The pQCT measurement seems to be useful in supplementing the integral, planar DXA data and obviously opens new possibilities for clinical practice and research.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. References

CURRENT EVALUATION OF MINERAL STATUS of the skeleton is mainly based on planar X-ray absorptiometric techniques, the dual-energy X-ray absorptiometry (DXA) being considered the method of choice for clinical practice and research.1 The DXA technique offers excellent precision in vivo, reasonable accuracy, flexibility in applications, short examination time and low radiation exposure to patients, and the ability to predict the fracture risk (i.e., to estimate bone strength).1–6 Despite the indisputable benefits of DXA, the utility of the DXA-derived areal bone mineral density (BMD) is restricted by the inherent planar nature of the measurement,1 thus rendering a true geometric assessment of a bone impossible and bone strength estimations grossly approximate. Further, a DXA measurement is unable to discriminate between the trabecular and cortical components of bone, which are, however, known to show different responses to aging, various diseases, and treatments.7–13

As a consequence of the above-noted limitations, some treatment effects on bone strength may well remain concealed or be misinterpreted if studied with DXA only.1,4,14 For example, such a situation may occur where a redistribution of bone mineral takes place within the existing structure.15 A standard whole body quantitative computed tomography (QCT) could basically cope with this situation, but given the high costs and technical requirements of the QCT instrumentation and investigation in conjunction with the large radiation exposure to the patient, the method is deemed to be of limited value in clinical bone research, especially if the study includes screening and repeated measurements of apparently healthy individuals.

Recently, special purpose peripheral QCT (pQCT) scanners have been introduced providing the clinical users and investigators some of the benefits of the large clinical QCT systems,1,14,16–21 but at a lower cost and considerably lower radiation exposure. The clinical applications of the first pQCT systems have mostly been limited to forearm measurements, the evaluated sites including the ultradistal and shaft regions of the radius.7–13,17,22 Given the apparently large body of potential applications in all appendicular bones,1,4,7–14,22–24 there is an obvious need for a more flexible and versatile pQCT system than the first forearm pQCTs. To meet this challenge, a new pQCT scanner (Norland/Stratec XCT 3000, (Stratec Medizintechnik GmbH, Pforzheim, Germany) was introduced recently allowing a broader range of applications at human limb bones in general. This study was carried out to evaluate the technical performance of this system, determine the in vivo precision of selected applications at the upper and lower limbs, and propose practical applications for clinical practice and research.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. References

Nineteen volunteers (12 men, 7 women) participated in the upper limb study (group A) and 36 volunteers (15 men, 21 women) in the lower limb study (group B). The mean ages of the study groups were 32 years (SD 8; range 23–47) and 39 years (SD 9; range 24–60); heights 175 cm (SD 8; range 161–186) and 171 cm (SD 8; range 158–186); and weights 72 kg (SD 15; range 53–99) and 68 kg (SD 11; range 52–93), respectively.

A new Norland/Stratec XCT 3000 scanner was used for the pQCT measurements. This scanner employs the similar translation-rotation principle as the manufacturer's forearm scanner (XCT-960)17–21 but is designed to provide a broader range of applications, especially in lower limbs. The XCT 3000 produces a narrow beam by means of an X-ray tube (focal spot size 250 × 250 μm) operating at 60 kV (0.3 mA) and coupled with a filter comprising 6 mm of aluminum and 0.5 mm of copper. The dimensions of the source collimator locating 44 mm above the focal spot are 1.5 mm (height) and 8.5 mm (width). The distance of the detector collimator from the source collimator is 343 mm. The detector system itself consists of 12 cadmium telluride detectors assembled in line, each of them equipped with a separate collimator aperture (4 mm [height] × 0.8 mm [width]). The total width of the detector collimator is 74.5 mm. The diameter of the central gantry opening is 300 mm.

The in vivo pQCT measurements and analyses were performed at distal radius and ulna, radial and ulnar shafts, and humeral shaft (Group A), and at distal tibia and fibula, tibial and fibular shafts, proximal tibia, distal femur, and femoral shaft (Group B). The repeated measurements were done with repositioning of the subjects and without side-to-side comparison to be able to evaluate the true operator effect on measurements. In the group A, both upper limbs (of 12 subjects) or only the right or left limb of the remaining 7 subjects were measured twice within 4 weeks by the same operator (V.K. or A.R.). In addition, 14 limbs were also measured a third time by the second operator, within the above-noted time period. In group B, either the right or left lower limb was measured twice within 4 weeks. Twenty-three of these 36 subjects were measured twice by the same operator, whereas the second measurement of the remaining 13 subjects was made by the other operator.

Scanner performance

The short-term in vitro scanner performance was evaluated by different phantom experiments. First, plastic containers (diameter 50 mm) filled with known concentrations (c) of K2HPO4 solution (i.e., 50, 100, and 250 mg/cm3)15 were measured 15 times consecutively. The linear attenuation coefficient (μ) and density values of the phantom cores, as determined by the analysis software (Version 5.20, Stratec Medizintechnik GmbH), were used for evaluating the short-term in vitro precision and the linearity of the pQCT measurement within the given range, which represents well the trabecular density levels from a severely osteoporotic bone up to a normal level. Second, the effect of soft tissue thickness on bone variables was assessed by measuring the quality assurance (QA) phantom (see below) separately and embedded in concentric, water-filled acrylic cylinders of different size (diameters 84, 100, and 140 mm). The effective soft tissue thicknesses expressed as water equivalents (10 mm of acryl corresponds to about 11 mm of tap water, our unpublished observation) were 26, 73, 100, and 140 mm, respectively. For each thickness, 10 repeated scans were performed at a constant site (30 mm from the endplate) of the phantom using exactly the same analysis procedure as used in actual in vivo measurements of long bone ends (see below). These data were also used for evaluating the short-term in vitro precision.

The long-term performance of the scanner was assessed by daily phantom measurements. The phantom provided by the manufacturer consisted of a polyvinylidenefluoride (PVDF) cone in line with three PVDF cylinders (diameter 24 mm) mimicking different bone compositions (the μ values were 0.442, 0.513, and 0.585 cm−1). The PVDF cylinders were covered by a 12-mm-thick acrylic cylinder, which acted as a soft tissue. The endplate of the cone acted as a reference point to which the distances of subsequent tomographic slices were automatically located. The longitudinal QA data used by us contained the daily μ values and cross-sectional areas (about 496 mm2 in all) of the above four slices, except the cone, for which only the area (about 700 mm2) was used. The μ value of the corresponding cone section was 0.546 cm−1.

pQCT applications in vivo

For the upper limb measurements, the subjects laid on their back on a 40 cm high, slightly cushioned bench, with the abducted upper limb resting inside a 35-cm-long, concentric acrylic cylinder (diameter 15 cm) at the central gantry (Fig. 1A). Before the measurement, the subject was asked to search for a convenient and natural position to avoid movement or discomfort during the scanning period totaling several minutes. An optional weight of 0.5 or 1 kg was kept in hand, and additional supports, whenever necessary, were used to stabilize the extended arm. Special care was taken to keep the supports away from the measurement site.

thumbnail image

Figure FIG. 1. Subject positioning for the upper limb measurements (A) and the lower limb measurements (B).

Download figure to PowerPoint

For the lower limb measurements, the subject sat on the scanner chair with the extended lower limb resting inside the concentric acrylic cylinder (diameter 18 cm) at the central gantry (Fig. 1B). As done before the upper limb measurements, the subject was asked again to search for a convenient and natural position. The weight of the resting lower limb was considered adequate for stable positioning.

Before measuring the tomographic scan, a 30 mm planar scoutview over the joint line of interest was performed at a speed of 30 mm/s and with a scan line spacing of 1 mm in order to define the anatomic reference line, to which the relative location of the subsequent slice was automatically adjusted. The appropriate location of the reference line (see below) was visually verified and manually relocated to meet the specifications of the reference line we routinely employ in our DXA applications.3,4 Identical to scanner performance measurements, single axial slices of 2.5 mm thickness (voxel size 0.5 mm) were then performed at a translational speed of 30 mm/s (a total of 15 translations in 12° steps of rotation).

In the upper limbs (Fig. 2A), the forearm slices were taken at 4% (FA4) and 30% (FA30) of the approximated segment length (0.146 × height) proximal to the distal endplate of ulna. The upper arm slices were taken at 20% (UA20), 50% (UA50), and 80% (UA80) of the approximated segment length (0.186 × height) proximal to the proximal endplate of the radius. In the lower limbs (Fig. 2B), the shank slices were taken at 5% (S5) and 50% (S50) of the approximated segment length (0.246 × height) proximal to the distal endplate of tibia and at 5% (S95) of the approximated segment length distal to the proximal lateral endplate of tibia. The thigh slices were taken at 5% (T5) and 50% (T50) of the approximated segment length (0.245 × height) proximal to the distal lateral endplate of the femur. The above approximations of body segment lengths were based on general anthropometric data.25 The total effective dose of a combination of any scoutview and subsequent tomographic scan remains lower than 0.1 μSv (according to technical inspection by Finnish Center for Nuclear and Radiation Safety, Helsinki, Finland).

thumbnail image

Figure FIG. 2. The anatomic sites (thick lines) of the pQCT measurements in the upper limbs (A) and in the lower limbs (B). The thin lines refer to the reference lines to which the relative location of the tomographic slice was adjusted. The number in the slice name (e.g., FA30) refers to a fraction of the approximated bone length. Further details are given in text.

Download figure to PowerPoint

The total time required for the scoutview and tomographic scan was about 5 minutes for an upper limb scan and about 6 minutes for a lower limb scan. Altogether, a single pQCT investigation including the patient preparation and data analysis took about 10–15 minutes.

For the analyses of the most peripheral sites (FA4 and S5), an iterative contour detection algorithm was used for calculating the cortical and trabecular density (CoD and TrD in mg/cm3) and corresponding cross-sectional areas (CoA and TrA in mm2). In short, the algorithm first determines the attenuation thresholds for cortical and trabecular bone by integrating over the region of interest (defined by the operator) and then searches for a voxel that apparently represents the cortex. The search for the “seed” cortical voxel starts from the midpoint of the region of interest in the upward direction and continues until two neighbored voxels are found. Then the iteration proceeds in a clockwise direction. Similar to a previously described algorithm,26 eight different voxel patterns can describe all the allowed candidates for a new voxel position. By this method, a minimum contour curvature is obtained. The iteration proceeds until the start point is reached again. A 3 × 3 median filter is used to eliminate the image noise and control for the continuity of the detected edge. The edge area corresponds to CoA and the enveloped area to TrA.

For the analysis of the knee region (S95 and T5), a constant threshold of 0.420 cm−1 was used to separate the soft tissue from the outer edge of bone. Out of the detected area, 20% of voxels were concentrically peeled off to define a core area, which was assumed to consist exclusively of the trabecular bone, and from which data the TrD and TrA values were calculated. The CoD and CoA values, in turn, were calculated from the attenuation data of the outer 20% region.

For the analyses of cortical long bone shafts (FA30, UA20, UA50, UA80, fibula from S5, S50, and T50), only the CoD and CoA values were calculated. A constant threshold of 0.690 cm−1 was used to separate the cortical area from the surrounding soft tissue and bone marrow regions.

In addition, a bone strength index (BSI in mm3) was determined for the above-noted skeletal sites, except the bones at the knee region. The BSI is defined as the density weighed polar section modulus of given bone cross-section. At the shaft sites, the threshold of 0.700 cm−1 was used to define the bone tissue assumed to contribute mainly to bone strength. A lower threshold of 0.500 cm−1 was used for the distal sites of radius and tibia. Also, bone mineral content (BMC in mg) was calculated as follows:

BMC = 0.0025 (TrD·TrA + CoD·CoA) (for long bone end sites)

BMC = 0.0025 (CoD·CoA) (for long bone shaft sites)

Data analysis

The mean, SD, and 95% confidence interval are given as descriptive statistics. The standard linear regression analysis was used for evaluating the linearity of the scanner and the long-term variability of the QA data. Pearson correlation coefficients (r) were used for evaluating the strength of association between skeletal variables at different sites. A p-value less than 0.05 was considered significant.

The short-term precision of the duplicate pQCT measurements was defined as the 95% limit of agreement for the differences observed in repeated measurements of given variable (i.e., average bias ± twice the standard deviation [SDmeas] of the differences).27 The 95% limit of agreement provides information on precision in absolute terms and on how large the change at the individual level should be to make one confident that a real change in given variable has occurred. The apparent benefit of this approach is that the variability between repeated measurements is assumed to be constant over the entire range of measurements and thus independent of the mean of the given variable in the studied population. In addition to 95% limits of agreement, two proportionate measures of precision, the average root mean square coefficient of variation (CVrms)28 and the reliability coefficient [R = 100(1 − SDmeas2/SDbiol2) in percentage] were determined. The R-value represents the error-free proportion of the intersubject variability (SDbiol) observed in a given population.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. References

Scanner performance

The short-term CVrms of the μ values as obtained from the K2HPO4 phantoms exhibited virtually a constant value (0.3%), whereas CVrms of the density values ranged from 0.5–1.5% in a clearly size-dependent fashion: the lower the density, the more reduced the precision. The absolute precision error (i.e., the 95% limit of agreement) of the K2HPO4 phantom measurements was virtually constant, being ± 0.002 cm−1 for the μ values and ± 2.7 mg/cm3 for the density. The relationships (Fig. 3) between the known K2HPO4 concentrations and measured μ and density values were both linear (r ≈ 1), the slope of the density equation (i.e., density = 0.963c + 50.2) being close to 1. The regression equation for the μ values was μ = 0.00062c + 0.261.

thumbnail image

Figure FIG. 3. Relationships between the concentration of the K2HPO4 phantoms and the linear attenuation coefficient (A) and density (B) values as determined by pQCT measurements.

Download figure to PowerPoint

The influence of the soft tissue thickness on bone variables measured at a fixed section of the QA phantom is illustrated in Fig. 4. The effect on TrD and CoD was linear (r > 0.99), the slopes being −3.7 and −7.8, respectively. In contrast, the effect on total cross-sectional area and BMC was apparently nonlinear (r > 0.99 for power-law fit), the exponents being −0.013 and −0.096, respectively. While the total cross-sectional area was slightly affected by the soft tissue thickness, the effect on TrA or CoA separately was not systematic (data not shown). The short-term in vitro CVrms values as obtained from the QA phantom were 0.5% for TrD, 0.6% for TrA, 0.8% for CoD, 2.2% for CoA, and 0.5% for BMC. The corresponding 95% limits of agreement were ± 3.1 mg/cm3, ± 4.1 mm2, ± 11.1 g/cm3, ± 5.1 mm2, and ± 4.4 mg. Both area measurements showed a clear reduction in precision at the thickest soft tissue level (1.0% vs. 0.4% for TrA, and 3.6% vs. 1.4% for CoA).

thumbnail image

Figure FIG. 4. The influence of the soft tissue thickness on trabecular density (solid circles) and cortical density (open circles) (A), total cross-sectional area (B), and BMC (C) measured at the fixed section of the quality assurance phantom. The whiskers refer to the 95% limits of agreement of given measurements. The test arrangement is described in detail in the methods section.

Download figure to PowerPoint

The long-term CVrms, as obtained from the QA phantom, varied between 0.1 and 0.2% for both the μ and cross-sectional area values. The slopes calculated from the 9-month QA data were insignificant, the order of their magnitude being <10−5 cm−1/month or <0.1 mm2/month, respectively (Fig. 5). Neither the direction of the slopes showed any systematic trend. The cone area showed a precision of 0.5%, the slope being insignificant and comparable to other QA data of cross-sectional area.

thumbnail image

Figure FIG. 5. Long-term data representing the linear attenuation coefficients (A) and cross-sectional areas (B) of the three slices (see the Methods section) obtained from the daily measurements of the quality assurance phantom.

Download figure to PowerPoint

In vivo pQCT data characteristics

Table 1 shows the descriptive data and detailed precision data of the upper limb pQCT measurements and Table 2 shows the corresponding data of the lower limb measurements. It is noteworthy that the TrD values of distal bone ends showed comparable ranges between different sites and that the CoD values of the long bone shafts were not only similar at different sites but also showed a very small variability (about 3% on average). As expected, BMC, cross-sectional areas, and BSI values of given bones exhibited a large intersite variability. The small-sized distal ulna, however, appeared to be a poor target for pQCT measurements, excluding its TrD measurement. Therefore, only the TrD data of this site were analyzed further in this study.

Table Table 1. Descriptive Data and Precision of pQCT Measurements in Different Sites of Upper Limbs
Thumbnail image of
Table Table 2. Descriptive Data and Precision of pQCT Measurements in Different Sites of Lower Limbs
Thumbnail image of

The TrD values of distal radius and ulna showed significant correlation (r = 0.79). In the lower limb sites, too, the same strength of association was observed, the r-value varying between 0.65 (between distal tibia and proximal femur) and 0.81 (proximal tibia and distal femur). Of note, according to data from 10 subjects who participated in both precision studies, the correlation between TrD of the mutually most distant sites (i.e., distal tibia and distal radius) was 0.78 (p < 0.01). The corresponding regression equation was TrDDistTib = 0.853TrDDistRad + 63.6, which implies quite similar TrD values at these sites.

The CoD values of the upper limbs showed significant correlations excluding the lowest r-value, the r-values varying between 0.29 (between ulnar shaft and midshaft of humerus, NS) and 0.79 (radial shaft and proximal shaft of humerus). In the lower limbs, the corresponding correlations ranged from 0.37 (between the midshafts of tibia and femur) to 0.86 (midshafts of tibia and fibula).

The CoA values of the upper limbs showed highly significant correlations, the r-values varying between 0.82 (between distal radius and ulnar shaft) and 0.96 (distal and midshafts of humerus). In the lower limbs, the corresponding correlations ranged from 0.54 (between the distal tibia and midshaft of fibula) to 0.84 (midshafts of tibia and femur).

The BSI values of the upper limbs showed highly significant correlations, too, the r-values varying between 0.79 (between distal radius and ulnar shaft) and 0.94 (distal and midshafts of humerus). In the lower limbs, the corresponding range was from 0.55 (between the midshafts of fibula and femur) to 0.81 (midshafts of tibia and femur).

The BMC values of the upper limbs showeed highly significant correlations, with the r-values varying between 0.82 (between ulnar shaft and proximal shaft of humerus) and 0.96 (distal and midshaft of humerus). In the lower limbs, the corresponding range was from 0.57 (midshaft of fibula and proximal tibia) to 0.93 (proximal tibia and distal femur).

In vivo precision of the pQCT

In general, the absolute and proportionate precision errors of the upper limb measurements were compatible to those of the lower limb measurements (Tables 1 and 2). There was no apparent size-dependence in the variability between the repeated measurements, and thus the condition regarding the constant 95% limits of agreement over the entire measurement range was met. Also noteworthy, the variability between the repeated measurements was virtually the same (F-tests of variance were insignificant), regardless of whether the measurements were done by the same or a different operator. The gross movement artifacts, instead, observed in about 20 (about 3%) out of the total of 740 pQCT measurements of this study, were the major source of large individual variability. At worst, the movement could result in an error whose magnitude exceeded even the SD of the given variable. There were 10 such cases in the CoD measurements, 1 in the CoA measurements, and 3 in the BSI measurements. Most of these artifacts occurred during the measurement of the most proximal sites (proximal shaft of humerus and midshaft of femur).

In the long bone end measurements (Tables 1 and 2), the CVrms values ranged from 0.9% (distal tibia) to 2.7% (distal femur) for TrD, from 1.8% (distal femur) to 7.6% (distal radius) for TrA, from 2.0% (distal tibia) to 6.8% (proximal tibia) for CoD, from 1.8% (distal femur) to 4.9% (proximal tibia) for CoA, from 4.2% (distal tibia) to 7.7% (distal radius) for BSI, and from 1.2% (distal femur) to 3.5% (proximal tibia). When the intersubject variability was taken into account, the TrD measurement turned out to be the best pQCT application for the long bone ends, the BMC and BSI values being appropriate variables, too.

In the long bone shaft measurements (Tables 1 and 2), the CVrms values ranged from 0.5% (midshaft of humerus) to 1.4% (midshaft of fibula) for CoD, from 1.7% (midshaft of tibia) to 4.6% (proximal shaft of humerus) for CoA, from 2.5% (midshaft of tibia) to 7.5% (proximal shaft of humerus) for BSI, and from 1.3% (distal shaft of humerus) to 4.9% (proximal shaft of humerus). When the intersubject variability was taken into account, the CoD measurement, despite its excellent CVrms values, appeared to be of limited value. In contrast, the BMC, CoA, and BSI measurements, irrespective of the bone size, provided reliable information. The above-mentioned most proximal sites, however, seemed to show reduced precision as compared with more distal sites.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. References

The performance evaluation of this study showed that the XCT 3000 is a highly precise, linear, and stable system for bone measurements, these points thus suggesting high reliability in clinical practice and research including repeated bone investigations. The further in vivo evaluation confirmed this and showed that this pQCT system provides good precision, not only in the lower limb bones, for which applications it was primarily designed, but also in the upper limb bones. However, to obtain reliable results, close attention must be paid on the potential confounding effect of the soft tissue thickness, selecting appropriate anatomic sites and measurement variables, proper site-specific data analysis, and precise and comfortable subject positioning so that possible movement artifacts can be minimized.

The technical performance of XCT 3000 compares well with that of the forearm pQCT scanner21,29 or DXA.1,2,29–33 The linearity is compatible to a clinical QCT within the evaluated range.15 Both the short-term and long-term variabilities or drift were minimal even when compared with those of DXA,29–33 the observations indicating a negligible role of the scanner technique per se as a potential source of imprecision. Given the observed excellent stability and precision in vitro, one could well expect a compatible performance in in vivo applications, too. However, in terms of relative precision, the CVrms values obtained with the pQCT were mostly not as good as those obtained with DXA,1–5,34 but when the variability and magnitude of the measured variables were considered, the superiority of DXA was not so clear any more. As compared with in vivo precision (CV) data of the forearm pQCT scanner1,11–13,17 or clinical QCT,15 the XCT 3000 provided comparable performance.

Although the soft tissue thickness had a clear effect on TrD, CoD, and BMC values, this problem would not compromize the reliability of pQCT measurements under a normal situation. As regards the long bone ends, normal joint regions are surrounded by a relatively thin (∼1–2 cm) soft tissue layer (mostly comprising tendons, ligaments, and skin), whose thickness remains relatively unchanged. In contrast, some abnormalities, like considerable swelling of the joint region due to any reason, may result in artefactual systematic changes in bone variables and thus confound longitudinal observations at the particular site. The nonlinear relationship (Fig. 4C) can even augment the relative adverse effect on BMC at initially thin soft tissue thicknesses. Also the possibility that a calcified tendon or ligament adjacent to bone becomes falsely detected as a part of the bone needs to be recognized. Another issue is the potential effect of soft tissue changes on pQCT measurements of shaft regions of long bones, where the surrounding muscles may become substantially atrophied or hypertrophied, or, e.g., the subcutaneous fat tissue thickness may change as well. However, due to the nonlinear relationship again, the effect on BMC in this case should not be as pronounced as it could be with the initially thin soft tissue layer. The obvious differences in CoD values between the most proximal shaft sites (humerus and femur) and distal sites (e.g., radius and tibia) may, however, partially stem from the soft tissue effect. Therefore, care must be taken when interpreting the differences between subject groups showing clearly different body habitus. Finally, given the observed imprecision, it is emphasized that the cross-sectional areas most likely remain virtually unaffected by the soft tissue changes or intergroup differences.

The gross movements of the subject during the scanning were the major source of imprecision. Fortunately, these events were quite rare and could be best reduced by keeping the total time of the consecutive measurements reasonably short (of note, in some of our subjects the total session of 10 measurements took almost 2 h, which is definitely too long for clinical practice), and apparently by avoiding the most proximal bone sites. It is obvious that a distorted pQCT scan does not convey true information within acceptable limits and needs thus to be omitted or repeated. However, to provide a realistic image of the precision attainable by the XCT 3000, we wished to evaluate the effect caused by the distorted data, too, and none of the measurements was excluded from the precision analyses. This was because in clinical practice it may not always be discernible, especially at the bone ends, whether the particular scan was really affected by a movement or not. According to our experience with about 25,000 DXA scans done at several skeletal sites,3,4 the body movements during the DXA scanning do not pose a real problem in adults or adolescents.

In contrast to observations with DXA,3,4,34 the operator performance seemed not to influence the precision of the pQCT measurements. This was probably due to the fact that both of the operators were intensively involved in developing the present pQCT applications from the very beginning of this project. Also, the deliberate choice not to measure the length of each bone directly but to employ the height-related adjustment for locating the tomographic slice may have contributed to this issue. Another issue is whether the operator's growing experience with pQCT will lead to further improvements in precision, an observation evident from our DXA experience.4 Before this study, the operators had no previous experience with tomographic measurements.

Inconsistencies in the alignment of the target bones with respect to direction of the tomographic slice most likely form the second largest source of imprecision. The subject positioning relied mainly on visual checking of the correct limb alignment, a procedure that was undoubtedly susceptible to some variability. These misalignments regarding the axial direction of bone or a change in the reference line,21 even how small they might be, can result in considerable differences at the physeal sites of long bones, where the cross-sectional geometry changes quite rapidly along the longitudinal axis of given bone. This effect was manifested as a reduced precision of the cross-sectional area measurements of the bone ends. To improve this situation, the current simple fixation tubes need refinement to allow a more consistent positioning, not only within a subject but also between subjects. The relatively straight aspects of the anterior tibia and medial ulna might serve as firm anatomic reference “lines,” against which the perpendicular tomographic scan is done. In the long bone shafts showing only smooth changes in the cross-sectional geometry, such a meticulous alignment of the shaft is probably not so crucial while attaining a good precision at the same.

Not only the above-mentioned issues of subject positioning and associated errors but also the analysis procedures can affect the precision of given pQCT measurements. Usually the TrD values are determined using a concentrical peeling procedure in which a certain area (e.g., 50% or 55% in the forearm applications, or 20% in our knee application) is subtracted off the total bone area to obtain the trabecular core area.7–9,22 However, because of a slightly poorer precision of this procedure (2.5 vs. 2.2%, data not shown in the result section) we felt it more reasonable to employ primarily the automatic contour detection algorithm to determine the TrD values. As could be expected, the respective TrD data correlated significantly (r = 0.93), but surprisingly they did not agree very well. The corresponding 95% limit of agreement (2.0 ± 36.2 g/cm3) was about twice that obtained from repeated measurements of distal radius and showed no bias, i.e., the other method did not systematically over/underestimate the data provided by another but rather induced random variability. This simply means that in prospective studies it is indispensable to use exactly the same analysis procedures as used in previous analyses of the same subject. At the knee region, it was necessary to use the simple peeling procedure because the thin cortices of proximal tibia and distal femur prevented the contour detection algorithm from proper operation in most cases. This, together with the quite complicated cross-sectional shapes, were the reasons the BSI values were not determined for these sites. This apparent lack of performance might have been solved by using a smaller voxel size and a slower scan speed at the same, but at the expense of longer scan times and thus increased possibility for body movement artifacts. Consequently, the present analysis procedures were anticipated to provide appropriate compromises between the reliability and practical efficiency of the XCT 3000. Further development in the scanner itself (e.g., leading to reduced investigation time), measurement procedures (stacks of multiple slices for improved precision, automatic procedures for clinical routine, firm fixation devices, etc.), and data processing and analysis algorithms, however, may well give reasons to modify the present procedures.

What are then the most feasible and reasonable variables the pQCT provides? In terms of in vivo precision and practicality, the TrD of adequately sized long bone ends (i.e., distal radius and tibia, and the knee region) offers a useful variable to be evaluated in any study design or setting. The TrD value corresponds to apparent density of the trabecular architecture of the given site and characterizes the bone tissue as a single lumped measure. Trabecular bone has large area-to-volume ratio, is metabolically active tissue, and exhibits large interindividual variability in density,8 and can therefore show marked changes in various skeletal disorders and considerable responses to various treatments.8–10,12,13 The CoD value, instead, despite the excellent CVrms values, is probably not a good variable to characterize a subject's skeletal status in general. To corroborate this argument, it is known (and as was implied by this study) that the density of a purely cortical bone remains virtually constant (i.e., shows a small variability) irrespective of bone size, age, or even the degree of osteoporosis.8 Further, especially in the case of thin cortices, the partial volume effect inherent to a tomographic measurement can result in underestimated CoD values.35 A useful target group for the CoD measurement, however, could be the patients with hyperparathyroidism who seem to show preferential loss of cortical bone36 and actually reduced CoD as well.8 Finally, in contrast to a somewhat limited value of the area measurements of long bone ends, the cross-sectional areas as well as the BSI values of long bone shafts are good targets of evaluation, however, excluding the measurements at the above-mentioned most proximal sites of long bones. Also, the crude BMC may appear a reasonable measure to describe age- or disease-related effects on bone mass in general.

Is it then necessary to measure more than one skeletal site? As discussed above, the answer apparently depends on the particular clinical or research question. If the treatment or skeletal disorder of interest is expected to have a systemic effect on the skeleton, then the distal tibia (as the most precise trabecular site), possibly complemented by a precise cortical site (tibial midshaft or distal shaft of humerus), would be the most logical choice to receive a reasonable description of the trabecular and cortical bone status. It is noteworthy that the TrD of the weight-bearing distal tibia was similar to that of the non–weight-bearing distal radius. This is condordant with previous findings9 and conforms well with the apparent fact37 that the cross-sectional area plays a major role in making the difference between the bones in terms of the absolute strength and experienced loading environment. However, in adults whose joints and physeal sites are not likely to substantially grow in size,38 the TrD (or BMC) remains virtually the only precisely measurable, adaptive element of the long bone ends. The apparently tight mechanical coupling of bone characterictics was further supported by the observed strong correlations between BMC, CoA, and BSI values of the functionally linked sites—from the most distal site up to the most proximal one in both the upper and lower limbs. In principle, the bone mass, size, and geometry of one site could therefore reflect well that of the other sites representing the same kinetic chain (e.g., tibia-femur), but probably only if the functional environment remained unchanged. It is known that various injuries with associated deloading of bones can result in considerable mineral losses in a site-specific fashion.39,40 Likewise, it is known that certain types of physical loading can lead to sport- and site-specific effects on the skeleton.41,42 Consequently, the pQCT measurements performed at reasonably selected multiple sites seem appropriate. It should be remembered here that one of the objectives of this study was to discuss the options that the XCT 3000 can provide for clinical practice and research.

So, as compared with DXA, what does the pQCT offer that is new for skeletal studies? First, the pQCT investigation can reduce the potential misinterpretations arising from the integrated and planar nature of DXA data. The striking consistency of TrD at the distal end of the tibia and radius demonstrates nicely how the DXA-derived, size-dependent4 BMD can give a false impression about the actual bone density. The DXA-derived BMD at the ankle can be about twice the BMD at the forearm,43 while the difference actually stems from the differences in the absolute bone sizes. As another example, the DXA-derived BMD at the distal femur can be even four times higher than that of the distal radius, while the TrD is actually similar.3,4,41,43 Second, in contrast to DXA, the pQCT allows a direct density measurement of the metabolically active trabecular bone with comparable precision, a fact that may render this measurement more sensitive than DXA to detecting actual treatment effects, or, e.g., the degree of osteoporosis. The third issue preferring the pQCT over DXA is that this method can give an actual description (shape and specific dimensions) of the cross-sectional geometry and bone composition (trabecular to cortical bone ratio) at a variety of skeletal sites in contrast to the planar and integral description of the same sites by DXA. Apparently, the pQCT can enhance the understanding of the relationships among bone density, bone composition, and bone strength with respect to demands of the eventual loading environment and thus help interpret the obtained data appropriately.37,44

Although in some applications the pQCT appears to be a more reasonable method than DXA, this does not lessen the clinical value of DXA. On the contrary, the pQCT can efficiently supplement the information of DXA by providing detailed information about the cross-sectional geometry and trabecular density of many peripheral skeletal sites. At this point, it must be noted that the current version of the XCT 3000, due to apparent technical and subject-related constraints, does not yet allow reliable measurement of the proximal femur in vivo. Instead, currently the DXA remains the only widely used method that can give reasonable, and even sophisticated information45 about the mineral status and estimated strength of this clinically most relevant anatomic site.1

Overall, this study provided, for the first time, independent precision data for the new XCT 3000 pQCT scanner in various applications of human long bones (radius, ulna, humerus, tibia, fibula, and femur) and gave practical guidelines and procedures on how to employ this versatile method in clinical and research applications. The stability and linearity of the tested system were excellent, and it allowed, with a low radiation dose, precise in vivo evaluation of trabecular and cortical density, cross-sectional area, and BMC of the selected skeletal sites. The potential effect of the soft tissue thickness on density and mineral content values need to be recognized. The pQCT measurement seems to be useful in supplementing the integral, planar DXA data, and obviously opens new possibilities for clinical practice and research.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. References

We are indebted to Taru Malminen and Saija Kontulainen for their help in recruiting the volunteers for the study. We also thank Dr. Harri Suominen (Department of Health Sciences, University of Jyväskylä, Finland) and Dr. Teppo Järvinen (Medical School and Institute of Medical Technology, University of Tampere, Finland) for providing us material for the performance testing, and Dr. Georg Tysarczyk-Niemeyer of Stratec for reviewing the technical details of the XCT 3000 system.

References

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. References
  • 1
    Genant HK, Engelke K, Fuerst T, Glüer C-C, Grampp S, Harris ST, Jergas M, Lang T, Lu Y, Majumdar S, Mathur A, Takada M 1996 Noninvasive assessment of bone mineral and structure: State of art J Bone Miner Res 11:707730.
  • 2
    Johnson J, Dawson-Hughes B 1991 Precision and stability of dual energy X-ray absorptiometry measurements Calcif Tissue Int 49:174178.
  • 3
    Sievänen H, Oja P, Vuori I 1992 Precision of dual-energy X-ray absorptiometry in determining bone mineral density and content of various skeletal sites J Nucl Med 33:11371142.
  • 4
    Sievänen H, Kannus P, Nieminen V, Heinonen A, Oja P, Vuori I 1996 Estimation of various mechanical characteristics of human bones using dual energy X-ray absorptiometry: Methodology and precision Bone 18:17S27S.
  • 5
    Beck TJ, Ruff CB, Mourtada FA, Shaffer RA, Maxwell-Williams K, Kao GL, Sartoris DJ, Brodine S 1996 Dual-energy X-ray absorptiometry derived structural geometry for stress fracture prediction in male U.S. Marine Corps recruits J Bone Miner Res 11:645653.
  • 6
    Marshall D, Johnell O, Wedel H 1996 Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures Br Med J 312:12541259.
  • 7
    Rüegsegger P, Durand EP, Dambacher MA 1991 Localization of regional forearm bone loss from high resolution computed tomographic images Osteoporos Int 1:7680.
  • 8
    Rüegsegger P, Durand EP, Dambacher MA 1991 Differential effects of aging and disease on trabecular and compact bone density of the radius Bone 12:99105.
  • 9
    Rüegsegger P, Dambacher MA, Rüegsegger E, Fischer JA, Anliker M 1984 Bone loss in premenopausal and postmenopausal women J Bone Joint Surg 66-A:10151023.
  • 10
    Rüegsegger P, Keller A, Dambacher MA 1995 Comparison of the treatment effects on ossein-hydroxyapatite compound and calcium carbonate in osteoporotic females Osteoporos Int 5:3034.
  • 11
    Rico H, Gonzales-Riola J, Revilla M, Villa LF, Gomez-Castresana F, Escribano J 1994 Cortical versus trabecular bone mass: Influence of activity on both bone components Calcif Tissue Int 54:470472.
  • 12
    Gatti D, Rossini M, Zamberlan N, Braga V, Fracassi E, Adami S 1996 Effect of aging on trabecular and compact bone components of proximal and ultradistal radius Osteoporos Int 6:355360.
  • 13
    Boonen S, Cheng XG, Nijs J, Nicholson PH, Verbeke G, Lesaffre E, Aerssens J, Dequeker J 1997 Factors associated with cortical and trabecular bone loss as quantified by peripheral computed tomography (pQCT) at the ultradistal radius in aging women Calcif Tissue Int 60:164170.
  • 14
    Genant HK, Lang TF, Engelke K, Fuerst T, Glüer CC, Majumdar S, Jergas M 1996 Advances in noninvasive assessment of bone density, quality and structure Calcif Tissue Int 59 (suppl.):S10S15.
  • 15
    Cheng S, Toivanen JA, Suominen H, Toivanen JT, Timonen J 1995 Estimation of structural and geometrical properties of cortical bone by computerized tomography in 78-year-old women J Bone Miner Res 10:139148.
  • 16
    Müller A, Rüegsegger E, Rüegsegger P 1989 Peripheral QCT: A low risk procedure to identify women predisposed to osteoporosis Phys Med Biol 34:741749.
  • 17
    Grampp S, Lang P, Jergas M, Glüer CC, Mathur A, Engelke K, Genant HK 1995 Assessment of skeletal status by peripheral quantitative computed tomography of the forearm: Short-term precision in vivo and comparison to dual X-ray absorptiometry J Bone Miner Res 10:15661576.
  • 18
    Louis O, Boulpaep F, Willnecker J, van den Winkel P, Osteaux M 1995 Cortical mineral content of the radius assessed by peripheral QCT predicts compressive strength on biomechanical testing Bone 16:375379.
  • 19
    Louis O, Willnecker J, Soykens S, van den Winkel P, Osteaux M 1995 Cortical thickness assessed by peripheral quantitative computed tomography: Accuracy evaluated on radius specimens Osteoporos Int 5:446449.
  • 20
    Louis O, Soykens S, Willnecker J, van den Winkel P, Osteaux M 1996 Cortical and bone mineral content of the radius: Accuracy of peripheral computed tomography Bone 18:467472.
  • 21
    Takada M, Engelke K, Hagiwara S, Grampp S, Genant HK 1996 Accuracy and precision study in vitro for peripheral quantitative computed tomography Osteoporos Int 6:207212.
  • 22
    Wapniarz M, Lehmann R, Reincke M, Schonau E, Klein K, Allolio B 1997 Determinants of radial bone density as measured by pQCT in pre- and postmenopausal women: The role of bone size J Bone Miner Res 12:248254.
  • 23
    Augat P, Reeb H, Claes LE 1996 Prediction of fracture load at different skeletal sites by geometric properties of the cortical shell J Bone Miner Res 11:13561363.
  • 24
    Augat P, Merk J, Genant HK, Claes L 1997 Quantitative assessment of experimental fracture repair by peripheral computed tomography Calcif Tissue Int 60:194199.
  • 25
    Drillis R, Contini R 1966 Body segment parameters, Technical Report no. 1166.03. School of Engineering and Science, New York University, New York, NY, U.S.A.
  • 26
    Seitz P, Rüegsegger P 1983 Fast contour detection algorithm for high precision quantitative CT IEEE Trans Med Imaging 2:136141.
  • 27
    Bland JM, Altman DG 1986 Statistical methods assessing agreement between two methods of clinical measurement Lancet i:307310.
  • 28
    Glüer CC, Blake G, Lu Y, Blunt BA, Jergas M, Genant HK 1995 Accurate assessment of precision errors: How to measure the reprodicibility of bone densitometry techniques Osteoporos Int 5:262270.
  • 29
    Wapniarz M, Lehmann R, Randerath O, Baedeker S, John W, Klein K, Allolio B 1994 Precision of dual X-ray absorptiometry and peripheral computed tomography using mobile densitometry units Calcif Tissue Int 54:219223.
  • 30
    Sievänen H, Oja P, Vuori I 1994 Scanner-induced variability and quality assurance in longitudinal dual-energy X-ray absorptiometry measurements Med Phys 21:17951805.
  • 31
    Orwoll ES, Oviatt SK, Biddle JA 1993 Precision of dual-energy X-ray absorptiometry: Development of quality control rules and their application in longitudinal studies J Bone Miner Res 8:693699.
  • 32
    Glüer CC, Faulkner KG, Estilo MJ, Engelke K, Rosin J, Genant HK 1993 Quality assurance for bone densitometry research studies: Concept and impact Osteoporos Int 3:227235.
  • 33
    Wahner HW, Looker A, Dunn W, Walters LC, Hauser MF, Novak C 1994 Quality control of bone densitometry in a national health survey (NHANES III) using three mobile examination centers J Bone Miner Res 6:951960.
  • 34
    Fuleihan GEH, Testa MA, Angell JE, Porrino N, Leboff MS 1995 Reproducibility of DXA absorptiometry: A model for bone loss estimates J Bone Miner Res 10:10041014.
  • 35
    Hangartner TN, Gilsanz V 1996 Evaluation of cortical bone by computed tomography J Bone Miner Res 11:15181525.
  • 36
    Heath H III 1996 Primary hyperparathyroidism, hyperparathyroid bone disease, and osteoporosis. In: MarcusR, FeldmanD, KelseyJ (eds.) Osteoporosis. Academic Press, San Diego, CA, U.S.A., pp. 885897.
  • 37
    Genant HK, Gluer CC, Lotz JC 1994 Gender differences in bone density, skeletal geometry, and fracture biomechanics Radiology 190:636640.
  • 38
    Frost HM 1997 Indirect way to estimate peak joint loads in life and in skeletal remains Anat Rec 248:475483.
  • 39
    Järvinen M, Kannus P 1997 Injury of an extremity as a risk factor for the development of osteoporosis J Bone Joint Surg 79-A:263276.
  • 40
    Sievänen H, Heinonen A, Kannus P 1996 Adaptation of bone to altered loading environment: A biomechanical approach using X-ray absorptiometric data from patella of a young woman Bone 19:5559.
  • 41
    Heinonen A, Oja P, Kannus P, Sievanen H, Haapasalo H, Manttari A, Vuori I 1995 Bone mineral density in female athletes representing sports with different loading characteristics of the skeleton Bone 17:197203.
  • 42
    Haapasalo H, Sievänen H, Kannus P, Heinonen A, Oja P, Vuori I 1996 Dimensions and estimated mechanical characteristics of the humerus after long-term tennis loading J Bone Miner Res 11:864872.
  • 43
    Kannus P, Järvinen M, Sievänen H, Oja P, Vuori I 1994 Osteoporosis in men with a history of tibial fracture J Bone Miner Res 9:423429.
  • 44
    Beck TJ 1996 On measuring bone to predict osteoporotic fracture: Moving beyond statistical inference Radiology 199:612614.
  • 45
    Fourtada FA, Beck TJ, Hauser DL, Ruff CB, Bao G 1996 Curved beam model of the proximal femur for estimating stress using dual-energy X-ray absorptiometry derived structural geometry J Orthop Res 14:483492.