- Email: [email protected]
Hip Fracture Discrimination Study
Journal of Clinical Densitometry, vol. 6, no. 2, 163–172, 2003 © Copyright 2003 by Humana Press Inc. All rights of any nature whatsoever reserved. 1094-6950/03/6:163–172/$20.00
Original Article
Hip Fracture Discrimination Study QUS of the Radius and the Calcaneum
Didier Hans,*,1 Laurence Genton,2 Samia Allaoua,1 Claude Pichard,2 and Daniel O. Slosman1 1Nuclear
Medicine, Geneva University Hospital, Geneva, CH; and 2Nutrition, Geneva University Hospital, Geneva, CH
Abstract As an emerging alternative to current radiation-based bone densitometry techniques, there is a growing interest in the use of quantitative ultrasound (QUS) measurements for the noninvasive assessment of fracture risk in the management of osteoporosis. However, there are also a multiplicity of technologically different QUS devices available on the market and, so far, no study has compared heel and radius QUS device for the discrimination of subjects with hip fractures. Our study evaluated the ability of three QUS devices (one calcaneal gel-coupled system, one calcaneal water-coupled system, and one radius system) to discriminate osteoporotic from controls subjects, using the same population. We also checked if the combination of two different skeletal sites (i.e., calcaneum and radius) improve the discriminatory ability of the QUS devices. Forty-five women aged 79.1 ± 7.1 yr with hip fractures within the last 4 d were used as the hip-fracture group and compared to 40 healthy controls from 65–87 yr. In addition, 47 young controls, aged 20–40 yr, were used as reference population to express some of the results as T-scores. QUS measurements were performed with the Hologic Sahara, Ge-Lunar Achilles+, and Sunlight Omnisense devices according to the manufacturer’s recommendations. Adjusted odds ratio results showed that a decrease in Omnisense SOS of 1 standard deviation (SD) was associated with a significant increase in fracture risk (OR adj. = 2.83) comparable with Sahara BUA (OR adj. = 2.42) and Achilles BUA (OR adj. = 3.29). However, given the large overlap between the 95% intervals of each odds ratio, no significant difference was found between the devices. Similarly, comparison between the areas under ROC curves did not show any significant difference between all the parameters. Considering the parameters provided per default by each QUS device, the Sahara, Achilles, and Omnisense devices classified correctly 70, 67.5, and 62.5% of the subjects, respectively. Although the OR of the combination of radius and calcaneum is improved (3.62 to 4.74) compared with either one of the single skeletal site, the large confidence intervals do not allow to claim a significant difference. The three QUS technologies tested against hip fractures seem to show the same discriminatory ability. However, there is some differences in the definition of the diagnostic threshold underlying the nonusability of the World Health Organization (WHO) osteoporosis definition. Finally, the combination of several site using two different devices is not clinically relevant. Potential interest could be seen in combining several sites using the same device, supposing that such device would measure differently active bone. Key Words: Hip fractures; QUS; SOS; BUA; radius; calcaneum. Received 03/06/02; Accepted 10/08/02. *Address correspondence to Dr. D. Hans, Head of R&D, Division of Nuclear Medicine, Geneva University Hospital, 1211 Geneva 14, Switzerland. E-mail: [email protected]
163
164
Introduction Osteoporosis represents a major worldwide public-health problem that will grow in importance in the coming decades as the population ages. Because of the importance of bone mineral density (BMD) in determining bone strength, many noninvasive techniques based on the attenuation of ionizing radiation have been developed to quantify BMD in the axial and peripheral skeleton (1). These include Single Photon Absorptiometry (SPA), Single X-Ray Absorptiometry (SXA), Dual X-Ray Absorptiometry (DXA), Quantitative Computed Tomography (QCT), and peripheral QCT (pQCT). The decision about which method to use is often based on cost and availability rather than technical considerations like predictive value, precision, or radiation dose. DXA is the most widely used method of BMD measurement, largely because of its low radiation exposure, high precision, and ability to assess several different sites in the skeleton. However, it is a relatively expensive approach that gives limited information on bone structure (2). As a consequence, accessibility to such technology is limited for many doctors, and subsequently to many women. Indeed, according to the International Osteoporosis Foundation report, 83% of women in Europe have never been screened for the potential risk of osteoporosis. Such forecasts have lead to the search for new, cost-effective methods for early detection, prevention, and treatment, such as peripheral systems (pDXA, PQCT, and so on). As an emerging alternative to current radiationbased bone-densitometry techniques, there is a growing interest in the use of quantitative ultrasound (QUS) measurements for the noninvasive assessment of fracture risk in the management of osteoporosis (2). Quantitative ultrasound is a simple, inexpensive, and relatively portable technique that is noninvasive and ionizing radiation-free. As such, it has a much greater potential for widespread application than standard bone densitometry. Several studies have demonstrated the ability of QUS to discriminate patients with osteoporotic fractures from age-matched controls (3–5) and more importantly to predict fracture risk (6–9). Therefore, QUS may be used as a screening method for osteoporosis (10) or as a diagnostic tool if clinically relevant threshold are defined. Though based on similar principles, QUS instruments from different manufacturers have significant Journal of Clinical Densitometry
Hans et al. differences, particularly in their calibration methods, sites of measurement, acquisitions, analysis software, and scanner designs (11). Moreover, it is difficult to assess whether all QUS devices have similar discriminatory ability because cross-sectional odd ratios are affected by the study populations. As a result, the Receiver Operating Characteristic Curve (ROC) as well as the odds ratio may vary largely between devices of different brands. As a result, results obtained with one device cannot be compared with another that is technologically different. Therefore, QUS devices ideally should be compared on the same population. A few studies have reported such results (5) for the calcaneum, but none of them compared ultrasound measurement at different sites except Hartl et al. (12). In their study, they tested the ability of heel vs phalangeal ultrasound devices in discriminating vertebral fracture. They concluded that QUS measurements of the proximal phalanges performed as well as QUS measurements of the calcaneum for the discrimination of subjects with multiple vertebral fractures. The combination of DXA and QUS has been studied several times, primarily to find if those two technologies would have an improving effect in diagnosing patient at risk of osteoporosis. Conflicting results do not seem to support the hypothesis. Similarly, some studies investigated if the combination of several skeletal sites assessed by ultrasound only would be beneficial as well (3,19). However, the combination of the calcaneum using a dedicated technology with the radius has never been investigated. An added effect of two skeletal sites would likely decrease the overlap between fractured and nonfractured patients, and could also be used to select a subpopulation at very high risk of osteoporosis. To our knowledge, no study has compared heel and radius QUS device for the discrimination of subjects with hip fractures. As a result, the first aim of our study was to evaluate the ability of three QUS devices (one calcaneal gel-coupled system, one calcaneal water-coupled system, and one radius system) to discriminate osteoporotic from controls subjects, using the same population. The second and third aims were to define the optimum discrimination threshold for each ultrasound parameter and to check how the combination of two different skeletal sites (i.e., calcaneum and radius) affects the discriminatory ability of the different QUS devices. Volume 6, 2003
Hip Fracture Discrimination Study
Materials and Methods Subjects The Geneva University Hospital enrolled 123 subjects. Thirty-eight women aged 80.0 ± 6.1 yr with hip fractures within the last 4 d were used as the hip-fracture group. Subjects were included in the fracture group only if the fracture resulted from a fall of standing height or less and involved no motion greater than walking. These fractures were defined as low-energy fractures and have been called “osteoporotic fracture.” They were compared to 38 nonfractured controls from 66–87 yr. In addition, 47 young controls, aged 20–40 yr, were used as reference population to express some of the results as T-scores. All participants underwent a questionnaire about their medical history. Women with hormone replacement therapy (HRT) use for more than 6 mo within the previous 10 yr, Paget’s disease, juvenile diabetes, renal failure, or malignant disease with metastatic tumor were excluded from the study. We also excluded women who took SERM, calcitonin, bisphosphonates, anabolic steroids, fluoride, or parathyroid hormone for more than 3 mo within the previous 3 yr or any other drug known to affect bone metabolism. Smoking was not an exclusion criteria. For each subject, height, weight, and body mass index (BMI = weight [kg]/height2 [m]) were obtained.
Ultrasound Devices and Measurements Measurements were acquired using three ultrasound devices, the Achilles+ (GE-Lunar Corporation, USA), the Sahara (Hologic Inc., USA) and the Sunlight Omnisense™ (Sunlight Medical Ltd., Israel). The first two systems measure ultrasounds at the calcaneum. The Achilles uses water for coupling and its transducers are placed at fixed distance from each other, whereas the second QUS device uses ultrasonic gel and its transducers are placed in contact with the calcaneum. The particularity of the third device is its measured skeletal site (the radius) and its mode of ultrasound propagation (axial). All systems measure speed of sound (SOSm/s). In addition the Achilles+ and the Sahara devices carry out the broadband ultrasound attenuation (BUA-dB/MHz) parameter and a combined index called Stiffness and QUI, respectively. The technical details and the precision of these devices Journal of Clinical Densitometry
165 have been described by Njeh et al. (13), Cheng et al. (14), and Hans et al. (15). One trained operator performed all the measurements according to the recommendations provided by the device manufacturer. Each subject underwent one single QUS measurement per device. QUS measurements were carried out on the left calcaneum (and radius), except for the fracture group where the measurements were performed at the heel of the nonfractured leg. Daily quality control (QC) was performed for all the ultrasound devices using acoustic phantoms provided by the manufacturers. No drift or shift was observed during the study with the exception of the Sahara device, for which the coupling pads were changed once.
Statistical Analysis Before any analysis, all values of each group were tested for normal distribution. All results were expressed as mean and standard deviation (SD). For all devices, the differences between QUS measurements of the fracture and nonfracture groups were compared by a Student’s t-test. Only p-values < 0.05 were considered as significant. Risk of hip fracture associated with low ultrasound measurements was estimated with logistic regression analysis. The presence of fracture was used as the binary response variable. Results were expressed as age- and weightadjusted odd ratios per SD decrease with 95% confidence intervals. A receiver operating curve (ROC) analysis was also conducted to assess the performance of bone-measurement parameters for the classification of subjects with and without hip fractures by calculating the respective area under the curve (AUC). Each ROC curve was compared using a bivariate Chi-square test. Finally, a discriminant analysis between fracture and nonfracture groups was performed to test the ability of the different ultrasound variables to classify study subjects according to their hip-fracture status. Resulting values in terms of sensitivity and specificity were used to identify the corresponding threshold value from the ROC. Combinations of two skeletal sites (radius and calcaneum) were performed for the QUI, Stiffness, and SOS (Omnisense only for the later) parameters using the average T-score. Similar statistics than the one used for a single skeletal site were performed. Data analysis was performed with the software Statistical Package for Social Sciences, Volume 6, 2003
166 Version 9.0.0. for IBM PC (SPSS, Inc., Chicago, IL) and Excel 2000 (Microsoft, Seattle, WA) program.
Results Demographics All ultrasound and anthropometric parameters showed a normal distribution and the characteristics of the three study groups are presented in Table 1. The age was significantly higher for the fractured subjects than the controls, whereas their weight and BMI were significantly lower. All bone parameters were significantly lower in the fracture group (see Table 1). As a consequence, results of the logistic regression were adjusted for age and weight. BUA, SOS, and the combined index (Stiffness and QUI) of the fractured subjects, expressed in Zscores and in t-scores using the results of healthy aged and young controls, respectively, are shown in Table 2. The Omnisense device showed the lowest T-score values (–3.0) compared to both Sahara and Lunar (–2.6).
Diagnostic Ability The ability of the three QUS devices to discriminate between control and hip-fractured patients is presented in Table 3, and in Figs. 1 and 2, and was expressed as both odd ratios and area under the ROC curve. The significancy of nonadjusted and adjusted logistic regression coefficients associated with each bone parameter (p < 0.05) demonstrated that all the ultrasound parameters were predictors of fracture risk. Adjusted odd ratios, which indicate the increase of fracture risk for 1 SD decrease in bone-measurement parameter, were all significantly higher than 1. Adjusted odds ratio results showed that a decrease in Omnisense SOS of 1 SD was associated with a significant increase in fracture risk (OR adj. = 2.7) but apparently lower than Sahara BUA (OR adj. = 4.8) and Achilles BUA (OR adj. = 3.5). However, given the large overlap between the 95% intervals of each odds ratio, no significant difference was found between the devices. Comparison between the areas under ROC curves did not show any significant difference between all the parameters in their ability to discriminate between fractured subjects and controls
Journal of Clinical Densitometry
Hans et al. for both calcaneum devices. In the upper left quarter of Fig. 2, which corresponds to the sensitivity and specificity of interest (both between 0.5 and 1), the curves are very close to each other. However, the radius SOS was significantly lower than any calcaneum ultrasound parameters measured with the Achilles and Sahara. This difference is likely owing to the age and weight differences between groups, which seems to affect more the Omnisense device than the two other calcaneal ones (see the results expressed as T-scores and the raw OR in Table 2 and 3, respectively).
Definition of Thresholds A discriminant analysis was applied for subjects with hip fractures and controls. Results in terms of sensitivity and specificity for the assessment of hip fractures for the different bone parameters are shown in Table 4 for QUS. The highest sensitivity (76.3%) and specificity (76.3) for the detection of hip fracture was seen for the calcaneal BUA values measured with the Sahara device. If we consider the parameters provided per default by each QUS device, the Sahara, Achilles, and Omnisense classified correctly 76.3, 69.7, and 62.0% of the subjects, respectively. With the sensitivity and specificity calculated in the discriminant analysis, it was possible to identify the corresponding optimum threshold values for the different bone parameters (from the ROC analysis), as shown in tables 4. The lowest T-scores for hip fracture discrimination were recorded with the Omnisense SOS (–2.5), which was matching the World Health Organization’s (WHO) definition of osteoporosis for central DXA.
Combination of Two Skeletal Sites The statistical results of the combination are given in Table 5. Although the OR of the combination (4.89–5.65) is improved compared with either one of the single skeletal sites, the large confidence intervals do not allow a claim of significant difference. Moreover, no difference was found for the combined or not combined AUC. In addition, the number of patients correctly classified was not improved when combining the Achilles and the Omnisense. Consequently, the number of false-positives and -negatives stayed stable.
Volume 6, 2003
167 38
Fracture (F) P
mean SD
mean SD
mean SD
0.00
80.0 6.1
74.9 6.0
31.7 5.0
0.01
58.3 9.1
65.4 12.6
58.7 7.6
Age Weight (yrs) (Kg)
0.263
160.5 4.4
159.3 5.1
165.8 6.9
Height (cm)
0.00
22.7 3.7
25.9 4.7
21.3 2.1
BMI
0.002
3834.6 115.0
3922.7 120.2
4115.8 95.2
0.000
55.0 13.1
74.9 15.0
98.4 16.6
0.000
41.9 14.0
61.2 13.7
77.5 13.6
0.000
1485.0 19.0
1514.2 23.8
1555.1 28.5
Omnisense Sahara Sahara Sahara SOS QUI BUA SOS
0.000
56.6 12.9
72.1 11.8
95.4 14.8
0.000
90.7 9.3
101.8 7.8
113.9 11.1
0.000
1486.0 26.1
1515.0 24.9
1570.3 32.8
Achilles Achilles Achilles stiffness BUA SOS
SD, Standard Deviation; P, significance probability; BMI, Body Mass Index in (Kg/m2); QUI, Quantitative Ultrasound Index; Stiffness, mathematical combination of SOS and BUA: Ultrasound Index; SOS, Speed of Sound in m/s; BUA, Broadband Ultrasound Attenuation in dB/MHz.
Test Student ( F vs AMC)
38
47
Strength
Age Matched control (AMC)
Young Control (YC)
Studied groups
Table 1 Characteristics of the Different Groups as a Mean and Standard Deviation as Well as Results of the t-Test between Fractured Subjects and Controls
168
Hans et al. Table 2 Mean Ultrasound Parameters of the Fracture Groups and Controls Assessed for All Devices and Expressed in Z-Score and in T-Scores Using the Same Young Population Measured parameters Sun Rad SR Stiff SR BUA SR SOS Lun Stiff Lun BUA Lun SOS
Age Matched Control (T-score and Z-score) –2.0 –1.4 –1.2 –1.4 –1.6 –1.1 –1.7
Fracture (T-score and Z-score)
0.0 0.0 0.0 0.0 0.0 0.0 0.0
–3.0 –2.6 –2.6 –2.5 –2.6 –2.1 –2.6
–0.9 –1.2 –1.4 –1.0 –1.0 –1.0 –0.9
QUI, Quantitative ultrasound index; SOS, speed of sound; BUA, broadband ultrasound attenuation.
Table 3 Odds Ratios for One Standard Deviation Decrease and Area Under the ROC Curve for Each Ultrasound Parameter Parameters
Odds ratios
(95% CI)
Adjusted Odds Ratiosa
(95% CI)
AUC
(95% CI)
Sun Rad SR Stiff SR BUA SR SOS Lun Stiff Lun BUA Lun SOS
2.28 5.93 5.18 5.65 4.51 4.70 3.97
(1.30;4.01) (2.59;13.57) (2.44;11.00) (2.48;12.89) (2.15;9.45) (2.23;9.88) (1.95;8.08)
2.72 4.76 4.10 4.54 3.50 3.62 3.10
(1.40;5.26) (1.98;11.46) (1.85;9.12) (1.89;10.91) (1.58;7.75) (1.62;8.08) (1.45;6.65)
0.70 0.84 0.84 0.83 0.82 0.83 0.80
(0.58;0.81) (0.75;0.93) (0.75;0.93) (0.74;0.92) (0.73;0.91) (0.74;0.93) (0.70;0.90)
a
Adjusted for age and weight. QUI, Quantitative ultrasound index; SOS, speed of sound; BUA, broadband ultrasound attenuation; AUC, area under the curve; CI, confidence interval.
Discussion Diagnostic Ability of Technologically Different QUS Device of the Radius and Calcaneum To our knowledge, this was the first study to compare heel and radius QUS devices for the discrimination of subjects with hip fractures. Overall, QUS devices involved in our study demonstrated similar ability to discriminate hip-fracture patients from age-matched controls (Table 3). The adjusted odds ratios varied from 2.21–3.84 and the areas under the ROC curve (AUC) ranged from 0.70–0.82. Whereas the Achilles+ appeared to have a higher OR, the dif-
Journal of Clinical Densitometry
ference was not significant because of the overlapping 95% CI. In addition, the 95% CI of the Achilles showed a much larger variability, which canceled completely this apparent advantage. Other studies have demonstrated that QUS can discriminate between controls and patients with hip fracture (5,16). Reported odds ratios for hip fracture have ranged from 1.4–3.7 for each SD decrease in BUA or SOS. Compared with our study, Njeh et al. (5) showed a lower AUC ranging from 0.65–0.71 and lower odds ratios (2.3–2.8), for the Sahara and Achilles devices. Similarly, Ekman et al. (17) found that the risks of hip fracture, estimated as odds ratios for every 1 SD reduction in heel Stiffness Index, was
Volume 6, 2003
Hip Fracture Discrimination Study
169
Fig. 1. Non-age-adjusted ROC curves for Omnisense Radius Speed Of Sound (SOS) compared with Achilles Stiffness and Sahara Quantitative Ultrasound Index (QUI). The area under the curve is significantly lower than both calcaneum device. However, the age difference between groups likely increased artificially the difference between the calcaneum and the radius AUC.
Fig. 2. Comparison of age- and weight-adjusted odd ratio (OR) of all ultrasound parameters with their respective 95% Confidence Interval (95% CI). Although the OR for the calcaneum systems are higher than that of the radius, the 95% CI of all devices are overlapping. Direct comparison is therefore difficult.
3.4 (95% CI 2.2–5.0) in 87 noninstitutionalized women aged 65–85 yr with a first hip fracture, compared to 195 randomly selected age-matched controls. For the radius SOS measured with the
Journal of Clinical Densitometry
Omnisense, Weiss et al. (18) found, in 50 hip-fracture elderly women compared to 130 elderly controls, an AUC of 0.79 (95% CI, 0.73–0.86) and an odds ratio of 1.92 (95% CI, 1.22–3.02). They concluded that
Volume 6, 2003
170
Hans et al. Table 4 Results of the Discriminant Analysis for Each Ultrasound Devices and Their Respective Parametersa Discriminant analysis in %
Optimum discrimination threshold value
Measured parameters
Specificity
Sensitivity
False negative
False positive
Raw value
T-score
Sun Rad SR QUI SR BUA SR SOS Lun Stiff Lun BUA Lun SOS
60.5 73.7 76.3 68.4 68.4 73.7 65.8
63.2 78.8 76.3 76.3 71.1 73.7 71.1
36.8 21.1 23.7 23.7 28.9 26.3 28.9
39.5 26.3 23.7 31.6 31.6 26.3 34.2
3880.0 64.8 51.9 1500.0 64.5 96.5 1501.5
–2.5 –2.0 –1.9 –1.9 –2.1 –1.6 –2.1
a In addition the optimum threshold corresponding to the optimum sensitivity/specificity was calculated for each parameters (expressed as raw value and T-score) QUI, Quantitative ultrasound index; SOS, Speed of sound; BUA, Broadband ultrasound attenuation.
Table 5 Combination of the Omnisence Radius SOS with the Sahara Heel QUI and Achilles Heel Stiffness, Respectively: Results of the Whole Statistical Analysis
Parameters Odds ratiosa (95% CI) AUC (95% CI) Specificity Sensitivity False negative False positive % correctly classified
Average T-score Average T-score Sun Radius SOS Sun Radius SOS and Sahara QUI and Achilles STI 5.65 (2.18–14.64) 0.84 (0.75–0.93) 76.30 76.30 23.70 23.70 76.30
4.89 (1.95–12.27) 0.81 (0.72–0.91) 73.7 71.1 28.9 26.3 72.4
a
Adjusted by age and weight. CI, confidence interval; AUC, area under the ROC curve; STI, stiffness index; QUI, quantitative ultrasound index; SOS, speed of sound.
SOS measured at the radius by Omnisense discriminated subjects with hip fracture from controls. Hans et al. (19) found that the best discriminators of hipfractured patients from controls were the distal radius and calcaneum measurements assessed with the Omnisense device (ORs = 2.4 and 3.0). Because in our study the fractured subjects were not perfectly age- and weight-matched with the conJournal of Clinical Densitometry
trols, it was necessary to adjusted our statistics for those factors. This largely influenced the OR, which decreased from 5.93 to 4.76, and 4.51 to 3.50 for the QUI and Stiffness, respectively, whereas the radius SOS increased (2.28 to 2.72). Taking this into consideration, it is likely that the AUC would react in the same way, i.e., increase for the radius and decrease for the calcaneum if we were able to make such adjustment (not possible with our statistical software). We could then hypothesize that no clinical difference would be observable between the different ultrasound devices tested in our study.
Definition of Discrimination Threshold for Each Ultrasound Parameter Considering our study and the other ones, all QUS devices tested against hip fracture seem to show the same discriminatory ability. However, some differences can be seen in the definition of the diagnostic threshold. Indeed, the main problem remains the definition of optimum thresholds per device for the diagnostic of osteoporosis. In general, the WHO definition of osteoporosis is not applicable to ultrasound devices (20,21). In our study, the Omnisense device seems to have the lowest discriminating t-score (–2.5) compared to both Lunar (T-score = –2.1) and Sahara (T-score = –2.0). Even if the use of cross-sectional studies is always problematic, because they do not separate causes from effects, one should calculate those thresholds based on meta-analysis to have a better and more accurate estimation, if no prospecVolume 6, 2003
Hip Fracture Discrimination Study tive studies are available. Combining several studies weighted according to their designs could also be an attractive alternative.
Effect of Combination of Two Different Skeletal Sites on Discriminatory Ability Logistic regression analysis of the heel (Stiffness or QUI) and radius (SOS) combined t-scores yielded slightly higher adjusted odds ratios of approx 4.89–5.65 (compared with 3.5–4.76 obtained for QUS alone) but no higher AUC. However, these increases were not clinically significant. Thus, the combination of radial and calcaneal QUS measurements assessed by two technologically different ultrasound devices did not significantly improve hipfracture discrimination compared with either method alone. Those results differ from those of Barkman et al. (3), who showed a significant increase of the ROC area to 0.95 (p < 0. 05) when combining the radius, metacarpal, and phalanx SOS. They concluded that the ability to measure a large variety of sites and the potential to combine these measurements were promising to optimize fracture-risk assessment. However, compared to our study, they used the same device to measure all skeletal sites and the outcome was all types of fracture, compared with hip fracture for our study. In addition, their combination did not include the calcaneum (but phalanx and radius). The only study thus far that combined the radius with the calcaneum was published by Hans et al. (19). They studied 79 postmenopausal women with hip fracture and 295 controls. Using a forward selective linear regression model, the discriminator values of combined assessment at two sites were investigated. There was moderate improvement in diagnostic value, but the best combination was the calcaneum with the distal radius, which improved the AUC by 3% and raised both the sensitivity and specificity to 94%. However, once again this combination was performed using the same device (not optimized for calcaneum measurement) and not two different devices of different technologies.
Limitations of the Study Although a comparison of several technologically different ultrasound devices was already performed for the calcaneum, no comparison was made including the radius. Nevertheless, there are some limitations besides the cross-sectional design of the study. Journal of Clinical Densitometry
171 Indeed, one could regret than no DXA was performed on all patients to confirm the osteoporotic status of our low-energy hip-fracture patients. The etiology of the hip fracture may not have an impact on the comparison between the ultrasound devices, but could certainly affect the overall individual results. The sample size of this study is relatively modest, leading to an overestimation of the OR and its 95% confidence interval.
Conclusion The three QUS technologies tested against hip fractures seem to show the same discriminatory ability, although the calcaneum tends to be a stronger determinant of the hip fracture compared with the radius. However, there is some difference in the definition of the diagnostic threshold underlying the nonusability of the WHO osteoporosis definition. Finally, the combination of several site using two different devices is not clinically relevant. Potential interest could be seen in combining several sites using the same device, supposing that such device would measure differently active bone (e.g., cortical vs trabecular bone, weightbearing vs nonweight-bearing bone, etc.). According to the literature and our results, we are confirming that the tested devices could be suitable for screening strategy in conjunction with other clinical factors.
Acknowledgment This study has been performed thanks to research funds received from Sunlight Medical Ltd. Special thank to M Barada, G Conicella, and C. Perron for their technical help as well as the Pr. R. Rizzoli and Drs. M. Delmi, H. Vuagnat, and S. TahintziZawadynski for their participation.
Reference 1. Genant HK, Engelke K, Fuerst T, et al. 1996 Noninvasive assessment of bone mineral and structure: state of the art. J Bone Miner Res 11(6):707–730. 2. Gluer CC. 1997 Quantitative ultrasound techniques for the assessment of osteoporosis: expert agreement on current status. The International Quantitative Ultrasound Consensus Group. J Bone Miner Res 12(8):1280–1288. 3. Barkmann R, Kantorovich E, Singal C, et al. 2000 A new method for quantitative ultrasound measurements at multiple skeletal sites: first results of precision and fracture discrimination. J Clin Densitom 3(1):1–7.
Volume 6, 2003
172 4. Greenspan SL, Bouxsein ML, Melton ME, et al. 1997 Precision and discriminatory ability of calcaneal bone assessment technologies [published erratum appears in J Bone Miner Res 1997 Nov; 12(11):1957]. J Bone Miner Res 12(8):1303–1313. 5. Njeh CF, Hans D, Li J, et al. 2000 Comparison of six calcaneal quantitative ultrasound devices: precision and hip fracture discrimination. Osteoporos Int 11(12):1051–1062. 6. Hans D, Dargent-Molina P, Schott AM, et al. 1996 Ultrasonographic heel measurements to predict hip fracture in elderly women: the EPIDOS prospective study. Lancet 348(9026):511–514. 7. Heaney RP, Avioli LV, Chesnut CHR, Lappe J, Recker RR, Brandenburger GH. 1995 Ultrasound velocity, through bone predicts incident vertebral deformity. J Bone Miner Res 10(3):341–345. 8. Bauer DC, Gluer CC, Cauley JA, et al. 1997 Broadband ultrasound attenuation predicts fractures strongly and independently of densitometry in older women. A prospective study. Study of Osteoporotic Fractures Research Group. Arch Intern Med 157(6):629–634. 9. Thompson PW, Taylor J, Oliver R, Fisher A. 1998 Quantitative ultrasound (QUS) of the heel predicts wrist and osteoporosis related fractures in women age 45–75 years. J Clin Densitom 1(3):219–226. 10. Glüer CC, Hans D. 1999 How to use ultrasound for risk assessment: a need for defining strategies [editorial]. Osteoporos Int 9(3):193–195. 11. Njeh CF, Boivin CM, Langton CM. 1997 The role of ultrasound in the assessment of osteoporosis: a review. Osteoporos Int 7(1):7–22. 12. Hartl F, Tyndall A, Kraenzlin M, et al. 2002 Discriminatory ability of quantitative ultrasound parameters and bone mineral density in a population-based sample of postmenopausal women with vertebral fractures: results of the BOS-study. J Bone Miner Res 17(2):321–330. 13. Njeh CF, Blake GM. 1999 Calcaneal quantitative ultrasound: water-coupled. In: Quantitative Ultrasound: Assessment of
Journal of Clinical Densitometry
Hans et al.
14.
15.
16.
17.
18.
19.
20.
21.
Osteoporosis and Bone Status. Njeh CF, Hans D, Fuerst T, Glueer C-C, Genant HK, eds. Martin Dunitz, London, 109–124. Cheng S, Hans D, Genant H. 1999 Calcaneal Quantitative Ultrasound Systems: Gel-coupled, in Quantitative Ultrasound: Assessment of Osteoporosis and bone status. Njeh CF, Hans D, Fuerst T, Glueer C-C. eds. Martin Dunitz, London, p. 125–144. Hans D, Fan B, Fuerst T. 1999 Non Heel Quantitative Ultrasound Devices, in Quantitative Ultrasound: Assessment of Osteoporosis and bone status. Njeh CF, Hans D, Fuerst T, Glueer C-C, Genant H, eds. Martin Dunitz, London, p. 145–162. Hans D, Allaoua S, Genton L, et al. 2002 Is time since fracture influencing the discrimination between hip fractured and non fractured subjects as assessed by three technologically different quantitative ultrasound devices of the calcaneum. Calcif Tissue Int 71(6):485–492. Ekman A, Michaelsson K, Petren-Mallmin M, Ljunghall S, Mallmin H. 2001 DXA of the hip and heel ultrasound but not densitometry of the fingers can discriminate female hip fracture patients from controls: a comparison between four different methods. Osteoporos Int 12(3):185–191. Weiss M, Ben-Shlomo A, Hagag P, Ish-Shalom S. 2000 Discrimination of proximal hip fracture by quantitative ultrasound measurement at the radius. Osteoporos Int 11(5):411–416. Hans D, Srivastav SK, Singal C, et al. 1999 Does the combination of quantitative ultrasound at multiple measurement sites improve discrimination of hip fractures? J Bone and Mineral Res 14(4):644–651. Faulkner KG, von Stetten E, Miller P. 1999 Discordance in patient classification using T-scores. J Clin Densitom 2(3):343–350. Frost ML, Blake GM, Fogelman I. 2000 Can the WHO criteria for diagnosing osteoporosis be applied to calcaneal quantitative ultrasound? Osteoporos Int 11(4):321–330.
Volume 6, 2003
Original Article
Hip Fracture Discrimination Study QUS of the Radius and the Calcaneum
Didier Hans,*,1 Laurence Genton,2 Samia Allaoua,1 Claude Pichard,2 and Daniel O. Slosman1 1Nuclear
Medicine, Geneva University Hospital, Geneva, CH; and 2Nutrition, Geneva University Hospital, Geneva, CH
Abstract As an emerging alternative to current radiation-based bone densitometry techniques, there is a growing interest in the use of quantitative ultrasound (QUS) measurements for the noninvasive assessment of fracture risk in the management of osteoporosis. However, there are also a multiplicity of technologically different QUS devices available on the market and, so far, no study has compared heel and radius QUS device for the discrimination of subjects with hip fractures. Our study evaluated the ability of three QUS devices (one calcaneal gel-coupled system, one calcaneal water-coupled system, and one radius system) to discriminate osteoporotic from controls subjects, using the same population. We also checked if the combination of two different skeletal sites (i.e., calcaneum and radius) improve the discriminatory ability of the QUS devices. Forty-five women aged 79.1 ± 7.1 yr with hip fractures within the last 4 d were used as the hip-fracture group and compared to 40 healthy controls from 65–87 yr. In addition, 47 young controls, aged 20–40 yr, were used as reference population to express some of the results as T-scores. QUS measurements were performed with the Hologic Sahara, Ge-Lunar Achilles+, and Sunlight Omnisense devices according to the manufacturer’s recommendations. Adjusted odds ratio results showed that a decrease in Omnisense SOS of 1 standard deviation (SD) was associated with a significant increase in fracture risk (OR adj. = 2.83) comparable with Sahara BUA (OR adj. = 2.42) and Achilles BUA (OR adj. = 3.29). However, given the large overlap between the 95% intervals of each odds ratio, no significant difference was found between the devices. Similarly, comparison between the areas under ROC curves did not show any significant difference between all the parameters. Considering the parameters provided per default by each QUS device, the Sahara, Achilles, and Omnisense devices classified correctly 70, 67.5, and 62.5% of the subjects, respectively. Although the OR of the combination of radius and calcaneum is improved (3.62 to 4.74) compared with either one of the single skeletal site, the large confidence intervals do not allow to claim a significant difference. The three QUS technologies tested against hip fractures seem to show the same discriminatory ability. However, there is some differences in the definition of the diagnostic threshold underlying the nonusability of the World Health Organization (WHO) osteoporosis definition. Finally, the combination of several site using two different devices is not clinically relevant. Potential interest could be seen in combining several sites using the same device, supposing that such device would measure differently active bone. Key Words: Hip fractures; QUS; SOS; BUA; radius; calcaneum. Received 03/06/02; Accepted 10/08/02. *Address correspondence to Dr. D. Hans, Head of R&D, Division of Nuclear Medicine, Geneva University Hospital, 1211 Geneva 14, Switzerland. E-mail: [email protected]
163
164
Introduction Osteoporosis represents a major worldwide public-health problem that will grow in importance in the coming decades as the population ages. Because of the importance of bone mineral density (BMD) in determining bone strength, many noninvasive techniques based on the attenuation of ionizing radiation have been developed to quantify BMD in the axial and peripheral skeleton (1). These include Single Photon Absorptiometry (SPA), Single X-Ray Absorptiometry (SXA), Dual X-Ray Absorptiometry (DXA), Quantitative Computed Tomography (QCT), and peripheral QCT (pQCT). The decision about which method to use is often based on cost and availability rather than technical considerations like predictive value, precision, or radiation dose. DXA is the most widely used method of BMD measurement, largely because of its low radiation exposure, high precision, and ability to assess several different sites in the skeleton. However, it is a relatively expensive approach that gives limited information on bone structure (2). As a consequence, accessibility to such technology is limited for many doctors, and subsequently to many women. Indeed, according to the International Osteoporosis Foundation report, 83% of women in Europe have never been screened for the potential risk of osteoporosis. Such forecasts have lead to the search for new, cost-effective methods for early detection, prevention, and treatment, such as peripheral systems (pDXA, PQCT, and so on). As an emerging alternative to current radiationbased bone-densitometry techniques, there is a growing interest in the use of quantitative ultrasound (QUS) measurements for the noninvasive assessment of fracture risk in the management of osteoporosis (2). Quantitative ultrasound is a simple, inexpensive, and relatively portable technique that is noninvasive and ionizing radiation-free. As such, it has a much greater potential for widespread application than standard bone densitometry. Several studies have demonstrated the ability of QUS to discriminate patients with osteoporotic fractures from age-matched controls (3–5) and more importantly to predict fracture risk (6–9). Therefore, QUS may be used as a screening method for osteoporosis (10) or as a diagnostic tool if clinically relevant threshold are defined. Though based on similar principles, QUS instruments from different manufacturers have significant Journal of Clinical Densitometry
Hans et al. differences, particularly in their calibration methods, sites of measurement, acquisitions, analysis software, and scanner designs (11). Moreover, it is difficult to assess whether all QUS devices have similar discriminatory ability because cross-sectional odd ratios are affected by the study populations. As a result, the Receiver Operating Characteristic Curve (ROC) as well as the odds ratio may vary largely between devices of different brands. As a result, results obtained with one device cannot be compared with another that is technologically different. Therefore, QUS devices ideally should be compared on the same population. A few studies have reported such results (5) for the calcaneum, but none of them compared ultrasound measurement at different sites except Hartl et al. (12). In their study, they tested the ability of heel vs phalangeal ultrasound devices in discriminating vertebral fracture. They concluded that QUS measurements of the proximal phalanges performed as well as QUS measurements of the calcaneum for the discrimination of subjects with multiple vertebral fractures. The combination of DXA and QUS has been studied several times, primarily to find if those two technologies would have an improving effect in diagnosing patient at risk of osteoporosis. Conflicting results do not seem to support the hypothesis. Similarly, some studies investigated if the combination of several skeletal sites assessed by ultrasound only would be beneficial as well (3,19). However, the combination of the calcaneum using a dedicated technology with the radius has never been investigated. An added effect of two skeletal sites would likely decrease the overlap between fractured and nonfractured patients, and could also be used to select a subpopulation at very high risk of osteoporosis. To our knowledge, no study has compared heel and radius QUS device for the discrimination of subjects with hip fractures. As a result, the first aim of our study was to evaluate the ability of three QUS devices (one calcaneal gel-coupled system, one calcaneal water-coupled system, and one radius system) to discriminate osteoporotic from controls subjects, using the same population. The second and third aims were to define the optimum discrimination threshold for each ultrasound parameter and to check how the combination of two different skeletal sites (i.e., calcaneum and radius) affects the discriminatory ability of the different QUS devices. Volume 6, 2003
Hip Fracture Discrimination Study
Materials and Methods Subjects The Geneva University Hospital enrolled 123 subjects. Thirty-eight women aged 80.0 ± 6.1 yr with hip fractures within the last 4 d were used as the hip-fracture group. Subjects were included in the fracture group only if the fracture resulted from a fall of standing height or less and involved no motion greater than walking. These fractures were defined as low-energy fractures and have been called “osteoporotic fracture.” They were compared to 38 nonfractured controls from 66–87 yr. In addition, 47 young controls, aged 20–40 yr, were used as reference population to express some of the results as T-scores. All participants underwent a questionnaire about their medical history. Women with hormone replacement therapy (HRT) use for more than 6 mo within the previous 10 yr, Paget’s disease, juvenile diabetes, renal failure, or malignant disease with metastatic tumor were excluded from the study. We also excluded women who took SERM, calcitonin, bisphosphonates, anabolic steroids, fluoride, or parathyroid hormone for more than 3 mo within the previous 3 yr or any other drug known to affect bone metabolism. Smoking was not an exclusion criteria. For each subject, height, weight, and body mass index (BMI = weight [kg]/height2 [m]) were obtained.
Ultrasound Devices and Measurements Measurements were acquired using three ultrasound devices, the Achilles+ (GE-Lunar Corporation, USA), the Sahara (Hologic Inc., USA) and the Sunlight Omnisense™ (Sunlight Medical Ltd., Israel). The first two systems measure ultrasounds at the calcaneum. The Achilles uses water for coupling and its transducers are placed at fixed distance from each other, whereas the second QUS device uses ultrasonic gel and its transducers are placed in contact with the calcaneum. The particularity of the third device is its measured skeletal site (the radius) and its mode of ultrasound propagation (axial). All systems measure speed of sound (SOSm/s). In addition the Achilles+ and the Sahara devices carry out the broadband ultrasound attenuation (BUA-dB/MHz) parameter and a combined index called Stiffness and QUI, respectively. The technical details and the precision of these devices Journal of Clinical Densitometry
165 have been described by Njeh et al. (13), Cheng et al. (14), and Hans et al. (15). One trained operator performed all the measurements according to the recommendations provided by the device manufacturer. Each subject underwent one single QUS measurement per device. QUS measurements were carried out on the left calcaneum (and radius), except for the fracture group where the measurements were performed at the heel of the nonfractured leg. Daily quality control (QC) was performed for all the ultrasound devices using acoustic phantoms provided by the manufacturers. No drift or shift was observed during the study with the exception of the Sahara device, for which the coupling pads were changed once.
Statistical Analysis Before any analysis, all values of each group were tested for normal distribution. All results were expressed as mean and standard deviation (SD). For all devices, the differences between QUS measurements of the fracture and nonfracture groups were compared by a Student’s t-test. Only p-values < 0.05 were considered as significant. Risk of hip fracture associated with low ultrasound measurements was estimated with logistic regression analysis. The presence of fracture was used as the binary response variable. Results were expressed as age- and weightadjusted odd ratios per SD decrease with 95% confidence intervals. A receiver operating curve (ROC) analysis was also conducted to assess the performance of bone-measurement parameters for the classification of subjects with and without hip fractures by calculating the respective area under the curve (AUC). Each ROC curve was compared using a bivariate Chi-square test. Finally, a discriminant analysis between fracture and nonfracture groups was performed to test the ability of the different ultrasound variables to classify study subjects according to their hip-fracture status. Resulting values in terms of sensitivity and specificity were used to identify the corresponding threshold value from the ROC. Combinations of two skeletal sites (radius and calcaneum) were performed for the QUI, Stiffness, and SOS (Omnisense only for the later) parameters using the average T-score. Similar statistics than the one used for a single skeletal site were performed. Data analysis was performed with the software Statistical Package for Social Sciences, Volume 6, 2003
166 Version 9.0.0. for IBM PC (SPSS, Inc., Chicago, IL) and Excel 2000 (Microsoft, Seattle, WA) program.
Results Demographics All ultrasound and anthropometric parameters showed a normal distribution and the characteristics of the three study groups are presented in Table 1. The age was significantly higher for the fractured subjects than the controls, whereas their weight and BMI were significantly lower. All bone parameters were significantly lower in the fracture group (see Table 1). As a consequence, results of the logistic regression were adjusted for age and weight. BUA, SOS, and the combined index (Stiffness and QUI) of the fractured subjects, expressed in Zscores and in t-scores using the results of healthy aged and young controls, respectively, are shown in Table 2. The Omnisense device showed the lowest T-score values (–3.0) compared to both Sahara and Lunar (–2.6).
Diagnostic Ability The ability of the three QUS devices to discriminate between control and hip-fractured patients is presented in Table 3, and in Figs. 1 and 2, and was expressed as both odd ratios and area under the ROC curve. The significancy of nonadjusted and adjusted logistic regression coefficients associated with each bone parameter (p < 0.05) demonstrated that all the ultrasound parameters were predictors of fracture risk. Adjusted odd ratios, which indicate the increase of fracture risk for 1 SD decrease in bone-measurement parameter, were all significantly higher than 1. Adjusted odds ratio results showed that a decrease in Omnisense SOS of 1 SD was associated with a significant increase in fracture risk (OR adj. = 2.7) but apparently lower than Sahara BUA (OR adj. = 4.8) and Achilles BUA (OR adj. = 3.5). However, given the large overlap between the 95% intervals of each odds ratio, no significant difference was found between the devices. Comparison between the areas under ROC curves did not show any significant difference between all the parameters in their ability to discriminate between fractured subjects and controls
Journal of Clinical Densitometry
Hans et al. for both calcaneum devices. In the upper left quarter of Fig. 2, which corresponds to the sensitivity and specificity of interest (both between 0.5 and 1), the curves are very close to each other. However, the radius SOS was significantly lower than any calcaneum ultrasound parameters measured with the Achilles and Sahara. This difference is likely owing to the age and weight differences between groups, which seems to affect more the Omnisense device than the two other calcaneal ones (see the results expressed as T-scores and the raw OR in Table 2 and 3, respectively).
Definition of Thresholds A discriminant analysis was applied for subjects with hip fractures and controls. Results in terms of sensitivity and specificity for the assessment of hip fractures for the different bone parameters are shown in Table 4 for QUS. The highest sensitivity (76.3%) and specificity (76.3) for the detection of hip fracture was seen for the calcaneal BUA values measured with the Sahara device. If we consider the parameters provided per default by each QUS device, the Sahara, Achilles, and Omnisense classified correctly 76.3, 69.7, and 62.0% of the subjects, respectively. With the sensitivity and specificity calculated in the discriminant analysis, it was possible to identify the corresponding optimum threshold values for the different bone parameters (from the ROC analysis), as shown in tables 4. The lowest T-scores for hip fracture discrimination were recorded with the Omnisense SOS (–2.5), which was matching the World Health Organization’s (WHO) definition of osteoporosis for central DXA.
Combination of Two Skeletal Sites The statistical results of the combination are given in Table 5. Although the OR of the combination (4.89–5.65) is improved compared with either one of the single skeletal sites, the large confidence intervals do not allow a claim of significant difference. Moreover, no difference was found for the combined or not combined AUC. In addition, the number of patients correctly classified was not improved when combining the Achilles and the Omnisense. Consequently, the number of false-positives and -negatives stayed stable.
Volume 6, 2003
167 38
Fracture (F) P
mean SD
mean SD
mean SD
0.00
80.0 6.1
74.9 6.0
31.7 5.0
0.01
58.3 9.1
65.4 12.6
58.7 7.6
Age Weight (yrs) (Kg)
0.263
160.5 4.4
159.3 5.1
165.8 6.9
Height (cm)
0.00
22.7 3.7
25.9 4.7
21.3 2.1
BMI
0.002
3834.6 115.0
3922.7 120.2
4115.8 95.2
0.000
55.0 13.1
74.9 15.0
98.4 16.6
0.000
41.9 14.0
61.2 13.7
77.5 13.6
0.000
1485.0 19.0
1514.2 23.8
1555.1 28.5
Omnisense Sahara Sahara Sahara SOS QUI BUA SOS
0.000
56.6 12.9
72.1 11.8
95.4 14.8
0.000
90.7 9.3
101.8 7.8
113.9 11.1
0.000
1486.0 26.1
1515.0 24.9
1570.3 32.8
Achilles Achilles Achilles stiffness BUA SOS
SD, Standard Deviation; P, significance probability; BMI, Body Mass Index in (Kg/m2); QUI, Quantitative Ultrasound Index; Stiffness, mathematical combination of SOS and BUA: Ultrasound Index; SOS, Speed of Sound in m/s; BUA, Broadband Ultrasound Attenuation in dB/MHz.
Test Student ( F vs AMC)
38
47
Strength
Age Matched control (AMC)
Young Control (YC)
Studied groups
Table 1 Characteristics of the Different Groups as a Mean and Standard Deviation as Well as Results of the t-Test between Fractured Subjects and Controls
168
Hans et al. Table 2 Mean Ultrasound Parameters of the Fracture Groups and Controls Assessed for All Devices and Expressed in Z-Score and in T-Scores Using the Same Young Population Measured parameters Sun Rad SR Stiff SR BUA SR SOS Lun Stiff Lun BUA Lun SOS
Age Matched Control (T-score and Z-score) –2.0 –1.4 –1.2 –1.4 –1.6 –1.1 –1.7
Fracture (T-score and Z-score)
0.0 0.0 0.0 0.0 0.0 0.0 0.0
–3.0 –2.6 –2.6 –2.5 –2.6 –2.1 –2.6
–0.9 –1.2 –1.4 –1.0 –1.0 –1.0 –0.9
QUI, Quantitative ultrasound index; SOS, speed of sound; BUA, broadband ultrasound attenuation.
Table 3 Odds Ratios for One Standard Deviation Decrease and Area Under the ROC Curve for Each Ultrasound Parameter Parameters
Odds ratios
(95% CI)
Adjusted Odds Ratiosa
(95% CI)
AUC
(95% CI)
Sun Rad SR Stiff SR BUA SR SOS Lun Stiff Lun BUA Lun SOS
2.28 5.93 5.18 5.65 4.51 4.70 3.97
(1.30;4.01) (2.59;13.57) (2.44;11.00) (2.48;12.89) (2.15;9.45) (2.23;9.88) (1.95;8.08)
2.72 4.76 4.10 4.54 3.50 3.62 3.10
(1.40;5.26) (1.98;11.46) (1.85;9.12) (1.89;10.91) (1.58;7.75) (1.62;8.08) (1.45;6.65)
0.70 0.84 0.84 0.83 0.82 0.83 0.80
(0.58;0.81) (0.75;0.93) (0.75;0.93) (0.74;0.92) (0.73;0.91) (0.74;0.93) (0.70;0.90)
a
Adjusted for age and weight. QUI, Quantitative ultrasound index; SOS, speed of sound; BUA, broadband ultrasound attenuation; AUC, area under the curve; CI, confidence interval.
Discussion Diagnostic Ability of Technologically Different QUS Device of the Radius and Calcaneum To our knowledge, this was the first study to compare heel and radius QUS devices for the discrimination of subjects with hip fractures. Overall, QUS devices involved in our study demonstrated similar ability to discriminate hip-fracture patients from age-matched controls (Table 3). The adjusted odds ratios varied from 2.21–3.84 and the areas under the ROC curve (AUC) ranged from 0.70–0.82. Whereas the Achilles+ appeared to have a higher OR, the dif-
Journal of Clinical Densitometry
ference was not significant because of the overlapping 95% CI. In addition, the 95% CI of the Achilles showed a much larger variability, which canceled completely this apparent advantage. Other studies have demonstrated that QUS can discriminate between controls and patients with hip fracture (5,16). Reported odds ratios for hip fracture have ranged from 1.4–3.7 for each SD decrease in BUA or SOS. Compared with our study, Njeh et al. (5) showed a lower AUC ranging from 0.65–0.71 and lower odds ratios (2.3–2.8), for the Sahara and Achilles devices. Similarly, Ekman et al. (17) found that the risks of hip fracture, estimated as odds ratios for every 1 SD reduction in heel Stiffness Index, was
Volume 6, 2003
Hip Fracture Discrimination Study
169
Fig. 1. Non-age-adjusted ROC curves for Omnisense Radius Speed Of Sound (SOS) compared with Achilles Stiffness and Sahara Quantitative Ultrasound Index (QUI). The area under the curve is significantly lower than both calcaneum device. However, the age difference between groups likely increased artificially the difference between the calcaneum and the radius AUC.
Fig. 2. Comparison of age- and weight-adjusted odd ratio (OR) of all ultrasound parameters with their respective 95% Confidence Interval (95% CI). Although the OR for the calcaneum systems are higher than that of the radius, the 95% CI of all devices are overlapping. Direct comparison is therefore difficult.
3.4 (95% CI 2.2–5.0) in 87 noninstitutionalized women aged 65–85 yr with a first hip fracture, compared to 195 randomly selected age-matched controls. For the radius SOS measured with the
Journal of Clinical Densitometry
Omnisense, Weiss et al. (18) found, in 50 hip-fracture elderly women compared to 130 elderly controls, an AUC of 0.79 (95% CI, 0.73–0.86) and an odds ratio of 1.92 (95% CI, 1.22–3.02). They concluded that
Volume 6, 2003
170
Hans et al. Table 4 Results of the Discriminant Analysis for Each Ultrasound Devices and Their Respective Parametersa Discriminant analysis in %
Optimum discrimination threshold value
Measured parameters
Specificity
Sensitivity
False negative
False positive
Raw value
T-score
Sun Rad SR QUI SR BUA SR SOS Lun Stiff Lun BUA Lun SOS
60.5 73.7 76.3 68.4 68.4 73.7 65.8
63.2 78.8 76.3 76.3 71.1 73.7 71.1
36.8 21.1 23.7 23.7 28.9 26.3 28.9
39.5 26.3 23.7 31.6 31.6 26.3 34.2
3880.0 64.8 51.9 1500.0 64.5 96.5 1501.5
–2.5 –2.0 –1.9 –1.9 –2.1 –1.6 –2.1
a In addition the optimum threshold corresponding to the optimum sensitivity/specificity was calculated for each parameters (expressed as raw value and T-score) QUI, Quantitative ultrasound index; SOS, Speed of sound; BUA, Broadband ultrasound attenuation.
Table 5 Combination of the Omnisence Radius SOS with the Sahara Heel QUI and Achilles Heel Stiffness, Respectively: Results of the Whole Statistical Analysis
Parameters Odds ratiosa (95% CI) AUC (95% CI) Specificity Sensitivity False negative False positive % correctly classified
Average T-score Average T-score Sun Radius SOS Sun Radius SOS and Sahara QUI and Achilles STI 5.65 (2.18–14.64) 0.84 (0.75–0.93) 76.30 76.30 23.70 23.70 76.30
4.89 (1.95–12.27) 0.81 (0.72–0.91) 73.7 71.1 28.9 26.3 72.4
a
Adjusted by age and weight. CI, confidence interval; AUC, area under the ROC curve; STI, stiffness index; QUI, quantitative ultrasound index; SOS, speed of sound.
SOS measured at the radius by Omnisense discriminated subjects with hip fracture from controls. Hans et al. (19) found that the best discriminators of hipfractured patients from controls were the distal radius and calcaneum measurements assessed with the Omnisense device (ORs = 2.4 and 3.0). Because in our study the fractured subjects were not perfectly age- and weight-matched with the conJournal of Clinical Densitometry
trols, it was necessary to adjusted our statistics for those factors. This largely influenced the OR, which decreased from 5.93 to 4.76, and 4.51 to 3.50 for the QUI and Stiffness, respectively, whereas the radius SOS increased (2.28 to 2.72). Taking this into consideration, it is likely that the AUC would react in the same way, i.e., increase for the radius and decrease for the calcaneum if we were able to make such adjustment (not possible with our statistical software). We could then hypothesize that no clinical difference would be observable between the different ultrasound devices tested in our study.
Definition of Discrimination Threshold for Each Ultrasound Parameter Considering our study and the other ones, all QUS devices tested against hip fracture seem to show the same discriminatory ability. However, some differences can be seen in the definition of the diagnostic threshold. Indeed, the main problem remains the definition of optimum thresholds per device for the diagnostic of osteoporosis. In general, the WHO definition of osteoporosis is not applicable to ultrasound devices (20,21). In our study, the Omnisense device seems to have the lowest discriminating t-score (–2.5) compared to both Lunar (T-score = –2.1) and Sahara (T-score = –2.0). Even if the use of cross-sectional studies is always problematic, because they do not separate causes from effects, one should calculate those thresholds based on meta-analysis to have a better and more accurate estimation, if no prospecVolume 6, 2003
Hip Fracture Discrimination Study tive studies are available. Combining several studies weighted according to their designs could also be an attractive alternative.
Effect of Combination of Two Different Skeletal Sites on Discriminatory Ability Logistic regression analysis of the heel (Stiffness or QUI) and radius (SOS) combined t-scores yielded slightly higher adjusted odds ratios of approx 4.89–5.65 (compared with 3.5–4.76 obtained for QUS alone) but no higher AUC. However, these increases were not clinically significant. Thus, the combination of radial and calcaneal QUS measurements assessed by two technologically different ultrasound devices did not significantly improve hipfracture discrimination compared with either method alone. Those results differ from those of Barkman et al. (3), who showed a significant increase of the ROC area to 0.95 (p < 0. 05) when combining the radius, metacarpal, and phalanx SOS. They concluded that the ability to measure a large variety of sites and the potential to combine these measurements were promising to optimize fracture-risk assessment. However, compared to our study, they used the same device to measure all skeletal sites and the outcome was all types of fracture, compared with hip fracture for our study. In addition, their combination did not include the calcaneum (but phalanx and radius). The only study thus far that combined the radius with the calcaneum was published by Hans et al. (19). They studied 79 postmenopausal women with hip fracture and 295 controls. Using a forward selective linear regression model, the discriminator values of combined assessment at two sites were investigated. There was moderate improvement in diagnostic value, but the best combination was the calcaneum with the distal radius, which improved the AUC by 3% and raised both the sensitivity and specificity to 94%. However, once again this combination was performed using the same device (not optimized for calcaneum measurement) and not two different devices of different technologies.
Limitations of the Study Although a comparison of several technologically different ultrasound devices was already performed for the calcaneum, no comparison was made including the radius. Nevertheless, there are some limitations besides the cross-sectional design of the study. Journal of Clinical Densitometry
171 Indeed, one could regret than no DXA was performed on all patients to confirm the osteoporotic status of our low-energy hip-fracture patients. The etiology of the hip fracture may not have an impact on the comparison between the ultrasound devices, but could certainly affect the overall individual results. The sample size of this study is relatively modest, leading to an overestimation of the OR and its 95% confidence interval.
Conclusion The three QUS technologies tested against hip fractures seem to show the same discriminatory ability, although the calcaneum tends to be a stronger determinant of the hip fracture compared with the radius. However, there is some difference in the definition of the diagnostic threshold underlying the nonusability of the WHO osteoporosis definition. Finally, the combination of several site using two different devices is not clinically relevant. Potential interest could be seen in combining several sites using the same device, supposing that such device would measure differently active bone (e.g., cortical vs trabecular bone, weightbearing vs nonweight-bearing bone, etc.). According to the literature and our results, we are confirming that the tested devices could be suitable for screening strategy in conjunction with other clinical factors.
Acknowledgment This study has been performed thanks to research funds received from Sunlight Medical Ltd. Special thank to M Barada, G Conicella, and C. Perron for their technical help as well as the Pr. R. Rizzoli and Drs. M. Delmi, H. Vuagnat, and S. TahintziZawadynski for their participation.
Reference 1. Genant HK, Engelke K, Fuerst T, et al. 1996 Noninvasive assessment of bone mineral and structure: state of the art. J Bone Miner Res 11(6):707–730. 2. Gluer CC. 1997 Quantitative ultrasound techniques for the assessment of osteoporosis: expert agreement on current status. The International Quantitative Ultrasound Consensus Group. J Bone Miner Res 12(8):1280–1288. 3. Barkmann R, Kantorovich E, Singal C, et al. 2000 A new method for quantitative ultrasound measurements at multiple skeletal sites: first results of precision and fracture discrimination. J Clin Densitom 3(1):1–7.
Volume 6, 2003
172 4. Greenspan SL, Bouxsein ML, Melton ME, et al. 1997 Precision and discriminatory ability of calcaneal bone assessment technologies [published erratum appears in J Bone Miner Res 1997 Nov; 12(11):1957]. J Bone Miner Res 12(8):1303–1313. 5. Njeh CF, Hans D, Li J, et al. 2000 Comparison of six calcaneal quantitative ultrasound devices: precision and hip fracture discrimination. Osteoporos Int 11(12):1051–1062. 6. Hans D, Dargent-Molina P, Schott AM, et al. 1996 Ultrasonographic heel measurements to predict hip fracture in elderly women: the EPIDOS prospective study. Lancet 348(9026):511–514. 7. Heaney RP, Avioli LV, Chesnut CHR, Lappe J, Recker RR, Brandenburger GH. 1995 Ultrasound velocity, through bone predicts incident vertebral deformity. J Bone Miner Res 10(3):341–345. 8. Bauer DC, Gluer CC, Cauley JA, et al. 1997 Broadband ultrasound attenuation predicts fractures strongly and independently of densitometry in older women. A prospective study. Study of Osteoporotic Fractures Research Group. Arch Intern Med 157(6):629–634. 9. Thompson PW, Taylor J, Oliver R, Fisher A. 1998 Quantitative ultrasound (QUS) of the heel predicts wrist and osteoporosis related fractures in women age 45–75 years. J Clin Densitom 1(3):219–226. 10. Glüer CC, Hans D. 1999 How to use ultrasound for risk assessment: a need for defining strategies [editorial]. Osteoporos Int 9(3):193–195. 11. Njeh CF, Boivin CM, Langton CM. 1997 The role of ultrasound in the assessment of osteoporosis: a review. Osteoporos Int 7(1):7–22. 12. Hartl F, Tyndall A, Kraenzlin M, et al. 2002 Discriminatory ability of quantitative ultrasound parameters and bone mineral density in a population-based sample of postmenopausal women with vertebral fractures: results of the BOS-study. J Bone Miner Res 17(2):321–330. 13. Njeh CF, Blake GM. 1999 Calcaneal quantitative ultrasound: water-coupled. In: Quantitative Ultrasound: Assessment of
Journal of Clinical Densitometry
Hans et al.
14.
15.
16.
17.
18.
19.
20.
21.
Osteoporosis and Bone Status. Njeh CF, Hans D, Fuerst T, Glueer C-C, Genant HK, eds. Martin Dunitz, London, 109–124. Cheng S, Hans D, Genant H. 1999 Calcaneal Quantitative Ultrasound Systems: Gel-coupled, in Quantitative Ultrasound: Assessment of Osteoporosis and bone status. Njeh CF, Hans D, Fuerst T, Glueer C-C. eds. Martin Dunitz, London, p. 125–144. Hans D, Fan B, Fuerst T. 1999 Non Heel Quantitative Ultrasound Devices, in Quantitative Ultrasound: Assessment of Osteoporosis and bone status. Njeh CF, Hans D, Fuerst T, Glueer C-C, Genant H, eds. Martin Dunitz, London, p. 145–162. Hans D, Allaoua S, Genton L, et al. 2002 Is time since fracture influencing the discrimination between hip fractured and non fractured subjects as assessed by three technologically different quantitative ultrasound devices of the calcaneum. Calcif Tissue Int 71(6):485–492. Ekman A, Michaelsson K, Petren-Mallmin M, Ljunghall S, Mallmin H. 2001 DXA of the hip and heel ultrasound but not densitometry of the fingers can discriminate female hip fracture patients from controls: a comparison between four different methods. Osteoporos Int 12(3):185–191. Weiss M, Ben-Shlomo A, Hagag P, Ish-Shalom S. 2000 Discrimination of proximal hip fracture by quantitative ultrasound measurement at the radius. Osteoporos Int 11(5):411–416. Hans D, Srivastav SK, Singal C, et al. 1999 Does the combination of quantitative ultrasound at multiple measurement sites improve discrimination of hip fractures? J Bone and Mineral Res 14(4):644–651. Faulkner KG, von Stetten E, Miller P. 1999 Discordance in patient classification using T-scores. J Clin Densitom 2(3):343–350. Frost ML, Blake GM, Fogelman I. 2000 Can the WHO criteria for diagnosing osteoporosis be applied to calcaneal quantitative ultrasound? Osteoporos Int 11(4):321–330.
Volume 6, 2003