各廠牌骨質密度儀比較表

---

 

廠牌	OSTEOMETER (DTX-200) （雙能DEXA）	Table Type (可能廠牌: Norland、Lunnar、Holigic…) （雙能DEXA）
X光劑量	低 (55kV, 0.3mA; patient dose 0.2 mGy per scan)	高
操作面積	1.5 平方公尺	20平方公尺(6坪)
操作容易度	簡單、容易，有操作步驟提示，最重要的是掃描測量計算分析起始點與全部計算面積，都是軟體自動偵測定位進行，不費吹灰之力，準確迅速又方便。	最重要的是掃描測量計算分析起始點與全部計算面積，都必須人為手工定位而後交給電腦計算分析，人為手工定位難免有誤差，既耗時間又不方便。
測量檢查部位 (單純或複雜)	前臂超遠端(Ultra Distal) 前臂遠端(Distal 8mm) (構造最單純，只有薄薄的一層皮與少許的肌肉，是沒有誤差干擾變數存在的位置)	主要測量檢查部位部份：脊椎骨(L2、L3、L4)、股骨(大轉子) 澄清誤解：一般稱呼此三種大型品牌的骨質密度儀是whole body，事實上並非如字面上之解釋，可以任意測量全身任何部位或是全身的骨骼都測量。因為不同的骨骼部位有不同的密度存在，要測量的部位必須事先建立正常值的資料庫，測量之後才能進行判讀診斷。(脊椎骨被兩旁的肌肉與前面的各種內臟(胃、腸)軟組織緊緊包圍著，猶如隻身處在叢林中，是誤差干擾變數最多的位置)
受測者測量前之準備與姿勢角度之要求	不須更衣，坐在儀器旁邊只要捲起衣袖，將手臂置於測量的凹槽內，輕輕握住握把即可。當受測者重複測量時，每次都能很容易的擺出相同的姿勢角度。(這是操作者輕鬆、受測者方便的檢查部位)	受測者要更衣，要上下檢查台，若遇到年老或行動不便者，操作者更要費力扶上扶下，當受測者重複測量時不可能每次都擺出同樣的姿勢角度。(這是操作者費力、耗時間，受測者不方便的檢查部份)
精 準 度 (Precision)	最佳精準度，誤差小於1% 理由：由於所選量檢查部位的單純特性(不會變形沒有誤差干擾變數存在)，再加上DTX-200絕佳的精密結構與電腦軟體的獨特功能，無論何人操作或是受測者重複測量，DTX-200的電腦軟體都能自動偵測到橈骨尺骨相距0.8cm處。以此點距離定位，向前計算至最遠端及向後計算2.4cm的面積，無論受測者握住握把是緊或鬆，每次都是這個位置不會偏差，除非橈骨尺骨變形扭曲。	精準度差，脊椎骨誤差大於3%，股骨誤差大於3%。 理由：脊椎骨其外形凹凸不規則，再加骨剌、脊椎側彎、動脈鈣化、各種內臟、腸胃內未消化的魚骨與各種結石，這些都是影響精準度誤差的干擾變數、非常難以克服，再加上在掃描後操作者再以手動方式移動螢幕上的游標線將L2、L3、L4框住，交給電腦計算分析，由於此位的複雜度不宜以自動定位的方式進行。所以，當不同的操作者可能產生不同的定位角度，若再加上受測者重複測量所擺出的姿勢角度又不一樣時，所測量脊椎的精準度就差多了。
早期骨質輕微流失的偵測能力	由於精準度最佳，不會有假陽性的高估，所以DTX-200能很正確敏銳的偵測反映出受測者是否有輕微的骨質流失現象，很容易達到早期發現早期治療的檢查目的。	由於誤差干擾變數最多，而且又不容易克服，所以測量脊椎部位容易得到假陽性的高估，不容易偵測反映出受測者是否有輕微的骨質流失現象，所以不容易達到早期發現早期治療的檢查目的。
動態骨質流失率計算功能	有此功能。 只要輸入受測的以生化標示劑檢測所得的數，DTX-200特殊的軟體即能計算出受測者當時體內正在進行中的動態骨質流失速率。一般的骨質密度檢查只是反映出受測者當時存在的骨質密度，是屬於靜態性，無法分辨出受測者是否屬於「快速骨質代謝性骨質疏鬆症」或是「非快速骨質代謝性骨質疏鬆症」，此二種急慢性症狀，將隨著醫藥科技之進步，在治療方法上將有所不同，所以兩者在臨床上都具有同等重要性，兩者兼備是先進或家對骨質疏鬆症的預防測量檢查、診斷治療與療效評估監視之最近發展趨勢。 罹患骨質疏鬆症的二大危險因子，在DTX-200的測量檢查計算分析之下都會現出原形。	無此功能

 

 

使用外部DXA系統比較脊椎和股骨的正常密度

---

A Comparison of a Peripheral DXA System with 
Conventional Densitometry of the Spine and Femur

 Rajesh Patel, MSC, Glen M Blake, PHD, Anita Jefferies, BSC,

Paul M. Sautereau-Chandley, BSC, and lgnac Fogelman, MD

 Department of Nuclear Medicine, Guy’s Hospital, London, UK; and Merck Sharpe

& Dohme Ltd, Hoddesdon, Hertfordshire, UK

Abstract

Because of the perceived high cost of dual-energy X-ray absorptiometry (DXA) studies of the spine and femur, there is renewed interest in small, low-cost X-ray devices for scanning the peripheral skeleton. We have compared forearm bone mineral density (BMD) measurements (distal and ultradistal sites) performed on a DTX-200 (Osteometer MediTech, Hoersholm, Denmark) with spine (Ll-L4) and femur (femoral neck and total hip sites) scans performed on a QDR-4500 (Hologic, Waltham, MA) in 172 white UK women aged 22-84 yr with a view to establishing differences caused by inconsistent reference ranges and different age-related changes in BMD. All BMDs were expressed as T-scores using the manufacturers’ reference ranges for the forearm and spine, and the National Health and Nutrition Examination Survey (NHANES) ranges for the femur. Linear regression between peripheral and axial sites gave correlation coefficients r = 0.71–0.74 and roof mean standard errors (RMSE) 0.88–1.14 in J-score units. Subjects were divided into the following five age groups: <40 yr; 40–49 yr; 50–59 yr; 60–69 yr and ≥70 yr. A large systematic difference between distal and ultradistal T-scores (mean AT= 0.59, SEM = 0.05) was found affecting all age groups. When the mean difference in T-score between each forearm site (distal, ultradistal) and each axial site (spine, femoral neck, total hip) was examined for premenopausal subjects (n = 58) the mean difference between forearm and axial T-score showed a consistent negative offset (AT = –0.41 to –0.48) for the distal forearm site and a consistent positive offset (AT = t0.30 to t0.37) for the ultradistal site. When interpreting results in postmenopausal women, age-related T-score changes in the forearm were in close agreement with the femoral neck region of interest (ROI), but systematic differences were found between the forearm and the spine and total hip sites. The two forearm and three axial sites were compared to evaluate the number of postmenopausal subjects identified as osteoporotic on the basis of the World Health Organization (WHO) Study Group criteria (T-score <–2.5). Although forearm and spine T-scores identified approximately equal numbers of subjects as osteoporotic (distal 38/114; ultradistal 31/114; spine 30/114), the two femur sites identified fewer subjects as osteoporotic (femoral neck 25/114; total hip 16/114). The number for the total hip site was statistically significantly smaller than the spine and forearm sites. In conclusion, we have identified systematic differences between T-score results for a peripheral and an axial DXA device that may have a significant effect on the interpretation of BMD measurements. 

Introduction

Over the past decade, the growing awareness of the effects of osteoporosis on the elderly population (1) and the consequent costs of health care (2) have stimulated a rapid growth in the clinical applications of bone densitometry (3). With its advantages of high precision, low radiation dose, and stable calibration, dual-energy X-ray absorptiometry (DXA) scanning of the spine and hip has become an essential part of the evaluation of patients at risk of osteoporosis. Bone mineral density (BMD) measurements are performed to detect low bone mass and therefore, assess the risk of fragility fracture to help make decisions on initiating preventive treatment and to evaluate response to treatment. By allowing intervention with preventive treatment in patients at the highest risk, bone densitometry may help reduce future fracture incidence by up to 50% (4).

Although the number of clinics providing bone densitometry services has increased, there are inadequate resources to meet demand, and conventional DXA scans of the spine and hip are not available to all patients who could conceivably benefit. However, osteoporosis is a systemic disease, and it has been shown that the risk of fracture can be assessed from BMD measurements obtained at peripheral sites (5-7). Because of the perceived high cost of DXA for axial measurements, in recent years, there has been a renewed interest in small, low-cost X-ray absorptiometry devices dedicated to scanning the peripheral skeleton (3, 8). Such systems are compact, simple to operate, and result in an exceptionally low radiation dose to the patient.

Originally, bone densitometry of the forearm was performed using the technique of single-photon absorptiometry (SPA), which utilized a radioactive ¹²⁵I source. However, following the technical advances in DXA, an X-ray tube with a low-voltage generator has replaced the radionuclide source, avoiding the need for frequent source replacement and recalibration, which was a serious drawback of SPA. Examination times for acquiring a projectional scan of the distal forearm that allows the measurement of BMD at both trabecular and cortical sites have been reduced to around 4-5 min. This methodology is usually referred to as peripheral DXA (pDXA).

In this study, pDXA measurements of BMD in the distal forearm were compared with conventional DXA measurements of the lumbar spine and femur in a group of white UK women. The object of the study was to compare the agreement between the interpretation of forearm DXA measurements with DXA of the spine and femur. Differences between measurements in the peripheral and axial skeleton might reflect inconsistencies in reference data (i.e., in the mean BMD and standard deviation [SD] for young normal subjects) or site-dependent differences in the variation of BMD with age. The object of this study was to characterize these differences and determine the degree of concordance between forearm and spine and hip DXA measurements.

 

Subjects and Method

Forearm BMD measurements at the distal and ultradistal sites were performed in the nondominant arm in 172 white UK women scanned using a DTX-200 pDXA system (Osteomerer MediTech). Scans were performed in duplicate with repositioning of the patients forearm between measurements. Conventional DXA scans of the lumbar spine (Ll-L4) and left proximal femur were performed using a Hologic QDR-4500 system (Hologic). One hundred forty-six subjects (mean age 59.6, range 38-84 yr) were patients referred by their general practitioner for a routine bone densitometry investigation. The remaining 26 subjects were young healthy premenopausal women (mean age30.8, range 22-38 yr) recruited to give data over a wider range of ages. The study was approved by the local research ethics committee.

BMD results from the following skeletal sites were recorded: lumbar spine (L1-L4), femoral neck, total hip, distal (radius plus ulna), and ultradistal forearm (radius only). The distal site is defined as the 24-mm long section of bone immediately proximal to the reference line where the separation between radius and ulna is 8 mm. It consists of 87% cortical bone and 13% trabecular bone. The ultradistal site is the area distal to the 8-mm reference line, and contains 45% cortical and 55% trabecular bone. All scan analysis was performed according to the manufacturers’ standard protocols. Forearm scans were analyzed using software version 1.54. The manufacturers’ reference ranges for the spine, distal, and ultradistal sites were used to convert the BMD results into T-score values according to the equation:

T-score = (Measured UMD – young adult mean BMD/young adult SD)   (1)

Reference range data based on the NHANES III study (9) were used to calculate T-score values for the femoral neck and total hip sites. The manufacturers’ reference range for the spine was used to convert the BMD results for this site into 2-score values according to the equation:

Z-score = (Measured BMD – age-matched BMD/age-matched SD)   (2)

Linear regression analysis was used to examine the relationship between the distal and ultradistal T-scores, and between each of the two forearm sites and each of the three axial sites for all 172 subjects.

Short-term precision of the distal and ultradistal BMD measurements was calculated as the coefficient of variation (CV = [SD/mean] × 100). Results for individual patients were combined as the root mean square (RMS) CV (10).

To investigate the consistency between T-score values for the distal and ultradistal sites, the subjects were divided into the following five age groups: <40 yr; 40–49 yr; 50–59 yr; 60–69 yr, and ≥70 yr, and the mean difference between the ultradistal and distal T-scores was plotted for each age group.

To investigate the consistency between the reference data for the forearm and axial sites, the mean difference among distal forearm T-score, ultradistal T-score, and the T-score for each of the three axial sites (spine, femoral neck, and total hip) was examined for the premenopausal women in the study. The statistical significance of the mean T-score difference between each pair of sites was evaluated using Student’s t-test. Premenopausal women were chosen for the comparison on the assumption that age-related BMD losses since the attainment of peak bone mass should be minimal. To test this latter assumption, linear regression analysis was performed between T-score and age for all five measurement sites for the premenopausal women.

To investigate the different age- and menopause-related changes in T-score at the different skeletal sites, the postmenopausal women were divided into age groups according to decade and the mean difference in T-score between each forearm site and each axial site plotted as a function of age group. Statistical significance of the difference from the mean T-score difference for the premenopausal women was tested using Student’s t-test.

 

Finally, the forearm and axial T-scores were used to evaluate the number of postmenopausal subjects identified as osteoporotic at each site on the basis of the World Health Organization (WHO) Study Group criteria (T-score <–2.5) (11). The statistical significance of the differences in the number of osteoporotic subjects at each measurement site was evaluated using Fisher’s exact test. All statistical evaluations were performed using a commercial software package (Stata, College Station, TX).

Results

The number of subjects in each of the five age groups is shown in Table 1 together with the mean 2-score for the lumbar spine and the mean T-score for all five sites studied in each age band. Results of linear, regression analysis between the distal and ultradistal T-scores gave a correlation coefficient of r = 0.92 (root mean standard error [RMSE]. σ_T = 0.68). Results of linear regression analysis between the distal and ultradistal forearm T-scores and T-scores derived from measurements at the three axial sites (Hologic spine, National Health and Nutrition Examination Survey [NHANES] femoral neck and NHANES total hip) gave correlation coefficients of r = 0.71–0.74 and RMSE of σ_T = 0.88–1.14. Precision of the forearm BMD measurements gave a coefficient of variation of 1.55% for the distal site and 2.99% for the ultradistal site. When expressed as the RMSD of the T-score figures, the precision was σ_T = 0.11 for the distal and σ_T = 0.21 for the ultradistal site.

When the mean difference ultradistal T-score-distal T-score was plotted as a function of age group, large and highly statistically significant systematic differences (δT= 0.43–0.71) were evident for all five age groups (Fig. 1). When the mean difference in T-score between each forearm site (distal, ultradistal) and each axial site (spine, femoral neck, total hip) was examined for the premenopausal subjects (n = 58), the mean difference between forearm and axial T-score showed a consistent negative offset (AT= –0.41 to–0.48) for the distal forearm site and a consistent positive offset (AT= +0.30 to +0.37) for the ultradistal site (Fig. 2). All the differences were highly statistically significant (p = 0.039 to p = 0.0004). When linear regression plots between T-score and age were examined for the same subjects, the slopes were small (AT= +0.05 to –0.31/p decade), and nor statistically significantly different from 0 at any of the five measurement sites (Table 2).

 

When the postmenopausal women (n = 114) were divided into age groups according to decade and the mean difference in T-score between each forearm site and each axial site plotted as a function of age group with the data for the premenopausal women, recognizable trends were evident for both the distal and ultradistal data, indicating systematic differences in age-related changes between forearm and the different axial sites (Fig. 3A-C). The numerical values of the mean differences in forearm and axial T-score shown in Fig. 3 are summarized in Table 3together with the standard error of the mean (SEM) and population SD. The SD figures in Table 3 indicate the limits of agreement when results for forearm T-score are used to predict T-score figures for sites in the axial skeleton. For premenopausal women, the SD values are all about 1.0. This gives 95% confidence intervals of approx ±2 for predicting axial T-score values from forearm measurements. For postmenopausal women, SD values, and therefore the limits of agreement, are lightly larger than for premenopausal women, and show a trend to increase with age.

The two forearm and three axial sites were compared to evaluate the number of postmenopausal subject identified as osteoporotic on the basis of the WHO Study Group criteria (T-score < –2.5) (Table 2). The results show that although forearm and spine T-scores identified approximately equal number of subjects as osteoporotic (distal 38/114; ultradistal 31/114; spine 30/114), the two femur sites identified fewer subjects as osteoporotic (femoral neck 25/114; total hip 16/114). The number for the total hip site was statistically significantly smaller than the spine and forearm sites (Table 2).

The introduction of new commercial devices for performing BMD measurements in the peripheral skeleton raises a number of legitimate concerns. First, since few centers providing a bone densitometry service have the resources to generate their own normative data, the accuracy of reference ranges provided by manufacturers for the calculation of T-score is often essential to the validity of the clinical interpretation of scans. This is an issue that continues to generate controversy, even for established densitometer systems that have been in use for many years (12, 13).

A second source of concern over using peripheral rather than axial measurement sites is that different effects of aging and the menopause at different sites might cause systematic discrepancies in scan interpretation in older subjects. Thus, although use of consistent young adult reference data should ensure agreement across manufacturers and sites for T-score results in younger (premenopausal) women, significant differences might nevertheless still arise for postmenopausal women and cause differences in the number of patients diagnosed as osteoporotic, and offered preventive treatment.

The object of the present study was to compare the agreement between T-score results from a new pDXA densitometer with a well-established and widely used instrument for measuring spine and femur BMD. Premenopausal and postmenopausal women over a wide range of ages were studied to differentiate between differences arising from inconsistencies in the manufacturers' reference ranges and those arising from different age-related changes at the different skeletal sites. The subjects were all white women referred by their general practitioners for bone mass measurements of the spine and femur, and who volunteered to have an additional scan on the DTX-200. To obtain data over a wider age range, numbers were augmented by volunteers from hospital staff and staff from local health centers. No attempt was made to exclude women with risk factors for osteoporosis. Although the mean 2-score for lumbar spine BMD was negative at each age group (Table l), the overall mean 2-score (–0.10 ± 0.14) was not statistically significantly different from 0. Therefore, we believe that on the basis of their BMD results, the women in the present study were statistically indistinguishable from normal subjects.

As would be expected for adjacent sites, T-score results for the distal and ultradistal regions of interest (ROIs) were highly correlated. Correlation coefficients between the two forearm sites and the three axial sites were aroundr ≈ 0.7 and were typical of the values expected between distant sites in the skeleton (14).

An unexpected result of this study was a highly significant systematic difference between T-score results for the distal and ultradistal sites (Fig. 1). To investigate this discrepancy T-score differences between the forearm and axial sites were examined for premenopausal women (Fig. 2). Age-related rates of loss in these women were negligible (Table 2), suggesting that they were representative of the young adult population used to define peak bone mass. The findings were consistent for all three axial sites, with the distal ROI showing a mean T-score difference of –0.45 and the ultradistal a mean difference of +0.32 (Fig. 2). This result points to a significant discrepancy between the DTX-200 reference data and the NHANES femur reference range (9), which has recently been recommended by the International Committee for Standards in Bone Measurement (ICSBM) as the basis for the interpretation of hip BMD measurements (15).

When the effects of aging and the menopause on the differences between forearm and axial site T-scores in postmenopausal women were investigated, systematic differences from the results for premenopausal women were noted to be suggestive of different age-related changes at different sites (Fig. 3). The difference between postmenopausal women aged 50-59 yr and premenopausal women is consistent with more rapid bone loss in the spine than the forearm following the menopause (Fig. 3A). For older subjects, the difference was reversed, probably because of the increasing effects of degenerative disease on spine BMD in elderly subjects. For the femoral neck site there was close agreement with forearm T-score across all age ranges (Fig. 3B), whereas for the total hip site, the differences seen were suggestive of lower age-related losses than in the forearm (Fig. 3C). Tables of T-score differences, such as those listed in Table 3 could be used to correct peripheral measurements and thus ensure consistency with treatment decisions based on results for axial sites. However, it is important to bear in mind that some clinical conditions may lead to significant bone mineral losses from axial sites that would not be predicted from peripheral measurements.

Another important consequence of changing skeletal measurement site is that a different group of patients will be identified as osteoporotic on the basis of the WHO Study Group criteria. As a consequence, the number of patients thought to be osteoporotic may also change. Because the WHO criteria define osteoporosis as a T-score result < –2.5, a difference in the number of subjects identified as osteoporotic at different sites may reflect either errors in reference range data or differences in age-related losses between measurement sites.

The principal limitation of the present study was that the subjects were not randomly selected from the general population. Twenty-six out of 58 premenopausal women were either hospital staff or volunteers from local health clinics. The remaining patients were referred by their general practitioners for bone densitometry studies for reasons that included help in deciding whether to start estrogen- replacement therapy, family history of osteoporosis, and radiological evidence of osteopenia. However, the mean Z-score figures (Table 1) were consistent with data from age-marched normal subjects included in earlier reference range studies conducted in our unit (13, 16).

In conclusion, we have shown that systematic differences in T-score between BMD measurements in premenopausal women on the DTX-200 and the Hologic spine and NHANES hip T-scores are strongly suggestive of errors in the DTX-200 reference data. When interpreting results in postmenopausal women, age-related T-score changes in the forearm were in close agreement with the femoral neck ROI, but systematic differences were found between the forearm and the spine and total hip sites. When patients were classified as osteoporotic on the basis of the WHO Study Group criteria, the number of patients identified as osteoporotic on the basis of forearm T-score was similar to the spine, but rather greater than the femoral neck and total hip sites.

 

Acknowledgment

The authors are grateful to Merck, Sharpe & Dohme Lrd for the loan of the DTX-200 forearm scanner.

 

Reference

1. Melton LJ, Thainer M, Ray NF, et al. 1997.Fractures attriburable to osteoporosis: report from the Narional Osteoporosis Foundation. J Bone Miner lles 1216-23.

2. Ray NF, Chan JK. Thanier M, Melton LJ. 1997 Medical expenditures for the treatment of osteoporosis fractures in the United States in 1975: report from the National Osteoporosis Foundarion. J Bone Miner Res 1224-35.

3. Blake GM, Wahner HW, Fogelman I. 1998 l h e Evaluarion of Osteoporosis: Dual Energy X-ray Absorptiometry and Ultrasound in Clinical Practice. Martin Dunirz, London.

4. Black DM. Cummiiigs SR, Karpf DB, et al. 1996 Randomised trial of the effect of alendronate on risk of fracture in womeii

with existing vertebral fractures. Lanccr 348: 1535-1 54 I .

5. Marshall I), ]ohnellO, Wcdel H. 1996 Metaanalysis of how well measures on bone mineral density predict occurrence of osteoporotic fractures. Br Med J 312:1254-1259.

6. Duppe H, Gardsell I? Nilsson B, Johnell 0. 1997 A single bone density nieasurement can predict fractures over 25 years. Calcif Tissue Int 60:171-174.

7. Cuniniings SR, Black DM, Nevitt MC, et al. 1993 Bone density at various sites for prediction of hip fractures. Lancet 341 :72-75.

8. Gluer C-C, ]ergas M, Hans D. 1997 Peripheral nieisurement techniques for the assessment of osteoporosis. Seminars Nuclear

Med 27:229-247.

7. Looker AC, Waliner HW, Dunn WL, et al. I978 Updated data on proximal femur hone mineral levels of US adults. Osteoporosis lilt (in press).

10. Gluer C-C, Blake GM, Lu Y, Blunt BA, jergas M, Genant HK. 1995 Accurate assessment of precision errors: bow 10 measure the reproducibility of bone densitometry techiiiques. Osteoporosis Inr 1795; 5:262-270.

1 1 . Rrport of n WHO Study Group. 1994 Assessnient of fracture risk and its applicatioii to screening for postnienopausal osreoporosis. WHO Technical Report Serirs H43. World Health Organisation. Geneva.

12. Faulkner KG, Rohcrrs LA, McClung MR. 1996 Discrepancies in normative darn between Lunar and Hologic DEXA systems.

Osteoporosis Int 6~432-436.

13. Ahmed AIH, Blake GM, Rymer JM, Fogelman 1. 1997 Screening for osteopenia and osteoporosis: do the accepted normal ranges lead to overdiagnosis? Osteoporosis Int 7:432-438.

14. Grampp S, Genant HK, Marhur A, et al. 1997 Comparisons of non-invasive bone niincral measurements in assessing age-relared loss. fraciure discrimination and diagnostic classification. J Rone Miner Res 12:697-71 1 .

15. Hanson J. 1997 Letter to die editor: stalldardisarion of femur BMD measurements. J Bone Miner Res 12:1316-1317.

16. Ryan PJ, Blake GM, Fogelman I. 1192 I’usmienopausal screening for vsteopenia. Br J Rheuinarol 31:823-828.

 

 

Impact of Soft Tissue on In Vivo Accuracy of Bone Mineral Measurements in the Spine

---

Impact of Soft Tissue on In Vivo Accuracy of Bone Mineral Measurements in the Spine, Hip, and Forearm: A Human Cadaver Study  OLE L. SVENDSEN, CHRISTIAN HASSAGER, VERA SKODT, and CLAUS CHRISTIANSEN  ABSTRACT The impact of soft tissue in vivo on accuracy of bone mineral density (BMD) measurements of the spine and hip by dual energy X-ray absorptiometry and of the forearm by single photo absorptiometry was assessed by use of 14 human cadavers. The in vivo accuracy errors (SEE%) were: forearm 3-5%, anteroposterior spine 5.3% lateral spine 10-12%, and femoral greater trochanter, neck, total, intertrochanteric, and Ward’s triangle 3%, 6.5%, 6.7%, 8%, and 11-13%, respectively. Except from the lateral spine and the greater trochanter, the slopes of the linear regressions of in vivo BMD against in vitro BMD were not significantly different from 1(p > 0.05). The calculated random accuracy error of BMD measurements due to fat inhomogeneity was estimated to 3-4% for the anteroposterior spine and 9-14% for the lateral spine (from abdominal computed tomography in 26 healthy women).In conclusion, acceptable accuracy errors below 6-7% (of soft tissue in vivo) of BMD measurements were obtained in the anteroposterior spine, the forearm, and the neck, greater trochanter, and total proximal femur. INTRODUCTION Low bone mineral density (BMD) is associated with increased risk of osteoporotic fractures. The precision errors of current bone densitometers, i.e., dual energy X-ray absorptiometry (DEXA), single energy X-ray absorptiometry (SEXA), and single photon absorptiometry (SPA) is about 1%. The accuracy error of bone mineral measurements is of clinical importance when a patient value is compared with a reference population. The lower the differences between subjects (biological variation) of a given parameter, the lower the accuracy error of the method has to be in order to differentiate between normal and abnormal values. Since the biological variation of bone density is relatively small (1SD is approximately 10-15%), the accuracy error becomes of crucial importance.Precious studies have shown a high in vitro accuracy of bone mineral measurements, i.e., comparison of bone mineral measurements, i.e., comparison of phantoms or already excised bones with dry- or ash weight. However, the accuracy of bone mineral measurements depend on the soft tissue mass and composition. Thus, to be of clinical value a study of accuracy must include the in vivo accuracy. i.e., measurement of bone mineral in vivo with soft tissue followed by excision of bones with in vitro measurements. The in vivo accuracy of DEXA measurements of spinal and hip bone density has not been assessed. In addition, a basic but unproven assumption for DEXA is that the percentage of fat (fat%) in a measurement point with bone is the same as the fat% in a measurement point without bone. The aim of the present study was to assess the in vivo accuracy of bone mineral measurements of the hip and spine by DEXA and of the forearm by SPA. MATERIALS Fourteen human cadavers were used in the present study. Permission for autopsy was given by a close relative. None had bone metastasis or earlier fracture of the bones studied. The study was carried out in accordance with the Declaration of Helsinki II and with approval from the Ethical Committee of Copnhagen Country. ACCURACY OF BONE MINERAL MEASUREMENTS METHODS Measurement of each bone was performed twice (without repositioning) before autopsy (in vivo), twice after excision and mechanical removal of soft tissue (in vivo 1), and twice after chemical removal of remaining soft tissue (in vitro 2). In vivo measurements Measurements of bone mineral of the lumbar spine and the left femur were performed by DEXA with a QDR-1000 X-ray bone densitometer, software version 4.29 (Hologic, Inc., Waltham, MA). In vivo measurements and all analyses were performed in accordance with Operators Manual and User’s Guide (QDR-1000, Hologic. Inc.). For the AP spine (anterior-posterior projection) bone mineral concentration (BMC), area (cm2) of interest (AREA), and bone mineral density (BMD) of the integrated vertebrae L2-L4 were measured, whereas for the LAT spine (lateral projection) BMC, AREA, and BMD of the vertebral body of L2 and L3, excluding the posterior cortical vertebral complex, were measured. LAT spine was measured with the subject in the decubitus position. The left proximal femur was measured in vivo with the leg rotated 45° inward. A drill was bored into the femoral shaft about 10 cm distal to the minor trochanter. The drill was positioned in level by mans of a spirit level to enable exact repositioning in 45° inward rotation at the later in vitro measurements of the excited proximal femur. BMC, AREA, and BMD of the left distal forearm were measured by 125I SPA (DT 100. Osteometer A.S. Rodovre, DK, software version 2.1). The site where the distance between the ulna and radius is 8mm is determined by an initial survey scan. BMC and BMD is then measured in 24mm of integrated bone proximal to the 8mm site. The results are corrected for fat. Before in vivo measurements, the forearm was positioned in the DT 100, and a drill was bored through the radius and ulna proximal to the site of measurement but within the visible scan area. The drill served as a reference point and controlled the supination/ pronation of the forearm. In vitro measurements After the in vivo measurements, the spine, the femur, and the forearm were excised and processed as follows. The lumbar spines and the left proximal femurs (including the drill and the femoral head) were removed from the cadavers. The soft tissue on the palmar side of the forearm from the radioulnacarpeal joint distally to the drill proximally was removed. Plaster casts were made with the forearm in the in situ measurement position for later repositioning for the in vitro measurements. Thereafter the radius and ulna were excised through the radioulnacarpeal joint and proximal to the drill. The soft tissue was carefully removed from the excised bones with a pair of scissors and scalpels. Thereafter, measurements of the bones were performed (in vitro 1). The remaining extra- and intraosseous soft tissue was then removed with a chemical procedure using three solution: (a) antiformin 5% for 18 h, (b) sodium carbonate 2% for 12 h, and (c) ether/acetone 50%/50% v/v for 24 h. The chemical macerated and defatted bones were dried at 60℃ to constant weight. Thereafter, measurements were performed once again (in vitro 2). The in vitro measurements by DEXA were performed with the spine and the proximal femur embedded in a 71.4% water/ethanol solution (w/w, liquid height 16-17cm) in Plexiglas models constructed for the purpose. In vitro measurements by SPA were performed in 100% water. Quality assurances were performed daily. Since the aim of the study was to investigate the accuracy, we attempted to minimize the influence of the precision error between different measuring situations. Thus, once a region of interest of a particular subject had been defined, all consecutive in vitro measurements of that particular specimen were related to this. The influence of soft tissue heterogeneity on BMD measurements In order to evaluate the magnitude of the theoretical error in measured spinal BMD due to fat inhomogeneity in the abdomen, we performed computed tomography scans(using a General Electric CT-9800Q computerized tomography (CT) scanner with the following radiographic factors: 120kVp, 140mA, 3s, slice interval and thickness: 10mm) from the first to the fourth lumbar intervertebral disk in 26 healthy postmenopausal women. This resulted in 9-12 horizontal 10mm thick CT scan per women. The women were examined in a supine position with their arms stretched above their heads. The women (aged 54.7 ± 2.4 years; mean ± SD) covered a wide range of body fatness with body mass index (BMI) ranging from 19.9 to 35.1 kg m2 (mean 27.9 ± 4.2kg/ m2). The CT Hounsfield Units were calibrated against fat% (determined by chemical analysis) by measurements on mixtures of lard and beef. With this in vitro calibration it is possible to calculate the fat% from the measured mean Hounsfield Units in a region of interest (ROI) by CT in vivo. The fat% was thus assessed by CT in ROIs corresponding to bone pixels and nonbone pixels, respectively, for both AP and LAT spinal BMD measurements by DEXA. Bone and bone marrow were excluded from the CT ROIs. For the AP spine, an ROI on the CT images that included the vertebral body in the width and the whole subject (soft tissue) in the height was defined. This ROI corresponds to the bone containing pixels for the AP spine by DEXA. This ROI was then moved laterally (without changing the height or width of the ROI) until the transverse process was no longer within the ROI. This ROI correspond to non-bone-containing pixels for the AP spine by DEXA. For the LAT pine, a ROI that included the vertebral body in the height and the whole subject (soft tissue) in the width was defined. This ROI correspond to bone-containing pixels for LAT spine by DEXA. The ROI was then moved anteriorly until the vertebral body was no longer within the ROI. This ROI corresponds to the non-bone-containing pixels for the LAT-spine by DEXA. For each women the fat% data from all slices were averaged, resulting in four data values for each woman: the soft tissue fat% in bone and non-bone pixels for both AP and lateral projection. Furthermore, the height and width of the abdomen were measured for each women on the CT-scan corresponding to the umbilicus level. CALCULATIONS AND STATISTICAL ANALYSIS Knowledge about thickness of the subjects and the difference in fat% between non-bone pixels and bone pixels makes it possible to calculate the error in measured spinal BMD caused by fat inhomogeneity in surrounding soft tissue. error(delta) of BMD (g/cm2) = Delta fat% × 1/100 × diameter of subject (in cm) × 0.051 The precision errors were calculated as coefficients of variation (CV%) of dublicate measurements in vivo, in vitro 1, and in vitro 2. The accuracy errors were calculated as the standard error of estimate (SEE%) of linear regressions between in situ and in vitro measurements (mean of double measurements). A significance level at p < 0.05 was chosen. All analysis was performed with SAS (SAS Institute Inc., Cary NC). RESULTS Table 1 gives the clinical data of the 14 human cadavers included in the study. The precision error (CV%) of the BMD measurements in vivo for the AP and LAT spine was 0.7 and 3.5%, respectively (not shown). It was 0.9% for the forearm and between 1.3 and 2.6% for the different femoral regions. The precision error in vitro (after mechanical removal of soft tissue) was 2-3 times lower, namely 0.4, 0.7, and 0.4% for the AP spine, the LAT spine, and the forearm, respectively. There was no systematic difference between the in vivo and the in vitro precision errors of the different femoral regions. In vivo versus in vitro measurements Figure 1 shows the relationships between the in vivo and the in vitro 1 BMDs (after macroscopic removal of soft tissue) of the AP spine and LAT spine measurements (left column). The linear regressions yielded r values of 0.93 and 0.77 for the AP spine and LAT spine, respectively. For the LAT spine, but not for the AP spine, the slope of the regression line was significantly different from 1 (p < 0.05). The right column shows the correlation between the in vitro 1 and 2 (chemical removal of remaining soft tissue) measurements. The r values were higher (0.95 and 0.98, respectively), and the slopes were not significantly different from 1 (p > 0.05). Figure 2 shows the relationships between the in vivo and the in vitro 1 BMD measurements in the other areas of interest (forearm, hip). The r values were generally higher than what was the case for the spine values. The slopes of the regression lines were, except from the greater trochanter, not statistically significantly different from 1 (p > 0.05). Figure 3 visualizes the accuracy errors of the BMD measurements from the in vivo to the in vitro 1 measurements. The lowest errors were found for the forearm (2.9%) and the trochanteric hip region (3.4%), but also the AP spine had a relatively low error (5.2%). The LAT spine and Ward’s triangle had relatively high accuracy errors exceeding 10%. As shown in Table 2, the accuracy values did not change significantly for the in vitro 2 measurements. Table 2 also gives the r values and accuracy errors of the BMC and AREA measurements. The accuracy errors of BMD measurements were, in general, lower than the BMC measurements and higher than the AREA measurements. The sum of the dry weight of L2, L3, and L4 was correlated with the in vivo AP spine BMC measurements with an r-value of 0.93 and an accuracy error of 8.3% (SEE%). The error due to fat inhomogeneity The relationship between the fat% in bone and non-bone pixels is illustrated in the left part of Fig.4. There was a high correlation between fat% in bone and in non-bone pixels for both the AP (r = 0.93) and the LAT (r = 0.95) projections. However, the fat% in non-bone pixels was higher than in bone pixels for the AP projection: 37.7% (9.2%) versus 31.2% (9.5%) (mean [SD]), p < 0.001, and lower for the LAT projection: 29.8% (10.0%) versus 31.1% (9.7%), p < 0.05. More importantly, the individual differences in fat% between non-bone pixels and bone pixels ranged from 1-15% fat for the AP projection and from -8-7% fat for the LAT projection. The right part of Fig. 4 shows the calculated theoretical errors in DEXA measured spinal BMD caused by fat inhomogeneity. The fat inhomogeneity resulted in a random accuracy error at about 0.03 g/cm2 (SD) for the AP projection, and about 0.06 g/cm2 for the LAT projector. Using previously published BMD data for normal pre- and post- menopausal women (mean AP spinal BMD = 1.03 and 0.81 g/cm2, and mean LAT spinal BMD = 0.68 and 0.43 g/cm2 respectively.) we thus find that the estimated random accuracy error due to fat inhomogeneity is about 3-4% for the AP projection and 9-14% for the LAT projection. DISCUSSION The accuracy error is the total error made in estimating the true value. It may be expressed as the standard error of estimate about a regression between the measured values and the true values if the estimated line is used. The true total accuracy error of bone mineral measurements is the square root of the sum of the squared accuracy error by soft tissue of measurements in vivo and the sum of the squared accuracy error of measurements in vitro. In the present study the accuracy in vivo was assessed because it gives a more true impression of the measurement error when bone mineral is measured in a patient. Others have shown a high in vitro accuracy by comparing bone mineral values obtained from measuring excised bone and the corresponding ash content. In the present study we have not determined the ash content. The obvious reason for this decision was that it is impossible to accurately isolate the region of interest, not only in the forearm and hip but also for the spin. When the AP spine is measured in vivo it is the integrated bone mineral of L2-L4 that is measured. The posterior processus spinosus from L1 is then within the measured ROI, whereas that of L4 is outside this ROI. Despite this a priori error, the regression of in vivo AP spine BMC on the dry weight of L2-L4 gave an accuracy error (SEE%) of 8.3%, which is only slightly higher than that from in vivo to in vitro 1 (6.1%) of in vitro 2 (7.6%). In the study by Ho et al., the SEE% of measurements in vitro versus dry weight was the same as in vitro versus ash weight (8-9%). Together, these findings suggest that the accuracy error of bone mineral measurements is predominantly determined by the accuracy error of soft tissue in vivo, which is further supported by low accuracy errors of measurements on bone chips or phantoms. High correlations were found between the in vivo and in vitro measurements, but there were accuracy errors ranging from 2.9% to more than 11%. Low accuracy errors appeared in the forearm measurements (3-5%). The AP pine (5.2%), and certain femoral regions (3-6%). The reliability of a diagnostic test depends not only on the accuracy of the test but also on the biological variation of the variable in the population on which the test is applied. The lower the differences between subjects, the lower the accuracy error has to be in order to reliably differentiate between normal subjects and those at risk (low BMD). It can be calculated that a biological variation at about 14% (as seen for BMD in normal perimenopausal women) demands an accuracy error below 6-7%. In addition to the accuracy error caused by soft tissue and variations in fat%, the accuracy may also be influenced by such degenerative changes that take place in the spine with increasing age. Formation of osteophytes, calcification of ligaments, etc., will result in systematic high values that do not reflect bone strength. Such errors can possibly explain why no bone loss is found in the spine in women over age 60. In the present study BMC and BMD were determined in the hip and spine with some of the most advanced technology. LAT spine BMD was, however, determined in the decubitus position and it is unknown whether the recently developed method to determine LAT spine BMD in the supine position would improve the results. Forearm BMC/BMD was measured by SPA using a multipath technique. SPA scanners are currently being replaced by SEXA scanners. The accuracy of SEXA has not been investigated yet, but since the correlation between SPA and SEXA on comparable equipment is very high, it is reasonable to assume that SEXA has the same accuracy error as SPA. Correspondingly, the introduction of DEXA did improve the accuracy error compared with DPA. The precision errors found in the present study represent the short-term in vivo and in vitro errors. For clinical use the long-term in vivo precision is the value of interest, but for method comparison also shot-term in vitro precision is of value. The fact that the precision errors was 2-3 times higher in situ than in vitro strongly suggests the soft tissue also plays a role in the precision.   CONCLUSION Acceptable accuracy errors below 6-7% (of soft tissue in vivo) of BMD measurements by DEXA and SPA were obtained for the AP spine the forearm, and for the neck, greater trochanter, and total of the proximal femur.   ACKNOWLEDGMENTS We thank Dr. Otto Brændstrup, head of the Department of Pathological Anatomy, Glostrup Hospital, for allowing us to perform the work in his department, and the staff who kindly helped us throughout the study.   REFERENCES 1.      Consensus Development Conference. Diagnosis, Prophylaxis. and Treatment of Osteoporosis 1993. An J Med 94; 646-650. 2.      Marslew U, Riis BJ. Christiansen C 1991 Desogestrel in hormone replacement therapy; Long-term effects on bone, calcium and lipid metabolism, climacteric symptoms, and bleeding. Eur J Clin Invest 21; 601-607. 3.      Hansen MA, Hassager C, Overgaard K. Marslew U, Riis BJ, Christiansen C 1990. Dual energy x-ray absorptiometry: A precise method of measuring bone mineral density in the lumbar spine. J Nucl Med 31: 1156-1162. 4.      Orwoll ES. Oviatt SK. The Nafarelin. Bone Study Group 1991. Longitudinal precision of dual-energy X-ray absorptiometry in a multicenter study. J Bone Miner Res 6: 191-197. 5.      Kelly TL. Crane G.. Baran DT 1994. Single X-ray absorptiometry of the forearm; Precision, correlation and reference data. J Bone Miner Res (in press). 6.      Hassager C, Gotfredsen A, Jensen SB, Christiansen C 1991. The impact of measurement errors of the diagnostic value of bone mass measurements. Theoretical considerations. Osteoporosis Int 1:250-256. 7.      Ho CP, Kim RW, Schaffler MB. Sartoris DJ 1990. Accuracy of dual-energy radiographic absorptiometry of the lumbar spine: Cadaver study. Radiology 176:171-173. 8.      Harrbo J, Gotfredsen A, Hassager C, Christiansen C 1991. Validation of body composition measurements by dual energy x-ray absorptiometry (DEXA). Clin Physiol 11:331-341. 9.      Svendsen OL, Marslew U, Hassager C, Christiansen C 1992. Measurement of bone mineral density of the proximal femur by two commercially available dual energy x-ray absorptiometry systems. Eur J Nucl Med 19: 41-46. 10.  Marslew U, Overgaard K, Riis BJ, Christiansen C 1992. Two new combinations of postmenopausal bone loss: Long-term effects on bone. Calcium and lipid metabolism, climacteric symptoms, and bleeding. Obstst Gynecol 79: 202-210. 11.  Hassager C, Borg J, Christiansen 1989. Measurement of the subcutaneous fat in the distal forearm by single photon absorptiometry. Metabolism 38: 159-165. 12.  Romeis B 1968. Mikroskopische technik. Munchen. Wien: R Oldenburg Verlag, p.215. 13.  Gotfredsen A, Podenphant J. Norgaard H, Nilas L Nielsen VAH, Christiansen C 1988. Accuracy of lumbar spine bone mineral content by dual-photon absorptiometry. J Nucl Med 29: 248-254. 14.  Nilas L, Norgaard H, Podenphant J, Gotfredsen A, Christiansen C 1987. Bone compostition in the distal foream. Scand J Clin Lab Invest 47: 41-46. 15.  Svendsen OL, Hassager C, Bergmenn I, Christiansen C 1993. Measurement of abdominal and intra-abdominal fat in postmenopausal women by dual energy x-ray absorptiometry and anthropometry : comparison with computed tomography, Int J Obesity 17: 45-51. 16.  Hangarten TN, Johnson CC 1990. Influence of fat on bone measurements with dual energy absorptiometry. Bone Miner 9:71-81. 17.  Nilas L, Hassager C, Christiansen C 1988. Long-term precision of dual photon absorptiometry in the lumbar spine in clinical settings. Bone Miner 3: 305-315. 18.  Slosman DO, Rizzoli R, Donath A, Bonjour JPL 1990. Vertebral bone mineral density measured laterally by dual energy x-ray absorptiometry. Osteoporosis Int 1:23-29. 19.  Watts NB, Harris ST, Genant HK Wasnich RD, Miller PD, Jackson RD, Licata AA, Ross P, Woodson GC, Yanover MJ, Mysiw WJ, Kohse L, Rao MB, Steiger P, Richmond B, Chesnut CH III 1990. Intermittent cyclical etidronate treatment of postmenopausal osteoporosis. N Engl J Med 323: 73-79. 20.  Riggs BL, Hodgson SF, O Failon WM, Chao EYS, Wahner HW, Muhs JM, Cedel SL, Melton LJ III 1990. Effect of fluoride treatment of the fracture rate in postmenopausal women with osteoporosis. N Engl J Med 322: 802-809. 21.  Overgaard K, Hansen MA, Jensen SB, Christiansen C 1992. Effect of saleatonin given inrranasally on bone mass and fracture rates in established osteoporosis: a does-response study. Br Med J 305: 556-561.     Address reprint requests to : Dr. Ole Lander Svendsen Center for Clinical and Basic Research Ballentp Byvej 222 DK-2750 ballenip. Denmark   Received in original form April 25, 1994: in recised form January 16,1995; Accepted January 17, 1995.