The effects of standardization and reference values on patient ...

2 downloads 0 Views 940KB Size Report
Classification for Spine and Femur Dual-Energy X-ray Absorptiometry. A. Simmons 1'2'3, D. E. Simpson 1, M. J. O'Doherty 1, S. Barrington 1 and A. J. Coakley l.
OsteoporosisInt (1997) 7:200-206 © 1997EuropeanFoundationfor Osteoporosisand the National OsteoporosisFoundation

Osteoporosis International

Original Article The Effects of Standardization and Reference Values on Patient Classification for Spine and Femur Dual-Energy X-ray Absorptiometry A. S i m m o n s 1'2'3, D. E. Simpson 1, M. J. O ' D o h e r t y 1, S. Barrington 1 and A. J. C o a k l e y l 1East Kent Osteoporosis Screening and Research Unit, Kent and Canterbury Hospital, Canterbury, Kent; 2Department of Clinical Neurosciences, Institute of Psychiatry, London; 3Department of Neuroimaging, Maudsley Hospital, London, UK

Abstract. The effect of two methods for standardizing dual-energy X-ray absorptiometry (DXA) measurements on patient classification by the T-score has been determined for a group of over 2000 patients. The methods proposed by the International DXA Standardization Committee and the European Community's COMAC-BME group were used in conjunction with young reference data from the major DXA manufacturers, the COMAC-BME group and the third US National Health and Nutrition Examination Survey (NHANES III). The two standardization techniques produced dissimilar classifications as measured by the kappa statistic (to = 0.34-0.90), especially for the femoral neck, with up to 24.3% of patients reclassified from osteopenic to normal and 18.6% reclassified from osteoporotic to osteopenic when the standardization method was changed. Considering the effects of both reference data and standardization techniques together, there was a wide variation of patient classification, with the number of patients classified as osteoporotic varying from 9.6% to 21.1% for the postero-anterior spine L2-4 region and from 2.3% to 27.6% for the femoral neck. The agreement between different classifications ranged widely, from very poor to excellent (~c = 0.02-0.98). The creation of standardized reference data must be an important priority in order to harmonize patient management using standardized BMD measurements. The choice of standardization technique, however, must be addressed in light of the results presented here.

Correspondence and offprint requests to: Dr Andrew Simmons, Neuroimaging Research, Institute of Psychiatry, De Crespigny Park, London SE5 8AF, UK. Fax: +44 17I 919 2477; e-mail: a.simmons @iop.bpmf.ac.uk,

Keywords: BMD; Dual-energy X-ray absorptiometry; Normal range; Osteoporosis; Standardized bone mineral density

Introduction Dual-energy X-ray absorptiometry (DXA) [1] is the most widespread technique for determination of bone mineral density (BMD) in vivo since it allows measurement of all the clinically relevant sites, is precise [1-3], has stable calibration [4], involves a low radiation dose [1,5,6] and is inexpensive. The lack of agreement between measurements of BMD made on different manufacturers' systems [4,7-10] is widely recognized as a difficulty with DXA [11]. Not only are multi-centre epidemiological or drug trials affected by these differences, but individual patient management may be complicated if a follow-up measurement is made on a different system from the original. This might be due to a machine replacement, geographical movement of the patient, or a change of service provider as a result of health reforms. Attempts to address this lack of agreement have been made both by the International DXA Standardization Committee (IDSC) [12-14] and the European Community's COMAC-BME programme [15-18] using different methodologies. The IDSC scanned the same group of women on one example of each manufacturer's scanner, and used a measurement of the final version of the European Spine Phantom (ESP) [19] to choose the absolute scaling of the standardization. The COMAC-BME methodology was based purely on measures of the ESP prototype [20] from a number of scanners for each manufacturer.

Effects of DXA Standardization Techniques on Patient Classification

201

Patient classification for DXA is made relative to reference data. An expert panel of the World Health Organization (WHO) has recently recommended that such classification should be made relative to young healthy adult reference data (the T-score) rather than age-matched reference ranges (e.g. the Z-score) claiming that "age-matched reference ranges . . . are flawed, because the incidence of osteoporosis would not rise with age even though bone mass was decreasing with age and the risk of fracture increasing" [21]. The Tscore is defined as the patient's BMD expressed in terms of the number of standard deviations from the young healthy adult BMD, whilst the Z-score is similarly defined in terms of standard deviations from the agematched mean. Percentages of age-matched means have also been used for patient classification. Although there is not complete consensus on the use of the T-score yet (e.g. [22]), it seems likely that it will be widely used. Previously it has been shown that differences between reference data can have an important effect on patient classification [23-26]. There are now published methods for standardizing BMD values [113-18] and a standard technique for patient classification (the T-score) [21]. The aim of this study was to evaluate whether the IDSC and COMAC-BME standardization techniques led to differences in patient classification for hip and spine data from over 2000 patients classified using reference data from the principal manufacturers of DXA equipment, the European Community COMAC-BME group and the third US National Health and Nutrition Examination Survey (NHANES III).

two standardization methods the following terminology is used here. BMD refers to the original measurements made on individual manufacturer' s systems, sBMD[IDSC] refers to standardized measurements determined using the approach of the IDSC [13], eBMD[IDSC] refers to Nortand equivalent BMD values calculated using the IDSC cross-calibration equations for the femoral neck, and sBMD[BME] refers to standardized measurements determined using the approach of the COMAC-BME group [18]. Both patient data and reference data from the three main manufacturers of DXA equipment (Hologic, Waltham, MA; Lunar, Madison, WI; and Norland, Fort Atkinson, WI) and the COMACBME group were standardized as follows. Posteroanterior (PA) L2-4 spine data were converted to sBMD[IDSC] values according to the relationships [13]:

Subjects and Methods Subjects and Scanning A sample of 2497 consecutive Caucasian female patients aged 22-90 years (mean 54 years) from the East Kent Osteoporosis Screening and Research Unit was considered. These patients were a mix of self-referrals and those referred by general practitioners and hospital consultants. Each patient had a postero-anterior spine L2-4 scan and most had hip scans using a Norland XR26 Mark II scanner (Norland, Fort Atkinson, WI) with 2.2 series software and positioning aids. The manufacturer's recommended daily quality control procedures were followed routinely to ensure the DXA system was within specification. A total of 2497 spine measurements and 2450 femoral neck measurements were acquired.

Standardized BMD Values Two techniques for standardization of BMD measurements have been proposed: one carried out by Genant et al. under the auspices of the IDSC [13] and one by the European COMAC-BME group [15-18]. In order to distinguish between measurements produced on individual manufacturer's systems and values produced by the

sBMD[IDSC] = 1.0761 Norland sBMD[IDSC] = 0.9522 Lunar sBMD[IDSC] = 1.0755 Hologic These spinal sBMD values have been implemented by the manufacturers, and are expressed in mg/cm2 instead of g/cm2 in order to distinguish them from uncorrected DXA results. Genant et al. [13] provide scanner cross-calibration equations for the femoral neck rather than standardization equations: Norland XR-26 Mark II = 0.961 Lunar - 0.037 Norland XR-26 Mark II = 1.030 Hologic + 0.058 These were used to calculate Norland equivalent values of bone density (eBMD[IDSC]). Neither standardization equations nor cross-calibration equations were provided for greater trochanter or Ward's triangle regions. Total hip standardization is likely to be implemented by the manufacturers in the near future. Values of sBMD[BME] were calculated for the postero-anterior spine L2-4 and femoral neck using the following expression: sBMD[BME] -- ln(3.5129) ln(3.5129- Hologic BMD) 0.3044 0.3044 sBMD[BME] - ln(4.8923) In(4.8923- LunarBMD) 0.2628 0.2628 sBMDIBME] --

Norland BMD 0.9317

In a minority of cases for Lunar systems, the COMACBME group found a better fit using linear regression than fitting by an exponential curve [18]. By contrast all Norland data were best fitted by linear regression and all Hologic data by an exponential curve. The term for Lunar above represents the exponential expression that provided the best fit for the large majority of Lunar systems. Standardization of peripheral BMD measurements has also taken place [t6]. Such measurements are less relevant than measurements of the axial skeleton, however, and for this reason axe not considered here.

202

A. Simmons et al.

Classification of' Patient BMD Measurements by Comparison of Standardized BMD Values with Standardized Reference Data According to the expert panel of the WHO three diagnostic categories of BMD measurements may be defined for adult women based on the T-score [21] as detailed in Table 1. The WHO also recommended that the term "established osteoporosis" (or "severe osteoporosis") is used to describe a value for BMD or bone mineral content that is more than 2.5 SD below the young adult value in the presence of one or more fragility fractures. Young reference data are provided by each of the principal manufacturers (Hologic, Lunar, Norland) with their systems. Norland supply different reference data for use in Europe and in the United States, whilst both Hologic and Lunar use the same reference data for Europe and the United States. The COMAC-BME group has published normal ranges collated from member countries of the European Community based on premenopausal women, postmenopausal women and women of all ages for the femoral neck [17], greater trochanter [17] and PA spine L2-4 [27]. Additionally, a recent paper has reported bone mineral levels for the proximal femur determined as part of NHANES III [28], a large study carried out using Hologic systems. The

mean and standard deviation of BMD for the three manufacturers and NHANES III axe given in Table 2. Peak BMD values were not explicitly given in the COMAC-BME publications and since non-linear fits were used, young data are not included in Table 2 for these reference ranges. Instead, thresholds separating normals, osteopenics and osteoporofics were calculated using published fits to the group of premenopausal women. The age at peak BMD was 20 years for the fit. For each of the sets of reference data the number of patients defined as normal, osteopenic and osteoporotic on the basis of standardized BMD values and T-scores was evaluated. The level of agreement between each set of classified data was quantified by calculating the kappa statistic, and by determining the concordance between pairs of data in terms of the percentage of patients identically classified. Patients classification for the IDSC and COMACBME standardization methods was compared for each set of reference data. Only the PA spine L2-4 and femoral neck regions were considered since the IDSC did not supply equations relating to the greater trochanter. The combined effect of varying standardization method and reference data was also determined.

Results Table 1. The three diagnostic categories of bone mineral density measurements defined for adult women by the expert panel of the World Health Organization based on the T-score [23] Category of BMD

Definition

Normal

A value of bone mineral density or bone mineral content that is not more than 1 SD below the young adult mean value A value for bone mineral density or bone mineral content that lies between 1 and 2.5 SD below the young adult mean value A value for bone mineral density or bone mineral content that is more than 2.5 SD below the young adult mean value

Osteopenia Osteoporotic

Table 2. Mean and standard deviation of BMD for young normals provided by the principal DXA manufacturers (Hologic, Lunar and Norland) and the NHANES III study Reference data

PA spine L2-4

Femoral neck

Hologic

1.079 (0.110) 1.200 (0.120) 1.085 (0.115) 1.164 (0.162) (-)

0.895 (0.100) 0.980 (0.120) 0.900 (0.120) 0.928 (0.131) 0.849 (0.109)

Lunar Norland Europe Nofland US NHANES III

Values are the mean (SD) in g/cm 2)

The percentage of women defined as normal, osteopenic and osteoporotic on the basis of T-scores calculated from standardized BMD measurements is given by Tables 3 and 4 for PA spine L2-4 and the femoral neck respectively. The agreement between classification produced using the different standardization methods and normal reference range data is given in terms of the kappa statistic for the spine by Table 5 and for the femoral neck by Table 6. The concordance between classifications in terms of the percentage of patients identically classified is given for the spine by Table 7 and for the femoral neck by Table 8. The agreement between classifications produced by the two standardization methods is excellent for the Hologic spine (~c = 0.90, concordance = 93.8%), but less good for the Lunar spine (to -- 0.60, concordance = 78.3%). There is poor agreement between classifications by the two standardization methods for the Hologic hip (to = 0.34, concordance = 57.1%) and the Lunar hip (~ = 0.37, concordance = 69.3%) however. There is perfect agreement for each Norland classification (to = 1.00, concordance = 100.0%), since all patient measurements originate from a Norland scanner, and changing standardization method results in a simple rescaling of both data and thresholds. The classifications following IDSC standardization show excellent agreement between Hologic, Lunar and Norland Europe (to = 0.94-0.98, concordance = 96.198.5%) for the spine. The agreement between the Norland US and Hologic, Lunar or Norland Europe classification is still good 0c = 0.76-0.80, concordance

203

Effects of DXA Standardization Techniques on Patient Classification Table 3. Patient classification for standardized PA spine L2-4 BMD measurements and the T-score using IDSC and COMAC-BME standardization methods and a variety of normal reference ranges Normal reference range

Standardization technique

Normal patients (%)

Patients with osteopenia (%)

Patients with osteoporosis (%)

Hologic

sBMD ~DSC] sBMD [BME] sBMD [IDSC] sBMD [BIvIE] sBMD [IDSC] sBMD [BME] sBMD [IDSC] sBMD [BME] sBMD [BME]

52.1 47,7 54.7 67.9 51,9 52.0 44.3 44,4 48.7

28.5 34.7 27.2 22.5 30.0 29.8 42.3 42.2 30.2

19.4 17.6 18.1 9.6 18.1 18.2 13.5 13.4 21.1

Lunar Norland Europe Norland US COMAC-BME

Totals may not equal 100% due to rounding.

Table 4. Patient classification for standardized femoral neck BMD measurements and the T-score using IDSC and COMAC-BME standardization methods and a variety of normal reference ranges Normal reference range

Standardization technique

Normal patients (%)

Patients with osteopenia (%)

Patients with osteoporosis (%)

Hologic

eBMD sBMD eBMD sBMD eBMD sBMD eBMD sBMD sBMD eBMD sBMD

32.5 56.8 55.9 80.2 58.1 58.1 53.5 53.6 57,6 46.9 71.9

39.9 34.2 35.5 17.4 35.9 35.9 40.3 40.2 32.1 39.3 25.0

27.6 9.0 8.7 2.3 6.0 6.0 6.2 6.2 10.3 13.8 3.1

Lunar Norland Europe Norland US COMACBME NHANES III

[IDSC] [BME] [IDSC] [BME] [IDSC] [BME] [IDSC] [BME] [BME] [IDSC] [BME]

Totals may not equal 100% due to rounding. Table 5. Kappa scores for comparisons of classifications using different standardization methods and normal reference range data for the posteroanterior spine Hologic

Hologic Lunar Norland Europe Norland US COMAC-BME

IDSC BME IDSC BME IDSC BME IDSC BME BME

Lunar

Norland Europe

Norland US

COMAC-BME

IDSC

BME

IDSC

IBME

IDSC

BME

IDSC

BME

BME

1.00

0.90 1.00

0.94 0.88 1.00

0.55 0.51 0.60 1.00

0.98 0.92 0.95 0.56 1.00

0.98 0.92 0.95 0.56 1.00 1,00

0.78 0.88 0.76 0.53 0.80 0.80 1.00

0.78 0.88 0.76 0.54 0.80 0.80 1.00 1.00

0.92 0.93 0.86 0.47 0.90 0.90 0.81 0.81 1.00

= 85.0-87.8%). For the hip, the IDSC classification shows poor agreement b e t w e e n Hologic and any other reference data (to = 0,28-0.57, concordance = 5 2 . 8 71.8%, with better agreement b e t w e e n the N H A N E S III reference data and other ranges (0.57-0.76, concordance = 7 1 . 8 - 8 5 . 9 % ) and good agreement b e t w e e n the r e m a i n i n g ranges (ic = 0.91, concordance = 95.1-95.2%).

The classifications following C O M A C - B M E standardization show good agreement for the spine between Hologic, N o r l a n d Europe, N o r l a n d US and the C O M A C B M E ranges (to = 0.80-0.93, concordance = 8 7 . 6 95.5%). The agreement b e t w e e n the L u n a r range and the Hologic, both N o r l a n d and C O M A C - B M E ranges is worse (~ = 0.47-0.56, concordance = 6 9 . 3 - 7 5 . 5 % ) . For

204

A. Simmons et al.

Table 6. Kappa scores for comparisons of classifications using different standardization methods and normal reference range data for the femoral neck Hologic

Hologic

IDSC BME IDSC BME IDSC BME IDSC BME BME IDSC BME

Lunar Norland Europe Norland US COMAC-BME NHANES III

Lunar

Norland Europe

Nortand US

COMAC-BME

NHANES III

IDSC

BME

IDSC

BME

IDSC

BME

IDSC

BME

BME

IDSC

BME

1.00

0.34 1.00

0.35 0.98 1.00

0.02 0,38 0.37 1.00

0.28 0.92 0.91 0.45 1.00

0.28 0.92 0.91 0.45 1.00 1.00

0.35 0.89 0.91 0.39 0.91 0.91 1.00

0.35 0.89 0.91 0.39 0.91 0.91 1.00 1.00

0.35 0.96 0.94 0.36 0.91 0.91 0,85 0.85 1.00

0.57 0.76 0.76 0.19 0.67 0.67 0.76 0.76 0.76 1.00

0.03 0.58 0.57 0.76 0.66 0.66 0.58 0.58 0.57 0.36 1.00

Table 7, Concordance between classifications using different standardization methods and normal reference range data for the postero-anterior spine in terms of the percentage of patients identically classified Hologic

Hologic Lunar Norland Europe Nofland US COMAC-BME

IDSC BME IDSC BME IDSC BME IDSC BME BME

Lunar

Norland Europe

Norland US

COMAC-BME

IDSC

BME

IDSC

BME

IDSC

BME

IDSC

BME

BME

100.0

93.8 100.0

96.1 92.5 100.0

74.4 71.8 78.3 100.0

98.5 95.3 97.2 75.5 100.0

98.7 95.1 97.2 75.5 99.8 100.0

86.3 92.5 85.0 72.5 87.8 87.6 100.0

86.3 92.5 85.0 72.7 87.8 87.6 99.8 100.0

94.9 95.5 91.0 69.3 93.8 93.8 88.0 88.0 100.0

Table 8. Concordance between classification using different standardization methods and normal reference range data for the femoral neck in terms of the percentage of patients identically classified Hologic

Hologic Lunar Norland Europe Norland US COMAC-BME NHANES III

IDSC BME IDSC BME IDSC BME IDSC BME BME IDSC BME

Lunar

Norland Europe

Norland US

COMAC-BME

NHANES III

IDSC

BME

1 D S C BME

IDSC

BME

IDSC

BME

BME

IDSC

BME

100.0

57.1 100.0

57.7 98.8 100.0

52.8 52.8 95.7 95.7 95.1 95.1 74.2 74.2 100.0 100.0 100.0

57.6 93.9 95.1 69.4 95.2 95.2 100.0

57.5 94,0 95.2 69.5 95.3 95.3 99.9 1130.0

57.6 97.9 96.7 69.4 95.2 95.2 91.8 91.9 100.0

71.8 85.3 85.9 55.2 81.0 81.0 85.8 85.7 85.8 100.0

36.1 79.0 78.4 90.9 83.3 83.3 78.5 78.6 78.7 64.3 100.0

34.8 69.9 69.3 100.0

the f e m o r a l neck there is p o o r a g r e e m e n t b e t w e e n the L u n a r range and the H o l o g i c , N o r l a n d Europe, N o r l a n d U S and C O M A C - B M E ranges (~c = 0.36-0.45, concordance = 6 9 . 4 - 7 4 . 2 % ) , better a g r e e m e n t b e t w e e n the N H A N E S III range and all other ranges (0.57-0.76, c o n c o r d a n c e = 7 8 . 6 - 9 0 . 9 % ) and e x c e l l e n t a g r e e m e n t

b e t w e e n the H o l o g i c , N o f l a n d Europe, N o r l a n d U S and C O M A C - B M E ranges (to = 0 . 8 7 - 0 . 9 8 , c o n c o r d a n c e = 91.9-97.9%). C o n s i d e r i n g the c o m b i n e d effect o f v a r y i n g standardization m e t h o d and reference data, there was a wide variation in a g r e e m e n t b e t w e e n the different classifica-

Effects of DXA Standardization Techniques on Patient Classification

205

tions both for the PA spine L2-4 region (~ = 0.47-0.98, concordance = 69.3-98.7%) and for the femoral neck (~c = 0.02-0.98, concordance = 34.8-98.8%).

these corrections to typical human subjects did not bring the human measurements into agreement, particularly at tow and high BMDs. The IDSC does not endorse a single phantom for the determination of standardized units [14] but states that a study on individuals spanning the clinical range of BMD in vivo is required. A drawback of the IDSC methodology, however, is its reliance on a single system for each manufacturer. The error in the estimate of sBMD due to intra-manufacturer variability is not taken into consideration. In fact, no error ranges are given for the parameters in the IDSC standardization equations. Peak bone mass was determined using a number of different methods tor the reference data considered here. For example, Hologic used the peak bone mass based on 5-year wide samples, whilst Lunar used the mean of women aged 20-39 years. There is still considerable debate in the literature regarding the age of peak bone mass. Whilst many studies would suggest the midthirties, others have suggested that femoral neck and lumbar spine bone mass accumulation can essentially be completed at ages as young as the end of the second decade [39-47]. The method used to determine peak bone density is likely to contribute to differences in patient classification. Variations in methods of subject recruitment, subject exclusion and the number of subjects used may additionally contribute to differences between reference data from the three principal manufacturers and the COMAC-BME group. The data presented here demonstrate that the two standardization techniques proposed by the IDSC and the European Community COMAC-BlVIE group produce dissimilar results, particularly for the femoral neck. Differences between reference data also affect the diagnosis of osteoporosis and osteopenia based on standardized BMD values and the T-score. We strongly advocate the creation of standardized reference data, especially for the femoral neck, perhaps on a geographical basis, in order to harmonize patient assessment using standardized BMD values.

Discussion It is clear that the two standardization methods proposed by the IDSC and the COMAC-BME group do not produce similar results. The agreement between Hologic spine classifications is excellent (tc = 0.90, concordance = 93.8%), that between Lunar spine classifications less good (to = 0.60, concordance = 78.3%) and both those for the Hologic femoral neck (~c = 0.34, concordance = 57.1%) and Lunar femoral neck: (to = 0.37, concordance = 69.3%) poor. There was perfect agreement for Nofland values since both patient and reference values underwent the same transform, it is likely, however, that if the data had been acquired on a Lunar or Hologic scanner and assessed using the Norland reference data thresholds, then similar differences in classification between the two standardization techniques would have been demonstrated. There are important differences between the two methodologies proposed by the IDSC and COMACBME groups for deriving standardization equations. The IDSC standardization is based on a cross-calibration study in vivo in which 100 women were scanned on a Norland XR-26 Mark II, a Lunar DPX-L and a Hologic QDR 20000. They found a high correlation between BMD values obtained on different systems for the spine, confirming other studies [29-36]. The IDSC used the intercept and slope of the patient's correlations and the value of the middle vertebra of the ESP as a reference point in a series of standardization formulas. By contrast, the COMAC-BME group employed only linear or exponential fits to measurements of the ESP prototype [20], providing weighted linear regression lines or weighted exponential curves from a number of systems from each manufacturer studied. At the time of writing the IDSC had given final approval to the standardization of PA spine measurements by DXA [14] as published by Genant et al. [13]. It had not given final approval to the standardization of femoral neck BMD, however, and was continuing work to expand standardization to other sites [141. Tothill [37] has criticized the application of the ESP phantom for standardization (as used by the COMACBME group) since relationships between BMD values for each vertebra scanned on different systems are not linear and observations for the highest density, in particular, are not close to the regression lines in vivo. Tothill notes that the COMAC-BME reference ranges have a wide variance and attributes part of this to standardization using the ESP. Nord [38] has used the data from Genant et al. [13] to evaluate the usefi~lness of the ESP in cross-correlating human data taken with different DXA instruments by using cm'ves generated from the ESP measurements to adjust results from one DXA system to match another. He found that applying

References 1. Cullum ID, Ell PJ, Ryder JP. X-ray dual-photon absorptiometry: a new method for the measurement of bone density. Br J Radiol 1989;62:587-92. 2. Johnson J, Dawson-Hughes B. Precision and stability of dualenergy X-ray absorptiometry measurements. Calcif Tissue Int 1991 ;49:174-8. 3. LeBlanc AD, Schneider VS, Engelbretson DA, Evans JH. Precision of regional bone mineral measurements obtained from total-body scans. J Nucl Med 1990;31:43-5. 4. Kelly TL, Slovik DM, Neer RM. Calibration and standardisation of bone mSneral densitometers. J Bone Miner Res 1989;4:663-9. 5. Laskey MA, Flaxman ME, Barker RW, Trafford S, Hayball MP, Lyttle KD, et al. Comparative performance in vitro and in vivo of Lunar DPX and Hologic QDR-1000 dual energy X-ray absorptiometers. Br J Radiol 1991;64:1023-9. 6. Lewis MK, Blake GM, Fogelman I. Patient dose in dual X-ray absorptiometry. Osteoporosis Int 1994;4:11-5.

206 7. Vainio P, Ahonen E, Leinonen K, Sievanen H, Koski E. Comparison of instruments for dual-energy X-ray bone mineral densitometry. Nucl Med Commun 1992;13:252-5. 8. Lai KC, Goodsitt MM, Murano R, Chestnut CH. A comparison of two dual-energy X-ray absorptiometry systems for spinal bone mineral measurement. Calcif Tissue Int 1993;50:203-8. 9. Morrita R, Orimo H, Yamamoto I, Fukunaga M, Shiraki M, Nakamurra T, et al. Some problems of dual-energy x-ray absorptiometry in the clinical use. Osteoporosis Int 1993;(Suppl 1):$87-90. 10. Svendsen OL, Marslew V, Hassager C, Christiansen C. Measurements of bone mineral density of the proximal femur by two commercially available dual eneregy X-ray absorptiometric systems. Enr J Nucl Med 1992;19:41-6. 11. Compston JE, Cooper C, Kanis JA. Bone densitometry in clinical practise. BMJ 1995;310:1507-10. 12. Nord RH. Work in progress: a cross-correlation study on four DXA instruments designed to culminate in inter-manufacturer standardization. Osteoporosis Int 1992;2:210-1. 13. Genant HK, Grampp S, Gluer CC, Faulkner KG, Jergas M, Engelke K, et al. Universal standardisation for dual X-ray absorptiometry: patient and phantom cross-calibration results. Bone Miner Res 1994;9:1503-14. 14. Steiger P. Standardization of spine BMD measurement [letter]. J Bone Miner Res 1995;10:1602-3. 15. Dequeker J, Reeve J, Pearson J, Bright J, Felsenberg D. Multicentre European COMAC-BME study on the standardisation of bone densitometry procedures. Technol Health Care 1993;1:127-31. 16. Pearson J, Ruegsegger P, Dequeker J, et aI. European semianthropomorphic phantom for the cross-calibration of peripheral bone densitometers: assessment of precision, accuracy and stability. Bone Miner 1994;27:109-20. t7. Pearson J, Dequeker J, Reeve J, Fetsenberg D, Henley M, et al. Dual X-ray absorptiometry of the proximal femur: normal European values standardised with the European spine phantom. J Bone Miner Res 1995;10:315-24. t8. Pearson J, Dequeker J, Henley M, Bright J, Reeve J, et aI. European semi-anthropomorphic spine phantom for the calibration of bone densitometers: assessment of precision, stability and accuracy. Osteoporosis Int 1995;5:174-84. 19~ Kalender WA, Fesenberg D, Genant HK, Nscher M, Dequeker J, Reeve J. The European spine phantom: a toot for standardization and quality control in spinal bone mineral measurements by DXA and QCT. Eur J Radiol 1995;20:83-92. 20. Kalender WA. A phantom for standardization and quality control in spinal bone mineral measurements by QCT and DXA: desi~ considerations and specifications. Med Plays 1992;19:583-6. 21. Kanis JA, Melton LJ, Christiansen C, Johnston CC, Khaltev N. The diagnosis of osteoporosis. J Bone Miner Res 1994;9:113741. 22. Eastell R, Peel NFA. Interpretation of bone density results. Osteoporos Rev 1994;2:1-3. 23. Laskey MA, Crisp AJ, Cole TJ, Compston JE. Comparison of the effect of different reference data in Lunar DPX and Hologic QDR-1000 dual-energy X-ray absorptiometers. Br J Radiol 1992;65:1124-9. 24. Pocock NA, Sambrook PN, Nguyen T, et al. Assessment of spinal and femoral bone density by dual X-ray absorptiometry: comparison of Lunar and Hologic instxuments. J Bone Miner Res 1991;7:1081-4. 25. Simmons A, Barrington S, O'Doherty MJ, Coakley AJ. DEXA normal reference range use within the UK and the effect of different normal ranges on the assessment of bone density. Br J Radiol 1995;68:903-9. 26. Simmons A, O'Doherty MJ, Barrington S, Coakley AJ. A survey of dual-energy x-ray absorptiometry normal reference ranges used within the United Kingdom and their effect on patient classification. Nucl Med Comm 1995;16:1041-53.

A. Simmons et al. 27. Dequeker J, Pearson J, Reeve J, Henley M, Bright J, Felsenberg D, et al. Dual X-ray absorptiometry: cross-calibration and normative reference ranges for the spine. Results of a European Community concerted action. Bone 1995;17:247-54. 28. Looker AC, Wahner HW, Dunn W%, Calvo MS, Harris TR, Heyse SP, et al. Proximal femur bone mineral levels of US adults. Osteoporosis Int 1995;5:389-409. 29. Lai KC, Goodsitt MM, Murano R, Chesnut CHIII. A comparison of two dual-energy x-ray absorptiometry systems of spinal bone mineral measurement. Calcif Tissue Int t992;50:203-8. 30. Reid DM, Lanharn SA, McDonald AG, Averell A, Fenner JAH, Boyle IT, Nuki G. Speed and comparability of three dual-energy x-ray absorptiometer (DXA) models. In: Overgaard K, Christiansen C (eds) Osteoporosis 1990, vol 2. Copenhagen: Osteopress, 1990:575-7. 31. Pocock W, Sambrooke P, Nguyen T, Kelly P, Freund J, Eisman J. Assessment of spinal and femoral bone density by dual x-ray absorptiometry: comparison of Lunar and Hologic instruments. J Bone Miner Res 1992;7:1081-4. 32. Arai H, Nagao K, Furumchi K. The evaluation of three different bone deusitometry systems: XR-26, QDR-1000 mad DPX. Image Technol Inform Display 1990;22:1-6. 33. Gundry GR, Miller CW, Ramos E, Moscona A, Stein JA, Mazess RB, et al. Dual-energy radiographic absorptiometry of the lumbar spine: clinical experience with two different systems. Radiology 1990;174:539-41. 34. Vainio P, Ahonen E, Leinonen K, Sievanen H, Koski E. Comparison of instruments for dual-energy x-ray bone mineral densitometry. Nucl Med Commun t992;13:252-5. 35. Laskey MA, Flaxman ME, Barger RW, et al. Comparative performance in vivo and in vitro of Lunar DPX and Hologic QDR-1000 dual energy x-ray absorptiometers. Br J Radiol 1991 ;64: t 023-9. 36. Tothill P, Fenner JAK, Reid DM. Comparisons between three dual-energy x-ray absorptiometers used for measuring spine and femur. Br J Radiol 1995;68:621-9. 37. Tothill P. Cross-calibration of DXA scanners for spine measurements. Osteopomsis Int 1995;5:410-11. 38. Nord RH. Performance of the European spine phantom: an evaluation from published data. Osteoporosis Int 1996;6:99. 39. Mazess RB, Barden HS, Ettinger M, et aL Spine and femur density using dual-photon absorptiometry in US white women. Bone Miner 1987;2:211-9. 40. Garn S, Rohmann CG, Wagner B.Bone loss as a general phenomenon in man. Fed Proc 1967;26:1729-36. 41. Goldsmith NF, Johnston JO, Picetti G, et al. Bone mineral in the radius and vertebral osteoporosis in an insured population. J Bone Joint Surg Am 1973;55:1276-93. 42. Riggs BL, Wahner HW, Melton LJ III, et al. Rates of bone loss in the appendicular and axial skeletons of women: evidence of substantial vertebral bone loss before menopause. J Clin Invest 1986;77:1487-91. 43. Geusens P, Dequeker J, Verstraeten A, et al. Age-, sex-, and menopause-related changes of vertebral and peripheral bone: population study using dual and single photon absorptiometry and radiogrammet~. J Nucl Med 1986;27:1540--9. 44. Bonjour JP, Theintz G, Law F, Slosman D, Rizzoti R. Peak bone mass. Osteoporosis Int 1994;4:7-13. 45. Recker RR, Davies KM, Hinders SM, et al. Bone gain in young adult women. JAMA 1992;268:2403-8. 46. Matkovic V, Jelic T, Wardlaw GM, et al. Timing of peak bone mass in Caucasian females and its implication for the prevention of osteoporosis. J Clin Invest 1994;93:799-808. 47. Sabatier JP, Guaydiersouquieres G, Laroche P, et al. Bone mineral acquisition during adolescence and early adulthood: a study in 574 healthy females 10-24 years of age. Osteoporosis Int 1996;6:141-8. Received for publication 25 March 1996 Accepted in revised form 8 November 1996