Journal of Clinical Densitometry, vol. 9, no. 1, 15–21, 2006� Copyright 2006 by The International Society for Clinical Densitometry1094-6950/06/9:15–21/$32.00DOI: 10.1016/j.jocd.2006.05.003

Position Statement

Skeletal Sites for OsteoporosisDiagnosis: The 2005 ISCD Official Positions

Didier Hans,*,1 Robert W. Downs Jr.,2 Francois Duboeuf,3 Susan Greenspan,4

Lawrence G. Jankowski,5 Gary M. Kiebzak,6 and Steven M. Petak7

1Geneva University Hospital, Geneva, Switzerland; 2Virginia Commonwealth University School of Medicine, Richmond,VA; 3Hopital Edouard Herriot, Lyon, France; 4University of Pittsburgh, Pittsburgh, PA; 5Center for Arthritis and

Osteoporosis, Morton Grove, IL; 6St. Luke’s Episcopal Hospital, Houston, TX; and 7Texas Institute for ReproductiveMedicine, Houston, TX

Abstract

The International Society for Clinical Densitometry (ISCD) has developed Official Positions to assist healthcareproviders in addressing some of the issues inherent with the use of bone mineral density (BMD) assessed by dual-energy X-ray absorptiometry (DXA) to diagnose osteoporosis, apply World Health Organization (WHO) T-scoreclassifications, and monitor BMD changes over time. Differences exist, however, between the ISCD Official Positionstatement and that of the International Osteoporosis Foundation with respect to WHO criteria for skeletal sites. Con-sequently, a subcommittee of the ISCD was directed to address the application of the WHO classifications to specificskeletal sites and regions of interest. In 2005, the ISCD Position Development Conference reviewed the findings andprepared Official Positions, which address whether or not: (1) the lowest T-score of the total proximal femur, femoralneck, trochanter, and spine should continue to be used for diagnosis; (2) the WHO classification may be applied toa single vertebral body T-score; and (3) the ISCD should endorse the use of the National Health and Nutrition Ex-amination Survey database for proximal femur T-score derivation. The resulting ISCD Official Positions, with theircorresponding rationales and evidence are provided here, as well as questions that will need to be addressed in thefuture.

Key Words: Bone mineral density; Densitometry; Standards; Dual-energy X-ray absorptiometry; DXA; Imaging;Fracture assessment; Position; Guideline.

Introduction

Bone mineral density (BMD) assessed by dual-energy X-ray absorptiometry (DXA) is used to diagnose osteoporosis,assess fracture risk, and monitor changes in BMD overtime. The World Health Organization (WHO) has proposeda set of operational criteria to define osteoporosis in postmen-opausal Caucasian women (1). The BMD value of an

*Address correspondence to: Didier Hans, PhD, MBA, Head ofR&D, Nuclear Medicine Division, Geneva University Hospital,Rue Micheli-du-Crest, 24, 1211, Geneva 14, Switzerland.

15

individual is expressed in terms of the number of standard de-viations from the mean of a healthy young-adult referencepopulation, commonly referred to as the T-score (1). Osteopo-rosis has been defined by a T-score that is equal to or less than-2.5. This classification was established based on referencevalues of BMD measured at the lumbar spine, hip* (also re-ferred to as proximal femur), or forearm. The socio-econom-ical emphasis on hip fractures, however, together with studiesshowing that BMD measured at the proximal femur has the

*Note: The term ‘hip’ and ‘proximal femur’ are used interchangeably in this

paper.

16 Hans et al.

strongest association with hip fracture (2), have focused someclinical guidelines for the assessment of individual patients onBMD measurements made by DXA at the hip (3–5). Further-more, the WHO working group did not specify how manyskeletal sites to measure, or which region(s) of interest(ROI) within a skeletal site should be used for diagnosis.

To help healthcare providers adequately utilize the multi-plicity of skeletal sites measurable by DXA, the ISCD hasrecommended that BMD should be measured for the purposeof diagnosing osteoporosis at two preferred skeletal sites, thehip and lumbar spine. A third site (33% or one-third radius ofthe non-dominant forearm) should be investigated if technicalproblems arise at any of these two primary sites (6). TheISCD recommended that osteoporosis be diagnosed on thebasis of the lowest T-score for BMD found at the spine, totalhip, femoral neck, and trochanter, and the distal one-third ofthe radius, if measured (6). This position differs from that ofthe International Osteoporosis Foundation (IOF), which rec-ommended the use of WHO criteria only at the hip, statingthat in all situations this skeletal site would predict osteopo-rotic fractures as well as for any other skeletal sites (3). Whiledivergence exists in handling discordances among skeletalsites, technical improvements in DXA open new perspectivesin the interpretation of DXA scans. The improvement in res-olution of the DXA scan renders the search for artifacts morereliable, enabling the exclusion of a single vertebra, for exam-ple. Awareness of technical differences among manufacturers,coupled with an increase in patient mobility (e.g., movingfrom one city to another), has resulted in manufacturers rec-ognizing the importance of finding equivalency links in orderto convert BMD values measured on one DXA system intoa value that is comparable to another. However, even withprogress in this area, problems arise. For example, it is com-mon to observe physicians applying the operational WHOdefinition of osteoporosis to any skeletal region of interest(ROI), or to convert densitometric GE data to Hologic datawithout careful consideration, although this is not recommen-ded by any medical society. Potential consequences of suchapproaches include an important change in the prevalenceof osteoporosis according to the choice of the skeletal siteschosen, and an obvious misdiagnosis of the patient (7), possi-bly leading to a decrease in the credibility of the field of bonedensitometry (3). It is therefore crucial that along with ad-vances in DXA technology, practitioners be given updatedrecommendations to ensure correct use of DXA in clinical pa-tient management.

The following questions regarding skeletal site selectionand reference databases were addressed at the 2005 ISCD Po-sition Development Conference, held in Vancouver, BritishColumbia, Canada, the results of which are presented here.

� Should the lowest T-score of the total proximal femur,femoral neck, trochanter, and spine continue to be usedfor diagnosis?� Should the ISCD endorse use of the National Health and

Nutrition Examination Survey (NHANES) database forproximal femur T-score derivation?

Journal of Clinical Densitometry

� Can the WHO classification be applied to a single verte-bral body T-score?

Methodology

The methods used to develop, and the grading system ap-plied to these ISCD Official Positions is presented in detail inthe Executive Summary that accompanies this paper. Briefly,all Positions were graded on quality of evidence (good, fair,poor), strength of the recommendation (A, B, or C, whereA is a strong recommendation supported by the evidence, Bis a recommendation supported by the evidence, and C isa recommendation supported primarily by expert opinion),and applicability (worldwide or variable limited).

I. Central DXA for Diagnosis of Osteoporosis

ISCD Official Positions:

� The WHO international reference standard for osteoporo-sis diagnosis is a T-score of -2.5 or less at the femoralneck:B The reference standard from which the T-score is calcu-

lated is the female, white, age 20–29 years NHANESIII database.

Grade: Good-A-1

� Osteoporosis may be diagnosed in postmenopausalwomen and in men age 50 and older if the T-score ofthe lumbar spine, total hip, or femoral neck is -2.5 orless*:B In certain circumstances the 33% radius (also called 1/3

radius) may be utilized.

Grade: Good-A-2

*Note: Other hip regions of interest, including Ward’sarea and the greater trochanter, should not be used for di-agnosis. Application of recommendation may vary ac-cording to local requirements.

Fracture Risk Assessment

� A distinction is made between diagnostic classificationand the use of BMD for fracture risk assessment.� For fracture risk assessment any well-validated technique

can be used, including measurements of more than onesite, where this has been shown to improve the assessmentof risk.

Grade: Good-A-1.

Rationale

In the past, the ISCD Official Position was to use the low-est T-score of the total proximal femur, femoral neck, tro-chanter, and posterior/anterior spine for the diagnosis ofosteoporosis. In light of current data, the ISCD Official Posi-tion has been updated.

Volume 9, 2006

Skeletal Sites for Osteoporosis Diagnosis 17

Although most studies have treated hip fractures as a ho-mogeneous condition (2,8,9), they can be separated into frac-tures of the cervical region (femoral neck fractures,intracapsular fractures), and fractures of the trochanteric re-gion (extracapsular). Mautalen et al recommended that thesetwo main types of hip fractures be considered independentlyin epidemiological or clinical studies. Women with trochan-teric fractures have more severe and generalized bone loss,especially in the trabecular component. Cervical fracturesseem to be more related to proximal femur morphology(10,11). There is also evidence suggesting that the mecha-nisms are different, possibly due to the trochanteric regionhaving a greater proportion of trabecular bone than the cervi-cal region (50% vs. 30%, respectively) (12). Uitewaal et alanalyzed bone structure in both types of hip fractures and re-ported significantly lower trabecular bone volume and surfacedensity in patients with trochanteric fractures than in patientswith cervical fractures (13). A significant difference in mor-phology and density between trochanteric and cervical hipfractures has been reported by Duboeuf et al in the EPIDOSprospective study (14). The morphologic difference has beenconfirmed by Partanen et al who reported that the dimensionsof the upper femur and the pelvis measured from radiographsare useful in the evaluation of hip fracture risk and fracturetype (15). Michaelson et al suggested that height and hor-monal factors affect the risk of the two types of hip fractures,and concluded that each fracture type should be consideredindependently when planning preventive intervention (16).This is even more important as it has been found that in recentyears there appears to be an increase in the incidence of tro-chanteric hip fractures compared with cervical fractures,(17,18) and that trochanteric fractures have been associatedwith up to twice the short-term mortality of cervical fracturesin the elderly (8).

Substantially lower femoral BMD amongst women withtrochanteric fractures as compared with cervical fractureshas been reported (19–22). Further, some studies have foundthat only trochanteric BMD is a better predictor of trochan-teric fractures as compared with cervical fractures. Othershave found no difference in BMD between these fracturetypes (23–25). Although the evidence concerning BMD fordifferent types of hip fracture is somewhat conflicting, it isgenerally agreed that BMD of the femoral neck best predictscervical fractures, whereas the trochanteric site best predictstrochanteric fractures, and that the total hip BMD best reflectsthe risk of any type of hip fracture. Without considering thetype of hip fracture, an in-house analysis performed on theEPIDOS cohort (in-house unpublished data: courtesy ofHans D) showed that the best discrimination between patientswith and without hip fractures was achieved by consideringthe lowest T-score of any hip measurements (i.e., femoralneck, greater trochanter and total hip), followed by the greatertrochanter. When discriminating all osteoporotic fracture,Hans D found that all hip sites were similar. In another studycomparing T-scores from different hip ROI in differentiatingpostmenopausal women with and without vertebral fracture,Vokes et al concluded that the best discrimination was

Journal of Clinical Densitometry

achieved with the femoral neck, the greater trochanter, orthe lowest of any hip ROI measurement (26). Similarly, ananalysis of the Study of Osteoporotic Fractures (SOF) cohortby Lu et al concluded that combining the results of BMD as-sessments at more than one ROI from a single hip scan signif-icantly increases prediction of hip and/or spine fracture risk,and elevates the relative risk with increasing numbers oflow BMD ROI (27).

Discussion

When extrapolating these data for use in clinical practice,differences in recommendations have arisen concerning thehip ROI to be used (femoral neck, trochanter, or total hip) be-tween the IOF (3) and the ISCD (28). On the technical aspect,one could argue that the lack of complete ROI agreementamongst manufacturers may impact the choice of hip sub-re-gions in the diagnosis of osteoporosis. Indeed, at this stage,the total hip ROI is one of the most consistent amongst man-ufacturers; however, this issue has been partially taken intoaccount by the introduction of the NHANES III database(29,30) for the total region.

In fact, the question, ‘‘Should the lowest T-score of the to-tal proximal femur, femoral neck and trochanter continue tobe used for diagnosis?’’ could even be extended to, ‘‘Shouldthe lowest T-score of the proximal femur or spine continue tobe used for diagnosis?’’

Recently, Kanis et al stated that the selection of patients onthe basis of a minimum value from two or more tests willmostly increase the apparent prevalence of osteoporosis, butnot improve fracture prediction (gradient of risk) (31). Froma theoretical point of view, there is unlikely to be any benefitfrom combining information from different types of bonedensitometry measurements unless they are completely unre-lated (32). These findings are at apparent variance with resultsfrom the SOF study, (33) which showed an increase in risk inindividuals with low BMD at more than one site. These ap-parent divergent findings are reconcilable. The identificationof individuals who have more than one site with low BMDwill identify those at higher risk for osteoporosis than patientswith a low BMD at one site: they lie at the tail end of the fre-quency distribution curve for fracture risks, and will be foundwith a much lower prevalence than individuals with an abnor-mality at one site only. Thus, specificity is traded-off onpoorer sensitivity, but the prognostication of that individualis enhanced. In contrast, when using cut-off points for diag-nostic classification, the trade-offs result in a larger propor-tion of the population selected.

The apparent divergent views on the topic are mostlyrelated to confusion of terms. What is the most relevantapproach to diagnose osteoporosis? Is it the gradient of riskor the number of patients selected based on a threshold ap-proach? In fact, these concepts are not in conflict with eachother. Indeed, a distinction is made between diagnostic clas-sification and the use of BMD for fracture risk assessment.For fracture risk assessment, any well-validated techniquecan be used, including measurements of more than one

Volume 9, 2006

18 Hans et al.

skeletal site, which has been shown to improve the assess-ment of risk. In this case, the lowest T-score of measured skel-etal sites would be relevant and could be used. The femoralneck is the emerging international reference standard; how-ever, based on a review of the current medical literature (2),osteoporosis may still be diagnosed if the T-score of the lum-bar spine, total hip, or femoral neck is -2.5 or less, although itshould be recognized that different patients would be charac-terized as having osteoporosis (34). The present recommenda-tion removes the trochanter from consideration asa diagnostic site in an effort to help reconcile the ISCD guide-lines with those of the IOF, National osteoporosis Foundation(NOF) and WHO, and to aid in the transition to fracture prob-ability as a basis for therapeutic intervention.

Additional Questions for Future Research

The following suggestions for additional research arelisted below, stated in the form of questions, most of whichcould be investigated with retrospective analyses:

� From meta-analyses comprised of prospective studies,does using the lowest T-score amongst femoral sites im-prove the fracture prediction?� Which one of the hip sub-regions better responds to cur-

rent treatment?� Due to the heterogeneous nature of their size and place-

ment, is there a difference in rates of response of allsub-regions between manufacturers?

II. Should the ISCD Endorse the Use of theNHANES III Database for Proximal FemurT-score Derivation?

ISCD Official Position

� The NHANES III database should be used for T-scorederivation at the hip regions.

Grade: Poor C-1

Rationale

Differences in the measurement of BMD and T-scoresamongst manufacturers are multifactorial. While technical di-versities and differences in ROI exist, the selection of stan-dardized reference databases for the calculation of T-scorescould ameliorate some of the differences (7). Unfortunately,different manufacturers use the same names for the hipROI, when the actual ROI placements and sizes are different.Until standardized ROIs are created by all manufacturers, therole of standardized BMD (sBMD) remains predominantlythat of a research tool used for combining large numbers ofpatients for clinical trials (33,35). Conversely, significantstandardization efforts have already been performed for thelumbar spine. The downside of using this latter site isthe lack of a current ‘universal’ reference database. TheNHANES III database is currently the best proximal femur

Journal of Clinical Densitometry

database, with a large number of subjects (36–38). Althoughbased only on measurements with Hologic instruments, sub-ject selection was performed with a minimum of statisticalbias, and comparison of different cohorts should be per-formed according to this point, referring back to the differentpapers. This database is, however, not only a reference data-base, but also a common platform from which manufacturerscan better apply equations of equivalence (33,35).

Discussion

It has been suggested that the NHANES III database, ob-tained in the United States, is not necessarily valid in othercountries. Furthermore, if one took a BMD value for the totalhip that equalled 22.5 on a Hologic machine and converted itto sBMD, then took a value on a Norland or GE Healthcaremachine that also equalled 22.5 and converted it to sBMD,the values would not be equal; therefore the NHANES deriva-tion based on sBMD may not be a valid approach. More spe-cifically, Binkley et al showed that a recent implementation ofNHANES hip database by GE Healthcare resulted in largechanges in T-scores at the hip, due to errors in its calculation.These errors were mostly related to an incorrect use of thecorrection formula. When the formula was correctly applied,the discordance in T-scores was reduced (29). Kiebzak et alreported an agreement of more than 90% in patient classifica-tion based on T-scores between Hologic (Delphi) and GEHealthcare (Prodigy) based on either total hip using theNHANES III database, or the spine using a non-NHANESIII manufacturer-specific database (30).

The question about the use of NHANES III for proximalfemur T-score derivation could easily be answered if patientsand providers were not regularly changing systems. The lackof a uniform database would be of less significance ifattempts were made to repeat DXA studies on a device ofthe same manufacturer and model. However, patient mobilityrepresents an increasing challenge to health care. Conse-quently, it is not unusual to have a bone measurement per-formed on a specific DXA device and a follow-up scan onanother DXA of a different brand. The importance of havinga common platform for comparison on the basis of T-scorecalculation (in default of having similar ROI) is, therefore,desired.

Additional Questions for Future Research

A series of questions should be investigated, mostly in ret-rospective analyses:

� Should we use a different approach based on the percent-age of the young normal mean that would be more appro-priate if NHANES cannot be used?� Would a standardized set of ROI definitions and anatomic

positioning (set internal rotation and abduction angles)between manufacturers improve BMD concordance atthe femur?

Volume 9, 2006

Skeletal Sites for Osteoporosis Diagnosis 19

III. Can the WHO Classification Be Applied toa Single Vertebral Body T-score?

ISCD Official Position

� BMD based diagnostic classifications should not be madeusing a single vertebra.� If only one evaluable vertebra remains after excluding

other vertebrae, diagnosis should be based on a differentvalid skeletal site.

Grade: Fair-C-1.

� Anatomically abnormal vertebrae may be excluded fromanalysis if:B They are clearly abnormal and non-assessable within

the resolution of the system; orB There is more than a 1.0 T-score difference between the

vertebra in question and adjacent vertebrae.� When vertebrae are excluded, the BMD of the remaining

vertebrae is used to derive the T-score.

Grade: Fair-C-1.

Rationale

In premenopausal women, where hip fracture risk is low, spi-nal BMD predicts the risk of any fracture as well as, or betterthan, hip BMD (2,3). However, in later life, spine BMD mea-surements are often confounded by osteoarthrosis (39), whereasthe hip is much less affected by these changes (40–42). Thelatest guidelines published by the ISCD clearly state thatwhen measuring the lumbar spine by DXA one should use allevaluable vertebrae (L1-L4), and only exclude vertebral bodies(VB) that are affected by local structural change or artefacts. Inthe worst case scenario, it is acceptable to diagnose osteoporo-sis based on a minimum of two vertebrae (6).

In an attempt to challenge this position statement, Hansenet al (43) studied the impact of picking the lowest vertebralbody T-score in the diagnosis of lumbar osteoporosis in men,using high resolution technology. They concluded that in clas-sifying men with prior fractures as being osteoporotic, the low-est lumbar T-score demonstrated similar sensitivity andspecificity to that of the lowest hip or wrist T-score, whereasthe average L1-L4 and ISCD-determined T-scores performedwith less sensitivity but better specificity. The study, however,was cross-sectional, limited to men, and employed a Lunar Ex-pert, a discontinued model with exceptionally high image res-olution that more readily identifies vertebral bodies worthy ofexclusion (44). This position has also been challenged by twoother studies which have recommended using the single verte-bra with the lowest T-score (29,45). Conversely, Vokes et al re-ported no added advantage in using the lowest vertebra inspine fracture discrimination, compared with the total (46).

Discussion

One should also be cautious not to introduce bias by a sys-tematic selection of the single lowest-density vertebral body,

Journal of Clinical Densitometry

with the assumption that the BMD of all other vertebral bod-ies is over-estimated due to artefacts. Indeed, it may well bethat the lowest vertebral body BMD is the consequence ofa metastasis or other structural problems that reduce the ap-parent BMD. In such a case, the specific vertebra with low ap-parent BMD should be excluded. The exclusion ofanatomically abnormal, non-assessable vertebrae wouldtherefore be reasonable as long as the resolution is sufficient,or if more than one SD is observed between individual verte-bral bodies (47). Since the ISCD recommendations are to beused by clinical densitometry facilities that may not be in-volved in the overall management of individual patients, itwill be important to have guidelines that apply generally. Be-cause use of the lowest vertebral body (VB) for diagnosis ofosteoporosis may require additional studies, such as plain X-ray to ensure that a single VB is in fact normal, the ISCD isreluctant to recommend the use of the lowest VB as a standardpractice for the diagnosis of osteoporosis.

Assuming that an assessable vertebra could be identified,the ISCD 2001 Position Development Conference suggestedthat it should not be used, as the measured area would betoo small. Indeed, the precision error is usually inversely cor-related to the measured area, meaning the smaller the areameasured, the larger the precision error will likely be. Onecan argue that small areas are used at other skeletal sitessuch as the femoral neck, so why not at the spine? (48) Whilethis point is well taken and in theory is indeed possible, theconsequences of using less precision, compared to what weexpect at the spine, will likely be multiple:

� Reducing the measured area will impact the least signifi-cant change (LSC) for single vertebral bodies (49).Jankowski et al (50) showed a strong correlation (R 5

20.89) between the root mean square (RMS)-standarddeviation (SD) and the measured area of the spine inmore than 100 patients (average age 71 years): the small-est measured area corresponded to the highest RMS error.Subsequently, the RMS coefficient of variation measuredat one single vertebra can equal twice the precision of thetotal L1-L4 (see Figure 1).� It would be necessary to increase the 22.5 T-score thresh-

old by the LSC amount prior to diagnosing osteoporosis,and raise the normal threshold by a similar amount if wewish to be consistent with the total lumbar spine diagno-sis. This would be very confusing to the referring physi-cian to have spinal data indicative of osteopenia, evenwith a single VB T-score of 20.7, or 22.8. In-housedata analysis based on more than 1,800 women (unpub-lished data: courtesy of Duboeuf F) shows that the errorof diagnosis based on one single vertebra versus the totalspine could lead to a misclassification of up to 5%. Onthis same analysis (subset of 900 patients) substantial var-iations in sensitivity and specificity amongst vertebraeversus total BMD were observed. The sensitivity rangedfrom 73% for L4 to 82% for L1 and L1-L4, while the cor-responding specificity ranged from 62% to 57%. Choos-ing the minimum T-score between L1 and L4 did not

Volume 9, 2006

20 Hans et al.

J

L1-L4

L1-L3

L2-L4

L1,L3-

L4

L1-L2,L

4

L1-L2

L1,L3

L3-L4

L2-L3

L2,L4

L1,L4 L1 L3 L2 L4

Possible combination of vertebrae

0.00

0.50

1.00

1.50

2.00

2.50

4 VB

3 VB2 VB

1 VB

Fig. 1. Precision as a function of the vertebrae level as well as different possible combinations of vertebrael.

improve the outcome, although a slight decrease of spec-ificity was reported (approximately 2%). These findingswere confirmed by Vokes et al (46). This misclassificationalso depends on the value of the SD, the normative data-base used, and the brand of the densitometer, as well asthe patient’s sex (51,52). Subsequently, the prediction ofosteoporotic fractures based on the total L1-L4 will differdepending on the single VB used. Sensitivity will increaseusing the lowest T-score of each of the lumbar vertebralbodies rather than the average value of the lumbar spine,but at the expense of reduced specificity (43).� The monitoring time interval (MTI) would change, mean-

ing that practitioners would need to adjust the time for thenext DXA examination according to the number of in-cluded vertebrae.

Given these data, we can sense that without consideringapplying the WHO definition or its concept to a single verte-bra, the decision to use a single specific vertebra should behandled with caution. An important general point is thatone will invariably find an abnormal result the more sitesone looks at.

Additional questions for future research

� Do prospective fracture data suggest that the lowest VBhas a different RR/SD than the total?� Does the vertebral body with the lowest T-score fracture

first?� Does L1 increase in BMD faster on current therapies than

L4?� Can any of the medications currently available increase

BMD in the spine by more than the LSC of a single ver-tebral body?

Summary

The ISCD Official Positions provided here address issueswith respect to the application of the WHO classificationsto specific skeletal sites and regions of interest. While theyare made in consideration of best available evidence, theyraise several questions that require further research. As

ournal of Clinical Densitometry

knowledge in the field of bone densitometry expands, the Of-ficial Positions can be periodically re-evaluated.

References

1. Kanis JA, Melton LJI, Christiansen C, Johnston CC, Khaltaev N.1994 The diagnosis of osteoporosis. J Bone Miner Res 9(8):1137–1141.

2. Marshall D, Johnell O, Wedel H. 1996 Meta-analysis of howwell measures of bone mineral density predict occurrence ofosteoporotic fractures [see comments]. BMJ (Clinical ResearchEd) 312(7041):1254–1259.

3. Kanis JA, Gluer CC. 2000 An update on the diagnosis andassessment of osteoporosis with densitometry. Committee ofScientific Advisors, International Osteoporosis Foundation.Osteoporos Int 11(3):192–202.

4. Kanis JA, Oden A, Johnell O, Jonsson B, de Laet C, Dawson A.2001 The burden of osteoporotic fractures: a method for settingintervention thresholds. Osteoporos Int 12(5):417–427.

5. Black DM, Steinbuch M, Palermo L, Dargent-Molina P,Lindsay R, Hoseyni MS, Johnell O. 2001 An assessment toolfor predicting fracture risk in postmenopausal women. Osteo-poros Int 12(7):519–528.

6. Lewiecki EM, Kendler DL, Kiebzak GM, et al. 2004 Specialreport on the official positions of the International Society forClinical Densitometry. Osteoporos Int 15(10):779–784.

7. Faulkner KG, von Stetten E, Miller P. 1999 Discordance in patientclassification using T-scores. J Clin Densitom 2(3):343–350.

8. Greenspan SL, Myers ER, Maitland LA, Kido TH,Krasnow MB, Hayes WC. 1994 Trochanteric bone mineral den-sity is associated with type of hip fracture in the elderly. J BoneMiner Res 9(12):1889–1894.

9. Kelsey JL, Hoffman S. 1987 Risk factors for hip fracture. [edi-torial]. N Engl J Med 316(7):404–406.

10. Mautalen CA, Vega EM. 1993 Different characteristics of cervicaland trochanteric hip fractures. Osteoporos Int 3(Suppl 1):102–105.

11. Mautalen CA, Vega EM, Einhorn TA. 1996 Are the etiologies ofcervical and trochanteric hip fractures different? Bone 18(3Suppl):133S–137S.

12. Riggs BL, Wahner HW, Seeman E, et al. 1982 Changes in bonemineral density of the proximal femur and spine with aging: dif-ferences between the postmenopausal and senile osteoporosissyndromes. J Clin Invest 70:716–723.

13. Uitewaal PJ, Lips P, Netelenbos JC. 1987 An analysis of bonestructure in patients with hip fracture. Bone Miner 3(1):63–73.

14. Duboeuf F, Hans D, Schott AM, et al. 1997 Different morpho-metric and densitometric parameters predict cervical and

Volume 9, 2006

Skeletal Sites for Osteoporosis Diagnosis 21

trochanteric hip fracture: the EPIDOS Study. J Bone Miner Res12(11):1895–1902.

15. Partanen J, Jamsa T, Jalovaara P. 2001 Influence of the upperfemur and pelvic geometry on the risk and type of hip fractures.J Bone Miner Res 16(8):1540–1546.

16. Michaelsson K, Weiderpass E, Farahmand BY, et al. SwedishHip Fracture Study Group. 1999 Differences in risk factor pat-terns between cervical and trochanteric hip fractures. OsteoporosInt 10(6):487–494.

17. Nydegger V, Rizzoli R, Rapin CH, Vasey H, Bonjour JP. 1991Epidemiology of fractures of the proximal femur in Geneva: in-cidence, clinical and social aspects. Osteoporos Int 2(1):42–47.

18. Nagant de Deuxchaisnes C, Devogelaer JP. 1988 Increase in theincidence of hip fractures and of the ratio of trochanteric to cer-vical hip fractures in Belgium. Calcif Tissue Int 42(3):201–203.

19. Nakamura N, Kyou T, Takaoka K, Ohzono K, Ono K. 1992Bone mineral density in the proximal femur and hip fracturetype in the elderly. J Bone Miner Res 7(7):755–759.

20. Vega E, Mautalen C, Gomez H, Garrido A, Melo L, Sahores AO.1991 Bone mineral density in patients with cervical and trochan-teric fractures of the proximal femur. Osteoporos Int 1(2):81–86.

21. Schott AM, Cormier C, Hans D, et al. 1998 How hip and whole-body bone mineral density predict hip fracture in elderly women:the EPIDOS Prospective Study. Osteoporos Int 8(3):247–254.

22. Nevitt MC, Johnell O, Black DM, Ensrud K, Genant HK,Cummings SR, Study of Osteoporotic Fractures ResearchGroup. 1994 Bone mineral density predicts non-spine fracturesin very elderly women. Osteoporos Int 4(6):325–331.

23. Chevalley T, Rizzoli R, Nydegger V, et al. 1991 Preferential lowbone mineral density of the femoral neck in patients with a recentfracture of the proximal femur. Osteoporos Int 1(3):147–154.

24. Bohr H, Schaadt O. 1983 Bone mineral content of femoral boneand the lumbar spine measured in women with fracture of thefemoral neck by dual photon absorptiometry. Clin Orthop RelatRes 179:240–245.

25. Dretakis EK, Papakitsou E, Kontakis GM, Dretakis K,Psarakis S, Steriopoulos KA. 1999 Bone mineral density, bodymass index, and hip axis length in postmenopausal Cretanwomen with cervical and trochanteric fractures. Calcif TissueInt 64(3):257–258.

26. Vokes T, Lovett J, Gillen D, Favus M. 2004 T-scores from differ-ent regions in subjects with and without vertebral fractures.J Clin Densitom 7:237.

27. Lu Y, Genant HK, Shepherd J, Zhao S, Mathur A, Fuerst TP,Cummings SR. 2001 Classification of osteoporosis based onbone mineral densities. J Bone Miner Res 16(5):901–910.

28. Abbott T, Barlas S, Brenneman S, Miller P. 2002 Minimal spineBMD predicts fracture better than total spine BMD. J Clin Den-sitom 5(Suppl):S49.

29. Binkley N, Kiebzak GM, Lewiecki EM, et al. 2005 Recalculationof the NHANES database SD improves T-score agreement and re-duces osteoporosis prevalence. J Bone Miner Res 20(2):195–201.

30. Kiebzak GM, Lewiecki EM, Miller P. 2004 Good DiagnosticAgreement Using T-score Between Dephi and Prodigy. J ClinDensitom 7:229.

31. Kanis JA, Seeman E, Johnell O, Rizzoli R, Delmas PD. 2005The perspective of the International Osteoporosis Foundationon the Official Position of the International Society for ClinicalDensitometry. J Clin Densitom 8(2):145–147.

32. Blake GM, Patel R, Knapp KM, Fogelman I. 2003 Does thecombination of two BMD measurements improve fracture dis-crimination? J Bone Miner Res 18(11):1955–1963.

33. Lu Y, Fuerst T, Hui S, Genant HK. 2001 Standardization of bonemineral density at femoral neck, trochanter and Ward’s triangle.Osteoporos Int 12(6):438–444.

Journal of Clinical Densitometry

34. Johnell O, Kanis J, Oden A, et al. 2005 A 2005 predictive valueof bone mineral density for hip and other fractures. J BoneMiner Res Online 20(7):1185–1194.

35. Hui SL, Gao S, Zhou XH, et al. 1997 Universal standardizationof bone density measurements: a method with optimal propertiesfor calibration among several instruments. J Bone Miner Res12(9):1463–1470.

36. Looker AC, Johnston CC Jr., Wahner HW, et al. 1995 Prevalenceof low femoral bone density in older US women from NHANESIII. J Bone Miner Res 10(5):796–802.

37. Looker AC, Orwoll ES, Johnston CC Jr., et al. 1997 Prevalenceof low femoral bone density in older US adults from NHANESIII. J Bone Miner Res 12(11):1761–1768.

38. Looker AC, Wahner HW, Dunn WL, et al. 1998 Updated data onproximal femur bone mineral levels of US adults. Osteoporos Int8(5):468–489.

39. Franck H, Munz M, Scherrer M. 1995 Evaluation of dual-energyX-ray absorptiometry bone mineral measurementdcomparisonof a single-beam and fan-beam design: the effect of osteophyticcalcification on spine bone mineral density. Calcif Tissue Int56(3):192–195.

40. Masud T, Langley S, Wiltshire P, Doyle DV, Spector TD. 1993Effect of spinal osteophytosis on bone mineral density measure-ments in vertebral osteoporosis. BMJ 307(6897):172–173.

41. Jones G, Nguyen T, Sambrook PN, Lord SR, Kelly PJ,Eisman JA. 1995 Osteoarthritis, bone density, postural stability,and osteoporotic fractures: a population based study. J Rheuma-tol 22(5):921–925.

42. Greenspan SL, Maitland-Ramsey L, Myers E. 1996 Classifica-tion of osteoporosis in the elderly is dependent on site-specificanalysis. Calcif Tissue Int 58(6):409–414.

43. Hansen KE, Vallarta-Ast N, Krueger D, Gangnon R,Drezner MK, Binkley N. 2004 Use of the lowest vertebralbody T-score to diagnose lumbar osteoporosis in men: is ‘‘cherrypicking’’ appropriate? J Clin Densitom 7(4):376–381.

44. Ryan PJ, Evans P, Blake GM, Fogeman I. 1992 The effect ofvertebral collapse on spinal bone mineral density measurementsin osteoporosis. Bone Miner 18(3):267–272.

45. Sahota O, Pearson D, Cawte SW, San P, Hosking DJ. 2000 Si-te-specific variation in the classification of osteoporosis, andthe diagnostic reclassification using the lowest individual lumbarvertebra T-score compared with the L1-L4 mean, in early post-menopausal women. Osteoporos Int 11(10):852–857.

46. Vokes TJ, Gillen DL, Lovett J, Favus MJ. 2005 Comparison ofT-scores from different skeletal sites in differentiating postmen-opausal women with and without prevalent vertebral fractures.J Clin Densitom 8(2):206–215.

47. Hansen KE, Binkley N, Christian R, Vallarta-Ast N, Krueger D,Drezner MK, Blank RD. 2005 Interobserver reproducibility of criteriafor vertebral body exclusion. J Bone Miner Res 20(3):501–508.

48. Schneider D, Bettencourt R, Barrett-Connor E. 2002 Should oneor more vertebral levels be used for fracture risk assessment.J Bone Miner Res 17(Suppl 1):S414.

49. Blake GM, Jagathesan T, Herd RJ, Fogelman I. 1994 Dual X-rayabsorptiometry of the lumbar spine: the precision of paired ante-roposterior/lateral studies. Br J Radiol 67(799):624–630.

50. Jankowski L, O’Brien-Schmack E, Broy S. 2003 What is the Magni-tude of Precision Error at the PA Spine by Central DXA When Dis-contiguous Vertebal Bodies are Selected. J Clin Densitom 6:185.

51. Smith J, Prestwood C, Dauser D, Kleppinger A, Kenny A. 2004Discordance in spine bone density of elderly men and women.J Clin Densitom 7(2):231.

52. McMahon K, Kalnins S, Freund J, Pocock N. 2003 Discordancein lumbar spine T-scores and nonstandardization of standarddeviations. J Clin Densitom 6(1):1–6.

Volume 9, 2006