clinical evaluation of a phalangeal bone mineral density assessment system

Journal of Clinical Densitometry: Assessment of Skeletal Health, vol. 13, no. 3, 292e300, 2010� Copyright 2010 by The International Society for Clinical Densitometry1094-6950/13:292e300/$36.00
DOI: 10.1016/j.jocd.2010.04.001
Original Article

Clinical Evaluation of a Phalangeal Bone Mineral DensityAssessment System

Rajesh Patel,*,1 Glen M. Blake,2 Elleny Panayiotou,2 and Ignac Fogelman2

1Department of Biosurgery and Surgical Technology, Division of Surgery, Academic Bone Densitometry Unit,Imperial College London, London, UK; and 2Osteoporosis Screening and Research Unit, King’s College London,

Guy’s Campus, London, UK

Abstract

Re*A

BiosuBoneCharinUnited

Because osteoporosis is common and usually managed in primary care, there is a requirement for cheap and con-venient methods of measuring bone mineral density (BMD). AccuDEXA (Lone Oak Medical Technologies, Doyles-town, PA) is a tabletop dual-energy X-ray absorptiometry (DXA) device that performs BMD measurements of thehand in the middle phalanges of the third finger. The aims of this study were to (1) evaluate the use of AccuDEXA inUK women; (2) investigate the concordance between AccuDEXA T-scores and DXA T-scores for central (spine andhip) sites; (3) investigate the comparative response of AccuDEXA measurements to clinical risk factors for osteo-porosis. Measurements of phalangeal and central BMD were performed in 620 women referred by their family doc-tors for bone densitometry (group 1) and 159 healthy female volunteers (group 2). For 65 women in group 2, aged39 yr or younger, the mean Z-scores for AccuDEXA and the central sites calculated from US reference ranges wereconsistent with the expected value of 0, whereas for the 62 group 2 women, aged 50 yr or older, the mean Z-scoresfor AccuDEXA and the central sites were in the range 0.4e0.7 and were statistically significantly different from 0.In both group 1 and group 2, the AccuDEXA T-scores in older and younger women were systematically higher thanthose in the central sites by up to 1 unit. Of the 157 women aged 50 yr or older, with osteoporosis, based on theircentral DXA results, only 34 (22%) had an AccuDEXA T-score less than or equal to �2.5, whereas 76 (48%) hadosteopenia and 47 (30%) were normal based on their AccuDEXA T-scores. When assessed by the effect of clinicalrisk factors on Z-scores, both AccuDEXA and central BMD were affected to a similar extent. We conclude that theconventional World Health Organisation T-score criteria for the diagnosis of osteoporosis should not be applied toAccuDEXA measurements in UK women. Clinical risk factors for low BMD were found to affect AccuDEXA mea-surements to a similar extent as central BMD measurements. AccuDEXA measurements could, therefore, provide analternative method for identifying individuals with low bone mass, provided care is taken in interpreting T-scores,perhaps, through the use of device-specific thresholds.

Key Words: Bone mineral density; clinical risk factors; osteoporosis; phalanges.

disability, loss of quality of life, and mortality (1,2). Dual-
Introduction
Osteoporosis is a major public health concern in theUnited Kingdom and around the world. The fractures associ-ated with osteoporosis have been shown to cause considerable

ceived 12/21/09; Revised 03/11/10; Accepted 04/03/10.ddress correspondence to: Rajesh Patel, PhD, Department ofrgery and Surgical Technology, Division of Surgery, AcademicDensitometry Unit, Imperial College Faculty of Medicine,g Cross Hospital, Fulham Palace Road, London W6 8RF,Kingdom. E-mail: [email protected]

292

energy X-ray absorptiometry (DXA) has become the mostwidely used technique for the assessment of osteoporosisbased on measurements of bone mineral density (BMD) atthe spine and hip. The reasons for this choice include thefact that hip BMD is the best predictor of hip fracture risk(3,4), the use of spine BMD for monitoring treatment (5),and the widespread consensus that spine and hip BMDresults should be interpreted using the World Health Organi-sation (WHO) definition of osteoporosis (T-score��2.5)(6,7). However, an estimated two-thirds of patients with

mailto:[email protected]

Osteoporosis and Phalangeal BMD 293

osteoporosis are not diagnosed until they have suffered a frac-ture, highlighting the need for the wider availability of bonedensitometry services and increased awareness of how indica-tions for osteoporosis can be used to select individuals at riskof future fracture. Because osteoporosis is common and usu-ally managed in primary care, there is a requirement for cheapand convenient methods of evaluating BMD. A variety of dif-ferent types of X-ray devices for measuring peripheral sites inthe skeleton are available (8). These include systems for mea-surements in the forearm, heel, or hand, as well as radio-graphic absorptiometry and radiogrammetry methods forquantifying BMD from plain radiographs of the hand (9).

AccuDEXA (Lone Oak Medical Technologies, Doyles-town, PA) is a tabletop dual-energy computed digital absorpti-ometry device that performs BMD measurements of the handin the middle phalanges of the third finger. Previous studieshave demonstrated that BMD measurements using Accu-DEXA correlate well with DXA measurements of the spine,hip, and forearm, and other peripheral technologies (10). Theability of the AccuDEXA device to predict vertebral fractureshas also been demonstrated (11). The principal aim of thisstudy was to evaluate the use of AccuDEXA in a UK popula-tion, including making an assessment of the appropriateness ofusing the manufacturer’s US reference range to interpret scanfindings in UK women. Additional aims were to evaluate theconcordance between AccuDEXA T-scores and DXA T-scoresat the spine and hip and to compare the responsiveness ofAccuDEXA measurements to clinical risk factors for osteopo-rosis, with conventional DXA of the central skeleton.

Methods

Patient Population

The study population consisted of 779 women who hadBMD measurements of the middle finger using the Accu-DEXA device in addition to DXA measurements of spineand hip BMD. Women across a wide age range (20e84 yr)were included, and all subjects participating in the study com-pleted a self-administered questionnaire. Six hundred andtwenty of the subjects (group 1) were women, referred bytheir family doctor for routine bone density evaluation, whoconsented to have an additional measurement on AccuDEXA.The remaining 159 women (group 2) were healthy volunteersrecruited through a circular e-mail sent to all staff of King’sCollege London. The study was approved by the Guy’s Hos-pital Research Ethics Committee. DXA scans of the lumbarspine and left proximal femur were performed usinga QDR4500A (Hologic Inc., Bedford, MA). Of the 779women participating in the study, 261 had no clinical risk fac-tors for osteoporosis. Two hundred and twenty seven of thewomen were younger than 50 yr.

Demographic, T-Score, and Z-Score Data

Subjects’ mean weight, height, and body mass index(BMI 5 Wt (kg)/Ht (m)2) and the mean BMD, T-score, and

Journal of Clinical Densitometry: Assessment of Skeletal Health

Z-score for AccuDEXA, and the lumbar spine (L1eL4), fem-oral neck, and total-hip sites were examined by dividing thesubjects into the following 6 age ranges: !30, 30e39,40e49, 50e59, 60e69, and 70þ yr. Manufacturers’ referenceranges derived from the US populations were used to calcu-late the T-scores and Z-scores for the spine and AccuDEXAmeasurements. The US NHANES III reference range wasused to calculate the T-scores and Z-scores for the hip sites(12). The analysis was carried out separately for women ingroup 1 and group 2.

Short-Term Precision

Duplicate AccuDEXA examinations with repositioning ofthe subject’s hand were made in a subset of patients to eval-uate the reproducibility of the BMD measurements. Results ofthe precision study were expressed as the standard deviation(SD) of duplicate measurements and the root mean square(RMS) SD of the individual patient results divided by themean BMD to calculate the coefficient of variation (CV).

Relationship Between Spine and Hip T-Scores

For all 779 women participating in the study, the Accu-DEXA and the spine and hip T-score data were compared us-ing linear regression analysis and Bland-Altman plots (13) toestablish the relationship between phalange and central BMDmeasurements. Statistical significance was assumed if the pvalue was less than 0.05.

T-Score Concordance

The AccuDEXA and the spine and hip T-score measure-ments were used to evaluate the percentage of postmeno-pausal women from group 1, aged 50 yr or older, identifiedas osteoporotic on the basis of the WHO definition of osteo-porosis (T-score��2.5). The statistical significance of thedifferences in the number of osteoporotic subjects at eachsite was evaluated using Fisher’s exact test (13). The samedata were used to analyze the frequency of T-score discor-dance between the phalange measurements and those at thecentral sites. T-Score discordance is the observation that theinterpretation of a patient’s BMD results using the WHO cat-egories of normal, osteopenia, and osteoporosis varies be-tween different measurement sites. Results were consideredconcordant if measurements at the phalanges and the lowestT-score at the 3 central sites placed the subject in the sameWHO category. Discordance was classified as minor whenthe difference between sites was no more than 1 WHO cate-gory (i.e., normal instead of osteopenia, or osteopenia insteadof osteoporosis) and as major when 1 site was osteoporoticand the other normal.

Clinical Risk Factors

The aim of this part of the study was to compare Accu-DEXA BMD measurements with spine and hip BMD mea-surements in a large group of women, some with no clinicalrisk factors and others with 1 or more risk factors for osteo-porosis. Women in the study were placed in the following7 groups according to which clinical risk factor they

Volume 13, 2010

294 Patel et al.

had: (1) atraumatic fracture since the age of 25 yr; (2) reportof X-ray osteopenia; (3) predisposing medical condition oruse of therapy known to affect bone metabolism; (4) prema-ture menopause before the age of 45 yr or a history of amen-orrhea of longer than 6-mo duration; (5) maternal history offracture; (6) BMI ! 20 kg/m2; and (7) current smoking habit.

The first 6 risk factors are listed in the UK Royal Collegeof Physicians guidelines for the prevention and treatment ofosteoporosis (14). For the present study, a maternal historyof fracture at the spine, hip, or forearm was used ratherthan hip fracture alone. A current smoking habit was in-cluded, as it is listed in the European Foundation of Osteopo-rosis and Bone Disease guidelines (15) and the NationalOsteoporosis Foundation guidelines (16) for identifying indi-viduals at risk of fracture. Women who were currentlyprescribed or had previously taken treatment for osteoporosiswere not excluded from the analysis. The group consisting ofwomen with a predisposing medical condition or therapyknown to affect bone metabolism included women takingoral corticosteroid therapy for longer than 6 mo.

Of the 779 women participating in the study, 261 had noclinical risk factors for osteoporosis. The remaining 518women had 1 or more risk factors, including low-trauma frac-ture. The relationship between Z-score values and clinical riskfactors was investigated using multivariate regression analysisto calculate the regression coefficient associated with each ofthe 7 risk factors using the following equation:

Z � score5 b0þX

i

biRFi ð1Þ

In Eq. (1), each risk factor is represented by an independentvariable RFi, which is set to 1 if the risk factor is present ina subject and 0 otherwise. When the equation is solved forall 779 women, the constant term b0 represents the meanZ-score for the women with no clinical risk factors, whereasthe regression coefficient bi represents the mean change inZ-score associated with the ith risk factor.

Results

A total of 779 white female subjects aged 19e84 yr wereincluded in the study. Demographic data were initially

TablePatient Demographic D

Age group(yr) n Age (yr)

!30 61 25.5 (2.9)30e39 62 34.7 (3.1)40e49 104 45.5 (2.8)50e59 202 54.8 (2.7)60e69 209 64.7 (2.9)70þ 141 74.5 (3.5)

Figures show the mean and standard deviation for each age group.


evaluated in 2 groups: first, the subjects referred by their fam-ily doctors for routine bone densitometry (group 1: n 5 620);and second, the women who responded to the recruitmente-mail for healthy volunteers (group 2: n 5 159). For eachgroup, subjects were divided into the 6 age ranges listed inTable 1. Differences in height, weight, and BMI betweenthe 2 groups were not statistically significant, and the resultsshown in Table 1 are for all 779 subjects. Short-term precisionfor AccuDEXA BMD measurements based on duplicate stud-ies in 32 women gave a CV of 1.8%.

T-Score Differences Between AccuDEXAand Central Sites

The mean BMD figures and their population SD for Accu-DEXA, lumbar spine, femoral neck, and total-hip sites ineach of the 6 age groups are listed in Tables 2A and 2B forgroup 1 and group 2, respectively. The same data are shownin Fig. 1, plotted as graphs of mean T-score against age to-gether with the appropriate US reference ranges (manufac-turers’ or NHANES III) for comparison. For patients ingroup 1, aged 50 yr or older, the mean T-scores and standarderror of the mean (SEM) for AccuDEXA, spine, femoralneck, and total-hip sites were �0.47 (0.07), �1.47 (0.06),�1.44 (0.05), and �1.08 (0.05), respectively. For women ingroup 2, aged 50 yr or older, the corresponding mean T-scoreswere þ0.38 (0.2), �0.69 (0.18), �0.83 (0.15), and �0.35(0.14), respectively. For subjects younger than 50 yr, themean T-scores were þ0.10 (0.10), �0.85 (0.11), �0.85(0.09), and �0.67 (0.10) in group 1 and þ0.55 (0.11),�0.04 (0.10), �0.09 (0.10), and 0.13 (0.09) in group 2,respectively. These data indicate a large systematic differencein the mean T-score between AccuDEXA and the spine andhip sites affecting older and younger subjects in both groups.This is illustrated in Fig. 2, which shows the mean T-scoredifference between AccuDEXA and each central BMD siteplotted by age for the pooled groups. AccuDEXA T-scoreswere systematically higher than the spine and hip sites byup to 1 unit. With the exception of the difference betweenAccuDEXA and total-hip T-score for the younger-than-30 yrage group, all the differences were statistically significant( p ! 0.05).

1ata (All Subjects)

Height (cm) Weight (kg) BMI

165.4 (6.2) 62.3 (10.0) 22.7 (3.5)165.2 (6.0) 64.8 (14.3) 23.7 (5.2)162.9 (6.5) 66.3 (14.0) 25.0 (5.1)163.2 (6.6) 68.3 (13.1) 25.6 (4.8)160.4 (6.2) 66.5 (11.7) 25.8 (4.7)156.2 (6.0) 60.4 (11.1) 24.7 (4.1)

Volume 13, 2010

Table 2AGroup 1: AccuDEXA, Spine, Femoral Neck, and Total-Hip BMD for Each Age Group

Age group(yr) n AccuDEXA PA spine Femoral neck Total hip

!30 23 0.503 (0.072) 0.961 (0.142) 0.784 (0.110) 0.884 (0.144)30e39 35 0.515 (0.049) 0.955 (0.125) 0.754 (0.117) 0.847 (0.133)40e49 72 0.526 (0.066) 0.950 (0.141) 0.744 (0.117) 0.858 (0.133)50e59 162 0.525 (0.077) 0.927 (0.144) 0.738 (0.127) 0.857 (0.131)60e69 194 0.493 (0.078) 0.884 (0.153) 0.691 (0.108) 0.823 (0.137)70þ 134 0.438 (0.077) 0.837 (0.147) 0.625 (0.109) 0.736 (0.128)


Table 2BGroup 2: AccuDEXA, Spine, Femoral Neck, and Total-Hip BMD for Each Age Group

Age group(yr) n AccuDEXA PA spine Femoral neck Total hip

!30 38 0.529 (0.056) 1.031 (0.119) 0.852 (0.100) 0.955 (0.110)30e39 27 0.538 (0.067) 1.035 (0.111) 0.830 (0.137) 0.948 (0.126)40e49 32 0.564 (0.055) 1.062 (0.096) 0.830 (0.093) 0.971 (0.098)50e59 40 0.551 (0.084) 1.000 (0.149) 0.798 (0.120) 0.947 (0.109)60e69 15 0.493 (0.090) 0.918 (0.182) 0.695 (0.153) 0.806 (0.187)70þ 7 0.518 (0.052) 0.919 (0.115) 0.661 (0.134) 0.824 (0.170)



Z-Score Differences Between AccuDEXAand Central Sites

Figure 3A shows the mean Z-score for each DXA devicefor patients aged 39 yr or younger and 50 yr or older forgroup 1. The same plot for the women in group 2 is shownin Fig. 3B. For patients in group 1 aged 50 yr or older, themean Z-scores (SEM) for AccuDEXA, spine, femoral neck,and total-hip sites were þ0.22 (0.08), þ0.25 (0.06), þ0.07(0.05), and þ0.13 (0.05), respectively. For volunteers in group2 aged 50 yr or older, the corresponding mean Z-scores(SEM) were þ0.52 (0.14), þ0.67 (0.17), þ0.43 (0.15), andþ0.57 (0.13), respectively. For the 65 healthy women ingroup 2 aged 39 yr or younger (and hence representinga group of young normal subjects), the mean Z-scores for Ac-cuDEXA, spine, femoral neck, and total-hip sites were þ0.25(0.14), �0.04 (0.13), þ0.04 (0.13), and þ0.13 (0.12), respec-tively. For the 58 patients in group 1 in the same age range,the figures were �0.18 (0.14), �0.72 (0.17), �0.62 (0.14),and �0.60 (0.16), respectively.

Relationship Between AccuDEXA and CentralBone Mineral Density

Scatter plots of AccuDEXA T-score against lumbar spineand total-hip T-scores for all 779 women showed correlationcoefficients of r 5 0.64 and 0.59, respectively. When Bland-Altman plots were drawn, the mean differences between


AccuDEXA and spine and hip T-scores (95% limits of agree-ment) were as followsdspine BMD: þ1.00 (�1.66 to þ3.58)and total-hip BMD: þ0.69 (�2.06 to þ3.21).

T-Score Concordance Between AccuDEXAand Central Sites

Figure 4 shows the percentages of postmenopausal womenwith AccuDEXA, spine, femoral neck, and total-hip T-scoresless than or equal to �2.5 for the 3 age ranges 50e59, 60e69,and 70þ yr. For the AccuDEXA measurements, 1% of 50- to59-yr-olds, 6% of 60- to 69-yr-olds, and 23% of those aged70þ yr had T-scores less than or equal to �2.5. The equiva-lent figures for spine BMD were 13%, 23%, and 35%, respec-tively, and for femoral neck BMD, 6%, 10%, and 34%,respectively. The figures for total-hip BMD (2%, 7%, and17%) were similar to those for AccuDEXA. For all 552 post-menopausal women aged 50 yr or older, a total of 48 (8.7%)had an AccuDEXA T-score less than or equal to �2.5 com-pared with 126 (22.8%) at the spine, 80 (14.5%) at the fem-oral neck, and 43 (7.8%) at the total-hip sites. Thedifferences between AccuDEXA and the central sites werestatistically significant for the spine ( p ! 0.001) and femoralneck ( p 5 0.006) but not for the total hip ( p 5 0.50).

Table 3 shows the results of the analysis of the concor-dance between AccuDEXA T-scores interpreted using theWHO definitions of osteoporosis and osteopenia and the in-terpretation of the hip and spine T-scores. The analysis

Volume 13, 2010

Fig. 1. Mean T-score vs. age plot for (A) AccuDEXA; (B) PA spine; (C) femoral neck; and (D) total-hip bone mineral density(BMD) for subjects in the 2 study groups. Error bars show �1 standard error of the mean. Dashed curves show the age-specificmean T-score calculated from BMD reference ranges. AccuDEXA and spine reference ranges are the manufacturer’s range.Femoral neck and total-hip reference ranges are the NHANES III range. Asterisks show the statistical significance of the dif-ference of data points from the reference range (*p ! 0.05; **p ! 0.01; ***p ! 0.001).

296 Patel et al.

showed that, in 38% of the patients, there was concordancebetween the interpretation of the AccuDEXA and centralsites. Minor discordance was seen in 52% of the patientsand major discordance in 10% of the patients. For all 47 sub-jects with major discordance and 236 out of 255 subjects(93%) with a minor discordance, AccuDEXA T-scores placed

Fig. 2. Bar chart showing the mean difference in bone min-eral density T-score between AccuDEXA and central sites.Error bars show �1 standard error of the mean.


patients in a higher category than hip and spine T-scores, sothat, based on the WHO definition, the diagnosis of osteopo-rosis was less likely to be made. Of 157 women with osteo-porosis based on their central DXA results, only 34 (22%)had an AccuDEXA T-score less than or equal to �2.5,whereas 76 (48%) had osteopenia and 47 (30%) were normalbased on their AccuDEXA T-scores.

Clinical Risk Factors

For all 779 women in the study, multivariate regressionanalysis was used to estimate the mean Z-scores for eachrisk factor group (Fig. 5). Women with clinical risk factorsfor osteoporosis had significantly lower AccuDEXA BMDmeasurements compared with women without risk factors.The Z-score changes associated with the various risk factorswere similar for AccuDEXA and central BMD sites. Accu-DEXA Z-score changes associated with a history of atrau-matic fracture (�0.35), X-ray osteopenia (�0.41), anda BMI less than 20 kg/m2 (�0.37) were all statistically signif-icant compared with women with no risk factors.

When the multivariate regression analysis was carried outseparately for women younger than 50 yr and 50 yr or older,similar results to the entire group were obtained, with both

Volume 13, 2010

Fig. 3. Bar chart of the difference of the mean Z-scorefrom 0 for each DXA device for patients aged 39 yr or youn-ger and patients aged 50 yr or older for subjects in (A) group1 and (B) group 2. Error bars show �1 standard error of themean.

Fig. 4. Percentage of women with bone mineral density(BMD) T-score��2.5 for AccuDEXA and the centralBMD sites.


groups showing significantly reduced AccuDEXA Z-scores inwomen with a history of atraumatic fracture, X-ray osteope-nia, and a low BMI, compared with women with no riskfactors.

Discussion

The AccuDEXA peripheral DXA (pDXA) device used inthis study is commercially available with the recommendationthat the results should be interpreted using T-scores and Z-scores derived from the manufacturer’s US reference data.The objective of this study was to compare the peripheralbone density measurements of the phalanges with the BMDof the lumbar spine and hip and to determine the clinicalvalue of finger BMD measurements in identifying womenwith osteoporosis in a UK population using the referencedata provided by the manufacturer. Healthy women recruited


through the e-mail (group 2) were studied primarily to helpevaluate the comparative effect of clinical risk factors onthe Z-scores of older (�50 yr) and younger (!50 yr) women.

The short-term reproducibility of AccuDEXA measure-ments expressed as the CV was found to be 1.8%. This figureis similar to that reported by other authors for the precision ofAccuDEXA measurements (11,17), although Mulder et al re-ported a figure of 0.98% (10).

Figure 2 illustrates the striking difference in mean T-scoresbetween AccuDEXA and the conventional spine and hipBMD sites, with AccuDEXA giving systematically higherT-score values for patients across all age ranges. From thedata shown in Fig. 1, this T-score difference appears to be as-sociated with AccuDEXA peak bone mass occurring at anolder age than central BMD measurements. The correspond-ing Z-score data in Fig. 3 are also of interest. For the healthywomen in group 2, aged 50 yr or older and without clinicalrisk factors for osteoporosis, the mean Z-scores for Accu-DEXA and the central BMD sites were in the range0.43e0.67 and were all statistically significantly differentfrom the expected figure of 0. A number of previous studieshave reported that the mean Z-scores for older UK women,calculated using US spine and hip reference ranges, averagearound þ0.5, suggesting that US reference ranges are not ap-propriate for calculating Z-scores for UK women (18e20).The present study supports these findings and suggests thatthey be applied to AccuDEXA Z-scores and to the spineand hip sites. In contrast, the mean Z-scores for the youngnormal women in group 2, aged 39 yr or younger, were notstatistically significantly different from the expected figureof 0 for any of the 4 sites studied.

For the women aged 50 yr or older in group 1, referred bytheir family doctors for routine bone densitometry, the meanZ-scores for AccuDEXA and the central BMD sites were inthe range 0.07e0.25 and were around 0.3e0.4 units lowerthan those for women aged 50 yr or older in group 2. This dif-ference reflects the effect of clinical risk factors on the BMD

Volume 13, 2010

Table 3Concordance and Discordance (Women in Group 1 Aged 50 yr or Older)

Agreement BetweenSites

Hip and spinediagnosis

AccuDEXAdiagnosis

Number ofpatients

Observedprevalence (%)

Concordance Normal Normal 92 18.8Osteopenia Osteopenia 62 12.7Osteoporosis Osteoporosis 34 6.9

Minor discordance Normal Osteopenia 7 1.4Osteopenia Normal 160 32.7Osteopenia Osteoporosis 12 2.4Osteoporosis Osteopenia 76 15.5

Major discordance Normal Osteoporosis 0 0Osteoporosis Normal 47 9.6

298 Patel et al.

results for the women in group 1. For the women aged 39 yror younger in group 1, the mean Z-scores were in the range�0.18 to �0.72 and were around 0.4e0.7 units lower thanthose for the women aged 39 yr or younger in group 2(Fig. 3). One explanation for the greater Z-score differencesfor the women aged 39 yr or younger could be that youngerwomen in group 1 were more likely to have significant clin-ical risk factors for osteoporosis before they were referred fora bone density investigation. However, the results of the mul-tivariate regression analysis of the effect of risk factors onZ-scores failed to show that younger women were signifi-cantly more at risk.

The correlations between AccuDEXA measurements andmeasurements at the spine and hip in postmenopausal womenolder than 50 yr were typical for BMD measurements be-tween distant sites. The correlation coefficient varied from

Fig. 5. Mean Z-score change associated with each clinicalrisk factor for all 779 women for AccuDEXA and the centralbone mineral density sites. Error bars show �1 standard errorof the mean.


r 5 0.59 for hip sites to r 5 0.64 for spine BMD. Similarvalues for the correlation coefficient have been reportedfrom previous studies (10,11,17). Bland-Altman plots demon-strated that BMD measurements at central BMD sites wereconsistently lower than AccuDEXA measurements. The larg-est difference was for the spine measurements, which were onaverage 1 SD lower than AccuDEXA measurements.

When examining the number of postmenopausal womenidentified as osteoporotic, the percentage of women witha T-score of �2.5 or less were 8.7%, 22.8%, 14.5%, and7.8% for AccuDEXA, spine, femoral neck, and total-hipBMD, respectively. Varney et al (21) carried out a similaranalysis for 115 ambulatory, community-dwelling, Caucasianpostmenopausal women. They reported similar results for thecentral sites, but a significantly higher number of patientswere identified using AccuDEXA finger measurements,with figures of 13%, 20%, 17%, and 6% for AccuDEXA,spine, femoral neck, and total-hip BMD, respectively, forthe percentage of women with a T-score less than or equalto �2.5.

These results demonstrate how the uncritical use of theWHO definition of osteoporosis can result in large differencesin the percentage of patients diagnosed with osteoporosis. Asapplied to hip BMD, the WHO criterion of a T-score less thanor equal to �2.5 was intended to identify the 16% of post-menopausal women who will suffer a hip fracture (7). How-ever, Lu et al reported that, based on the WHO definition,the prevalence of osteoporosis in the Study of OsteoporoticFractures study population ranged from 3% to 60% depend-ing on the measurement site chosen (22).

One approach to this problem is the use of device-specificthresholds defined by a set sensitivity and specificity (23,24).The advantage of device-specific thresholds is that they canbe applied to all types of peripheral equipments and guaranteea reproducible level of diagnostic accuracy, irrespective of is-sues, such as systematic differences between T-scores fordifferent types of measurements. In the United Kingdom,the National Osteoporosis Society recommends a triage ap-proach, in which patients with pDXA results are interpreted

Volume 13, 2010


using upper and lower thresholds specific to each type ofperipheral device (23). The thresholds are defined so that pa-tients with osteoporosis at the hip or spine are identified with90% sensitivity and 90% specificity. Patients with a peripheralT-score less than the lower threshold are likely to have oste-oporosis at the hip or spine, and patients with a peripheralT-score greater than the upper threshold are unlikely tohave osteoporosis, whereas those between the 2 thresholds re-quire a hip and spine BMD examination for a definitive diag-nosis. For AccuDEXA, the device-specific thresholds forexaminations in UK women are an upper T-score thresholdof þ0.1 and a lower threshold of �1.6 (24).

The WHO approach to DXA scan interpretation has nowbeen supplemented by the fracture risk assessment tool(FRAX), which enables physicians to use information abouta patient’s clinical risk factors in combination with hipBMD result to assess the 10-yr probability of fracture for in-dividual patients (25). Although combining BMD with clini-cal risk factors predicts fracture risk better than BMD or riskfactors alone, the FRAX algorithm was developed solelyusing hip BMD measurements and does not permit the useof BMD data measured at peripheral skeletal sites.

Previous studies investigating the relationship betweenspine and femur BMD and the reason for referral fora DXA scan have confirmed that certain indications, suchas premature menopause, X-ray osteopenia, and glucocorti-coid therapy, are associated with low BMD and increasedprevalence of osteoporosis (26). A recent study demonstratedthat clinical risk factors for low BMD affect forearm BMDmeasurements to a similar extent as central BMD (27), sug-gesting that forearm DXA could become an alternativemethod for identifying individuals with low bone mass andthose who are possible candidates for therapeutic interven-tion. One aim of the present study was to determine whetherBMD measurements of the hand using AccuDEXA areaffected by clinical risk factors in a similar manner. Toaccount for the possible differences between premenopausaland postmenopausal women, the analysis of AccuDEXAmeasurements was initially carried out separately for womenyounger than 50 yr and those 50 yr or older. As there were nosignificant differences in the results for younger and olderwomen, the results presented are for all the 779 womenwho participated in the study.

The results of the multivariate regression analysis showedthat, for AccuDEXA measurements, Z-score changes associ-ated with a history of atraumatic fracture, X-ray osteopenia,and a BMI of less than 20 kg/m2 were all statistically signif-icant compared with women with no risk factors. Figure 5demonstrates that AccuDEXA measurements vary in a similarmanner to Z-scores at central sites. There are no previousstudies reporting the relationship of phalangeal BMD withclinical risk factors for osteoporosis.

One issue that remains unexplored with phalangeal BMDmeasurements is the ability to monitor changes in bonedensity. Although there is relative agreement that peripheraldevices can be effective at identifying patients with lowbone mass, their ability to monitor changes associated with


aging, disease, or therapy, remains controversial (28). Studiesexploring this issue at peripheral sites have been limited toforearm BMD studies and have shown this site to be insensi-tive for monitoring response to treatment (29). Because of theabsence of data for phalangeal BMD measurements, no rec-ommendations can be made about its use for monitoringBMD changes over time.

Conclusions

We have demonstrated that phalangeal BMD measure-ments acquired on AccuDEXA correlate with spine and hipmeasurements and are able to detect bone losses related toage and other clinical risk factors for osteoporosis. In a groupof healthy young UK women, mean Z-scores for AccuDEXAand spine and hip sites calculated from US reference rangeswere consistent with the expected value of 0. However, forhealthy UK women aged 50 yr or older, mean Z-scores forAccuDEXA and the central sites were in the range 0.4e0.7and were statistically significantly different from 0. The con-ventional WHO T-score criteria for the diagnosis of osteopo-rosis should not be applied to AccuDEXA measurements inUK women. However, using a triage approach, device-specific T-score thresholds may be used to aid with diagnosticclassification. Clinical risk factors for low BMD were foundto affect AccuDEXA measurements to a similar extent as cen-tral BMD. AccuDEXA measurements could, therefore, be-come an alternative method for identifying individuals withlow bone mass, provided care is taken in interpreting theT-scores.

References

1. Cooper C, Campion G, Melton LJ. 1992 Hip fractures in theelderly; a world-wide projection. Osteoporos Int 2:285e289.

2. Johnell O, Kanis JA. 2006 An estimate of the worldwide preva-lence and disability associated with osteoporotic fractures.Osteoporos Int 17:1726e1733.

3. Marshall D, Johnell O, Wedel H. 1996 Meta-analysis of howwell measures of bone mineral density predict occurrence of os-teoporotic fractures. BMJ 312:1254e1259.

4. Stone KL, Seeley DG, Lui L-Y, et al. 2003 BMD at multiplesites and risk of fracture of multiple types: long-term resultsfrom the Study of Osteoporotic Fractures. J Bone Miner Res18:1947e1954.

5. Eastell R. 1998 Treatment of postmenopausal osteoporosis.N Engl J Med 338:736e746.

6. National Osteoporosis Society. 2002 Position statement on thereporting of dual x-ray absorptiometry (DXA) bone mineral den-sity scans. National Osteoporosis Society, Bath, England.

7. WHO Technical Report Series 843. 1994 Assessment of fracturerisk and its application to screening for postmenopausal osteopo-rosis. World Health Organisation, Geneva, Switzerland.

8. Genant HK, Engelke K, Fuerst T, et al. 1996 Noninvasive assess-ment of bone mineral and structure: state of the art. J BoneMiner Res 11:707e730.

9. Gl€uer C-C, Jergas M, Hans D. 1997 Peripheral measurementtechniques for the assessment of osteoporosis. Semin NuclMed 27:229e247.

Volume 13, 2010

300 Patel et al.

10. Mulder JE, Michaeli D, Flaster ER, Siris E. 2000 Comparison ofbone mineral density of the phalanges, lumbar spine, hip andforearm for assessment of osteoporosis in postmenopausalwomen. J Clin Densitom 3:373e381.

11. Ohtsuka M, Michaeli D, Wasnich RD. 2002 Relationship be-tween phalangeal bone density and risk of vertebral fracture.J Clin Densitom 1:11e15.

12. Looker AC, Wahner HW, Dunn WL, et al. 1998 Updated data onproximal femur bone mineral levels of US adults. Osteoporos Int8:468e489.

13. Altman DG. 1991 Practical statistics for medical research. Chap-man & Hall, London, UK.

14. Royal College of Physicians. 1999 Strategic considerations. In:Osteoporosis: clinical guidelines for prevention and treatment.Royal College of Physicians, London, UK.

15. Kanis JA, Delmas P, Burckhardt P, et al., on behalf of the Euro-pean Foundation for Osteoporosis and Bone Disease. 1997Guidelines for diagnosis and management of osteoporosis.Osteoporos Int 7:390e406.

16. National Osteoporosis Foundation. 2008 Risk assessment. In:Clinician’s guide to prevention and treatment of osteoporosis.National Osteoporosis Foundation, Washington DC.

17. Fiter J, Nolla JM, Gomez-Vaquero C, et al. 2001 A comparativestudy of computed digital absorptiometry and conventionaldual-energy x-ray absorptiometry in postmenopausal women.Osteoporosis Int 12(7):565e569.

18. Petley GW, Cotton AM, Murrills AJ, et al. 1996 Referenceranges of bone mineral density for women in southern England:the impact of local data on the diagnosis of osteoporosis. Br JRadiol 69:655e660.

19. Holt G, Khaw KT, Reid DM, et al. 2002 Prevalence of osteopo-rotic bone mineral density at the hip in Britain differs substan-tially from the US over 50 years of age: implications forclinical densitometry. Br J Radiol 75:736e742.


20. Noon E, Singh S, Cuzick J et al. Significant differences inUK and US female bone reference ranges. Osteoporos Int.In press.

21. Varney LF, Parker RA, Vincelette A, Greenspan SL. 1999 Clas-sification of osteoporosis and osteopenia in postmenopausalwomen is dependent on site specific-analysis. J Clin Densitom2:275e283.

22. Lu Y, Genant HK, Shepherd J, et al. 2001 Classification of oste-oporosis based on bone mineral densities. J Bone Miner Res 16:901e910.

23. National Osteoporosis Society. 2004 Position statement onthe use of peripheral x-ray absorptiometry in the manage-ment of osteoporosis. National Osteoporosis Society, Bath,England.

24. Blake GM, Chinn DJ, Steel SA, et al. 2005 A list ofdevice-specific thresholds for the clinical interpretation of pe-ripheral x-ray absorptiometry examinations. Osteoporos Int 16:2149e2156.

25. FRAX-WHO fracture risk assessment tool. Available at, http://www.shef.ac.uk/FRAX. Accessed on March 10, 2010.

26. Ahmed AIH, Ilic D, Blake GM, et al. 1998 Review of 3530 re-ferrals for bone density measurements of spine and femur: evi-dence that radiographic osteopenia predicts low bone mass.Radiology 207:619e624.

27. Patel R, Blake GM, Fogelman I. 2007 Peripheral and centralmeasurements of bone mineral density are equally strongly asso-ciated with clinical risk factors for osteoporosis. Calcif TissueInt 80:89e96.

28. Gl€uer CC. 1999 Monitoring skeletal changes by radiologicaltechniques. J Bone Miner Res 14:1952e1962.

29. Bouxsein ML, Parker RA, Greenspan SL. 1999 Forearm bonemineral density cannot be used to monitor response to alendro-nate therapy in postmenopausal women. Osteoporosis Int 10:502e509.

Volume 13, 2010

http://www.shef.ac.uk/FRAX

http://www.shef.ac.uk/FRAX

clinical evaluation of a phalangeal bone mineral density assessment system

Documents