Unequal decrease in bone density of lumbar spine and ultradistal radius in Colles' and vertebral fracture syndromes

(1)

Unequal decrease in bone density of lumbar

spine and ultradistal radius in Colles' and

vertebral fracture syndromes.

R Eastell, … , L J Melton 3rd, B L Riggs

J Clin Invest.

1989;

83(1)

:168-174.

https://doi.org/10.1172/JCI113854

.

We measured bone mineral density (BMD) at the lumbar spine (LS-BMD) and ultradistal

radius (UDR-BMD) in 42 postmenopausal normal women and in 108 postmenopausal

osteoporotic women (55 with vertebral fracture, 34 with Colles' fracture, and 19 with both

fractures). By receiver operating characteristic analysis, LS-BMD was better than UDR-BMD

(P less than 0.01) as an indicator of vertebral fracture; the converse was true for Colles'

fracture (P less than 0.01). Although UDR-BMD and LS-BMD were lower in each of the

three fracture groups than in controls (P less than 0.01), the pattern of bone loss differed (P

less than 0.001, analysis of variance): with vertebral fracture, LS-BMD decreased relatively

more than UDR-BMD; with Colles' fracture, UDR-BMD decreased relatively more than

LS-BMD; and with both fractures, decreases in LS-BMD and UDR-BMD were similar. We

conclude that both types of fracture are caused by excessive bone loss but the difference in

bone loss at the two sites is a major factor in determining which will fracture.

Research Article

Find the latest version:

(2)

Unequal Decrease

in

Bone Density of Lumbar Spine and Ultradistal Radius

in

Colles' and Vertebral Fracture Syndromes

Richard Eastell, Heinz W. Wahner, W. Michael O'Fallon, Peter C. Amadio, L. Joseph Melton 111, and B. Lawrence Riggs

Endocrine Research Unit, Section ofDiagnostic Nuclear Medicine, Department ofHealthSciencesResearch, and Department of Orthopedics, Mayo Clinic and MayoFoundation, Rochester, Minnesota 55905

Abstract

We measured bone mineral density (BMD) at thelumbar spine (LS-BMD) and ultradistal radius (UDR-BMD) in 42 post-menopausal normal women andin108postmenopausal osteo-porotic women (55 with vertebral fracture, 34 with Colles'

fracture, and 19 with both fractures). By receiver operating characteristicanalysis, LS-BMDwas better than UDR-BMD (P < 0.01) as an indicator of vertebral fracture; the converse was true for Colles' fracture (P < 0.01). AlthoughUDR-BMD andLS-BMDwere lower in each of the threefracture groups thanin controls (P < 0.01), the pattern of bone loss differed (P

< 0.001, analysis of variance): with vertebral fracture, LS-BMD decreased relatively more than UDR-LS-BMD, withColles'

fracture, UDR-BMD decreased relatively more than LS-BMD; and with both fractures, decreases in LS-BMD and UDR-BMD were similar. We conclude that both types of fracture are caused by excessive bone loss but thedifferencein bone loss at the twosites isamajor factorindetermining which

willfracture.

Introduction

Involutional osteoporosishasbeen divided intotwotypes on thebasis ofdifferences inbonedensity, inthe age-and

sex-spe-cific incidencepatternof the associatedfractures, and in

mech-anisms of bone loss (1, 2). In type I osteoporosis, there is disproportionateandaccelerated loss of trabecular bone,and

fractures characteristically occur at skeletal sites containing large amountsoftrabecular bone: the vertebrae and distal fore-arm(Colles'

fracture).

In type II

osteoporosis,

thereisa more

gradualthinning of both trabecularandcortical bone, leading

tofractures ofthehip, pelvis,andproximal humerus and to

multiplewedgefractures ofthe vertebrae ("dowager's hump"). Type I osteoporosis mainly affects women within 25 yr of menopause and is believed to result from factorsrelated to estrogen

deficiency.

Type II osteoporosis affects men and womenolder than75 yr andis believedto resultfromfactors related toaging(1).

Bonelossfromthe vertebraeinwomen withtype I

osteo-porosisand vertebralfracturehas been extensively studied (1). Much less is known about the extent of bone loss from the

Addressreprintrequests to Dr.Riggs,Endocrine Research Unit, Mayo

Clinic, 200 First Street SW, Rochester, MN55905.

Receivedfor publication16February 1988 and inrevised form 6

July1988.

ultradistal radius

(UDR),I

i.e.,the distal 3cmoftheradius,in women with type I osteoporosis and Colles' fracture or the relationship of lossatthis site to loss in the vertebrae. Most

densitometricmeasurementshave been madeproximaltothe

usual

site ofColles'

fracture,

in aregion

containing

5-50% trabecular bone(3-9).

In the presentstudy, we measured bonemineral density (BMD) of the lumbar spine (LS-BMD) and UDR

(UDR-BMD)inwomenwith type I osteoporosisin orderto answer

two questions. The first question was: is measuring UDR-BMD as effective asmeasuring LS-BMD fordetecting bone loss from the vertebrae in womenwithvertebral fracture? If

trabecular bone loss were uniform throughout the skeleton

and if

cortical

bone

density

wereunchanged,thensimilar

re-sults wouldbe obtained at measurement sites with similar

proportions of trabecular bone. Ifmeasurement of UDR-BMDprovidedagoodestimateof LS-BMD, thiswould have

practical advantages. The

radius

is

considerably

easierto

mea-surethanthe spine and isnotbeset with artifacts suchasspinal deformities (bonespurs, vertebral

fracture)

andaortic

calcifi-cation that may confoundthe measurement of LS-BMD in

olderwomen. Weaddressed thisquestion by performing

re-ceiveroperating characteristic (ROC) analysis.

The second

question

was:if allwomenwith typeI osteopo-rosis undergoa phase of accelerated trabecular bone loss (1),

why do some ofthemhave Colles'fracture andothers have vertebral fracture?Wetestedthehypothesisthatunequalbone loss predisposessome osteoporotic women to vertebral

frac-ture andotherstoColles' fracture.Thenull

hypothesis

isthat

the degree of bone loss is similar at sites ofpredominantly trabecular bone, and that the typeof fracture is

determined

only by externaleventssuchas trauma toonesiteortheother.

Bonelossfromthe lumbarvertebraecan bemeasuredby dual-photon absorptiometry. However,

in

order to measure

the relevant site in the UDR, itwas necessary to develop a

methodformeasuringBMD atthe distal 3 cmof the radius. This site is composed ofpredominantly trabecular bone (10)

andis where Colles' fracture occurs.Themethoduses single-photonabsorptiometryandcomputer-assisted image

process-ing for accurate

repositioning

andselection oftheregion of interest.

Methods

Experimental subjects

Normal women. We studied 42 healthy postmenopausal women (TableI),aged 50-75yr, whowere1-31 yrpostmenopausal.Nonehad hadmenopausebefore theageof 40yr,hadanydiseaseor wastaking

1.Abbreviations used in thispaper:LS-BMD,lumbarspinebone

min-eral density; UDR-BMD, ultradistal radius bone mineraldensity; ROC, receiver operatingcharacteristic.

168 Eastelletal. J.Clin.Invest.

©TheAmerican Societyfor ClinicalInvestigation,Inc.

0021-9738/89/01/0168/07 $2.00

(3)

anymedication knowntoaffectbonedensity,orhad ahistoryof pain

orstiffnessof the wrist.Nosubject hadavertebralfracture evidenton

anteroposterior or lateral radiographs ofthe lumbar and thoracic

spine.

Fracturegroups. The patients with type Iosteoporosiswere divided intothree groupsaccordingto the type of fractures they had.

The firstgroup (55 women, aged 54-75 yr) had only vertebral fracture. All had three or more mild wedge fractures (anterior vertebral

height75-85%of posterior height)or one ormore severe wedge frac-tures(anterior height<75%ofposterior height)or both. Allfractures

had occurred after minimal or no trauma. Factors that might have resulted indecreasedbonedensity includedahistory of thyrotoxicosis inone,chronic obstructive airwaydisease inone, andsurgically

in-ducedpremature menopauseinseven. Nonehad hadprevious treat-ment with fluoride,but 28 were being treated withcalcium supple-ments(0.5-2gof elemental calciumdaily) and 10 hadpreviouslyused orcurrentlywereusingestrogen.

Thesecond group (34women, aged 53-75 yr) had onlyColles'

fracture.Thefractures hadoccurred 1-28 mo (mean, 11.7mo)before

measurement; theyaffectedthedominant forearm in17 andthe

non-dominant forearm in 17. Factorsthatmighthavecontributedto

de-creased bonedensity includedahistory of thyrotoxicosis intwo,

chronicobstructive airway disease intwo,andsurgically induced pre-maturemenopausein three.None werebeingtreatedforboneloss.

Thethirdgroup(19women, aged59-75 yr) had both Colles'and vertebral fractures. The Colles' fracturehad occurred 1-31 yr (mean, 15yr) before measurementand affected thedominant forearm in 8 andthenondominant forearm in1 1. Thecriteriafor vertebral fracture

were the same asforthefirstgroup. Factors that may have resultedin decreasedbonedensity includedahistory ofthyrotoxicosis inoneand surgically inducedpremature menopause in one. None had had

pre-vioustreatmentwith fluorideorestrogen, but 10 werebeingtreated withcalciumsupplements (0.5-2 g of elementalcalcium daily).

Bone

densitometry

UDR-BMD was measured in the nondominant forearm (orin the

uninjured forearm in the Colles' fracture group) by single-photon

ab-sorptiometry withcomputer-assisted image processing. Theforearm

was positioned in the scanningapparatus, and the wristwas

sur-rounded byawater bag to ensure constantthicknessovertheentire scanning path.Scanningofthedistal 3cmof the radiusbeganatthe radial styloidprocess(identified by palpation) and movedproximally

in2-mm steps(that is, 15 scanlines).Theimage (Fig. 1A)was dis-playedon amicrocomputer(IBM-PC)toconfirmcorrectpositioning

ofthewrist and thenwasstoredon afloppydiscforsubsequent

pro-cessing. Theradiation sourcewas 1251I (photopeak,27 keV) and the collimatorwas2mm X2mm.

Theimagedisplayprogram allowedselection ofregions of interest

anddetection of boneedges.Theareaof interestwas1cmlong(Fig. 1

B), andthe distal end of thisareawas4mmproximaltothemedial edge ofthedistalarticular surface of the radius(whichwasalwaysthe

line of peak bone density).Wechosethisparticularsite because it is the sitethrough which Colles' fractureoccurs.Wedetermined that Colles'

TableL CharacteristicsofPatientGroups

Osteoporoticwomen

Vertebral Colles' Both

Variable Normalwomen fracture fracture fractures

Number 42 55 34 19

Age(yr) 62±8 67±5 65±6 68+5

Postmenopause(yr) 12±8 19±7 16±7 21±9

Weight (kg) 68±13 63±11 70±13 65±10

Values given asmean±SD.

fractureoccurs at a meandistance of22mm_(±4mm,SD)

proximal

to

thetip ofthe radial

styloid,

based on measurementsmade on 20

radiographsof Colles' fracture with minimal

displacement.

Precision ofUDR-BMD, determinedbyduplicatemeasurements on20premenopausalhealthysubjects,was 17%.2The valueforthe

dominant radiuswas 3% higher, on average, than that forthe

non-dominant radiusasdeterminedbymeasurementsmadeonanother 20

healthypremenopausalwomen.Thus,inwomenwith Colles'fracture ofthe nondominant forearm, the dominant forearm was measured anda3%correctionwasmade.

LS-BMDwasdeterminedbydual-photon

absorptiometry

of

verte-braeL-2toL-4 with thescanningapparatus

previously

described(

11)

and

'"Gd

as the source (photopeaks, 44 and 100 keV). Fractured vertebraewerenotmeasured. Theprecisionwas2.2%.

Anatomic

studies

of

the radius

Radii removed from four cadaverswerestudiedtodetermine the

accu-racyof themeasurementand the

proportion

of trabecularboneatthe

site ofmeasurement.TodetermineBMDof the bone

specimens,

after removal ofsurroundingtissuetheywere_air-dried,

_defatted,

and

em-beddedinmethylmethacrylate.Threetofive

pieces

2-7mmwidewere

cutwithajeweler'ssaw (Isomet, BuehlerLtd., LakeBluff, IL), and

200-pm

sections from the proximalend ofthepieceswere_groundto

100-pmsectionsbeforemicroradiography;theproportionof trabecu-lar bonewasestimatedby point counting (Table II; Fig. 2).The

re-mainingpiecesofbonewerepartiallyashed(500'Cfor24_{h); then,}

trabecularbone was_separated from corticalbonebydissectionand

ashed inamuffle furnaceat600'Cfor 24 h. Theamountofeachtype of bonewasdeterminedby weighing.

Theresults ofthe accuracy

analysis

areshown in

Fig.

3. The re-gression line does not intersectthe

yaxis

because the cadaver radii

werescannedwiththemarrowfat in situ: thismethod allows

correc-tion fortheeffectofmarrowfat in vivo.The

regression

linefordefatted

boneis shown for

comparison

(I12).The accuracyofthe

technique (the

standarderroroftheestimate ofthe

_regression

X

_100/mean)

was8%.

Statistical

analysis

TheabilityofthetwoBMDmeasurementstodiscriminatebetween the

normalsubjectsand womenwith vertebral fracture or women with

Colles' fracturewasevaluated

by

applying

theROCcurve

approach

(I13, 14).Foreach ofthetwoBMDmeasurementsandfor each fracture

group,all

possible

cut-off

points

weredefinedand the

proportion

of

healthysubjectsabove(the

specificity)

and the

proportion

of osteopo-roticsubjectsbelow(the

sensitivity)

each

point

werecalculated.This yieldsanROCcurvethat

displays

the

relationship

between

sensitivity

and

specificity

for eachBMDmeasurementasadiscriminator between

thenormal and fracturegroups. TheareabetweentwoROCcurves

contraststhe

ability

oftwoBMDmeasurestodiscriminate.Theareas

wereestimatedandtested

(null

hypothesis

is thatarea

equals 0) by

a technique developed byWieandetal.3

Theresults of LS-BMDandUDR-BMDwere

reported

in grams

per squarecentimeter and in the form ofZscores.Zscoreswereused

fortwo reasons. First, the BMD resultwas

adjusted

for factors that

differbetween individuals,suchas_{age and}

weight. Secondly,

this ap-proachconvertsthedeviation from normal ofBMDvaluesateachof thetwo

_scanning

sites into SD unitsandthusallowsestimation ofthe

relative deficitin BMDin each fracturegroup.

Zscoreswerecalculatedbyatwo-stepprocedure. First,theeffects

ofage,

(age)2, (age)3, body weight, height,

and yearspostmenopauseon

BMD in the 42normalwomen wereassessed

by multiple regression

analysis.BothLS-BMDand UDR-BMDcorrelatedmost

closely

with 2. Precision = _(SD _of_differences_between

paired

BMD measure-ments)/(mean BMD)X 100.

3.Wieand,H.S.,K.James,B.James,and M. H.Gail. 1987. Nonpara-metricproceduresfor

_{comparing diagnostic}

testswith

paired

or

(4)

A

I I

B

0% 10 20 30 40 50 60 70 80 90 100%

percentage of Radiallength

age andbody weight: R2 = 0.36 (P< 0.001)and0.29 (P<0.01),

respectively. Secondly,theregression equations wereused topredict

BMDbasedon anindividual's age and bodyweight. The Z score for

BMD ateither sitewas thencalculatedas Zscore =(observedBMD

Table II. TrabecularBoneContentinRadiifromFourCadavers

Cadaver Trabecular bone

Ageatdeath Sex By weighing By point counting

yr %

33 M 53 72

52 M 74 74

74 F 60 *

94 M 56 66

Mean 61 71

*_Samplesfractured during preparation for microradiography.

Figure 1. (A) Computer-as-sistedimage of distal 3cm

of radius. Use of horizontal andverticalcursorsallows

selection ofregion of

inter-est.Upperpaneldisplays

BMDprofile(-axisis log-arithm ofattenuation;

x-axis isscandistance)

se-lectedby horizontal cur-sors.(B)Tracing of radiograph of forearm bonesexcised postmortem from 94-yr-old manto

showsite ofmeasurement

of UDR-BMD.Inprevious

studies ( 11)BMD was

mea-suredatthemidradius (50%), one-thirdsite, and distal site (10%); these pro-portions relatetodistance along the ulna,notalong theradius.

- predicted

BMD)/(sy.X)

in which

sy.,

is the standard deviation of

BMDvaluesabout theregressionline.Thus,bydefinition,themeanZ scorein normalsubjectswould be0,and95% ofnormalsubjectswould haveZ scoresbetween -2 and +2.

The Z score calculationswereusedto answer twoquestions.First,

did BMD in thefracture groups differ fromBMDin the normalgroup?

Under thenullhypothesisthat thefracture groupsare the sameasthe

normal group, themean Z scoreshouldnotdiffersignificantlyfrom0;

thishypothesiswastestedbyusingone-samplet tests. Secondly,was

thererelativelygreater loss of LS-BMD in the vertebral fracture group andrelativelygreater loss of UDR-BMD intheColles'fracturegroup?

Thishypothesiswastestedbycalculatingthe meandifference between LS-BMDZscore and UDR-BMD Z scoreamongeachofthefracture groups. Thus, ifthere wasrelativelygreaterbone lossatLSthanat

UDR, the difference LS-BMD Zscore minus UDR-BMD Zscore

would benegative;conversely, if there wasrelativelygreater bone loss

atUDRthanLS,thedifference LS-BMDZ scoreminus UDR-BMDZ scorewould bepositive. This derived valuewasusedas abasis for statisticalcomparison of the relative degree of bone lossatthetwo

scanningsites among groups. Aftertestingfor overall differences amongthegroupsbyone-wayanalysisofvariance,wecompared dif-ferencesbetweengroupsby two-samplet tests.

170 Eastell etal.

a ""6.P

-

-%,amp,

a

0.

ol

:-i.

I. .a

.%10

1% tt

:%I

a

k,

9

k

p

il

(5)

Figure2.Microradiograph of

_l00-,gm

sections from the middle of theregionof interest in cadaver radiusspecimens.(Left)Section from 33-year-oldman;(right) from94-year-oldman.

Results

Trabecular bonecontentof UDR. At theUDRsite the mean

percentageof trabecular bonewas71% byvolume and 61% by

weight (Table II).

Discrimination offracturegroups.The ROCcurvesfor

LS-BMDand UDR-BMD in women with vertebral fractureare

shown in Fig. 4A.Forthis analysis, thecurvethatisnearest to

thetopleftcorner(thiscornercorrespondsto 100%sensitivity and 100%specificity)representsthe besttest. Thus, LS-BMD discriminates osteoporotic women with vertebral fracture

from age-matched normal women better than UDR-BMD

does(P<0.001). Theareasunder thecurvesfor LS-BMD and

UDR-BMDwere91% and 78%, respectively. The ROCcurves

forLS-BMD andUDR-BMD forwomenwithColles'fracture are shown inFig. 4 B. Here UDR-BMD discriminates better

betweenwomenwithand those without Colles' fracture than

does LS-BMD (P < 0.01). The areas under the curves for

UDR-BMD and LS-BMD were 73% and 61%, respectively.

Thesensitivity and specificity of thetwo measurementsinthe

2

vertebral fracture and Colles' fracture groups are given

nu-merically in Table III.

Unequal bone

loss atsites

offracture.

The LS-BMD and UDR-BMD values are given in Table IV. The percent de-creasesofmeanLS-BMD from normalwomenwere25%, 7%,

and 25%, respectively, in the groups with vertebralfracture, Colles' fracture, orboth fractures. The

respective

percent de-creasesofmean UDR-BMD from normalwomen were 15%,

12%, and 20%. Allofthesedecreases werestatistically

signifi-1.0

0.8

0.6

0.4

0.2

1.5 r

1.0

1-C

._I

c

a)

Un Regression line

(untreatedbone)

r=0.97

0.0

1.0

/-O

-V1 0.8

0.6

Ad Regression line (defatted bone) 0.5

1-0

0.4

0.2

0 4 6

Area undercurve, units

Figure3.Relationshipbetween estimated BMD(areaundercurve

[see Fig. 1],inarbitrary units)and ashweightperunitlength(g/cm) infour cadaver radiusspecimens.Regressionline intersects waxis

abovetheorigin. Equationusedtocalibrate the densitometerwas:

ashcontent=_0.161 X(bone density)+0.3 18.

0.0

1.0 0.8 0.6 0.4 0.2 0.0

Specificity

Figure4.ROCcurvesshowing sensitivityandspecificityofLS-BMD

(thin line)andUDR-BMD(thick line). (A)Inwomenwithvertebral

fractures.(B)InwomenwithColles'fracturesonly. E

C.)

-0

C

0

(6)

TableIII.Sensitivityand

Specificity

of UDR-BMDand

LS-BMDasTests for Vertebral Fracture and Colles' Fracture

Sensitivity* Specificityt

Measurement At90% At50% At90% At 50%

For vertebral fracture

UDR-BMD 45 90 50 83 LS-BMD 69 96 76 98

ForColles'fracture

UDR-BMD 28 88 47 74

LS-BMD 26 53 27 72

*_At_{specificities of 90% and 50%.}

*_At_{sensitivities of 90% and 50%.}

cant(TableIV).Inthe groupwithvertebralfracturethere was a relatively greaterdecrease in LS-BMD, in the group with

Colles' fracture there was a relatively greater decrease in UDR-BMD, and in the groupwithboth fractures there were

similardecreasesinboth LS-BMD andUDR-BMD.

The womenwith vertebral fracture had the most negative meandifference(-0.76) between LS-BMD Z score and UDR-BMD Zscore, thosewithColles' fracture had the most positive mean difference(0.40), and those with both fractures had an

intermediate mean difference (-0.22). These differences among the three mean valueswerestatistically significant (P

<0.001, analysis of variance).Thewomen with vertebral frac-turealoneand thosewith both fractureshad a relatively greater

decrease at LS-BMD than at UDR-BMD compared with

women with Colles' fracture (P < 0.001, P < 0.02, respec-tively). Inbothgroupsofwomen with vertebral fracture,the meandecreasesinLS-BMD were similar (Table IV); however, thosewithbothvertebral and Colles' fractureshad a

margin-allygreaterreduction in UDR-BMDthan did those with

ver-tebralfracturealone(P= 0.06).

Individualvaluesfortheserelativedifferences are shown in

Fig.5. Forthegroup with vertebral fracture only, 80% of the

points fall below thelineof identity,indicating that bone loss wasrelativelygreater at LS than at UDR. For the group with

Colles'

fracture,

68%of the points fallabove the lineof

iden-tity, indicating that bone loss was relatively greater at UDR

than at LS. Forthe group with both fractures, 9 points fall

above the line of identityand 10pointsfallbelow,

indicating

similar relativedecreasesatLSand UDR.

Discussion

Weencounteredtwotechnical problems in developinga

repro-ducible andaccurate

technique

formeasurement atthe UDR. First, repositioningwas foundtobe ofcriticalimportanceat

this

site because there were large

changes

in bone mineral

contentand in the

proportion

of trabecularbone over a short

distance. We overcame this by

utilizing computer-assisted

imageprocessingtoidentifytheareaof interestonthe inten-sity-modulated bone mineral image. Some investigatorshave

attemptedtosolve thisproblem by

scanning

at afixed distance between the radius and ulna

(for

example, at5 mm[15] or8

mm [16]). However, at thesesites the bone is lessthan 50% trabecularandthe

proportion

of trabecularboneisvery vari-able. Others have

performed

a

preliminary

"scout scan" by computed tomography (5, 17, 18).

Thesecond problemwas that the high

proportion

of

tra-becularbone(Fig. 2) is associated withalargeamountof fat in the marrow space, and this produces a

systematic

error in results of scanning with a single-energy photon source.This

problem was appreciated by

Karjalainen

(19) but has been

ignored by subsequent workers (13, 14).Wecorrected for itby measuring UDR-BMD incadaverradiusspecimens with mar-row fat in situ and then defatting the bone andashing the

specimen. This regression ofash content on area under the curve wasthenused toestimate UDR-BMD. The presenceof fat hasamarkedeffecton UDR-BMD becauseattenuation of thephoton beam islessthrough fat than

through

water;thus, failuretocorrectfor fat within themarrowspaceresults inan

underestimate ofUDR-BMD.

Most

proximal

tothe fracture site(3, 4,

6-9),

wherethe boneis

mainly cortical;

at

those sitesbonemineralcontent was5-7%below normal.The one

exception

wasthe

study by Hesp

etal.

(5)

inwhich UDR-BMD wasmeasured by acombined

computed

tomography/

single photon

absorptiometry technique.

Itis notablethat

they

reported a relatively large decrease

(12%)

in UDR-BMD as

compared withage- andsex-matched normal

subjects.

Inthe presentstudy,we alsofoundthatthemeandecrease in UDR-BMD inColles' fracture

patients

was 12%.

TableIV.BoneDensity ValuesinNormalWomenandWomen With Fractures

Osteoporoticwomen

Variable Normalwomen Vertebral fracture Colles' fracture Both fractures

Absolute bone mass

LS-BMD

(g/cm2)

1.04±0.02 0.78±0.02 0.97±0.03 0.78±0.03

LS-BMD

(%

below

_normal)

25 7 25

UDR-BMD

(g/cm2)

0.41±0.01 0.35±0.01 0.36±0.01 0.33±0.01 UDR-BMD(% belownormal) 15 12 20

Z-scores

LS-BMD (SD units) -1.61±0.13* -0.60±0.18t -1.63±0.20* UDR-BMD(SD units) -0.85±0.13* -0.97±0.13* -1.40±0.22* Valuesgiven asmean±SE.

*_For_{difference from}_normal_women,_P_<_0.001. t_{For difference from}normalwomen,P<0.01.

(7)

3

2

-1

-2

-3

2

a)

L-o

C.)

d

0

-1

-2

-3

-4

3

2

1

0

-1

-2

-3

-4 K

-2 -1 0 1

UDR-BMD, Z-score

2 3

Figure 5.LS-BMDand UDR-BMDexpressedasZ-scores,in indi-vidualwomen.Obliquelineindicates line ofidentity. (A)Women

with vertebralfractures.(B)Women withColles'fractures.(C)

Women withbothvertebral and Colles' fractures.

In our _study, BMD was measured at sites composed mainly oftrabecular bone. Our anatomic studies indicated

that the UDR-BMDmeasurement sitecontainedabout 70%

trabecular bone, and others (10, 20, 21) obtained similar values. Moststudieshave found thatvertebraecontainabout

70% trabecularbone (22, 23). However, ina recent studyof

cadavers of_{eight elderly}women_(aged60-86years),Nottestad

et al.(24)found theproportionoftrabecular bonetobe42% in

thebodyand 24% intheentire vertebrae. This lowproportion

oftrabecularbone may have been related toageandtobone

loss_duringtheterminal illness.

Our findings provide answers to the twoquestionsthatwe

posed in the Introduction. In answer to the first question,

UDR-BMD isnot an appropriate substitute for LS-BMD in

assessingthe extent of bone loss from the vertebrae in women with vertebral fracture. ROCcurveanalysisshowedthat LS-BMD measurements were much more sensitive and specific

than UDR-BMD measurements for separating

osteoporotic

patients with vertebral fracture from age- and sex-matched

normalcontrols, despite the similarity in content oftrabecular bone at the twosites. Conversely, UDR-BMD measurements were moresensitiveandspecificthanLS-BMDmeasurements forseparating osteoporotic patients with Colles' fracture from normals. Thus, in order to predict fracture risk in a bone, the

BMDof that bone should bemeasured-measurementsmade

atother parts of the skeleton have less predictive value. The results of LS-BMD measurements inwomenwith

ver-tebral fracture were similar to those in our earlier report( 11);

in thatstudy,45% of women with vertebral fracture had LS-BMDvalueslessthan the 5thpercentileofnormals, compared with 49%inthe present study (Z score < 1.645). In that study, 7% had distal radius BMC values less than the 5th percentile, compared with 25% with UDR-BMD values less than the 5th

percentile

(Zscore< 1.645).Thehigher proportion oftrabecu-lar boneattheUDRthan at thedistal radiusmay allow better

separationof women with and without vertebral fracture.

These results dodiffer fromthose in two recent reports. Ott

etal.(25) measured LS-BMD andBMCofthedistal radius in

postmenopausalwomenwith and without vertebral fracture.

UsingROC curveanalysis, they foundthatthetwo

measure-mentsites gaveapproximatelyequalsensitivityand specificity.

Thus, for asensitivity of90% (i.e., the value we use as our "fracturethreshold"),thespecificity of LS-BMD in theirstudy

was 21%, whereas in ourstudy it was 76%; the specificity of distal radius BMC in their study was 35%,whereas for

UDR-BMD in our

study

it was 50%. Thus, the major difference

between our

findings

and those of Ott et al. (25) is better

specificity

ofLS-BMDinourstudy. Nilasetal.(26)reported

BMDresults in 28 women with vertebralfracture.The mean LS-BMD Zscore was -0.43 (itwas -1.61 in the present

study),

andthemeanUDR-BMD Z score was-0.51(-0.85in the presentstudy).

Theresultsof Ottetal.

(25)

andNilasetal.

(26)

maydiffer

fromourresults fortworeasons.We used strictercriteria for

diagnosing

osteoporosis,

requiring

the presenceofatleast three

mild

anteriorly

wedged vertebrae; Nilas et al. (26) required onlyone suchfracture and Ott et al. (25) required onlytwo such fractures.

Recently,

theDanishgrouphavefurther

ana-lyzed

their data. By

dividing

their osteoporotic subjects into those with and

without compression fractures,

Podenphantet al.

(27)

nowreporta

relatively

greaterlossofbone

density

at

the lumbar

spine

in

subjects

with

compression

fractures.The

precision

of LS-BMD in thepresentstudywas2.2% (11); Nilas

etal.

(26) reported

a

precision

of6.2%,andOttetal.(25)did

notreport the

precision

of their method,but the intrapopula-tion standard deviaintrapopula-tion was greater than thatreported by

others

(28).

Withrespecttothesecond

question,

differentialbone loss atthe site of fracture may explain, in part, why somewomen

with type I

osteoporosis

have vertebral fracture and others

have Colles'fracture.

Why

is there

relatively

greater bone loss

atthe LS inwomenwith vertebral fracture and greater bone lossattheUDR inthosewithColles'fracture, giventhat these

sitesareboth

composed

mainly

oftrabecularbone? One expla-B

1

-

'A

A

AAA A A

A

AA

A

AAAtA hA, A

A AA

(8)

r-nation is that the differences were present atskeletal maturity ("peak bone mass") and that subsequent rates ofbone loss

were similar. An alternative explanation is that there were

differences in the subsequent ratesof bone lossatsitesamong

individuals. The rates of bone lossmaydifferbetweenthe LS

and the UDR because of the different forces to which these bones are subjected: bone loss from the spine may be

de-creased byweight-bearingexercise and obesity. Also, therate

of bone turnover in the LS may begreaterthan that in the UDR because the vertebrae contain both red cellularmarrow

(containing theprecursorsof osteoclasts and osteoblasts) and yellow fatty marrow whereas the UDR contains only the latter. Theproportion of redtoyellowmarrowinthe vertebrae may

differbetween individuals and couldaccountfor differences in rate of bone loss.

Osteoporotic women with only vertebral fracture had a

much larger mean decrease in LS-BMD than did those with onlyColles' fracture. However, the mean decrease in UDR-BMD was only slightly larger in women with Colles' fracture than in those with vertebral fracture. Thus, thewomen with vertebral fracture are also at increased risk forColles'fracture but have not yet sustained thenecessar'traumasuchasfalling on the outstretched hand. An alternative explanation is that women with Colles' fracture are more likely to fall than those with vertebral fracture. In this regard, Crillyet al.(7)reported that women with Colles' fracture hadmoreposturalinstability

thanage-matched controls.

In summary, although patients with type I osteoporosis lose excessive amounts of bone from both the LS and the UDR, the amounts of bone lost at these two sitesvaryamong individuals. The relative decreases at thesetwositesmay

con-tributetothe developmentof the different fracturesyndromes. Whateverthe underlying cause of the unequal bone loss in type I osteoporosis, it appears thatsite-specific measurements of BMD are the best way to study these fracture syndromes

and the best way to estimate fracture riskprospectively. Acknowledgments

WethankJoanM.Muhsfor coordinatingthepatient studies;Darla J.

Jech,Robert L.Carlson, and WilliamL.Dunnforassisting withbone

densitymeasurements; andCherylK.Collins forpreparingthe

manu-script.

This investigation was supported inpart by research grant

AR-27065from the National Institutes ofHealth.

References

1.Riggs,B. L.,andL. J.Melton III. 1986.Involutional osteoporo-sis.N.Engl.J.Med. 314:1676-1686.

2.Riggs, B. L., andL. J.MeltonIII.1983. Evidencefortwodistinct syndromes of involutional osteoporosis.Am.J.Med. 75:899-901.

3. Nilsson, B. E., and N. E. Westlin. 1974. The bone mineral contentintheforearm ofwomen with Colles'fracture. Acta Orthop.

Scand.45:836-844.

4.Jensen, G. F., C. Christiansen,J. Boesen, V. Hegedus, andI.

Transbol. 1983. Relationshipbetween bone mineral content and

fre-quency of postmenopausal fractures. Acta Med. Scand. 213:61-63.

5. Hesp, R., L. Klenerman, and L. Page. 1984. Decreased radial

bonemassin Colles'fracture.ActaOrthop. Scand.55:573-575. 6.Harina, M.,andP.

_Karjalainen.

1986. Trabecular osteopenia in

Colles'fracture. Acta Orthop.Scand. 57:38-40.

7. Crilly, R.G., L. Delaquerriere Richardson, J. H. Roth, A. A. Vandervoort, K. C. Hayes, and R. A.Mackenzie.1987. Postural

stabil-ity and Colles' fracture. Age Ageing. 16:133-138.

8.Nordin, B. E. C., R. G. Crilly,andD. A. Smith. 1984.

Osteoporo-sis.InMetabolicBoneand Stone Disease. 2nd edition. B. E. C.Nordin,

editor. ChurchillLivingstone, Edinburgh. 1-70.

9. Krolner, B., E. Tondevold, B. Toft, B. Berthelsen, and S. P.

Nielsen. 1982.Bone massoftheaxial and theappendicularskeleton in

women with Colles' fracture: its relation to physical activity. Clin.

Physiol.

(Oxf). 2:147-157.

10. Schlenker,R.A., andW. W. VonSeggen. 1976. The distribu-tion ofcortical and trabecularbonemassalongthelengthsofthe radius andulna and theimplications forin vivobonemassmeasurements.

Calcif TissueRes.20:41-52.

11.Riggs,B. L., H. W.Wahner,W. L.Dunn,R. B.Mazess,K. P.

Offord, andL.J. MeltonIII. 1981. Differentialchangesinbone min-eraldensity of the appendicularand axial skeleton withaging:

rela-tionshiptospinal osteoporosis.J. Clin.Invest.67:328-335.

12. Wahner, H.W.,B. L.Riggs, andJ. W.Beabout. 1977. Diag-nosis ofosteoporosis: usefulness of photon absorptiometry atthe radius.J.Nuci.Med. 18:432-437.

13. Hanley,J. A.,andB.J.McNeil. 1982.Themeaning anduseof theareaunderareceiveroperatingcharacteristic(ROC)curve.

Radiol-ogy. 143:29-36.

14. Hanley,J.A.,andB. J.McNeil. 1983.Amethod ofcomparing

theareasunderreceiveroperating characteristiccurvesderived from thesame cases.Radiology. 148:839-843.

15. Awbrey, B. J., P.C.Jacobson, S.A.Grubb,W.H.McCartney,

L.M. Vincent, andR. V.Talmage. 1984. Bonedensityinwomen: a

modified procedure for measurement of distal radial density. J.

Orthop.Res.2:314-321.

16. Nilas, L., J. Borg, A.Gotfredson,andC.Christiansen. 1985.

Comparison of single- and dual-photonabsorptiometryin postmeno-pausal bone mineral loss.J.Nuci. Med.26:1257-1262.

17.

Ruegsegger,P., M.Anliker, andM.Dambacher. 1981. Quanti-fication of trabecularbonewith low dose computed tomography.J.

Comput.Assist. Tomogr. 5:384-390.

18. Hangartner, T. N., and T. R. Overton. 1982. Quantitative measurement of bone density usinggamma-ray computed tomogra-phy. J. Comput.Assist. Tomogr.6:1156-1162.

19.

_Kadjalainen,

P. 1973. A method for determination of the

min-eralcontentand mineraldensity of the distal radius usinggamma ray

attenuation.Ann.Clin.Res. 5:231-237.

20. Melsen, F.,H. E. Nielsen,P. Christensen, andL.Mosekilde. 1978. Some relations betweenphoton-absorptimetric and histomor-phometric measurements ofbone massin the forearm.Am.J.

Roent-genol. 131:541.(Abstr.)

21.Nilas,L., H.Norgaard,J.Podenphant,A.Gotfredsen,and C.

Christiansen. 1987.Bonecomposition inthedistalforearm. Scand.J.

Clin. Lab.Invest.47:41-46.

22. Johnson, L. C. 1964. Morphologic analysis in pathology:the

kinetics of disease and general biology ofbone. In BoneBiodynamics.

H. M.Frost, editor. Little, Brown andCompany, Boston. 550.

23.Pesch, H.-J.,H.-P.Scharf, G.Lauer, and H. Siebold. 1980. Der

altersabhangige Verbundbau derLendenwirbelkorper: eine Struktur-und Formanalyse. VirchowsArch. A Pathol. Anat. Histopathol. 386:21-41.

24. Nottestad,S. Y., J. J. Baumel, D. B.

Kimmnel,

R. R. Recker, and R. P. Heaney. 1987. Theproportion of trabecularbone inhuman vertebrae. J. Bone Miner. Res.2:221-229.

25.Ott, S. M.,R.F.Kilcoyne,andC.H.ChesnutIII. 1987.Ability of four differenttechniques of measuring bonemass todiagnose

verte-bral fractures inpostmenopausalwomen. J. Bone Miner. Res. 2:20 1-210.

26. Nilas, L.,J. Podenphant, B. J. Riis, A. Gotfredsen, and C.

Christiansen. 1987.Usefulness of regional bonemassmeasurements in

patients with osteoporosisfracturesof the spine and distal forearm.J.

Nucl. Med. 28:960-965.

27.Podenphant, J.,V. A.H.Nielsen,B. J.Riis,A.Gotfredsen,and

C.Christiansen. 1987. Bonemass,bonestructureand vertebral

frac-turesinosteoporotic patients.Bone.8:127-130.

28. Mazess, R. B. 1988.Diagnostic sensitivityof bonedensitometry (letter).J.Bone Miner. Res.3:121-122.