Genetic determinants of bone mass in adults A twin study

Genetic determinants of bone mass in adults. A

twin study.

N A Pocock, … , P N Sambrook, S Eberl

J Clin Invest.

1987;

80(3)

:706-710.

https://doi.org/10.1172/JCI113125

.

The relative importance of genetic factors in determining bone mass in different parts of the

skeleton is poorly understood. Lumbar spine and proximal femur bone mineral density and

forearm bone mineral content were measured by photon absorptiometry in 38 monozygotic

and 27 dizygotic twin pairs. Bone mineral density was significantly more highly correlated in

monozygotic than in dizygotic twins for the spine and proximal femur and in the forearm of

premenopausal twin pairs, which is consistent with significant genetic contributions to bone

mass at all these sites. The lesser genetic contribution to proximal femur and distal forearm

bone mass compared with the spine suggests that environmental factors are of greater

importance in the aetiology of osteopenia of the hip and wrist. This is the first demonstration

of a genetic contribution to bone mass of the spine and proximal femur in adults and

confirms similar findings of the forearm. Furthermore, bivariate analysis suggested that a

single gene or set of genes determines bone mass at all sites.

Research Article

Find the latest version:

Genetic Determinants of Bone Mass in Adults

A Twin

Study

Nicholas A. Pocock,* JohnA.Eisman,*John L.

Hopper,*

Michael G.Yeates,1 PhilipN.Sambrook,*and StefanEberI *Garvan Institute ofMedical Research, Sydney, Australia;

IDepartment

ofNuclearMedicine, St. Vincent'sHospital, Sydney, Australia; andtEpidemiology Unit, Faculty ofMedicine, University ofMelbourne, Melbourne,Australia

Abstract

The

relative importance

of

genetic

factorsin

determining

bone

mass in different parts of the skeleton is

poorly

understood.

Lumbar spine and proximalfemur bone mineral

density

and forearm bone mineral content were measured byphoton

ab-sorptiometry

in38

monozygotic

and 27

dizygotic

twinpairs. Bone mineral

density

was

significantly

more

highly

correlated

inmonozygotic than in

dizygotic

twins for the

spine

and proxi-mal femur and in the forearm ofpremenopausal twin pairs,

which

is

consistent

with

significant genetic

contributions to

bone mass at all these sites. Thelessergeneticcontribution to

proximal femur and distal forearm bonemass

compared

with

the spine suggests that environmental factors are ofgreater

importance

inthe

aetiology

of

osteopenia

of the

hip

andwrist.

This is the first demonstration ofa

genetic

contributiontobone

mass of the

spine

and

proximal

femur in adults and confirms similar

findings

of the forearm.

Furthermore,

bivariate

analy-sis suggested

thata

single

geneor setof genes determinesbone

mass at all sites.

Introduction

Osteoporosis

isa major health

problem

of Western societies

that affects up to halfof the

elderly

female

population (1).

Osteoporosis-related

fractures are an

increasing problem

in

aging

menand women. In women, the sudden declinein

es-trogen

production

atthe menopauseis animportant

aetiologi-cal

factor

in the

subsequent

accelerated rateof bone mineral loss

(2,

3).

However, the

significant

prevalence ofthe disease

in

aging

men and

its

absence in some

postmenopausal

women

indicate

that

other factors

are also

important

in

determining

bone mass and hence the risk of

osteoporotic

fractures.

Envi-ronmental

factors

such as tobacco and alcohol

use,

physical

activity,

and

body

weight

have all been

reported

to

influence

bone mass

(4-10).

A

genetic

contribution

to

bone

mass has

previously been

reported

for

the

bones

of

the upper

limb

(1

1-13).

One study

found

a

genetic

component

of spinal

bone mass

in

individuals

less than

25

years

old, but

not

in

subjects

older than 25

(11).

To our

knowledge,

no data have been

published

in

relation

to a

possible genetic contribution

to bone massinthe

clinically important

site

of

the proximal femur.

Address correspondence to Dr. Eisman, GarvanInstituteof Medical Research, St. Vincent's Hospital, Sydney, NSW 2010, Australia.

Receivedfor publication 27 August 1986 and in revised form 30 March 1987.

The study of twins providesaunique methodtoinvestigate the

importance

ofgenetic and environmental factors in

deter-mining bone

mass.Wereporthere the results ofatwin

study

examining heritability of bone massin the lumbar spine,

proximal

femur,

anddistal forearm.

Methods

The twinpairs studiedwerevolunteersobtainedthroughthe Austral-ianNational Health and Medical Research Council Twin Registry and from appeals through the media. Only adult twins ofthe same sex were

studied,andinformed writtenconsentwasobtained from all partici-pants. Twins were only excluded from analysis on the basis of disease, suchasrheumatoidarthritis,or useofmedications,suchas corticoste-roids, which may have affected bone density inone orbothtwins. 65 twinpairswerestudied in full.

Thezygosity ofthe twinswasdeterminedfrom theirown classifica-tion.This has been showntobeaccuratetowithin5%andis compara-ble withclassification bymoresophisticated and extensive investiga-tion (14).

The study groupcomprised 32 female and 6 male monozygotic

(MZ)'twin pairs, as wellas26female and 1 male dizygotic (DZ) twin

pairs. 13of the MZ and4of theDZfemale twinpairswere postmeno-pausal. Two other DZ twin pairswere discordant for menopausal status;onetwinof eachpairwaspremenopausal and the other post-menopausal(four years in each case).

Bone mineral density (BMD, grams per square centimeter)was measured in the lumbarspine (L2-L4)andright proximal femur using aDP3 dual photonabsorptiometer (Lunar Radiation, Madison, WI). Dualphotonabsorptiometryutilizes the relative transmission of pho-tonsoftwoenergies (44and100 keV), emitted fromaGadolinium 153 radiation source, todetermine bone mineral content (BMC) (15). BMDwasderived by dividing the BMC ofeach region by the projected bonearea.Intheproximalfemur three sitesweremeasured: the femo-ralneckat atrans-cervical position,

the

trochanteric region, and Ward's triangle within thefemoral neck. All femoral scans were

checked,without knowledge of zygosityorbonedensity estimate, to ensure there was no difference in positioningof region of interest between twinpairs. The radiation dose to the skin and gonads was <200 and 100

gGy,

respectively (10-20 mrad). As we have described previously the coefficient of variation was1.4%over36 determinations (weekly)with cadaveric vertebrae and 2.6% on repeated determina-tionsin five normal volunteers(16), consistent with published val-ues(17).

Forearm BMC(units percentimeter) of the distal radius and ulna wasmeasured using a single photon densitometer (Molsgaard Instru-ments, Copenhagen,Denmark). Scanning was commenced at a site

correspondingto8mmseparation between the radius and the ulna; five subsequent scans were performed, each 4 mm more proximal. ForearmBMC was calculated from the mean of the six scans. At the

8-mmsite the radiusandulnaare comprised of- 20and 12%

trabecu-1. Abbreviations used inthispaper: BMC, bone mineralcontent; BMD, bone mineral density; DZ, dizygotic;MZ, monozygotic;

V02max, maximumoxygenuptake.

706 Pocock, Eisman, Hopper, Yeates, Sambrook,andEberl J.Clin. Invest.

© The American Society for Clinical Investigation, Inc.

0021-9738/87/09/0706/05

$2.00

lar bone, respectively, whileat the mostproximalsite the radius and ulnaareboth - 5% trabecular bone(18). The coefficient ofvariation

onrepeatedmeasurementsinnormal volunteers was 1.5%. Measure-mentsofBMD were made without knowledge as to the zygosity of the twins.

Lumbarspine radiographswereobtainedin all subjects older than 40years.Scansofeachtwinwere analyzed with reference to relevant X rayswithout knowledge ofthe sibling's results. The lumbar BMD estimateswereexcluded fromanalysis for six twin pairs because ofthe presenceof spinal osteoarthritis, whichmayfalselyelevate estimates of spinal bone density. Vascular calcificationwas not aproblem inany subject.Nosubjectin the study hadapriorhistory of renal diseaseand all had normalrenalfunctionasassessedby creatinine clearance (avail-able in 100subjects)and/oranormalserumcreatinine. Weight (kilo-gram) and height (meter)weremeasured inallsubjectsand body mass index (kilogrampersquared meter)wascalculated. Physical fitnesswas estimated in 26 MZand 23DZ twinpairs by measurementof pre-dicted maximaloxygenuptake (VO2max, litersperminute) according tothe criteria of AstrandandRyhming (19). Subjectswereexercisedat aknown work loadon abicycle ergometer. Theplateau pulse rate, steady forat least 2 min, wasused in conjunction with the loadto estimate the V02max accordingto the nomogram ofAstrand and Ryhming (19).

Statistical

Methods

Analyses of twin studiesassume thatintrapair variance ofMZand hencegenetically identical twins isdue toenvironmental factorsand measurement error,while intrapairvariance inDZtwins is addition-ally affected by genetic factors.Itis also assumedthat common envi-ronmental factorsaresharedto asimilarextentbetweenMZ and DZ twins.Comparison of the correlation ofBMDinMZtwinpairs with that inDZtwinpairscanthereforeprovidea meansofdetermining the genetic contributiontoobservedvariation inBMD.Thus,atany par-ticular location:BMD=, +G+C+E,whereMis themeanBMDfor

thatlocation (andmayitself beafunction of measured variables such asage, sex, years postmenopause,etc.), Grepresentsgenetic determi-nantsof bonemass,Crepresents commonenvironmental factors

af-fectingbone mass,andE representsenvironmental factorsparticular

toanindividualincludingmeasurementerror.Furthermore,for each twin, G, C, andEareassumedtobeindependent,normally distributed random variables with meansequal tozeroand non-negative vari-ances

_a82,

0

2,

and

Oa2,

respectively. TherelationshipofBMDintwin pairs

(BMD1

,BMD2)asmeasuredby covariance (Cov)canthereforebe expressed as:

Cov(BMD1,BMD2)

=

_Cov(G,,G2)

+ Cov(C,,C2). For MZpairs the correlation of

GI

andG2 = 1, while forDZpairsthe correlation of

GI

and G2=1/2(20).Theclassical twin modelassumes

thecorrelation ofCl and C2= 1 independentofzygosity.Hencefor MZpairs: Cov(BMD,,BMD2)=_Cov(MZ) =

_ag2

+a2, while forDZ pairs

Cov(BMD1

,BMD2)=_Cov(DZ)=1/2

ag2

+

Cf 2,

andtherefore

Cov(MZ)=Cov(DZ) if and only if

Cfg2

=0.

In summary,foranytrait (e.g.,BMD),demonstration ofasignificant

difference incovariance,and hence in thecorrelation,betweenMZ twin pairsandDZtwinpairsisconsistent withasignificant genetic

determinant of variationin thattrait.

Estimationofparameters. Thecorrelations ofBMDand BMCat

different sitesin MZ and DZtwinpairswerecalculatedby maximum likelihoodasoutlinedbelow. In theanalysesthemeanwasfitted

inde-pendently ofzygosity andas alinear functionofageformales and premenopausal females,oralinearfunctionofyears postmenopause for postmenopausal females. Inallanalysesthemean wasallowedto

differbetweenlocations.Testsoffit for outlierswereperformedafter HopperandMathews(21):In noinstancewasthereevidenceofapoor fit.For eachtwinpair,ateachlocation,it is assumedthat BMD hasa

bivariate normaldistribution with covariancesthat can be expressed as above intermsof

_CTg2

andoCT2. Under the restrictionthat all are non-negative, the variancecomponents

CT82,

0C,2,

and

_,2

areestimated by maximum likelihood(22)usingthealgorithm Fisher (23, 24). After conventionwedefineheritabilitytobe h=

_aCg2/02,

andsimilarly define

c =

_oj2/12,

ande=

CT2/oa2,

where

or2

=

og2

+aT2+aC.Thatis,h, c, and e

aretheproportionsoftotal variation explained by genetic,common

environmental,and otherfactors, respectively. Thecorrelations be-tween twins were estimated by maximum likelihood as rMz

=

_Cov(MZ)/la2,

and rDz = Cov(DZ)/a2, with both Cov(MZ) and Cov(DZ)notconstrained.

Thelikelihood ratiocriteriaisused to test thenullhypothesisOrg2

=0,i.e., heritability=_0.The maximum log likelihoodwascalculated undertwomodels with

_Tg2

=_{0 and}

_aCg2

_{20.Under the null}hypothesis, twice the absolute difference inlog likelihood between these models willbeasymptotically distributedas a 50:50 mixture of a x2 variate and

apointmassattheorigin(21).

Bivariateanalyses. ThesimilaritybetweenthefactorsG,C, andE

determiningBMD atdifferent skeletal siteswasinvestigated. This sta-tisticalanalysiswascarriedoutinthepremenopausal female twin pairs whoconstituted the largest subgroup.Allpossiblepairs of skeletal sites wereanalyzed using bivariate analysis.Thus,foreachlocation(k) and twin(i)(i= _1,2),

_BMDA,

=_Ak+

Gk,

+Cki+Eki,with corresponding variance components

rgk(2,

0CT2 and

aek2

as above. For each pairof skeletal siteskandm,we calculatedPg, Pc, andPe, thecorrelations between theG,C, andE components,respectively,atdifferent loca-tions in thesameindividual.Thatis,Pg=

_{Corr(Gkj,Gmj),}

independent ofi, sothat, fortwoindividuals, i and

j,

ifi #jfor MZ

pairs,

Cov(Gki,Gmj) =

_pgCagkagm,

while forDZ

pairs Cov(Gki,Gmj)

= ½/2

pgCagkgC,. Also, Cov(CkCmj) =Pcackar, and Cov(EkiEmj) = PeCekcaem

independentofzygosity. Thecorrelationsp =

_{Cor(Yki,Ymi),}

between BMDatdifferent sites in thesameindividualwerealsocalculated

by

maximum likelihoodasoutlinedabove.

Results

Themeanage

of the

MZ

twins

was47 yr

(range,

24to75

yr),

and

for the

DZ

twins

40 yr

(range,

24to65

yr).

Themean

body

mass

index

of

the MZ

twins

was 23

kg/M2

(range,

17 to 31

kg/m2)

and

for

the DZ

twins,

23

kg/mr2 (range,

18to36

kg/M2).

There

was no

significant

difference between

the

correlation of

body

mass

index

in

the

MZ

twins

and inthe DZ

twins

(0.39

and

0.25, respectively).

Themean

VO2max

of the female MZ

twins

was 34

ml/kg

per min

(range,

18to 57

ml/kg

per

min)

and

for the

DZ

twins,

36

ml/kg

per min

(range,

21 to

62

ml/kg

per

min).

Therewas no

significant

difference

between the

in-trapair

correlation of V02max, in

the MZ

twins

(r

=

_0.88)

and

in the

DZ

twins (r

=

_0.81).

Therewas no

significant

difference

between the mean BMD or BMC at any

site

in the

twins

compared with

values

from

age-matched

controls.

Also,

forno

skeletal site

were the means or

interpair

variances

different

between MZ

and

DZ

pairs.

The

correlations of

BMDat the

different

skeletal

sites

are

shown

in Table

I

(top)

for all

MZ and DZ

twin

pairs.

Atall

sites

the

correlation

between MZ

twins

wasgreater than that between DZ

twins. Estimates of

heritability

derived from

maximum likelihood

analysis

arealso listed.Fromthis

analy-sis

we

found

that

heritability

was

significantly

greater thanzero

atall

locations

except the

distal

forearm.

Analysis

excluding

thetwo

twin

pairs

discordant for years postmenopause didnot

alter the results.

However,

analysis

of the

correlations of

BMD at the different skeletal

sites confined

to the

premenopausal

female twin

pairs (Table I, bottom)

resulted in somewhat

dif-ferent estimates

of

heritability.

Inthis

_subgroup

atall sites the

correlation

between MZ

twins

was greater than between DZ

TableI. TwinCorrelation andHeritability ofBone Mass

Parameter rMZ rDZ Heritability

All twin pairs

LumbarBMD 0.92 0.36 0.92 (P<0.001)

Proximal femurBMD

Femoral neck 0.73 0.33 0.73 (P<0.005)

Ward'striangle 0.85 0.36 0.85 (P<0.001)

Trochanteric 0.75 0.47 0.57 (P<0.02)

Forearm BMC 0.71 0.50 0.42 (P<0.08)

Premenopausal twinpairs

Lumbar BMD 0.92 0.36 0.92 (P<0.001)

Proximal femurBMD

Femoral neck 0.77 0.56 0.46 (P<0.08)

Ward'striangle 0.88 0.48 0.87 (P<0.001)

Trochanteric 0.88 0.56 0.66 (P<0.005)

Forearm BMC 0.88 0.42 0.88 (P<0.001)

Intrapaircorrelation coefficients forMZ(rMZ) andDZ(rDZ) twins andheritabilityfor the five skeletalsites,determinedfrom analysis of alltwinpairs(top) and fromanalysisof only premenopausal twin pairs (bottom).Heritabilitywascalculatedusingmaximum log likeli-hood with the constraint that

ol,

ac2,and

ael

wereall greater thanzero.

and

two

sites

inthe

proximal

femur

(i.e.,

Ward's

triangle

and

the trochanteric

region). However,

heritability

didnot reach

significance

atthe

femoral

neck. Forearm

BMC,

onthe other

hand, showed

significant

heritability

in

the

premenopausal

women. The

correlations

of

lumbar

spine

and

femoral

neck BMD as

well

as

forearm BMC in all the

MZ

and

DZ

twins

are

shown in

Figs. 1-3, respectively.

The range

of

BMDobserved

E

-a

SE 1.616"

E

C~

cs

1.2-0

a

co

Z

0.8-0

z .8

E

a

IL 0.4

Ne 1.6-E

N

S 1.2

4-0 z

8 0.8

0

E

S U.L

A

a.

o

op. o

1.6

0.4 0.8 1.2

Femoral Neck BMDTwin1-g/cm

Figure 2.Correlation of femoral neckBMDbetween (A)MZtwins and(B) DZ twinsshowingmale(m) and female (o) twin pairs. The line ofidentity is shown.

in

the

twins studied

was the same as we have

observed in

a

normal

population. The close

correlation between the twin

pairs

occurs

throughout

the range

of

BMD.

Table

II shows the cross-correlations of BMD in

premeno-pausal

women at

different sites

in the same individual and the

estimated

cross

correlations

Pg,

Pc,

and Pe

for

each pair of

locations. The estimates of

pg

arepositive between all sites but

E

U

m

I-0

U.

2.01 A

1.6-

1.2-am

a

aa

a0

00

NE

c

0

a

m E

3

E

.2

c

0

a m

0

U.

0.6 1.0 1.4 1.8

2 Lumbar BMD Twin 1-g/cm

Figure1.Correlation of lumbar spine BMD between (A) MZ twins and(B)DZtwins showing male (-) and female

(u)

twin pairs. The line ofidentityis shown.

2.01 B

1.6 - a a a

a a

a

2.0

0.8 1.2 1.6

Forearm BMC Twin1-U/cm

Figure 3. Correlationof forearm BMC between (A) MZ twins and (B)DZtwins showing male

(i)

and female (o) twin pairs. The line of identity isshown.

Table II. Correlations between DeterminantsofSkeletal Mass at

Different

Skeletal Sites

Femoral Ward's

neck triangle Trochanteric Forearm

Lumbar

P 0.59 0.57 0.52 0.58

Ps

0.80 0.77 0.62 0.41

Pc 0.82 0.64 0.60 0.53

Pc -0.55 -0.50 -0.27 0.56

Femoral neck

p 0.91 0.84 0.43

Pg 0.98 0.97 0.69

Pc 0.98 0.95 0.55

Pe 0.64 0.36 -0.47

Ward'striangle

p 0.80 0.38

PS

0.80 0.64

Pc 0.91 0.31

Pe 0.58 -0.37

Trochanteric

P 0.44

PS 0.48

Pc 0.55

Pe -0.21

Correlations (p) betweenBMDatdifferent skeletal sites. The correla-tions between genetic

(p),

commonenvironmental(Pc),and individ-ualenvironmental(Pe)components of bone mass at different skeletal sites in thesameindividualareshown.

are greater between the

three

proximal femur

sites than be-tween

these sites and

the

lumbar

spine

and forearm. The esti-mates

of

pe are

positive between femoral

neck, Ward's

triangle,

and

trochanteric, and between

lumbar and forearm, but are

negative

between the

former

three

locations

and the

latter

two.

Estimates of heritability from bivariate analyses,

which

impose

a greater structure on

the

cross-covariances,

are lower than

those estimated from univariate

analysis

butnevertheless are

consistently

between 0.4 and 0.6.

Discussion

These

data

demonstrate for

the

first time

in

adults

a strong

genetic

component to

the determination of

bonemassin the

spine

and

proximal

femur.

They

also confirm the

previously

reported

genetic contribution

to bonemassin

the

appendicu-lar skeleton of the

upper

limb ( 11,

13).

Analysis of

all

twin

pairs

studied,

aswellasseparate

analy-sis of

the

premenopausal twins,

suggestsagreater

genetic

deter-minant

of

BMD inthe lumbar

spine

than in

either

the

proxi-mal

femur

or the

distal forearm.

The apparent greater

contri-bution of

genetic

factors

in

determining

forearm

bonemassin

premenopausal

women

compared with

thegroup as a whole suggests that

environmental factors

are

of

increasing

impor-tance

after

the menopause

and/or with

advancing

ageat

this

site.

The

bivariate

analysis of

BMD in the

premenopausal

women,

demonstrating

a

significant correlation

between the

genetic components of BMD at different locations, is consis-tent with one gene or a single set of genes determining bone mass at all skeletal sites measured. There were also strong correlations between common environmental components across the different locations. However, there was a negative association between individual environmental components of the sites in the proximal femur and both the lumbar and the forearm locations. This suggests that although there are genetic andenvironmental components common for bone density at all

locations,

there are some environmental factors specific for bone density at the hip compared with the lumbar spine or forearm.

A genetic contribution to bone mass in the upper limb has been previously suggested by single photon absorptiometry

(11,

13) and by

metacarpal

morphometry (12). Dequeker et al. (11) have shown, by absorptiometry, a genetic contribution to spinal bone mass in individuals younger than 25

yr,

as well as to distal radius BMC in individuals > 25 yr. The values for heritability reported by that study, i.e., 0.88 and 0.75, respec-tively, were calculated as described by Holzinger (25): H

=(DZ intrapair variance - MZ intrapair variance)/DZ intra-pair

variance.

Similar values derived from our data are 0.85 for the lumbar spine, and in the premenopausal females, 0.78 for the distal forearm, which is virtually identical to those reported byDequeker et al. (1 1).

To our knowledge, there are no previous studies that have

examined

the role of genetic factors in determining bone mass of the proximal femur, the site of the most clinically severe osteoporosis-related fractures. That our data suggest a smaller genetic

contribution

to bone mass of the proximal femur and forearm compared with the lumbar spine imply that environ-mental factors play a more dominant role in determining vari-ation of bone mass at these sites. This is consistent with the

observation

that skeletal sites in the forearm and proximal femur are exposed to large individual variations in mechanical loading; e.g., the load on the upper and lower limbs are highly variable between

occupations.

However, the spine bears the weight of the upper half of the body with relatively little varia-tion during waking hours except in

occupations

involving heavy labor. Even the

transition

from a sitting position to a walk does not result in a large change in

mechanical

load on the spine, but results in a major change in

mechanical

stress exerted on the hip. We have previously demonstrated (9) that physical fitness

(VO2max),

and by

implication

habitual physi-cal activity (26, 27), is

important

in

determining

BMD in the femoral neck. The high correlation of

VO2max

between MZ twins was similar to that between DZ twins. This suggests that both MZ and DZ twin pairs are

concordant

for habitual physi-cal activity. This

concordance

could contribute to the relative

similarity

of intrapair

correlation

of proximal femur BMD in MZ and DZ twins.

In

summary,

this study has

demonstrated

for the first time a significant genetic

contribution

to bone mass in the spine and proximal femur in adults and confirms the previously reported genetic

contribution

to upper limb bone mass. The smaller genetic

determinant

ofbone density in the hip and forearm compared with the lumbar spine, and by implication the

importance

of

environmental

factors, supports the findings of our previous study (9) and suggests a greater

potential

for

lifestyle intervention in

achieving

a reduction in the

incidence

studies (e.g., restriction fragment length polymorphism studies of appropriate genes)toidentifyindividuals at risk.

Identifica-tion of such individuals would allow early life style (e.g.,

exer-cise, diet) and other therapeutic interventions and possibly

preventthe development of clinical disease.

Acknowledaments

We areindebtedtothe Australian National Health and Medical Re-searchCouncil TwinRegistryfor assistance inrecruitingvolunteers for this study andtoDr.Barry Wren andTheCenterfor the Management of the Menopause for theuseof thesingle photon bone densitometer. Thework was supported by SandozAustralia,the Australian and NewSouth Wales Dairy Corps., the National Health and Medical Research Council,and by the Garvan Research Foundation.

References

1. Stevenson,J.C., andM.I. Whitehead. 1982. Postmenopausal osteoporosis.Br.Med.J.282:585-588.

2. Genant,H.K., C. E.Cann,B.Ettinger,andG.S.Gordan. 1982. Quantitativecomputed tomographyof vertebral spongiosa:asensitive method fordetecting early bone loss after oophorectomy.Ann. Intern. Med. 97:699-705.

3.Krolner,B.,and S.PorsNielsen. 1982.Bonemineralcontentof thelumbarspine in normal and osteoporoticwomen: crosssectional andlongitudinalstudies. Clin. Sci. 62:329-336.

4.Aloia,J.F.,S.H.Cohn,A.Vaswani,J. K.Yeh,K. Yuen, and K. Ellis. 1985. Risk factors for postmenopausal osteoporosis.Am. J.Med. 78:95-100.

5. Brewer, V., B. Meyer, M. S. Keele, S. J. Upton, and R. D. Hagan. 1983. Roleof exercise inpreventing involutional bone loss. Med. Sci. Sports Exercise. 15:445-449.

6.Daniell,H.W. 1976.Osteoporosis of the slender smoker. Verte-bralcompressionfractures and loss of metacarpalcortexinrelationto postmenopausalcigarette smokingandlackof obesity.Arch.Intern. Med. 136:298-304.

7.Nilsson,B.E., and N. E.Westlin.1971. Bonedensity in athletes. Clin. Orthop.Relat. Res. 77:179-182.

8.Oyster, N., M.Morton, and S. Linnell. 1984.Physical activity

andosteoporosisin postmenopausalwomen.Med. Sci.SportExercise. 16:44-50.

9. Pocock,N. A., J. A.Eisman,M. G. Yeates, P. N. Sambrook, and S. Eberl. 1986.Physical fitness isamajor determinant of femoralneck and lumbarspine bone mineral density.J.Clin.Invest. 78:618-621.

10.Vernejoul,M.C.,J.Bielakoff,M. Herve, J. Gueris, M. Hatt, D. Madrowski,D.Kuntz,L.Mirovet,and A. Ryckewaert. 1983. Evidence fordefectiveosteoblast function.Aroleforalcohol and tobacco con-sumption in osteoporosis in middle aged men. Clin. Orthop. Relat. Res. 179:107-115.

11. Dequeker, J.,J. Nijs,A. Verstraeten, P. Geusens, andG. Gevers. 1987. Genetic determinants of bone mineralcontent atthe spine and radius:atwinstudy.Bone(NY).Inpress.

12. Moller, M., A.Horsman,B.Harvald,M.Hauge,K. Henning-sen, and B. E. C.Nordin. 1978. Metacarpal morphometry in monozy-gotic and dizymonozy-gotic elderly twins. Calcif Tissue Res. 25:197-201.

13.Smith, D.M.,W. E.Nance, K.W.Kang, J. C.Christian, and C. C. Johnston. 1973. Genetic factors indetermining bone mass.J. Clin.Invest.52:2800-2808.

14.Hrubec, Z., and C.D. Robinette. 1984. Thestudy of human twins in medicine and research.N.Engl. J. Med. 310:320-339.

15.Kr0lner,B.,and S.PorsNielson. 1980.Measurementof bone mineralcontent(BMC) ofthe lumbar spine.1.Theory andapplication

ofa new twodimensional dual photon attenuationmethod.Scand.J. Clin.Lab.Invest.40:653-663.

16. Pocock,N. A.,J.A.Eisman, M. G. Yeates,P. N.Sambrook, S. Eberl, and B. G.Wren.1986.Limitations of forearm bone densitome-try as anindex ofvertebral orfemoralneckosteopenia. J.BoneMin. Res. 1:369-375.

17. Riggs, B.L.,H. W.Wahner,E.Seeman, K. P.Offord, W.L. Dunn,R. B.Mazess, K.A.Johnson, andL. J.Melton.1982. Changes in bone mineraldensityof the femur and spine with aging.J. Clin. Invest. 70:716-723.

18.Schlenker,R. A. 1976. Percentages of corticaland trabecular bonemineralmassin theradius and ulna.Am.J.RadioL.

126:1309-1312.

19. Astrand,P.O., andI.Ryhming. 1954.A nomogramfor calcu-lation of aerobic capacity (physical fitness) from pulse rate during submaximal work.J.AppL.Physiol. 7:218-221.

20. Fisher,R. A. 1918. Thecorrelation betweenrelativesonthe supposition of Mendelian inheritance. Trans.R.Soc.Edinb. 52:399-433.

21.Hopper, J. L.,and J. D. Mathews. 1982.Extensionsto multi-variate normal models for pedigree analysis. Ann. Hum. Genet. 46:373-383.

22.Kendall,M.G., andA.Stuart. 1973. The advanced theory of statistics, Vol. 2. Charles Griffin & Co. Ltd.,London.234-272.

23. Lange, K.,J.Westlake,and M. A.Spence. 1976. Extensionsto

pedigree analysis. Variancecomponentsby thescoring method.Ann. Hum.Genet.39:485-491.

24. Lange, K., M. Boehnke, andD. Weeks. 1986.Programsfor Pedigree Analysis. Department of Biomathematics, UCLASchoolof Medicine.

25.Holzinger,K.J. 1929. The relative effect ofnatureand nurture influencesontwindifferences. J. Educ.PsychoL. 20:241-248.

26. Astrand,P.O.,andK. Rodahl.1970.Physicalworkcapacity.In Textbook ofWork Physiology. McGraw Hill, Inc., New York. 279-318.

27.Cotes,J.E., C.T.M.Davies,0.G. Edholm, M. J. R. Healy, and J. M. Tanner. 1969. Factorsrelatingtotheaerobiccapacity of 46 healthy British malesandfemalesaged 18 to28 years. Proc. R. Soc. Lond. (Biol.) B174:91-114.