Appl. Radiat. lsot. Vol. 49. No. 5/6, pp.

685-686, 1998

Pergamon
P I h S0969-8043(97)00090-0

@ 1998 ElsevierScienceLtd. All rights reserved

Printed in Great Britain
0969-8043/98 $19.00+ 0.00

Variation in Bone Mineral Density Between

Different Anatomical Sites in a Normal Local
Population
C. F. N J E H

a n d C. M . B O I V I N *

Department of Nuclear Medicine, Queen Elizabeth Hospital, Birmingham B15 2TH, U.K.
Bone mineral density (BMD) measured using dual energy X-ray absorptiometry (DXA) can be expressed
in terms of standard deviations, above or below mean young adult T-scores and above or below
age-matched Z-scores. The differences between the left neck of the femur (LN), right neck of the femur
(RN) and lumbar spine (L2 L4) were statistically significant, irrespective of whether expressed as T-scores
or Z-scores. Therefore skeletal status assessment should involve as many sites as practically possible.
Considering the low radiation dose and short scanning time, it is suggested that at least the two femurs
and lumbar spine BMD be used in routine osteoporosis risk assessment. ~) 1998 Elsevier Science Ltd.
All rights reserved

Introduction
Osteoporosis is characterised by low bone mass and
micro-architectural deterioration of bone tissue,
leading to increased risk of fragility fractures. Bone
mineral density (BMD) measured using D X A is
currently the most widely used method of quantifying
the development and degree of osteoporosis. Some of
the clinically available machines such as the Lunar
D P X - L have the capability to measure B M D at
various sites, including spine-PA, lateral spine,
proximal femur, hand and total body. Using the
W H O definition (WHO, 1994), a patient is osteoporotic if their B M D is 2.5 SD or more below the
young adult mean (i.e. T-score < - 2.5). It has been
reported that using a T-score of - 2.5 as a cut-off
point, the number of patients classified as osteoporotic varied between sites (Njeh et al., 1996).
Also, it has been reported that degenerative joint
disease affects B M D measurements (Yu et al., 1995).
To reduce variation due to pathological conditions
such as rheumatoid arthritis and osteoarthritis, we
restricted our study to a local normal population. The
most commonly measured sites are the spine and
proximal femur. However, the number of sites
measured varies between centres. This study investigated retrospectively site differences in measured
B M D in a local normal population.

Method
B M D at the spine (L2 L4), right and left proximal
femur (neck of femur, Ward's triangle and
trochanter) were measured using a Lunar D P X - L
*To whom all correspondence should be addressed.
685

densitometer. The manufacturer's recommended

positioning and analysis protocols were used. Local
ethical approval was granted for this study. Healthy
volunteers were recruited from the local population
by sequential referral by G P and the measurements
were carried out with their informed consent. 1740
male and female subjects with a mean age of
51.3 + 13.0 yr were recruited. Each subject was given
an explanation of the procedure and X-ray radiation
dose involved. A questionnaire was used to eliminate
those with pathological conditions that might affect
their BMD. Subjects were also excluded if they
regularly used certain medications, had a history of
fracture, alcohol abuse or prolonged immobilisation.
The B M D values were expressed as T- and Z-scores.
Simple linear correlations were computed for
different regions. The difference between sites was
examined using the Student paired t-test.

Resultsand Discussion
There were highly significant correlations between
B M D at all regions in both left and right femurs with
the coefficients varying from 0.74 to 0.93 (p < 0.0001)
(Table 1). There was a consistently better correlation
between identical regions on the two femurs
(r = 0.92, SEE = 0.057 g/cm2). A moderate correlation between femur regions and the lumbar spine
(r = 0.69-0.70, p < 0.0001, SEE = 0.106-0.136 g/
cm 2) was observed. These results are consistent
with previous reports (Faulkner et al., 1995a).
The mean ( + SD) T-scores of the left femoral
neck, right femoral neck and L2-L4 lumbar
spine were - 0.480__ 1.198, - 0.430 + 1.197 and
- 0 . 5 8 0 _+ 1.574, respectively. The differences between these T-scores were statistically significant

686

C.F. Njeh and C. M. Boivin

Table 1. Linear regression correlations between regions in the left femur, right femur and lumbar spine (L2-L4), p < 0.0001
Right
Left
Spine
Ward's triangle
Trochanter
Neck
Ward's triangle
Trochanter
L2-L4

Right neck
Right Ward's triangle
Right trocbanter
Left neck
Left Ward's triangle
Left trocbanter

0,93

0.83
0,79

Table 2. A Student paired t-test demonstrating the difference

between the T-scoresand Z-scores of the left neck of the femur (LN),
right neck of the femur (RN) and lumbar spine (L2-L4). At 95%
confidence, absolute significant t-value is > 1.96
T-scores
Z-scores
t-Value
p - V a l u e t - V a l u e p-Value
LN RN
-4.276
<0.0001 - 4.09
< 0.0001
LN L2 L4
3.624
0.0003
15.02
< 0.0001
RN L2-L4
5.430
<0.0001
17.02
< 0.0001

(Table 2). There was still significant differences when

B M D was expressed as Z-scores. The distribution of
the differences was skewed towards older age.
The observed variation between sites could be due
to measurement errors a n d localised anatomical
variation. M e a s u r e m e n t errors are r a n d o m ( p h o t o n
variation) a n d systematic (possible asymmetric
positioning of left and right legs and asymmetry
between femurs in the analysis algorithm). Short term
precision (CV) in this clinic has been reported to be
1.6 _+ 0.8% for the neck of the femur (Lilley et al.,
1991). So, if the difference between necks o f femurs
was solely due to m e a s u r e m e n t error, then 95% o f the
studied population would have a difference o f no
more t h a n 4.5%. (This is because the m i n i m u m
detectable significant (5% level) difference between
two measurements in a single subject is 2 , ~ CV).
However, a difference of more t h a n 4.5% was
observed in 4 2 % of the volunteers. Hall et al. (1991)
also observed a mean percentage difference in B M D
between the two femurs greater t h a n the CV o f hip
measurement. A l t h o u g h F a u l k n e r et al. (1995a) did
not observe a significant difference between the m e a n
values of left and right femurs, they noted that the
observed individual left/right differences were
occasionally large.
The differences between the spine and neck of the
femur could be accounted for by a n u m b e r of factors.
The two sites might be exposed to varying applied
loads and differences in the a m o u n t of cortical and
cancellous bone. Undiagnosed pathological conditions such as osteoarthritis could influence the
variation. Lastly, the generation of Z-scores using the
m a n u f a c t u r e r ' s n o r m a l range may have an inherent
variation between femur a n d l u m b a r spine ( F a u l k n e r
et al., 1995b).

0.92
0.87
0.80

0.88
0.92
0.75
0.93

0.79
0.74
0.92
0.83
0.79

0.70
0.70
0.70
0.70
0.70
0.69

Conclusions
One would expect to observe a difference in
T-scores between hip and spine, but not Z-scores
because of the different rates of bone loss. The
differences observed here indicate t h a t either there is
a genetic variation or there are pathological causes
which were not/could not be scrutinised by the
questionnaire. We conclude that B M D expressed as
T-scores or Z-scores may vary between sites and
therefore skeletal status assessment should involve as
m a n y sites as practically possible. Considering the
low radiation dose and short scanning time, it is
suggested that at least the two femurs a n d l u m b a r
spine B M D be used in routine osteoporosis risk
assessment. This will increase the sensitivity of B M D
for diagnosis of osteoporosis. In cases where only one
femur was measured, longitudinal studies should be
carried out on the same femur otherwise false
deductions might be made. There is the need to
reassess the use of Z-scores and T-scores in
osteoporosis risk assessment, in particular to see if
left and right hip measurements better predict
fractures o f the same hip.

References
Faulkner, K. G., Genant, H. K. and McClung, M. (1995a)
Bilateral comparison of femoral bone density and hip axis
length from single and fan beam DXA scans. Calc(f.
Tissue Int. 56, 26-31.
Faulkner, K. G., Roberts, L. A. and McClung, M. R.
(1995b) Discrepancies in normative data between Hologic
and Lunar systems. J. Bone Miner. Res. 10 1, S146.
Hall, M. L., Heavens, J. and Ell, P. J. (1991) Variation
between femurs as measured by dual energy X-ray
absorptiometry (DEXA). Eur. J. Nucl. Med. 18, 3840.
Lilley, J., Waiters, B. G., Health, D. A. and Drolc, Z. (1991)
In vivo and in vitro precision for bone density measured
by dual energy X-ray absorption. Osteo. Int. 1, 141-146.
Njeh, C. F., Boivin, C. M. and Green, A. (1996)
Inconsistency in osteoporosis risk assessment due to
BMD dependence upon sites measured: A retrospective
study. Nucl. Med. Commun. 17, 281.
WHO Geneva (1994) Assessment of fracture risk and its
application to screening for postmenopausal osteoporosis. Report of WHO Study Group, WHO Technical
Report Series 843, Geneva.
Yu, W., Gluer, C. C., Fuerst, T., Grampp, S., Li, J., Lu, Y.
and Genant, H. K. (1995) Influence of degenerative joint
disease on spinal bone mineral measurements in
postmenopausal women. CalciJ~ Tissue Int. 57, 169-174.