Heel bone densitometry: device specific thresholds for the
assessment of osteoporosis
1E MCCAULEY, MSc,1A MACKIE, PhD,3D ELLIOTT, RGN, ONC and2A CHUCK, FRCP
1Regional Medical Physics Department and2Department of Rheumatology, University Hospital of
North Durham, North Road, Durham DH1 5TW and3Durham and Chester-le-Street Primary Care
Trust, John Snow House, Durham University Science Park, County Durham DH1 3YG, UK
ABSTRACT. A diagnosis of osteoporosis is facilitated by bone mineral density (BMD)
measurement of the lumbar spine and hip using dual energy X-ray absorptiometry
(DXA), interpreted in accordance with criteria published by the World Health
Organization (WHO). The use of peripheral DXA is growing in primary care and
guidance on its use has recently been published by the National Osteoporosis Society
(NOS), recommending a triage approach using thresholds specific to each type of
peripheral device. However, no data currently exist for the Norland Apollo heel
densitometer (Cooper Surgical, Trumbull, USA). 215 women between 50 years and
75 years of age (mean age 64.6 years) referred for hip and spine BMD measurements
also had a heel BMD measurement. Device specific upper and lower thresholds were
calculated for the Norland Apollo heel densitometer to give a 90% sensitivity and 90%
specificity for osteoporosis at the hip or spine. Patients with a heel T-score of above –1.2
are very likely to have normal bone density on axial densitometry, whilst patients with
heel T-score of below –2.2 are very likely to have osteoporosis at the hip or spine.
Patients whose measurements lie between the thresholds should be referred for axial
Received 30 August 2005
Revised 14 October 2005
Accepted 31 October 2005
’ 2006 The British Institute of
A diagnosis of osteoporosis or osteopenia can be made
using dual energy X-ray absorptiometry (DXA) of the
lumbar spine or femur. The defining criterion for the
diagnosis is the value of the T-score, which is the number
of standard deviations the measured bone mineral
density (BMD) is above or below the young adult
Organization (WHO) criteria , the patient is osteo-
porotic if the T-score # –2.5 and osteopenic if the T-score
.–2.5 but # –1.
Axial densitometers, which are designed to measure
BMD at the hip and/or spine, are large, expensive and
require a permanent location, ordinarily in a secondary
care setting. Significant research into the development of
alternative methodologies has resulted in a plethora of
techniques and alternative measurement sites . The
measurement of peripheral bone density is growing
outside secondary care centres, principally because of the
much lower cost of these devices compared with axial
densitometers. They are also very easy to use, requiring
minimal set up and positioning, and are relatively
Heel densitometers produce results in terms of BMD
(g cm22) and a heel T-score. Manufacturers’ data
accompanying devices often quote the WHO criteria
indicating that a heel T-score of # 22.5 would be
indicative of osteoporosis. It is well established, how-
ever, that it is inappropriate to interpret any peripheral
T-score as one would for axial densitometry [3, 4] and
WHO T-score criteria applied to heel DXA would
significantly underestimate the prevalence of osteopo-
rosis [3, 5].
In addition, it is well known that measurements of
BMD vary from scanner to scanner and between scanner
types; particularly between manufacturers because of the
variation in acquisition and analysis techniques as well
as the absence of a standard calibration technique [6–10].
The International Society for Clinical Densitometry
(ISCD) state that although peripheral DXA measure-
ments can theoretically be used to identify patients
unlikely to have osteoporosis, and also identify patients
who should be treated, this cannot be applied in clinical
practice until device-specific cut-off points are estab-
The National Osteoporosis Society (NOS) has recently
produced a revised position statement concerning the
recommended use of peripheral densitometry . NOS
recommend a triage approach using T-score thresholds,
which are specific to each type of peripheral device.
These thresholds are defined so that patients with
osteoporosis at the hip or spine are identified with 90%
sensitivity and 90% specificity. Patients with a peripheral
T-score below the lower threshold are very likely to have
osteoporosis at the hip or spine, whereas patients with a
peripheral T-score above the upper threshold are
unlikely to have osteoporosis. Those patients who have
a heel T-score that falls between the two thresholds
require a hip and spine BMD examination for a definitive
The NOS statement gives device specific threshold
data for five peripheral devices, two of which are heel
densitometers. Data in this form are not yet published for
the Norland Apollo heel densitometer (Cooper Surgical,
The British Journal of Radiology, 79 (2006), 464–467
464The British Journal of Radiology, June 2006
In 2001, the Durham and Chester-le-Street Primary
Care Trust (PCT) purchased a Norland Apollo DXA
scanner for the measurement of heel bone density within
the primary care setting. The device was to be operated
by the osteoporosis prevention specialist nurse within a
protocol agreed between primary and secondary care.
The aim of this study was to establish thresholds for
Apollo heel BMD T-scores for use in risk stratifying
patients within a primary care management protocol.
Ethical committee approval was obtained to measure
heel BMD with the Apollo in women attending for axial
bone density measurements; all fulfilled at least one
accepted referral criterion . 76 women aged between
50 years and 75 years (mean age 64.0 years) gave
informed written consent. Axial BMD was measured
for the lumbar spine, L2–4, and for the non-dominant
femur using a Lunar DPX-IQ (General Electric/Lunar
Corp., Madison, WI) densitometer. Heel BMD, of the
dominant side, was measured with a Norland Apollo
portable densitometer. Measurement of the dominant
side was acceptable since published data demonstrates
no significant difference between BMD measurements of
the heel between the right and left sides [5, 14].
Patients were then categorised as either not osteo-
porotic (T-score . –2.5) or osteoporotic (T-score # –2.5),
dependent upon their lowest axial T-score, i.e. total
femur or spine (L2–4). 23 of these patients were found to
have normal axial T-scores, 34 were osteopenic and 19
patients were osteoporotic. Threshold values were
initially calculated using this preliminary data so that
the Apollo DXA scanner could be used in clinical
practice as quickly as possible, although insufficient
patients had been examined for the data to achieve the
statistical accuracy recommended by NOS, i.e. 95%
confidence that the true sensitivity and specificity do
not fall below 80%. The initial lower threshold was a
T-score of –2.2. The initial upper threshold was 0.1,
although this value was obviously very inaccurate since
there was a large difference in the T-scores of the patients
at the cut off point and those patients immediately above
and below them, as is demonstrated in Figure 1.
Patients were seen in a primary care setting by an
osteoporosis prevention specialist nurse, who performed
heel DXA measurements on patients considered at risk
and referred those patients for axial densitometry with
heel T-scores below the upper threshold value of 0.1.
Heel and axial T-scores were recorded for every patient
referred in this way. Data for a further 139 patients
between the ages of 50 years and 75 years were recorded,
mean age 64.9 years.
In total, heel and axial DXA T-scores were measured
for 215 women, mean age 64.6 years. Of these, 71 patients
were found to be osteoporotic on axial densitometry and
144 patients were not osteoporotic.
Figure 2 shows a scatter plot of heel T-scores for
patients in the two diagnostic categories. The line
Figure 1. Scatter plot of heel T-scores for the initial group of
patients; 90% of the 57 not osteoporotic patients lie above
the lower threshold of 22.2, whilst 90% of the 19 patients
who are osteoporotic lie below the higher threshold of 0.1.
The limited data points in the osteoporotic group do not
allow an accurate cut off value to be selected.
Figure 2. Scatter plot of all heel T-scores for patients in the
two diagnostic categories; 90% of patients who have a T-
score above the lower threshold of 22.2 do not have
osteoporosis, whilst 90% of patients who have a T-score
below the upper T-score threshold of 21.2 are osteoporotic.
Patients with heel T-score between the two threshold values
should be referred for axial densitometry.
Heel bone densitometry
The British Journal of Radiology, June 2006465
through the not osteoporotic group shows the lower
threshold, a T-score of 22.2; 90% of patients who have a
T-score above this threshold do not have osteoporosis.
The line through the osteoporotic group shows the upper
threshold, a T-score of 21.2; 90% of patients who have a
T-score below this threshold are osteoporotic.
The information provided by the manufacturers of the
Norland Apollo heel densitometer states that a heel T-
score of # 22.5 is indicative of osteoporosis. This
assumption that the WHO criterion for axial DXA can
be applied to heel DXA has been previously demon-
strated to be incorrect. Only 15 of the 71 patients (21.1%)
who were found to be osteoporotic on axial densitometry
had heel T-scores of # 22.5.
Data were initially acquired so that a working thresh-
old value could be quickly calculated and used within
the primary care setting. These original 76 patients had
their hip, spine and heel DXA measurements all
performed on the same day. Subsequently, patients were
then referred for axial DXA on the basis of established
referral criteria or on the value of their heel DXA
measurement. These latter results will therefore intro-
duce some bias in the data, as patients who were found
to have high T-scores at the heel, above the higher
threshold value, were not referred for axial DXA, despite
the presence of other risk factors. These patients, with a
very low risk of axial osteoporosis, were therefore
excluded from the study as it was the intention of the
initial threshold values to avoid unnecessary referrals for
axial DXA. Since patients with high heel T-scores were
excluded, this bias will have the effect of lowering the
average T-score measured in the latter group of patients.
These patients represent 65% of the total patient group.
This measurement bias prevents these data from being
incorporated into the NOS published data for device
specific thresholds. Also, NOS consider only women
between the ages of 55 years and 70 years, whereas these
data have been acquired for patients between 50 years
and 75 years of age. However, the acquisition of truly
unbiased data within a more narrow age range and with
enough patients for statistical accuracy, i.e. 95% con-
fidence that the true sensitivity and specificity do not fall
below 80%, would take more than 10 months with this
department’s current workload. This estimation assumes
that only half of all patients referred are within the
required age range (based on 2004 data) and less than
one third of patients in this age group are likely to be
These biased data therefore, may be considered an
approximation of the true thresholds for the Norland
Apollo heel densitometer. The true values are likely be
slightly higher, due to the exclusion of patients with high
T-scores, therefore caution is advised with heel T-scores
just above the upper threshold of 21.2. These patients
should be considered to have an equivocal result and be
referred for axial DXA to prevent misdiagnosis. Patients
with heel T-scores just above the lower threshold (22.2)
will be referred for axial densitometry rather than be
given a diagnosis of osteoporosis from the heel DXA
alone. In this case, no patients should be misdiagnosed,
but the number of referrals to axial DXA will therefore be
higher than if the threshold values were unbiased.
The lower threshold of 22.2 was unchanged from that
calculated using the initial group of 57 patients. The
higher threshold was significantly lowered to 21.2 when
a total of 70 osteoporotic patients were considered,
compared with the initial value of 0.1 which was
calculated from only 19 patients. The difference in these
figures is not surprising, and is simply due to the
variability of T-scores within small numbers of patients.
Thresholds have been calculated for a Norland Apollo
heel densitometer; patients with a heel T-score of above
21.2 are very likely to have normal bone density on axial
densitometry, whilst patients with heel T-score of below
22.2 are very likely to have osteoporosis at the hip or
spine. Patients with heel T-scores that lie between these
two thresholds should be referred for axial densitometry.
Users of a Norland Apollo heel densitometer may wish
to implement these threshold values, with the caveat that
they are an approximation but are the best data currently
available for this scanner.
1. World Health Organization. Assessment of fracture risk
and its application to screening for postmenopausal
osteoporosis: Report of WHO Study Group. Technical
report series 843. Geneva, Switzerland, WHO 1994.
2. Placide J, Martens MG. Comparing screening methods for
osteoporosis. Curr Womens Health Rep 2003;3:207–10.
3. Fordham JN, Chinn DJ, Kumar N. Identification of women
with reduced bone density at the lumbar spine and femoral
neck usingBMD at the
4. Miller PD, Njeh CF, Jankowski LG, Lenchik L. International
Society for Clinical Densitometry Position Development
Panel and Scientific Advisory Committee. What are the
standards by which bone mass measurement at peripheral
skeletal sites should be used in the diagnosis of osteoporo-
sis? J Clin Densitom 2002;5Suppl.:S39–45.
5. Michalska D, Zikan V, Stepan J, Weichetova M, Kubova V,
Krenkova J, et al. X-ray densitometry and ultrasonography
of the heel bone – sensitivity and comparison with
6. Genant HK, Grampp S, Glueer CC, Faulkner KG, Jergas M,
Engelke K, et al. Universal standardisation for dual x-ray
absorptiometry: patient and phantom cross-calibration
results. J Bone Miner Res 1994;9:1503–14.
7. Pearson J, Dequeker J, Reeve J, Felsenberg D, Henley M,
Bright J, et al. Dual x-ray absorptiometry of the proximal
femur: normal European values standardized with the
European spine phantom. J Bone Miner Res 1995;10:315–24.
8. Formica CA. Standardization of BMD measurements
(editorial). Osteoporos Int 1998;8:1–3.
9. Grigorian M, Shepherd JA, Cheng XG, Njeh CF, Toschke JO,
Genant HK. Does osteoporosis classification using heel BMD
agree across manufacturers? Osteoporos Int 2002;13:613–7.
10. Faulkner KG, Roberts LA, McClung MR. Discrepancies in
normative data between Lunar and Hologic DXA systems.
Osteoporosis Int 1996;6:432–6.
11. Leib ES, Lewiecki EM, Binkley N, Hamdy RC. Official
Densitometry. J Clin Densitom 2004;7:1–6.
E McCauley, A Mackie, D Elliott and A Chuck
466The British Journal of Radiology, June 2006
12. Blake G, Chinn D, Steel S, Patel R, Panayiotou E, Thorpe J, Download full-text
et al. The Revised NOS Position Statement on peripheral
X-ray Absorptiometry: A listing of device specific T-score
thresholds for the clinical interpretation of pDXA examina-
tions. NOS, 2005.
13. Royal College of Physicians. Osteoporosis: Clinical guide-
lines for prevention and treatment. London: RCP, 1999;7:66.
14. Williams ED, Daymond TJ. Evaluation of calcaneus bone
densitometry against hip and spine for diagnosis of
osteoporosis. Br J Radiol 2003;76:123–8.
Heel bone densitometry
The British Journal of Radiology, June 2006467