Diagnosing osteoporosis by using dental panoramic radiographs:

The OSTEODENT project

Hugh Devlin, BDS, BSc, MSc, PhD,a Kety Karayianni, DDS, PhD,b Anastasia Mitsea, DDS, MSc,c
Reinhilde Jacobs, LDS, MSc, PhD,d Christina Lindh, DDS, Odont Dr,e Paul van der Stelt, DDS, PhD,f
Elizabeth Marjanovic, BSc, MSc, PhD,g Judith Adams, MBBS, FRCR, FRCP,h Susan Pavitt, BSc, PhD,i
and Keith Horner, BChD, MSc, PhD, FDSRCPS Glasg, FRCR, DDR,j Manchester, England; Athens,
Greece; Leuven, Belgium; Malmö, Sweden; and Amsterdam, The Netherlands
UNIVERSITY OF MANCHESTER, UNIVERSITY OF ATHENS, KATHOLIEKE UNIVERSITEIT LEUVEN, MALMÖ
UNIVERSITY, AND ACADEMIC CENTRE FOR DENTISTRY

Objectives. Measurement of cortical thickness and subjective assessment of cortical porosity on panoramic radiographs are
methods previously reported for diagnosing osteoporosis. The aims of this study were to determine the relative efficacy of
the mandibular cortical index and cortical width in detecting osteoporosis, both alone and in combination, and to
determine the optimal cortical width threshold for referral for additional osteoporosis investigation.
Study design. Six hundred seventy-one postmenopausal women 45 to 70 years of age were recruited for this study.
They received dual energy x-ray absorptiometry (DXA) scans of the left hip and lumbar spine (L1 to L4), and dental
panoramic radiographic examinations of the teeth and jaws. Three observers separately assessed the mandibular
cortical width and porosity in the mental foramen region of the mandible. Cortical width was corrected for
magnification errors. Chi-squared automatic interaction detection analysis (CHAID) software was used (SPSS
AnswerTree, version 3.1, SPSS Inc., Chicago, IL).
Results. Chi-squared automatic interaction detection analysis showed that the cortical porosity was a poorer predictor
of osteoporosis than mandibular cortical width. For the 3 observers, a mandibular cortical width of ⬍3 mm provided
diagnostic odds ratios of 6.51, 6.09, and 8.04. The test is therefore only recommended in triage screening of
individuals by using radiographs made for purposes other than osteoporosis.
Conclusion. When evaluating panoramic radiographs, only those patients with the thinnest mandibular cortices (i.e.,
⬍3 mm) should be referred for further osteoporosis investigation. (Oral Surg Oral Med Oral Pathol Oral Radiol Endod
2007;104:821-8)

Cortical width and porosity on dental panoramic radio-

graphs have been shown to be potentially useful meth-
ods of assessing an individual’s risk of systemic osteo-
porosis.1 These radiographs are primarily taken as an
This work was supported by a research and technological develop- aid to the diagnosis of oral and dental disease but may
ment project grant from the European Commission Fifth Framework
Programme, Quality of Life and Management of Living Resources
also provide information about a patient’s osteoporotic
(QLK6-2002-02243; OSTEODENT). status. The OSTEODENT project was a collaboration
a
Reader, School of Dentistry, University of Manchester. between 5 European centers to determine the best ra-
b
Professor, Dental School, University of Athens. diographic and clinical method of identifying those
c
Postgraduate doctoral student, Dental School, University of Athens. individuals most at risk of osteoporosis.
d
Oral Imaging Centre, School of Dentistry, Oral Pathology and
Maxillofacial Surgery, Katholieke Universiteit Leuven.
A thin mandibular cortical width has been shown to
e
Associate Professor, Faculty of Odontology, Malmo University. be correlated with reduced skeletal bone mineral den-
f
Professor, Academic Centre for Dentistry. sity, but controversy surrounds the issue of what con-
g
Research Associate, Imaging Science and Biomedical Engineering, stitutes a “thin” cortical threshold, as this affects the
University of Manchester. sensitivity and specificity of the diagnostic test. It has
h
Professor, Imaging Science and Biomedical Engineering, University
of Manchester.
been recommended that a cortical width ⱕ4.5 mm
i
Cochrane Senior Research Fellow, School of Dentistry, University of should be used as an indicator of high osteoporosis
Manchester. risk.2 Although choosing a ⱕ4.5 mm threshold will
j
School of Dentistry, University of Manchester. produce a high-sensitivity test, it will produce a large
Received for publication Sep 15, 2006; returned for revision Nov 7, number of false-positives and unnecessary further con-
2006; accepted for publication Dec 22, 2006.
1079-2104/$ - see front matter
firmatory bone mineral density examinations. Using
© 2007 Mosby, Inc. All rights reserved. this threshold in this study, a sensitivity of 89.5% and
doi:10.1016/j.tripleo.2006.12.027 specificity of 33.9% (diagnostic odds ratio ⫽ 4.4) was

821

822 Devlin et al.
achieved for identifying spinal osteoporosis in post-
menopausal women. Diagnostic odds ratio is the ratio
of the positive likelihood ratio and the negative likeli-
hood ratio.
Horner et al.3 found that thinning of the mandibular
cortex below 3 mm at the mental foramen was associ-
ated with low skeletal bone mass (or osteopenia) at the
spine, femoral neck, or forearm. This provided a diag-
nostic test with high specificity but low sensitivity
(specificity, 98.7%; sensitivity, 8%; diagnostic odds
ratio, 6.6). In another study, the 3-mm cortical-width
threshold had specificity of 93.6% and sensitivity of
25.9% (diagnostic odds ratio, 5.1) in detecting osteo-
porosis at the lumbar spine, femoral neck, or forearm.4
The diagnostic odds ratio, a summary statistic of a
diagnostic test’s performance, would indicate that mea-
surement of the radiographic cortical width has a sim-
ilar moderate diagnostic ability across several studies.
Others have found that mandibular cortical width does
not predict spinal osteoporosis, but they used a small
population of postmenopausal women.5
The mandibular cortical index (CI)6 describes the
porosity of the mandible and is related to the mandib-
ular bone mineral density.7 The cortical bone at the
lower border of the mandible on panoramic radio-
graphs, analyzed bilaterally distal to the mental fora-
men, is subjectively classified as follows: (1) CI 1:
the cortical endosteal margin appears even and regular;
(2) CI 2: the endosteal margin appears to have semilu-
nar defects or 1 to 3 layers of cortical endosteal resi-
dues; and (3) CI 3: the cortical layer has numerous (⬎3)
endosteal residues and is clearly porous. The mandib-
ular cortical index is a simple 3-point index with fairly
good reproducibility, the higher CI 3 category indicat-
ing a substantially greater risk of osteoporosis than the
lower CI 1 category.6 Many observers have found this
index to be a useful method of osteoporosis screen-
ing.8-12
The aims of this study were to determine the relative
efficacies of the mandibular cortical index and cortical
width in detecting osteoporosis at either total hip, fem-
oral neck, or lumbar spine; to establish in what circum-
stances a combination of the 2 measurements would
provide an optimum categorization of osteoporosis; and
to determine the optimal cortical width threshold for
osteoporotic diagnosis.
MATERIAL AND METHODS
The study population
The patients invited to take part in the study were 45-
to 70-year-old women. These subjects were either at-
tending for routine dental care or were interested health
care staff who heard about the study through local
publicity. Subjects were recruited from 4 centers in
OOOOE
December 2007
Athens (Greece), Leuven (Belgium), Manchester (En-
gland), and Malmo (Sweden), and ethical approval was
obtained for the research from the relevant committees
in each country. Informed consent was obtained from
the patients in the study after the nature of the proce-
dures had been fully explained. Women in the relevant
age range who had undergone a bone density scan in
the last year and who had been identified as having
osteoporosis were also recruited into the study.
Exclusion criteria were age outside the 45- to 70-year
age range, incomplete bone density measurements, in-
adequate radiographic material, local destructive le-
sions of the mandible, and systemic disease (such as
secondary osteoporosis, poorly controlled thyrotoxico-
sis, primary hyperparathyroidism malabsorption, liver
disease, or alcoholism) that might influence bone min-
eral density.
Each of 3 observers independently made measure-
ments.
Bone mineral density measurements
Dual energy x-ray absorptiometry (DXA) scans of
the left hip and lumbar spine (L1 to L4) were performed
using either a Hologic QDR 4500, Hologic Discovery
(Hologic Inc., Bedford, MA) or a GE Lunar Prodigy
(GE Lunar Corporation, Madison, WI). Shewart’s rules
were used to monitor quality assurance throughout the
study period.13 The European spine phantom was used
to standardize measurements between different manu-
facturers’ DXA machines.14
Subjects with a T score value 2.5 standard deviations
or more below the young female adult mean bone
mineral density value at any one of total hip, femoral
neck, or lumbar spine were classified as osteoporotic,
and all others as normal.
Radiographic measurements
Dental panoramic radiography in Athens and
Manchester was performed on each subject by using a
Planmeca PM2002CC (Planmeca Oy, Helsinki, Fin-
land). In Leuven and Malmo, radiography was carried
out using a Cranex 3DC (Soredex, Tuusula, Finland).
In Leuven, a photostimulable phosphor plate system for
image capture and read was used (ADC Solo, Afga,
Mortsel, Belgium), and the digital images were printed
using a Drystar 2000 dry printing machine (Agfa, Mort-
sel, Belgium). The other centers used a conventional
film/cassette combination. The subjects bit on plastic
bite blocks, enclosing 3.175-mm-diameter ball bear-
ings, during the radiographic exposures. Measurements
were made using the films and prints. All radiographic
cortical width measurements were corrected for mag-
nification errors by using measurements of a ball-bear-
ing reference image.
OOOOE
Volume 104, Number 6
Table I. Mandibular cortical width and porosity
Observer 1 Observer 2
No. of observations cortical width cortical width
Valid 645 653
Missing 8 0
Complete measurements of all radiographs were not completed in some cases.
CI, cortical index.
Three experienced radiologists independently made
measurements on the dental panoramic radiographs of
mandibular cortical width in the region of the mental
foramen and of the mandibular cortical index. The
mandibular cortical width was measured in the mental
foramen region, perpendicular to a tangent to the lower
border of the mandible. Measurements were made us-
ing an eyepiece graticule (Graticules Ltd., Tonbridge,
England) with a ⫻6 magnification lens and recorded to
the nearest 0.1 mm. Measurements from both sides of
the mandible were averaged. The number of patients
for whom cortical width and CI measurements were
made by each observer are shown in Table I. In some
cases, observers were unable to complete measure-
ments of all radiographs.
The 3 expert radiologists independently used the
cortical index (CI) to classify mandibular cortical ero-
sions. With this index, the most severe porosity of the
cortical bone distal to the mental foramen on both sides
of the mandible is recorded. All measurements were
made with the observers blinded to the reference os-
teoporotic diagnosis.
Statistical analysis
Chi-squared automatic interaction detection
(CHAID) analysis software provides information about
which category combination of a predictor variable
yields the highest percentage or the optimum prediction
of a desired outcome (SPSS AnswerTree, version 3.1,
SPSS, Inc.). Cortical widths and CI values were entered
simultaneously into a CHAID analysis. By default,
CHAID converted mandibular cortical width data into
ordinal data with 10 categories, each with a similar
number of cases. The division of any continuous vari-
able into categories produces a range of values that do
not produce a perfectly optimal classification. Chi-
squared automatic interaction detection analysis com-
pared the different categories and merged those cate-
gories that showed no differences on the outcome by
using a likelihood ratio chi-square test. The significance
value for merging categories was P ⫽ .05. Chi-squared
automatic interaction detection partitioned the cortical
width data into categories that show significant differ-
ences as they relate to the presence of osteoporosis at
Devlin et al. 823
Observer 3
cortical width Observer 1 CI Observer 2 CI Observer 3 CI
653 649 652 653
0 4 1 0
any site. The categories were reanalyzed until they
could not be further divided into significantly different
subcategories or would contain less than 50 subjects. A
Bonferroni adjustment was used so that across the
range of tests a 0.05 type 1 error rate was maintained.
After the tree analysis was complete, each end node
was a subset of the study sample, containing cases with
a certain range of the predictor variables (mandibular
cortical width and/or cortical index). The risk of mis-
classification was calculated as the proportion of all
cases that were classified incorrectly by the tree anal-
ysis. The percentage response indicated the percentage
of osteoporotic individuals that were in the particular
category at each node. The percentage index was the
ratio (⫻100) between the percentage of osteoporotic
individuals at that node to the overall percentage of
osteoporotic individuals in the overall sample. An op-
timum threshold would therefore have the highest cu-
mulative value of percentage index and percentage
response. Previous work had recommended a cortical
width of less than 3 mm4 and a cortical index ⬎16 as
the optimum threshold.
The diagnostic odds ratio is a summary statistic of a
diagnostic test’s performance. It was calculated for all
3 observers by using either cortical index or cortical
width to diagnose osteoporosis. A diagnostic odds ratio
is the ratio of positive likelihood ratio (true positive
rate/false-positive rate) and negative likelihood ratio
(false-negative rate/true negative rate).
RESULTS
Six hundred seventy-one consecutive women were
initially recruited to the study, but 8 subjects were
excluded because they were less than 45 years of age.
Two patients were excluded because total hip– bone
mineral density had not been measured, and a further 8
subjects were excluded because the dental panoramic
radiographs were damaged, lost, or had unacceptable
image quality. Of the 653 subjects included in the
study, 141 (21.6%) were classified as having osteopo-
rosis involving at least 1 site. Observers differed
slightly in the number of measurements that were pos-
sible on the radiographs (Table I).
 OOOOE
824 Devlin et al. December 2007

Fig. 1. Chi-squared automatic interaction detection analysis (observer 1 data). Four nodes were formed using cortical width as the
categorical variable. The figures above the row of boxes refer to the category limits of cortical width. The boxes contain the
resulting numbers (and percentage) of normal and osteoporotic individuals falling into that particular category range. The total row
contains the number of individuals (and percentage of entire sample) in that category. Normal indicates no osteoporosis is present
at any site. Osteopor, osteoporosis present at femoral neck, total femur, or lumbar spine.

Table II. Summary statistics using CHAID analysis Table III. Summary statistical data from CHAID anal-
from data collected by observer 1 ysis for observer 2
Cortical width Response Index Cortical width Response Index
Observer 1 range (mm) (%)* (%)† Observer 2 range (mm) (%)* (%)†
Node 1 ⬍2.7 64.1 296.2 Node 1 ⱕ2.9 52.3 242.1
Node 2 2.7-3.1 36.9 170.7 Node 2 2.9-3.9 17.7 82.1
Node 3 3.1-3.7 23.6 109.1 Cortical width
Node 4 ⬎3.7 and missing 9.1 42.3 ⬎3.9 mm
and CI 1 or
CHAID, chi-squared automatic interaction detection.
⬎1
*Indicates the percentage of osteoporotic individuals that are in the
Node 5 11.9 55.2
particular category at each node.
Node 4 1.2 5.4
†The index is the ratio (⫻100) between the percentage of osteopo-
rotic individuals at that node to the overall percentage of osteoporotic Nodes are tabulated in order according to their response percentage.
individuals in the overall sample. An optimum threshold would CHAID, chi-squared automatic interaction detection; CI, cortical
therefore have the highest cumulative value of index percentage and index.
response percentage. *Indicates the percentage of osteoporotic individuals that are in the
particular category at each node.
†The index is the ratio (⫻100) between the percentage of osteopo-
rotic individuals at that node to the overall percentage of osteoporotic
individuals in the overall sample. An optimum threshold would
therefore have the highest cumulative value of index percentage and
Observer 1 data
response percentage.
Using CHAID analysis, the mandibular cortical
width had a stronger, more significant relationship with
osteoporosis than cortical index (Fig. 1). Mandibular
cortical width was partitioned into 4 categories that
gave significantly different proportions of osteoporotic For nodes 1 and 2 (cortical width ⬍3.1 mm), the
individuals from each other (P ⬍ .00001; chi-square ⫽ cumulative response was 50.4% and the cumulative
107.6). The highest prevalence of osteoporosis (64.1%) index was 233.0% for detecting osteoporosis. At the
was found in node 1 (cortical width ⱕ2.7 mm), which higher threshold of ⬍3.7-mm mandibular cortical
was about 3 times higher than that of the whole sample width, the cumulative response was reduced to 34.3%
(21.6%). The cortical index was not a significant vari- and the cumulative index was 158.4%. For verification
able in predicting osteoporosis and was excluded from of the model, the CHAID analysis was repeated on the
the tree model. The estimate of the overall misclassifi- same radiographs for cortical width and CI data col-
cation risk was 0.189 (SE ⫽ 0.015). lected by the other 2 experienced radiologists.
OOOOE
Volume 104, Number 6 Devlin et al. 825

Fig. 2. Chi-squared automatic interaction detection analysis (observer 2 data). The figures above the middle row of boxes (nodes
1, 2, and 3) refer to the category limits of cortical width. The boxes contain the resulting numbers (and percentage) of normal and
osteoporotic individuals falling into that particular category range of cortical width. At node 3 only, mandibular cortical index
provided a significant further subdivision of subjects (CI values either ⬎1 or 1), providing categories (nodes 4 and 5) with a
significantly different proportion of osteoporotic individuals. The total row contains the number of individuals (and percentage of
entire sample) in that category. Normal indicates no osteoporosis is present at any site. Osteopor, osteoporosis present at femoral
neck, total femur, or lumbar spine.

Observer 2 data
Table IV. Summary statistics using CHAID analysis
Cortical width was found to be a more significant
from data collected by observer 3
variable than cortical index and was included first in the
Cortical width Response Index
tree analysis. The first split below the root node is due Observer 3 range (mm) (%)* (%)†
to cortical width because this variable had a stronger
Node 1 ⱕ2.6 74.6 345.6
relationship with osteoporosis than cortical index. At Node 2 2.6-2.9 40.0 185.2
node 1 (threshold cortical width ⬍2.9 mm), the re- Node 3 2.9-3.1 23.4 108.5
sponse was 52.3%, (i.e., about half of the patients were Node 4 3.1-4.4 12.5 57.7
osteoporotic), and the index was 242.1% (Table III). Node 5 ⬎4.4 1.6 7.2
Node 1 contained the group with the greatest likelihood CHAID, chi-squared automatic interaction detection.
of a patient being osteoporotic. The cortical width value *Indicates the percentage of osteoporotic individuals that are in the
⬎3.93 mm was split into 2 nodes based on cortical particular category at each node.
†The index is the ratio (⫻100) between the percentage of osteopo-
index (Fig. 2), but only 11.9% of those individuals with
rotic individuals at that node to the overall percentage of osteoporotic
a cortical width greater than 3.93 mm and a cortical individuals in the overall sample. An optimum threshold would
index greater than 1 were osteoporotic (Table II). This therefore have the highest cumulative value of index percentage and
is much less than the 21.6% for the sample population. response percentage.
 OOOOE
826 Devlin et al. December 2007

Fig. 3. Chi-squared automatic interaction detection analysis (observer 3). Subjects were categorized according to their mandibular
cortical width. The figures above the row of boxes labeled nodes 1 to 5 refer to the category limits of cortical width. The boxes
(nodes 1-5) contain the resulting numbers (and percentage) of normal and osteoporotic individuals falling into that particular
category range. The total row contains the number of individuals (and percentage of entire sample) in that category. Normal
indicates no osteoporosis is present at any site. Osteopor, osteoporosis present at femoral neck, total femur, or lumbar spine.

Table V. Diagnosis of osteoporosis assuming a threshold of ⬍3 mm for cortical width

Observer Cortical width Mandibular cortical index
⫹LR ⫺LR Ratio ⫹LR ⫺LR Ratio
1 4.23 0.65 6.51 1.35 0.36 3.75
2 3.53 0.58 6.09 1.32 0.32 4.13
3 3.86 0.48 8.04 1.35 0.25 5.4
This analysis was separately repeated for mandibular cortical index with the threshold of cortical erosion present (CI ⱖ 1) indicating osteoporosis.
Diagnosis of osteoporosis was made when the T score at total hip, femoral neck, or spine was less than ⫺2.5. Ratio indicates diagnostic odds ratio.
⫹LR, positive likelihood ratio; ⫺LR, negative likelihood ratio; CI, cortical index.

The estimate of the overall misclassification risk was
0.207 (SE ⫽ 0.016).
Observer 3 data
In the tree analysis, only cortical width was found to
be a significant variable, and cortical index was ex-
cluded (Table IV and Fig. 3). In node 1 (subjects with
a cortical width of ⱕ2.6 mm), 74.6% of patients were
osteoporotic. For both nodes 1 and 2 (cortical width
⬍2.9 mm), the cumulative response for detecting os-
teoporosis was 57.6% and the cumulative index was
266.6%. The estimate of the overall misclassification
risk was 0.165 (SE ⫽ 0.014).
Diagnostic odds ratio
Therefore, data from the 3 experienced observers
was partitioned in similar ways with cortical width
thresholds less than about 3 mm, segmenting the data in
an optimal manner. The diagnostic odds ratios were
separately calculated for a threshold of mandibular corti-
cal width (ⱕ3 mm and ⬎3 mm) and a threshold cortical
index of any erosion versus no erosion (Table V). The
diagnostic odds ratios were greater for cortical width
(median, 6.51) than cortical erosion (median, 4.13).
DISCUSSION
Chi-squared automatic interaction detection analysis
is used widely in market research to identify those
groups of individuals that would most likely respond to
advertising campaigns, mail shots, or promotions for a
particular product or service. Those people who would
be most likely to want the product in question are
contacted, and costs incurred by inefficiently contacting
nonresponders are minimized. In the medical field,
CHAID has been used to determine those variables that
best predict survival after major traumatic injury, and
then those variables are segmented to establish their
significantly different outcome predictability.15 In den-
tistry, this technique has been used infrequently but has
great potential for audit projects such as identifying
OOOOE
Volume 104, Number 6
those categories of patients most likely to fail to keep
appointment dates16 or who would be expected to use
independent dental services.17
In the present study, CHAID analysis showed that
the cortical index was a poorer predictor of osteoporo-
sis than mandibular cortical width. The sensitivity of
any diagnostic test is dependent on the threshold of the
test above which a diagnosis of disease is made, but the
cortical index does not provide an accurate method of
osteoporosis diagnosis. Halling et al.18 found that in
their highest cortical index category (CI 3), the sensi-
tivity in detecting a reduced bone mineral density was
only 50%. In a study by Sutthiprapaporn et al.,19 when
the criterion for a positive finding of reduced bone
mineral density was the presence of any cortical erosion
(CI ⬎ 1), then sensitivity was improved to 73.0% at the
expense of a reduced specificity of 49.0%. However,
comparisons are difficult between these studies, be-
cause the criterion for low bone density differed be-
tween them.
In using the cortical index to diagnose osteoporosis,
Klemetti et al.6 found the sensitivity and specificity
were low. The presence of any cortical erosion gave a
sensitivity of 71% and specificity of 40% (diagnostic
odds ratio, 1.63); skeletal osteoporosis was diagnosed
in a subject when either value of the bone mineral
density from their femoral neck and lumbar spine were
in the lowest quintile of their population (n ⫽ 355).
(Since 1994, most authors use the reference standard of
the World Health Organization for diagnosing osteopo-
rosis, i.e., dual energy x-ray absorptiometry T score
measurement less than ⫺2.5).20 Combining cortical
index (CI ⫽ 1 vs. CI ⬎ 1) and cortical width (CW ⬍ 4
mm vs. ⬎ 4 mm), so that a patient must fail both
thresholds to be classed as osteoporotic, increased the
specificity to 99%, but reduced the sensitivity to 10%.6
The sensitivity of a diagnostic test is not constant
across populations with differing severity of disease. In
many published studies, patients were recruited follow-
ing referral for bone mineral density assessment, and it
is possible that the spectrum of severity of osteoporosis
may not relate to that in primary care. Taguchi et al.2
found that for normal postmenopausal women (n ⫽ 159),
with any cortical erosion (CI ⬎ 1), the sensitivity ⫽
86.8% and specificity ⫽ 63.6% for diagnosing spinal
osteoporosis (diagnostic odds ratio, 11.49). These high-
sensitivity values may not be possible in a general
practice setting where the severity of disease may be
expected to be less. Taguchi et al.21 found that the
positive likelihood ratio for identifying women with
osteoporosis was 6.40 for a thin cortical width and 7.11
for a severely eroded cortex. If the study population
now consists of patients who are mildly affected by
osteoporosis, positive likelihood ratios will move to-
Devlin et al. 827
ward unity. Similarly, the positive predictive values for
mandibular cortical erosion will also depend on the
spectrum of osteoporotic disease in the study sample. In
the present study, the population consisted of a mixture
of healthy volunteers and patients recruited from an
osteoporotic clinic.
As well as the pain and suffering to the individual
concerned, the treatment of hip fractures is a heavy
burden to any health care system, with an average cost
per hip fracture of £13,000 in the first year and £7,000
for the subsequent year in the UK.22 Diagnosing pa-
tients at the greatest risk of osteoporosis is an effective
identification strategy that avoids the unnecessary costs
of confirmatory axial DXA scanning on healthy indi-
viduals. White et al.10 used a classification and regres-
sion tree analysis to analyze the clinical and radio-
graphic features for identifying osteoporosis and found
that the thickness of the mandibular cortex was one of
the most clinically useful risk factors, which is in
agreement with this study.
But these radiographic indices do not provide suffi-
cient diagnostic evidence by themselves for definitive
diagnosis and are only useful in triage screening of
individuals by using radiographs taken for other pur-
poses. Determining which radiographic index is the
best discriminator for osteoporosis involves analyzing
sensitivity and specificity, which are interdependent
and trade off with one another. The diagnostic odds
ratio measures the performance of a test and is the ratio
of the odds of a positive test result in a patient with
disease compared with one without disease (or positive
likelihood ratio divided by negative likelihood ratio).
Using cortical width (⬍4.3 mm) for diagnosing os-
teopenia and osteoporosis at either the lumbar spine or
femoral neck, Taguchi et al.21 found a positive likeli-
hood ratio of 1.65 and a negative likelihood ratio of
0.22, giving a diagnostic ratio derived from this data of
7.5, which is similar to the data presented in our study.
But diagnostic ratio values need to be above 20 to
indicate strong diagnostic evidence for a test.23 In our
study, both cortical width and cortical index measure-
ments from all 3 experts had positive likelihood ratios
of ⬍5. This would indicate that the tests had a fairly
small influence on diagnosis because there was limited
change in the pretest to posttest probability of osteopo-
rosis,24 but the use of radiographic indices can be
justified if the radiographs are taken for other purposes
than osteoporosis diagnosis.
Some authors have not found any relationship be-
tween axial or femoral osteoporosis and mandibular
bone quality; for example, the evidence linking osteo-
porosis and implant failure in the jaws is poor.25 De-
spite this, endosteal cortical thinning has been observed
in the tibia26 and mandible27 and may be a generalized

828 Devlin et al.
age- and sex-related phenomenon. We have described a
statistical relationship between mandibular cortical
thickness and osteoporotic status, but the physiological
linking mechanism may involve a complex interaction
between local factors, estrogen deficiency, the individ-
ual’s peak bone mass, and many other factors.
In conclusion, the study demonstrated that mandib-
ular cortical width has better efficacy than the mandib-
ular cortical index in detecting osteoporosis. There was
no evidence for any benefit associated with combining
the 2 measurements to detect osteoporosis. Only those
with the thinnest mandibular cortices (ⱕ3 mm) should
be referred for further osteoporosis investigation, be-
cause it is this group that has the highest likelihood of
osteoporosis.
REFERENCES
1. Horner K, Devlin H, Alsop CW, Hodgkinson IM, Adams JE.
Mandibular bone mineral density as a predictor of skeletal os-
teoporosis. Brit J of Radiol 1996;69:1019-25.
2. Taguchi A, Suei Y, Sanada M, Ohtsuka M, Nakamoto T, Sumida
H, et al. Validation of dental panoramic radiography measures
for identifying postmenopausal women with spinal osteoporosis.
Am J Roentgenol 2004;183:1755-60.
3. Horner K, Devlin H, Harvey L. Detecting patients with low
skeletal bone mass. J Dent 2002;30:171-5.
4. Devlin H, Horner K. Mandibular radiomorphometric indices in
the diagnosis of reduced skeletal bone mineral density. Osteo-
poros Int 2002;13:373-78.
5. Yasar F, Akgunlu F. The differences in panoramic mandibular
indices and fractal dimension between patients with and without
spinal osteoporosis. Dentomaxillofac Radiol 2006;35:1-9.
6. Klemetti E, Kolmakov S, Kroger H. Pantomography in assess-
ment of the osteoporosis risk group. Scand J Dent Res 1994;
102:68-72.
7. Horner K, Devlin H. The relationships between two indices of
mandibular bone quality and bone mineral density measured by
dual energy X-ray absorptiometry. Dentomaxillofac Radiol
1998;27:17-21.
8. Klemetti E, Kolmakow S. Morphology of the mandibular cortex
on panoramic radiographs as an indicator of bone quality. Den-
tomaxillofac Radiol 1997;26:22-5.
9. Bollen AM, Taguchi A, Hujoel PP, Hollender LG. Case-control
study on self-reported osteoporotic fractures and mandibular
cortical bone. Oral Surg Oral Med Oral Pathol Oral Radiol
Endod 2000;90:518-24.
10. White SC, Taguchi A, Kao D, Wu S, Service SK, Yoon D, et al.
Clinical and panoramic predictors of femur bone mineral density.
Osteoporos Int 2005;16:339-46.
11. Nakamoto T, Taguchi A, Ohtsuka M, Suei Y, Fujita M, Tani-
moto K, et al. Dental panoramic radiograph as a tool to detect
postmenopausal women with low bone mineral density: un-
trained general dental practitioners’ diagnostic performance. Os-
teoporos Int 2003;14:659-64.
12. Taguchi A, Suei Y, Ohtsuka M, Otani K, Tanimoto K, Ohtaki M.
Usefulness of panoramic radiography in the diagnosis of post-
menopausal osteoporosis in women. Width and morphology of
inferior cortex of the mandible. Dentomaxillofac Radiol
1996;25:263-7.
13. Orwoll ES, Oviatt SK. Longitudinal precision of dual-energy
All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and read them immediately.
OOOOE
December 2007
x-ray absorptiometry in a multicenter study. The Nafarelin/Bone
Study Group. J Bone Miner Res 1991;6:191-7.
14. Pearson J, Dequeker J, Henley M, Bright J, Reeve J, Kalender W,
et al. European semi-anthropomorphic spine phantom for the
calibration of bone densitometers: assessment of precision, sta-
bility and accuracy. The European Quantitation of Osteoporosis
Study Group. Osteoporos Int 1995;5:174-84.
15. Hill DA, Delaney LM, Roncal S. A chi-square automatic inter-
action detection (CHAID) analysis of factors determining trauma
outcomes. J Trauma 1997;42:62-6.
16. Moles DR, Bedi R. A simple technique for data management in
general dental practice audit. Prim Dent Care 1997;4:61-5.
17. McGrath C, Moles D, Bedi R. Who uses independent dental
services? Findings from a national survey. Prim Dent Care
1999;6:157-60.
18. Halling A, Persson GR, Berglund J, Johansson O, Renvert S.
Comparison between the Klemetti index and heel DXA BMD
measurements in the diagnosis of reduced skeletal bone mineral
density in the elderly. Osteoporos Int 2005;16:999-1003.
19. Sutthiprapaporn P, Taguchi A, Nakamoto T, Ohtsuka M, Mallick
PC, Tsuda M, et al. Diagnostic performance of general dental
practitioners after lecture in identifying post-menopausal women
with low bone mineral density by panoramic radiographs. Den-
tomaxillofac Radiol 2006;35:249-52.
20. Kanis JA, WHO Study Group. Assessment of fracture risk and its
application to screening for postmenopausal osteoporosis: syn-
opsis of a WHO report. Osteoporos Int 1994;4:368-81.
21. Taguchi A, Tsuda M, Ohtsuka M, Kodama I, Sanada M, Naka-
moto T, et al. Use of dental panoramic radiographs in identifying
younger postmenopausal women with osteoporosis. Osteoporos
Int 2006;17:387-94.
22. Torgerson D, Iglesias C, Reid DM. The effective management of
osteoporosis. In: Barlow DH, ed. The economics of fracture
prevention. London: Aesculapius Medical Press; 2001. p.
111-21.
23. Deeks JJ. Systematic reviews of evaluations of diagnostic and
screening tests. In: Egger M, Smith GD, Altman DG, editors.
Systematic reviews in health care: meta-analysis in context. 2nd
ed. London: BMJ Publishing Group; 2001. p. 255-6.
24. Jaeschke R, Guyatt GH, Sackett DL. Users’ guides to the medical
literature. VI. How to use an article about a diagnostic test. B:
what are the results and will they help me in caring for my
patients? The Evidence-Based Medicine Working Group. JAMA
1994;271:703-7.
25. Mombelli A, Cionca N. Systemic diseases affecting osseointegra-
tion therapy. Clin Oral Implants Res 2006;17(Suppl 2):97-103.
26. Amorim MA, Takayama L, Jorgetti V, Pereira RM. Comparative
study of axial and femoral bone mineral density and parameters
of mandibular bone quality in patients receiving dental implants.
Osteoporos Int 2006;17:1494-500.
27. Bollen AM, Taguchi A, Hujoel PP, Hollender LG. Case-control
study on self-reported osteoporotic fractures and mandibular
cortical bone. Oral Surg Oral Med Oral Pathol Radiol Endod
2000;90:518-24.
Reprint requests:
Hugh Devlin, PhD, MSc, BSc, BDS
School of Dentistry
University of Manchester
Higher Cambridge Street
Manchester, M15 6FH, United Kingdom
hugh.devlin@manchester.ac.uk