OSTEOPOROSIS III

Osteoporosis III

Diagnosis of osteoporosis and assessment of fracture risk

John A Kanis

The diagnosis of osteoporosis centres on the assessment of bone mineral density (BMD). Osteoporosis is defined as a
BMD 2·5 SD or more below the average value for premenopausal women (T score <–2·5 SD). Severe osteoporosis
denotes osteoporosis in the presence of one or more fragility fractures. The same absolute value for BMD used in
women can be used in men. The recommended site for diagnosis is the proximal femur with dual energy X-ray
absorptiometry (DXA). Other sites and validated techniques, however, can be used for fracture prediction. Although
hip fracture prediction with BMD alone is at least as good as blood pressure readings to predict stroke, the predictive
value of BMD can be enhanced by use of other factors, such as biochemical indices of bone resorption and clinical risk
factors. Clinical risk factors that contribute to fracture risk independently of BMD include age, previous fragility
fracture, premature menopause, a family history of hip fracture, and the use of oral corticosteroids. In the absence of
validated population screening strategies, a case finding strategy is recommended based on the finding of risk factors.
Treatment should be considered in individuals subsequently shown to have a high fracture risk. Because of the many
techniques available for fracture risk assessment, the 10-year probability of fracture is the desirable measurement to
determine intervention thresholds. Many treatments can be provided cost-effectively to men and women if hip fracture
probability over 10 years ranges from 2% to 10% dependent on age.

Introduction and assessment of risk, which in turn implies a distinction

As prevalence and awareness of osteoporosis increases, between diagnostic and intervention thresholds.
and treatments of proven efficacy become available, the
demand for management of patients with the disease will Measurement of bone mineral content
also rise. Such demand will, in turn, require widespread Single and dual X-ray absorptiometry
development of facilities for the diagnosis and assessment Single and dual X-ray absorptiometry (DXA) are used
of osteoporosis. Measurement of bone mineral density to assess mineral content of the entire skeleton and that
(BMD) is the central component of any provision that of specific sites, including those most vulnerable to
arises from the internationally agreed definition of fracture.2 Bone mineral content is the amount of mineral
osteoporosis: a systemic skeletal disease characterised by in the specific site scanned and, when divided by the
low bone mass and microarchitectural deterioration of area measured, can be used to derive a value for BMD.
bone tissue, with a consequent increase in bone fragility Both techniques provide a two-dimensional, areal
and susceptibility to fracture.1 The diagnosis of picture, rather than a true volumetric density. Thus, the
osteoporosis thus centres on assessment of bone mass and size of the bone affects the apparent density, since the
quality. There are no satisfactory clinical means to assess relation between area and volume is non-linear.
bone quality. Diagnosis of osteoporosis, therefore, Paradoxically, this error can improve the value of BMD
depends on the measurement of skeletal mass. as a predictor of fracture risk, since bone size is also a
The clinical significance of osteoporosis rests with the determinant of skeletal strength.
fractures that arise as a consequence of the condition, and
their attendant morbidity and mortality. Low bone mass is Panel 1: Sources of error in the diagnosis of
an important component of the risk of fracture, but other osteoporosis by dual X-ray absortiometry4
abnormalities arise in the skeleton that contribute to
skeletal fragility. Furthermore, various non-skeletal Incorrect diagnosis of osteoporosis caused by:
factors, such as the liability to fall, contribute to fracture Osteomalacia
risk. Thus, ideally, assessment of fracture risk should Osteoarthritis (of spine but also of the hip)
encompass all these aspects. There is, therefore, a Soft tissue calcification (especially aortic calcification for spine
distinction to be made between diagnosis of osteoporosis measurements)
Overlying metal objects
Contrast media
Search strategy Previous fracture (spine, hip, and wrist)
This article is based on review of international publications Severe scoliosis
collected by the author during his time working in the specialty. Extreme obesity or ascites
Vertebral deformities due to osteoarthritis, Scheuermann’s
disease
Lancet 2002; 359: 1929–36 Inadequate reference ranges
Inadequate operating procedures—eg, calibration, region
WHO Collaborating Centre for Metabolic Bone Diseases, selection, acquisition mode, positioning)
University of Sheffield Medical School, Sheffield S10 2RX, UK Reproduced from reference 4 by permission of Osteoporosis
(Prof J A Kanis MD) International.
(e-mail: [email protected])

THE LANCET • Vol 359 • June 1, 2002 • www.thelancet.com 1929

For personal use. Only reproduce with permission from The Lancet Publishing Group.
OSTEOPOROSIS III

The accuracy of DXA at the hip exceeds 90%;3 several applications of densitometry—eg, diagnosis of
residual errors arise for various reasons, related to the osteoporosis, assessment of prognosis (fracture
technique itself and the manner in which it is applied. prediction), monitoring of the natural history of the
Panel 1 shows some of the causes of diagnostic errors of disorder, and assessment of response to treatment.
clinical importance.4
BMD is an index of bone mass only when bone is fully Diagnosis of osteoporosis
mineralised. The presence of osteomalacia, often a The easiest way to diagnose osteoporosis by bone density
complication caused by poor nutrition in elderly people, measurements is to define a threshold—namely, a cutoff for
will, therefore, result in underestimation of bone mass. BMD, that encompasses most patients with osteoporotic
Osteoarthritis at the spine or the hip is common in the fractures. Bone density measurements are, however, also
elderly, and contributes to the density measurement but used to assess future risk of fracture, so that more than one
not necessarily to skeletal strength. Heterogeneity of cutoff is needed.
density because of osteoarthritis or previous fracture can
often be detected on the scan, and in some instances can Thresholds
be excluded from the analysis. In the case of the hip, other Skeletal mass and density remain fairly constant, once
regions of interest such as the femoral neck can be selected growth has stopped, until about age 50 years.13 The
to exclude the joint. distribution or density of bone mineral content in young
healthy adults (peak bone mass) is appoximately Gaussian
Ultrasonic measurement of bone normal irrespective of the technique used. Because of the
Skeletal status in osteoporosis can be measured with Gaussian distribution, bone density values in individuals
quantitative ultrasound methods. The most widely can be expressed in relation to a reference population in
assessed methods are broad-band ultrasound attenuation standard deviation (SD) units. This ability reduces the
and speed of sound (or ultrasound velocity) at the heel. difficulties associated with differences in calibration
Because these techniques do not involve ionising radiation between instruments. When SDs are used in relation to the
and could provide some information with respect to the young healthy population, this measurement is referred to
structural organisation of bone in addition to bone mass, as the T score.
there is much interest in their use. For reasons outlined For women, four general diagnostic categories have
below, these techniques cannot be used to diagnose been proposed by WHO and modified by the International
osteoporosis, but evidence5–8 lends support to their use for Osteoporosis Foundation, for assessments done with
the assessment of fracture risk in elderly women. DXA:3,12

Normal—hip BMD greater than 1 SD below the young

Computed tomography adult female reference mean (T score ⭓–1).

Quantitative computed tomography has been applied both
Low bone mass (osteopenia)—hip BMD greater than

to the appendicular skeleton and to the spine.9–11 1 SD below the young adult female mean, but less than
Conventional whole body computed tomography scanners 2·5 SD below this value (T score <–1 and >–2·5).
need calibration to convert their results into units relevant
Osteoporosis—hip BMD 2·5 SD or more below the

to BMD. Quantitative computed tomography is most young adult female mean (T score ⭐–2·5).
useful in the assessment of cancellous bone density
Severe osteoporosis (established osteoporosis)—hip
because it provides a measure of true volumetric density, BMD 2·5 SD or more below the young adult mean in
rather than an area-adjusted result (as is the case with the presence of one or more fragility fractures.
DXA). Cancellous bone is more responsive than cortical In women, bone loss occurs predominantly after the
bone to many interventions. Computed tomography can, menopause. In the young healthy population, 15% of
therefore, be used to monitor the effect of treatment.2 women have a T score of less than –1 and thus have low
Additionally, the technique avoids the effect of bone mass or osteopenia (figure 1).14,15 Because of the
degenerative disease, a particular drawback to DXA at the normal distribution for BMD, about 0·5% of women fall
spine. The main disadvantages of computed tomography into the osteoporotic range, with a T score of –2·5 or less.14
are high exposure to radiation, difficulties with quality Furthermore, the proportion of women affected by
control, and high cost compared with DXA.

Radiography Proportion of population (%)

Osteoporosis can often be diagnosed by looking at simple 0·6 15 50 85 >99
radiographs, albeit with low sensitivity. Furthermore, there Osteoporosis Osteopenia Normal
are several characteristic features of osteoporosis that can
be seen with this technique, which help in diagnosis or in
differential diagnosis. Subclinical vertebral fracture is a
strong risk factor for subsequent fractures, for example,
both at new vertebral sites and at other sites susceptible to
osteoporosis. There is, therefore, great interest in the
identification of vertebral deformities due to osteoporosis
that might not have otherwise come to clinical attention.
Of the many techniques that have been developed to
assess bone mass, bone mineral content, or other related
aspects of skeletal mass or structure, the technique that has
been paid the greatest amount of attention in terms of
technical development and biological validation is DXA, –4 –3 –2 –1 0 1 2 3 4
which is regarded as the gold standard for diagnosis.12 The Bone mineral density (T score)
adoption of DXA as a reference standard provides a
technique against which the performance characteristics of Figure 1: Distribution of bone mineral density in healthy
less well established methods can be compared for the women aged 30–40 years

1930 THE LANCET • Vol 359 • June 1, 2002 • www.thelancet.com

For personal use. Only reproduce with permission from The Lancet Publishing Group.
 OSTEOPOROSIS III

intersite correlations, though usually significant, are

inadequate for prediction, giving rise to errors of
misclassification18,19 because of biological variations in
BMD, and because of technical errors of accuracy. The
Age (years) same holds true in principle for many other multifactorial
50 diseases. For example, in hypertension, measurements
made at the leg can differ substantially from measurements
made at the arm. In the specialty of hypertension, to select a
standardised site for diagnosis is appropriate, though such
standardisation should not negate the use of other sites (or
techniques) for risk assessment.
For diagnosis, measurement at the hip is the gold
standard in terms of site, since it has the highest predictive
50–59 value for hip fracture,20 which is the most severe
complication of osteoporosis and predicts risk of all
fractures as well as other techniques. These considerations
should not be taken to infer that other techniques and
measurements with DXA at other sites are not useful.
Indeed they are for risk assessment rather than for
diagnosis.
60–69 Diagnosis in men
Suitable diagnostic cutoff values for men are less well
defined than for women. Many studies21–25 have examined
fracture risk in men and women and have variously
concluded that the risk rises with decreasing BMD, or the
fracture threshold is the same or differs between sexes.
There are several reasons for these discrepancies. First, the
relation between BMD and fracture risk changes with
70–79 age,26,27 so that age adjustment is required. Second, a
difference between sexes in the gradient of risk (relative risk
per SD increase in BMD) could be a result of differences in
the SD of measurements. Third, data from referral
populations of osteoporotic men and women could be
biased. These difficulties are overcome by sampling
populations at random and expressing risk as a function of
BMD or standardised T scores, with age adjustment. The
>80 few studies available28,29 show that the risk of hip fracture is
similar in men and women for any given BMD. Such
studies indicate that a similar cutoff value for hip BMD that
is used in women can be used in the diagnosis of
–6 –5 –4 –3 –2 –1 0 1 2 3 osteoporosis in men—namely, a value for BMD 2·5 SDs or
Bone mineral density (T score) more below the average for women.12

Figure 2: Distribution of bone mineral density in women of Reference ranges

different ages, and the prevalence of osteoporosis (blue)14 Of particular importance is the type of normal reference
T score below –2·5=osteoporosis. Reproduced from reference 14 by range used, which should be taken from appropriate
permission of the American Society for Bone and Mineral Research. populations. Small differences between ranges have a large
effect on the number of individuals with BMD below a
osteoporosis at any one anatomical site increases greatly diagnostic threshold. The current recommendation of the
with age in much the same way as fracture risk increases International Osteoporosis Foundation and WHO is to use
with age (figure 2).14 Indeed, the increase in prevalence is the National Health and Nutrition Examination Survey
roughly exponential and conforms to the known pattern of (NHANES) reference database in women aged 20–29 years
frequency of many osteoporotic fractures in ageing women. as the reference range.12
When measurements are made at one site, for example at
the hip, then the prevalence of osteoporosis of the hip in Assessment of fracture risk
white women aged 50 years or more is about one in six, DXA and quantitative ultrasound
which is close to the life-time risk of hip fracture.16 The clinical consequence of osteoporosis is the fractures
that arise. There is, therefore, a great deal of interest in
Sites and techniques the prognostic use of bone mineral measurements—
The T score cannot be used, for diagnosis, interchangeably namely, their ability to predict the likelihood of fractures.
with different techniques or be based on measurements In this sense the accuracy of the techniques is not how
taken from different sites, since the same T score derived closely they measure BMD, but their sensitivity and
from different sites and with different techniques yields specificity to predict fractures. Many well controlled
different information on fracture risk. Reasons for this prospective studies with DXA, particularly in elderly
variation include differences in the gradient of risk for women, indicate that the risk of fracture about doubles for
techniques to predict fracture, discrepancies in the each SD reduction in BMD (table 1).20
population SDs, and differences in the apparent rates of The gradient of risk (increase in fracture risk for specific
bone loss with age.12,17 A further difficulty is that the change in BMD) depends on the technique used, the site

THE LANCET • Vol 359 • June 1, 2002 • www.thelancet.com 1931

For personal use. Only reproduce with permission from The Lancet Publishing Group.
OSTEOPOROSIS III

Forearm fracture Hip fracture Vertebral fracture All fractures

Site of measurement
Distal radius 1·7 (1·4–2·0) 1·8 (1·4–2·2) 1·7 (1·4–2·1) 1·4 (1·3–1·6)
Femoral neck 1·4 (1·4–1·6) 2·6 (2·0–3·5) 1·8 (1·1–2·7) 1·6 (1·4–1·8)
Lumbar spine 1·5 (1·3–1·8) 1·6 (1·2–2·2) 2·3 (1·9–2·8) 1·5 (1·4–1·7)
Adapted from reference 20 by permission of the BMJ publishing Group.
Table 1: Age-adjusted relative increase in risk of fracture (with 95% CI) in women for every 1 SD decrease in bone mineral density
(absorptiometry) below the mean value for age20

measured, and the fracture of interest. In general, site- phosphatase, osteocalcin, and the procollagen propeptides
specific measurements show the higher gradients of risk for of type I collagen. The most widely used markers of bone
their respective sites. For example, measurements at the hip resorption are hydroxyproline, and the pyridinium
predict hip fracture with greater power than do crosslinks and their associated peptides.
measurements at lumbar spine or forearm. Gradients of risk Bone markers are increased after the menopause, and the
range from 1·5 to 3·0 for each SD decrease in bone mineral results of several studies indicate that the rate of bone loss
measurement (table 1). Gradients of risk for different varies according to the marker value. Thus a potential
sites are independent of age. The clinical application of biochemical
performance characteristics of Panel 2: Risk factors for osteoporotic indices of skeletal metabolism is in
ultrasound are similar for prognostic fractures assessment of fracture risk. Findings
use. Results of most studies suggest of prospective studies indicate an
that measurements of broad band Female sex association between osteoporotic
ultrasound attenuation or speed of Premature menopause fracture and indices of bone
sound are associated with a 1·5-fold Age* turnover, independent of bone
to 2·0-fold increase in risk for each Primary or secondary amenorrhoea density in women at the meno-
SD reduction in BMD.5 Findings of Primary and secondary hypogonadism in man pause35,36 and in elderly women.37 In
some, but not all, studies suggest that Asian or white ethnic origin elderly women with values for
ultrasound might measure some Previous fragility fracture* resorption markers that exceed the
aspects of skeletal status and Low bone mineral density reference range for premenopausal
fragility that cannot be measured Glucocorticoid therapy* women, fracture risk is increased
with absorptiometric techniques High bone turnover* about two-fold after adjustment for
alone.7,30,31 Family history of hip fracture* BMD. These results suggest that a
Estimation of fracture risk by Poor visual acuity* combined approach, with BMD and
BMD measurements is similar to the Low bodyweight* indices of bone turnover, could
assessment of the risk of stroke by Neuromuscular disorders* improve fracture prediction in post-
blood pressure readings. Blood Cigarette smoking* menopausal women.38
pressure values are continuously Excessive alcohol consumption
distributed in the population, as is Long-term immobilisation Clinical risk factors
BMD. In the same way that a patient Low dietary calcium intake Many risk factors for osteoporosis
above a cutoff for blood pressure is Vitamin D deficiency have been identified (panel 2). In
diagnosed as hypertensive, the general, risk factor scores show poor
*Characteristics that capture aspects of fracture
diagnosis of osteoporosis is based on specificity and sensitivity in
risk over and above that provided by bone mineral
a value for BMD below a cutoff prediction of either BMD or fracture
density.
threshold. As is the case for blood risk.39–42 Moreover, some risk factors
pressure, there is no threshold of vary in importance according to age.
BMD that discriminates absolutely between those who will For example, risk factors for falling—eg, visual impairment,
or will not have a clinical event. The ability to predict hip reduced mobility, and treatment with sedatives—are
fracture by measurement of BMD is, however, at least as more strongly predictive of fracture in the elderly than in
good as that of blood pressure in predicting stroke, and younger individuals.43
considerably better than the use of serum cholesterol to Hypogonadism is an important risk factor for
predict coronary artery disease.3,32 Nevertheless, that a osteoporosis in both sexes. In young women, hypogonadism
normal BMD measurement is no guarantee that fracture can be primary or secondary to conditions such as anorexia
will not occur should be recognised—only that the risk is nervosa, exercise-induced amenorrhoea, chronic illness,
reduced. Conversely, if BMD is in the osteoporotic range, hyperprolactinaemia, and gynaecological disorders.
then fractures are more likely. At age 50 years, the Premature menopause, either spontaneous or induced by
proportion of women with osteoporosis who will fracture surgery, chemotherapy, or radiotherapy is also associated
their hip, spine, or forearm or proximal humerus in the next with increased risk of osteoporosis. In men, hypogonadism
10 years—ie, positive predictive value—is about 45%. The can be caused by various disorders, including Klinefelter’s
detection rate for these fractures (sensitivity) is, however, syndrome, hypopituitarism, hyperprolactinaemia, and
low, and 96% of such fractures would arise in women castration—for example, after prostatic surgery.
without osteoporosis.33 Low sensitivity is one of the reasons Glucocorticoids are an important cause of osteoporosis.
why widespread population-based screening is not recom- Bone loss is believed to be most rapid in the first few
mended in women at the menopause.3 months of treatment and affects both the axial and
appendicular skeleton, but is most pronounced at the spine,
Biochemical assessment of fracture risk where cancellous bone predominates. Bone loss can be
Biochemical indices of bone turnover can be divided into avoided by inhaled glucocorticoid therapy.44 Although the
two groups: markers of resorption and markers of skeletal response to glucocorticoids varies between
formation.34 The principal markers of bone formation are individuals, high doses are generally associated with
total alkaline phosphatase, the bone isoenzyme alkaline greater adverse skeletal effects, whereas daily doses of

1932 THE LANCET • Vol 359 • June 1, 2002 • www.thelancet.com

For personal use. Only reproduce with permission from The Lancet Publishing Group.
 OSTEOPOROSIS III

prednisolone below 7·5 mg are less likely to result in

increased rates of bone loss and fracture.45 Panel 3: Risk factors that provide indications for
Furthermore, a history of fragility fracture is an the diagnostic use of bone densitometry4
important risk factor for further fracture—eg, risk of • Presence of strong risk factors
subsequent fracture of the hip is increased by more than
two-fold after previous fracture of the hip or spine. This Oestrogen status
risk is reduced though still present if the site of previous Premature menopause (<45 years)
fracture is the forearm (risk increased by 1·9) or the Long-term secondary amenorrhoea (>1 year)
proximal humerus (2·0). Risk of fracture of the spine is Primary hypogonadism
similarly increased after a previous break at the hip (2·5), Corticosteroid therapy
the spine (4·4), the forearm (1·7), or the proximal Prednisolone (or equivalent) 7·5 mg per day or more with an
humerus (1·9).46 The presence of two or more prevalent expected use of more than 6 months
vertebral fractures is associated with a 12-fold increase in Maternal family history of hip fracture
fracture risk for any specific BMD.47
Results of case-control studies of hip fractures in men Low body-mass index (<19 kg/m2)
and women show an increased risk of fracture with Other disorders associated with osteoporosis
disorders associated with secondary osteoporosis, such as Anorexia nervosa
previous hyperthyroidism, gastric surgery, and Malabsorption syndromes, including chronic liver disease and
hypogonadism.39,48–50 There is also a greater risk of hip inflammatory bowel disease
fracture with conditions related to an increased risk of Primary hyperparathyroidism
falling, such as hemiparesis, Parkinson’s disease, Post-transplantation
dementia, vertigo, alcoholism, and blindness.24,48 Chronic renal failure
Of the risk factors shown in panel 2, smoking, alcohol, Hyperthyroidism
and poor calcium nutrition are weak risks. Complete Long-term immobilisation
immobilisation leads to rapid bone loss at the affected Cushing's syndrome
sites, but evidence that lesser degrees of physical inactivity
increase the risk of osteoporosis is not so well • Radiographic evidence of osteopenia or vertebral
documented. A low body-mass index is an important risk deformity or both
factor for osteoporosis and fractures, probably because of • Previous fragility fracture, especially of the spine or wrist
its association with bone size. Finally, a parental history of
hip fracture is an independent risk factor for fracture. For • Loss of height, thoracic kyphosis (after radiographic
any specific BMD, hip fracture risk is increased about confirmation of vertebral deformities)
two-fold.40 Reproduced from reference 4 by permission of Osteoporosis
International.
Identification of individuals for treatment
At present there is no universally accepted policy for
screening to identify individuals at high risk of fracture. women at the menopause. In the absence of a validated
The test used to diagnose osteoporosis, bone screening strategy, a case finding strategy is recommended
densitometry, has low sensitivity (detection rate) at in which patients are identified because of a fragility
acceptable specificity. Thus, the risk of fracture is very fracture or by the presence of other strong risk factors for
high when osteoporosis is present, but the risk of fracture fracture, and thereafter assessed by BMD measurement
is by no means negligible when BMD is normal (figure 3). (panel 3). This strategy forms the basis of a case-finding
Because factors in addition to BMD can be measured that approach widely used in Europe and the USA.4,51,52
contribute independently to fracture risk, screening In Europe, treatment is recommended in the presence
strategies for fracture prediction might be developed in of osteoporosis, but in the USA less stringent criteria are
the future. Such a strategy is likely to be targeted to used. Not all patients with such risk factors need
elderly people, in whom fracture probability is higher, as diagnostic assessment. For example, patients with more
too is the prevalence of many of the risk factors, than in than one vertebral fracture should be offered treatment
irrespective of their bone density, although the
measurement could help to monitor treatment effect.
50 The use of risk factors that add information on fracture
risk independently of BMD (see panel 2) improves the
Lifetime risk (%)

40 sensitivity of the assessment for any specificity.33,53 Such

factors can be used to enhance the case-finding strategy,
30
although some care is needed in the use of density
20 independent risk factors to identify individuals for drug
treatments that affect skeletal metabolism. For example,
10 inhibitors of bone turnover might not be effective in
Osteoporosis Osteopenia Normal populations selected on the basis of a fall. With this
0 proviso, independent risk factors could be used to
0·6 0·7 0·8 0·9 1 1·1 1·2 enhance the predictive value of BMD.
Femoral neck BMD (g/cm2) An example of the use of independent risk factors is
provided by the interaction of biochemical markers of
–3 –2 –1 0 +1 skeletal turnover and BMD. Results of the EPIDOS
T score (SD units) study54 show independent contributions of BMD and
urinary resorption markers on hip fracture risk in elderly
Figure 3: Remaining lifetime risk of hip fracture in women women with a mean age of 81 years.54 At this age, the
aged 50 years, according to bone mineral density (BMD) or average remaining 10-year risk of hip fracture is 15%. By
T score at the hip selecting women with osteoporosis, the 10-year risk for

THE LANCET • Vol 359 • June 1, 2002 • www.thelancet.com 1933

For personal use. Only reproduce with permission from The Lancet Publishing Group.
 OSTEOPOROSIS III

Age (years) view that intervention thresholds should be based on

50 60 70 80
fracture risk rather than on any particular level of BMD.
The use of high gradients of fracture risk by combining
Relative risk risk factors not only improves the detection rate, but also
Hip fracture
Men
enlarges the population that can be selected with a
1 0·84 1·26 3·68 9·53 particular threshold of risk. Say, for example, one wanted
2 1·68 2·50 7·21 17·89 to identify individuals with a risk of fracture three times
3 2·51 3·73 10·59 25·26 greater than the average risk for the general population, a
4 3·33 4·94 13·83 31·75 test with a gradient of risk of 2·0 SD would identify 2·7%
Women
1 0·57 2·40 7·87 18·0
of the population, but a test with a gradient of risk of 3·0
2 1·14 4·75 15·1 32·0 SD would detect 6·1% of the population.55
3 1·71 7·04 21·7 42·9
4 2·27 9·27 27·7 51·6 Intervention thresholds
Hip, clinical spine, humeral, or Colles’ fracture Absolute risk depends on age and life expectancy as well as
Men
present relative risk.56 Estimates of lifetime risks are of
1 3·3 4·7 7·0 12·6
2 6·5 9·1 13·5 23·1 value in considering the burden of osteoporosis in the
3 9·6 13·3 19·4 13·9 community and the effects of intervention strategies. For
4 12·6 17·3 24·9 39·3 several reasons they are less relevant for assessment of risk
Women of individuals in whom treatment might be envisaged.
1 5·8 9·6 16·1 21·5 First, treatments are not prescribed for a lifetime, because
2 11·3 18·2 29·4 37·4
3 16·5 26·0 40·0 49·2
of side-effects of continued treatment—eg, hormone
4 21·4 33·1 49·5 58·1 replacement therapy—or low compliance. Moreover, the
feasibility of lifelong interventions has not been tested with
Reproduced from reference 33 by permission of Elsevier Science.
either high risk or global strategies.52 Second, the
Table 2: 10-year probability of fracture (%) in men and women predictive value of low BMD and some other risk factors
from Sweden, according to age and risk relative to the for fracture risk is attenuated over time.57 Finally, the
average population33 confidence in estimates falls with time because of
uncertainties about future mortality trends. For
hip fractures rises to 25%. Selection of women on the basis this reason, the International Osteoporosis Foundation
of values for urinary markers above the premenopausal and WHO recommend that risk of fracture should
range shows a risk of 33%. With the combination of low be expressed as an absolute risk—ie, probability—over
hip BMD and high resorption markers, the 10-year 10 years.12 This period covers the probable duration of
fracture risk is 49%.38 Such considerations have led to the treatment and benefits that might continue once treatment
is stopped.58
100 Table 2 shows the 10-year fracture probabilities for the
common osteoporotic fractures, according to the
T score
population relative risk.33 Note that the probabilities are
–4
derived from Swedish data, which show a high incidence
30 of fracture and should be adjusted downwards for
countries where the age-specific risk is low. Probabilities
–3 can similarly be derived from the T-score result of BMD
assessment (figure 4).26,59
10 Since the aim of the assessment of fracture risk is to
target interventions accurately to those at highest risk, and
–2
avoid treatment of those at low risk, an important question
Probability (%)

is what is the cut off value for relative risk, BMD, or 10-
3 year risk that provides an intervention threshold. This
–1 issue is complex and depends on clinical practice,
effectiveness of treatment (compliance, continuance, and
efficacy), side-effects of treatment, the type of fracture
1 expected, and the costs of treatment. Several agencies in
0 Europe and the USA have constructed evidence-based
practice guidelines in which intervention thresholds are
based on health economic analyses.4,51,52 Although there are
0·3
important differences between the approaches,60 these
agencies would agree that for most interventions
envisaged, individuals with osteoporosis should be offered
treatment, which can be justified from a health economics
perspective.
0·1
When hip fracture alone is considered, a 10-year
50 60 70 80
probability of 10% or more provides a cost-effective
Age (years)
threshold for women in Sweden.61 However, many
fractures other than hip fracture also contribute to
Figure 4: 10-year probability of hip fracture in Swedish men and
morbidity, particularly in the young in whom hip fractures
women, according to T scores assessed at the femoral neck
are rare. When account is taken of such fractures,62 cost-
by dual X-ray absorptiometry
effective intervention probabilities decrease, especially in
Probability scale is logarithmic. Green dotted line=probability at which young individuals (figure 4).59 Note that cost-effective
interventions are cost effective.26,59 Reproduced from reference 26 by interventions can be provided to most women with
permission of Osteoporosis International. osteoporosis.

1934 THE LANCET • Vol 359 • June 1, 2002 • www.thelancet.com

For personal use. Only reproduce with permission from The Lancet Publishing Group.
 OSTEOPOROSIS III

Conclusions invasive bone mineral measurements in assessing age related loss,

The diagnosis of osteoporosis is generally based on fracture discrimination, and diagnostic classification. J Bone Miner Res
1997; 12: 697–711.
assessment of BMD at the proximal femur by DXA. By
20 Marshall D, Johnell O, Wedel H. Meta-analysis of how well measures
contrast, intervention thresholds should be based on of bone mineral density predict occurrence of osteoporotic fractures.
fracture probability. Several clinical risk factors for BMJ 1996; 312: 1254–59.
fracture with and without BMD allow the more accurate 21 Orwoll E. Assessing bone density in men. J Bone Miner Res 2000; 15:
stratification of risk than the use of BMD alone. In the 1867–70.
absence of validated screening strategies, a case-finding 22 Selby PL, Davies M, Adams JE. Do men and women fracture bones at
similar bone densities. Osteoporos Int 2000; 11: 153–57.
approach is advocated for individuals with strong risk
23 Lunt M, Felsenberg D, Reeve J, Benevolenskaya L, Cannata J,
factors who are referred for BMD assessment. Dequeker J. Bone density variation and its effect on risk of
Intervention is best targeted to those in whom fracture vertebral deformity in men and women studied in thirteen
probability exceeds a threshold of reversible risk, based on European Centres: the EVOS Study. J Bone Miner Res 1997; 12:
cost-effectiveness. 1883–94.
24 Nguyen T, Sambrook SP, Kelly P, Jones G, Freund J, Eisman J.
Prediction of osteoporotic fractures by postural instability and bone
Conflict of interest statement density. BMJ 1993; 307: 1111–15.
J A Kanis has received research funding and done ad-hoc consultancies 25 Melton LJ, Atkinson EJ, O’Connor MK, O’Fallon WM, Riggs BL.
for most companies with an interest in osteoporosis, and has equity in Bone density and fracture risk in men. J Bone Miner Res 1998; 13:
British Biotec, UK; Celtrix, USA; Glaxo, UK; Lilly, USA; Merck, USA; 1915–23.
Novo Nordisk, Denmark; Procter and Gamble, USA; Shire, UK; Strakan,
26 Kanis JA, Johnell O, Oden A, Dawson A, De Laet C, Jonsson B.
UK; and Unigene, USA.
Ten year probabilities of osteoporotic fractures according to BMD
and diagnostic thresholds. Osteoporos Int 2001; 12: 989–95.
References 27 Hui SL, Slemenda CW, Johnston CC. Age and bone mass as
predictors of fracture in a prospective study. J Clin Invest 1988; 81:
1 Consensus Development Conference. Diagnosis, prophylaxis and
1804–09.
treatment of osteoporosis. Am J Med 1993; 94: 646–50.
28 De Laet CEDH, Van Hout BA, Burger H, Hofman A, Weel AEAM,
2 Genant HK, Engelke K, Fuerst T, et al. Non invasive assessment of
Pols HAP. Hip fracture prediction in elderly men and women:
bone mineral and structure: state of the art. J Bone Miner Res 1996;
validation in the Rotterdam Study. J Bone Miner Res 1998; 13:
11: 707–30.
1587–93.
3 WHO. Assessment of fracture risk and its application to screening for
29 Kanis JA, Johnell O, Oden A, De Laet C, Mellstrom D. Diagnosis of
postmenopausal osteoporosis: technical report series 843. Geneva:
osteoporosis and fracture threshold in men. Calcif Tissue Int 2001; 69:
WHO, 1994.
218–21.
4 Kanis JA, Delmas P, Burckhardt P, et al. Guidelines for diagnosis
30 Heaney RP, Kanis JA. The interpretation and utility of ultrasound
and management of osteoporosis. Osteoporos Int 1997; 7:
measurements of bone. Bone 1996; 18: 491–92.
390–406.
31 Ross P, Huang C, Davis J, et al. Predicting vertebral deformity using
5 Glüer CC, for the International Quantitative Ultrasound Consensus
bone densitometry at various skeletal sites and calcaneous ultrasound.
Group. Quantitative ultrasound techniques for the assessment of
Bone 1995; 16: 325–32.
osteoporosis: expert agreement on current status. J Bone Miner
Res 1997; 12: 1280–88. 32 Cooper C, Aihie A. Osteoporosis: recent advances in pathogenesis and
6 Gregg EW, Kriska AM, Salamone LM, et al. The epidemiology of treatment. Q J Med 1994; 87: 203–09.
quantitative ultrasound: a review of the relationship with bone mass, 33 Kanis JA, Johnell O, Oden A, De Laet C, Jonsson B, Dawson A.
osteoporosis and fracture risk. Osteoporos Int 1997; 7: 89–99. Ten year risk of osteoporotic fracture and the effect of risk factors
7 Hans D, Dargent-Molina P, Schott AM, et al. Ultrasonographic heel on screening strategies. Bone 2001; 30: 251–58.
measurements to predict hip fracture in elderly women: the EPIDOS 34 Delmas PD, ed. The use of biochemical markers of bone turnover in
prospective study. Lancet 1996; 348: 511–14. the management of post–menopausal osteoporosis. Osteoporosis Int
8 Porter RW, Miller CG, Granger D, Palmer SB. Prediction of hip 2000; 11 (suppl 6): S1–76.
fractures in elderly women; a prospective study. BMJ 1990; 301: 35 Garnero P, Sornay-Rendu E, Claustrat B, Delmas PD. Biochemical
638–41. markers of bone turnover, endogenous hormones and the risk of
9 Genant HK, Cann CE, Ettinger B, Gordan GS. Qualitative computed fractures in postmenopausal women. J Bone Miner Res 2000; 15:
tomography of vertebral spongiosa: a sensitive method for detecting 1526–36.
early bone loss after oophorectomy. Ann Intern Med 1982; 97: 36 Hansen M, Overgaard K, Riis B, Christiansen C. Role of peak bone
699–705. mass and bone loss in postmenopausal osteoporosis: 12 year study.
10 Lang T, Augat P, Majumdar S, Ouyang X, Genant HK. Non-invasive BMJ 1991; 303: 961–64.
assessment of bone density and structure using computed tomography 37 Garnero P, Hauser E, Chapuy MC, et al. Markers of bone turnover
and magnetic resonance. Bone 1998; 2 (suppl): 149–53. predict hip fractures in elderly women. J Bone Min Res 1996; 11:
11 Ruegsegger P, Elsasser U, Anliker M, Gnehm H, Kind H, Prader A. 1531–38.
Quantification of bone mineralisation using computed tomography. 38 Johnell O, Oden A, De Laet C, Garnero P, Delmas PD, Kanis JA.
Radiology 1976; 121: 93–97. Biochemical markers and the assessment of fracture probability.
12 Kanis JA, Glüer CC, for the Committee of Scientific Advisors, Osteoporos Int (in press).
International Osteoporosis Foundation. An update on the diagnosis 39 Johnell O, Gullberg B, Kanis JA, et al. Risk factors for hip fracture in
and assessment of osteoporosis with densitometry. Osteoporos Int 2000; European women: the MEDOS study. J Bone Miner Res 1995; 10:
11: 192–202. 1802–15.
13 Bonjour JP, Rizzoli R. Bone acquisition in adolescence. In: Marcus R, 40 Cummings SR, Nevitt MC, Browner WS, et al. Risk factors for hip
Feldman D, Kelsey J, eds. Osteoporosis. San Diego: Academic Press, fracture in white women. N Engl J Med 1995; 332: 767–73.
1996: 465–76. 41 Compston JE. Risk factors for osteoporosis. Clin Endocrinol 1992; 36:
14 Kanis JA, Melton LJ, Christiansen C, Johnston CC, Khaltaev N. 223–24.
The diagnosis of osteoporosis. J Bone Miner Res 1994; 9: 42 Ribot C, Pouilles JM, Bonneu M, Tremollieres F. Assessment of the
1137–41. risk of postmenopausal osteoporosis using clinical risk factors.
15 Kanis JA, McCloskey EV, de Takats D, Pande K. Clinical assessment Clin Endocrinol 1992; 36: 225–28.
of low bone mass, quality and architecture. Osteoporos Int 1999; 43 Kanis JA, McCloskey EV. Evaluation of the risk of hip fracture.
9 (suppl 1): 24–28. Bone 1996; 18 (suppl 3): 127–32.
16 Kanis JA, Johnell O, Oden A, et al. Long-term risk of osteoporotic 44 Van Staa TP, Leufkens HGM, Cooper C. Use of inhaled
fracture in Malmo. Osteoporos Int 2000; 11: 669–74. corticosteroids and risk of fractures. J Bone Miner Res 2001; 16:
17 Faulkner KG, von Stetten E, Miller P. Discordance in patient 581–88.
classification using T-scores. J Clin Densitometry 1999; 2: 45 Van Staa TP, Leufkens HGM, Abenhaim L, Zhang B, Cooper C.
343–50. Use of oral corticosteroids and risk of fractures. J Bone Miner Res
18 Arlot ME, Sornay-Rendu E, Garnero P, Vey-Marty B, Delmas PD. 2001; 15: 993–1000.
Apparent pre- and postmenopausal bone loss evaluated by DXA at 46 Klotzbuecher CM, Ross PD, Landsman PB, Abbot TA, Berger M.
different skeletal sites in wom en: the OFELY cohort. J Bone Miner Patients with prior fractures have increased risk of future fractures:
Res 1997; 12: 683–90. a summary of the literature and statistical synthesis. J Bone Miner Res
19 Grampp S, Genant HK, Mathur A, et al. Comparisons of non- 2000; 15: 721–27.

THE LANCET • Vol 359 • June 1, 2002 • www.thelancet.com 1935

For personal use. Only reproduce with permission from The Lancet Publishing Group.
OSTEOPOROSIS III

47 Ross PD, Genant HK, Davis JW, Miller PD, Wasnich RD. Predicting Osteoporos Int 1998; 8: 563–69.
vertebral fracture incidence from prevalent fractures and bone density 55 De Laet CED, Oden A, Johnell O, Jonsson B, Kanis JA. Case-finding
among non black, osteoporotic women. Osteoporos Int 1993; 3: and opportunistic screening: a mathematical approach. J Bone Miner
120–26. Res 2001; 16 (suppl 1): S414.
48 Poor G, Atkinson EJ, O’Fallon WM, Melton LJ III. Predictors of hip 56 Oden A, Dawson A, Dere W, Johnell O, Jonsson B, Kanis JA.
fractures in elderly men. J Bone Miner Res 1995; 10: 1900–07. Lifetime risk of hip fracture is underestimated. Osteoporos Int 1999;
49 Stanley HL, Schmitt BP, Poses RM, Deiss WP. Does hypogonadism 8: 599–603.
contribute to the occurrence of a minimal trauma hip fracture in 57 Kanis JA, Johnell O, Oden A, Jonsson B, DeLaet C, Dawson A.
elderly men? J Am Geriatr Soc 1991; 39: 766–71. Prediction of fracture from low bone mineral density measurements
50 Kanis JA, Johnell O, Gullberg B, et al. Risk factors for hip fracture in overestimates risk. Bone 2000; 26: 387–91.
European men. Osteoporos Int 1999; 9: 45–54. 58 Jonsson B, Kanis JA, Dawson A, Oden A, Johnell O. Effect and offset
51 National Osteoporosis Foundation. Analyses of the effectiveness and of effect of treatments for hip fracture on health outcomes.
cost of screening and treatment strategies for osteoporosis: a basis for Osteoporos Int 1999; 10: 193–99.
development of practice guidelines. Osteoporos Int 1998; 8 (suppl 4): 59 Kanis JA, Johnell O, Oden A, De Laet C, Oglesby A, Jonsson B.
1–88. Intervention thresholds for osteoporosis. Bone (in press).
52 Royal College of Physicians. Clinical guidelines for the prevention and 60 Kanis JA, Torgerson D, Cooper C. Comparison of the European and
treatment of osteoporosis. London: RCP, 1999. US practice guidelines for osteoporosis. Trends Endocrinol Metab 2000;
53 Kanis JA, Johnell O, Oden A, Jonsson B, Dawson A, Dere W. 11: 28–32.
Risk of hip fracture in Sweden according to relative risk: an 61 Kanis JA, Dawson A, Oden A, Johnell O, De Laet C, Jonsson B.
analysis applied to the population of Sweden. Osteoporos Int 2000; Cost-effectiveness of preventing hip fracture in the general female
11: 120–27. population. Osteoporos Int 2001; 12: 356–61.
54 Garnero P, Dargent–Molina P, Hans D, et al. Do markers of bone 62 Kanis JA, Oden A, Johnell O, Jonsson B, De Laet C, Dawson A.
resorption add to bone mineral density and ultrasonographic heel The burden of osteoporotic fractures: a method for setting intervention
measurement for the prediction of hip fracture in elderly women? thresholds. Osteoporos Int 2001; 12: 417–27.

Uses of error
The dangers of conformity
John Swales

Let me start by stating categorically that there is no totally dependent on her. On this occasion the ritual of
shortage of clinical errors in my professional career. A spell clinical enquiry and his blood pressure measurement was
as a pre-registration casualty officer in a London hospital over when she murmured in a curiously forced, yet off-
allowed me to contribute a sizeable quota to the hand way that she had been suffering from headaches. I
misdiagnosed chest pains and overlooked fractured had no difficulty in recognising the hidden importance of
scaphoids that finally led to the merciful abolition of such what was being said by my patient’s sister. My response
junior posts. I learnt one thing from these experiences: the was professionally correct—to suggest as sympathetically
fundamental mistake is failure to recognise when someone as I could that she should make an appointment to see her
is genuinely ill or in pain. Without the ability to do this, doctor as soon as possible. I do not know whether she did
textbook knowledge is singularly futile. My memorable or not. At his next visit, my patient came alone. His sister
error is somewhat different, although I hope that the lesson had died of a cerebral haemorrhage and uncontrolled
is also broadly applicable. In the early 1970's malignant hypertension shortly after seeing me.
hypertension was much more common than it is today. What I did was consistent with professional etiquette.
Hypertension clinics cared for substantial numbers of such She did not obviously need emergency treatment on the
patients who unquestionably owed their lives to day I saw her and she was already under her GP’s care. To
complicated and often fairly unpleasant medication. Their employ that language of possession that we owe to the
survival was in some ways a source of pride at a time when nineteenth century, she was not my patient. Of course, I
treating milder forms of hypertension was still being could have done more, even within the limits of
questioned. professional guidance. I could have spoken to her GP and
One of the patients who had been rescued by anti- offered help, and he may well have welcomed some
hypertensive treatment had presented a couple of years advice. But that is not my major concern. My regret is that
before to the emergency room with malignant I allowed a rigid view of professional etiquette to dictate
hypertension. He differed from the other patients as he was what I did against my initial instinct and this ruined two
a chronic schizophrenic and would express rather pathetic lives. It would have been quite straightforward to take her
gratitude for our interest, but otherwise said little. By the blood pressure and look at her fundi. Medicine at times
time I came to know him he owed, I felt, less to us than to demands that the wrong thing is done for the right
his sister, who lived with him, looked after him and reasons. I should have done the wrong thing. That was
invariably accompanied him to hospital. He was clearly my error.

Faculty of Medicine, University of Leicester, Leicester LE2 7LX, UK (Prof J Swales)

Prof J Swales died on October 17, 2000

1936 THE LANCET • Vol 359 • June 1, 2002 • www.thelancet.com

For personal use. Only reproduce with permission from The Lancet Publishing Group.