Does This Woman Have Osteoporosis?

You recommend screening densitometry to a healthy 64-year-old woman. She will have to drive an hour to the nearest testing center and doesn’t believe that she needs the test. To further assess her risk, you note that she weighs 49 kg (108 lb). What can you tell this patient about her probability of osteoporosis?

A frail, 79-year-old woman is admitted to the hospital with a diverticular bleed. On examination, you note that she has significant kyphosis. When she stands upright against a wall, she cannot touch the back of her head to the wall. You wonder whether she has vertebral fractures.

A 58-year-old woman presents for her annual examination. She experienced physiologic menopause 8 years ago but is asymptomatic and has no other risk factors for osteoporosis. On examination you note that her rib-pelvis distance is 1 fingerbreadth. She tells you that she has developed a humped back. Should this patient be referred for densitometry?

Osteoporosis causes 1.5 million fractures per year in the United States.¹ As the population continues to age, this number is expected to double by 2040.² Half of all postmenopausal women and 15% of white men older than 50 years will have an osteoporosis-related fracture in their lifetime, with 15% of those occurring in the hip. Pain, loss of independence, impaired ambulation, depression, and nursing home admission are common sequelae.^{3 - 8}

In 1995, health care spending for osteoporotic fractures in the United States was $13.8 billion and is estimated to be $31 billion to $62 billion by 2020.⁹ The US Preventive Services Task Force (USPSTF) recommends that women 65 years of age or older be screened routinely for osteoporosis and that women younger than 65 years be screened if they have risk factors.¹⁰ There are no current guidelines on when to screen perimenopausal women with few to no risk factors or men.

The physical examination may assist clinicians in preventing osteoporotic fractures in several ways. First, it may identify patients with low bone mineral density (BMD), in whom routine screening is not currently recommended or has not been completed. It may also identify patients at very low risk of osteoporosis, in whom BMD testing is unnecessary. While it is an imperfect indicator of fracture risk, BMD measurement is widely used both in randomized controlled trials and in clinical practice as the primary criterion for initiating osteoporosis therapies.

Second, the physical examination could identify patients with occult vertebral fracture. Two thirds of vertebral fractures are clinically silent but are associated with a 2- to 3-fold increased risk of further fractures. Several osteoporosis therapies reduce the risk of further fractures in women with vertebral fractures, and the National Osteoporosis Foundation algorithm suggests that patients found to have vertebral fracture should be treated regardless of their BMD measurement.¹¹ Thus, the objective of this review was to identify clinical examination findings that improve the identification of patients with low BMD or occult vertebral fractures who would benefit from therapy or in whom further screening with BMD testing is unnecessary.

Osteoporosis is a skeletal disorder characterized by compromised bone strength predisposing a person to an increased risk of fracture. For this review, we used the World Health Organization’s definition of osteoporosis based on BMD that compares a patient’s density to normative values for a population of 20- to 40-year-olds in terms of the number of standard deviations from the mean value. Osteoporotic bones have a density that is more than 2.5 SDs below the mean (T score <−2.5). Osteopenic bones have a T score that is between –2.5 and –1. Normal bones have a BMD T score of −1 or higher.¹²

Vertebral fractures are compression deformities that reduce vertebral body height by 20% or more on imaging studies; most of the articles included in this review used a semiquantitative technique to diagnose vertebral fractures on plain lateral radiographs of the spine. Spinal fractures are classified by the maximal percentage of vertebral body height loss as follows: grade 1, 20% to 24%; grade 2, 25% to 39%; and grade 3, 40% or more.¹³

The prevalence of osteoporosis in large population-based studies allows an estimation of the pretest probability in women of varying ages. The prevalence of BMD-defined osteoporosis at the spine, wrist, or hip in white women in the United States by decade is as follows: for age 50 to 59 years, 14.8%; 60 to 69 years, 21.6%; 70 to 79 years, 38.5%; and 80 years or older, 70.0%.¹⁴ For nonwhite women older than 50 years, the prevalence of BMD-defined osteoporosis in the Third National Health and Nutrition Examination Survey was reported as follows: non-Hispanic black women, 12%; Mexican Americans, 18.6%; and women in other ethnic groups, 28.3%.¹⁵ In special populations, the prevalence of osteoporosis can be much higher. For example, in residents of skilled nursing facilities older than 75 years, the prevalence of osteoporosis exceeds 50% for all residents, regardless of race and sex.¹⁶

Occult vertebral fractures are also common and increase with age (Table 1). Grade 2 vertebral deformities are found in 6.6% of women aged 55 to 59 years and in 49.2% of women aged 80 to 84 years.¹⁷ Clinical characteristics or historical items that might increase a clinician’s pretest probability of osteoporosis or vertebral fracture include older age, low activity level, family history, hypogonadism (men), and exposure to glucocorticoids and alcohol. The pretest probability threshold for testing BMD depends on the anticipated benefit of treatment for an individual patient and the patient’s desire for treatment.

The pathophysiology of osteoporosis is related to physical examination findings in several ways. The loading or mechanical forces on bone tend to increase bone formation and bone mass through osteoblast stimulation. Thus, increasing body weight and muscle strength is inversely related to osteoporosis. Type I collagen is a major constituent of both bone and skin that is reduced with advancing age and low estrogen levels.^{18 - 20} Skinfold thickness may therefore reflect skeletal collagen content. Similarly, tooth loss is influenced by mandibular alveolar bone quality and may provide an easily observed marker of bone health in the rest of the skeleton.

The sequelae of clinically occult vertebral fractures can also lead to physical examination findings that may become apparent before a symptomatic fracture occurs. Height loss resulting from vertebral compression fractures can be measured in the clinic over time or using the patient’s recalled maximal adult height. Vertebral fractures affect height but not armspan, so armspan-height differentials may identify individuals with occult vertebral fractures.²¹ Thoracic kyphosis can result from anterior compression fractures in the thoracic spine (“dowager’s hump”). Kyphosis can be measured on physical examination using a curved ruler such as an architect’s rule or by measuring the wall-occiput distance. The wall-occiput distance describes the difference between the wall and the patient’s occiput when he/she stands straight with heels and back against the wall. Lumbar fractures also result in decreased rib-pelvis distance that can be measured in finger breadths on examination.

Data for several physical examination signs are included in this review. Weight and height are routinely measured in the clinical setting. Aside from clinic notes, height change can be documented from alternate sources (such as a driver’s license) or from the patient’s memory of height at age 25 years.^{22 - 24} Several studies have shown good to excellent correlation between elderly patients’ recalled maximal height and previous health records.^{25 - 27} A stadiometer (an upright bar marked with a height scale with a sliding notch to designate height) is the most accurate method of height measurement, although it is often not available in clinical settings.

Armspan-height differential is determined by subtracting a patient’s height in centimeters from the armspan in centimeters measured with arms at a 90° angle from the trunk. The armspan is the distance between the tips of the middle fingers while the patient faces forward with the arms fully extended and palms facing forward.

Measurements of thoracic kyphosis can be made indirectly on radiographs but can also be directly measured by applying an architect’s semiflexible rule, called a flexicurve, to the patient’s back.²⁸ The flexicurve is a device that can be bent in 1 plane only and retains its shape when applied to the curvature of the back between C7 spinous process and S2 spinous process level. The outline is traced on paper and the maximal angle measured with calipers or a ruler.²⁹ The kyphosis index is the ratio of thoracic curvature to the length of the upper back and is calculated as 100 times the maximum horizontal distance divided by the vertical length of the upper back curve. Flexicurve measurements, while painless, inexpensive, and safe, are time consuming.³⁰

Another measure that quantitates the degree of kyphosis is wall-occiput distance. It is measured while the patient stands straight with his/her back against the wall and heels touching the wall (Figure). While the head faces forward so that an imaginary line connecting the lateral corner of the eye to the superior junction of the auricle of the ear is parallel to the floor, the distance between the occipital prominence and the wall is quantified using a tape measure.³¹ For the purpose of this review, the inability to touch the wall with the back of the head is a positive finding.

Rib-pelvis distance is a measure of lumbar fracture. The patient stands erect with arms outstretched at 90°. The examiner stands behind the patient and inserts his or her fingers into the space between the inferior margin of the ribs and the superior surface of the pelvis in the midaxillary line. The rib-pelvis distance is the closest whole number of fingerbreadths between these structures.³²

Skinfold thickness is measured at the back of the hand with calipers.^{18 - 20} The back of the hand is a convenient site for measurement in the clinic.¹⁹ The fourth metacarpal longitudinal fold site was used in the studies of skinfold thickness included in this review.

Hand grip strength is measured using a small hydraulic hand grip or isometric dynamometer and is defined as the maximal force recorded while the patient squeezes the device with arms straight to the side.^{33 - 34}

We searched MEDLINE for articles from 1966 through August 2004 with a search strategy similar to that used by other authors in this series.³⁵ We used several National Library of Medicine Medical Subject Headings to encompass osteopenia, osteoporosis, and spinal fracture disease states: exp osteoporosis, exp spinal fracture, exp metabolic bone disease (for osteopenia), and exp bone density. The MEDLINE search was supplemented with a manual review of the bibliographies of all identified articles, additional review articles including recent osteoporosis guidelines, 4 clinical skills textbooks,^{36 - 39} and contact with experts in the field. Two authors (A.D.G. and M.T.D.) independently executed the MEDLINE search strategy and reviewed titles and abstracts from the search results. Two authors (A.D.G. and C.C.E.) then independently reviewed and extracted data from articles or abstracts identified as relevant. We contacted authors for original data when articles reported data on the precision of signs in diagnosing osteoporosis or spinal fracture but did not include enough information to calculate likelihood ratios (LRs).

We included studies in our review if they included original data on the accuracy or precision of the history or physical examination in diagnosing osteoporosis, osteopenia, or spinal fracture. We required that the gold standard comparison for the clinical examination parameters be bone densitometry at any site or documented vertebral fracture using either a semiquantitative technique or vertebral morphometry. When BMD values were reported directly, the corresponding T score was obtained using sex-appropriate tables provided by the manufacturers of the densitometer used in the study. Articles were excluded if they contained insufficient data to allow calculation of LRs. We included in our tables and results only the physical examination parameters that are feasible to perform in a clinical setting.

Two authors (A.D.G. and C.C.E.) independently assessed the methodological quality of included articles using criteria adapted from other authors in this series.⁴⁰ Level 1 evidence classifies articles that were independent (neither the test result nor the gold standard result was used to select patients for the study), studied consecutive patients representative of a population where the test is likely to be used, were blinded, measured the gold standard (BMD measurement or documented fracture) in all patients, and included at least 100 study participants. Level 2 evidence met criteria for level 1 evidence but fewer than 100 patients were studied. Level 3 evidence was the same as level 2 evidence but the population was nonconsecutive or nonrepresentative. Studies of lower levels of evidence were excluded. Disagreements were resolved by discussion and consensus.

We used raw data from reported studies that met our inclusion criteria to calculate values and 95% confidence intervals for sensitivity, specificity, and positive and negative likelihood ratios (LR+ and LR−) using SAS statistical software, version 8.0 (SAS Institute Inc, Cary, NC).

We identified 246 articles with our search strategy and an additional 79 from reference lists and expert consultation. Fourteen studies met inclusion criteria and were identified for final review (Table 2 and Table 3).

Table 4 lists reported precision estimates for the physical examination maneuvers. Interrater reliability was not reported for studies of height and weight included in this review. Differences in sensitivity and specificity for the same maneuver across different studies could be related to examiner differences that were not reported.

The most clinically relevant cut points and their associated LRs for the physical examination maneuvers are listed in Table 5 for osteoporosis and Table 6 for vertebral fracture. In general, the patient populations were women, with the majority of patients from osteoporosis clinics and/or older than 65 years. Translating these results to younger women might yield error that is difficult to quantify. Because many of the examination findings may be measuring similar or identical physiologic phenomena, we do not recommend using the LRs in series.

For postmenopausal women, prediction rules using osteoporosis risk factors, such as the Simple Calculated Osteoporosis Risk Estimation⁴¹ or the Osteoporosis Risk Assessment Instrument,⁴² have some predictive value in selected populations (Table 7).^{11 ,43 - 46} Variables included in these prediction rules include age, weight, and race, which overlap with the clinical examination. An exhaustive review of prediction rules for the diagnosis of osteoporosis or fracture was not attempted in this study because reviews already exist in the literature.⁴² While the positive LRs of the prediction rules are not clinically informative (1.2-1.7), the negative LRs are far superior to the physical examination maneuvers listed here (0.02-0.3), making prediction rules much more useful for ruling out osteoporosis or fracture. Thus, clinical prediction rules are the most useful means of identifying women who are at low risk of fracture, in whom BMD screening can safely be deferred.

Three studies of postmenopausal women using recalled heights found an association between height loss and vertebral fractures, with 2 of the studies including enough data to calculate LRs (Table 5).^{23 ,47 - 48} In the first study, a height loss of more than 3 cm was useful in classifying patients with and without low BMD (LR+, 3.2; LR–, 0.4).²³ However, the study population was nonconsecutive female patients with rheumatoid arthritis. In a study of women in the general population, Dargent-Molina et al⁴⁷ did not find a strong association between height loss of more than 3 cm and osteoporosis (LR+, 1.1; LR−, 0.6). The third study, based on 13 732 women in the Fracture Intervention Trial, reported that a self-reported height loss greater than 4 cm since age 25 years was associated with an odds ratio of 2.8 for vertebral fractures.⁴⁸ Thus, although height loss is a potentially useful examination tool, the generalizability of this measure is uncertain.

Versluis et al²¹ reported that with age, height declined at twice the rate of armspan. The mean difference in armspan and height was 1.4 cm in women aged 55 to 59 years and increased to 3.2 cm in women aged 80 to 84 years. Finding an armspan-height difference of 5 cm or greater yielded an LR+ of 1.6 and an LR− of 0.8 for spinal fracture based on these data (Table 6). Verhaar et al⁴⁹ reported that an armspan-height difference cutoff of 3 cm resulted in a sensitivity of 58% and a specificity of 56% for BMD-diagnosed osteoporosis, for an LR+ of 1.3. Wang et al⁵⁰ found no association between armspan and vertebral fractures in both men and women (LR+ for men, 1.0; LR+ for women, 0.9). We conclude that the armspan-height difference does not predict vertebral deformities or BMD-diagnosed osteoporosis.

For women, the relationship between both low weight and body mass index and osteoporosis has been consistently reported.⁴⁴ In cohort studies examining clinical risk factors in women, weight less than 70 kg (154 lb) is the single best predictor of low BMD^{11 ,45 - 46} and is an important variable in 4 of the 5 prediction rules reviewed here. Bedogni et al⁵¹ reported that body weight allowed a better classification of BMD than did body mass index, with women weighing less than 51 kg having a much greater risk for osteoporosis than women weighing more (LR+, 7.3; Table 5).

The cross-sectional survey by Michaelsson et al⁴⁴ demonstrated that body weight was the best predictor of BMD among measures of body size in women. In this study, women weighing less than 60 kg had a greater risk for osteoporosis than women who weighed more (LR+, 3.6). Women weighing 60 to 70 kg or more than 70 kg had a lower risk for osteoporosis (LR+, 0.3, and LR+, 0.2, respectively). Study limitations included a 20% participation rate and a low prevalence of osteoporosis.

Dargent-Molina et al⁴⁷ found current body weight to be the strongest predictor of very low bone mass (defined as a T score <−3.5 SDs). When BMD was measured in the 50% of women who weighed the least (<59 kg), the LR+ was 1.9 and the LR− was 0.3.

Although all of the studies that met our inclusion criteria were samples of women, other studies that were excluded because they reported only regression analysis data found similar associations between BMD at all sites and weight in both men and women, with weight having a similar impact in each sex.^{11 ,51 - 52}

Thus, body weight less than 59 kg appears to be a simple and reasonably sensitive but nonspecific measure for selecting women for further diagnostic testing. Heavier patients have a lower likelihood of osteoporosis. However, osteoporosis cannot be ruled out based on weight greater than 59 kg alone because of the broad range of LRs across the 3 studies (Table 5).

Flexicurve measurements in women were converted into kyphosis index values by Ettinger et al,²⁸ with the highest decile of kyphosis index used to classify patients as kyphotic. Ettinger et al reported that kyphosis was associated with reduced BMD and significant height loss. The presence of kyphosis was specific but not sensitive for osteoporosis (LR+, 3.1; LR−, 0.8), and it is not clear if the clinician’s simple observation of kyphosis without sophisticated measurements would yield the same result (Table 5).

Self-reported humped back was reported by Kantor et al⁵³ to be highly specific for hip osteoporosis in more than 2000 women referred for densitometry, with an LR+ of 3.0. The absence of self-reported humped back is not useful, however (LR−, 0.85; Table 5).

Siminoski et al³¹ reported in abstract form that a kyphosis angle greater than 43° or wall-occiput distance greater than 7 cm in women rules in a thoracic fracture with a high degree of accuracy and a kyphosis angle less than 20° or wall-occiput distance of 0 cm reduces the chance of thoracic fracture but does not reliably rule it out. The 0-cm cutoff seems most pragmatic, with an LR+ of 4.6 for thoracic fracture when a patient cannot place the back of her head to the wall (Figure and Table 6). In a sample size of 60 elderly women, however, Balzini et al⁵⁴ did not find a relationship between wall-occiput distance and vertebral fractures (data were not presented for calculating LRs).

Rib-pelvis distance of less than or equal to 2 fingerbreadths was calculated to have an LR+ of 3.8 and an LR− of 0.6 for detecting occult lumbar fractures (Table 6).³² Adjusting for patient height does not affect the operating characteristics of this test and is unnecessary. The positive LRs for vertebral fracture in a woman with 0 and 4 finger breadths of rib-pelvis distance are 11.5 and 0.1, respectively. Thus, a low rib-pelvis distance may increase the posttest probability of lumbar fracture to a level at which further testing is warranted.

Of the common measures of muscle strength, grip strength is most feasible to evaluate in the typical primary care clinic. Di Monaco et al³³ reported a positive association between grip strength and distal radius BMD in postmenopausal women in multiple regression analysis adjusted for age, years since menopause, years of ovarian activity, body height, body weight, body mass index, and calcium and alcohol dietary intake, with an LR+ of 1.5 (Table 5).

Foley et al³⁴ examined the relationship between hand grip strength and femur BMD with the goal of canceling out the effects of other anthropometric data and did not find a relationship between grip strength and proximal femur BMD for men. In women, it was thought that weight was related both to grip strength and femur BMD, with an LR+ of 1.3 for osteoporosis when a cutoff of less than 60 lb on the dynamometer was used.

Several other studies reported a positive association between grip strength and BMD, although reported data were not sufficient to calculate LRs.^{55 - 59} Overall, grip strength has insufficient sensitivity and inconsistent results for specificity.

Orme and Belchetz²⁰ studied the skinfold thickness in consecutive women in an osteoporosis clinic compared with normal, younger control women and reported odds ratios for a range of skinfold thickness of 1.5 to 2.1. These odds ratios corresponded to an LR+ of 1.2 and an LR− of 0.4 (Table 5). Although simple to perform, skinfold thickness does not appear to be useful in the diagnosis of osteoporosis.

Several studies have not shown a relationship between tooth loss and osteoporosis,^{60 - 63} but inclusion of younger patients may have limited their ability to detect an association.⁶⁴ It is not clear whether population studies reveal women with poor dental hygiene and tooth loss or tooth loss from osteoporosis.

Inagaki et al⁶⁴ reported that among postmenopausal women, the proportion of women with fewer than 20 teeth increased from 7% in the normal BMD group to 32% in the very low BMD group. The age-adjusted odds of having fewer than 20 teeth were significantly greater among women in the very low BMD group compared with the normal BMD group. The LR+ for having very low BMD if less than 20 teeth are counted is 3.4, but choosing a threshold of less than 22 teeth provides no additional clinical information (Table 5).⁶⁰

In a retrospective study, Astrom et al⁶⁵ found that elderly women with the least remaining teeth had twice the risk of hip fracture when compared with women with the most teeth. For men, the risk was more than 3-fold. Unfortunately, the cut-point number of teeth dividing the patients was not provided. May et al⁶⁶ found an association between self-reported tooth loss and BMD of the hip and spine using bone densitometry in older men that was independent of age, body mass index, and cigarette use. Other population-based studies reviewed demonstrated variable positive correlations between tooth counts and BMD.^{67 - 72} Overall, tooth counts are easy to do, and less than 20 teeth can reasonably lead the clinician to screen further for osteoporosis.

No single physical examination finding or combination of findings is sufficient to rule in osteoporosis or spinal fracture without further testing. The risk factor prediction rules for osteoporosis quoted in this article have more informative negative LRs than any of the physical findings and may reduce the need for testing in low-risk women. Several convenient examination maneuvers including low body weight (<51 kg), inability to place the back of the head against a wall when standing upright, low tooth count, self-reported humped back, and rib-pelvis distance can significantly increase the likelihood of osteoporosis or spinal fracture and identify additional women who would benefit from earlier screening ( Article ).

Wall-Occiput Distance
Inability to touch occiput to the wall when standing with back and heels to the wall

Weight
Less than 51 kg

Rib-Pelvis Distance
Less than 2 fingerbreadths between the inferior margin of the ribs and the superior surface of the pelvis in the midaxillary line

Tooth Count
Fewer than 20 teeth

Self-reported Humped Back
Patient report that back has become humped

Although the major osteoporosis clinical focus has been on women, the hip fracture incidence in 80-year-old men is similar to that in 75-year-old women.⁷³ A review of male osteoporosis suggests that the risk factors for men are the same (eg, BMD and body weight), although the level of risk is different from that for women.⁷⁴ Because osteoporosis develops at a later age in men, meaningful research is needed to determine whether the examination findings have similar properties in men or whether there is an age at which men should be screened for BMD similar to the recommendations for women.

The reluctant 64-year-old woman has a pretest probability of approximately 21.6% for osteoporosis at any site (Table 1). Her low weight (<51 kg) infers an LR of 7.3, thus increasing her posttest probability of osteoporosis to 67%. She decides that this level of risk makes the drive to the testing center worthwhile.

The prevalence of grade 2 or grade 3 vertebral deformities in women aged 75 to 79 years is approximately 29%.⁷⁵ The LR+ of a positive wall-occiput maneuver is 4.6, resulting in a posttest probability of 65%. This 78-year-old patient is very likely to have vertebral fractures. If spine films confirm the presence of vertebral fractures, then she should be considered for osteoporosis therapy. Bone mineral density testing to confirm osteoporosis is not required but may help guide therapeutic decisions.

Although the 58-year-old woman does not meet current screening guidelines for dual-energy x-ray absorptiometry, the low rib-pelvis distance detected on her physical examination increases the probability that she already has occult vertebral fracture from 3.4% to 12%.⁷⁵ Her self-reported humped back increases the probability that she has osteoporosis from 14.8% to 37%, prompting early assessment of her bone density.