Medicine For People!

March 2010

Coming Months

Bone Health: Understanding the DEXA Scan

Mary came to my office because she wasn't convinced she needed her prescription. A healthy-looking 55-year-old women, she laid a copy of her bone density report on my desk, and asked, "what am I gong to do?" According to her previous doctor, the test said she had osteopenia, a thinning of the bones that is the precursor to osteoporosis. He wanted to start her on alendronate (Fosamax). The report also said that she had poor bone density only in the spine, not the hip. Why were they different? Did she have osteopenia or not? Mary's instinct was that she did not need the drug. She felt that, with all the hiking and dancing she did, her bones must be strong. But she was worried. Was she going to end up crippled with a broken hip or bent spine?

Answering such questions isn't easy. I've put people on treatment for osteoporosis only to see their bone density report become worse. I've seen women who've undertaken no treatment have follow-up scans showing bone density going up and down with no apparent reason. And I know that bones are much more than just hunks of calcium. They are alive.

This newsletter is the second in a series that will help answer Mary's questions and also, hopefully, any of yours. This month we tackle the questions:

  • How does a DEXA scan work?
  • How do they determine what is normal and what isn't?
  • How accurate is the scanner?

We start with the spine, because bone strength varies from bone to bone. For this reason, a scan of the wrist or heel, such as is offered in commercial community screenings, tells you little about the areas of concern, your hip and spine. I recommend against such scans.

Before we begin, you may wish to look back at our previous newsletter illustrating the dynamic nature of your living skeleton.

The DEXA Scan and How it Works

DEXA stands for Dual Energy X-ray Absorptiometry, and is our standard test for bone strength. DEXA scanners throw an x-ray beam at the lumbar vertebrae and the hip and measure the shadow cast by the bones. Software in the machine estimates the amount of calcium in the bone based on the darkness of the shadow. The result is expressed as a number of grams per square centimeter. DEXA scans are not three-dimensional like CT scans, but two-dimensional, like a plain X-ray. They produce an estimate, which can be erroneous in several ways.

DEXA Results Vary With Body Size

Put two different women in front of the DEXA beam. One is 5'10 and she has nice, big vertebrae. Her friend is 5'2", with small vertebrae, but both women have the exact same concentration of calcium in their bones, the same number of grams per cubic centimeter. They have equal strength in their bones, and all else being equal, an equal risk of fracture. To a DEXA scanner, they look quite different. The X-ray beam must travel further to get through the vertebrae of the large woman than those of her smaller friend.

bone health

bone health

Source: Br J Radiol. 1998 Sep;71(849):934-43. 1

The smaller bones throw a lighter shadow, so the machine concludes that the petite woman's bones contain less calcium. It's pretty easy for the machine to err, because it does not measure volume, but only the greyness of a shadow.

Bone Density Varies in Healthy Women

A Danish study found that healthy premenopausal woman can have a bone density varying by as much as 9 percent in the forearm, 6 percent in the spine, and around 7 percent for the hip.2

Different Brands of DEXA Scanners Produce Different Results

A DEXA scanner has a tough job. The people coming through the scanner, and their bones, come in various sizes and shapes. The scanner has to maintain an artificial reference point of the healthy twenty-year old, and compare each actual person with this internal reference. If you want to buy a scanner and check out its workings so you can make an informed decision, you can't. The manufacturers keep their algorithms secret.

So it should come as no surprise that scanners from different manufacturers produce results that vary. When I say "vary," I don't mean by some inconsequential amount. Some researchers claim that bone density readings can vary as much as 20 percent from one model scanner to another.3

If you find that hard to believe, the details are even less heartening. One way researchers calibrate DEXA scanners is to scan cadavers. They remove the vertebral body and measure the volume, then use high heat to burn off everything in the bone except the minerals. By measuring the weight of the ash, they find the mineral content of that bone. Using this technique, a newer generation DEXA scanner underestimated spinal density by up to 33 percent.4 5

How to Read a DEXA Graph

Now you are ready to look at Mary's DEXA report. First, let's look at the diagram that plots her estimated bone density against her age and compares this with so-called "normal" bone density.

bone health

This graph plots Bone Mineral Density (BMD), going up the left edge against age along the bottom. The blue band across the graph shows the range of normal BMD across the span of a lifetime. You can see that bone density is highest between ages 20 to 45, then decreases. The little white square stands for Mary's estimated BMD (1.036) at her current age (55). The DEXA scanner puts the box at the lower end of the normal blue zone for her age.

How to Read The DEXA Report

The clinical report below compares Mary to both young women and women of her own age.

bone health

The Z Score: The written report is what doctors are trained to look at. This DEXA scan report says that Mary's lumbar spine density compared to women her age is 0.7 standard deviations below average. A standard deviation is unit of measure for how much something deviates from the norm. The number 0 stands for the norm, so -0.7 indicates the shadow of Mary's bones is a little bit lighter than the norm. (Wikipedia has a very clear explanation of standard deviation at .)

The T-Score: This number compares Mary's bone density with a twenty-year old female and shows that she is 1.4 standard deviations below mean. Most younger women have denser bones than Mary.

The assumption behind this comparison is that if, somehow, without endangering Mary, we could keep her bones as strong as a twenty-year-old's, that she would be healthier and less likely to have a fracture. I disagree with this assumption. I don't believe it makes sense to use a 20-year-old woman as a standard against which to measure a 55-year-old woman.

Impression: Finally, the scanner is programmed to define this bone density as osteopenia, a deficiency of bone. The scanner then looks in its table of risk and tells Mary that she is 2.6 times as likely as other women her age to suffer a compression fracture of a vertebral body. What does that actually mean for Mary? In fact very few healthy women her age get spinal fractures. The bottom line is that 2.6 times a tiny number is still a tiny number.

In reality, the impression is what most women care about. Few of them wish to know how many milligrams of calcium reside in each cubic centimeter of the bones. They want to know whether they are at risk for fractures. What matters is how well this test predicts future fractures.

Vertebral Compression Fracture

A vertebral compression fracture is a cracking or "squishing" of the vertebral body caused by pressure. People with weakened bones are susceptible to these injuries. This is the fracture we'd like to prevent if we can. Such a fracture can occur anywhere in the spine, but most often in the thoracic or lumbar area. Most compression fractures cause no symptoms so go undetected without X-ray. Other times these fractures can cause significant pain for a month or two. We tend to lose height as we age from drying and shrinking of the inter-vertebral disks, and the addition of a few compression fractures can turn this into significant loss of height. Surgery is available, but in my opinion carries too much risk of worsening the problem.

vertebral compression fracture

Does DEXA Score match reality?

Let's look at a Canadian survey6 of actual women and see if Mary's supposed risk seems a reasonable estimate.

bone health

bone health

Source: CMAJ. 2007 Sep 11;177(6):575-80

In the graph above, 16,500 Manitoba women with an average age of 65 are separated by T-score. Their actual fracture count over a three year period is indicated by the rising black line and the scale on the right. The white arrow points to Mary's score. On the average, women with ages and scores near hers suffered about 1 fracture per 1000 women per year. (About a sixth of those fractures occurred in the women under 65, about 5/6th occurred in women over the age of 65). Mary's report stated that her risk was 2.6 times higher than it would be if she had a normal T-score (the black arrow above). Friends, the black line in the graph does not rise 2.6 times as high between the two arrows. The actual fracture rate among those 16,500 Canadian women was similar whether their T-score was a normal -1.0 or an osteopenic -1.4.

Why does the DEXA scanner give a result that doesn't jibe with actual real-world knowledge? I can't tell you. The software details are a commercial secret.

Who Benefits from a Yearly Scan?

Notice that the DEXA scanner is programmed to suggest another scan in a year. This will not benefit Mary. The error of the scan is greater than any meaningful change that will occur in her bones in a year. Most doctors, even when concerned, repeat the test no sooner than every two years.

Factors to Consider Beside the Scan

I asked Mary whether her mother had lost height. Mary replied that her mother had lived to 81, and after 75 seemed to be a bit shorter, but maintained normal posture and an active life. Had her mother ever complained of pain in the back? Mary didn't think so. I was looking for evidence that her mother had suffered a significant number of spinal compression fractures. Had Mary's mother a significant height loss, or if Mary smoked or had generally poor health, we'd have reason to worry. Given the story behind the DEXA scan above, Mary decided not to take alendronate. She decided to address her issues of bone health using the foundational knowledge from last month's newsletter, using methods we'll outline in the coming months.

Coming Months

  • Aging Bones
    • We will review Mary's DEXA report on her hip, and unearth more limitations of this common and poorly understood test.
    • You will learn more about hip fracture, allowing you to replace unrealistic fear with accurate knowledge.
    • We will give you tools to estimate your risk of osteoporotic fracture.
    • We will discuss measures, natural as well as pharmaceutical, to strengthen bone and reduce fracture.
  • Multiple vitamins
  • Colon cancer screening


1 Diagnosis of osteoporosis by planar bone densitometry: can body size be disregarded? Pors Nielsen S et al. Br J Radiol. 1998 Sep;71(849):934-43. Department of Clinical Physiology, Hillerød Hospital, Denmark.

Bone densitometry using dual energy X-ray absorptiometry (DXA) is frequently used to diagnose osteoporosis and to identify patients at risk of later fractures. The parameters of interest are bone mineral content (BMC) and bone mineral areal density (BMD). Bone densitometry results have a large overlap between normals and patient with fractures. This would suggest that other factors are important for the development of fractures or that bone densitometry is not used optimally. It is generally believed that the conversion of BMC to BMD by division of the former by the projected bone area is a good normalization procedure. Other normalization procedures have been attempted in the past with little success. We hypothesized that this might be due to a blurring effect of time since menopause, and that body size could be demonstrated to have an effect on measured BMC and BMD, if this time effect could be eliminated. The results of this study, comprising 1625 early post-menopausal women studied at virtually the same time since menopause, confirm that this is the case. Body surface area was the parameter among conventional body size variables showing the highest correlation with BMC and BMD. It was clearly shown that low values of BMD were seen more often in the lowest than in the highest body surface area quartile. The difference between quartiles was statistically significant. Simple division of BMC by actual body surface area or division of BMD by the square root of body surface removed the uneven distribution between the body surface area quartiles for lumbar spine and femoral neck measurements, and reduced it at peripheral measuring sites. It is suggested that BMC and BMD of the lumbar spine and the femoral neck should be normalized as described to avoid overdiagnosis of osteoporosis in persons of petite body stature and underdiagnosis in tall ones.

2 The fallacy of BMD: a critical review of the diagnostic use of dual X-ray absorptiometry. Nielsen SP. Clin Rheumatol. 2000;19(3):174-83. Department of Clinical Physiology and Nuclear Medicine, Hillerød Hospital, Denmark.

Abstract: The diagnostic use of BMD should be cautious as BMD is not an ideal measure of true bone density; it is not an ideal measure of bone strength; it does not predict fractures well; and it has inherent problems of accuracy and linearity. The limitations of BMD, based on the physical deficiencies of DXA, are further obscured by the introduction of T-scores. It is suggested that BMD and BMC, when used diagnostically and for fracture risk classification, be used after correction for body size and/or bone size, age and sex, and that measured values be evaluated in the light of established mean fracture incidence data. BMD is not a parameter of sufficient validity to be the sole indicator of present and future fracture risk. A low BMD should be regarded one of several fracture risk factors. It seems that there is a need to redefine the T-score based definition of osteoporosis.

3 The fallacy of BMD: a critical review of the diagnostic use of dual X-ray absorptiometry. Nielsen SP. Clin Rheumatol. 2000;19(3):174-83. Abstract above.

4 The accuracy of volumetric bone density measurements in dual x-ray absorptiometry. Sabin MA et al. Calcif Tissue Int. 1995 Mar;56(3):210-4. Division of Anatomy and Cell Biology, UMDS (Guy's Campus) London, UK.

New developments in dual x-ray absorptiometry (DXA) allow the performance of high precision anteroposterior (AP) and lateral scans of spinal bone mineral density (BMD, units: g/cm2) without the patient moving from the supine position. Data from both projections may be combined to give an estimate of the true volumetric bone mineral density (VBMD, units: g/cm3) of the lumbar vertebral bodies. This report presents a cadaver study designed to validate DXA measurements of volumetric bone density. Sections of whole lumbar spine were scanned in AP and lateral projections in a water tank to simulate soft tissue. Individual vertebrae were then divided to separate the vertebral body from the neural arch, and vertebral body volume was measured using the displacement of sand. The bone mineral content (BMC) of vertebral bodies and neural arches was measured by ashing at 250 degrees C for 60 hours followed by 500 degrees C for a further 24 hours. The results showed that DXA scanning systematically underestimated ashing data by 14% for AP BMC, 33% for vertebral body BMC, 23% for vertebral body volume, and 12% for VBMD. Despite these significant systematic errors, the DXA measurements and ashing values were highly correlated (r = 0.979-0.992). The results suggested that after allowing for the systematic errors, lateral DXA parameters related closely to true BMC, volume, and VBMD. [ Note: this says that newer technology, not generally available, does measure a true volumetric density, yet nonetheless underestimates by up to 33%. Correction factors, not used with current machines, reduced the errors significantly. Current machines do not use lateral measurements. ]

5 Linearity and accuracy errors in bone densitometry. Pors Nielsen S et al. Br J Radiol. 1998 Oct;71(850):1062-8. Hillerød Hospital, Department of Clinical Physiology, Denmark.

This investigation was undertaken to quantify accuracy errors and identify possible linearity errors in dual energy X-ray absorptiometry (DXA) of bone, based on studies of commercially available bone densitometers for planar densitometry. The following was found in a combination of in vitro phantom studies and in vivo investigations of human volunteers: (1) Pronounced differences between the instruments when measuring vertebral size and contours of the projected bone regions. (2) Falsely low bone mineral content (BMC in terms of g) in cases of low nominal bone mass, due to the fact that edge regions were omitted by the calculation software of some devices. (3) An increase in the projected bone area secondary to an increase in nominal bone mass with some instruments. (4) Clinically and statistically significant errors of accuracy of BMC and to a lesser extent bone mineral density (BMD). (5) Substantial linearity errors with some osteodensitometers for BMC, a phenomenon that reduces the usefulness of this parameter. It is concluded that DXA devices are affected by a combination of accuracy errors and linearity errors, some more than others, and that linearity errors influence their ability to monitor change in BMC and to a lesser extent in BMD, making system intercomparison difficult.

6 Low bone mineral density and fracture burden in postmenopausal women. Cranney A et al. CMAJ. 2007 Sep 11;177(6):575-80. Clinical Epidemiology Program, Ottawa Health Research Institute, Ottawa, Ont.

BACKGROUND: The study objectives were to determine fracture rates in relation to bone mineral density at various central skeletal sites, using the World Health Organization definition for osteoporosis (T-score -2.5 or less), and to contrast fracture patterns among women 50 to 64 years of age with those among women 65 years of age and older. METHODS: Historical cohort study with a mean observation period of 3.2 (standard deviation [SD] 1.5) years. The study group (16,505 women 50 years of age or older) was drawn from the Manitoba Bone Density Program database, which includes all bone mineral density results for Manitoba. Baseline density measurements for the lumbar spine and hip were performed with dual-energy x-ray absorptiometry. Outcomes included the percentage of osteoporotic fractures and the rates of fracture and excess fracture (per 1000 person-years) among postmenopausal women with osteopenia and osteoporosis relative to those with normal bone mineral density (according to the classification of the World Health Organization). RESULTS: The mean age was 65 (SD 9) years, and the mean T-scores for all sites fell within the osteopenic category. There were 765 incident fractures (fracture rate 14.5 [95% confidence interval, CI, 13.5-15.6 [per 1000 person-years). Fracture rates were significantly higher among women 65 years of age or older than among women 50-64 years of age (21.6 [95% CI 19.7-23.4] v. 8.6 [95% CI 7.5-9.7] per 1000 person-years, p < 0.001). Although fracture rates were significantly higher among women with osteoporotic T-scores, most fractures occurred in women with nonosteoporotic values (min-max: 59.7%-67.8%). INTERPRETATION: In this study, most of the postmenopausal women with osteoporotic fractures had nonosteoporotic bone mineral density values. This finding highlights the importance of considering key clinical risk factors that operate independently of bone mineral density (such as age) when assessing fracture risk.


Medicine for People! is published by Douwe Rienstra, MD at Port Townsend, Washington.