Osteoporosis and The Assessment of Bone Mass in Men
Although the majority of osteoporotic fractures occur among women, fractures among elderly men are also quite common and will become increas-ingly frequent as life expectancy increases. In fact, one study estimated that 29% of men and 56% of women will experience fractures during their remaining lifetime, if they are currently 60 years old and receive no preventive measures (Jones et al., 1994). With the advent of nonhormonal therapies for prevention and treatment of osteoporosis, it is now appropriate to use BMD for evaluating fracture risk among men to assist with patient management decisions.
Bone mineral density (BMD) is widely used to identify which female patients should be given therapy for prevention or treatment of osteoporosis and also to monitor the efficacy of treatment. However, many physicians are uncertain about how to interpret BMD in men. This post summarizes the clinical application of BMD measurements, including a review of measurement techniques, comparison of age-related bone loss patterns in men and women, and discussion of the association between low BMD and increased fracture risk.
II. Clinical Interpretation of BMD
Maximum (peak) levels of BMD occur around age 30, or sooner; although this is fairly well-established for women, there are fewer data for men. Elderly men and women have low BMD, and a high risk of fractures. Therefore, it is not very useful to compare current BMD measurements for older people to average BMD values among people of similar age. Instead, the results are often expressed relative to peak BMD, as T scores. The T-score represents the number of standard deviations (SD) above or below the mean for young healthy people; positive numbers represent higher values, and negative numbers represent BMD below the average for young people.
The World Health Organization (WHO) has developed criteria for inter-preting BMD which are used widely (Kanis et al., 1994). In this system, patients with BMD that is at least 2.5 SD below the young adult mean (T-score < —2.5) have osteoporosis, and those with BMD between 1 and 2.5 SD below the young adult mean ( — 2.5 < T-score < —1.0) are classified as having low bone mass (or osteopenia). The goal of therapy when treating osteoporosis (patients with T-scores of —2.5 or less) is to increase BMD, or at least maintain current BMD levels. For patients with low bone mass (T-scores between — 1.0 and -2.5), the goal is to prevent further declines in BMD, or to at least slow bone loss considerably.
The standardization and interpretation of BMD measurements has been complicated by the variety of skeletal sites that can be measured and by differences in calibration between manufacturers. Nevertheless, as will be shown later, most measurements are able to predict spine and nonspine fracture risk to a similar extent. One exception is hip BMD, which is somewhat better than other measurements for predicting hip fractures. However, hip fractures account for less than 10% of all nonviolent fractures among the elderly, and certain other fracture types, such as wrist and spine, often occur prior to hip fractures. Thus, most measurements are probably suitable for basing treatment decisions. The spine and hip (especially the trochanter) appear to be superior for measuring treatment response; however, the need for such monitoring is still controversial.
Perhaps the best hip BMD reference data for the United States were derived from the third National Health and Nutrition Examination Survey (NHANES III) (Looker et al., 1995). Although the WHO BMD cutoffs for defining osteoporosis and osteopenia based on the female and male reference ranges differ only slightly (Table I), the estimated prevalence of male osteopenia is twice as high if the male cutoff is used instead of the female cutoff (Looker et al., 1997). However, because the relative distributions of BMD are different in men and women, the number of men with osteoporosis is essentially the same regardless of whether male or female cutoffs are used (Looker et al., 1997). Using the male cutoffs, approximately 3-6% (1-2 million) of U.S. men 50 years and older are estimated to have osteoporosis and 28-47% (8-13 million) to have osteopenia. For comparison, 13-18% (4-6 million) and 37-50% (13-17 million) women ages 50 and older have osteoporosis and osteopenia, respectively (Looker et al., 1997). A European study yielded similar estimates; the prevalence of osteoporosis was 6% in men and 23% for women, ages 50 years and older (Kanis et al., 1994). The greater peak BMD, slower rate of loss, and shorter life expectancy in men probably contribute to the lower prevalence of osteoporosis, relative to women.
TABLE I Mean Femoral BMD of 20- to 29-Year-Old Non-Hispanic White Men and Women, and Cutoff Values for Osteopenia and Osteoporosis, Using WHO Definitions
Standard deviation (g/cm1)
III. Techniques for Measuring Bone Mineral Density
There are numerous technologies for measuring BMD; all are applicable to both men and women (Table II). In addition, quantitative ultrasound (QUS) provides an estimate of bone density and may also provide some information about bone structure or other aspects of bone quality. These methods for measuring bone mass are relatively safe, with low radiation exposures compared to standard x-ray techniques (there is no ionizing radiation for QUS). For example, a typical DXA scan represents less than 0.1% of the annual radiation exposure from natural sources (Rizzoli etal., 1995).
TABLE II Characteristics of Methods for Evaluating Bone Density and Quantitative Ultrasound
Examination duration (min)
Single-photon (isotope) absorptiometry
Single-energy x-ray absorptiometry
Dual-photon (isotope) absorptiometry
Dual-energy x-ray absorptiometry
Quantitative computed tomography
Peripheral quantitative computed tomography
Characteristics of these techniques are summarized in Table II and discussed in greater depth later.
Bone mass measured with absorptiometry is usually reported as BMD (g/cm2), which is the measured bone mass in grams divided by the area of the bone region of interest; bone mineral content (BMC) is sometimes reported in units of grams, or as grams per centimeter. The QCT technique provides a measure of true volumetric density (g/cm3), whereas rectilinear techniques do not (because the bone is scanned in only two dimensions). Nevertheless, the term bone density is often used when referring to either grams per cubic centimeter or grams per square centimeter measurements.
Measurements such as the spine, hip, and wrist represent skeletal sites where osteoporotic fractures commonly occur; a variety of other sites can also be measured, depending on the technique and equipment manufacturer. Most measurement sites, but not all, contain a substantial proportion of trabecular bone. The rate of turnover in trabecular bone is usually greater than that in cortical bone; as a result, disturbances in bone balance generally produce larger changes in predominantly trabecular sites.
The procedures to acquire and analyze bone scans are relatively simple. Many densitometers display an image of the bone, and indicate the Region of Interest (ROI) within which the bone mass is measured; a small amount of operator judgment and manual intervention is required. It is very important that the manufacturer’s instructions be followed carefully when choosing or changing ROIs. This is especially important when monitoring changes over time, because subtle differences in ROI placement can alter the results by several percent, or more, which is comparable in magnitude to the average BMD changes over 1 to 2 years. It is also important to watch for artifacts which may cause measurement errors and to perform quality control procedures daily to verify proper calibration and watch for measurement drifts. For these reasons, formal training or certification of operators is recommended.
A. Dual-Photon Absorptiometry and Dual-Energy X-Ray Absorptiometry
Dual-energy x-ray absorptiometry (DXA) is considered by some to be the current “gold standard” for measurement of BMD. It replaces an older, similar method, dual-photon absorptiometry (DPA), in which the photon source of the dual-energy beam is gamma radiation from gadolinium (mGd); DXA thus obviates the expense and difficulty of working with radioactive isotopes (Genant et al., 1993; Gliier et al., 1990). The largest manufacturers worldwide are Hologic (Waltham, Massachusetts, USA) and Lunar (Madison, Wisconsin, USA), with fewer instruments being produced by Norland (Ft. Atkinson, Wisconsin, USA) and Sopha (Buc, France). Although these machines have some different features, they are very similar in their basic
principles of operation. To some extent, normal ranges differ between man-ufacturers (Arai et al., 1990; Pocock et al., 1992). For example, spine BMD measured with Lunar densitometers is approximately 16% higher than that using machines from other manufacturers.
I. Spine BMD
A commonly measured BMD site is the postero-anterior (PA) lumbar spine. Increases in bone density during treatment with antiresorptive agents are generally greatest at the spine. The precision of spine BMD measurements using DXA is excellent, with a coefficient of variation (CV) of approximately 1-2% (Lees and Stevenson, 1992). However, a number of factors including osteoarthrosis, osteosclerosis, and aortic calcification can introduce significant measurement errors causing spine BMD to be overestimated; these sources of interference increase with age and are common after age 65 (Genant etal., 1996; Dawson-Hughes and Dallal, 1990).
Lateral spine BMD measurements have been developed to exclude the posterior spinous elements (which consist predominantly of cortical bone), thereby limiting measurements to the vertebral bodies, which contain a higher component of trabecular bone. The supine position is preferred because of its superior reproducibility compared to the earlier decubitis position (Genant et al., 1996; Mazess et al., 1991). However, lateral spine BMD is limited by overlapping ribs and pelvis in many patients.
2. Hip BMD
Treatment effects are generally smaller at the hip than the spine, possibly because the proportion of cortical bone is higher for the hip. Also, the precision of BMD measurements at the hip (CV of approximately 2%) is not as good as for the spine, which has a CV about 1% (Genant et al., 1996; Wilson et al., 1991). The hip is subdivided into several regions of interest: femoral neck, trochanter, intertrochanteric area, and Ward’s triangle. Total hip BMD (combined femoral neck, trochanter, and intertrochanteric areas) is also available on some machines. Ward’s triangle is a small (approximately 1 cm2) subregion located near the base of the femoral neck; the precision of this measurement is not as good as the other hip sites. As noted earlier, the ability to predict fracture risk is similar for most sites, but the trochanter may be better than other hip sites with regard to measuring changes over time.
3. Total Body BMD
Measurement of total body BMD is usually confined to research. As the name indicates, it measures BMD of the entire skeleton. For example, one research application has been to confirm that increases in BMD at a specific site (e.g., spine) during antiresorptive treatment are not simply caused by a redistribution of bone mineral from one region to another within the skeleton but that increases occur throughout the skeleton (Hosking et al., 1998). Total body BMD has better precision (short-term in vivo CV of about 0.5-1%) than most other sites (Genant et al., 1996). However, the skeleton consists predominantly (>80%) of cortical bone (Parfitt, 1980). Because of the low rate of bone turnover at cortical sites (approximately 3% per year compared to 20-30% per year in trabecular bone), smaller changes in BMD are typically seen for total body BMD, compared to the spine (Parfitt, 1980).
4. Peripheral BMD
Heel (calcaneus) and distal (wrist) and mid-forearm bone density mea-surements using SPA and SXA have been used in many studies during the past 20 years. The forearm represents nonweight-bearing bone, in contrast to the heel, spine, and hip. The one-third forearm site is located at one-third the ulnar length, from the wrist, and the mid-forearm is located halfway; both are predominantly cortical bone, whereas ultradistal measurements (at the distal end of the radius) contain a substantial proportion of trabecular bone. Total forearm measurements are also available on some machines, representing a larger region from the wrist to the middle of the forearm. Forearm BMD measurements sometimes include both the radius and ulna, or only the radius. The heel and forearm sites are easily accessible and have little covering soft tissue so that single- (rather than dual-) energy absorptiometry is adequate and provides precise estimates of bone density (Weinstein et al., 1991). Precision (CV) is generally around 1-2%. Response to treatment is often small at peripheral sites, especially in the mid- or one-third radius sites, which are almost completely (>95%) cortical bone (Schlenker and VonSeggen, 1976) . The heel is predominantly trabecular bone; as with the spine and hip, large changes are observed during bed rest and disuse.
5. Radiographic Absorptiometry
Radiographic absorptiometry (RA) uses a radiographic film image, usually of the hand or fingers, to measure bone mass by comparing the optical density of the region of interest (bone) to a calibration standard (such as an aluminum wedge), which is included in the exposure field (Yates et al., 1995; Ross, 1997). Some RA methods use existing radiographic facilities to obtain images on film, which are then analyzed on-site using an optical analyzer, or mailed to a central lab. Other methods acquire the bone image digitally, and provide density results within minutes without any film. Typical measurement sites are the phalanges (mixed cortical and trabecular regions), metacarpal (cortical), and radius (cortical). The precision (CV) ranges from good (2%) to excellent (1 %).
6. Single-Energy X-Ray Absorptiometry
Like RA, single-energy x-ray absorptiometry (SXA) offers the potential of cheaper and more accessible bone mass measurement relative to DXA. Also, like RA, SXA is limited to peripheral sites, such as the forearm and heel, because it generally requires immersion in water to provide a uniform soft-tissue equivalent. Advantages of machines dedicated to peripheral mea-surements (either DXA or SXA) are their smaller size and cost, compared to large DXA machines required for measuring spine and hip.
7. Quantitative Computed Tomography
Quantitative computed tomography (QCT) can provide information on vertebral body bone density measurement that is similar to DXA, or it can be used to measure true volumetric (g/cc) trabecular bone density inside the vertebral body (excluding the cortical shell). For example, this technique has been used in research to investigate the relative contributions of trabecular bone versus cortical bone to the overall strength of the vertebral body. Peripheral QCT (pQCT) has also been used to measure trabecular and cortical volumetric (g/cc) bone density at peripheral sites, such as the radius and tibia.
8. Quantitative Ultrasound
Quantitative bone ultrasound measurements correlate with bone density measurements and may also measure aspects of bone quality which are in-dependent of bone density (Faulkner et al., 1994). Heel ultrasound appears to be equivalent to measurements of BMD for predicting the risk of spine and nonspine fractures (Huang et al., 1999; Ross et al., 1995; Gregg et al., 1997) . There is very little data on the usefulness of ultrasound for monitoring treatment effects or other changes over time. The precision of most QUS techniques are not as good as the better BMD methods, but recent improvements have put the CV for some QUS machines in the range of 1-2%. As with other peripheral measurements, the small size and low cost are attractive features.
IV. Age-Related Changes in Bone Mass
Both cortical and trabecular bone mass decline with age in men and women. We recognize that (with the exception of QCT) densitometry techniques are not able to measure either cortical or trabecular bone exclusively, but have summarized cortical and trabecular bone separately to the extent possible, because there appear to be some differences related to these two types of bone, such as the rate of change, and response to treatment.
A. Cortical Bone
Men have approximately 7-10% (about 0.5-0.7 T-score) (Looker etal., 1995, 1997) higher femoral bone density in early adulthood (ages 20-29) compared to women of the same race (Table I). Some studies have reported that cortical bone mass begins to decline in the third decade of life and that this decline continues at a relatively steady rate throughout the remainder of life in both men and women (Looker et al., 1995; Hannan et al., 1992; Steiger et al., 1992; Davis et al., 1991; Gotfredsen et al., 1987). However, most of those studies were cross-sectional, and other studies (including prospective studies) have indicated that bone loss at cortical sites is faster soon after menopause in women and may accelerate after about age 60 or 70 in both men and women (Ensrud et al., 1995; Wishart et al., 1995; Jones et ai, 1994; Tobin et al., 1993; Garn et al., 1992; Orwoll et al., 1990; Mazess etal., 1990; Blunt et al., 1994; Orwoll and Klein, 1995; Riggs et al., 1981; Davis etal., 1991).
Compared to women, most studies have reported that the rate of cortical bone loss is slower in men, leading to progressively larger gender differences in cortical bone mass with age (Davis et al., 1991; Blunt etal., 1994; Orwoll and Klein, 1995; Riggs et al., 1981; Kalender et al., 1989). However, other studies (longitudinal and cross-sectional) found little difference between men and women in the pattern of decline in femoral bone density after age 60 (Hannan et al., 1992; Jones et al., 1994). Furthermore, data from more recent, longitudinal studies have consistently shown that the rate of cortical bone loss in men may be considerably more rapid (0.5-1% per year) (Davis et al., 1991; Jones et al., 1994; Tobin et al., 1993; Orwoll et al., 1990; Slemenda et al., 1992) than previously estimated from earlier, cross-sectional studies (0.1-0.3% per year) (Hannan etal., 1992; Mazess et al., 1990; Blunt et al., 1994; Orwoll and Klein, 1995; Riggs et al., 1981; Kalender et al., 1989). Cross-sectional studies may not yield accurate estimates of bone loss rates because peak bone density may have been different for older generations than later generations; thus, the longitudinal data are probably more reliable. Among Caucasians, femoral BMD is approximately 25% lower among men and 33% lower among women at ages 80-85, compared to peak BMD for men and women at ages 20-29, respectively (Looker etal., 1995).
A number of structural factors influence the ability of long bones to withstand forces, including the length, cross-sectional shape, and the distribution of mineral relative to the applied force (Melton et al., 1988; Mosekilde and Mosekilde, 1990). The combination of smaller declines in bone mass and larger bone size in men may explain to a large extent the lower risk of fractures in men, which is typically half that compared to women of similar age.
B. Trabecular Bone
Spinal trabecular bone density (g/cm!) measured by QCT is similar for young men and women. Spinal areal BMD (g/cm2) measured by DXA is greater for men than women, partly because this measure does not fully account for the larger bone size of men, and possibly also because it includes a substantial amount of cortical bone (Orwoll and Klein, 1996; Genant et al., 1994). Vertebral bone density declines with age, possibly beginning as early as the third decade of life, both in men and in women (Orwoll et al., 1990; Riggs et al., 1981; Meier etal., 1984). Trabecular connectivity and the number and thickness of trabecular struts are decreased among older men, similar to postmenopausal women (Aaron et al., 1987; Mosekilde, 1989). These histologic findings, coupled with evidence of increased bone turnover, suggest that the pathophysiologic mechanism underlying age-related osteoporosis in men is similar to that in postmenopausal women (Orwoll and Klein, 1995).
Some cross-sectional studies reported that spine BMD did not decline after age 70 (or after age 50 in one study) among men (Mazess et al., 1990; Blunt et al., 1994). This is most likely an artifact caused in part by selection bias (men with poor health at older ages may not participate) and by overestimation of BMD at older ages related to osteoarthritis and other degenerative changes (Dawson-Hughes and Dallal, 1990). For example, a cross-sectional study of postmenopausal women found that the rates of BMD decreases with age were similar at almost all sites examined (including predominantly cortical sites such as the metacarpal and mid-radius and predominantly trabecular sites such as the calcaneus), with the exception of the spine, which exhibited the smallest difference with age (Ross et al., 1995). Moreover, cross-sectional studies using QCT measurements of the spine, which selectively measure trabecular bone, reported that spinal trabecular bone density declines with age among men (Meier et al., 1984; Cann and Genant, 1982). Progressive declines with age were also demonstrated among men with longitudinal measurements of predominantly trabecular calcaneus BMD; in fact, the rate of decline appeared to increase after age 75 (Davis etal., 1991).
The age-related reduction in vertebral BMD measured by DXA appears to be greater in women than men. This may be partly because DXA measures both cortical and trabecular bone, and cortical bone loss is more rapid in women. One study reported that spine BMD decreased by 14% in men and 47% in women from youth to old age (Riggs etal., 1981). However, as noted earlier, measurement errors and selection bias may have caused these data to underestimate the true declines. When measured by QCT, trabecular bone loss at the spine was similar in men and women (minus 37 and 48%, respectively), corresponding to a yearly rate of 0.72% per year decline for men (Genant et al., 1988). A different study using QCT found even greater lifetime decreases of more than 50%, or 1.2% per year (Meier et al., 1984). Although single-energy QCT may overestimate the reduction in spine BMD (Seeman, 1993), the finding of similar trabecular spinal bone loss in men and women measured by QCT is also consistent with the available histomorphometric data (Mosekilde and Mosekilde, 1990).
As with the spine, the age-related decrease in iliac crest trabecular volume is only marginally greater in women than in men, as measured by histomorphometric data (Aaron et al., 1987; Parfitt et al., 1983) and by QCT measurements (Kalender et al., 1989; Meier et al., 1984). Also, in the Framingham study cohort, the rate of bone loss at the largely trabecular femoral trochanter site was only slightly slower among men, compared to women ( — 0.45% per year and —0.53% per year, respectively) (Hannan etal., 1992). In contrast, rates of bone loss at the trabecular calcaneus appear to be approximately twice as great among women as men (Davis et al., 1991; Cheng et al., 1997).
In summary, there is a large body of evidence showing that both cortical and trabecular bone loss occurs with age in both men and women. Declines in spinal trabecular bone appear to be similar for both genders, amounting to about 50% over a lifetime. Declines in BMD at cortical sites and the trabecular calcaneus are somewhat less for men than for women, amounting to approximately 25% for men and 33% for women at the femoral neck (Looker et al., 1995). Even so, these figures may underestimate true declines in bone mass among untreated patients if participants in epidemiologic studies tend to be healthier than the general elderly population.
Both cross-sectional and prospective studies have demonstrated the im-portance of BMD for predicting fracture risk in women. Osteoporosis tends to be associated with low bone mass throughout the skeleton, so that bone mass at a given site, such as the hand or the forearm, can provide information concerning the status of the skeleton as a whole. Consequently, BMD at numerous skeletal sites (including spine, hip, forearm, phalanges, and calcaneus) have been demonstrated to predict fracture risk (Marshall etal., 1996; Mussolino et al., 1997; Mussolino, 1998; Ross et al., 1995; Huang et al., 1999) . There is less information for men, but the data available are all con-sistent with those for women; studies in both genders are summarized here.
A. Hip Fracture Risk
Studies of hip fracture in men are mostly cross-sectional and show that, as seen in women, BMD is reduced at the hip, spine, and forearms, compared to age-matched controls (Greenspan et al., 1994a,b; Chevalley etal., 1991; Johnell and Nilsson, 1984). In a prospective study of 2879 white men and 1559 white women aged 45 to 74 who were followed for up to 16 years, the risk of hip fracture increased 1.7 to 1.9 times for each SD decrease in phalangeal BMD measured by RA (Mussolino, 1998; Mussolino et al., 1997). Another prospective study of men reported that the relative risk of hip fractures increased by 3.9 times (95% Cl 1.3-11.6) for each 1 SD decrease in forearm BMD (Nyquist et al., 1998). However, it does not appear that this analysis was adjusted for age, which would probably reduce the magnitude of the association.
B. Risk of All Types of Fractures
The Dubbo Osteoporosis Epidemiology Study (DOES) invited all men and women aged 60 years and older living in the Dubbo region of Australia; follow-up was approximately 3 years (38 months). The overall incidence of nonviolent fractures was 1.9% per year in men and 3.3% per year in women. Each 1 SD decrease in femoral neck BMD increased the risk of fractures by 2.0 times (95% Cl 1.5-2.6), after adjusting for muscle strength and balance (Nguyen et al., 1993). The 25% of men with the lowest femoral neck BMD (^ 0.82 g/cm2) had almost three times greater fracture incidence (3.3% per year) than the 25% of men with the highest BMD (incidence = 1.2% per year) (Nguyen etal., 1993). A later publication from the same study reported a weaker association; the risk increased 1.4 times (95% Cl = 1.2-1.6) for each 1 SD decrease in BMD (Nguyen et al., 1996).
In another prospective study, baseline forearm bone mineral content was lower in Scandinavian men who went on to sustain osteoporotic fractures (hip, pelvis, forearm, proximal humerus, vertebra, and tibial condyle) during 11 years of follow-up (Gardsell et al., 1990). The incidence of fractures was approximately six times higher in the 20% of men with the lowest distal forearm BMC, compared to the highest 20% (Gardsell etal., 1990). Another Scandinavian study of men found similar results; each 1 SD decrease in forearm BMD increased the risk of all fractures by 1.8 (95% Cl 1.1-2.8) times during 7 years of follow-up (Nyquist et al., 1998). Cheng et al. (1997) reported that the incidence of new fractures (at any site) increased progressively with declining levels of baseline calcaneus BMD and that fracture risk is comparable for both men and women at a given level of BMD. The same report demonstrated that decreases in calcaneus BMD during follow-up were associated with increased risk of nonspine fractures among women; a lack of association among men was attributed to the smaller number of fractures.
C. Vertebral Fracture Risk
Cross-sectional studies have shown that men with vertebral fractures have reduced mean levels of bone mass at several skeletal sites, including the spine (Orwoll et al., 1990; Riggs et al., 1981; Resch etal., 1995; Vega etal., 1994; Mann et al., 1992; Odvina et al., 1988; Cann et al., 1985), hip (Vega etal., 1994; Mann ctf al., 1992; Francises/., 1989), and total body (Hamdy et al., 1992). For example, the prevalence of vertebral deformities was 6.7 times higher among men with a femoral neck BMD more than 1 SD below the normal mean, compared to those with BMD more than 1 SD above the mean (Mann et al., 1992). The number of vertebral deformities was also negatively correlated with both spine and femoral neck BMD. Other large, population-based, cross-sectional studies have also reported that the prevalence of vertebral deformity is increased when BMD is low at the spine or hip, among men and women ages 50 and older (Lunt et al., 1997; Jones etal., 1996).
A prospective study of BMD and vertebral fracture incidence was conducted in the population-based Hawaii Osteoporosis Study (HOS) (Kim, 1996; Heilbrun etal., 1991). The HOS men (mean age 68 years) were several years older than the women (mean age 63 years) at baseline, and the mean follow-up was 9.6 years for women and 6.4 years for men. New vertebral fractures were defined as vertebral height decreases of more than 15% on serial spine radiographs; 151 of 964 women and 41 of 1008 men experienced new vertebral fractures during follow-up. The rate of new vertebral fractures was approximately three times greater among women than men in each age group. Accordingly, the age-adjusted fracture rate among women was 2.8 (95% Cl = 1.9, 4.0) times higher, compared to men.
Bone density (BMD and BMC) was a significant predictor of vertebral fractures in both HOS men and women (Table III). The risk of having a new fracture during follow-up increased by 1.5 to 2.0 times for each successive 1 SD decrease in baseline BMD for both women and men, depending on the BMD measurement site. When the incidence of new fractures was examined as a function of calcaneus BMD, men and women had equivalent risks of fractures for a given level of BMD. Similar results were observed for the distal and proximal radius BMD measurements.
TABLE III Associations of Baseline Bone Density with New Vertebral Fractures
|Relative risk- (95% C.I)|
1.66 (1.40, 1.97)
1.58 (1.33, 1.87)
1.63 (1.37, 1.93)
1.53 (1.31, 1.80)
Lumbar spine (Ll-4)
In contrast to BMD, there were substantial differences in fracture risk between men and women at similar levels of BMC; men had greater risk than women for a given level of BMC. This is probably because men have larger bones, and the mineral would be distributed over a larger bone area for men, resulting in weaker bone strength (compared to an equivalent amount of mineral in a smaller bone size for women). Fortunately, the larger bones of men are accompanied by greater amounts of bone mineral than women, on average. Adjusting for body size reduced the differences in fracture risk between men and women for a given level of BMC, yielding results similar to those for BMD. Thus, BMC values do not appear to account for differences in body size between men and women. These data suggest that using BMD allows evaluation of vertebral fracture risk using a single scale for both genders. However, it is not known whether this would also hold true for fracture risk at nonvertebral sites.
The relationship between changes in BMD and fracture risk were also examined in the HOS. For these analyses, rates of change in BMD were calculated up to the end of follow-up or time of the first fracture (whichever came first). Adjusting for bone loss rate did not measurably alter the associ-ations between BMD and fracture risk shown in Table III. Furthermore, changes in BMC and BMD were significant predictors of new vertebral fractures after adjusting for baseline bone density (Table IV), indicating that initial bone density and bone loss rate both contribute independently to fracture risk. The results indicate that faster declines in bone density are associated with increased risk of subsequent fractures. The magnitude was similar for both men and women, but did not attain statistical significance for men, probably because of the smaller sample size for men. The single exception was spine BMD; increases in spine BMD during follow-up (rather than decreases as observed for other bone density measurements) were associated with higher risk of fractures (spine BMD data were available for women only). As mentioned earlier, the increases in spine BMD probably do not reflect true changes in bone density, but rather the effects of arthritis and other health problems.
TABLE IV Associations of Changes in Bone Density with New Vertebral Fractures
|Relative risk’ (95% Cl)|
1.40 (1.15, 1.71
0.94 (0.78, 1.14)
1.23 (1.04, 1.46)
Lumbar spine (11-4)
1.13 (0.79, 1.60)
1.45 (1.04, 2.04)
1.13 (0.91, 1.41)
1.18 (0.84, 1.66)
1.18 (0.84, 1.65)
Prevalent (preexisting) vertebral fractures are an important predictor of future vertebral fractures, independent of BMD (Ross et al., 1991). The prevalence of vertebral fractures (defined as vertebrae with height dimensions more than 3 SD below the normal range) increased with age in both men and women, but the increases with age were greater among women . The risk of new vertebral fractures increased progressively with the number of prevalent fractures existing at baseline, for both men and women in the HOS (Table V). The magnitude was somewhat stronger for women than men, but the confidence intervals overlap. The smaller magnitude for men may be a statistical artifact related to the smaller number of fracture cases, or it may be a real effect if some of the fractures existing at baseline were caused by violent forces in the past (such as occupational injuries), and therefore do not represent true osteoporotic fractures. The higher vertebral fracture prevalence prior to age 65 among men than women in the HOS has also been reported by others, and suggests that vertebral fractures among men at younger ages are related to violent causes. Adjusting for BMD did not meaningfully alter the magnitudes of associations for prevalent fractures (Table 5), indicating that BMD and prevalent fractures provide complementary information about fracture risk.
TABLE V Associations of Preexisting Vertebral Fractures” with New Vertebral Fractures
Age-adjusted relative risk
Age- and BMD-adjusted relative risk
2.39 (1.06, 5.40)
2.18 (0.94, 5.02)
VI. Future Research Directions
Additional research is warranted to determine the extent to which reference ranges for men may differ between geographic localities and ethnic groups, as have been reported for women (Looker etal., 1995; Dequeker etal., 1995; Flicker et al., 1995). Further research is also needed to evaluate the usefulness of BMD measurements for monitoring individual responses to therapy, because few data are available in this regard for men. Nevertheless, BMD can accurately be measured in men, and low BMD levels are associated with increased fracture risk. It follows that identifying men with low BMD, and prevention of declines in BMD are important goals in the prevention of osteoporotic fractures among men.