Bone Mineral Density and Blood Flow to the Lower Extremities: The Study of Osteoporotic Fractures


  • Presented in part at the 16th Annual Meeting of the American Society for Bone Research in Kansas City, Missouri, U.S.A., 1994.


This study tests the hypothesis that reduced blood flow to the lower extremities may affect bone remodeling, resulting in a decrease in bone mineral density (BMD). BMD was measured in the axial and appendicular skeleton of 1292 elderly women (mean age, 71 years) enrolled in the Study of Osteoporotic Fractures. The ratio of the posterior tibial and brachial systolic blood pressures, the ankle/arm index, was used as a measure of blood flow to the legs. In the cross-sectional analysis, this index was positively correlated with BMD at the radius, calcaneus, and hip, but not at the spine. A decrease in the index of 2 standard deviations (SD) (as might occur in patients with moderate arterial disease) was associated with a decrease of 3.7% (95% CI, 1.7%, 5.8%) in hip BMD. The effect size at the hip decreased from 3.7 to 1.8% (and was not statistically significant) when adjustment was made for smoking and/or body mass index (BMI). In the prospective analysis, the rate of bone loss at the hip and calcaneus was greater (p < 0.05) among women whose annual decrease in ankle/arm index was more than 1 SD greater than the mean decrease. This increase was independent of estrogen use, smoking, BMI, pattern of fat distribution, history of diabetes, exercise, and ability to walk. The results from this prospective community-based study provide the first evidence that among relatively healthy older women decreased vascular flow in the lower extremities may be associated with an increased rate of bone loss at the hip and calcaneus.


Severe symptomatic lower extremity arterial disease (LEAD) is often accompanied by decreased bone mineral density (BMD) in the affected limb.1 Surprisingly, even the bony tissue proximal to the arterial stenosis can show pathologic changes including loss of osteocytes and infarction of the marrow or bony cortex.1 It is not known whether these findings represent a specific response to the chronic intermittent tissue ischemia due to LEAD or to the decreased mobility common in these patients. In patients with acute traumatic injuries, there is evidence that fracture healing can also be delayed if one or more of the major arteries of the leg are occluded or injured.2,5

Epidemiologic evidence also suggests that there may be an association between cardiovascular disease and osteoporosis particularly in postmenopausal women. Both conditions share several important risk factors. For instance, smoking is associated with both increased incidence of cardiovascular disease, including LEAD, and with low BMD.6,7 Conversely, older women who receive hormone replacement therapy are at decreased risk for cardiovascular disease and osteoporosis.8–10 However, obesity is an important counterexample in that it is associated with an increased risk for cardiovascular disease and LEAD and decreased risk for osteoporosis.

This study has been designed to investigate the relationship between blood flow in the lower extremities, BMD in the axial and appendicular skeleton, and the annual rate of bone loss. The 1292 elderly women evaluated are enrolled in the Pittsburgh Clinic of the community-based Study of Osteoporotic Fractures (SOF).


Study population

This study is ancillary to the multicenter SOF, a prospective study of the risk factors for fractures in elderly women. Women living in four geographical areas (Baltimore, MD, U.S.A., Minneapolis, MN, U.S.A., Monongahela Valley near Pittsburgh, PA, U.S.A., and Portland, OR, U.S.A.) were recruited from population-based listings to participate in this study. During a 2-year time period (October, 1986–October, 1988) 2401 women aged 65 years or older were enrolled in the SOF clinic in a rural area near Pittsburgh. To identify potential participants, mass mailings were made to all women in this age group listed on the voter's registration list and living within approximately 25 miles of the clinic. Women who were unable to walk without the help of another person, were institutionalized, and those who had had bilateral hip replacement were ineligible to participate in SOF. African-American women were also excluded because they are at low risk for osteoporotic fractures.

At the baseline visit, approximately two-thirds of the enrollees at the Pittsburgh clinic (1601/2401) also participated in an ancillary study designed to investigate the prevalence, correlates, and natural history of LEAD. The demographic characteristics and medical conditions reported by the 1601 women enrolled in the ancillary study were similar to those reported by the 800 women who did not participate. The institutional review board at the University approved this ancillary study in 1986.

Measurement of lower-extremity arterial disease

The systolic blood pressures in the right and left posterior tibial artery and the right brachial artery were measured twice after the subjects had been resting in the supine position for at least 5 minutes. The pulses were detected using a hand-held Doppler flowmeter (8 MHz Huntleigh Dopplex, Cardiff, U.K.); blood pressure was measured using a standard mercury manometer. Interobserver and intraobserver variability was not statistically significant in our clinic.7,11

The resting ankle/arm blood pressure index in each leg was calculated as the average of the two measurements of the posterior tibial pressure divided by the average of the two brachial pressure measurements. The index could not be calculated for 104 women. Most of these women refused to allow duplicate measurements of the tibial systolic pressure either due to time constraints or to discomfort during the procedure; in about 10%, the tibial pulses could not be located due to obesity or leg edema. These women (n = 104) were older (74.3 vs. 71 years) and more likely (20 vs. 9%) to be diabetic than the women for whom indices were obtained.

The American Heart Association Council on Epidemiology has endorsed the use of ankle/arm index measurements to assess arterial stenosis and the adequacy of blood flow through the major arteries of the lower extremities.12 Previous studies have shown that individuals with a resting index of 0.94 or higher have no arteriographic evidence of arterial disease.13–15 With increasing stenosis, the index falls and an index of 0.5 or less is indicative of severe arterial disease. Women were excluded from the data analysis (n = 5) if their ankle/arm index was greater than 1.5. Such high values are found in subjects with rigid incompressible arterial walls and are considered to be measurement artifacts.14 The ankle/arm index was repeated at the fourth clinic visit which occurred about six years after the baseline visit.

Measurement of BMD

At the baseline visit the bone mineral density (g/cm2) of the distal and proximal radius and the calcaneus were measured using a OsteoAnalyser (Siemens-Osteon, Wahiawa, HI, U.S.A.) following a standard protocol.16 Coefficients of variation at these sites were 1.5, 2.0, and 1.3%, respectively. BMD of the proximal femur and spine was measured in the anteroposterior projection at the 2-year follow-up visit using a Hologic QDR 1000 scanner (Hologic, Inc., Waltham, MA, U.S.A.). Coefficients of variation at these sites were 1.2 and 1.5%, respectively. All measurements were made on the right side unless the subject reported a stroke, fracture, or severe injury involving the right limb; measurements were made on the left side in these women.

All women were invited to participate in the fourth clinic visit (between 5 and 6 years after the baseline visit) and the BMD of the calcaneus (single-energy X-ray absorptiometry) and the hip (Hologic QDR 1000 scanner) were again measured.10 The intraindividual correlation coefficients between the hip measurements made at the second and fourth clinic visit was 0.96; and for the two calcaneus measurements made at the baseline and fourth clinic visit was 0.94. Assessment of the longitudinal performance of the scanners revealed a mean BMD shift of −3 SD during a 6-month period (July–December 1989) at the Pittsburgh Clinic.10 The calculations in this paper were performed twice: once with the hip and spine BMD data acquired during this time period excluded and once with this data included. The results were very similar and therefore the data for the entire cohort is presented in this paper. The range of rates of bone loss reported are 2- to 4-fold greater than could be accounted for by methodologic imprecision.

Covariate measurement

At baseline all women completed a questionnaire and information was obtained on age, education, history of selected medical conditions (arthritis, diabetes, angina, myocardial infarction, stroke, osteoporosis, claudication), history of cigarette use, exercise, and use of medications (estrogen, diuretics). The women were also asked whether they could walk two to three blocks on level ground and if they walked regularly for exercise. Subjects were weighed on a balance beam scale and height (without shoes) was measured with a stadiometer (Harpenden, Holtain Ltd., Crymych, Dyfed, U.K.). Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Waist and hip girths were measured with a steel tape as described previously,10 and the ratio of waist and hip girth was used as an indicator of body fat distribution.

Statistical analysis

The data from 200 women who did not have their hip and spine BMD measured at the 2-year follow-up visit was excluded from this analysis; typically these women had either died prior to the clinic visit or were confined to the home because of personal illness or the illness of their spouse. Analysis of the data with these 200 subjects included did not appreciably change the strength of the relationship between baseline appendicular BMD and ankle/arm index. The final number included in the cross-sectional analysis was 1292.

Age, BMD, BMI, waist/hip ratio, and ankle/arm index were used as continuous variables in the analysis except where otherwise indicated. Other covariates were used as categorical variables: estrogen use (current/former/never), walking for exercise (yes/no), insulin or noninsulin dependent diabetes (yes/no), smoking (never/former/current), ability to walk 2–3 blocks on level ground (yes/no). The values of ankle/arm index and the BMD used in the analysis were measured on the same side of the body, i.e., the left index was used when the left hip was scanned to measure BMD. Partial correlation coefficients were calculated to assess the age-adjusted correlation between the index and BMD at the various sites. Multivariate linear regression was used to determine the percentage change in BMD per unit change in ankle/arm index adjusted for other covariates. The percentage change in BMD associated with each unit change in a covariate was calculated as (β [unit] ± 1.96 [standard error])/mean BMD of women in study.

Across the entire SOF cohort, 77% of the participants who were still living (6.7% of the cohort had died) chose to return for the fourth clinic visit and to have their hip and calcaneus BMD remeasured. The percentage of women in this ancillary study who returned for clinic visit four were similar. Ninety five percent had technically adequate measures of calcaneus BMD at this time. Data on both the annual rate of decrease in blood flow and the annual rate of bone loss were obtained on a total of 780 women (n = 769 for hip, n = 767 for calcaneus). The annual change in bone mass was calculated as the difference between the BMD at baseline (calcaneus) or the second visit (hip) and the BMD measured at the fourth clinic visit (mg/cm2/year), divided by the time elapsed between visits. The annual change in ankle/arm index was calculated in a similar manner. The age-adjusted correlation between the baseline ankle/arm index, annual change in the index, and the annual bone loss at the hip, and calcaneus was determined by calculation of the partial correlation coefficient. Multivariate linear regression was used to determine the rate of bone loss stratified by annual change in blood flow and adjusted for age and other covariates.


The average age of the women in this study at the baseline visit was 70.6 years, 8.4% were diabetic, approximately one-quarter had ever used estrogen and about one-half reported walking for exercise on a regular basis. The mean ankle/arm index was identical in both legs (Table 1).

Table Table 1. Baseline Characteristics of the Population of Elderly Women (N = 1292)
original image

The BMD at the distal radius, proximal radius, calcaneus, and total hip were weakly but positively correlated with the baseline ankle/arm index (p < 0.05) after adjustment for age. Total spine BMD was not correlated with the index (p > 0.05) (Table 2). The age-adjusted percentage increase in BMD for each unit increase in ankle/arm index varied from 3.7% at the hip to 1.9% at the proximal radius (unit decrease in ankle/arm index = 2 SD). These associations remained unchanged when estrogen use, waist/hip ratio, history of diabetes, ability to walk two to three blocks on level ground and walking for exercise were entered into the model. The effect size at the hip decreased from 3.7 to 1.8% per unit change in index when adjustment was made for smoking and/or BMI (Table 3). By comparison in the full model current smoking was associated with a decrease in hip BMD of 4.4% (95% CI, 1.6%, 7.3%); a five-year increase in age with a decrease of 3.7% (95% CI, 2.7%, 4.6%), current estrogen use with an increase of 10.1% (95% CI, 7.1%, 12.9%); and a 1 SD increase in body mass index with an 8.8% (95% CI, 7.9%, 9.7%) increase in hip BMD. Inclusion of the ankle/arm index in the model did not change the parameter estimates for any other variables in the model.

Table Table 2. Correlation Between Ankle/Arm Index and Bone Mineral Density
original image
Table Table 3. Percentage Difference in Bone Mineral Density Per Unit* Difference in Ankle/Arm Index Among Elderly Women at Baseline
original image

Examining the data by tertiles of baseline ankle/arm index, the BMD at the hip, calcaneus, proximal, and distal radius increased by tertile and the mean BMD in the highest tertile was higher (p < 0.05) than that in the other two groups. (Fig. 1) After stratification of the data by smoking status, the mean BMD showed no statistically significant variation across tertiles of ankle/arm index although the mean BMD for smokers was lower than that for never smokers. Similarly, the mean BMD did not vary across tertiles after stratification by tertiles of BMI.

Figure FIG. 1.

Age-adjusted mean BMD at sites in the axial and appendicular skeleton by tertile of ankle/arm index (n = 1292).

The average absolute rate of bone loss at the hip was 5.3 (95% CI, 4.5, 6.1) mg/cm2/year, or an annual percentage loss of 0.71% (95% CI, 0.60%, 0.81%); for the calcaneus, these rates of loss were 3.6 (95% CI, 3.2, 4.0) mg/cm2/year and 0.89% (95% CI, 0.79%, 0.99%). The correlation between baseline ankle/arm index and the annual change in BMD was low and not statistically significant (Table 4). Stratification of the data by tertiles or quintiles of baseline ankle/arm index revealed no statistically significant difference between the annual bone loss in each group of women (data not shown).

Table Table 4. Correlation Between Ankle/Arm Index at Baseline, Annual Change in Ankle/Arm Index and Annual Change in Bone Mineral Density at the Hip and Calcaneus
original image

The annual change in the ankle/arm index also showed a low correlation with change in BMD but the age-adjusted correlation coefficient for the hip was statistically significant (p < 0.05) (Table 4). The annual change in ankle/arm index ranged from a decrease of 0.097 to an increase of 0.088 with a mean annual change close to zero (0.0004 ± 0.020). After stratification by amount of change in this index, women whose index decreased the most (more than 1 SD above the mean annual change, n = 113) showed an age-adjusted annual decrease in hip BMD that was more than twice that for women whose index change was within one standard deviation of the mean (n = 561) (Fig. 2). Thus, women who had decreased blood flow to the lower extremities during the follow-up period lost more bone than women whose blood flow remained relatively stable during follow-up. Women whose annual index decrease (n = 95) was more than 1 SD below the mean also showed a higher annual bone loss, but this rate of loss was not statistically significantly different from the loss found in the other groups (Fig. 2). Inclusion of other covariates (smoking, use of estrogen, ability to walk two to three blocks without difficulty, walking for exercise, BMI, waist/hip ratio, and history of diabetes) in the linear regression model (individually or in combination) did not change the rates of bone loss. The annual change in calcaneal BMD was similarly influenced by changes in ankle/arm index over time but the magnitude of the effect was smaller (Fig. 2).

Figure FIG. 2.

Mean rates of percentage annual change in BMD at the hip and calcaneus adjusted for age and stratified by annual change in ankle/arm index. The 95% confidence intervals for the rates are shown in parentheses. The change in ankle/arm index is stratified into four groups: Group 1: increased blood flow, ankle/arm index decline more than one standard deviation below the mean (range −0.0971 to −0.0197, hip n = 95, calcaneus n = 95); Group 2: slightly increased blood flow, ankle/arm index decline equal to or less than one standard deviation below the mean, but less than the mean (range −0.0196 to +0.0003, hip n = 282, calcaneus n = 283); Group 3: slightly decreased blood flow, ankle/arm index decline equal to or greater than the mean loss but not greater than one standard deviation above the mean (range +0.0004 to +0.0204, hip n = 279, calcaneus n = 281); Group 4: decreased blood flow, ankle/arm index decline more than one standard deviation above the mean loss (range equals +0.0205 to +0.0879, hip n = 113, calcaneus n = 108).


These results provide the first evidence from a community-based study that a decrease in blood flow to the lower extremities (measured as a decrease in the ankle/arm index) may be associated with an increase in the annual rate of bone loss at the hip and calcaneus. Women who had the largest decrease (greater than 1 SD above the mean decrease) in ankle/arm index between the baseline visit and the fourth clinic visit (5–6 years later) lost bone at the hip at approximately twice the rate as women with smaller changes in this index. Annual bone loss at the calcaneus was increased about 30% in this group of women. These findings were independent of all other covariates tested, including baseline ankle/arm index.

The cross-sectional data from this study indicate that among elderly women the ankle/arm index is weakly but positively correlated with BMD at the radius, calcaneus, and hip. Women with increased blood flow to the lower extremities have a higher bone mass at the hip and in the appendicular skeleton. This association is independent of estrogen use, pattern of fat distribution, history of diabetes, ability to walk, and exercising. However, after adjusting for either the smoking status or BMI, the association was markedly reduced and was not statistically significant. Similiarly, the annual rate of bone loss at the hip and calcaneus was not correlated with the baseline level of blood flow to the legs.

Our results confirm previous findings in cross-sectional studies that cigarette smoking, BMI, and use of estrogen may affect the development of osteopenia in postmenopausal women.6,8,10,18–21 However, we found no evidence from our cross-sectional analysis to suggest that the effect of these variables on BMD was mediated by variations in blood flow. Smoking is considered to be the major risk factor for lower extremity arterial disease (or a low ankle/arm index)7,17 and also a moderately important correlate of decreased BMD.6,18–21 When a variable for ankle/arm index was added into a multivariate model for hip BMD, the parameter estimates for smoking were unchanged. Conversely, the addition of smoking to a model for hip BMD resulted in a marked decrease in the parameter estimate for ankle/arm index. This latter effect suggests that in the cross-sectional analysis the association between decreased arterial blood flow in the lower extremities (as assessed by measurement of ankle/arm index) and BMD reflects confounding by smoking or correlates of smoking. Similar results were obtained when the variable for BMI and ankle/arm index were added to models for hip BMD in the same way. It is interesting that the age-adjusted correlation between ankle/arm index and BMD is not site specific and occurs not only in the hip and calcaneus but also in the radius. Although ankle/arm index is considered an indicator of overall cardiovascular health status, it is not correlated with reduced blood flow to the upper extremity. This provides additional evidence that the correlations in the cross-sectional analysis observed may be artifactual, reflecting confounding by other variables.

Cross-sectional analysis may also be subject to other potential sources of bias.10,22,23 In community-based studies, it is usually the most healthy surviving individuals who volunteer to participate. These persons are likely to have both higher BMD and better cardiovascular status than the nonparticipants. This may lead to truncated distributions of these variables influencing the correlation coefficients obtained and leading to an underestimation of the relationship between BMD and blood flow.

The interpretation of the results from the longitudinal data must also be done with caution. No association was found between the baseline ankle/arm index and annual rate of bone loss at either the hip or calcaneus. In addition, the relationship between the annual change in blood flow (ankle/arm index) and annual rate of bone loss was nonlinear. As hypothesized, women with a marked decrease in annual flow had a larger rate of bone loss than those with relatively stable blood flow. But women with increased blood also showed a larger (but statistically nonsignificant) increased rate of bone loss.

There is a growing body of evidence from animal research, epidemiologic, and clinical studies indicating an association between osteopenia and cardiovascular disease. Research with dogs points to a close correlation between bone blood flow and bone formation and mineralization.24 A highly reactive vasomotor control mechanism appears to be operating in bone,25,26 and under ischemic conditions the intraosseous vessels become more sensitive to circulating catecholamines. In a small cross-sectional study of men (n = 17) with unilateral symptomatic LEAD,27 the BMC of the femoral neck on the affected side (ankle/arm index 0.64) was 3.3% lower than on the healthy side (ankle/arm index 0.84). These differences (which are similar in magnitude to the changes reported in this paper) were attributed to the effects of ischemic atherosclerotic disease. Additionally, total bone blood flow has been reported to be slightly lower than normal in crush fracture osteoporosis.28

Browner et al. found that osteopenia is related to an increased risk of morbidity and mortality from stroke among the elderly women enrolled in SOF.29,30 They concluded that this association was indirect, mediated by some other process and suggested that calciotropic hormones, diet, estrogen deficiency, or elevated levels of plasma homocysteine may play a role in linking bone remodeling and vascular disease.31–35 The results from our study provide supportive data of a link between the cardiovascular system and the development of osteopenia.


Supported in part by NIH Grants 5T32HL7011, AR35582, HL40489, and AG05407.