Bone Mineral Density and Bone Loss Measured at the Radius to Predict the Risk of Nonspinal Osteoporotic Fracture



Low bone mineral density (BMD) and, probably, the rate of bone loss (RBL) are associated with the risk of osteoporotic fractures. To estimate the risk of nonspinal fracture in osteoporotic women, we measured BMD and RBL in a prospective study (average follow-up, 5.38 years) in 656 postmenopausal women. The women were considered in three groups: group A (whole population), group B (women under the age of 65years) and group C (women over the age of 65 years). At the beginning of the study, BMD was measured at the distal radius (DR) and at the proximal radius (PR) using a single-energy densitometer. BMD measurements made 2 years previously in the same patients were used to calculate RBL. Then patients were checked annually for nonspine fracture due to minor trauma. During follow-up, 121 nonspinal fractures were detected. Women with fractures were older and had lower BMD. With the Cox regression, age-corrected BMD at both DR and PR predicts fracture risk in groups A and B but not in group C. After correction for potential confounders, DR still predicts fractures in groups A and B whereas PR predicts fractures only in group B. In group C, only the RBL at the PR was predictive of the fracture risk as well as in the other two groups. Specific types of fractures are predictable in the whole population at the wrist. In conclusion, radial BMD predicts the risk of nonspine fractures except in women over the age of 65 years. The RBL at the PR is an effective predictor of fracture risk also in women over the age of 65 years.


PROSPECTIVE STUDIES have indicated that low bone mineral density (BMD) is a major risk factor for osteoporotic fractures.(1–4) It also has been documented that the sensitivity of the BMD in identifying patients with osteoporotic fractures depends on the skeletal site of measurement, the type of fracture, and the patients' age.(5–8) The pros and cons of measurement sites have been evaluated accurately and the diagnostic sensitivity of peripheral sites have been debated deeply, being of interest not only for BMD but also for ultrasound quantitative measurement of bone. Although forearm BMD has been found to predict spine and not spine fractures,(9–11) it is particularly useful in predicting wrist fractures(12) but less sensitive than spine and hip BMD in predicting vertebral(5) and proximal femur(8) osteoporotic fractures. Furthermore, the ability of forearm BMD to identify the risk in elderly people is debatable; it has been proven by some authors and confuted by others.(1, 2, 13) More recently, however, it has been found that the rate of bone loss (RBL) is just as important as BMD itself in predicting the risk of osteoporotic fracture in postmenopausal women.(14) Therefore, RBL might be considered an additional method for predicting osteoporotic fractures by densitometry. The aim of this prospective study was to assess whether RBL actually is associated with the risk of nonspine osteoporotic fractures and to contribute further to the evaluation of the predictive ability of forearm BMD in both young and old postmenopausal women.


Six hundred fifty-six postmenopausal women attending our center for their routine yearly check of BMD and having had two BMD measurements 2 years apart were included in this observational study. Those excluded were patients with endocrine, renal, pulmonary, or cardiac diseases; major rheumatic diseases; malignancies; immobilization; and long-term treatment with cortisone, anabolic steroids, or other drugs known to affect bone. Patients taking calcitonin, bisphosphonates, calcium supplements, or vitamin D were not excluded. Women underwent two BMD measurements using a single-energy densitometer (model 2780; Norland Corp., Fort Atkinson, WI, USA). All the measurements were taken by the same instrument, which was calibrated daily according to the manufacturer's instructions. BMD was measured at the proximal radius (PR-BMD; one-third site) and at the distal radius (DR-BMD; at the radioulnar separation of 5 mm) according to the technique described by Abwrey.(15) The reproducibility of the measurements at DR have been described previously.(16) At PR repeated measurement CV with patients' repositioning, estimated in 20 volunteers in the same way as for DR, was 1.92%. The reported BMD values are the average of three repeated scans on the point of measurement and are expressed as milligrams per square centimeter. The baseline information included RBL (calculated as half the difference between the two yearly BMD measurements), the baseline BMD (corresponding to the second BMD measurement), age, age of menopause, weight, height, previous fractures, and treatment. Fractures that occurred during the observation period, after the second BMD measurement, and the way they occurred were reported at the yearly visit of the patient and recorded. Only fractures due to minor trauma, such as falling from a standing position or walking, were included. Only nonspine fractures were considered, excluding those of the fingers. The diagnosis of fracture was accepted only when confirmed by an expert radiologist. All research methods were conducted in accordance with the Declaration of Helsinki. The study protocol has been approved by the local ethical committee.

Data analysis

The two-sample t-test was used to compare fractured and not fractured women. The frequencies of true positive (patients with incident fracture having a T score < −2.5) and false negative (patients with incident fracture having a T score > −2.5) cases at the two measurement points were compared using the χ2 test with Fisher's exact test.

The comparative ability of each measurement site to separate correctly women with and without fracture was tested by comparing the areas under the receiver operating characteristic (ROC) curves generated with their SE for PR and DR measurement.

The relative risk (RR) of the first fracture with the corresponding 95% CI was estimated for 1 SD decrease in the BMD value using the Cox regression. The time elapsing from the baseline observation and the fracture event was recorded as follow-up duration for the patient with incident fracture.


The selected population had a mean age of 57.08 ± 7.58 years. Mean follow-up was 5.38 ± 1.66 years. The baseline biological characteristics for fractured and healthy women are reported in Table 1 dividing the population into three groups: group A (whole population), group B (women under the age of 65 years), and group C (women over the age of 65 years). In groups A and B, the women with fractures were significantly older and had a lower DR and PR-BMD as well as DR and PR T scores than the healthy control group. Women with fractures also had a significantly higher RBL at DR in group A. In group C, only RBL at PR, being higher in women with fractures, approaches statistical significance (p = 0.054).

Table Table 1.. Baseline Biological Characteristics for Healthy and Fractured Women of the Three Groups Examined
original image

During the collection period, 121 nonspinal fractures (18.44%) were observed. The total number and the type of incident fractures for each of the three groups are reported in Table 2. Fracture rates and prevalence by age-adjusted quartiles of BMD are reported in Table 3. As expected, there was an inverse relationship between BMD values and fractures.

Table Table 2.. Number and Type of Fractures (% of the Population) of the Three Age Groups Examined
original image
Table Table 3.. Fracture Prevalence and Incidence Rates of Age Adjusted Quartiles of BMD (Number of Fractures)
original image

T scores were calculated from the BMD values (DR-BMD = 380.58 ± 54.07 mg cm−2, PR-BMD = 699.74 ± 58.76 mg cm−2) of 69 young fertile women (mean age, 39.84 ± 8.40 years). No statistically significant differences were found comparing DR-BMD with PR-BMD T score values in healthy women and in women with fractures, except for those not fractured in group C (DR-BMD T score = −1.77 ± 1.01 and PR-BMD T score = −2.23 ± 1.46; p < 0.018). Considering T score −2.5 as a cut-off value for osteoporosis from young healthy women, as proposed by the World Health Organization, at DR-BMD there were 15.70% women with incident fractures below and 84.30% above the cut-off value, while at the PR-BMD the corresponding rates were 28.10% (below) and 71.90% (above). The rate of true positive to false negative is significantly different between the two measurement points (χ2 = 4.735; p = 0.029). The proportion of women with fractures having at least one T score below the cut-off was 33.89%.

There were no statistically significant differences in the ROC curve areas between PR (area = 0.650; SE = 0.027) and DR (area = 0.680; SE = 0.026) measurement sites (Kendal Tau = 0.42; Z = 1.023). Curves are very close to each other. In fact for specificity values of 0.25, 0.50, 0.75, and 0.90 the sensitivities were 0.57, 0.72, 0.89, and 0.96 at DR and 0.47, 0.71, 0.86, and 0.95 at PR.

The fracture RRs were calculated for a “1 SD” decrease from healthy women mean value for baseline BMD values and RBL. Only the first fracture of patients with multiple fractures was evaluated to determine the fracture RR. The estimated age-corrected RRs were significantly associated with BMD at both measurement sites in group A (DR-BMD: RR, 3.54, and 95% CI, 1.81-6.90; PR-BMD: RR, 1.98, and 95% CI, 1.14-3.41) and group B (DR-BMD: RR, 2.90, and 95%CI, 1.29-6.50; PR-BMD: RR, 2.33, and 95% CI, 1.32-4.09), whereas they were not in group C (DR-BMD: RR, 1.83, and 95% CI, 0.73-4.56; PR-BMD: RR 1.06, and 95% CI, 0.38-2.94). After the correction for potential confounders, DR-BMD was still significantly associated with the fracture risk in groups A and B, whereas PR-BMD was only significant in group B (Table 4).

Table Table 4.. Cox Regression Measuring the RR of Osteoporotic Fractures for 1 SD Change in BMD Values in the Three Age Groups Considered, Adjusted for Age, Age of Menopause, Height, Weight, Previous Fracture, and Treatment
original image

In group A the RR was then calculated for specific types of fracture (wrist, rib, and ankle) sufficient in number for statistical analysis after correction for potential confounders. There was a significant association between BMD at both DR and PR and the risk for wrist fracture. No other statistically significant associations were found (Table 5).

Table Table 5.. Cox Regression Measuring the RR of Osteoporotic Fractures in the Whole Population for Specific Type of Fracture
original image

At PR there was a significant association of the RBL with the fracture risk, after correction for potential confounders, in group A (RR, 1.99; 95% CI, 1.13-3.51), in group B (RR, 2.00; 95% CI, 1.07-3.71), and in group C (RR, 2.98; 95% CI, 1.03-8.58). The relationship was persistent also after correcting for PR-BMD (group A: RR, 2.50, and 95% CI, 1.42-4.41; group B: RR, 1.90, and 95% CI, 1.05-3.43; group C: RR, 3.55, and 95% CI, 1.26-10.00). At DR, the RBL was not related to the fracture risk in any of the age groups. After correction for age, previous fracture was significantly related to the fracture risk in group A (RR, 1.50; 95% CI, 1.03-2.19) and in group B (RR, 1.58, 95% CI, 1.01-2.48) but not in group C (RR, 1.46; 95% CI, 0.73-2.92). Correcting for potential confounders, previous fracture was no longer predictive of the future fracture in any of the considered groups (Table 4).


In this study we assess the ability of radial BMD and RBL to predict the risk of nonspine fragility fractures in postmenopausal women. We found that age-corrected BMD, both at the PR and at the DR, significantly predicts the risk when a large range of ages is examined or when women younger than 65 years are considered separately. After correcting for potential confounders, PR-BMD is no longer an effective predictor of the fracture risk in the whole population but is an effective predictor in the younger age group of women. Our data seem to indicate that in postmenopausal women there are no significant differences between PR and DR-BMD in predicting the risk of nonspine fractures, in agreement with other authors.(13, 17) In women over 65 years of age we found that radial BMD does not predict fracture risk either at the proximal or at the distal measurement site. These findings agree with those of other authors, indicating that the ability of BMD to predict fracture becomes weak in the elderly(1, 13, 18) but is in contrast with other reports.(2, 11) Differences among authors about this topic might depend not only on the differences in BMD between healthy and selected elderly women with fractures but also on their physical characteristics such as propensity to fall or poor response to the neuromuscular function tests that are not generally considered in studies on the ability of BMD to discriminate osteoporotic fractures.

When specific types of fracture are examined, both radial BMD measurements corrected for potential confounders have a significant relationship with the risk of wrist fracture, whereas none of them are associated with risk of rib or ankle fracture. These data confirm the ability of radial BMD to predict the risk of the nonspine fracture commonly related to bone osteoporosis.(19, 20)

RBL at PR was significantly associated with the fracture risk in each age group considered. In our experience, in the elderly this is the only effective predictor of the fracture risk. Because it is persistent also after correcting for BMD, it seems to indicate further a direct and independent role of the rate of BMD loss on the risk of fracture in elderly people. This matches the observation of other authors,(21, 22) who report a relationship of osteoporotic fracture with high bone turnover, which is in turn directly correlated to the RBL.(23, 24) Furthermore, our data are in accordance with those of Riis,(25) who found that the RBL is significantly associated with the risk of fracture in a selected subpopulation of early postmenopausal women with low bone mass, which decreased at a high rate. On the contrary, no relationship was found between the fracture RR and RBL at the DR where cancellous bone mainly is represented.(26) This seems to suggest that radial cortical bone, which almost fully constitutes bone at the PR-BMD, is a marker of fragility at other skeletal sites. This is consistent with the better association of BMD with fracture risk at the middle radius rather than at the DR reported by other authors.(19)

The T score mean values were not significantly different in the two measurement sites in the whole population. Nevertheless, the rate of subjects with fractures having a different classification according to the cut-off value of −2.5 T score is significantly different between the two sites of measurement. Furthermore, we found that less than 34% of the patients with incident fractures had a baseline T score below the cut-off for osteoporosis at least at one of the measurement sites. These data reflect the difficulty of providing a univocal clinical classification to subjects by multisite BMD measurements(27) and show that at the radius the T score threshold has moderate sensitivity for prognostic decision-making.

Our study has some limitations. The population is a sample and not population-based and the BMD measurements have been limited to the radius so that no conclusions can be drawn about fracture predictability of site-specific BMD measurements for lower limb fractures. Moreover, photon densitometry is now obsolete. Nevertheless, our data can be relevant for defining the indication and limits of the radial site measurements by X-ray instruments.

In conclusion, our data show that the radial BMD predicts the risk of osteoporotic nonspinal fracture with a better predictability for radial fractures and that its predictive ability becomes weak in older people. RBL at PR predicts the risk of fragility fractures not only in the younger people but also in the elderly. At this age, we found that it is the only predictor of fracture risk. T score classification has a moderate sensitivity for clinical purposes.


This study was funded by a grant from Rizzoli Orthopaedic Institute.