Osteoporosis International

, Volume 20, Issue 12, pp 2079–2085

The prevalence of significant left–right hip bone mineral density differences among black and white women

Authors

    • Division of Endocrinology, Diabetes and Medical GeneticsMedical University of South Carolina
  • D. L. Kamen
    • Division of RheumatologyMedical University of South Carolina
  • K. L. Hermayer
    • Division of Endocrinology, Diabetes and Medical GeneticsMedical University of South Carolina
  • J. Fernandes
    • Division of Endocrinology, Diabetes and Medical GeneticsMedical University of South Carolina
  • J. Soule
    • Division of Endocrinology, Diabetes and Medical GeneticsMedical University of South Carolina
  • M. Ebeling
    • Division of Pediatric EpidemiologyMedical University of South Carolina
  • T. C. Hulsey
    • Division of Pediatric EpidemiologyMedical University of South Carolina
Original Article

DOI: 10.1007/s00198-009-0915-0

Cite this article as:
Alele, J.D., Kamen, D.L., Hermayer, K.L. et al. Osteoporos Int (2009) 20: 2079. doi:10.1007/s00198-009-0915-0

Abstract

Summary

In a cross-sectional retrospective study, we examined the prevalence of significant opposite hip bone mineral density difference among white and black women. Left–right hip bone mineral density difference was a common finding in both races, raising the possibility that osteoporosis can be missed if only one hip is imaged.

Introduction

We examined the prevalence of significant left–right hip bone mineral density (BMD) difference among black and white female subjects and its implications on the diagnosis of osteoporosis.

Methods

This was a retrospective review of dual energy X-ray absorptiometry (DXA) data in black and white subjects age 50 years and older. One thousand four hundred seventy-seven scans obtained using a GE Lunar Prodigy scanner in dual hip mode were analyzed (24% black, 76% white). Significant left–right hip BMD difference was considered present when the subregion least significant change (LSC) was exceeded. Its prevalence was determined, along with consequences on the diagnosis of osteoporosis.

Results

Significant differences in BMD were common in both races; the LSC was exceeded in 47% of the patients at the total hip, 37% at the femoral neck, and 53% at the trochanter. Diagnostic agreement was lower when the LSC was exceeded than when it was not. The LSC was exceeded in a statistically significant number of black and white patients with normal or osteopenic spines and unilateral hip osteoporosis.

Conclusions

Significant left–right hip BMD difference is a common finding among black and white women and can result in osteoporosis being missed if only one hip is imaged.

Keywords

Black womenBone mineral densityDual energy X-ray absorptiometryHipOsteoporosisWhite women

Introduction

Osteoporosis is a major and increasing public health problem [1]. Among sequelae of osteoporosis, fracture of the hip causes the most morbidity, mortality, and medical expenditure [13]. With a progressively aging population, hip fracture will take an increasing economic and personal health toll. Therefore, early detection of those at highest risk of fracture is an essential factor in the prevention of hip fracture.

The best way to diagnose osteoporosis is by measurement of bone mineral density (BMD) using dual energy X-ray absorptiometry (DXA) scans [4, 5]. Based on epidemiological studies, each standard deviation of hip BMD below the peak BMD (T-score) predicts a two- to threefold increased risk of fracture [68].

The preferred method for diagnosing osteoporosis at many centers is by measuring BMD at the lumbar spine (L1–4) and one hip, usually the non-dominant side, the opposite side of handedness (i.e., if the patient is right-handed, the left hip is measured). This practice is based on the assumption that there are no significant BMD differences between the two hips, a notion which is supported by two prior studies concluding that differences between the left and right hip subregions were not clinically significant [9, 10].

However, a few studies report disagreement between opposing hip subregion BMD [1113]. These reports lead some to suggest BMD measurement in both hips to reduce the chances of missing osteoporosis. In a study that examined opposing hip subregion BMD difference, Hamdy et al. found that significant left–right hip BMD differences were common among white women, resulting in a significant number of women with osteoporosis being classified differently when only one hip was scanned [14].

The current study was conducted to compare the prevalence of significant left–right hip BMD discrepancy and the resulting diagnostic differences between female white and black subjects in our patient population. We tested the hypothesis that left–right hip BMD difference would most impact women with normal or osteopenic spines because of the potential of a non-imaged hip being the only site meeting the threshold for osteoporosis.

Methods

Study design

This study is a retrospective review of outpatient DXA data collected at the Medical University of South Carolina (MUSC) Bone and Joint Center from 2000 to 2006. It was approved by the MUSC Institutional Review Board. All scans were performed using a GE (Lunar Prodigy) scanner. Data on all white and black women >50 years old were selected for analysis. All patients had spine and hip scans done in the same session. Hip scans were acquired in dual hip mode. For women who had more than one DXA scan, only the first DXA was included in the analysis. Scans with anatomical defects were excluded from analysis. A total of 1,477 scans were selected for analysis.

BMD was reported in grams per square centimeter for all cases. Spine T-scores were calculated using normative data provided by the manufacturer, while hip T-scores were calculated using NHANES III normative data [15]. Patients were classified as normal, osteopenic, or osteoporotic depending on the lowest measured T-score [16]. To allow for a direct comparison with Hamdy et al., we included the trochanter in our analysis, despite International Society for Clinical Densitometry recommendations to use only total hip and femoral neck for the diagnosis of osteoporosis [17]. Overall, osteoporotic range BMDs (spine and both hips, the spine only, or both hips only) were detected in 2.5%, 5.6%, and 4.7% of the patients, respectively. It was detected on the left or right hip only in another 5.5% and 6.6% of the patients, respectively.

Hip subregion (total hip, femoral neck, and trochanter) precision calculation was performed following duplicate hip scanning in 30 patients with repositioning in between measurements. There was no significant difference between the left and right hip subregion precision based on the F test, and pooled precision values were used to determine the least significant change (LSC).

Opposing hip BMD difference was considered significant when the BMD difference between the left and right hips exceeded the subregion LSC (0.033 g/cm² at the total hip, 0.046 g/cm² at the femoral neck, and 0.034 g/cm² at the trochanter, respectively, 95% confidence interval).

Data analysis

Prevalence of significant left–right hip BMD difference among white and black subjects was determined by the percentage in which the left–right hip BMD difference exceeded the subregion LSC. We then compared the percentage of discrepant hip measurements in white and black cases to assess for any racial differences. We also compared the hip subregion diagnostic agreement (normal, osteopenic, or osteoporotic) between the two races, based on the frequency that subjects’ contralateral hips were discrepant for meeting the threshold for diagnosis.

We then determined the cases among whom a diagnosis of osteoporosis depended on the bone status of one hip only as was done by Hamdy et al. Patients in each race category were sorted by spine status (normal, osteopenic, and osteoporotic) and by age (50–64 and >65 years) followed by identification of hip pairs in which one side was normal or osteopenic while the other side was osteoporotic using the lowest measured T-score. Finally, we determined the number and percentage of pairs in which hip BMD exceeded the LSC for each race. We then compared the percentage of cases among whom the LSC was exceeded between white and black cases to determine if there were any differences between the races.

Results

We found significant left–right hip subregion differences in BMD to be common, with the LSC being exceeded in 47% of the patients at the total hip, 37% at the femoral neck, and 53% at the trochanter.

Overall, left–right hip BMDs were highly correlated at the total hip, femoral neck, and trochanter (r = 0.95, 0.91, and 0.91 respectively, p < 0.0001). Left–right hip BMDs were highly correlated among white and black subjects (r = 0.94, 0.90, and 0.90 vs. 0.96, 0.93, and 0.93, p < 0.0001) at the total hip, femoral neck, and trochanter, respectively. No significant differences were detected when opposing hip subregion BMD correlations were compared among blacks or whites.

Left vs. right mean BMDs (grams per square centimeter, SD) were 0.939 (0.17) vs. 0.929 (0.17), 0.885 (0.16) vs. 0.887 (0.16), and 0.745 (0.15) vs. 0.737 (0.14) at the total hip, femoral neck, and trochanter, respectively. There were no significant differences in mean left vs. right hip subregion BMD for blacks or whites. Grouped by race to determine any black–white differences in the prevalence of opposing hip subregion BMD differences, there were more cases in which the left–right hip subregion differences exceeded the LSC among blacks than there were among whites at all subregions (52% vs. 45% at the total hip, 43% vs. 34% at the femoral neck and 58% vs. 51% at the trochanter; p = 0.019, 0.002, and 0.003, respectively; Table 1).
Table 1

Percent of cases in which the left–right hip subregion exceeded the LSC with patients sorted by spine status, age, and ethnicity

 

Total hip, %

Neck, %

Trochanter, %

Spine

 Osteoporosis (n = 82)

37.8

23.2

45.1

 Osteopenia (n = 368)

43.2

32.3

50.0

 Normal (n = 1,027)

48.1

38.2

54.1

 

p = 0.08

p = 0.006

p = 0.15

Age

 50–64 (n = 857)

44.9

36.1

51.7

 65+ (n = 620)

48.7

35.6

53.9

 

p = 0.21

p = 0.87

p = 0.41

Ethnicity

 Black (n = 360)

51.7

42.8

57.5

 White (n = 1,117)

44.6

33.7

51.0

 

p = 0.019

p = 0.002

p = 0.03

Comparing younger (<65 years) and older (>65 years) age groups, we found no difference in the prevalence of left–right hip BMD discrepancy. Significant left–right hip BMD differences were more common among blacks than whites in all three areas of measurement, and this difference was statistically significant in the femoral neck among those 50–64 years old (p = 0.03) and in the femoral neck and trochanter in those >65 years old (p = 0.02 and p = 0.047, respectively) (Table 2).
Table 2

Within-age category percent of cases in which the left–right hip subregion exceeded LSC, grouped by ethnicity

Total hip, %

Neck, %

Trochanter, %

50–64 years old

 Black (n = 223)

50.2

42.1

55.2

 White (n = 634)

43.1

33.9

50.5

 

p = 0.06

p = 0.03

p = 0.2

65+ years old

 Black (n = 137)

54.0

43.8

61.3

 White (n = 483)

46.6

33.3

51.8

 

p = 0.12

p = 0.02

p = 0.047

We then examined the effect of significant opposing hip subregion BMD difference on diagnostic agreement. There were 95%, 94%, and 91% agreement when the LSC was not exceeded vs. 80%, 77%, and 70% when it was exceeded at the total hip, trochanter, and femoral neck, respectively (p < 0.0001). Diagnostic agreement was consistently higher among blacks compared to whites (Table 3). When the LSC was exceeded, younger subjects showed better agreement (83.6%, 80.8%, and 74.1%) compared with subjects >65 years of age (72.2%, 71.0%, and 64.2%) at the total hip, trochanter, and femoral neck subregions, respectively.
Table 3

Percent agreement between left and right classification for all women 50+ irrespective of spine status and by race

Hip subregion

<LSC (%)

>LSC (%)

Chi-square p

All women

 Total hip

95.0

79.7

<0.0001

 Trochanter

93.7

76.6

<0.0001

 Femoral neck

90.6

70.0

<0.0001

Black

 Total hip

96.6

87.6

0.002

 Trochanter

96.1

79.2

<0.0001

 Femoral neck

94.7

83.8

0.0007

White

 Total hip

94.5

75.3

<0.0001

 Trochanter

93.1

75.6

<0.0001

 Femoral neck

89.5

64.4

<0.0001

We then limited the analysis to women with normal (n = 1,027) and osteopenic (n = 368) spines and excluded those who had already met the diagnosis of osteoporosis based on spine BMD (Table 1). Hip status was defined by the subregion with the lowest T-score. Our purpose was to identify the absolute numbers and percentages of women in these spine categories who had unilateral hip subregion osteoporosis (Figs. 1 and 2). For women with normal spines, 18 (1.8%) had osteoporosis of the left hip only and another 19 (1.9%) had osteoporosis of the right hip only. Among women with osteopenic spines, 29 (7.9%) had osteoporosis of the left hip only while 35 (9.5%) had osteoporosis of the right hip only (Table 4). For each group, the absolute number of women with left–right hip BMD differences exceeding the LSC was statistically significant (significantly greater than zero). This trend persisted when the patients were grouped by race, although lower numbers (especially for blacks) imply a need for cautious interpretation (Tables 5 and 6).
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-009-0915-0/MediaObjects/198_2009_915_Fig1_HTML.gif
Fig. 1

Left–right hip diagnostic agreement in women with normal spines. Scatter plot was derived by plotting the minimum T-scores (total hip, femoral neck, or trochanter) for the left and right hips for women with normal spines; each dot represents one hip pair. Women with discordant hips are located in the left upper and right lower quadrants. Women with osteoporotic left and right hips are located in the left lower quadrant, while women with osteopenic to normal hips are located in the right upper quadrant

https://static-content.springer.com/image/art%3A10.1007%2Fs00198-009-0915-0/MediaObjects/198_2009_915_Fig2_HTML.gif
Fig. 2

Left–right hip diagnostic agreement in women with osteopenic spines. Scatter plot was derived by plotting the minimum T-scores (total hip, femoral neck, or trochanter) for the left and right hips for women with normal spines; each dot represents one hip pair. Women with discordant hips are located in the left upper and right lower quadrants. Women with osteoporotic left and right hips are located in the left lower quadrant, while women with osteopenic to normal hips are located in the right upper quadrant

Table 4

Discordant hip pairs in women with normal and osteopenic spines

Spine status

Osteoporosis

Number of women

Number of women (%) with left–right difference >LSC

Percent of women with left–right difference >LSC for group defined by spine statusa

Normal

Left hip

18

6 (33.3%)

0.6

Normal

Right hip

19

7 (36.8%)

0.7

Osteopenic

Left hip

29

5 (17.2%)

1.4

Osteopenic

Right hip

35

11 (31.4%)

3.0

The absolute number of women with left–right hip BMD differences exceeding the LSC was statistically significantly greater than zero for each group, p ≤ 0.05

an = 1,027 women with normal spines and n = 368 women with osteopenic spines

Table 5

Discordant hip pairs in women with normal and osteopenic spines—blacks (statistics should be interpreted with caution due to small samples)

Spine status

Osteoporosis

Number of women

Number of women (%) with left–right difference >LSC

Percent of women with left–right difference >LSC for group defined by spine statusa

Normal

Left hip

3

2 (66.7%)

0.7

Normal

Right hip

3

2 (66.7%)

0.7

Osteopenic

Left hip

4

0 (0.0%)

0.0

Osteopenic

Right hip

5

1 (20.0%)

1.7

The absolute number of women with left–right hip BMD differences exceeding the LSC was statistically significantly greater than zero for each group, p ≤ 0.05, except osteopenic, right hip (p = 0.3)

an = 285 black women with normal spines and n = 60 black women with osteopenic spines

Table 6

Discordant hip pairs in women with normal and osteopenic spines—whites (statistics should be interpreted with caution due to small samples)

Spine status

Osteoporosis

Number of women

Number of women (%) with left–right difference >LSC

Percent of women with left–right difference >LSC for group defined by spine statusa

Normal

Left hip

15

4 (26.7%)

0.5

Normal

Right hip

16

5 (32.2%)

0.7

Osteopenic

Left hip

25

5 (20.0%)

1.6

Osteopenic

Right hip

30

10 (33.3%)

3.3

The absolute number of women with left–right hip BMD differences exceeding the LSC was statistically significantly greater than zero for each group, p ≤ 0.05

an = 742 white women with normal spines and n = 308 white women with osteopenic spines

Discussion

To our knowledge, this is the first study to compare the prevalence of opposing hip subregion BMD difference between black and white patients. Significant left–right hip subregion BMD difference was common, being detected in 33–50% of white vs. 41–55% of black hips, with all hip subregions being affected. In both races, diagnostic agreement between the left and right hip subregions was lower when the LSC was exceeded compared to when it was not.

These findings have potential implications on the diagnosis and treatment of osteoporosis. Since the presence of significant BMD difference between the left and right hip subregions can translate into different diagnostic classification between the two hips based on WHO criteria, unilateral hip imaging can result in underdiagnosis of osteoporosis, meaning that treatment would be withheld when it should be provided. Potent therapies for osteoporosis are now available, with proven efficacy in both black and white patient populations, raising the possibility of effective hip fracture prevention in both races. Given the well-documented morbidity, mortality, and expenditure associated with hip fracture, underdiagnosis of osteoporosis can translate into significant personal and societal consequences.

Because the diagnosis of osteoporosis is based on the lowest measured T-score, our findings are expected to have implications for patients with normal and osteopenic spines only, since those with osteoporosis of the spine would be treated. In our study, osteoporosis was detected in 4% of women whose spine T-score was normal and 17% of women with spine T-scores in the osteopenic range. Of these, the LSC was exceeded in 35% of women with normal and 25% of those with osteopenic spines. Similar conclusions were reached when we analyzed our data by race, except that the sample sizes were small, especially for the black patients. These findings are similar to those reported in a white cohort in an earlier study, and underscore the superiority of bilateral vs. unilateral hip scanning in the diagnosis of osteoporosis of the hip [14].

Important aspects of hip osteoporosis have emerged from this and other studies. First, whereas earlier reports suggested that there was no added benefit to bilateral vs. unilateral hip imaging (given the high BMD correlation between the two hips), we and others [14] found significant left–right hip BMD to be common even though the mean BMD between the left and the right hips were not different. This implies that unilateral hip BMD cannot be relied on to predict BMD of the contralateral hip. Secondly, although the overall prevalence of unilateral hip osteoporotic range T-score is low in the general population, there are certain subgroups at greater risk, and would therefore benefit more that the general population from bilateral hip imaging. These include the elderly [14] and those with previous non-hip low trauma fracture [13]. Finally, the report by Hamdy et al. that the presence of unilateral hip osteoporotic range T-score affects both hips with equal frequency implies that there is no way for the clinician or radiologist to predict its location a priori [14]. The current practice allows for imaging of the non-dominant hip even though there is no biological basis for it. Indeed, there is no correlation between lower limb dominance and BMD [9, 11].

It is worth pointing out that measuring both hips identifies a slightly larger patient population than would be identified by measuring only one hip. As such, it is very likely that the relationship between hip BMD and hip fracture (and possibly the anti-fracture effects of osteoporosis treatments) is altered when bilateral hip imaging is done. Current data on the relationship between hip BMD and hip fracture risk (and fracture risk reduction with medical treatment of osteoporosis) are based on unilateral hip BMD measurement, and a lack of corresponding data for bilateral hip BMD measurement could be a reason why bilateral hip imaging is not widely accepted in clinical practice. Based on recent data, bilateral hip imaging should be particularly considered among the elderly and among patients with prior non-hip low trauma fracture.

Hip asymmetry results from factors that preferentially alter bone formation or bone resorption in one hip relative to the other. It may originate from genetic factors but disease states like stroke and unilateral leg immobilization have also been implicated [14]. In addition, artifacts such as those seen in osteoarthritis can cause apparent asymmetry of the opposing hips. Finally, technical factors related to scan acquisition and analysis have also been implicated. For example, excessive internal or external rotation of the hip can cause up to a 10% BMD variation over normal positioning, which can translate into a T-score difference of up to 1.0 [18]. In a small study examining differences between opposing femoral neck BMD differences in a normal population, there was poor correlation between femoral neck BMD difference and height, weight, and age [11].

There is no evidence that technical factors played a significant role in our findings, and scans with anatomic defects were excluded from analysis. Furthermore, we did not detect any significant difference between the left and right hip subregion precision values. Moreover, our precision values are within the ISCD recommended values [17]. Therefore, our findings lend support to the assertion that real differences exist between the opposing hip BMD in a significant percentage of the population.

An intriguing finding of our study is the observation that left–right hip subregion BMD differences were more common among blacks than among whites, and yet diagnostic agreement between the two hips was also higher among blacks than it was among whites at all subregions. This dichotomy is most probably due to the fact that the normative data used for analysis of the DXAs from black subjects were derived from white subjects, whose mean BMD is significantly different from that of the reference population. As a result, the T-scores for blacks were positively shifted compared to whites [18]. Additionally, diagnostic disagreement is more likely for those whose BMD is closer to diagnostic cutoff, and white mean BMDs are generally lower than black BMDs.

Finally, this study has several limitations worthy of mention. First, we have examined opposing hip BMD differences among black and white females aged 50 years and older. We do not know if our findings are generalizable to younger women, women of other races, and to men. Secondly, this is a cross-sectional study and our conclusions regarding fracture risk in unilateral vs. bilateral hip imaging need to be confirmed in a longitudinal study. Third, we report a higher prevalence of hip BMD difference among black than white subjects, without a corresponding increase in diagnostic disagreement. It would be interesting to see if these differences are altered when race-specific normative data are used for T-score calculation among blacks. Fourth, we included the trochanter in our analysis, which increases the number of hip subregions analyzed. We do not know if including this subregion truly enhances our capacity to capture osteoporosis. Finally, we used a GE Lunar equipment with “dual hip” acquisition, which often results in one or both hips being significantly abducted, thus affecting BMD and T-score results. We do not know the extent to which our results have been impacted by this feature of the equipment.

In conclusion, BMD difference between opposing hips is a common finding among black and white females, and the practicing clinician should be aware of its impact on the management of osteoporosis. Based on our findings, measuring only one hip for routine DXA would miss the diagnosis of osteoporosis in one of every 48 scans performed. This risk is higher among patients with osteopenic spines compared to those with normal spines (1/28 vs. 1/79, respectively) and can result in the undertreatment of osteoporosis.

Conflicts of interest

None.

Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2009