Morphologic differences between the hips of Chinese women and white women: Could they account for the ethnic difference in the prevalence of hip osteoarthritis?




Hip osteoarthritis (OA) is a common disabling disease, which has a much higher prevalence in whites than in Asians. The reasons for this ethnic difference in prevalence are unknown. Hip OA is often thought to be secondary to morphologic abnormalities. This study was undertaken to examine whether particular abnormalities predisposing to hip OA occur more frequently in whites and whether these differences in hip shape account for differences in the prevalence of OA.


A morphometric study was performed on 400 hips of 200 female participants without OA from 2 studies, the Beijing OA Study and the Study of Osteoporotic Fractures from the US. We focused on measures of hip dysplasia and impingement (lateral center–edge angle, impingement angle, acetabular slope, femoral head–to–femoral neck ratio, and the crossover sign) and compared data from the hips of Chinese and white women.


Compared with their Chinese counterparts, white women had a lower mean impingement angle (83.6° versus 87.0°; P = 0.03) and were more likely to have center–edge angles suggestive of impingement (>35°; 11% of hips in Chinese versus 23% of hips in whites, P = 0.008). In contrast, low center–edge angles suggesting dysplasia (<20°) were found more often in Chinese women (22% of hips in Chinese versus 7% of hips in whites, P = 0.005).


In a study of elderly women without signs of OA, the morphometry of impingement and asphericity was more common in the hips of white women compared with Chinese women. Our findings suggest that whites may be at higher risk of hip OA than Chinese because of morphologic findings that predispose whites to femoroacetabular impingement.

Symptomatic disabling hip osteoarthritis (OA) occurs in ∼3% of persons ages 30 years and older in the US (1). Among the factors responsible for the development of OA are genetic factors, obesity, overuse, and traumatic injury (2–6). Overt forms of deformity caused by developmental problems such as developmental dysplasia, Perthes disease, and slipped capital femoral epiphysis cause hip OA in early adulthood (7–9). Murray (9), Stulberg et al (10), and Harris (7) recognized that in addition to these major deformities, milder forms of these or other deformities may be responsible for much adult-onset hip OA. To date, only mild hip dysplasia has been demonstrated to be a risk factor for incident hip OA among elderly white women (11).

Among the mild developmental abnormalities suggested to cause hip OA is a nonspherical “pistol grip” deformity of the femoral head and neck, which can predispose to femoroacetabular impingement. Femoroacetabular impingement occurs when, during hip motion, the femoral head or neck impinges against the acetabulum, causing damage to structures inside the hip joint. Such impingement could occur because of femoral head/neck deformities (so-called “cam impingement”) or because of acetabular deformities such as overcoverage or retroversion (so-called “pincer impingement”) (12, 13). Recently, Ganz et al demonstrated that pistol grip bony deformity causes significant damage in hips even before there is radiographic evidence of OA (14, 15).

Currently, it is thought that for long-lasting, pain-free functioning of a hip joint, optimum femoral head coverage by the acetabulum is required (7, 14, 16). Lack of such coverage, as occurs in acetabular dysplasia, can lead to instability and overloading of the articular cartilage, which may then lead to joint degeneration. Conversely, overcoverage, as occurs in acetabular protrusion or acetabular retroversion, may also cause symptoms of pain in the hip due to the pincer-type internal impingement between the femoral head–neck junction and the acetabulum. Even with optimal acetabular coverage, a femoral head–neck junction that is too broad or aspherical may cause symptoms of pain due to a cam-type impingement. These subtle anatomic abnormalities often occur in combination.

Hip OA has a heterogeneous geographic distribution, perhaps accounted for by racial and ethnic disparities in occurrence (17, 18). Large-scale population-based studies that assessed hip pain and obtained radiographs of participants have shown that the prevalence of hip OA in Chinese populations is approximately one-tenth the prevalence of hip OA in groups of white subjects of the same age and sex (5, 19). Even the rate of total hip replacements is much higher in whites than in Asians, as shown in a mixed ethnic population from Hawaii (20, 21). A previous study showed that whites in San Francisco had an incidence of total hip replacement for OA that was 15–20 times the rate for Asians there (17). Hip OA is most common in Western countries such as the US and countries in northern Europe. The hip OA prevalence difference is not mirrored by similar prevalence differences for other joints, suggesting that there is not a similar difference in generalized OA. Knee OA is actually more prevalent in Chinese subjects compared to white subjects (22), and there is a similar prevalence of hand OA (19).

The ethnic differences in hip OA prevalence could be explained by subtle morphologic differences in the hip joint, with morphologic features that are likely to predispose to hip OA being more common in white populations. The aim of this study was to compare morphometric differences in non-OA hip joints between Chinese and white cohorts, using radiographs obtained in population-based surveys of elderly women from Beijing, China and elderly white women from the US. We hypothesized that subtle differences in acetabular coverage and orientation and femoral head sphericity could be detected on radiographs and may be responsible for the high prevalence of hip OA seen in white relative to Asian populations. We focused on hips without OA because the bony remodeling that occurs with disease development may mimic some of the morphologic changes of interest. We limited our study to women because we had used identical protocols to acquire hip radiographs in studies of white women and Chinese women, making it unlikely that any differences we saw in morphologic features would be due to different radiographic techniques.


Study cohorts.

The Study of Osteoporotic Fractures (SOF) (11) is a multicenter cohort study of risk factors for fractures in 9,704 white women ages ≥65 years recruited from September 1986 to October 1988. Details of this study have been described previously (23, 24). At the baseline SOF examination, standard anteroposterior (AP) radiographs of the pelvis were obtained for all subjects, with the feet fixed at 15–30° of internal rotation and the x-ray beam centered on the symphysis pubis (25, 26).

The Beijing OA (BOA) Study, consisting of door-to-door recruitment of all persons ≥60 years old from randomly selected neighborhoods in Beijing, was carried out to investigate the prevalence of OA of the hands, knees, and hips among elderly residents in Beijing. As part of the study, standard supine AP radiographs of the pelvis were obtained for all subjects. The sampling frame and recruitment approach have been described previously (22). Pelvis radiographs were obtained at Peking Union Medical College Hospital (Beijing, China) using the SOF protocol described above.

In these studies, hip radiographs were read using a standard approach for OA (27). The BOA Study and SOF radiographs were read by the same panel of readers using the same approach (7, 24). We selected hip radiographs in which no features of OA were present in either hip. For both cohorts, none of the following features could be present: osteophyte score of ≥2 on a 0–3 scale, joint space narrowing score of ≥2 on a 0–3 scale, minimal joint space <2.5 mm, or any combination of ≥2 individual radiographic features (sclerosis, cysts, osteophytes, and narrowing scored semiquantitatively) (28).

We measured morphometric parameters on 100 such pelvic radiographs of persons without radiographic OA in either hip that were randomly selected from each of the study cohorts (a total of 200 radiographs containing 400 hip joints). The cohort of Chinese women was obtained from the BOA Study, while the cohort of white women was obtained from the SOF. Radiographs from the 2 studies were mixed together to limit the effect of measurement drift on comparative results.

Morphologic assessment of hips.

Femoral head coverage by the acetabulum was assessed by the center–edge angle as defined by Wiberg (29) and the acetabular slope of Tönnis (also called Tönnis angle) (30) (Figure 1). Acetabular retroversion was assessed by the crossover sign (31). Femoral head asphericity was assessed using the impingement angle (32) (Figure 2) and the femoral head ratio as defined by Murray (9) (Figure 3). Details regarding the radiographic measurement technique are provided in the figure legends. We did not include figures for center–edge angle and crossover signs, since they are widely used. Center–edge angle is the angle between a vertical line extending up from the femoral head center and another line also originating at the femoral head center that passes to the edge of the acetabulum. A positive crossover sign (33–38) exists when the anterior and posterior walls of the acetabulum cross over the femoral head. (A hip with a normal pelvic inclination should have the anterior and posterior lips join at the edge of the acetabulum.) A positive crossover sign may indicate acetabular retroversion or anterior overcoverage.

Figure 1.

Acetabular slope of Tönnis (16, 30, 55, 56). A horizontal line is drawn connecting the femoral head centers. The angle formed by the horizontal line and the line connecting the edges of the acetabular sourcil is the acetabular slope of Tönnis.

Figure 2.

Impingement angle (32, 37). A perfect circle that encompasses the inferior and superior portions of the femoral head is drawn. A line passing through the center of both femoral heads is found and used as a reference line. An angle is formed by drawing a line perpendicular to the horizontal reference line and a line connecting the center of the femoral head to the first part of the lateral femoral head, which is outside the perfect circle. An angle of <70° is defined as pathologic, assuming that a neck shaft angle in adults is on average 130°.

Figure 3.

Femoral head ratio of Murray (9). The ratio MH:LH of the femoral head width lying on each side of a line drawn through the middle of the femoral neck and middle of the line connecting the greater and lesser trochanters is determined. A larger ratio implies a more “pistol grip” shape of the femoral head. The mean femoral head ratio is 0.92 (range 0.66–1.36) in women and 1.17 (range 0.62–1.92) in men.

The radiographs were mixed and read independently by 2 orthopedic surgeons (MD and YJK). Prior to reading, the readers were trained on the measurement protocol. Both the interreader reliability and intrareader reliability were high, with intraobserver and interobserver intraclass correlation coefficients ranging from 0.90 to 0.97.

In addition to the measures of dysplasia and impingement described above, we used accepted cut points of continuously measured angles to determine whether the hips of Chinese subjects differed from those of white subjects in terms of measures of either dysplasia or impingement. We divided the criteria for thresholds that constituted abnormal hip morphology into measures of hip dysplasia (acetabular undercoverage) and measures of impingement. Two different sets of criteria (one that was less strict and one that was more strict) were used. The criteria for hip dysplasia included a center–edge angle of <20° (29) or <16° (31) and an acetabular slope of >10° or >15° (16, 30). The criteria for impingement/overcoverage included a center–edge angle of >35° or >40° (suggests pincer-type impingement) (16) and an acetabular slope of Tönnis of <0°, which is considered to be acetabular overcoverage (16) (suggests pincer-type impingement). Impingement angle thresholds were based on the thresholds used for the alpha angle of Notzli and normal neck shaft angle (39). A hip with an impingement angle of <70° was considered to have a cam-type impingement. The impingement angle is related to but different from the alpha angle of Notzli. The impingement angle also takes into account proximal femoral varus and valgus angulation and is decreased in varus neck angulation, which, for the same proximal femoral deformity with the same alpha angle, accentuates impingement.

Statistical analysis.

We calculated mean values of center–edge angle, acetabular slope, and femoral head ratio for Chinese and white subjects. We carried out a linear regression to evaluate whether continuously defined center–edge angle, acetabular slope, and femoral head ratio were different between Chinese and white subjects. We also calculated the proportion of measures of impingement (center–edge angle [>35°], impingement of acetabular slope [<0°], and femoral head ratio [>1.35]) among Chinese and white subjects and compared the difference in the proportion of each measure between Chinese and white subjects using a logistic regression model. The dependent variable was the morphometric measure; the independent variable was ethnic group. Since data were available for both left and right hips for each subject, we used generalized estimating equations with robust standard errors to account for the correlation between 2 sides within a subject.


The subjects selected for our study were similar in age, educational attainment, and body mass index (BMI) to the larger cohorts from which they were drawn (Table 1). The hip radiographs of one of the Chinese subjects were not clear enough for the hips to be measured, leaving 99 subjects from Beijing. All of the subjects selected from the SOF (and almost all of the subjects in the parent study) were white. The mean age of the subjects in our sample was 71.0 years in the white group and 70.7 years in the Chinese group (P = 0.66 by t-test). The mean BMI was 26.5 in the SOF and 25.5 in the BOA study (P < 0.001). We selected subjects who did not have radiographic hip OA in either hip. All hips had Croft scores of 0 or 1. Croft grade 1 represents possible osteophytes or possible narrowing. The hips with a Croft score of 1 had either medial narrowing scores or acetabular osteophyte scores of 1, neither of which is typically indicative of OA.

Table 1. Characteristics of the women in the SOF and the BOA Study*
 All SOF subjects (n = 9,704)SOF subjects included in this study (n = 100)All BOA subjects (n = 1,505)BOA subjects included in this study (n = 99)
  • *

    SOF = Study of Osteoporotic Fractures; BOA = Beijing Osteoarthritis; NA = not available; BMI = body mass index.

  • Percent of subjects with pain for at least 1 month in the past 12 months.

Age, mean ± SD years71.7 ± 5.371.0 ± 4.867.5 ± 6.170.7 ± 5.2
Education ≥12 years, no. (%)7,461 (76.9)82 (82)168 (11.2)NA
BMI, mean ± SD kg/m226.5 ± 4.726.8 ± 4.225.5 ± 4.222.5 ± 4.2
Hip pain, %30.43112.77.2

As shown in Table 2, the mean center–edge angle of Wiberg for white women (30.4°) was significantly greater than that for Chinese women (25.5°) (P < 0.0001). Similarly, Chinese women had a higher acetabular slope of Tönnis than white women, suggesting that the hips of the Chinese women were less covered than those of the white women.

Table 2. Morphologic differences between the hips of Chinese and white women, adjusted for height, age, and body mass index*
Radiographic measurementWhite subjectsChinese subjectsP
  • *

    Except where indicated otherwise, values are the mean degrees.

Impingement angle83.687.00.03
Femoral head ratio0.890.890.99
Acetabular slope3.86.6<0.0001
Center–edge angle30.425.5<0.0001
Femoral head ratio >1.35, %

No difference was observed in the femoral head ratio of Murray between the 2 ethnic groups. However, the mean impingement angle was 83.6° in the white women and 87° in the Chinese women (P = 0.03), suggesting that the hips of the white women were more aspherical.

The threshold values for center–edge angle, acetabular slope of Tönnis, impingement angle, and femoral head ratio were used to define the percentage of hips that were thought to be normal versus abnormal (Table 3). Using center–edge angle >35°, acetabular slope <0°, and impingement angle <70° as the criteria, hips that would be predisposed to impingement were more commonly seen in the white women. Of these, high center–edge angle and low acetabular slopes point to pincer-type impingement, and low impingement angle is compatible with cam-type impingement. In the hips of the Chinese women, a center–edge angle of <20° and an acetabular slope of >10°, factors suggesting undercoverage, were more prevalent. However, when the more strict criteria of lateral center–edge angle of <16° and acetabular slope of >15° for undercoverage were used, there was no significant increase in the number of Chinese women versus the number of white women with hips with undercoverage. The crossover sign (acetabular retroversion) was slightly more prevalent in Chinese than in whites, but the difference was not statistically significant (Table 3).

Table 3. Comparison of hip anatomic features in Chinese and white women*
 No. of hipsNo. (%) with featureAdjusted prevalence ratio (95% CI)P
  • *

    The frequencies of impingement and instability and other hip abnormalities were determined using accepted cut points in radiographs of the hips of 100 white subjects and 99 Chinese subjects. 95% CI = 95% confidence interval.

  • The prevalence ratios were derived from logistic regression in which the dependent variable was the morphologic abnormality dichotomized and the independent variables were ethnicity, age, height, and body mass index.

Center–edge angle    
 >35° impingement    
  Whites20046 (23.00)1.0 
  Chinese19822 (11.11)0.4 (0.2–0.8)0.008
 >40° impingement    
  Whites20018 (9.00)1.0 
  Chinese1988 (4.04)0.3 (0.1–1.0)0.051
 <20° instability    
  Whites20014 (7.00)1.0 
  Chinese19843 (21.72)2.7 (1.4–5.3)0.005
 <16° instability    
  Whites2005 (2.50)1.0 
  Chinese19813 (6.57)2.2 (0.7–6.5)0.16
Acetabular slope    
 <0° impingement    
  Whites20026 (13.00)1.0 
  Chinese1986 (3.03)0.2 (0.1–0.5)0.002
 >10° instability    
  Whites20015 (7.50)1.0 
  Chinese19837 (18.69)2.2 (1.1–4.6)0.04
 >15° instability    
  Whites2002 (1.00)1.0 
  Chinese1989 (4.55)4.7 (0.8–26.5)0.08
Impingement angle <70°    
 Whites19924 (12.06)1.0 
 Chinese1987 (3.54)0.3 (0.1–1.0)0.04
Femoral head ratio of Murray >1.35    
 Whites1998 (4.02)1.0 
 Chinese1966 (3.06)1.0 (0.3–3.1)0.99
Acetabular retroversion    
 Whites200162 (81.00)1.0 
 Chinese198172 (86.87)1.0 (0.9–1.1)0.95


The prevalence of hip OA has been found to be much higher in white than in Asian populations in several studies (5, 12, 13, 40). However, the reasons for these differences are unclear. What puts whites at risk? This study demonstrated that in white women, the hips are more often overcovered and less spherical than in Chinese women, suggesting that white women are more susceptible to femoroacetabular impingement. Of note, mild dysplasia was not more common in the hips of white women. If anything, it was more prevalent among the Chinese subjects.

Recent studies suggest that structural abnormalities in terms of acetabular undercoverage, acetabular overcoverage, and femoral head asphericities are important factors in the development of hip OA. In particular, femoroacetabular impingement syndrome is increasingly recognized as a factor in the development and progression of hip OA (14, 15, 41). Although nonspherical head shape has been suggested as a risk factor for hip OA in one cross-sectional study (42), it remains unclear as to whether morphologic variants thought to lead to impingement predispose to hip OA. Our data showing a higher prevalence of the morphologic features related to femoroacetabular impingement syndrome in whites who, in turn, have higher rates of hip OA, provide circumstantial evidence that these morphologic changes can also be implicated as causes of hip OA.

Acetabular dysplasia can lead to premature OA (11, 43), but the importance of milder forms of this anatomic abnormality as a major cause of OA has been debated. Some studies suggest that as much as 40% of hip OA may be due to hip dysplasia (7, 9, 44). However, large prospective studies have shown hip dysplasia to be a minor risk factor for hip OA in women (11). Other studies showed that in Chinese subjects the prevalence of hip dysplasia was as high as that in white subjects, so that the low rate of OA of the hip is likely not caused by a rarity of hip dysplasia in the Chinese population (45). As in our study of Chinese, a study of the Japanese population showed a higher prevalence of dysplasia but a low prevalence of hip OA (13).

A possible explanation for the protection against hip OA in Asian populations is that Chinese subjects have a more shallow acetabulum, resulting in more space for motion of the hip. In addition, our results show that in Chinese women the impingement angle is significantly greater, so the femoral head is more spherical than in white women. The shallow acetabulum and the more spherical femoral head may protect against impingement syndrome and in turn, against OA. Biomechanical modeling of hips with various amounts of coverage and femoral asphericity suggests that less acetabular coverage protects a hip with an aspherical femoral head from impinging (46).

Femoroacetabular impingement has 2 overlapping subtypes, cam and pincer deformities, and of these, pincer deformities may be more prevalent among women (47). Our measures of pincer deformity (large center–edge angles and acetabular slope impingement) showed a higher prevalence of overcoverage in white women. Interestingly, even though cam impingement is not necessarily as common in women as in men, even our morphometric measure of this, the impingement angle measure, showed that cam deformities were more prevalent in white than Chinese women.

What else could account for the marked difference in hip OA prevalence between these ethnic groups? Ethnic differences in BMI may explain the variations (48). An increased BMI increases the risk of hip OA, but the association of obesity with hip OA is weaker than that with knee OA (49), with at most a 2-fold increase in risk among those who are obese (compared with the 10-fold higher risk of hip OA in white subjects compared to Chinese subjects) (50). The modest effect of obesity on hip OA risk could not be expected to account for this marked ethnic difference (19).

Genetic factors may also explain ethnic differences in the occurrence of hip OA. Hip OA has been shown to have a heritability approaching 60% in some studies (28, 51–54). In fact, the heritability of hip OA is likely to be accounted for by anatomic/morphologic factors. For example, hip OA runs in families separately from OA in other joints (53). One compelling explanation for such heritability would be the inheritance of a morphologic variant that would predispose some members of a family to hip OA. Such a variant could also account for some of the ethnic difference in hip OA (21, 53).

In this study, we found significant differences for acetabular slope, center–edge angle, and impingement angle but not for the femoral head ratio of Murray, another measure that would point to femoroacetabular impingement. The femoral head–to–femoral neck ratio may not be measured accurately enough on an AP hip radiograph to distinguish between morphologic differences in the hips of Chinese and white subjects.

There were important limitations to our study. For one, it was not longitudinal. Ideally, an investigation of the causes of hip OA implicating the morphologic findings we discussed should track these findings longitudinally and document that they cause OA. We did not do that, although numerous studies now point to femoroacetabular impingement and its measures as probable causes of hip OA and clear-cut causes of disease progression.

A prospective study of the morphometric causes of hip OA in the Chinese population is probably not feasible given the rarity of the disease. In a large-scale survey (of 1,800 older subjects with pelvis radiographs), we could not find enough cases of hip OA to carry out risk factor studies, and any retrospective study would be limited since the bony remodeling that occurs as a consequence of hip OA would make premorbid hip anatomy impossible to assess in cases with extant OA. Given the rarity of hip OA in the Chinese population and the lack of validity of retrospective approaches, our approach of examining unaffected persons to evaluate predisposing hip morphology may be among the best possible approaches to gain insight into the reasons for this ethnic difference. Studies using new imaging modalities that are more sensitive to mild changes of OA may be an alternative approach.

Our study is also potentially limited by the fact that it included only elderly women. We had radiographs obtained using comparable techniques only for women from cohorts in the US and China. We note that the prevalence of hip OA is low in both sexes in China and much higher in both sexes in the Western world and that impingement is not a sex-specific finding. Also, similar pelvic radiographic views were obtained for female subjects. Studies in men need to be carried out. In addition, to the extent that morphometric abnormalities would have been expected to exert their influence on disease development at an earlier age than that of the women included in our study, we might have missed important effects.

In summary, we have demonstrated that white women have more overcoverage and less sphericity of the hips than Chinese women. This may put them at higher risk of hip OA compared with Chinese women. These findings suggest more frequent femoroacetabular impingement in the hips of white subjects. It is possible that this impingement may cause hip OA and account for the high rate of hip OA in white subjects.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Felson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Kim, Zhang, Nevitt, Felson.

Acquisition of data. Dudda, Kim, Nevitt, Xu, Goggins.

Analysis and interpretation of data. Dudda, Kim, Zhang, Nevitt, Niu, Doherty, Felson.