Very low prevalence of hip osteoarthritis among Chinese elderly in Beijing, China, compared with whites in the United States: The Beijing osteoarthritis study




To compare the prevalence of osteoarthritis (OA) of the hip among elderly persons in China and the US.


We recruited a population-based sample of 1,506 persons (82% of those enumerated) ages ≥60 years living in Beijing, China. Subjects answered questions about joint symptoms and underwent radiography of the pelvis. Radiographs of the Beijing subjects were intermingled with hip radiographs of white women ages ≥65 years from the Study of Osteoporotic Fractures (SOF) and white men and women ages 60–74 years from the First National Health and Nutrition Examination Survey (NHANES-I) and were then interpreted. Radiographic hip OA was defined as the presence of 1 of the following 3 findings in either hip: minimum joint space of ≤1.5 mm, definite osteophytes and joint space narrowing, or ≥3 radiographic features of OA. Symptomatic hip OA was defined as both radiographic OA and hip pain.


The crude prevalence of radiographic hip OA in Chinese ages 60–89 years was 0.9% in women and 1.1% in men; it did not increase with age. Chinese women had a lower age-standardized prevalence of radiographic hip OA compared with white women in the SOF (age-standardized prevalence ratio 0.07) and the NHANES-I (prevalence ratio 0.22). Chinese men had a lower prevalence of radiographic hip OA compared with white men of the same age in the NHANES-I (prevalence ratio 0.19). There were no cases of symptomatic hip OA in the Chinese men and only 1 case in the Chinese women; 35 cases were expected in both sexes.


This is the first population-based study of hip OA in China to use standardized radiographic methods and definitions. We found that hip OA was 80–90% less frequent than in white persons in the US. Identification of the genetic and environmental factors that underlie these differences may help elucidate the etiology and prevention of hip OA.

Osteoarthritis (OA) is a leading cause of disability among older persons in white populations of Western societies (1). In the US, hip OA affects ∼5% of the population over the age of 60 years (2) and is often highly disabling and progressive. OA accounts for most total hip replacements in Western countries. Little is known of its causes.

Major differences in the racial or geographic distribution of disease may provide valuable clues about potential etiologic factors. Surveys of OA prevalence have been limited primarily to northern and western Europe and the US (3). There have been few population-based studies of OA in Chinese or other Asian populations. The studies that do exist have suggested that while knee OA is as prevalent or more prevalent in Chinese as in elderly white persons (4), the prevalence of hip OA may be much lower among Chinese (5–9). In general, however, only 1 study has used standardized and clearly described criteria for grading radiographic findings (8), and none has examined hip symptoms or evaluated a population sample. The absence of standardized data and of internationally accepted standards for characterization of hip symptoms or radiographs has impeded a definitive determination of whether genuine differences in prevalence exist.

To compare the prevalence of hip OA in China and the US, we conducted a population-based survey among elderly persons in Beijing, China, using the same instruments and protocols to ascertain and define hip OA as in a previous population-based study of elderly persons in the US (10, 11). Using these methods, we also assessed the prevalence of hip OA among subjects in the First National Health and Nutrition Examination Survey (NHANES-I) (12) in the US.


The Beijing OA (BOA) study.

We studied a random sample of men and women ages ≥60 years from 3 central districts of Beijing, China. The sampling frame and recruitment approach have been described in detail elsewhere (4). Pertinent elements of recruitment and assessment included door-to-door recruitment of all persons ages ≥60 years in randomly selected neighborhoods in central Beijing, and home administration of a survey about joint symptoms. Subjects were invited to the central examination site at Peking Union Medical College Hospital, and pelvis radiographs were obtained.

The Study of Osteoporotic Fractures (SOF).

The SOF is a multicenter cohort study of risk factors for fractures in a cohort of 9,704 primarily (>99%) white women ages ≥65 years who were recruited from September 1986 to October 1988 using population-based lists in 4 cities: Baltimore, Pittsburgh, Minneapolis, and Portland. Detailed descriptions of this study have been published elsewhere (10, 13). The baseline examination included pelvis radiographs and questions about hip pain. In the present study, we used previously performed readings of baseline pelvis radiographs of ∼8,000 women (11, 14).


To provide a comparable population-based sample of hip radiographs in elderly white men and women, we obtained pelvis radiographs from the NHANES-I, a US National Center for Health Statistics survey that was conducted from 1971 to 1975. The NHANES-I was a multistage probability sample chosen to be representative of the US population (12). We interpreted pelvis radiographs of 316 white subjects (156 men and 160 women) ages 60–74 years, which represented all the radiographs of persons this age from sampling units that could be found in the NHANES archives.

Survey instrument.

Survey questions in English were translated into Chinese and back into English using methods described previously (4). The specific question assessing hip symptoms was derived from the SOF and NHANES-I: “Have you ever had pain in or around either hip joint, including the buttock, groin, or either side of the upper thigh, on most days for at least a month?” This was followed by questions about the last occurrence and the severity of the pain. We also used a simpler question to ask about hip pain: “In the past 12 months, have you had pain, aching, or stiffness lasting at least a month in your hip?” Results were essentially the same using either question.

Examination protocols.

Radiography protocols from the SOF were used in the BOA study. Protocols were translated into Chinese, and two investigators (MCN and DTF) held on-site training for radiography technicians. A supine anteroposterior pelvis radiograph was obtained on all subjects. The subjects' feet were placed at 15–30° of internal rotation, the central ray was centered on the symphysis pubis, and a 101-cm focal film distance at 70–80 kilovolt (peak) was used. The NHANES-I radiograph was a supine anteroposterior view centered around L3 that included the pelvis and lumbar spine.

Interpretation of radiographs.

The radiographs from the BOA and NHANES-I studies were read by two rheumatologists (NEL and MCH) using the same methods that they had previously used for reading the pelvis films from the SOF. An initial interpretation by one reader (NEL) identified films with possible joint space narrowing or osteophytes (42% of films from the BOA and 46% from the NHANES-I); these were then reevaluated by the two readers jointly in a consensus session, with the initial scores available. Prior to each session, readers reviewed a batch of pelvis films from the SOF, with the previous readings provided, so as to calibrate the new with earlier readings. In addition, we commingled samples of previously scored films from the SOF to test reliability.

Using calipers and reticule, we measured the minimum joint space (in millimeters) according to a previously described method (15, 16). Pelvis radiographs were scored for individual radiographic features of OA based on atlas photographs (17, 18). Each hip was scored for joint space narrowing (range 0–4, where 0 = none and 4 = complete loss of joint space) at 2 locations (superolateral and superomedial) and for osteophytes (range 0–3, where 0 = none and 3 = severe) at 4 locations (superior femoral, lateral acetabular, and inferior femoral and acetabular). Subchondral sclerosis, cysts, and femoral neck deformity were also graded (range 0–3, where 0 = none and 3 = severe). A global grade of OA (range 0–4) for each hip was assigned, based on the number of individual radiographic features present (11, 16).

We compared the consensus results from the readings of 200 hips from the SOF, which were done for the present study and 5 years previously. Intraclass correlations for the measures of joint space and osteophytes ranged from 0.80 to 0.90. Kappas for dichotomous values (e.g., osteophytes and joint space narrowing scores ≥2 and minimum joint space ≤1.5 mm) ranged from 0.68 to 0.79 and, for the definition of radiographic OA used in this report (see below), κ = 0.74 (95% confidence interval 0.61–0.87). Disagreements were not in any particular direction, which suggests that there was no likelihood of bias in the estimates.

Definitions of hip OA.

A subject was characterized as having radiographic hip OA if either hip met at least 1 of the following 3 radiographic criteria: 1) a minimum joint space of ≤1.5 mm, 2) an osteophyte of grade 2 or higher in any location and either (a) superolateral joint space narrowing of grade 2 or higher or (b) superomedial joint space narrowing of grade 3 or higher, or 3) any 3 or more radiographic features of OA. Symptomatic hip OA required the presence of these radiographic findings plus 1 month of hip pain during the previous 12 months.

A review of the discharge abstract or operative record was performed on 125 of the 134 subjects in the SOF who had a total hip arthroplasty (THA) at baseline. Subjects were classified as having hip OA if the record stated that treatment was for OA of the hip and there was no mention of avascular necrosis or hip fracture. These patients are included in our prevalence data for the SOF. No subject with a THA was encountered in the NHANES-I, and only 4 subjects (3 men and 1 woman) had a THA in the BOA study after excluding those with a previous hip fracture. Since we did not obtain medical records for patients with THA in the BOA study, we tested the effect of assuming that they all had OA.

We examined the sensitivity of our results to several alternative definitions of hip OA, including alternative cut points for minimum joint space based on estimated trimmed means and SDs of normal values in each race- and sex-specific sample (19). Race- and sex-specific cut points were defined at 2.5 adjusted SDs below the adjusted mean.

Statistical analysis.

We excluded subjects who self-reported rheumatoid arthritis and whose hand radiographs revealed typical radiographic evidence of rheumatoid arthritis in the BOA (n = 4) and SOF (n = 99) studies (hand radiographs were available in these studies), as well as subjects who had a history of hip fracture (137 in the SOF, 10 in the BOA study, and 2 in the NHANES-I). To compare the prevalence of hip OA between Chinese women in the BOA study and white women in the SOF, we confined our analysis to subjects who were ages 65–89 years, which are age groups common to both studies. For comparisons with the NHANES-I, we similarly confined the analysis to subjects who were ages 60–74 years. These ages were then divided into 5 groups: 60–64 years, 65–69 years, 70–74 years, 75–79 years, and 80–89 years. The age-specific prevalence of hip OA in the SOF and NHANES-I was then applied to the age distribution in the BOA study, and an age-standardized prevalence ratio and its 95% confidence interval (20) were calculated.


Of the age-eligible subjects contacted during door-to-door canvassing in Beijing from January 1998 to March 2000, a total of 1,646 (89.6%) completed a home interview. Among subjects who completed a home interview, 1,506 (91.5%) had a clinical examination and an evaluable pelvis radiograph. Compared with those who had an examination and a radiograph, subjects who had only the home interview were older (71.7 versus 67.6 years; P < 0.001) and more likely to be women (71.4% versus 59.0%; P < 0.01), but there was no difference in the prevalence of hip pain (13.3% versus 10.1%; P = 0.25). On average, Chinese women were shorter and thinner than their US counterparts (Table 1). The mean minimum joint space was slightly larger in Chinese men compared with white men, but there were no consistent differences among women.

Table 1. Characteristics of the study subjects*
BOA (n = 614)NHANES-I (n = 156)BOA (n = 878)SOF (n = 7,998)NHANES-I (n = 158)
  • *

    BOA = Beijing Osteoarthritis study; NHANES-I = First National Health and Nutrition Examination Survey; SOF = Study of Osteoporotic Fractures; BMI = body mass index; MJS = minimum joint space.

  • Subjects in the SOF were 65–69 years of age.

  • Subjects in the NHANES-I were 70–74 years of age.

  • §

    P < 0.05 versus BOA group, adjusted for age.

  • Excludes hips with osteoarthritis, as determined by radiographic criteria.

Age group, no. (%)     
 60–69 years382 (62.2)117 (75.0)604 (68.8)3,580 (44.8)114 (72.2)
 70–79 years190 (30.9)39 (25.0)232 (26.4)3,769 (47.1)44 (27.9)
 80–89 years42 (6.8)42 (4.8)649 (8.1)
Height, mean ± SD cm166 ± 6172 ± 7§154 ± 5159 ± 6§159 ± 6§
Weight, mean ± SD kg70 ± 1175 ± 12§61 ± 1067 ± 12§69 ± 12§
BMI, mean ± SD25.2 ± 3.425.4 ± 3.826.0 ± 4.026.5 ± 4.6§27.1 ± 5.0§
BMI ≥30.0, no. (%)50 (8.1)14 (9.0)127 (14.5)1,609 (20.1)40 (25.2)
Hip pain in past year, no. (%)34 (5.6)8 (5.1)96 (11.1)2,397 (30.3)§18 (11.0)
MJS, mean ± SD mm3.03 ± 0.512.93 ± 0.49§2.84 ± 0.443.04 ± 0.59§2.76 ± 0.42§
Total hip replacement, no. (%)3 (0.5)0 (0.0)1 (0.1)133 (1.7)§0 (0.0)

The overall crude prevalence of radiographic hip OA in Chinese ages 60–89 years was ∼1% in both men and women (Table 2). Prevalence tended to increase with age in Chinese men, but not Chinese women. The crude prevalence of radiographic hip OA in whites ranged from 3.8% to 5.5% (Table 3). The age-standardized prevalence of hip OA in Chinese was 80–90% less than that observed in the white male and white female subjects (P < 0.001).

Table 2. Age-specific prevalence of hip osteoarthritis (OA) among elderly persons living in Beijing, China
Age groupMenWomen
No. of subjectsNo. (%) with radiographic OANo. (%) with symptomatic OANo. of subjectsNo. (%) with radiographic OANo. (%) with symptomatic OA
60–69 years3823 (0.8)0 (0.0)6046 (1.0)1 (0.2)
70–79 years1903 (1.6)0 (0.0)2322 (0.9)0 (0.0)
80–89 years421 (2.4)0 (0.0)420 (0.0)0 (0.0)
All subjects6147 (1.1)0 (0.0)8788 (0.9)1 (0.1)
Table 3. Prevalence of radiographic hip OA in Chinese living in Beijing, China, compared with that in Caucasians in the NHANES-I and SOF studies*
GroupCrude prevalence of radiographic hip OA (%)Age-standardized prevalence ratio (95% CI)
  • *

    OA = osteoarthritis; NHANES-I = First National Health and Nutrition Examination Survey; SOF = Study of Osteoporotic Fractures; BOA = Beijing Osteoarthritis study; 95% CI = 95% confidence interval.

 Ages 60–74 years0.84.50.19 (0.05–0.49)
 Ages 65–89 years0.45.50.07 (0.01–0.25)
 Ages 60–74 years0.83.80.22 (0.08–0.48)

The crude prevalence of symptomatic hip OA in women in the SOF was 2.2%. There were too few subjects in the NHANES-I sample to provide reliable estimates of symptomatic hip OA prevalence. No Chinese men and only one Chinese woman had symptomatic hip OA (Table 2), compared with an expected total of 35 in both sexes based on the SOF prevalence in women.

In sensitivity analyses, when we assumed that all THAs in Chinese subjects were for hip OA, this had a negligible effect on the prevalence in women, but it increased the crude prevalence to 1.2% in men ages 60–74 years (standardized prevalence ratio BOA men versus NHANES-I 0.25; 95% confidence interval 0.09–0.55). Additional analyses included classifying as OA any hips with grade 2 superomedial narrowing plus osteophytes and using cut points for minimum joint space based on race- and sex-specific distributions. This did not substantively change our results, with age-standardized prevalences being 50–90% lower in elderly subjects in Beijing compared with the US.

The age-standardized prevalence of selected individual radiographic features suggested that a minimum joint space of ≤1.5 mm, a joint space less than or equal to the race- and sex-specific cut point, superolateral joint space narrowing of grade 2 or higher or superomedial narrowing of grade 3 or higher, and femoral osteophytes all occurred much less frequently in Chinese compared with whites. In contrast, more nonspecific and prevalent findings, such as a minimum joint space ≤2.5 mm, grade 2 superomedial narrowing, and acetabular osteophytes, were common in both the Beijing and US subjects (present in 10–30% of hips) and were not consistently different between groups.


In this first population-based study of radiographic hip OA in China, the age-standardized prevalence of radiographic hip OA among Chinese elderly subjects was 80–90% lower than in white US elderly subjects. In Beijing, only 1 of >1,500 subjects had symptomatic hip OA. To our knowledge, this is the first study in which the comparison of hip OA prevalence between Chinese and white subjects was incorporated into the study design, cases were sampled from the general population, and carefully standardized methods were used to assess OA in both populations. Thus, the difference in hip OA prevalence between the two racial groups in our study is unlikely to be explained by sampling or assessment bias.

Although studies of symptomatic OA are rare, our results are consistent with previous studies of radiographic hip OA in Asian samples, none of them population-based and with uniform evaluations across populations. Among a hospital and clinic population of Hong Kong Chinese ages ≥55 years, Hoaglund and colleagues (6) reported that Kellgren/Lawrence grade 3 or 4 OA was uncommon in the hip (0.3%), much lower than in a previously studied English population the same age (5.5%). Hoaglund and colleagues have also reported that rates of hip replacement for OA among Chinese- and Japanese-surnamed Americans living in San Francisco (7) and Hawaii (9) are 65–90% lower than in whites.

It is possible that there are differences in hip OA prevalence between Beijing and other regions of China, especially rural areas. Studies of hip OA in other regions of China are needed to confirm the generalizability of our findings. Lau and colleagues (8) reported a low prevalence of radiographic hip OA in urban Hong Kong men ages 60–75 years undergoing intravenous urograms in which a pelvis film was obtained. Using definitions of radiographic OA similar to ours, hip OA was 65–85% less common in the Chinese men than in previously studied British men (21). A similar low prevalence of radiograph hip OA was found in elderly persons living in rural Japan (21).

Although not compared directly with the prevalence of hip symptoms in other societies, studies suggest that hip pain is uncommon in Chinese elderly, consistent with the near absence of symptomatic hip OA in the Beijing sample. In a random sample of Hong Kong men and women ages ≥70 years who were asked about musculoskeletal pain in the previous 12 months, too few reported hip symptoms to be enumerated in the results, yet knee pain was reported by more than one-third of them (22). In a study of symptomatic OA among adults in 3 regions in Taiwan, hip OA comprised <1% of all cases of OA (23). Similarly, in a survey of joint symptoms in >4,000 adults living in a suburban village near Beijing, hip pain was infrequently elicited during a range of motion examination (3%), but knee pain was common (36%) (24).

It is possible, but unlikely, that some of the Chinese subjects who reported hip pain had OA that was not detected by radiography. Using less stringent radiographic criteria for hip OA in the Beijing subjects did not increase the frequency at which hip pain and radiographic OA occurred together. In addition, 80% of those reporting hip pain in Beijing described it as mild and said that it did not interfere with their usual activities. It is difficult to localize the hip joint as the anatomical source of pain in the hip region, and pain referred from the spine and structures near the hip, such as trochanteric bursa, is common.

Possible explanations for the low risk of hip OA in Chinese compared with whites do not include a low risk of OA in general. In the present sample of Beijing elderly, the age-adjusted prevalence of radiographic knee OA in men is about the same as, and in the women it is higher than, the age-adjusted prevalence of radiographic knee OA among whites in the Framingham study (4). To our knowledge, the only study to compare the prevalence of hand OA in Chinese and white subjects reported a similar prevalence in Hong Kong men and British men and a somewhat lower prevalence in Hong Kong women than in British women (6). Another often suggested explanation for the low rate of hip OA in China is a low rate of subclinical acetabular dysplasia, in which the acetabulum is unusually shallow, among Asians; some investigators have postulated that a shallow acetabulum is a frequent cause of hip OA among white subjects (6, 25, 26). However, recent studies have found that shallow acetabula are actually more common in elderly Chinese and Japanese than in elderly whites (8, 21). It is possible that other racial differences in the anatomy of the hip play a role in the contrasting risks of OA and deserve further investigation.

Other potential explanations for a racial difference in hip OA include differences in physical activities, a lower prevalence of obesity in Chinese, and genetic factors. Squatting is a traditional resting and working posture in China and is required for the use of non-Western toilets. Squatting utilizes an extreme range of motion that may engage areas of hip cartilage that are not loaded during upright stance (6), possibly stimulating turnover and regeneration of cartilage that is otherwise subject to disuse-related thinning and more vulnerable to stress (27, 28). On the other hand, squatting is probably much less common among Chinese residing in the US, yet they have a much lower risk of total hip replacement for OA than US whites (7, 9). Studies of the role of obesity in hip OA have yielded conflicting results, but suggest a weak association (2). Therefore, the relative thinness of the Chinese women compared with white women in our samples (the prevalence of obesity in the male samples was similar) is unlikely to account for the large differences in prevalence we observed. Finally, there is convincing evidence of a strong genetic component in hip OA (29, 30), with heritability estimates from a classic twin study of ∼60% (31). Genetic factors affecting the anatomy and biomechanics of the hip or joint-specific cartilage metabolism may contribute to the low risk of hip OA in Chinese and yet not confer protection from knee OA.

Several strengths and potential limitations of our study deserve mention. In both China and the US, we sampled the source populations and not clinical patients who might represent referral or disease populations. We carefully standardized disease ascertainment methods in all groups, making it highly unlikely that the differences seen here are due to differences in the methods of assessment. We asked Chinese men and women about hip pain using the same questions as in the US studies, and we adopted a rigorous approach to translating the questionnaires. While the SOF radiographs were read 5 years prior to the NHANES-I and Beijing radiographs, we found good to excellent agreement between these earlier readings and re-readings done at the time of the present study. The pelvis radiography protocol used in the NHANES-I centered the x-ray beam more proximally (L2–L3) compared with the BOA and SOF studies (symphysis pubis), which may affect the assessment of the joint space (32) and the estimated prevalence of hip OA. Only anteroposterior pelvic views were obtained in all of the studies. The addition of other views (e.g., lateral) may increase the prevalence of radiographic findings slightly, but it is likely that an increase would occur similarly in all groups and, thus, not alter our basic findings.

We used published atlases to score the individual radiographic features of hip OA (17, 18) and measured the minimum joint space in the hip, a surrogate for cartilage thickness (33) that has been validated for defining hip OA (15, 16, 34). Since there is no consensus definition of radiographic hip OA, we utilized a combination of the most widely used and validated radiographic definitions to overcome limitations inherent in any one of them (16, 35). Our radiographic criteria were more stringent than in some previous studies (16, 31, 36) but similar to several others (8, 29, 37–39). Sensitivity analyses showed that our results were unchanged using alternative radiographic definitions. Our prevalence estimates for whites (4.5% for men ages 60–74 years, 3.8% for women ages 60–74 years, and 5.5% for women ages 65–89 years) are in the middle of the range from previous studies (3, 16, 36, 39–41) and confirm a similar prevalence in men and women (2). Some individual radiographic features, including acetabular osteophytes and moderate superomedial joint space loss, were common in the hips of Chinese as well as whites and may reflect age-related changes that are not specific for hip OA (27, 28, 42).

In conclusion, we found a very low prevalence of hip OA among elderly persons, both men and women, living in Beijing as compared with elderly white persons living in the US. Explanations, including racial variation in normal hip anatomy, other genetic factors, physical activity, and other environmental factors, must be sought. Like other studies comparing disease occurrence across geographic regions, our findings may provide clues to important pathogenic factors in OA, factors which have likely not yet been fully appreciated.


We are indebted to Yechiam Ostchega of the National Center for Health Statistics for allowing us access to the NHANES-I hip radiographs and to Clara Yeung for valuable assistance throughout the project.