Comparison of the prevalence of knee osteoarthritis between the elderly Chinese population in Beijing and whites in the United States: The Beijing osteoarthritis study




To estimate the prevalence of radiographic and symptomatic knee osteoarthritis (OA) in a population-based sample of elderly subjects in Beijing, China and compare it with that reported in the Framingham (Massachusetts) OA Study.


We recruited a sample of persons age ≥60, using door-to-door enumeration in randomly selected neighborhoods in Beijing. Subjects completed a home interview including questions on knee symptoms and a hospital examination including knee radiographs obtained during weight bearing. The protocol was identical to that used in the Framingham OA Study. A reader read intermingled Beijing and Framingham Study films to ensure high reliability. We defined a subject as having radiographic knee OA when the Kellgren/Lawrence grade was ≥2 in at least 1 knee. Symptomatic knee OA was recorded as present when knee pain was reported and the symptomatic knee had radiographic OA. We estimated the prevalence of these entities in elderly subjects in Beijing and compared it with OA prevalence in Framingham, using an age-standardized prevalence ratio.


Of 2,180 age-eligible Beijing subjects contacted, knee radiographs were obtained in 1,787 (82.0%). The prevalence of radiographic knee OA was 42.8% in women and 21.5% in men. Symptomatic knee OA occurred in 15.0% of women and 5.6% of men. Compared with women of the same age in Framingham, women in Beijing had a higher prevalence of radiographic knee OA (prevalence ratio 1.45, 95% confidence interval 1.31–1.60) and of symptomatic knee OA (prevalence ratio 1.43, 95% confidence interval 1.16–1.75). The prevalence of knee OA in Chinese men was similar to that in their white US counterparts (for radiographic OA, prevalence ratio 0.90; for symptomatic OA, prevalence ratio 1.02).


Using identical methods and definitions to evaluate the prevalence of OA across populations, we found, surprisingly, that older Chinese women have a higher prevalence of knee OA than women in Framingham, Massachusetts. The prevalence in men was comparable. Possible explanations for these differences range from genetic differences to heavy physical activity among Chinese.

Osteoarthritis (OA) is a leading cause of disability among elderly whites in Western societies (1), but its occurrence in other populations has been studied little. Major differences in the racial or geographic distribution of a disease often provide valuable clues about potential etiologic factors. Indeed, there is already strong evidence for racial differences in several musculoskeletal diseases (2–7). For example, rheumatoid arthritis (RA) is much more common in some American Indian tribes than in US whites, and it is somewhat less prevalent in selected Asian populations (2); prevalence of ankylosing spondylitis varies from population to population depending, in part, on HLA–B27 prevalence (3); and systemic lupus erythematosus is more common in blacks than in whites (4).

There is a remarkable paucity of information as to whether the occurrence of OA differs across racial groups or geographic regions. Most surveys have been limited to northern or western Europe, the US, and some sites in Africa and the Caribbean (8). There have been few population-based studies of OA in Chinese or in other Asian populations. However, the studies that do exist present tantalizing evidence that rates of OA may differ greatly from those seen in white populations. Findings of several investigations have suggested that the prevalence of hip OA is very low among Chinese in Hong Kong (6, 9), among elderly persons in Japan (10), and among Japanese and Chinese living in Hawaii (11) and San Francisco (7). If hip OA is uncommon in Chinese and Japanese populations, this may reflect a low occurrence of OA in general, leading to the expectation that the prevalence of knee OA would also be low. Furthermore, a powerful effect of obesity on the risk of knee OA (8) and the relative thinness of most Chinese persons (12) would also suggest that elderly persons living in China should have less knee OA than those in the US.

To date, relatively few data concerning knee OA among the Chinese have been reported. A study in Hong Kong (6) showed that elderly Chinese had a lower prevalence of radiographic knee OA than what had been reported in white populations, but the study did not include any internal non-Asian comparison groups. In contrast, results of a community survey in a suburban Beijing village (13) showed that the prevalence of symptomatic and radiographic knee OA was 30.1% and 60.5%, respectively, among those age ≥50. These estimates are higher than those derived from most European or US studies. In the suburban Beijing study (13), however, the number of subjects age ≥50 was small (n = 150). Furthermore, in the above-mentioned reports, the questions for soliciting information on knee symptoms, the protocols used to obtain radiographs, and the criteria for grading radiographic change were not clearly described. It is widely recognized that the reported prevalence of OA varies greatly depending on questions used to solicit symptom information, on ways that radiographs are obtained, and on methods used to read these radiographs. Such variability demands careful standardization when investigators are attempting to determine racial or geographic differences in OA prevalence. In addition, a comparison will be valid only if the subjects among whom prevalence is evaluated are representative of the populations being investigated.

The occurrence of disabling diseases that affect the elderly, such as OA, will soon be of major importance in China, the world's most populous country and one with a burgeoning elderly population. In addition, Chinese and other Asian populations constitute the fastest growing group of immigrants in the US (14), suggesting that the occurrence of OA in Asians will be of relevance in the US.

We conducted a population-based survey among elderly Chinese in Beijing, using the same instruments and protocols to ascertain and define OA as were used in the Framingham (Massachusetts) OA Study, a population-based study of elderly subjects in the US. We estimated the prevalence of knee OA in the elderly Chinese population and compared this prevalence with that in the elderly US white population.


Subject recruitment

The Beijing OA Study. We studied a random sample of men and women age 60 years or older from 3 central districts of Beijing, China. Each of the districts includes 9–24 health sections, with each section affiliated with 1 hospital for the purposes of public health services. We randomly chose 1 health section from each of the 3 central districts and then randomly selected administrative units (neighborhoods) from that health section. All of the streets in the selected neighborhoods within a given health section were listed in random order. Beginning with the first street on the list, interviewers went door-to-door to residents to enumerate and interview all men and women age ≥60 years who were self-described residents of Beijing. Street-by-street recruitment continued until interviewers met a district quota for each sex and age stratum (60–69, 70–79, ≥80). The number of subjects recruited in each district was determined as the proportion of the residents age ≥60 years in a particular district to the total population age ≥60 years in all central districts, based on 1995 National Intercensus data (∼445,000 people in this age group lived in central districts in Beijing in 1995). Within a district, the number of subjects in each age stratum was determined according to the sex and age distribution in that district as indicated in the census data.

Subjects were interviewed at their homes or site of identification. Trained health professional interviewers administered a standardized questionnaire that focused on joint symptoms, previous diagnoses of arthritis, and possible risk factors for OA. At the end of the interview, subjects were invited to the central examination site at Peking Union Medical College Hospital for a clinical examination, laboratory testing, and radiography. Transportation to the hospital was provided.

The Framingham OA Study. The Framingham Study cohort was assembled in 1948–1951 as a random sample of adults, ages 28–61 years, in Framingham, Massachusetts. Its original purpose was to evaluate risk factors for cardiovascular disease. Subjects have been examined biannually since then. At examination 18 (1983–1985), weight-bearing anteroposterior radiographs of both knees of surviving participants were obtained. All subjects were asked about knee pain. A detailed description of that study has been published elsewhere (15). All subjects in the Framingham OA Study were white.

Surveys and examinations

To ensure comparability of the assessment of knee OA in the 2 studies, we standardized instruments for the Beijing OA Study so as to produce definitions of knee OA comparable with those used in the Framingham OA study.

Survey instrument. First, we compiled an English version of the questionnaire for the Beijing OA Study. Questions on knee symptoms were derived from the Framingham OA Study. Two bilingual translators whose native language was Mandarin Chinese, working independently, translated these questions into Chinese. After that, 2 other translators whose native language was American English back-translated the questions into English. Finally, all instruments were presented to a translation committee consisting of these and other professional Chinese–English translators as well as study investigators. Discrepancies in translation were resolved to preserve contextual meaning. The specific question used to assess the presence or absence of knee symptoms was that administered as part of the Framingham OA Study: “Have you ever had pain in or around a knee on most days for at least a month?” Also, as in the Framingham Study, if the subject responded “yes” to this question, a battery of questions was administered, inquiring further about the site of pain, its duration, the last time the pain occurred, and its severity.

Examination protocol. We adopted the same approaches described above to develop the examination protocols for the Beijing OA Study. After the examination protocols were finalized, 2 of the investigators (DTF and MCN) held on-site training for the examiners and x-ray technicians. Subjects had a single weight-bearing anteroposterior knee radiograph obtained using the Framingham OA Study protocol.

Height was measured twice for each subject, using the wall-mounted Harpenden stadiometer (Holtain, Crosswell, Wales, UK). If the difference between 2 measurements was ≥4 mm, examiners were required to obtain an additional 2 measurements. The average of the measurements was then used in the current analysis. Weight was assessed using a balance beam scale with a precision to 0.1 kg. Calibration of the scale and stadiometer were checked regularly, using external standards.

Reading of radiographs. One bone and joint radiologist (PA) read anteroposterior knee films according to the reading protocols of the Framingham OA Study. Specifically, prior to the arrival of films from the Beijing OA Study, the radiologist re-read a batch of radiographs from the Framingham OA Study so as to calibrate his reading with earlier Framingham OA Study readings. Preliminary readings of batches of randomly selected films, which consisted of both OA and non-OA cases, continued until the reader reached a high level of agreement with previous readings (no specific cutoff was used). At that point, the reader began to read Beijing knee films. For each batch of Beijing OA Study films, films from the Framingham OA Study were used in 3 ways: first, they were used as a “calibrator” for the reader to review prior to reading each batch; second, Framingham OA Study films (with readings provided) were commingled randomly with Beijing OA Study films so the reader could continually recalibrate his readings during a session; and finally, Framingham OA Study films were commingled, at random intervals, with readings that had not been provided to test reliability. In addition, a sample of 5–10% of previously read knee radiographs from the Beijing OA Study was fed back to the reader to test reliability.

The weighted kappa on Kellgren/Lawrence grade (16) for reliability readings in all batches was 0.83 (95% confidence interval 0.77–0.90) for interreader reliability and 0.79 (95% confidence interval 0.73–0.84) for intrareader reliability. The small number of disagreements did not occur in any particular direction, suggesting that there was no likelihood of bias in estimates.

Each knee was evaluated for the presence of osteophytes, joint space narrowing, sclerosis, and cysts, and was graded for overall evidence of radiographic OA using the Kellgren/Lawrence scale (graded 0–4, where 0 = none; 1 = possible osteophytes only; 2 = definite osteophytes and possible joint space narrowing; 3 = moderate osteophytes and/or definite joint space narrowing; and 4 = large osteophytes, severe joint space narrowing, and/or bony sclerosis). We characterized a knee as having radiographic OA if it had a Kellgren/Lawrence grade of ≥2, and severe radiographic OA if the Kellgren/Lawrence grade was 3 or 4. We used the same definition of symptomatic OA as in the Framingham OA Study, i.e., 1) a subject reporting knee pain lasting at least 1 month with pain having last occurred within the current or previous year; and 2) radiographic OA in that painful knee.

Statistical analysis

We divided the subjects into 5 age groups: 60–64, 65–69, 70–74, 75–79, and ≥80 years. We calculated the age-specific prevalence of radiographic and symptomatic knee OA for Chinese men and women separately. In both the Framingham and the Beijing studies, we excluded subjects who self-reported RA and whose hand or knee radiographs revealed typical radiographic evidence of RA (posteroanterior hand radiographs were available in both studies). In comparing the prevalence of knee OA between Chinese subjects in Beijing and whites in Framingham, we confined our analysis to individuals age ≥65 years, because the number of subjects under age 65 in the Framingham OA Study was small (11 men and 17 women). We applied the age-specific prevalences of radiographic knee OA in the Framingham OA Study to the age distribution of the Beijing OA Study and calculated age-standardized prevalence ratios and their 95% confidence intervals. The same approaches were used to examine the difference in prevalence of symptomatic knee OA between Chinese and whites.


Of 2,180 age-eligible subjects contacted during door-to-door canvassing, 1,953 (89.6%) completed the home interview. Among subjects with a home interview, 1,787 (91.5%) had a clinical examination and knee radiographs. Of those, 3 subjects who had RA and 3 subjects whose radiographs could not be assessed were excluded from the analysis. Only 1 subject had a (unilateral) total knee replacement. Compared with those who had completed both a home interview and a clinical examination, subjects who had only the home interview were older (mean age 72.2 years versus 67.5 years; P < 0.001) and more likely to be female (72.4% versus 59.0%; P < 0.007). There was no difference in the prevalence of knee pain between the 2 groups (28.6% versus 27.9%).

Demographic and anthropometric characteristics of participants in the Beijing OA Study are presented in Table 1. Elderly Chinese subjects in Beijing were, on average, thinner than elderly subjects in many white populations (mean body mass index 25–26 kg/m2); and most had low levels of formal education.

Table 1. Demographic and anthropometric characteristics of the elderly Chinese subjects in the Beijing Osteoarthritis Study*
 Men (n = 730)Women (n = 1,051)
  • *

    Except where indicated otherwise, values are the number (%).

Age, years
 60–69465 (63.7)737 (70.1)
 70–79219 (30.0)270 (25.7)
 ≥8046 (6.3)44 (4.2)
Height, mean ± SD cm167 ± 5.9154 ± 5.4
Weight, mean ± SD kg69.9 ± 10.561.6 ± 10.4
Body mass index, mean ± SD kg/m225.2 ± 3.426.0 ± 4.0
Years of education
 040 (5.5)281 (26.7)
 1–6333 (45.6)486 (46.2)
 7–9129 (17.7)130 (12.4)
 10–12102 (14.0)85 (8.1)
 13–1695 (13.0)56 (5.3)
 ≥1731 (4.2)13 (1.2)

Table 2 shows the age-specific prevalences of knee OA in elderly Chinese subjects in Beijing. Radiographic knee OA was common among the elderly Chinese. The frequency of radiographic knee OA increased with age, and it was more common in women than in men. The prevalence of symptomatic knee OA was also high, especially among women and among subjects age ≥70.

Table 2. Prevalence of knee osteoarthritis (OA) among elderly subjects in Beijing, China
Age, yearsMenWomen
nRadiographic OA, %Symptomatic OA, %nRadiographic OA, %Symptomatic OA, %

Figure 1 depicts the age-specific prevalences of radiographic and symptomatic knee OA in Chinese men in Beijing and white men in Framingham. The prevalence of radiographic knee OA was lower in Chinese than in whites under the age of 70, was similar in the 2 groups between ages 70 and 79, and was slightly higher in Chinese than in whites age 80 and older. Except for subjects under age 70, the prevalence of symptomatic knee OA among Chinese men appeared slightly higher than that among white men. Overall, for men, no statistically significant difference was found in the prevalence of either radiographic or symptomatic knee OA between the 2 racial groups. However, Chinese men seemed less likely to have unilateral radiographic OA or severe radiographic knee OA compared with white men (Table 3).

Figure 1.

Prevalence of radiographic osteoarthritis (ROA) and symptomatic OA (SxOA) of the knee in elderly Chinese and white men.

Table 3. Comparison of the prevalence of knee osteoarthritis (OA) among Chinese and white participants age ≥65 in Beijing, China and Framingham, Massachusetts, respectively*
OA statusCrude prevalence, %Age-standardized prevalence ratio (95% CI)
  • *

    95% CI = 95% confidence interval; K/L = Kellgren/Lawrence.

 Radiographic OA (K/L grade ≥2)27.630.80.90 (0.75–1.07)
  Unilateral10.016.00.63 (0.46–0.83)
  Bilateral17.514.91.19 (0.95–1.47)
 Severe radiographic OA (K/L grade ≥3)7.715.90.50 (0.35–0.69)
 Symptomatic OA7.16.91.02 (0.70–1.42)
 Radiographic OA (K/L grade ≥2)46.634.81.45 (1.31–1.60)
  Unilateral12.515.20.86 (0.68–1.08)
  Bilateral34.119.71.92 (1.67–2.20)
 Severe radiographic OA (K/L grade ≥3)15.715.51.10 (0.90–1.34)
 Symptomatic OA15.411.61.43 (1.16–1.75)

In contrast, Chinese women had a higher prevalence of both radiographic and symptomatic knee OA than their white counterparts (Figure 2). The age-standardized prevalences of radiographic and symptomatic knee OA in Chinese women were 45% and 43% higher, respectively, than in white women (Table 3). The difference in radiographic knee OA between Chinese and white women was mostly due to the higher prevalence of bilateral radiographic knee OA among the Chinese women. However, no statistically significant difference was found in the prevalence of severe radiographic knee OA between the 2 racial groups.

Figure 2.

Prevalence of ROA and SxOA of the knee in elderly Chinese and white women. See Figure 1 for definitions.


In this population-based comparative study, we found that knee OA is very common among elderly Chinese in Beijing. Chinese women have a substantially higher prevalence of both radiographic and symptomatic knee OA than their white counterparts; the prevalence of knee OA among Chinese men is roughly similar to that among white men.

Several characteristics of this study are noteworthy. First, the disease assessment methods were carefully standardized across the 2 comparison groups, which makes it unlikely that the differences seen are attributable to differences in the methods of assessment. Second, in both the Framingham and the Beijing studies, we were able to recruit randomly a large percentage of the target population. Thus, we do not believe that our results are due to differential or selected sampling of the target population. Finally, we did not rely on only one definition of knee OA; we were concerned that both radiographic changes alone and radiographic changes plus symptoms be represented, since they are both valid disease definitions. Our comparative prevalence results are supported by the finding that similar differences were observed with both definitions of disease.

Earlier studies have not suggested consistent differences in knee OA occurrence between Chinese and white populations. Hoaglund et al (6) studied OA prevalence among a hospital and clinic Chinese population in Hong Kong, and found that the prevalence of knee OA was low. In that study, however, no non-Asian population comparison group was evaluated, and the number of subjects in each of the age deciles was small.

Zhang and colleagues (13) studied OA prevalence in a suburban community outside of Beijing, a geographic region much closer to the area in which the present investigation was carried out. While that study surveyed >2,000 subjects, only a few of the subjects were over age 60. The prevalence of radiographic and symptomatic knee OA among subjects over age 50 was 60.5% and 30.1%, respectively. While these prevalence estimates are higher than those reported for white populations, neither detailed information on the protocol for reading of radiographs nor specific questions used to assess knee pain was reported.

Woo et al (17) conducted a population-based survey among elderly Chinese (age ≥70) in Hong Kong. They found that 48% of women and 27% of men reported having knee pain. These estimates are higher than what has been found among age comparable subjects in the Framingham study. Investigators, however, did not list the criteria for knee pain assessment.

It is possible that there are differences in OA prevalence between northern and southern China; studies of knee pain would suggest it is more prevalent in the North (18). This may explain the marked difference in the prevalence of radiographic OA seen in Hong Kong (6) and suburban Beijing (13). Ultimately, however, knee OA in China has simply not been studied well enough to enable one to make definitive statements about geographic differences in prevalence.

It should be noted that bilateral disease was prevalent among elderly Chinese in Beijing, especially among women. Given the relative thinness of the Chinese subjects, their high prevalence of bilateral knee OA is even more remarkable. The higher prevalence of bilateral knee OA could be due to the much more physically active lifestyle of the Chinese compared with US whites, especially among those who are elderly. Among the participants in the Beijing OA Study, 39.1% of men and 34.1% of women reported that the job they had held longest involved moderate to heavy labor. If physical activity were the sole explanation for the difference in the prevalence of bilateral knee OA, this would imply that a lifetime of physical activity is likely to be a strong risk factor for knee OA. However, one would expect the difference in knee OA prevalence between the 2 racial groups to be greater in men than in women. The men in China not only had worked at manual laboring jobs for many years, but a large percentage of them continued to be very physically active (for example, riding bicycles and helping with the shopping for their families) and also spent significant amounts of time in a squatting position, which can create stress on the knees. That the differences between Chinese and white men were not as great as those found in women suggests that physical activity alone is not a sufficient explanation for the high prevalence of bilateral knee OA among Chinese.

In contrast, severe radiographic OA was more prevalent among the subjects in the Framingham study compared with those in Beijing. While explanations for this discrepancy are not obvious, it is possible that high body weight predisposes to more severe structural change rather than to disease occurrence per se.

This study has certain limitations. Perhaps the most important one is that we cannot necessarily generalize our location-specific findings to overall racial differences in prevalence. Our results do not necessarily equate to a comparison of all Chinese and all whites. As noted above, there may even be differences in the prevalence of knee OA among different regions in China. If occupation and physical activities are factors explaining the high prevalence of knee OA in our Beijing sample, prevalence may be even higher in other areas of China that are less urbanized or where more people are engaged in heavy manual labor. The validity of our prevalence comparisons could theoretically be threatened by selective participation of those with knee OA in China, but this seemed unlikely because participation rates in Beijing were extraordinarily high, and knee symptoms were equally prevalent among participants and nonparticipants in both the Beijing and the Framingham OA studies (15).

In conclusion, we found a surprisingly high prevalence of knee OA among elderly subjects in Beijing, especially women, when compared with elderly subjects in Framingham, Massachusetts. These findings were unanticipated, especially because of the thinness of the elderly Chinese. Like other investigations comparing disease occurrence across geographic regions in the world, our study may provide clues to important pathogenetic factors in OA: factors that have likely not yet been fully recognized. In addition, our results suggest that OA of the knee is likely to become a major public health problem in China given the fact that there are already 150 million persons age ≥60 in that country, and the population there, as in many Western nations, is aging. Our findings also suggest that the high rate of immigration of Asians to the US could have an impact on the occurrence of knee OA in this country.