We recently reported that elderly subjects in Beijing, China have a high prevalence of both radiographic and symptomatic knee osteoarthritis (OA) (1). When compared with women of the same age from Framingham, Massachusetts, Beijing women had a roughly 50% higher prevalence of both radiographic and symptomatic knee OA and had a 2-fold higher prevalence of bilateral radiographic disease. We found roughly equal prevalences of knee OA in Chinese and Framingham men. The differences in prevalence among women are especially notable, in that obesity is a potent risk factor for knee OA, especially in women (2), and Chinese women are substantially thinner than women from Framingham. Thus, the comparatively high prevalence of knee OA in Chinese women is unexplained. In rheumatic and other diseases, different patterns or characteristics of disease among different racial groups have often provided important clues about disease pathogenesis. For example, the high prevalence of renal disease among African Americans as compared with Caucasians with systemic lupus erythematosus (SLE) has supported the notion of a unique pattern of genetic predisposition to SLE in African Americans (3, 4).
One important characteristic of knee OA is the compartment affected; the medial compartment is affected more often than any other tibiofemoral compartment in clinical samples (5, 6), to the point where clinical teaching has suggested that OA knees can be differentiated from rheumatoid arthritis (RA) knees because of the bow-legged postures of patients with varus-knee OA. The excess loading of the medial compared with the lateral compartment has been put forward as a major reason for the widely accepted predilection for the preferential occurrence of OA in the medial compartment.
Tang et al have suggested that the risk of knee OA is increased in Chinese subjects because their knees are more varus angulated than are the knees of Caucasians, predisposing them to medial knee OA (7). Varus alignment increases medial compartment loading during gait by increasing the external adduction moment (8, 9). This increased adduction moment correlates, over time, with medial joint space injury (10), with the development of OA in animal models (11), and with low likelihood of recovery from realigning tibial osteotomy (12). Thus, it is likely that varus alignment among the knees of Chinese subjects could produce a higher prevalence of OA, and that the OA would create a predilection for damage in the medial tibiofemoral compartment.
We undertook 2 related investigations to evaluate whether this explanation for the high prevalence of OA in Chinese subjects was true. First, we compared the prevalence of medial and lateral tibiofemoral OA in the 2 racial groups from Framingham and Beijing. Second, to try to better understand the differences that we uncovered, we used fully extended, anteroposterior (AP), weight-bearing knee radiographs from the China and Framingham cohorts to compare the anatomic alignments with those in knees of subjects without OA.
PATIENTS AND METHODS
- Top of page
- PATIENTS AND METHODS
Determining the prevalence of medial versus lateral knee OA in Beijing versus Framingham cohorts. The Framingham and Beijing OA Studies were population-based studies of elderly subjects recruited from the community. The Framingham OA Study, a substudy of the Framingham Heart Study, was carried out in 1983–1985 and subjects were reevaluated in 1992–1993. Evaluations consisted of an AP, weight-bearing, 14 × 17–inch, fully extended knee radiograph. For this study, to create groups comparable in age to those from Beijing, we used films from the first evaluation (1983–1985). The Beijing OA Study (subjects ages 60 years and older), conducted from 1997 to 2001, was designed to provide exactly the same information about knee OA as the Framingham OA Study. Thus, radiographs were obtained using exactly the same protocol. Full-limb films were not obtained in either study. More complete descriptions of the recruitment and assessment techniques used in these studies can be found elsewhere (1).
Framingham and Beijing films were read for the overall severity of OA using the Kellgren and Lawrence (K/L) scale (13), and joint space narrowing (JSN) was read for both the medial and the lateral compartments on a 0–3 scale using the Framingham OA Study Atlas (14). Radiographs were read by 1 reader, an academically-based bone and joint radiologist (PA). Reproducibility of the intrareader assessments and interreader assessments (a team of readers that included the reader for the Beijing Study having read earlier Framingham films) were high (for OA versus no OA intraobserver κ = 0.79, interobserver κ = 0.83; both P < 0.001; for medial OA, intraobserver κ = 0.79, and for lateral OA, κ = 0.91; both P < 0.001).
We evaluated the prevalence of medial and lateral OA in all subjects from both studies. To identify those subjects with RA in both populations, which are urban and well-served areas, we asked subjects whether they had ever been diagnosed with RA. In China, we also asked subjects about another diagnosis, rheumatic arthritis, which we thought might be reported by subjects as RA. If subjects reported having RA and taking second-line drugs, we characterized them as having RA. In addition to the primary reading of hand and knee radiographs, which included screening for RA, in all those with self-reported rheumatic arthritis and in a 20% random sample of rheumatic arthritis subjects, the bone and joint radiologist (PA) reread all hand radiographs in separate batches to look for any evidence of RA. Those with any evidence were also characterized as having RA. Because no cases were found among those with self-reported rheumatic arthritis, we did not read the remainder. Those characterized as having RA were excluded.
Radiographic OA was defined as present when the K/L grade was ≥2. Medial OA was defined when a knee had a K/L grade ≥2 (definite osteophytes) and medial JSN score of ≥1 on a 0–3 scale. Lateral OA was also defined as a K/L grade ≥2 and lateral JSN score of ≥1 on a 0–3 scale. We tested an alternate definition of medial and lateral OA which required a medial and lateral JSN score of ≥2 on a 0–3 scale. No knees had both medial and lateral disease.
Using the knee as the unit of analysis, we addressed 2 questions: 1) How did the prevalence of medial and lateral knee OA compare in China and Framingham? 2) Of those knees with radiographic OA (K/L grade ≥2), what proportion had medial and lateral disease in the 2 populations?
To compare the prevalence of medial OA between Beijing and Framingham, we performed a logistic regression analysis with medial OA versus no OA in the knee as the dependent variable (those with lateral OA or OA without compartment narrowing were excluded). We evaluated Framingham versus Beijing as an independent variable. Confidence intervals were adjusted for the correlation between knees using generalized estimating equations. We used the same approach to compare the prevalence of lateral OA in the populations. For analyses in which we evaluated the proportion of knees with OA affected by medial (or lateral) OA, we used as the outcome the proportion of medial (or lateral) OA knees versus all other OA knees.
Recent work by Sharma et al (15), which linked obesity to medial JSN in varus knees (as compared with lateral narrowing in valgus knees), has suggested that obesity may increase the risk of only OA characterized by medial JSN. Furthermore, because Beijing subjects were thinner than those from Framingham (1), we also performed all of these analyses after adjusting for the subject's body mass index (BMI). This permitted us to test whether any difference in the prevalence of medial or lateral OA by population was explained by the difference in body weight between populations.
Comparing anatomic alignment in Framingham and Beijing subjects. To determine whether a difference in alignment might help explain any differences seen in medial/ lateral OA, we compared anatomic alignment in the Framingham and Beijing knees. We used the 14 × 17–inch, fully extended knee radiographs, which were obtained using the same protocol in both studies. Because alignment can be affected by OA, we chose to measure it in subjects who had no evidence of OA (bilateral K/L grade 0). We randomly selected 25 subjects from each of 4 groups without knee OA: Beijing women, Beijing men, Framingham women, and Framingham men. Films from each study were mixed. We drew lines through the middle of the femoral shaft and through the middle of the tibial shaft. The angle subtended at the point at which these 2 lines met in the center of the tibial spines was characterized as the anatomic angle according to the methods of Moreland et al (16). We rounded anatomic alignment to the closest degree. Intraobserver reproducibility was 0.97. Alignment measurements were done by a person (BB) different from the person who performed the radiograph readings, and they were done at a later time.
The analysis in this substudy on alignment consisted of a sex-specific comparison of the anatomic alignment of knees from the 2 racial groups. We first generated knee-specific data, using data from both knees of each person. Because the knees of an individual are not independent, we used generalized estimating equations to adjust for interknee correlations and compared the racial groups in terms of alignment.
In all analyses, the P values presented are 2-sided.
- Top of page
- PATIENTS AND METHODS
Prevalence of medial versus lateral knee OA between cohorts. Our Beijing subjects were 60–88 years of age and those from Framingham were ages 63–93 years at the time of the study. Framingham women and men were slightly older, on average, and the men weighed more compared with the Beijing subjects (Table 1). Radiographs of both knees of each person were read after removing the radiographs of knees with unreadable JSN, those with total knee replacements, and those from persons with RA, and we found that 24.6% (275 of 1,117) of the knees of the Framingham women had OA versus 37.0% (738 of 1,997) of the knees of the Beijing women. The prevalence of OA among the men's knees in the Framingham Study was 137 of 626 (21.9%) versus 243 of 1,393 (17.4%) among the knees of the Beijing men.
Table 1. Characteristics of the Beijing and Framingham Osteoarthritis (OA) Study subjects by sex*
| ||Framingham (n = 1,084)||Beijing (n = 1,781)|
| Age, years||71.9 ± 5.9||67.1 ± 6.1|
| Weight, kg||64.9 ± 12.2||61.6 ± 10.4|
| BMI||26.2 ± 4.6||26.0 ± 4.0|
| % subjects with radiographic OA in at least 1 knee||31.5||42.8|
| Age, years||70.7 ± 5.0||68.1 ± 6.1|
| Weight, kg||79.9 ± 12.1||69.9 ± 10.5|
| BMI||27.0 ± 3.6||25.2 ± 3.4|
| % subjects with radiographic OA in at least 1 knee||31.2||21.4|
We next evaluated the prevalence of medial and lateral radiographic OA in Caucasian subjects in Framingham versus Chinese subjects in Beijing (Table 2). When we examined the women, we found that the overall prevalence of medial knee OA was similar in the 2 groups (13.7% of knees in Framingham versus 16.6% in Beijing), whereas the prevalence of lateral knee OA was much higher in Beijing women (2.7% of knees in Framingham versus 10.6% in Beijing; P < 0.001). Framingham men had a notably higher prevalence of medial knee OA (14.9% of knees) than did Beijing men (7.1% of knees) (P < 0.001), but Beijing men had a much higher prevalence of lateral knee OA (5.7% of knees versus 1.9% in Framingham; P = 0.002) (Table 2).
Table 2. Prevalence of medial and lateral knee osteoarthritis (OA) in Framingham versus Beijing populations*
| Prevalence of medial OA||153/1,117 (13.7)||349/2,101 (16.6)|
| Prevalence of lateral OA||30/1,093 (2.7)†||222/2,101 (10.6)|
| Prevalence of medial OA||93/626 (14.9)†||103/1,459 (7.1)|
| Prevalence of lateral OA||12/621 (1.9)‡||83/1,459 (5.7)|
We next restricted our analyses to knees with K/L scores of at least 2, to evaluate the proportion of knee OA affecting the medial and lateral compartments. Milder cases of OA could have just osteophytes and no JSN, and could therefore be characterized as having OA without falling into either the medial or lateral OA categories. Of women's knees with radiographic OA (Table 3), the proportion with medial disease was slightly higher in the Framingham population (55.6%) than in the Beijing population (44.8%). What differed most, however, was the proportion with lateral OA, which was much less prevalent among the Framingham women than among the women from Beijing (P < 0.001).
Table 3. Among knees with osteoarthritis (OA), proportion with medial or lateral OA in Framingham versus Beijing*
| Proportion with Medial OA||153/275 (55.6)†||349/779 (44.8)|
| Proportion with Lateral OA||30/272 (11.0)‡||222/779 (28.5)|
| Proportion with Medial OA||93/137 (67.9)‡||103/257 (40.1)|
| Proportion with Lateral OA||12/136 (8.8)‡||83/257 (32.3)|
Among the men whose knees had radiographic OA, we again uncovered racial differences in the compartment affected. In the Framingham men, most (67.9%) of the knees with OA were affected in the medial compartment, whereas in the Beijing men, only 40.1% of the knees were affected in this compartment. In contrast, lateral knee OA was much less prevalent among the Framingham men (8.8% of knees) than among the Beijing men (32.3% of knees) (P < 0.001).
We next evaluated the odds of medial and lateral OA in Framingham subjects by sex using Beijing knees as the standard, both unadjusted and adjusted for the subject's BMI. For the Framingham women, the odds of medial knee OA were slightly less than in Beijing women (odds ratio [OR] 0.81, 95% confidence interval [ 95% CI] 0.66–1.00; adjusted for BMI OR 0.72, 95% CI 0.56– 0.93). However, for lateral OA, the odds in Framingham women were much lower (OR 0.24, 95% CI 0.16–0.35; adjusted for BMI OR 0.22, 95% CI 0.14– 0.34).
For medial OA in Framingham men, the OR was 2.30 (95% CI 1.71– 3.12), a value that dropped to 1.86 after adjustment for BMI (95% CI 1.27–2.71). For lateral OA, Framingham men had lower odds of OA than did Beijing men (OR 0.33, 95% CI 0.18–0.61), and this remained after adjusting for BMI (OR 0.29, 95% CI 0.14–0.35).
We tested an alternate definition of medial and lateral OA that required more severe narrowing (≥2 on a 0–3 scale). Among all knees of women with radiographic OA, this more severe form of lateral OA was present in 8.9% of Chinese women's knees versus 4.4% of Framingham women's knees (P = 0.03). Among the knees of men with OA, this severe lateral OA occurred in 10.1% of Chinese men's knees versus 3.7% of Framingham men's knees (P = 0.06). Results for medial OA were similar to those for milder disease.
Alignment in Framingham and Beijing subjects. Among Framingham and Beijing subjects with knees showing no evidence of OA, we compared alignment by sex. Among women, the mean was 4.5° of valgus alignment in the Framingham cohort (95% CI 4.0–5.1) and 4.7° in Beijing (95% CI 4.1–5.3) (comparison of 2 groups P = 0.68 by generalized estimating equation). Among men, we found a modest difference, with knees of Caucasian Framingham men being slightly more varus in angulation than Chinese men's knees (Framingham 2.7° valgus, 95% CI 2.0–3.4 versus Beijing 4.5°, 95% CI 3.9–5.1; comparison for 2 groups P < 0.001).
- Top of page
- PATIENTS AND METHODS
Our results suggest that the increased occurrence of tibiofemoral OA among Chinese women and the comparable prevalence among Chinese men compared with women and men of the Framingham cohort is not due to varus alignment. Medial knee OA, either in overall prevalence or as a proportion of a total OA, was not increased among Chinese subjects, as would be predicted if they had more varus angulation. Ironically, but perhaps more consistent with the observed alignment, lateral knee OA was substantially more prevalent in Chinese subjects from Beijing than in Caucasian subjects from Framingham.
We believe that our findings have important implications about the etiology of OA and suggest that there may be fundamentally different etiologies in different cultures. There is a substantial difference between the prevalence of lateral knee OA in Chinese subjects in Beijing versus Framingham subjects, with both Beijing women and Beijing men having much higher prevalence of lateral-compartment disease. In China in both men and women, lateral knee OA is almost as prevalent as medial knee OA. This finding violates many assumptions about the inevitable injuriousness of the relatively high medial-compartment load and suggests that the adduction moment may be less of a factor in causing tibiofemoral OA in Chinese than in Caucasians.
Our results raise questions about risk factors for OA in China and their possible compartment-specific effects. For example, is heavy labor more likely to predispose to lateral disease? In a study of those who already had knee OA, Sharma et al (15) recently suggested that obesity was correlated with the severity of medial, and not of lateral, JSN in those with OA. Racial differences in compartment involvement were not much affected by adjustment for BMI difference between the 2 populations. If the difference in weight accounted for the difference in compartment involvement, such adjustment would have attenuated or even eliminated the racial differences we found.
Our results directly contradict a hypothesis suggested recently by Tang and colleagues (7) that varus knee alignment in the Chinese might explain their increased prevalence of OA of the knee. Not only did we not find comparative varus alignment in Beijing compared with Framingham subjects, even more importantly, we could not document an increased prevalence of medial knee OA. The directionality of the compartment involvement with knee OA is the opposite of what would be predicted by Tang et al. Indeed, our own measures of alignment are highly concordant with the compartment involvement that we uncovered. Specifically, our comparison of anatomic alignment in the 2 racial groups suggests that it is roughly comparable in women, and in men is actually slightly more valgus among the Chinese. This would correspond to our findings that among women, medial knee OA prevalence is roughly comparable and among men, medial knee OA is, if anything, less prevalent in the Chinese.
We also provide the first data on the prevalence of medial and lateral knee OA. Clinical case series have suggested for many years that medial knee OA in the US is more prevalent than disease in the lateral compartment. Indeed, among women with OA from Framingham, the ratio of medial to lateral disease is roughly 5 to 1. Among men, it is even higher, at 8–9 to 1. Our data suggest that this ratio may be different in other parts of the world.
Our findings raise important questions about the relationship between alignment and the occurrence of OA, questions that are impossible to definitively answer without longitudinal data. It is not known, for example, whether alignment per se is a risk factor for OA and whether it operates similarly in both racial groups. It is not known whether alignment interacts with other risk factors to increase risk and whether those relationships or associations are specific to a racial group. It is also unknown whether a particular level of abnormal alignment increases the risk of disease. While our results suggest a concordance between alignment and lateral disease in the Chinese, they by no means prove this relationship to be causal.
The difference between racial groups in the overall prevalence of medial and lateral OA is a function of differences in OA prevalence, which we have previously reported. Our analysis focusing on the proportion of OA that is medial versus lateral gets around the dependency on overall disease prevalence and provides perhaps a purer comparison of compartment-specific involvement. The predilection for lateral OA among the knees of Beijing subjects emerges from both of these analyses.
We excluded persons with diagnosed RA. Undiagnosed RA among Chinese subjects is unlikely to explain our results. RA is less prevalent in Asian populations who have been surveyed and examined (17) than in Caucasian populations similarly studied. Even if the prevalence were similar (0.8 in Caucasians, ranging up to 2% in older women) and if all of those with RA in China had lateral knee disease (and not diffuse narrowing), there would still be an excess of lateral knee OA cases in China. For example, using data from Table 2, assuming that 2% of the women had RA and all had lateral knee disease, 8.6% of Chinese women would have lateral OA versus 2.7% of Framingham women.
Our study has many strengths. We were able to compare compartment-specific disease prevalence because we had obtained carefully standardized radiographs evaluating the occurrence of compartment-specific OA identically in both Beijing and Framingham and had these films read by the same reader using the same protocol. Indeed, our calibration of Beijing Study radiographic readings so that they were similar to those from Framingham, detailed elsewhere, included 3 elements: feedback of Framingham readings at the beginning of every reading batch, Framingham reading results fed back to the reader during the reading of every batch, and a third set of Framingham films included in every batch to test the reader's reliability. Compartment-specific differences reported here are not likely to be the results of either reader or radiographic-acquisition variability. Another important strength of this study is that both samples were drawn from a population base and are representative of that base. Participant rates from the general community were high in both studies, and it is unlikely that compartment-specific OA would be a function of biased participation.
There are important limitations to this work. Perhaps most importantly, we did not obtain full-limb films for accurate measurement of mechanical alignment in the population studies. Femoral bowing can be more common in the Chinese. Furthermore, although the anatomic axis of the tibia is supposed to be straight (7), it is possible that bowing curvature of the tibia would lead to differences between anatomic alignment using the entire tibia and that using a truncated image of it. Lastly, the correction factor needed to go from anatomic to mechanical alignment could be different in the Chinese based on that bowing. Tang et al (7) compared anatomic and mechanical alignments in Chinese subjects and suggested the correction factor was indeed different from that in Caucasian subjects, by <1°. If Tang et al are correct in both assumptions, about varus alignment in Chinese subjects and the correction when going from mechanical to anatomic alignment, the net result would be a Chinese anatomic alignment even more varus for Chinese knees than that reported by Tang et al using mechanical alignment. We could not confirm these racial differences.
One reasonable concern might be whether our results from isolated areas of China (Beijing) and of the US (Framingham) represent the larger racial or geographic areas from which they originate. This answer is unknowable without further descriptive epidemiologic analyses of OA. Based on the clinical case series that describe compartment-specific OA, and the well-known representativeness of the Framingham Heart Study (18–20), we believe that the Framingham results are probably generalizable to US Caucasians. We have less confidence in the data from China, since this is one of the first population-based studies of OA in China and the first that we know of to look at compartment-specific involvement. Northern and Southern Chinese may differ in their prevalence of knee OA (21) and this may be due, in part, to alignment differences.
In conclusion, Chinese women and men from Beijing have a prevalence of medial knee OA similar to that in Caucasian men and women of the same age from Framingham. Disease in the lateral compartment is considerably more prevalent in Beijing subjects than in those from Framingham. Alignment differences may account, in part, for some of these racial differences in compartment-specific prevalence. An understanding of why Chinese subjects have such a high prevalence of lateral disease might help elucidate specific causes of OA.