The prevalence of RKOA in this population sample of women whose data were obtained from general practice registers in the UK increased during the 15 years of followup, from 14.7% to 48.7%. We also noted the higher prevalence of OA in the right knee (versus the left), suggesting involvement of mechanical factors in RKOA etiologic development. The novel finding of this study was a consistent association of IL-6 circulating levels with the prevalence and incidence of RKOA as assessed by K/L grade. IL-6 levels were consistently found to be associated with the RKOA prevalence in the specific year and when the 3 visits were combined. Two other variables that were consistent predictors of the appearance of RKOA were age and BMI (Tables and ).
Although the evidence for an association with serum TNFα levels was negligible, CRP levels were also significantly associated with RKOA when considered separately (Table ). However, the effect of CRP disappeared after adjustment for age and IL-6 levels in multivariate analysis (Tables and ), likely because the CRP concentration and the levels of IL-6 are related biologically. In our study, the serum concentrations of CRP and IL-6 showed a significant correlation, which varied between 0.22 and 0.48 (P
< 0.001), depending on the year of the visit. CRP levels showed very low, nonsignificant correlations with age. This may account for our earlier findings of a significant predictive effect of CRP in the same sample (16
). In the present sample, when the IL-6 effect was considered alone, it showed a highly significant association with RKOA (P
< 0.001). As a result, the prevalence comparison between the first and fourth quartiles of IL-6 levels yielded an OR of 2.74 (95% CI 1.94–3.87; P
< 0.001) for those with K/L grade ≥2 knees, and an OR of 3.59 (95% CI 1.93–6.68; P
< 0.001) when only those with severe RKOA (K/L grade >2) were considered.
A growing body of evidence suggests that development of OA, even at the early stages, is often accompanied by inflammation (24
). This is consistent with the findings from epidemiologic studies in which the severity and progression of tibiofemoral cartilage damage were observed to be more common and severe in patients having reactive and inflamed synovial fluid (25
). Moreover, elevated levels of cytokines such as IL-1β and TNFα, both of which are regulators of inflammation, are found in the disease process, although their precise cell of origin is not clear (17
). Increased levels of CRP have been associated with the prevalence and progression of knee and hip OA in several studies (16
), although in other studies, this association became nonsignificant after adjustment for BMI (19
). Of note, however, to our knowledge, there is no published study examining the association between OA and circulating IL-6 levels. Thus, the present study is the first to assess this association.
One potential source of these inflammatory cytokines is adipose tissue, which accompanies obesity (28
) and which is known to be associated with high levels of biomarkers of inflammation, including IL-6 (7
). In the study by Das (29
), the results suggest that synthesis of TNFα (and leptin) in adipose tissue could induce the production of IL-6 and the acute-phase reactant CRP. Increased levels of these cytokines may, in turn, contribute to the development of OA. Our data support this observation. In our study, we observed significantly higher BMI values in individuals affected with RKOA. Moreover, the correlation between BMI and RKOA was found not only with the weight-bearing knee joints but also with hand OA in the present sample (results not shown). The link between obesity and hand OA was noted some time ago (e.g., see ref.30
). Although the mechanism remains to be debated, one of the most popular theories is that fat tissue causes systemic release of proinflammatory cytokines. Circulating IL-6, as mentioned above, is known to stimulate CRP production by the liver. Indeed, we observed consistent correlations between the CRP levels and IL-6 levels, and significantly higher CRP levels in RKOA-affected individuals.
There are several limitations to this study. First, there is a discrepancy in the timing of the study assessments of RKOA and the cytokine assays (Table ). Serum samples were not available at baseline when the first set of radiographs was obtained. Thus, it is not known how many individuals had RKOA at year 5 (when the first serum samples were assayed). Likewise, the number of subjects having elevated IL-6 levels at baseline, when the first radiographic examination was conducted, is not known. This was the main reason for testing the association between the cytokine levels at year 5 with RKOA at year 10.
The second limitation is that inflammatory cytokines were measured in the circulation and not in synovial fluid. It is possible that serum cytokines could originate from a different tissue. In addition, the decreasing sample size with each followup visit, albeit inevitable in a longitudinal study of this kind, did contribute to a loss of statistical power. We found, however, no evidence of a systematic bias resulting from the loss of baseline subjects at followup. First, only a relatively small proportion of the sample was lost during the followup period. Second, IL-6 levels and OA prevalence rates at the baseline examination did not differ between those who subsequently dropped out and the total group.
The effect of limited sample size was pronounced when multivariate analyses were undertaken, since the small sample numbers at baseline would not allow us to specifically examine progressive disease. However, we believe that incidence and progression may be interchangeable in this kind of study. For example, since we defined RKOA as a K/L grade ≥2 in either knee, those with a K/L grade ≥1 at baseline whose disease then progressed would be considered an incident case, although really they should be considered “progressors.” Indeed, when we compared IL-6 levels according to K/L grade (unaffected = K/L grade <2, minor RKOA = K/L grade = 2, and severe RKOA = K/L grade >2), a consistent association of continuous increases in these parameters was observed (Figure ), which was consistent with the ORs for RKOA by quartiles of IL-6 levels. Similarly, if we had used magnetic resonance imaging or another more sensitive measure of disease, then most incident cases would probably be those with minor OA detectable at baseline. Clearly, the study would have benefited from current radiologic standards and techniques for the knee, although any misclassification due to the extended knee position would have acted to obscure the results.
It is also important to note that this study did not examine the effect of OA comorbidity in the other joints. This would lead to a much more complex analysis of the data, which was beyond the scope of this study.
A further, minor limitation is that the study was undertaken using data only from female subjects. OA is increasingly recognized to be site- and sex-specific, and therefore the present results may not be directly extrapolated to men. There is little evidence, however, to suggest that cytokine expression differs significantly between the sexes.
In conclusion, the present findings suggest that a higher BMI and increased circulating levels of IL-6 are associated with the development of RKOA. IL-6 is therefore a potential new biomarker for OA development and may provide useful information in the prediction of disease outcome. Moreover, highlighting the potential key role of IL-6 in the disease process may lead to studies of IL-6 blockade as a therapeutic strategy in OA.