Methods:
Study subjects were those enrolled in the Dutch Combinatietherapie Bij Rheumatoide Artritis (COBRA) trial with baseline and follow-up serum samples. Details of the study methods have been described here previously. Baseline serum levels of RANKL and OPG were assayed from stored samples and compared to the 5-year progression rate for erosions and joints space narrowing for radiographs of the hands and feet, scored using the Van der Heijde modification of the Sharp scoring method (Sh/VdH). Shorter time intervals were used in secondary analyses. First-year time-averaged ESR values were used as the marker of baseline systemic inflammation.
Results:
92 of the 155 subjects enrolled in COBRA were included. Subjects were primarily female (59%) with a mean age of 50 years and a mean disease duration at entry of 5 months. Seventy-five percent of subjects were rheumatoid factor seropositive, with 44% demonstrating evidence of baseline radiographic damage (defined as > 4 Sh/VdH units). Baseline RA disease activity included a mean ESR of 51 mm/hr, and a mean DAS score of 6.1.
Significant positive correlations with 5-year radiographic progression were found with baseline ESR and first-year time averaged ESR levels (r = 0.44 and 0.50, respectively; p <0.001 for both). Significant negative correlation with 5-year radiographic progression was found with the ratio of OPG to RANKL (r = -0.38; p = 0.001). No significant correlation was found between the individual baseline OPG or RANKL levels and radiographic progression. Statistical interaction modeling demonstrated a synergism of baseline inflammation and OPG:RANKL, such that those subjects with the highest first-year time-average ESR and the lowest OPG:RANKL had the greatest risk of radiographic progression compared with those in whom the opposite relationships were seen.
In multivariate analyses controlled for RF status, ESR, baseline radiographic damage, and treatment, subjects with the highest baseline RANKL were at a greater than 4-fold higher risk of 5 year radiographic progression compared with those with the lowest baseline levels (OR 4.4 (95% CI 1.5 13)). In contrast, those with the highest baseline OPG levels were protected against radiographic progression (OR 0.29 (95% CI 0.10 0.85)). Subjects with high RANKL and low OPG at baseline had an average radiographic progression score of 8 Sh/VdH units per year compared to 2 Sh/VdH units per year in those with low RANKL and high OPG at baseline.
Conclusions:
The ratio of OPG to RANKL is clinically predictive of long-term radiographic progression in early untreated RA, particularly when baseline inflammation is also considered.
Editorial Comment:
This is a creative and very elegant clinical study that helps to clarify the key molecular participants in the pathway to erosive damage in RA. In particular, it may partially explain the disconnect between inflammation and joint damage that is sometimes seen subsets of RA patients. In these subsets, high levels of inflammatory activity may at times be associated with relatively little joint damage whereas in others, relatively mild inflammatory activity may be associated with accelerated erosive disease. Previous work from this group, using subjects from this same cohort, has emphasized the importance of fluctuations in inflammation over time and RF seropositivity on radiographic progression. Thus, further work will undoubtedly examine the effect of fluctuations in RANKL:OPG over time (particularly with respect to treatment with biologic and non-biologic DMARDs) and the role of RF, CCP antibodies, and the interaction of the inherited HLA-DR4 shared epitope.
One concern with the present study is that serum levels of RANKL and OPG may not be truly reflective of synovial levels (i.e. those directly leading to joint damage). However, given the robust results detected here using serum levels, results using synovial levels would likely be equal, if not superior, to those presented here.
On a practical note, RANKL is not a commercially available at this time for routine measurement in the clinic. Additional work will be needed to determine whether routine measurement and determination of OPG and RANKL dynamics has any role in predicting disease outcomes on an individual patient basis.
