Article Text
PDF
Abstract
Objective To determine to what extent baseline measurements of the ratio receptor activator of nuclear factor-κB ligand (RANKL):osteoprotegerin (OPG) and C-terminal cross linking of type-I and type-II (CTX-I and CTX-II), in addition to traditional markers of disease severity, could predict annual radiological progression.
Methods A cohort of 155 patients with early, active, untreated rheumatoid arthritis (RA) who participated in the Combination Therapy in Early Rheumatoid Arthritis trial (COBRA trial) was followed up for 11 years. Urine was sampled at baseline and after 3 months from the start of treatment and analysed for CTX-I and CTX-II. Baseline serum samples were analysed for RANKL and OPG. Available traditional markers of disease severity included baseline measurements of erythrocyte sedimentation rate, rheumatoid factor and baseline radiological damage. A digital database of frequent radiographs was available, scored according to the Sharp/van der Heijde method. Individual annual progression rates were calculated and used as outcome variable. Multiple linear regression analyses identified the strongest predictors of annual radiological progression.
Results In multivariable analyses the RANKL:OPG ratio and CTX-I or CTX-II proved to be independent predictors of annual radiological damage over 11 years. The prediction of annual radiological progression was strongest when the RANKL:OPG ratio and CTX-I or CTX-II were evaluated in the same model (36–39% explained variance). Adding the effect of treatment at 3 months to the baseline models improved the predictive ability of the models up to 44–46%.
Conclusion Unfavourable baseline levels of the RANKL:OPG ratio as well as CTX-I and CTX-II in patients with early, active, untreated RA are strong independent predictors of rapid and persistent damage progression over the 11-year follow-up. Early improvement in bone markers by treatment predicts a better outcome.
Statistics from Altmetric.com
Introduction
Rheumatoid arthritis (RA) is a chronic autoimmune disease characterised by inflammation, primarily, of the joints, followed by cartilage degradation and subchondral bone erosion.1 Cartilage and bone damage can be visualised on radiographs and quantified using several different scoring techniques, resulting in a total damage score.2,–,4 The rate of radiological progression is different for every patient and substantial research has been done to identify markers that can best predict future radiological progression. Traditional predictors of future joint damage include measures of disease activity, such as the erythrocyte sedimentation rate (ESR), the disease activity scale, serum C-reactive protein levels, rheumatoid factor (RF) positivity, anti-cyclic citrullinated peptide (anti-CCP) and baseline radiological damage.5,–,7 It is generally believed that inflammation is the cause of future radiological damage, but there are indications that joint inflammation and destruction may be partly independent, as destruction may continue despite effective suppression of inflammation, or may stop during treatment in the face of persistent inflammation.8 9 In this context, markers of bone and cartilage destruction are of high interest.
The Combination Therapy in Early Rheumatoid Arthritis (COBRA) study was a randomised trial that compared the combination of sulfasalazine (SSZ), methotrexate (MTX) and prednisolone with SSZ alone.10 Combination therapy proved to be more effective in suppressing the inflammation process than SSZ monotherapy. This rapid suppression of inflammation resulted in a sustained reduction of the progression of radiological joint damage, sustained over a 5-year follow-up, despite similar treatment and disease activity in the follow-up period.11
Over the past years, a series of studies in the COBRA follow-up dataset have increased our understanding of the importance of bone and cartilage markers. We showed that high baseline levels of urinary excretion of C-terminal cross linking of type-I and type-II (CTX-I and CTX-II) and the CTX-II response after 3 months of treatment independently predicted an increased risk of radiological progression over 4 years, especially in patients without radiological damage at baseline.12 13 Further, we showed that a direct longitudinal relationship exists between clinically perceptible signs of arthritis and degradation of CTX-I and especially CTX-II.14 These and other studies15,–,17 suggest that in RA urinary excretion of CTX-II and other cartilage degradation markers most likely represents joint cartilage, whereas excretion of CTX-I, less prominently associated with clinical signs of arthritis, most likely represents both generalised and localised bone loss.18 19
Another marker for bone loss or, more specifically, for osteoclast activation is the ratio between receptor activator of nuclear factor-κB ligand (RANKL) and osteoprotegerin (OPG) in serum.20 RANKL induces osteoclastic bone destruction and OPG is its decoy receptor that prevents bone destruction by preventing the binding of RANKL with its receptor RANK. Because bone resorption is regulated by the relative expression and production of RANKL and OPG levels, the ratio of these (RANKL:OPG) can be used as a marker of osteoclast activation.21 In the final study of the COBRA dataset we demonstrated that low baseline OPG:RANKL ratios also predicted 5-year radiological progression.22
Recently, we completed the 11-year follow-up of the COBRA trial, including long-term outcome of radiological progression in patients with RA for 11 years.23 These data provide a unique opportunity to confirm the predictive ability of bone markers CTX-I, CTX-II and the RANKL:OPG-ratio as well as traditional disease activity-related markers for annual radiological progression over an 11-year follow-up.
Methods
The COBRA cohort
The multicentre, randomised, double blind controlled COBRA trial included 155 patients with early, active RA, in which COBRA combination therapy was compared with SSZ monotherapy. Details of the initial trial, the 5- and the 11-year follow-up have been reported previously.10 11 23 COBRA combination therapy is a step-down treatment strategy comprising SSZ (2 g/day), MTX (7.5 mg/week) and an oral pulse of prednisolone, starting with 60 mg in the first week and tapered to 7.5 mg in the seventh week. At 28 weeks, prednisolone is tapered and withdrawn and after 40 weeks, MTX is withdrawn. Patients in the SSZ group received SSZ (2 g/day) and placebo MTX and prednisolone. Medication was given according to protocol up to 56 weeks, and after this period, treatment decisions were made at the discretion of the treating rheumatologists.
Measurements of traditional predictors of radiological progression during the trial period that were available for this study included ESR, RF positivity, baseline total Sharp/van der Heijde (SvH) score and the Health Assessment Questionnaire (HAQ).
All participating patients gave their written informed consent to join in the 11-year follow-up study and the proposal was approved by the medical ethical committee of the VU University Medical Center.
Radiology
A digital, radiological database of the COBRA cohort was available, comprising all available radiographs of every patient since the start of the COBRA trial. This database contained radiographs of hand and feet, scored by two independent assessors according to the SvH method.4 24 Assessors were blinded for group allocation but aware of time sequence. Scores can be stable, increase or decrease. The mean of the two observers' scores was used to calculate the total SvH score at every time point.
Radiological progression for each patient was expressed as the annual progression rate (expressed as total SvH score per year). This annual progression rate was calculated for each patient individually, by linear regression analysis on every available radiograph of that patient with time as the independent variable, thus providing the most accurate estimate for the regression coefficient. A mean of eight (range 2–11) radiographs was available for each patient to calculate the annual radiological progression rate.
Bone markers
Urine samples (second morning void, non-fasting) were obtained at baseline and 3 months after the start of the trial and kept frozen at −20°C. Urinary excretion of CTX-I and CTX-II was measured at baseline and 3 months after the start of the trial—CTX-I by the CrossLaps ELISA (Osteometer Biotech, Herlev, Denmark) and CTX-II by a CartiLaps ELISA based on a mouse monoclonal antibody raised against the EKGPDP sequence of human type 2 collagen C-telopeptide (Osteometer Biotech). Details of this assay are described elsewhere.25 26
Serum was obtained at baseline and kept frozen at −20°C. Baseline levels of RANKL and OPG in serum were measured by a capture ELISA, with two antibodies detecting different epitopes.22 RANKL was measured by personnel at Amgen (Thousand Oaks, California, USA), using two in-house-produced antibodies, and OPG was measured with a commercially available kit (Biomedica Medizinprodukte, Vienna, Austria). The levels of RANKL and OPG were expressed as a ratio, RANKL:OPG, reflecting bone destruction if >1 and bone formation/preservation when <1.
Analysis
Distribution of all the variables CTX-I, CTX-II, RANKL:OPG in relation to annual radiological progression improved after a log-transformation (natural logarithm) of the predictors as well as the outcome measure. All the results of these variables presented in this manuscript relate to the log (ln)-transformed data.
Scatter plots and Pearson's correlation coefficient of 0.65 (p<0.001) confirmed the high correlation between CTX-I and CTX-II reported earlier.13 For this reason and the possibility that these markers might measure similar processes, they were analysed in separate models. The correlations between CTX or CTX-II and RANKL:OPG were weak and it was thus decided that these markers can be analysed in the same model.
Radiological progression was defined as the mean yearly progression rate, calculated through individual linear regression over the available observations of a patient. Not all patients had CTX-I, CTX-II and RANKL:OPG measurements available. For that reason, the CTX-I and CTX-II models were investigated both with and without RANKL:OPG. Furthermore, RANKL:OPG was evaluated with and without CTX-I or CTX-II. Table 1 gives an overview of the patients and available measurements for each dataset.
Characteristics of the study participants
To investigate the strength of the relationship between all biologically plausible predictors and annual radiological progression individually, we first studied all variables separately using univariate linear regression. Subsequently, in a backwards selection procedure, one variable at a time was taken out based on its lack of statistical significance, with p>0.10 as a cut-off point for exclusion. The full model contained the variables: baseline damage (dichotomous, SvH ≤0.5 or >0.5), RF positivity, baseline ESR, 3-month change from baseline ESR, baseline CTX-I, 3-months change from baseline CTX-I, baseline CTX-II, 3-month change from baseline CTX-II, OPG:RANKL, HAQ at baseline and sex. Backwards regression analysis was done in two steps: first, including only baseline variables and second, including baseline as well as available 3-month follow-up measurements.
Results
Univariate linear regression analysis showed that all three bone markers were significantly related to annual radiological progression (depending on the dataset used). RANKL:OPG was numerically the strongest univariate predictor followed by CTX-II at 3 months and RF (table 2).
Univariate relationship between measures of disease activity, bone markers at baseline or during treatment and annual radiographic progression, expressed as percentage explained variance (R2) of the model, in different subsets of the database
Owing to the poor predictive ability of HAQ and sex in our datasets, these variables were excluded from the multivariate analysis.
In multivariate analysis, in both the CTX-I and CTX-II datasets, radiological progression was explained by a high baseline ESR and a positive test for RF (table 3). These models explained 18% and 22% of the variation in annual radiological progression, respectively.
β Coefficients (95% CI) of the best models based on backwards regression analysis
The dataset optimised for bone marker RANKL:OPG, without CTX-I or CTX-II (n=100), provides a model explaining 32% of annual radiological progression and is determined by a high baseline ESR, a positive test for RF and a high RANKL:OPG ratio.
The models combining CTX-I or CTX-II with RANKL:OPG predict 36–39% of variance and include RANKL:OPG ratio, CTX-I or CTX-II, ESR and RF for the CTX-I dataset together with baseline damage for the CTX-II dataset.
Including follow-up measurements of CTX-I, CTX-II and ESR at 3 months improves the predictive ability of the different models. With 44–46% explained variance, the models containing both RANKL:OPG and CTX-I or CTX-II best predict radiological progression. In the CTX-I dataset, annual radiological progression is best predicted by the presence of baseline damage, a high ESR 3 months after start of treatment, a positive test for RF at baseline, a high CTX-I at baseline and a high RANKL:OPG ratio at baseline. For the CTX-II dataset, annual radiological progression is best determined by the presence of radiological damage at baseline, a high ESR 3 months after start of treatment, a high CTX-II 3 months after this start of treatment and a high RANKL:OPG ratio.
These analyses suggest that a high baseline RANKL:OPG ratio in serum, in particular, strongly predicts a high annual progression rate. Figure 1 illustrates the model including both CTX-II and the RANKL:OPG ratio, with the predicted radiological damage of patients with an unfavourable level of either RANKL:OPG, ESR, CTX-II or baseline damage.
Effect of modelling prognostic factors on radiological progression over 11 years in patients with early rheumatoid arthritis (RA). The grey line shows progression in a patient with all factors at their mean. The other lines show the effect on progression of increasing one single factor to the value of the 90th centile. For example, the highest dashed line represents the radiological progression when C-terminal cross linking of type-II (CTX-II)3 months and erythrocyte sedimentation rate (ESR)3 months are at their median, baseline damage is absent, but the receptor activator of nuclear factor-κB ligand (RANKL):osteoprotegerin (OPG) ratio is at its 90th centile. The vertical box on the right represents the distribution from the 25th to 75th centile of radiological damage 11 years after inclusion in the Combination Therapy in Early Rheumatoid Arthritis trial (COBRA trial), if every patient had started with zero baseline damage. Regression equation: ln (annual radiological progression + 1) = 0.4×lnCTX-II 3 months+0.01× ESR3 months+0.2×lnRANKL:OPG+0.33×baseline damage (no/yes) lnCTX-II3 months (median, 90th centile)=3.07, 6.54; ESR (16, 46), lnRANKL:OPG (−0.249, 1.77), baseline damage (no 0, yes 1). SvH, Sharp/van der Heijde score.
Discussion
This study confirms the importance of RANKL:OPG and CTX-I and CTX-II as predictors of long-term radiological progression. In addition to earlier work, this study combined measurements of CTX-I or CTX-II with RANKL:OPG in one prediction model and found an even stronger relationship with annual radiological damage than for these markers modelled separately. As shown before, these bone markers measured at baseline are stronger predictors of progression of radiological damage than traditional predictors, such as baseline measurements of ESR, radiological damage and RF.
The baseline RANKL:OPG ratio as well as CTX-I and CTX-II were measured at an early and active stage of the disease, in mostly disease-modifying antirheumatic drugs naïve patients. The relationship between these baseline markers and long-term radiological progression suggests that the course of radiological progression is presumably already determined at a very early stage of the disease. However, the predictive ability of the 3-month change from baseline in CTX-I and CTX-II levels shows that effective treatment can alter this course. Identification of patients with high baseline levels of CTX-I, CTX-II and especially, a high RANKL:OPG ratio can help to estimate the severity of the disease and indicate the need for aggressive treatment. More research is needed to investigate the effect of treatment on the RANKL:OPG ratio.
Stusy of the explained variance from models including both CTX-I or CTX-II and RANKL:OPG compared with the separate models shows that the markers CTX-I or CTX-II with RANKL:OPG complement each other. This supports the assumption that those measurements reflect different disease mechanisms, with CTX-I reflecting generalised bone loss, CTX-II reflecting cartilage degradation and a high RANKL:OPG ratio reflecting inflammation-driven, local bone loss around the joints.
The strong predictive value of RANKL:OPG on annual radiological progression is supported by several different findings: RANKL inhibition is correlated with a marked suppression of bone erosion in several animal studies.27 28 Furthermore, the RANKL:OPG ratio has been shown to have a central role in bone resorption in postmenopausal osteoporotic women as well as during inflammation.21 29 Crotti et al showed that synovial RANKL expression is increased in patients with RA with active disease compared with patients with quiescent disease.30 The production of inflammatory cytokines in inflamed joints is likely to increase RANK and RANKL presentation, thus promoting osteoclast development and subsequent local bone destruction.31 The increased levels of RANKL in inflamed joints lead to a high RANKL:OPG ratio, reflecting bone destruction, which is predictive of increased radiological progression.22 This fundamental research, in combination with our own observations point to the crucial role of RANKL:OPG in the development of bone destruction and erosion in RA. In this context, it would be of great interest to study whether RANKL inhibition might prevent any new erosion occurring in human RA if given early in the disease, before erosions have occurred. Indeed, Cohen et al have recently shown that denosumab, a human monoclonal IgG2 antibody that binds RANKL and thus inhibits its activity, delayed structural damage, improved bone mineral density and suppressed bone turnover in patients with established RA.32
Young-Min et al studied the predictive ability of several biomarkers, including CTX-II, and compared their performance with traditional markers of radiological progression.17 This group concluded that CTX-II predicted radiological progression better than traditional markers, including C-reactive protein. However, a letter from Tchetverikov et al reminded readers that the currently most important marker of radiological progression is anti-CCP and was not included in the comparison, a limitation our study shares with that of Young-Min et al.33 Syversen et al showed that anti-CCP, ESR, RF and female gender are strong predictors of 10-year radiographic progression, with anti-CCP as the strongest contributor to the prediction model.34 However, this group recently showed that levels of RANKL and OPG measured in serum did not contribute to the predictive ability of current predictors such as anti-CCP, baseline damage and signs of inflammation.35 A large difference between the studies reported by Syversen et al and our study is that they studied a population with mild RA, while we studied patients with early aggressive RA. Further, we used the RANKL:OPG ratio, while Syversen et al used the separate components. Nevertheless, it is a shortcoming of our study that anti-CCP was not measured in this cohort since anti-CCP is one of the strongest predictors of radiological damage.5 6 36 37 The predictive ability of CTX-I, CTX-II and the RANKL:OPG ratio in a multivariate model including anti-CCP should be further investigated.
Because the COBRA trial was not designed to investigate bone markers as a predictor in RA outcome, these measurements were done on residual sera that was available about 7 years after the trial period. Unfortunately, not every patient had enough urine or serum available for the CTX measurements and even fewer patients had enough material left to do additional RANKL and OPG measurements. Literature on the stability of serum RANKL and OPG samples is scarce, although Chan et al showed that concentrations of serum RANKL and OPG significantly decreased after storage for 6 months at −20 or −70°C.38 However, sample collection, storage and measurements were all done in a blinded fashion, without knowledge of disease activity or long-term radiological outcome, making selection bias unlikely. Furthermore, baseline characteristics of all COBRA patients compared with the selection of patients in the different subgroups of the database used for this study, were highly comparable.10
Another point of consideration is the value of RANKL and OPG measurements in serum as compared with synovial fluid; a review by Rogers and Eastell showed that there is a lack of consistency between studies on the outcome of circulating OPG and RANKL.39 Owing to variation and variability in assays, the source of OPG and RANKL in different disease states is unknown and it is suggested that a proportion of circulating OPG and RANKL is from non-skeletal sources. For this reason, the RANKL:OPG ratio is a better marker, reflecting the net effect on osteoclast activity.39 40
The Outcome Measures in Rheumatology soluble biomarker group has formulated 12 validation criteria for soluble biomarkers to be regarded as valid with respect to reflecting structural damage end points in RA.41,–,43 The current research strengthens the case for CTX-I, CTX-II and RANKL:OPG as valid biomarkers based on their independent association with change in structural damage end points in this clinically well-defined prospective cohort with 11-year follow-up.
Our finding that the bone markers CTX-I, CTX-II and RANKL:OPG can predict long-term radiological progression is of major clinical importance, since patients with a poor prognosis can be identified very early in the disease and a decision about aggressive treatment can be better informed. However, this finding needs first to be studied in other datasets of patients with early RA.
In summary, measures of osteoclast activity reflected in the RANKL:OPG ratio and markers of bone and cartilage degradation CTX-I and CTX-II in patients with early, aggressive, untreated RA continue to predict annual radiological progression over an 11-year follow-up.
Acknowledgments
The authors would like to thank Maal van Everdingen MD, PhD and Arco Verhoeven MD, PhD for scoring the radiographs.
References
Footnotes
-
Funding This study was partly financed by the Dutch Arthritis Association, The Netherlands.
-
Ethics approval This study was conducted with the approval of the VU Medical Center.
-
Competing interests None.
-
Provenance and peer review Not commissioned; externally peer reviewed.
-
Patient consent Obtained.