Association between biomarkers of tissue inflammation and progression of osteoarthritis: evidence from the Rotterdam study cohort

Osteoarthritis (OA), the most common form of arthropathy, is characterized by alteration of joint structure including progressive cartilage destruction, synovial inflammation, and changes to the subchondral bone [1]. The etiology of OA is not well understood although the knowledge in this respect has accumulated during the past decades. Beside genetic variation and biomechanical mechanisms [1], altered lipid metabolism [2] and inflammation [3] might also be important drivers of the molecular mechanism underlying OA.

It is clear that OA is heterogeneous in its etiology and disease course. Recent efforts are now focused on identifying subgroups of patients with distinct disease pathology, which will allow the development of new targeted therapies [1]. Circulating biochemical markers (biomarkers) have the potential to serve as a measure of the different pathological processes linked to OA. However, very few biomarkers have been identified that can predict the course of OA. Up to now, only urinary C-terminal telopeptide of collagen type II (uCTX-II) and serum cartilage oligomeric protein (COMP) (both markers of cartilage and bone metabolism) have shown discriminative ability for diagnosis and prognosis of OA [4‐10].

Compared to rheumatoid arthritis (RA) or seronegative spondyloarthritis, inflammation is less prominent in OA. There is no marked infiltration of inflammatory cells into joint tissues, and the synovial fluid usually contains few neutrophils [11]. C-reactive protein (CRP), a systemic biomarker for inflammation, has been shown to be elevated in some OA patients yet the evidence is conflicting [12]. Recent studies have suggested that local inflammation plays a prominent role in the pathogenesis, symptoms, and progression of OA [3, 13, 14]. Recently, a newly developed CRP measure was described. Matrix metalloproteinase (MMP)-dependent degradation of CRP (CRPM) can be measured in serum to quantify CRP fragments released from the inflamed tissue, after CRP has been synthesized in the liver and deposited in the joint and degraded by the proteolytic burden [15]. It has been shown that MMP-degraded CRP provides a more discriminative diagnostic potential compared to that of full-length CRP in ankylosing spondylitis (AS) patients [16].

Synovial inflammation (synovitis) is a common characteristic of inflammatory OA and is believed to stimulate the connective tissue turnover. Type I collagen, the most abundant structural collagen of the human body, is a collagen of the connective tissue, including the synovial membrane. The collagen biomarker C1M is a measure of MMP-driven soft tissue destruction [17]. In a recent study among RA patients, C1M levels were associated with progression of RA and were also shown to correlate with RA activity [18]. Moreover, elevated levels of C1M were found in OA patients with higher CRP [15] and CRPM levels [16]. It seems that inflammation is important in disease pathology in a subset of the OA population; however, the pathology of this subset is not well described and few longitudinal analyses of these patients have been presented [19, 20].

Due to a general lack of consistent evidence, differences between the populations studied (clinical trials vs. population-based cohorts), and differences in sample collection, future research is needed to validate the existing OA markers and to identify new candidates. The aims of the present study were to explore the prognostic value of two biomarkers of tissue-inflammation: C1M and CPRM. In addition, we examined the extent to which these two biomarkers could be considered to be independent of well-known markers of uCTX-II and COMP and demographic characteristics such as age, sex, and body mass index (BMI) for incidence and progression of radiographic OA in a large population of men and women. Moreover, we investigated whether CRPM has prognostic value for OA progression compared to the full-length CRP.

The descriptive characteristics of study participants are presented in Table 1. Our study population was slightly younger, less obese, less diabetic, and drank more alcohol than total population. OA cases at baseline and follow-up were older, more obese and diabetic, and smoked less than subjects who remained healthy at follow-up time.

Baseline levels of most of the biomarkers were higher in OA cases compared with subjects who remained free of OA during follow-up. At baseline, there were 68 and 185 hip OA and knee OA cases, respectively. At 5-year follow-up, 41 hips (32 new cases) and 111 knees (68 new cases) with incident OA were identified. Overall progression of OA was found in 170 out of 1295 participants who were eligible for OA progression analyses [knee OA cases (n = 122) and/or hip OA cases (n = 53)].

Biomarkers levels were significantly correlated to BMI and age except for uCTX-II with BMI (p = 0.37) and CRPM with age (p = 0.54) (data not shown). Moreover, there were significant differences in C1M (p = 0.003) and uCTX-II levels (p <0.0001) between men and women. For age, gender, and BMI-corrected data, it was found that most of the biomarkers at baseline were significantly associated with each other, with the highest correlation between CRP, CRPM, and C1M (Table 2). Urinary CTX-II was not correlated to any of the inflammatory markers.

We here evaluated two novel biomarkers for tissue inflammation (C1M and CRPM) for their ability to predict incidence and progression of OA after 5 years of follow-up in a population-based setting. We show that CRPM is associated with incidence and progression of OA. Moreover, CRPM had prognostic value for progression of OA independent from uCTX-II and COMP. In the full statistical model, CRPM and CRP proved to have similar prognostic value as uCTX-II and COMP. Together the biomarkers added significant prognostic power to the known risk factors of age, gender and BMI.

To date, several studies have reported that levels of uCTX-II and COMP are associated with OA and progression of OA, suggesting them as the most promising OA biomarkers [5, 7‐10, 28‐31]. Consistently, we found that uCTX-II and COMP levels were correlated with increased risk and progression of OA, confirming our previous finding among participants of RS-I [7, 10]. However, despite substantial clinical investigations, none of these markers have yet been proven to be of definite clinical value [32]. Different outcome measures and study designs as well as limited sample size of these studies have resulted in inadequate discriminating ability to differentiate between individual patients and controls (diagnosis) or between patients with different disease severities (burden of disease), and to predict prognosis in individuals with or without osteoarthritis (prognosis) [33].

Our study showed that CRP levels were related to the risk and progression of OA. In our previous study with RS-I participants, an association between CRP levels and incidence of hip OA was seen as well [12]. Data on hip OA progression was limited in that review. A recent review and meta-analysis of 32 studies showed no significant associations between CRP levels and progression of OA [defined as either exacerbation of joint space narrowing (JSN) or TJR], while CRP levels were significantly associated with pain and decreased physical function. Therefore, it was suggested that CRP may be elevated in OA patients, but probably plays a greater role in symptoms rather than radiographic changes in OA [14]. In our study, we observed no significant associations between knee and hip pain and CRP levels among prevalent cases at baseline when controlled for BMI. It should be noted that there was significant heterogeneity due to the quality and methodology of the studies included in the meta-analyses and the result should be regarded with caution.

It has been shown that the CRPM level is elevated in OA patients compared to the healthy adult reference range, and that the inflammatory burden is independent of disease severity (KL score) [15]. In our study, we observed progression of OA to be associated with higher levels of CRPM. Moreover, the association seemed to be independent of CRP. Previous studies did not support the role of systemic inflammation in OA etiology and progression [14], however, CRPM, a degradation product of CRP, and a possible biomarker of chronic tissue inflammation, might point toward a local low-grade inflammation to play a role in a subset of individuals with OA. Our results suggest that this subset is more prone to radiographic OA progression.

We observed no association between levels of C1M and incidence and progression of OA in both biomarker-specific and full models. The collagen biomarker C1M was found to be higher in OA patients with an elevated inflammatory burden as measured by high-sensitivity CRP [15] and CRPM [16]. Previous analysis of C1M in RA has shown that the biomarkers were associated with structural progression measured by JSN [15]. Moreover, a recent ex vivo experiment showed that under pro-inflammatory conditions serological biomarkers C1M, MMP-mediated degradation of collagen type III (C3M), and active MMP-3 may originate from the inflamed synovial membrane (Kjelgaard-Petersen CF, Bay-Jansen AC, Christiansen T, Ladel C, Karsdal MA, Siebuhr AS. Novel synovitis biomarkers: TNF-α and IL-1β include MMP-mediated degradation of collagen type I and III and active MMP-3 and -9 in synovial membrane explants. 2014, Submitted). Our study showed significant positive linear association between C1M levels and CRPM, and CRP levels adjusted for age, sex, and BMI among baseline OA cases (data not shown). These findings illustrates that connective tissue turnover in OA (i.e., C1M levels) is increased with inflammation and, together with our findings of CRP and CRPM, provide more support to the importance of local inflammation in OA pathogenesis in a subgroup of patients.

Identification of subjects at a high risk of OA is necessary for preventive strategies. Biochemical markers might help in identification of subgroups of OA patients with higher risk of progression. Age, gender, and BMI are already rather strong predictors of OA risk and progression at older age in the population studied and in other investigations [10]. In a previous study among participants of RS-I, a prognostic model for incident knee OA was developed based on clinical, genetic and biochemical (uCTX-II) risk factors [34]. The study showed that these risk factors combined had a relatively low predictive value for knee OA. In contrast, a model including doubtful minor radiographic degenerative features (KL score = 1) reached a good predictive value. Consistently, in our study, baseline KL score of one was the best predictor of future OA, and even better than age, gender and BMI alone. Addition of well-established biochemical markers together with tissue inflammation biomarkers added moderate predictive value to most of our models. Moreover, for the first time, we showed that a marker of MMP-dependent inflammation is able to predict progression of OA independent of established biomarkers uCTX-II and COMP. Future investigations are needed to identify the potential usefulness of the tissue inflammatory biomarkers in OA prediction in a clinical setting.

Radiographic OA bears little relationship to the illness characterized by joint pain and functional impairment [35, 36]. Different factors (i.e., bone marrow lesions, joint effusion, psychological factors, comorbidities) and mechanisms of joint pain (i.e., nociceptive pain, neuropathic pain, and central pain sensitization) have been described in OA patients [37, 38]. A recent study among OA patients showed the correlations between central pain mechanisms (temporal summation and pain modulation) and CRPM, independently of age, gender, BMI, and hsCRP [37]. Consistently, we found a positive association, albeit nonsignificant, between hip pain and CRPM levels, and between knee pain and CRP and C1M levels. These trends support the role of inflammation as a possible underlying mechanism of joint pain in OA, but more power is needed to definitely prove this suggestion.

There were considerable correlations between different markers reflecting turnover of cartilage (CTX-II, COMP, and C1M) and inflammatory markers (CRP and CRPM). Therefore, simultaneous modeling of several markers could provide more insight into the role of individual markers in OA progression. Moreover, due to the longitudinal design of the study, we were able to assess the potential predictive value of the markers to predict disease risk and progression. Controlling for confounder factors including age, sex, BMI, smoking status, education, and diabetes is another strength of our study. Our study, however, had some limitations that must be taken into account. First of all, biomarkers were only assessed in baseline samples. Serial biomarker assessments could also be very informative on the natural dynamics of each biochemical marker and on disease status. Moreover, the biomarker contributions from OA need to be distinguished from the contributions of normal and age-related bone and cartilage turnover, and other conditions affecting the biomarkers levels; this is currently a primary limitation for all systemic (serum and urine) biomarker measures [33]. However, individuals with rheumatoid arthritis or other inflammatory arthropathies were excluded from our analyses. Additionally, we adjusted the analyses for BMI and age, two major factors which affect the biomarkers levels. Second, OA definition was based on KL grades, which conflate osteophytes, and JSN and some studies have shown that biomarker associations can be different with respect to these related but different features [39]. Third, knee OA was defined using anteroposterior radiographs of the knee. Therefore, patellofemoral joint OA was not taken into account in the study. Moreover, uncontrolled occurrence of misalignment of the medial tibial plateau and central X-ray beam in standing anteroposterior radiographs of the knee might result in an underestimation of the rate and homogeneity of JSN in knee OA [40]. Fourth, pain was assessed by questionnaire and not by more precise methods such as Western Ontario and McMaster Universities Arthritis Index pain score or visual analog scale. Fifth, controls were free of OA at the joint site studied but were allowed to have OA at other joint sites. For example, if knee OA was studied, controls had to be free of knee OA but were allowed to have hip OA. In addition, the inference of the relevant relationships between serum or urine biomarker concentrations and disease of specific joints is complicated by the fact that our OA patients might have disease in other joints such as hands or spine joints. Sixth, several studies have suggested that risk factors for incidence of OA may be different from risk factors for progression [41]. This could be due to heterogeneity in OA structural pathology (e.g., bony proliferation versus cartilage loss), limitations of imaging, which may result in different sensitivities to the structural features, limited sample size of the studies that examined the risk factors, and that the risk factors may affect disease differently at different disease stages [41]. Although we had a reasonable sample size for total OA incidence and progression analyses, the statistical power was limited to identify which factors operate at different disease stages and for analyses of OA phenotypes leading to wider confidence limits. Finally, although we covered several confounding variables, we could not exclude bias due to unmeasured confounding factors as well as health-based selection bias as we used a subset of RS-II participants who had follow-up data. These subjects were probably more mobile to visit the center and survived in the follow-up period.

Our study confirmed that the uCTX-II and COMP concentrations are associated with incidence and progression of radiographic OA. Moreover, we showed for the first time that a MMP-dependent tissue-inflammation marker predicts the risk of OA progression independent of established biomarkers uCTX-II and COMP at the 5-year follow-up. This indicates that inflammation is associated with disease progression in OA and that inflammation has pathological relevance in OA. Further prospective studies are needed to confirm this association for OA.