Cartilage damage – regardless of its cause – will lead to changes in matrix composition. Cartilage matrix is composed of a mesh of type II collagen (COL2), the most abundant cartilage protein. Collagen consists of three chains that form a triple helix, with the exception of the nonhelical N-terminal and C-terminal ends, the telopeptides. In the extracellular matrix, the collagen molecules are linked to one another by cross-linking molecules such as pyridinoline; this cross-linking involves the telopeptides. While the nonhelical parts are degradable by many enzymes, the helical portion can only be degraded by mammalian collagenase at a specific site that yields fragments of one-quarter and three-quarter lengths, respectively. During cartilage degradation, therefore, different collagen fragments – such as crosslinked C-terminal and N-terminal telopeptides – are released (Table 1). Likewise, alpha chain fragments of collagenase degradation are set free and can be measured, such as COL2 three-quarter-long collagen via the generated carboxyterminal neoepitope.

Within the COL2 network, other components of the cartilage matrix are interspersed. These components are degraded in the course of cartilage damage. Among these are aggrecan, the breakdown products of which can be detected via core protein, keratan sulfate or chondroitin sulfate epitopes, and cartilage oligomeric matrix protein.

Bone destruction is an important hallmark of RA, and therefore the ability to measure its surrogates is important. The major structural protein of bone is type I collagen, which like COL2 forms triple helical, cross-linked structures and is degraded in a similar manner as COL2. Type I collagen C-terminal telopeptide (CTX-I) and type I collagen N-terminal telopeptide (NTX-I), but also free pyridinoline crosslinks, therefore reflect type I collagen degradation. Just like cartilage, bone contains noncollagenous proteins, such as bone sialoprotein, which are released during bone damage (Table 1).

Landewe and colleagues have shown a significant correlation between early changes in type II collagen C-terminal telopeptide (CTX-II) and prediction of long-term radiographic progression as a result of therapy; in contrast, changes of CTX-I were not related to joint damage 1415. The changes in CTX-II levels, however, were broadly overlapping with baseline values, and were therefore predictive primarily on a group level. Subanalysis of radiographic changes by joint space narrowing and erosion scores did not reveal major differences between these two components of the radiographic score in relation to these biomarkers 14. This analysis indicates that CTX-II levels may at least partly reflect the inflammatory response (which leads to collagen degradation via activation of matrix metalloproteinases), although there was no statistical evidence for such relationship. In contrast, MMP3 levels appear significantly associated with progression of joint damage 1617.

Assessment of COL2 three-quarter-long collagen showed decreasing levels during treatment with methotrexate, and particularly with anti-TNF + methotrexate 16. Moreover, baseline COL2 three-quarter-long collagen concentrations were significantly associated with cytokine levels 16.

Evaluation of cartilage oligomeric matrix protein levels revealed that these are much higher in patients whose joint damage progresses when compared with those whose radiographic changes do not progress 18. These observations as well as the CTX-II data were recently confirmed 17, although there was again broad overlap in biomarker levels of patients with and without progression of joint damage. Yet none of these markers performed better than C-reactive protein (CRP), swollen joint counts or composite disease activity indices 17.

Another set of markers potentially useful to assess bone damage relates to osteoclast differentiation and function. Osteoclasts that derive from bone marrow progenitor cells via the monocyte lineage are the cell population responsible for erosive changes 1920 and are pivotally dependent on the triggering of receptor activator of NFκB. The receptor activator of NFκB ligand (RANKL), a member of the TNF family, is expressed on various cell populations and also exists as a soluble molecule 21. The ligand's action can be inhibited by osteoprotegerin, a decoy receptor. Both osteoprotegerin and RANKL can be measured in serum, and the osteoprotegerin/RANKL ratio may be an indicator of osteoclast differentiation and activation. In fact, a low osteoprotegerin/RANKL ratio, reflecting high RANKL activity, appears to be associated with increased progression of radiographic joint damage 22. Osteoclasts mediate their special function in degrading calcified bone by several mechanisms, including secretion of cathepsin K. Consequently, cathepsin K levels correlate with joint destruction 23 and may be reflective of effective therapy 24.

Despite some interesting, although mostly inconclusive, data in the literature, there are a number of confounding factors related to the usefulness of many of these bone and cartilage breakdown products in helping to subgroup and treat RA. First, comorbidity, such as osteoarthritis and especially generalized osteoporosis 25, may lead to collagen degradation and thus obscure the subtle changes observed. Another factor is that circadian variations in bone and cartilage marker levels have been described (that is, the timing of the measurement is critical) 2627. Third, physical activity may change biomarker concentrations significantly 28. A fourth factor is that renal or hepatic disease may influence circulating and/or urine levels of biomarkers 29; especially, urine biomarkers should always be related to creatinine levels 11. Finally, the tissue content and/or levels of certain proteins that constitute useful biomarkers may be genetically determined, and this heritability may have to be accounted for 30. These factors contribute to the complexity of monitoring cartilage and bone turnover in RA as they relate to predicting joint damage 31. In an individual patient, however, short-term changes in the course of therapeutic interventions aimed at interfering with the RA process ought to reflect the modification of that process.

Another set of markers that is not derived from cartilage or bone may be predictive of joint damage: autoantibodies. It has long been recognized that radiographic progression of joint destruction is much higher in patients positive for rheumatoid factor when compared with seronegative patients 32333435. This is particularly true in patients with high-titer rheumatoid factor; that is, rheumatoid factor ≥ 50 IU/ml 36. Autoantibodies to citrullinated proteins have been shown more recently to be predictive for the occurrence of erosions 3637; the autoantibodies are broadly overlapping with high-titer rheumatoid factor, and the latter appears to change with effective therapy more than the autoantibodies to citrullinated proteins 35363839. These markers of the autoimmune response characteristic of RA surpass the value of most cartilage and bone breakdown products in predicting joint damage. At present, the combination of autoantibody and acute phase protein assessment may constitute the most reliable way to predict severe erosive RA 3340.

The clinical manifestations of RA are the consequences of synovial inflammation and the subsequent degradation and destruction of cartilage and bone. These pathways comprise various cell populations characteristically involved in RA synovitis as well as cytokines and the products of their action.

When looking at arthroscopic biopsies, the composition of the cellular infiltrate is heterogeneous 4142, and the most consistent reflection of active disease is the presence of high numbers of macrophages in the synovium 4344. Potent anti-inflammatory agents such as glucocorticoids lead to a reduction in synovial macrophage cellularity (but little other cellular changes) 45. Interestingly, blocking other mediators such as the chemokine macrophage chemo-attractant protein MCP-1 was not associated with changes of synovial tissue composition or with clinical benefit 3946. In addition, the degree of B-cell depletion in synovial tissue in the course of rituximab therapy was not significantly associated with the clinical response and B cells may be present in the synovium even if depelted in peripheral blood 3947. This finding also indicates that the periphery often does not reflect the events occurring in the joint, but these by themselves are also not sufficiently predictive.

With the exception of macrophage infiltrates, therefore, the composition and extent of the cellular infiltrate may not be related to clinical manifestations, in line with observations of histologic synovitis in the absence of clinical joint involvement 48. Likewise, the immunohistochemistry of cytokines may not sufficiently reflect the disease activity of RA. The use of gene expression profiling currently does not appear to provide much additional information in this respect 4950, although the technique may be useful to detect unforeseen changes. Serum levels of proinflammatory cytokines – including the most abundant one, IL-6 – are not highly correlated with measures of disease activity and progression 5152; moreover, baseline IL-6 concentrations may vary almost 100-fold between different individuals, and can increase with exercise 53.

Qualitative and/or quantitative differences exist among individual patients both on the cellular level and the cytokine level 5455, and pathohistologic analyses have failed to reveal changes that are pathognomonic for RA. Synovial cytokine expression in RA, even at a group level, does not significantly differ qualitatively from other arthritic disorders 565758. Nevertheless, there appear to be differences in the quantity of cytokine expression when looking at other inflammatory joint diseases, such as psoriatic arthritis, ankylosing spondylitis or inflammatory osteoarthritis. Patients with these conditions have lower synovial TNF and IL-6 levels than RA patients 5859. These differences may have some important bearing given that RA is a rapidly and highly destructive joint disease, while the other disorders are usually slowly or not as destructive.

All these facets reveal the complexity we face when trying to search for good biomarkers. The complexity is further illustrated by observations that changes found in these markers early in the course of RA may be quite different from those seen in late disease 60, regardless of the level of activity or inflammation in either early or late disease.

To better appreciate the complexity, we present four hypothetical patients in whom different cytokines or different cell populations may appear to predominate, even though in all patients many cells or soluble molecules are at least partly activated (Figure 2).

The resolution of this mosaic will be difficult: changes of biomarker levels in the course of therapy may give helpful insights, but this knowledge is insufficient to provide prediction rules for the employment of specific therapies.

In contrast to the markers mentioned above, the serum levels of CRP, a molecule induced by proinflammatory cytokines, especially IL-6 61, not only reflect the extent of disease activity 6263 but are also cumulatively associated with joint destruction 176364. Moreover, CRP levels before the start of disease-modifying therapy can predict the degree of subsequent radiographic joint damage, and greater reduction from baseline in CRP levels was associated with less progression of joint damage and higher trough serum concentrations of anti-TNF antibody 6566. While other acute phase reactants, such as serum amyloid A protein, also reflect disease activity, CRP determination is widely available and of low cost 67, making it the preferred biomarker of disease activity (and even joint destruction) not yet surpassed by other markers.

The erythrocyte sedimentation rate could be regarded as an alternative to measuring CRP. However, it may be influenced by various other factors not primarily related to inflammation 68. Successful therapeutic interventions with glucocorticoids and disease-modifying antirheumatic drugs (DMARDs) including biological agents are usually associated with a fall in CRP levels, and are often also paralleled by reductions in IL-6 serum concentrations 246970.

It is somewhat disappointing that levels of other marker molecules, such as various cytokines, cytokine receptors or matrix metalloproteinases, have not been shown to exhibit better correlations with actual disease manifestations or outcomes of RA than acute phase reactants. No biological marker assessed hitherto has therefore been shown to be related better to disease activity and joint damage than CRP.

No biomarker currently allows one to predict the extent of clinical improvement in response to therapy, although the degree of radiographic progression on traditional disease-modifying agents can be foretold by baseline and cumulative levels of CRP, swollen joint counts or overall disease activity using respective composite scores 65.

In this context it needs to be borne in mind that TNF and IL-6 are cytokines that promote osteoclast differentiation and activation 202174757677. Importantly, under stable low concentrations of RANKL, increasing amounts of proinflammatory cytokines, such as TNF, will lead to increasing osteoclast differentiation 78. It is therefore conceivable that this relation of TNF levels and osteoclast activation also exists in vivo. In support of this notion, treatment of RA patients with a TNF inhibitor plus methotrexate leads to a dissociation of the close relationship between joint damage and the inflammatory response (exemplified by disease activity measures) 79, and these data were meanwhile confirmed with another TNF-blocker 80. These findings have led to the threshold hypothesis shown in Figure 3. According to this hypothesis, TNF will lead to joint damage especially once its levels exceed a particular threshold that lies above the threshold needed for the activation of the inflammatory response. Blocking of TNF may inhibit its bioactivity fully (Figure 3b), may reduce bioactivity to levels below the putative threshold of destruction with residual signs and symptoms but no destruction (Figure 3c), or may reduce bioactivity to levels above that threshold but still significantly retard joint damage compared with other treatment modalities, without preventing progression in full (Figure 3d). This hypothesis is in line with results from clinical trials where the use of TNF inhibitors significantly retarded or halted joint destruction despite residual active disease 818283.

Whilst the focus of attention with respect to joint damage has primarily been the osteoclast, it should be borne in mind that bone repair mechanisms may be deficient in TNF-mediated arthritis and can be induced by osteoblast activation 84. Biomarkers of osteoblast function, including proteins involved in the Wnt signaling pathways 85, may therefore constitute further interesting markers for the future.

Importantly, however, achieving low disease activity or remission with traditional DMARDs will lead to highly effective reduction of progression of joint damage 86. Another important aspect in our attempts to predict outcomes is therefore the estimation of the clinical response to treatment.

Recent analyses in a large cohort of clinical trial patients have revealed that baseline disease activity is already somewhat related to disease activity at 1 year of therapy, especially with methotrexate treatment 87. Irrespective of the type of therapy or disease duration, however, 3 months after initiation of treatment the disease activity – as assessed by the Simplified Disease Activity Index, the Clinical Disease Activity Index or the Disease Activity Score employing 28 joint counts – was highly correlated with disease activity at the end of the observation period 87 (Figure 4). These data were further validated by studying an observational cohort of RA patients 88. The probability to attain remission or low disease activity was more than 75% for patients achieving low disease activity by the Simplified Disease Activity Index already after 3 months of treatment, while it amounted to only 25% for patients having high disease activity at that point in time 87.

Needless to say, achieving lower disease activity after 3 months of therapy will also result in better outcomes in the other components of the triad – physical function and joint damage. Moreover, starting effective treatment early, and especially before initial damage has occurred, constitutes the optimal strategy (Figure 5). This optimization, however, requires early referral and early diagnosis 8990. Moreover, there is sufficient evidence to allow stating that DMARDs will interfere with the disease process at any point in time and will prevent progression of disease regardless of patient age or disease duration – it is the past damage that has to be carried on.

The ability to ulitilize biologic markers and or clinical markers to predict disease outcomes as well as therapeutic response early in the course of the disease, so as to achieve remission, is our ultimate goal in RA. We have indicated that predicting disease outcome and therapeutic response may be difficult due to the heterogeneity of the disease regarding both its clinical manifestations as well as putative pathogenic characteristics. Contributing to this has been the observed variability of biomarker levels. Data presently reveal that different therapeutic agents, including targeted therapies directed against different molecules, lead to similar therapeutic responses. To which degrees these responsive populations overlap, however, is currently unclear: do most of the patients achieving an American College of Rheumatology 70% response (or remission) to one drug have different characteristics than patients responding to a similar extent to another agent, or are they overlapping? Answers to these questions are mandatory for institution of proper treatment to gain control of disease and to induce remission.