Structural damage in rheumatoid arthritis, psoriatic arthritis, and ankylosing spondylitis: traditional views, novel insights gained from TNF blockade, and concepts for the future

Structural changes of bone and cartilage are a hallmark of inflammatory joint diseases such as rheumatoid arthritis (RA), psoriatic arthritis (PsA), and ankylosing spondylitis (AS). Despite certain similarities – in particular, inflammation as the driving force for structural changes – the three major inflammatory joint diseases show considerably different pathologies. Whereas RA primarily results in bone and cartilage resorption, PsA combines destructive elements with anabolic bone responses, and AS is the prototype of a hyper-responsive joint disease associated with substantial bone and cartilage apposition. In the present review we summarize the clinical picture and pathophysiologic processes of bone and cartilage damage in RA, PsA, and AS, we describe the key insights obtained from the introduction of TNF blockade, and we discuss the future challenges and frontiers of structural damage in arthritis.

bone and cartilage damage in RA. Both the severity of infl ammation -whether measured by C-reactive protein, the number of swollen joints, or the duration of morning stiff ness -and the duration of infl ammation have therefore emerged as important predictors of structural damage in RA [3,4]. Autoantibodies such as rheumatoid factor and anti-citrullinated protein antibodies, and -in close connection to anti-citrullinated protein anti bodies -the presence of the shared epitope in the HLA-DRB1 region, also predict the risk for bone erosions, which is probably related to a close association between autoantibodies and the chronicity of arthritis [5,6]. Molecularly, the tight interaction between infl ammation and bone/cartilage loss in RA is explained by the production of enzymes such as aggrecanases and matrix metalloproteinases, which degrade articular cartilage and bone as well as molecules that support the diff erentiation of osteoclasts [7].
Bone and cartilage loss has traditionally been a main diagnostic, monitoring, and outcome parameter in patients with RA in both clinical trials and routine clinical practice. Bone and cartilage damage is rapid and dynamic after disease onset and aff ects the majority of RA patients within the fi rst year [8]. Th e severity of bone and cartilage damage in RA is closely related to physical function in RA patients, suggesting that structural damage indeed impairs physical function [9][10][11]. Finally, eff ective control of infl ammation by conventional disease-modifying antirheumatic drugs (DMARDs) or combination therapies of DMARDs and glucocorticoids retards structural damage in RA. Structure-sparing eff ects have been documented for methotrexate (MTX), sulfasalazine, and lefl unomide individually and in combination [12][13][14][15]. It is not clear, however, whether MTX, sulfasalazine, lefl unomide, and hydroxychloroquine directly aff ect bone and cartilage damage, or indirectly benefi t joints by reducing infl ammation.

Novel insights gained from use of TNF blockers in RA
Th e introduction of TNF blockers as a therapeutic option in RA has challenged our view not only of synovitis but also of progression of structural damage. One of the most consistent eff ects of TNF-blocking agents in RA patients is a profound and sustained inhibition of bone erosion. In fact, all fi ve TNF blockers approved for the therapy of RA strongly retard or even arrest structural damage [16][17][18][19][20][21][22][23][24]. Th is strong structure-preserving eff ect is partially due to profound and rapid control of infl ammation. Also apparent, however, is that anti-resorptive eff ects may occur despite a lack of clinical response to a TNF blocker [24]. TNF-blocking agents thus combine a strong antiinfl ammatory poten tial, which controls synovitis, with direct protection of bone and cartilage (Table 1).
In this context it is noteworthy that TNF is an impor tant inducer of osteoclast formation and thus is a key molecular link between infl ammation and bone damage [7]. Addition of TNF to monocyte cultures challenged with macrophage colony-stimulating factor and receptor activator of NF-κB ligand (molecules that activate osteo clasts, which are the cells involved in bone resorption) enhances the formation of osteoclasts, and overexpres sion of TNF in mice entails increased formation of osteo clasts resulting in systemic bone loss as well as local bone erosions [25][26][27]. With respect to cartilage damage, TNF also is an inducer of matrix enzymes such as aggrecanases and metallo proteinases, particularly MMP-1, MMP-2, and MMP-3, which are produced by synovial fi broblasts, neutrophils, and chondrocytes and degrade the cartilage matrix. A specifi c protective eff ect of TNF blockade on articular cartilage is therefore conceivable; current evidence is circumstantial, however, and is not backed by suffi cient data. Direct assessment of the cartilage of small peripheral joints is still technically challenging and, to date, TNF blockers have shown little if any eff ect on the cartilage [28].

Future needs and unanswered questions in RA
TNF-blocking agents have undoubtedly enriched our therapeutic options for blocking structural damage in RA. Nonetheless, several aspects remain enigmatic. Th e lack of adequate spontaneous joint repair and better strategies to induce joint repair will be a central fi eld of future basic and clinical research. Indeed, any potential for erosion self-healing is still poorly characterized. Exami na tion of sequential radiographs from clinical studies suggest that individual lesions can improve, especially when there is no or reduced swelling in the joint [29]. Other studies indicate that joint repair and erosion healing is rare despite eff ective therapy with TNF inhibitors [30]. More detailed imaging techniques such as ultrasound, magnetic resonance imaging (MRI), and computed tomography may provide better information in the future. Refi lling bone erosions might be an important clinical goal, if the technique could restore ligament and enthesial function. If such repair proves possible, it must be followed by an assessment of joint function.
Future frontiers in RA also will include the interaction between infl ammation and structural progression. With improved treatment options and tighter control of infl ammation, more patients will have low disease activity or will be in remission. Even patients who are considered to be in clinical remission, however, can progress in structural damage [31][32][33]. How much residual synovitis is necessary to allow structural progression is not yet clear. Even subclinical synovitis may suffi ce to trigger a progression of cartilage and bone damage followed by a decrease in joint function. Improved detection of synovitis with ultrasound and MRI may allow a better understanding of the eff ect of subclinical synovitis on joint structure [34][35][36][37][38].

Original thoughts on structural damage in PsA
For a long time, PsA was not recognized as a specifi c entity but rather was considered a subtype of RA that occurred in combination with skin psoriasis. Even after formal recognition, PsA was considered to be a mild disease with a benign course. Research in PsA has long lagged behind RA research in terms of diagnosis, prog nosis, and treatment. Th e diagnostic criteria of Moll and Wright, although not based on patient-derived data and omitting key features of PsA such as nail disease and dactylitis, were widely used [39]. Th ese criteria did not mitigate diffi culties classifying study patients, and there fore research remained limited. Th e Classifi cation Criteria for Psoriatic Arthritis now provide sensitive and specifi c classifi cations for PsA [40]. Research is still limited in early disease, however, as the Classifi cation Criteria for Psoriatic Arthritis were built on data from patients with long-standing disease.
PsA patients suff er signifi cant joint damage and disability over time. In accordance with RA, PsA is an erosive disease leading to the resorption of cortical bone. In addition, however, PsA shows morphological features discordant with RA; that is, the formation of bony spurs along the insertion sites of the entheses (enthesiophytes) [41]. Depending on the scoring system used, the damage and disability in PsA is less pronounced than in RA [42] or is equal to RA [43] with equivalent disease duration. Patients with RA and patients with PsA have similar functional and quality-of-life impairment [42].
Data from longitudinal cohort studies have helped identify severe disease with poor structural outcome. High infl ammatory activity and joint damage at the time of presentation are considered the most important predictors of future clinical and radiologic joint damage [44,45]. For instance, a high erythrocyte sedimentation rate at baseline and the presence of joint swelling suggest a poor prognosis with respect to structural outcome [44,46]. Moreover, patients with axial disease have more severe peripheral joint disease [47].
Previously, therapies for PsA were borrowed from RA, often without any specifi c studies to assess their eff ectiveness in this diff erent condition. Th ere is a surprising lack of randomized controlled trials evaluating the impact of DMARD therapy on PsA. Observational studies of patients receiving traditional DMARD therapy, however, have shown little control of structural damage. Observational controlled studies with sulfasalazine and gold have shown no reduction in long-term joint damage [48,49]. An observational cohort study of 23 patients who received 2-year MTX therapy concluded that MTX treatment did not reduce radiologic progression compared with matched controls [50]. However, a more recent analysis of the same cohort -but without controls -has suggested otherwise [51]. Chandran and colleagues found that since the mid-1990s MTX had been prescribed earlier and in higher doses, resulting in a signifi cant decline in actively infl amed joint count and psoriasis, and some decrease in progression of radiologic joint damage [51]. Th ere is no direct evidence, however, showing that DMARD therapy aff ects joint damage.

Novel insights gained from use of TNF blockers in PsA
TNF blockers have provided the fi rst evidence-based treatment for PsA with proven eff ects on arthritis, skin disease, enthesitis, dactylitis, and spinal disease. Th ese agents are highly eff ective in PsA, and they are the fi rst with proven effi cacy at reducing both active joint infl am mation and radiographic damage in randomized con trolled trials of PsA [52][53][54][55][56] (Table 2). Th e vast majority of PsA patients treated with TNF blockers showed no worsening in radiographic damage scores [52,[55][56][57][58]. Since the scoring systems used for the assessment of radiographic damage of PsA are the same as those used for RA, however, our knowledge about TNF-blocker eff ects on structural damage are confi ned to the erosive component of the disease, and it is unclear whether these agents also aff ect enthesiophyte formation.

Future needs and unanswered questions in PsA
Th e next step in investigating structural damage in PsA is to search for evidence of a direct link between infl ammation and joint damage. Imaging studies have elegantly demonstrated this link in RA, using a combination of MRI, ultrasound, and conventional radiography [63,64]. Such data, however, are currently unavailable in PsA. Also of interest is the link between infl ammation and new bone formation, which is typical for PsA but is not encountered in RA. As in RA, we must evaluate the use of anti-infl am ma tory therapies in PsA and investigate their ability to prevent long-term damage. If there is a direct link between infl ammation and damage in PsA, then tight control of infl ammation may arrest damage in PsA -as has been demonstrated in RA [65,66]. Does this also apply to enthesiophyte formation? Th e answer is unclear and at least doubtful, as the formation of bony spurs in AS is not infl uenced by TNF blockade. In this context it also will be important to defi ne treatment targets based on either clinical outcomes or imaging. For instance, a new clinical measure for minimal disease activity encompassing remission and low disease activity has been developed, but needs further testing in prospective studies [67].

Original thoughts on structural damage in AS
Low back pain is the earliest clinical manifestation for AS and indicates infl ammation in the sacroiliac joints and the spine, which can be identifi ed by MRI [68]. Disease progression is characterized by ongoing back pain lead ing to skeletal changes in the sacroiliac joints, identifi able by plain radiography. Th e diagnosis of AS has long hinged upon evidence of structural damage; the modifi ed New York criteria require the presence of radiographic sacro iliitis to give a defi nite diagnosis [69]. Studies have shown that it can take up to 10 years for these changes to become visible on plain radiographs [70], but radiographs are still widely used in established disease. Skeletal changes in the sacroiliac joints in AS are characterized by the concomitant presence of catabolic changes such as erosions as well as new bone formation leading to progressive ankylosis.
Spinal syndesmophytes are thought to appear at a later stage [71], although this hypothesis remains unclear. Th is concept is supported by two facts: patients in the preradiographic stages of AS can suff er just as much pain and stiff ness as those already diagnosed [72]; and patients treated early with TNF blockers demonstrate a better response than those treated later in their disease course [73]. Treatment should be started in the early stages of the disease process, before irreversible structural damage has occurred; that is, before the modifi ed New York criteria are fulfi lled. With the new classifi cation criteria of spondylarthritis it will be possible to start eff ective medication earlier, which may yield a considerable change of the disease course in the future.
Spinal structural changes in AS are quantifi ed radiographically using the modifi ed Stoke Ankylosing Spondylitis Spine Score, which grades the cervical and lumbar spine for the presence of erosions, squaring, sclerosis, syndesmophytes, and bony bridging at each site [74]. Th e  [61] week 12

Etanercept
Mease and ACR20 at mTSS at x x x x x colleagues [58] week 12 months 6 and 12 Adalimumab Mease and ACR20 at mTSS at x x x x x x x colleagues [54] week 12 week 24 Golimumab Kavanaugh and ACR20 at NA x x x x x x colleagues [62] week 14 inability, however, of the modifi ed Stoke Ankylosing Spondylitis Spine Score to assess the thoracic spine [75], which is the most commonly involved area as shown on MRI studies [76], limits the score's sensitivity to assess change. Spooren berg and colleagues have demonstrated that a minimum 2 years of follow-up is necessary to reliably detect radio graphic change, meaning that studies assessing radio graphic damage must be of signifi cantly longer duration than similar studies in peripheral arthritis [77]. Traditionally, AS has been treated with regular physiotherapy and nonsteroidal anti-infl ammatory drugs (NSAIDs). Indeed, the only evidence for a reduction in radiographic progression in patients with AS is from a trial of continuous versus on-demand treatment with NSAIDs. Patients on continuous NSAID therapy had signifi cantly reduced radiographic progression compared with those who took the therapy only when serious symptoms were present [78]. Both groups experienced similar eff ects on symptoms, infl ammation, and spinal mobility.

Novel insights gained from use of TNF blockers in AS
Th e advent of TNF inhibitors has greatly improved the treat ment options for AS. Th ey allow treatment of patients with severe disease who do not fully respond to NSAIDs. Th ree TNF blockers are licensed and approved worldwide, and a fourth blocker (golimumab) was recently approved. Criteria have been set by various regulatory bodies for their use [79,80].
Similar effi cacy has been found for all of the TNF blockers, although studies have consistently shown that patients relapse with cessation of therapy [81,82] ( Table 3). Despite the strong anti-infl ammatory eff ect of TNF blockers in AS, these agents do not infl uence new bone formation in AS [83][84][85]. Only one small study showed reduced radiographic progression in AS patients treated with infl iximab in comparison with a historical cohort, but these results have to be considered with caution as the baseline Bath Ankylosing Spondylitis Disease Activity Index score was diff erent between the groups [86,87].
Th is lack of a structure-sparing eff ect of TNF blockers in AS unravels the diff erent pathophysiologic mechanisms underlying RA, PsA, and AS. RA is typically charac ter ized by bone erosion, whereas the main structural out come in AS is bony spur formation based on bone formation.
Radiographic damage at baseline is a major predictor of future radiographic progression; in particular, the presence of syndesmophytes or ankylosis [86,88]. MRIevident sacroiliitis and positivity for HLA-B27 have been shown to predict the development of radiographic sacroiliitis in patients with early infl ammatory back pain at 8-year follow-up, with a sensitivity and specifi city of 77% each [89]. MRI has now been incorporated into the new Assessment of Spondyloarthritis International Society classifi cation criteria for axial spondyloarthritis [90].
Th e relationship between infl ammation and new bone formation in AS remains unclear. Recent MRI studies have shown that active corner infl ammatory lesions (also known as Romanus lesions or bone edema) predict the development of syndesmophytes [91]. Th ese studies also demonstrated formation of syndesmophytes at the exact location of resolved infl ammatory lesions. One explanation is that there may be persistent mild infl ammation not detected by MRI. Th e discovery that syndesmophytes were more likely to develop at the sites of resolved corner infl ammatory lesions rather than those of persistent lesions, however, led to the TNF brake hypothesis. Th is hypothesis suggests that TNF triggers pathways leading to new bone formation, but that while there is active infl ammation TNF suppresses new bone formation via dickkopf-1 (a regulator of joint remodeling) [92]. When patients are treated with TNF inhibitors, therefore, infl ammation resolves, the brake is released, and tissue repair and new bone formation occur [91]. Th is process may account for radio graphic progression in patients who appear to otherwise respond well to anti-TNF therapy. Evidence for uncoup ling between infl ammation and new bone formation is supported by a mouse model of spondyloarthritis, which showed no eff ect of etanercept on the severity and incidence of joint ankylosis [93]. An independent study of patients receiving anti-TNF agents found that an infl amed vertebral edge at baseline had a threefold increased risk to develop a syndesmophyte than a non-infl amed vertebral edge [94]. Th ese results contrast with those of Maksymowych and colleagues [91]. MRI scans were performed only at baseline and 2 years, however, so it is possible that infl ammation had occurred and resolved between scans. Bennett and colleagues described fatty Romanus lesions in the spine, which they suggest may be the postinfl ammatory phase between osteitis on MRI and sclerotic bone formation on radiographs [95].

Future needs and unanswered questions in AS
Despite the effi cacy of TNF blockers for symptomatic control and improved quality of life in patients with AS, the lack of effi cacy for radiographic progression is noteworthy. Longer studies may be needed, because the process linking infl ammation and new bone formation is slow. Eff ective suppression of infl ammation may thus still reduce radio graphic progression.
Additional research is needed to analyze whether progression is due to persistent, low-grade infl ammation or to the release of the TNF brake once infl ammation is eff ectively treated. Th e answer to this question will guide future therapies. Perhaps a dual approach will be necessary: one therapy to treat infl ammation and another to prevent new bone formation.
MRI has facilitated the study of early disease, and the treatment response of patients in the pre-radiographic stage may help determine whether suppression of infl ammation can prevent early onset of structural damage. Since not all patients with MRI-evident sacroiliitis develop AS, however, treatment must be carefully targeted. Lastly, there is the issue of late presentation of AS patients to rheumatologists, which can be improved by the education of both doctors and patients.

Competing interests
ZRA has received research grants from Schering-Plough Corporation, now Merck Sharp & Dohme. PGC has received research grants from Pfi zer, and has been a speaker for Merck & Company, Inc. and AstraZeneca. The remaining authors declare that they have no competing interests.