The traditional view of rheumatoid arthritis (RA) is that an unknown event or a combination of circumstances triggers synovial inflammation. This synovitis leads to pain, soft tissue swelling, stiffness, and the characteristic erosions and radiographic joint destruction seen in a large proportion of patients. Clinicians have suspected for some time that this model of RA is not consistent with the experience of many patients. This view can now be strengthened by evidence from a variety of sources, including therapeutic clinical trials, findings in animal models of arthritis, histologic studies in humans, and community-based investigations. The report by Molenaar and colleagues in this issue of Arthritis & Rheumatism (1) offers further evidence that the link between inflammation and erosions is not clear-cut.
Most patients with RA who attend specialty clinics have moderate to severe disease manifested by persistent (although variable) signs of joint inflammation and progressive (although variable) joint destruction. It would be natural to assume that these manifestations of disease are aspects of a single pathology, but closer examination of the data suggests otherwise. Among patients with unequivocal clinical synovitis, about one-fourth will never have radiographic joint destruction (2). Furthermore, patients in the community in whom mild disease develops may never be referred to specialty clinics and are even less likely to experience development of erosions (3). Among patients in whom radiographic changes do develop, the correlation between symptoms and signs of inflammation and erosive progression is only ∼0.7 (4,5). This correlation occurs only in the patient as a whole but not in individual joints, reflecting the association of these 2 consequences of RA but not necessarily proving that one causes the other.
Molenaar et al (1) report on 187 patients with established RA (median disease duration 7 years) in clinical remission (according to modified American College of Rheumatology criteria or a disease activity score of <1.6) who were followed up prospectively for 2 years, with reassessment of their clinical status every 3 months. In approximately half of the patients, disease remained in remission throughout the followup period, but a statistically significant deterioration in erosion scores was observed in this group. Although less progression was observed in patients whose disease remained in remission than in patients in whom remission was not maintained, in some of the former patients the rate of progression was very high. Indeed, 2 of the patients whose disease was in remission scored the third and fourth highest rates of radiographic progression in the cohort. This finding—that patients with little inflammatory activity continued to experience joint destruction—fits with the weak correlation between clinical synovitis and erosions in individual joints (6).
Further evidence for a dissociation between clinical signs of inflammation and progression of joint destruction comes from randomized, controlled trials of different treatment regimens. Use of low doses of prednisolone taken in addition to traditional therapy in patients with early RA has now been shown to substantially reduce radiographic progression (7). An important aspect of the effects of glucocorticoids is that such therapy improves the signs and symptoms of RA for only a few months, but the ability of glucocorticoids to suppress joint destruction persists for at least 2 years (7). Thus, antiinflammatory agents will suppress clinical synovitis but will not prevent erosions, some disease-modifying antirheumatic drug therapies have a partial effect on erosion progression (8), while low-dose prednisolone will affect swelling and tenderness only transiently but does control progression of erosion in the longer term. These observations show that the 2 aspects of RA, clinical synovitis and radiographic joint destruction, respond differently to therapeutic intervention and support the notion that these 2 outcomes must result from different disease processes.
Inflammation of the synovium has been seen primarily as a lymphocytic phenomenon, and the accumulation of lymphocyte aggregates and germinal centers in the synovium is considered to be the hallmark of RA (9). However, these changes are not pathognomonic and can be recorded in other types of arthritis, including osteoarthritis (10). In one animal model of RA (11), human RA synovial tissue was implanted with normal human articular cartilage into SCID mice (which cannot reject or react to foreign tissues). When implanted subcutaneously or under the renal capsule (and in the absence of human or animal T cells and B cells and circulating human blood components), these implants develop proliferating fibroblast-like cells at sites of cartilage and bone invasion and destruction that bear a remarkable resemblance to the human articular cartilage lesion in RA (12).
The presence of synovial macrophages at sites of joint destruction has led to the suggestion that these macrophages may play an important role there. This view is corroborated by observations of synovial samples obtained at the time of arthroscopy from patients with progressive, erosive disease (13). Macrophage numbers in the synovial lining and sublining layers correlated with radiographic outcome in RA, while the presence of lymphocytes in the sublining layer did not do so. In contrast, lymphocytes did correlate with clinical and serologic assessments of synovitis.
It seems that a body of evidence from a variety of sources, including the latest report by Molenaar et al (1), points away from the concept that RA causes inflammation, which in turn causes joint damage. An alternative interpretation that can accommodate this evidence points to at least 2 pathologic processes working simultaneously in the joint, one leading to signs and symptoms of inflammation, and the other leading to direct joint destruction by synovial cells. This suggestion was first made >20 years ago, based on the comparative actions of nonsteroidal antiinflammatory drugs and disease-modifying antirheumatic drugs (14), but the more recent evidence from trials of glucocorticoids and developments in histologic techniques and animal models, and the new report by Molenaar et al, add a powerful drive for it to be seriously considered.
The implications of this simple change in perspective for understanding the disease and developing new treatments are fundamental. However, a clearer understanding of this model of RA depends on the recognition that there are assumptions implicit in the current approach to the evaluation of clinical outcome that tend to pass unnoticed. Neither clinical synovitis nor radiographic joint damage is representative of a single process, although they are often treated as if they were so. The inflammatory component of synovitis is best recognized by the simultaneous presence of both swelling and tenderness, but accumulation of the destructive synovial tissue at the site of cartilage and bone invasion will also cause swelling. Radiographic joint damage in RA has several components, including juxtaarticular osteoporosis, joint space narrowing due to generalized cartilage loss, and focal bone and cartilage erosions. Persistent synovitis may result in the release of inflammatory mediators into synovial fluid, in a number sufficient to cause some generalized cartilage loss (15), and most radiographic scoring systems concentrate erosions and joint space loss into a single overall score for damage. The disease model shown in Figure 1 allows for these complications.
If this model of RA more accurately represents the true nature of the disease, then a variety of predictions deriving from the model should prove to be correct. For example, the model predicts that within the natural history of progressive radiographic damage to joints, there will be times when joint space narrowing, representing generalized cartilage loss in response to persistent inflammation, will occur at a rate different from that of progression of discrete bony erosions. Other predictions are that clinical signs of inflammation will relate more strongly to joint space narrowing than to erosion development, that the therapeutic response to prednisolone will uncouple the weak correlation between synovitis and erosions in the patient as a whole, and that other therapeutic interventions might affect clinical synovitis differently from the manner in which they affect erosive progression. Some of these effects have now been observed (7, 16).
Other predictions are that it will be possible to distinguish between 2 types of synovial hyperplasia by using some joint imaging techniques, which has already been reported (17), that histologic evidence of macrophage numbers will correlate with the response to glucocorticoid treatment, and that serum markers of macrophage activation (e.g., production of metalloproteinases) will correlate with erosive progression. Such a relationship has been reported previously (18, 19). In the SCID mouse/human cartilage model of RA, we might expect cartilage destruction to be suppressed by glucocorticoids. Finally, we may find different risk factors and prognostic indicators for patients with erosive disease, those with nonerosive disease, and those with more or less clinical inflammation. Such factors may include a genetic predisposition to low thresholds for activating the different tissue reactions involved in this disease.
There are important implications for therapeutic studies and the introduction of new treatments. Clinical trial results, their interpretations as outlined above, and the new report by Molenaar and colleagues (1) make it imperative that the reasons for and expectations from therapeutic intervention in RA are clearly enunciated, in both routine clinical practice and therapeutic studies of new and existing agents. Adding low-dose prednisolone to the treatment of early disease will help control symptoms rapidly but only transiently, while continued treatment will go on to suppress erosive progression in spite of continuing symptoms. Such a distinction may be missed if composite indices of clinical outcome are used. For new therapies, the assumption that the control of inflammation, even by direct attack on the cells thought responsible, will result in suppression of erosions must be resisted. However, this model of RA raises the possibility of new therapeutic targets, such as macrophage or fibroblast control. The HLA system explains only part of the genetic contribution to the incidence or severity of RA. This model raises the possibility that the key to erosive disease may lie elsewhere in the genome (20).
Finally, the model shown in Figure 1 does not preclude further elaboration as understanding of the pathophysiologic processes in the joint develops. This may involve greater appreciation of the role of mast cells (21) or a recognition of the part played by polymorphs, which seem to be under attack within the inflamed joint (22). It is interesting that in antigen-induced arthritis in the rabbit (23), reduction in the availability of polymorphs, which are thought to be a component of the inflammatory response responsible for joint pain and swelling, actually results in greater joint destruction (24). Furthermore, by measuring the serum enzyme cytidine deaminase in the study of low-dose prednisolone, the investigators observed an increase in joint polymorph turnover in those patients with reduced erosive progression (25). It is possible that several pathologic processes are proceeding in parallel in the rheumatoid joint, and that polymorphs may have a different role than previously thought.