The need for new classification criteria for rheumatoid arthritis


Over the past decade, there has been breathtaking progress in the evolution of both basic research and clinical insights into rheumatoid arthritis (RA): new insights into pathophysiology, new measures to determine outcome, new therapeutic strategies, and new therapeutic compounds have emerged (1–13). In parallel, the recognition that RA needs to be diagnosed early and treated promptly with disease-modifying antirheumatic drugs (DMARDs) to most successfully interfere with the disease process and its progression to major damage and disability has become a new paradigm (14–20). This paradigm, combined with novel treatment options, has translated into improved prognosis for patients, with significant success with respect to improvement of global disease activity, retardation of joint damage, prevention of disability, and reduction in mortality (21). Nevertheless, in many patients, disease activity and progression cannot be reduced to a clinically meaningful extent once RA is established, and in many who experience significant improvement, there is residual activity (22). While remissions are rare in established disease, remission is the ultimate aim of therapeutic intervention, and the improvements achievable today have put this goal on the agenda.

Interference with the disease process may be easier in the very early stages of the disease. In fact, most recent evidence obtained in early-arthritis clinics indicates that very early DMARD treatment, i.e., instituted within 3 months of symptom onset, may be highly effective in leading to remission and preventing major radiographic progression (23–25). This suggests the existence of a “therapeutic window of opportunity.” However, as much as scientists working in early-arthritis clinics attempt to establish a correct diagnosis of RA, they usually do so based on subtle findings due to lack of appropriate criteria even though, as noted above, prompt diagnosis has prognostic implications given the greater likelihood of progressive joint destruction and disability in patients who are not treated early.

The currently widely used 1987 RA classification criteria of the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) (26) were developed using patients with established RA (mean disease duration close to 8 years) and had a sensitivity of 91% and a specificity of 89%. Among the 7 components of these criteria, 1 is a radiologic criterion that is met only rarely in early disease; among patients with very early arthritis (symptom duration <3 months [median 8 weeks]) who all developed RA, erosions were found in only 13% (27), whereas joint erosions are present in 50–70% of RA patients 2 years after onset (28). Another component of the 1987 criteria is serologic in nature, i.e., rheumatoid factor (RF), which also is seen less frequently in early than in established disease (29). Among the 5 clinical components of the criteria set, rheumatoid nodules are very rare in early RA. Thus, only 4 of the 7 components of the ACR criteria set are clinical and joint related and may therefore have a reasonable probability of being fulfilled in early RA. Moreover, 1 of these remaining 4 clinical criteria relates to involvement of at least 3 joint areas whereas RA frequently initially presents as a mono- or oligoarticular disease, further reducing the diagnostic potential of these criteria in early RA.

Not surprisingly, therefore, the sensitivity of the ACR classification criteria in early RA ranges from 40 to 60% and the specificity is no better than 80–90% (15, 27, 30–33). Patients with many non-RA arthritides, such as systemic lupus erythematosus, polymyositis, or viral arthritides, may fulfill the 1987 criteria, which do not list a series of exclusions. Taken together, these data indicate that the usefulness of the ACR criteria in early RA is low.

The 1958 criteria (34) contained 11 components, 2 of which were histologic in nature and 1 of which required synovial fluid analysis. Like the 1987 criteria, they also included radiographic changes, rheumatoid nodules, and RF as components. The remaining 5 were related to signs and symptoms of joint involvement; 2 of those even allowed a finding of pain or swelling in a single joint. Interestingly, in various studies (35) the sensitivity and specificity of the 1958 criteria were very similar to or even higher than those of the 1987 criteria. The 1958 criteria were subsequently modified in Rome in 1961 (36, 37), at which time the 2 histologic components and the synovial fluid analyses were omitted, leaving a set of 8 criteria. A patient was considered to have definite RA if 5 of the 8 (or 11) components of the criteria set were met. In addition, a series of exclusion criteria were added, mainly relating to other defined disorders. Moreover, with these older criteria, a patient would be classified as having probable RA if 3 or 4 of the components were met and the exclusion criteria were fulfilled.

What is needed at the present time, however, is a set of criteria that are useful for the stage of the disease at which physicians most frequently need the help of criteria, namely, very early in its course, when the decision to initiate a clinically beneficial but potentially harmful therapy must be made. Such criteria have never been elaborated consensually for early arthritis. They would have important implications regarding therapy, and inaccurate classification could therefore lead to undertreatment or overtreatment aside from simply misinforming the patient and other clinicians. This implies that these criteria should have prognostic implications—to ultimately prevent the occurrence of an unfavorable outcome by increasing the likelihood of appropriate early intervention. Initial attempts using prospectively studied patient populations have been made (32, 33), but they need further validation in different cohorts of patients with very early arthritis followed up prospectively.

Prognostication in early arthritis relates primarily to persistence of disease and also to the propensity of persistent disease to cause joint damage. The latter is the more serious sequela of joint inflammation requiring the most intensive therapeutic intervention, although persistent inflammation even in the absence of a major destructive component is also associated with disability. While the presence of radiologic changes is associated with an unfavorable prognosis (38–42), these changes are rare in the very early stages of disease that develops into (destructive) RA (27, 28, 43). However, new imaging modalities have entered the field, such as sonography and magnetic resonance imaging, which allow discernment of soft tissue changes with more accuracy than clinical examination and might be useful for early detection of erosions (44).

Like radiologic abnormalities, RF is one of the components of both the 1958 (1961) and the 1987 criteria for RA. And like the former abnormalities, RF has been regarded as a poor prognostic marker (17, 42, 45). Significant evidence has accumulated indicating that the mere presence of RF has only limited sensitivity and specificity in early RA. However, use of a cutoff value of 50 IU/ml greatly increases specificity while only mildly reducing sensitivity (46–48). Another recently recognized addition to the spectrum of autoantibodies in RA is the group of antibodies directed at citrullinated peptides, which have a similar sensitivity to and somewhat higher specificity than RF (33, 47, 49). Given that these autoantibodies, as well as high-titer RF, are associated with both persistence and destructiveness of RA (33, 43, 47, 49), it might be worthwhile to consider them as an addition to future diagnostic or classification criteria for (early) RA. Whether the association of the shared epitope with RA is independent or is related to the presence of autoantibodies is still a matter of debate (45, 50–52).

Although there has been so much activity and progress in RA research and management since 1987—not only the development of novel therapeutic agents and strategies making remission a realistic therapeutic goal, but also the establishment of new disease activity criteria (5, 7, 9, 53, 54) and the advances in serology and imaging mentioned above (55)—disease criteria have not changed. Would it not be timely to reevaluate the classification (or, as they were called in 1958, diagnostic) criteria for RA? To assess the usefulness of a set of criteria for persistent and/or destructive arthritis that might better address the issues arising during very early stages of joint inflammation? To examine if we have sufficient knowledge to distinguish such very early RA populations among patients with very early arthritis? To perform thorough evaluations and subsequent validations in prospectively collected patient cohorts?

Thus, new criteria will ideally need to shift from classifying established RA to classifying a subtype of early arthritis that is projected to become persistent and erosive (rheumatoid) disease. This persistence of symptoms is an explicit part of the current ACR criteria, and new criteria will have to replace this requirement of already present persistence with reliable prediction of persistence. However, it will be necessary to critically assess whether sufficient evidence for prognostic markers already exists (and which ones these would be) or whether this needs to be part of a future research agenda fostered by developing preliminary criteria based on current insights. What one would call such criteria and whether one would include the term “rheumatoid” are of secondary importance.

Such criteria would not be meant to define a new subset of the disease but rather to help to identify persistent (and likely erosive) arthritis as early as possible, allowing it to be distinguished from other forms of early arthritides. The utility of new criteria for very early RA could be manifold: first, they would provide the clinician with tools to improve patient care in accordance with the treatment paradigm changes of recent years; second, they could allow the study of various groups of patients fulfilling different components of the criteria to further analyze the heterogeneity and consequently possible underlying pathways of RA; and third, the ultimate therapeutic goal, i.e., curing RA or preventing the development of erosive (or persistent) disease, could be tested better by attempting to reach this goal in patients who fulfill the criteria.

It would, in fact, be timely for the European League Against Rheumatism (EULAR) and the ACR to act in liaison to revise the criteria for RA—the ACR with its long-term history of developing criteria and EULAR with its long-term devotion to the study of early arthritis. It is also timely to provide rheumatologists worldwide with a scheme that would allow them to escape the current difficulty in classifying RA early, or, better, to subclassify patients with any type of early inflammatory joint disease according to prognosis. Such new criteria might materialize within a few years and constitute a good basis for the celebration of the 20th anniversary of the 1987 (or the 50th anniversary of the 1958) criteria that have served rheumatologists well for decades but whose era has come to an end.