The diagnosis and prognosis of early arthritis: Rationale for new prognostic criteria


Accurately predicting which patients will have severe arthritis and need maximal treatment and, by implication, which patients will have mild disease and need minimal treatment is an aspiration of all rheumatologists. Although most specialists rely on clinical acumen and personal experience to make this assessment, a recurring theme of clinical investigators is identifying objective biomarkers that will reliably predict the outcome of early arthritis. In this issue of Arthritis & Rheumatism, Visser and colleagues report their experience in predicting outcomes at 2 years in their Early Arthritis Clinic in Leiden (1). Their report provides an opportunity to reflect on the relevance of such prognostic studies. This editorial reviews 3 related areas: the clinical relevance of erosive damage at 2 years, prognostic factors for erosive rheumatoid arthritis (RA), and the need for new classification criteria.

Dividing patients into erosive and nonerosive patterns of disease seems inherently sensible because it is inconceivable that joint destruction does not inevitably cause some disability. However, the relationships between radiographic damage and disability in RA are complex (2). Damage and disability are unrelated in early disease. It takes ∼5 years for significant relationships to appear. In established RA, there are significant correlations with joint damage, accounting for 25% of disability.

At all stages of RA, disease activity has a crucial role in disability. The relationship between damage and disability has been delineated by Drossaers-Bakker and colleagues (3, 4), who observed the evolution of disability in 105 RA patients. After 12 years, there were strong correlations between joint damage and Health Assessment Questionnaire (HAQ) scores, with dominant effects from large-joint damage. However, even in longstanding RA, disease activity was the main determinant of disability. Similar findings have been reported by Wolfe (5) based on >30,000 observations in 1,843 RA patients. He found that HAQ scores were high at disease onset, increased slowly with time (0.03 units per year), and were more influenced by disease activity, pain, and psychosocial factors than by structural abnormalities. These studies suggest that disability will be limited if erosive damage can be prevented, provided that there are concomitant therapeutic strategies aimed at decreasing disease activity.

Most experts concur that 2 years is a reasonable time point at which to divide erosive from nonerosive RA. In a review from 1995, van der Heijde reported that ∼75% of patients with early RA develop joint erosions, and most of these patients had developed erosions within the first 2 years of RA (6). Three subsequent long-term, prospective observational studies of radiographic progression all indicated that 2 years is a watershed in the development of erosions. However, these reports provide differing perspectives about the progression of erosive damage. Plant and colleagues (7) described erosive progression in 126 patients with early RA who were followed up for 8 years. Twenty-nine of these patients had nonerosive disease, 51 showed linear progression, and small numbers showed lag or plateau patterns. After 2 years, the course of radiologic progression was variable and not predicable. Another Danish report (8), which discussed the findings in 109 RA patients who were followed up for up to 30 years, also identified several patterns of progression. These patterns included no progression, which was rare, progression of slow onset and then exponential or linear growth, and progression of rapid onset that subsequently stabilized. Wolfe and Sharp (9) reported progression in 256 RA patients seen within the first 2 years of disease and followed up for up to 19 years. They found that radiographic damage persisted at a constant rate, and was neither greater in early RA nor reduced in late disease. The balance of evidence suggests that patients who have nonerosive disease by 2 years are unlikely to show subsequent erosive damage, but that not all patients with erosions at 2 years subsequently have progressive joint damage.

Many studies have attempted to predict outcomes and develop prognostic markers, especially in early RA. A comprehensive review of prognostic markers for severe RA was completed during the preparation of the Scottish Intercollegiate Guidelines Network (SIGN) guidelines for the management of early RA (10). Although these guidelines highlight the difficulties in predicting the outcome of RA in individual patients at disease onset, they suggest several indicators of poor outcome. These include many joints with active disease, a high initial erythrocyte sedimentation rate or C-reactive protein (CRP) level, positivity for rheumatoid factor, early radiologic erosions, poorer function at disease onset, and adverse socioeconomic circumstances and lower education level.

Another important resource for potential prognostic markers is the Norfolk Arthritis Register (NOAR), a community-based inception cohort of patients with inflammatory polyarthritis in rural England. Using regression analyses, Brennan and colleagues (11) identified several key risk factors for developing erosions. The most important were rheumatoid factor positivity, a history of arthritis exceeding 3 months, the involvement of 5 metatarsophalangeal (MTP) joints, and the involvement of 2 or more large joints. A comparison of the findings of Visser and colleagues (1) with the conclusions of the SIGN guidelines and the NOAR is shown in Table 1.

Table 1. Comparison of prognostic factors identified in cohort studies*
Finding, outcome predictorVisser et al (ref. 1)Brennan et al (ref. 11)Scottish Intercollegiate Guidelines Network (ref. 10)
  • *

    MTP = metatarsophalangeal; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; CCP = cyclic citrullinated peptide.

 Joints with active disease≥3 joints≥2 large jointsMany joints
 Rheumatoid factor positivityPositivePositivePositive
 Radiologic erosionsImportantNot assessedImportant
 Duration of diseaseProlonged symptoms>3 monthsNot assessed
 MTP involvementImportantImportantNot assessed
 Morning stiffnessImportantNot importantNot important
 High ESR/CRP levelNot importantNot importantImportant
 Poor functionNot importantNot importantImportant
 Social deprivationNot importantNot assessedImportant
 Anti-CCP antibodiesImportantNot assessedNot assessed

It is widely believed that an elevated acute-phase response, especially a high CRP level, predicts subsequent radiologic damage (12). Van Leeuwen and colleagues (13) showed, in a prospective study of 149 patients with early RA, that in individual patients, time-integrated CRP values correlated closely with radiologic progression. However, variations between patients with similar radiographic scores make it difficult to generalize from initial single CRP values in individual cases. Such individual variation may explain why neither data from the Leiden study nor data from the NOAR show that high CRP levels predict outcome. A comparable study from Leeds (14), in which 63 patients with early RA were followed up for 6 months, also showed that high initial CRP levels did not predict the persistence of arthritis at 6 months. The conventional view that high CRP levels indicate a poor prognosis may not apply in early RA.

In contrast to the findings with CRP, there is little doubt that rheumatoid factor positivity in early RA predicts the development of erosions. This has been confirmed in many cross-sectional and longitudinal cohort studies. The best prognostic measure of rheumatoid factor is debatable. Some reports suggest high titers of rheumatoid factor are most predictive. Others favor the use of IgA rheumatoid factor for prediction. At present, the optimal method remains uncertain (15).

Should rheumatoid factor positivity be supplemented by anti-cyclic citrullinated peptide (anti-CCP) enzyme-linked immunosorbent assay (ELISA) tests? This question needs to be placed within its historic context, which started with the identification of antikeratin antibodies (AKA). It has been known for more than 20 years that AKA are specifically associated with RA, but because they can only be detected by immunofluorescence, their value as a routine diagnostic test has always been dubious. Subsequent research showed that the antigen for AKA is filaggrin, which is found intracellularly in the normal buccal mucosa. Thereafter, more specific antifilaggrin antibody tests were developed. Aho and collaborators (16) defined the interrelationships between antifilaggrin antibodies and rheumatoid factor positivity in a case-control study nested within a Finnish cohort of nearly 20,000 adults without arthritis. After 10 years of observation, 226 patients had developed RA; 89 were rheumatoid factor positive and pre-illness serum levels of antifilaggrin antibody were directly proportional to the risk of rheumatoid factor-positive RA. Antifilaggrin antibodies did not predict the development of rheumatoid factor-negative RA.

The subsequent development of synthetic peptides containing citrulline, an amino acid present in filaggrin, enabled the introduction of an accurate ELISA. This assay has similar specificity as that used to identify AKA but overcomes standardization problems associated with immunofluorescence. Initial reports using the anti-CCP ELISA (17) suggested that it has an excellent specificity for the diagnosis of RA. An earlier report from Leiden found that combining anti-CCP and IgM rheumatoid factor ELISAs gave a high positive predictive value for RA of 91%, with a similarly high value for predicting erosive disease at 2 years (18).

Despite the high specificity of AKA and anti-CCP ELISAs for RA, their value as an additional diagnostic criterion is uncertain. In 1993, Von Essen and colleagues (19) reported a 3-year prospective study of 308 hospital patients with recent-onset inflammatory arthritis, comparing AKA with rheumatoid factor. The latex test for rheumatoid factor was the most sensitive (72%) but the least specific (86%). The AKA test was most specific (96%) but the least sensitive (33%). However, the actual classification impact achieved using AKA compared with using rheumatoid factor as the sole laboratory criterion was very modest, mainly because the criteria were strongly interrelated. A similar study by Kroot and colleagues (20) using an anti-CCP ELISA found that almost 70% of RA patients were positive for anti-CCP in early RA and they had more radiologic damage. However, multiple regression analyses demonstrated that the additional predictive value given by anti-CCP ELISAs was only moderate (20). Another report, by Van Jaarsveld et al (21), evaluated the clinical value of the anti-CCP ELISA in combination with rheumatoid factor status in 249 patients with early RA. Van Jaarsveld and colleagues concluded that the prognostic value of combining both tests lies mainly in their ability to predict mild disease. The report by Visser and colleagues (1) also suggests that anti-CCP ELISAs have limited value as a diagnostic criterion for early RA.

In one final area of prediction, using genetic markers, the evidence remains tantalizing, but incomplete. Visser and colleagues found a modest benefit in using the shared epitope and DQRA, which did not merit its consideration as a classification criterion. Mottonen and colleagues have also found no impact of genetic markers on the outcome of treatment in early RA (22). In contrast, the Leeds group has reported that genetic assessments can be used to discriminate between self-limiting and persistent arthritis (14). There have been many other studies about genetic markers with conflicting results and, for the present, it seems premature to use genetic markers as either diagnostic or prognostic criteria.

The most crucial question raised by Visser and colleagues (1) is whether or not the American College of Rheumatology (ACR) 1987 classification criteria for RA (23) are suitable in early RA. Other groups have also highlighted the weakness of these criteria. For example, experience with the NOAR has shown that the ACR criteria do not perform well for identifying which patients with early inflammatory polyarthritis will later develop RA (24). Despite this, the NOAR group believes that because the ACR criteria provide consistency in patient selection between studies, it is premature to discontinue their use (25). The argument for and against developing new criteria is finely balanced. Points in favor of changing the criteria include, first, the importance of accurately assessing early RA, for which purpose the ACR criteria perform poorly. Second, as new tests and imaging methods become available, it will be possible to subdivide patients with early arthritis in different ways. Finally, expensive disease-modifying therapies, especially immunotherapy, should only be given to patients at a high risk of severe disease, and the current ACR criteria do not help in identifying such cases.

The advantages in favor of changing the ACR criteria are counterbalanced by arguments against change. First, those studies suggesting that new criteria are needed come from highly selected cases in research programs run by extraordinarily motivated clinicians. They often use clinical methods (such as the MTP compression test) or laboratory investigations (such as anti-CCP assays) that are not widely used. Their results may not be generalizable to routine practice. Second, changing diagnostic criteria makes it difficult to compare current results with data from past studies collected using previous criteria. Finally, there is no evidence that the development of new criteria will change disease management or improve RA outcomes, and it is important to avoid change for change's sake.

Given the complexity of these issues, it is possible to produce other equally cogent arguments for and against change. Overall, it seems unwise to entirely resist change when so many experts in the field express concern about using the current ACR criteria in early polyarthritis. It therefore seems sensible to suggest the following points.

  • 1It is not necessary to introduce new diagnostic criteria for RA, since this may produce more problems than benefits. Disadvantages of changing criteria include loss of continuity with previous studies, unknown effects on established, as opposed to early, RA, and potential confusion for nonexperts in the field. Changing diagnostic criteria should be deferred until there is better knowledge about the cause of RA.
  • 2It is sensible to produce new criteria to define prognosis in early arthritis, based on the need to differentiate patients likely to have progressive destructive disease from those with self-limiting or nonerosive disease. Such criteria should be derived from research studies led by experts but applicable in routine clinical practice, using methods available to all practicing clinicians. Special tests such as the ELISA for anti-CCP antibodies should not form part of the prognostic criteria because they are not widely available.
  • 3Developing such criteria requires international collaboration. The criteria will need revision when new methods became widely available for use in routine clinical practice. The existence of high-cost treatments makes it important to utilize such prognostic criteria to avoid giving unnecessarily expensive treatments to patients with early arthritis whose disease is unlikely to progress.

Initially, simple agreements on standardization are needed. The importance of such standardization is illustrated in 2 reports. One is a comparison of 4 different inception cohorts of patients with early RA (26). In that report, Albers and colleagues showed that, despite similar inclusion criteria, there were significant differences in demographic factors and baseline disease activity between cohorts. There were also major differences in the types of treatment received. The other report, by Berthelot and colleagues, is a review of 34 reports on 23 cohorts of patients with early arthritis (27). That review showed that the methods used were often poor, there were numerous inclusion and exclusion biases, followup periods were often too short, and there was imprecision in making a diagnosis. Agreement regarding the optimal data to collect in prospective observational studies would be the first step along the road to international standardization.