Rheumatoid arthritis is a systemic autoimmune disease of unknown origin. To prevent joint destruction, early diagnosis and treatment is required. The diagnosis can be made by the 1987 classification criteria of the American College of Rheumatology (ACR),¹ but these criteria have a low sensitivity in early arthritis.²

There are a few rheumatoid arthritis specific antibodies. These include the so called antiperinuclear factor (APF), anti-filaggrin antibodies (AFA), and antikeratin antibodies (AKA). The epitopes recognised by APF, AFA, and AKA were found to be generated by a post-translational modification—namely deimination of the natural amino acid arginine to the amino acid citrulline by activity of peptidylarginine deiminase.³ Based on that knowledge, Schellekens and co-workers developed an enzyme linked immunosorbent assay (ELISA) using a cyclic citrullinated peptide (CCP) derived from the sequence of human filaggrin as substrate.⁴ The assay was later improved (the second generation anti-CCP test) and sensitivities of 70–80% at specificities of 98–99% have been reported in established rheumatoid arthritis and controls.^5,6 Sensitivity in early arthritis cohorts for the diagnosis rheumatoid arthritis varies between 40% and 70%.^{4,7,8,9,10,11} Although the anti-CCP ELISA has a reasonable sensitivity, the cyclic citrullinated peptide is not the physiological target of the autoantibodies.

Citrulline containing antigens are expressed in rheumatoid arthritis synovium.^12,13 Moreover, B cells that actively secrete anti-CCP are specifically present in bone marrow and synovial fluid of anti-CCP seropositive patients with rheumatoid arthritis.^14,15 When searching for the nature of citrullinated proteins in rheumatoid synovial tissue, Masson-Bessière et al identified citrullinated α and β chains of fibrin as the target for APF, AFA, or AKA positive sera.¹⁶ Subsequently, antibodies to in vitro citrullinated fibrinogen, a soluble precursor of fibrin, have been described as a serological criterion for the early diagnosis of rheumatoid arthritis when compared with rheumatoid factor (RF) and the first generation anti-CCP assay.¹⁷

To study the diagnostic and prognostic value of anti-citrullinated fibrinogen (ACF) in early arthritis in comparison with IgM-RF and the second generation anti-CCP test, two studies were undertaken: a cross sectional analysis of patients with established rheumatoid arthritis and non-rheumatoid controls, and a diagnostic and prognostic study on patients from an early arthritis clinic (EAC).

METHODS

RESULTS

The basic characteristics of the ACF test, the anti-CCP test, and the IgM-RF test in the patients with established rheumatoid arthritis and the controls are given in table 1⇓. At a specificity of 99%, the sensitivities of the ACF test and the anti-CCP test were 67.8% and 71.1%, respectively. At this specificity, the IgM-RF test was only 41.8% sensitive. As a control for ACF, IgG depleted fibrinogen was coated at 10 µg/ml and the ELISA was carried out as described above. Subtracting the extinction obtained from the fibrinogen coat from that of citrullinated fibrinogen before calculation did not influence the specificity and sensitivity (data not shown). Cut off values of the tests at 99% specificity were 140 U/ml for ACF, 25 U/ml for anti-CCP, and 45 IU/ml for IgM-RF. To compare all tests at the same specificity, these high cut off values were chosen for further analysis in an early arthritis population.

Table 2⇓ shows the baseline characteristics of that population. The group of patients with rheumatoid arthritis was significantly older (p<0.01), had higher mean ESR and C reactive protein levels (p<0.001), a higher mean DAS (p<0.001), a higher median radiographic damage score (p<0.01), and a worse mean HAQ (p<0.001) than the group of patients with undifferentiated arthritis.

The sensitivity, specificity, and PPV for the diagnosis of rheumatoid arthritis of the tests described are shown in table 3⇓. Sensitivities varied between 44.6% and 57.8%, and specificities between 92.6% and 96.7%. About 30% of the IgM-RF negative early arthritis patients were positive for ACF or anti-CCP or both. The ACF and anti-CCP tests had a very high agreement in early arthritis. ACF and anti-CCP were single positive in 29 of 379 patients (κ = 0.84, data not shown); 16 patients were single positive for anti-CCP (81.3% rheumatoid arthritis) and 13 patients were single positive for ACF (61.5% rheumatoid arthritis).

Complete two year follow up data were available from 296 of the 379 early arthritis patients (78.1%). These patients used a median of one DMARD (range one to five) during the period of follow up; 62% of the patients used methotrexate. The reasons for loss to follow up were: non-compliance (n = 31); discharge from the clinic because of remission (n = 17); moving home (n = 10); death (n = 9); and miscellaneous reasons (n = 16). The group of patients lost to follow up had similar baseline characteristics as the group which completed the follow up, except for the median baseline Sharp/van der Heijde score which was higher in non-completers than in completers (6 v 1, p<0.001). Also, the non-completers were less often positive for IgM-RF (18.1% v 35.1%), anti-CCP (25.6% v 45.1%), or ACF (26.8% v 43.7%) (p<0.01 for all tests) than the completers.

In the univariate analysis, all baseline variables were significantly associated with the diagnosis of rheumatoid arthritis at one year (p<0.05, data not shown). Variables predictive of the diagnosis of rheumatoid arthritis in the logistic regression analysis were anti-CCP, IgM-RF, ACF, DAS28, and VAS pain (table 4⇓). Because of the very high agreement between the ACF test and the anti-CCP test, two alternative predictive models were calculated with the same independent variables, but one without anti-CCP and one without ACF (data not shown). In the model without anti-CCP, the ACF test was the best predictor of diagnosis rheumatoid arthritis (odds ratio (OR) = 10.3; 95% confidence interval (CI), 3.9 to 26.7) and in the model without ACF, the anti-CCP predicted diagnosis rheumatoid arthritis best (OR = 10.6; 95% CI, 4.1 to 27.8).

Baseline variables with a significant association with radiographic progression at two years of follow up were ACF, anti-CCP, IgM-RF, ESR, C reactive protein, DAS28, HAQ, and the Sharp/van der Heijde score (p<0.001, data not shown). Variables predictive of radiographic progression in the logistic regression analysis were anti-CCP, ESR, and the Sharp/van der Heijde score at baseline, with anti-CCP as the best predictor (OR = 14.8) (table 5⇓). The ACF test was removed by this model, owing to the very high agreement between the ACF test and the anti-CCP test. Thus a second model for predicting radiographic progression was calculated without anti-CCP as an independent variable (table 6⇓). In this model, ACF, ESR, and the Sharp/van der Heijde score at baseline were most predictive of radiographic progression, with the ACF test as the best predictor (OR = 12.1).

DISCUSSION

The diagnostic and prognostic value of antibodies directed against citrullinated fibrinogen was compared with that of the second generation anti-CCP test in an early arthritis cohort. For diagnosing rheumatoid arthritis, the ACF test was as sensitive as the second generation anti-CCP test and more sensitive than the IgM-RF test. About 30% of the IgM-RF negative patients with early arthritis were positive for the ACF test and therefore this test will be useful, especially in IgM-RF negative early arthritis patients.

Despite the higher sensitivity of the ACF test and the second generation anti-CCP test compared with the IgM-RF test for diagnosing rheumatoid arthritis, the specificity of the IgM-RF test was slightly higher. This reflects a small percentage of patients diagnosed with undifferentiated arthritis and having autoantibodies to citrullinated proteins. Such patients could eventually develop rheumatoid arthritis, as has been suggested by the high positive predictive value of anti-CCP in a prospective study of patients with early arthritis.²⁴ In an earlier study in the same cohort, Jansen et al found a sensitivity of 42.6% and a specificity of 97.5% for the first generation anti-CCP test.¹⁰ In this early arthritis population, both the ACF test and the second generation anti-CCP test were more sensitive for the diagnosis rheumatoid arthritis. In an early rheumatoid population, Nogueira et al found a sensitivity for antibodies to citrullinated fibrinogen of 64.6% at 98.5% specificity,²⁵ which is in line with the results of the present study. In multivariate analysis, we found that the anti-CCP test was the best predictor of the diagnosis rheumatoid arthritis, followed by the IgM-RF test and the ACF test (odds ratios around 4.5). Because of the high agreement of ACF and anti-CCP, it will not be useful to combine the two tests to predict the diagnosis of rheumatoid arthritis. Therefore, two other models were calculated with the same independent variables, but with only one of the two tests. In these models, baseline ACF and anti-CCP were similarly good predictors of the diagnosis rheumatoid arthritis one year later, with odds ratios of approximately 10.5.

The prognostic value of the ACF test was evaluated with multivariate logistic regression analyses using two year follow up data. ACF was a good predictor of radiographic progression at the two year follow up, nearly as good as the anti-CCP test (OR = 12.1 v 14.8). Compared with previous reports on the prognostic value of citrulline specific autoantibodies, including the first and second generation anti-CCP test, an odds ratio of 12–14 is remarkably high.^8,11,26–,²⁸ The baseline characteristics of the non-completers were similar to those of the completers, except for the Sharp/van der Heijde score and the three antibody tests. The non-completers had a higher median Sharp/van der Heijde score than the completers at baseline, although they were positive less often for IgM-RF, anti-CCP, and ACF. There may have been a coincidental selection of patients with a high Sharp/van der Heijde score at baseline, and a subsequent mild course of the disease, resulting in remission and loss to follow up. As the Sharp/van der Heijde score predicts radiographic progression, the selective loss to follow up may have led to an overestimation of the odds ratios of the anti-CCP test and the ACF test in predicting radiographic progression in our early arthritis population.

The results of this study underline the high disease specificity for antibodies to citrullinated proteins and peptides. However, the present data provide no explanation of how the antibody response develops in rheumatoid arthritis. There was no difference in sensitivity between the anti-CCP test and the ACF test in early arthritis and in established rheumatoid arthritis. In the vast majority of patients both ACF and anti-CCP were found. ACF and anti-CCP were both single positive in 8% of EAC patients. The agreement between the tests is surprising as citrullinated fibrin (fibrinogen is the soluble precursor of fibrin) has been described as a physiological substrate for antibodies recognising citrulline containing epitopes.¹⁶ As has been shown for several other autoimmune disorders (reviewed by Doyle and Mamula²⁹), antibodies might preferentially recognise a modified physiological target—that is, citrullinated fibrin—early in the disease. Later the antibody response could spread towards less restricted epitopes.³⁰ Responses to uncitrullinated fibrinogen could be detected in our study population and were higher in patients than controls. However, they never reached the degree of positivity that was found with citrullinated fibrinogen. It is known that rheumatoid arthritis specific antibodies can be detected several years before the onset of clinical symptoms.^31–,³³ Although the participants from the EAC in Amsterdam had a short disease duration at the time of testing for ACF and anti-CCP, differences between the two responses might be difficult to detect. The ACF response in patient samples taken before clinical signs of the disease could shed light on how the antibody response develops. Alternatively, other citrullinated proteins—for example, vimentin—might trigger the initial immune response in rheumatoid arthritis.⁵ Citrulline containing peptides, derived from the sequence of vimentin, have been shown to be efficiently presented by the rheumatoid arthritis associated HLA-DRB1*0401 MHC class II molecule to T cells in a transgenic mouse model.³⁴ The data point towards an important role of citrulline as an anchor amino acid. Whether the overall sequence might be of less importance has to be elucidated in further studies.

In conclusion, the ACF test is useful for establishing the diagnosis of rheumatoid arthritis and is a good predictor of radiographic progression in early arthritis, comparable to the second generation anti-CCP test. Both tests are especially valuable in IgM-RF negative early arthritis.