Patients and methods

Patients

An inception cohort of patients with very early arthritis was followed prospectively for up to 5 years. The first 200 consecutive patients of the early arthritis clinic were included in the present analysis. Details of this patient population have been described previously.9 Briefly, all patients fulfilled the following criteria: symptom duration <3 months, non-traumatic synovitis of at least one joint, and erythrocyte sedimentation rate >20 mm/h and/or C-reactive protein >0.5 mg/dl. Initial diagnoses were based on the appropriate diagnostic or classification criteria and clinical judgment. When such preliminary diagnosis could not be established the disease was classified as undifferentiated arthritis (UDA).

Patient disposition from preliminary diagnosis at baseline to final diagnosis after 1 year is shown in fig 1. At first visit, 75 patients had a preliminary diagnosis of RA, 67 were classified as having UDA, 38 reactive arthritis and 3 osteoarthritis; among the remaining patients 5 had a connective tissue disease and 12 had other diagnoses. Of the 75 patients with suspected RA, 71 were finally diagnosed as having RA while the remaining 4 patients had other diagnoses assigned (2 UDA, 1 reactive arthritis, 1 connective tissue disease). Of the 67 patients with UDA, 26 developed RA. Among the 58 patients with an initial diagnosis other than RA or UDA, 5 were classified as RA within the first year. This resulted in 102 patients with a final diagnosis of RA. Among these, 42 had fulfilled the American College of Rheumatology (ACR) criteria at baseline and all fulfilled them cumulatively within the first year. Among the 98 patients without RA, 37 were diagnosed as having UDA and 61 had different diagnoses including reactive arthritis (n = 36), osteoarthritis (n = 4), connective tissue diseases (n = 6) and various other diagnoses (n = 15). For further analysis, these 98 patients without RA were considered as a single group. All patients were available for baseline and first follow-up analysis, and 82 patients with RA as well as 35 patients without RA remained for analysis at last follow-up.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Study outline. Preliminary diagnosis at baseline and development until final diagnosis within the first year is shown. For analysis, the resulting 98 patients with a final diagnosis other than rheumatoid arthritis (RA) were considered as a single group (98 without RA) and compared to 102 patients with a final diagnosis of RA. All patients were available for baseline and first follow-up analysis, while at last follow-up 82 patients with RA and 35 patients without RA remained for analysis. dx, diagnosis; UDA, undifferentiated arthritis.

The proportions of patients who were women (73% and 66%, respectively) and mean (SD) symptom durations until first presentation (6.4 (3.6) and 5.5 (3.9) weeks, respectively) were similar in patients with RA and patients without RA.9 Within 6 weeks of presentation, 68% of the patients (and by 1 year, 98%) finally classified as RA were treated with disease-modifying antirheumatic drugs (DMARDs). Initial distribution of DMARDs was similar in patients with erosive and non-erosive disease, but DMARD switches were significantly more frequent in patients who developed erosive disease.

Detection of autoantibodies

IgM-RF was measured by nephelometry, which preferentially detects this subtype; a level of >50 IU/ml was considered positive (RF50) in accordance with previous findings.9 13 ACPA were measured by a second generation anti-cyclic citrullinated peptide (CCP) ELISA (Axis Shields Diagnostics, Dundee, UK) and considered positive above a cut-off value of 5 arbitrary units (AU). Anti-RA33 was assessed by immunoblotting using the recombinant antigen as described.14 ELISAs for determination of IgG-RF and IgA-RF subtypes were from Hycor Biomedical (Garden Grove, California, USA). Antibodies were determined at first visit and additionally at 3 to 12 months from baseline and every 6 months thereafter. Cut-off values were determined for each individual autoantibody using receiver operating characteristic (ROC) curves. The cut-off values were 21 AU/ml for IgA-RF and 30 AU/ml for IgG-RF.

Clinical and radiographic assessments

Clinical and laboratory data were collected every 3 months according to the core set of disease activity measures for RA clinical trials.15 In patients with RA, initial and yearly follow-up radiographs of hands and feet, blinded for group and sequence, were assessed by a rheumatologist (KPM) and a radiologist (MU). In order to account for any fortuitous findings of singular erosions, only patients with at least two erosions in at least two different joints were classified as erosive. Erosions were defined essentially as described previously.9 In addition, radiographs were scored according to the Larsen method on 42 joints in the hands and feet: firstly, films were scored using the traditional Larsen scoring with 0 (normal) to 5 (mutilating changes) for each individual joint16; secondly, grade 1 scores were abandoned and the total scores recalculated.17 Only the latter modified Larsen scores are shown in our analyses.

Statistical analysis

The SPSS statistical software package was used for statistical analysis (SPSS, Chicago, Illinois, USA). The reported p values are the results of two-sided tests; a p value ⩽0.05 was considered significant. The presence of erosions in patients with RA with at least 1 year of follow-up and complete sets of x rays of hands and feet was used as prognostic outcome variable (erosive vs non-erosive disease). In addition, modified Larsen scores at baseline and the radiographic evaluation closest to last follow-up were analysed. Binary variables and correlation between individual autoantibodies were analysed using the χ² test.

Results

Follow-up visits

Clinical and serological data were collected from the 200 original patients for up to 5 years. All patients had at least 1 follow-up visit at 3 months to 1 year (mean 5 months for patients with RA and patients without RA) and 117 patients (82 with RA and 35 without RA) were seen at least once after the first year: the last visit in the 5-year observation period of each patient was chosen for analysis as “last follow-up visit” (mean 35 months for the patients with RA, 29 months for the patients without RA). Further details are shown in table 1.

Autoantibody profiles at baseline and during follow-up

Baseline data for all autoantibodies are shown in table 2 including the previously published data on RF50 (ie, IgM-RF>50 IU/ml as determined by nephelometry), ACPA and anti-RA33.9 IgA-RF and IgG-RF were less prevalent than RF50 being detectable in only 29% and 14%, respectively, of patients with RA as compared to a sensitivity of 45% of RF50. Nevertheless, since IgA-RF and IgG-RF were rarely detected in patients without RA, their positive predictive values (PPV) were rather high (90% and 93%, respectively) and thus similar to the PPV of RF50 (92%).

At first follow-up visit autoantibody profiles were almost identical with the baseline profiles (data not shown). During further follow-up prevalence of RF (all subtypes) and anti-RA33 slightly decreased, whereas prevalence of ACPA increased. Although, presumably due to the relatively small number of patients, neither of these changes reached the level of statistical significance, it was interesting to note that the proportion of patients with RA who were positive for RF50 and/or ACPA remained constant throughout the observation period (fig 2): thus, at baseline 59% (60 of 102 patients) and at last follow-up 57% (47 of 82 patients) were positive for either antibody. In detail, of the 28 patients initially positive for RF50 and ACPA, 16 stayed positive for the 2 antibodies throughout the study period, 5 became RF50 negative and 7 were lost to follow-up. Of the 18 patients initially RF50 positive/ACPA negative, only 6 showed this constellation throughout the study period while 4 patients additionally developed ACPA, 4 became RF50 negative and 4 were lost to follow-up; however, 2 of the patients who were RF50 negative were ACPA positive at last visit. Of the 14 patients initially RF50 negative/ACPA positive, 7 stayed in this category throughout, 3 became RF50 positive and 4 were lost to follow-up. Four patients initially negative for both antibodies became positive for either one or both antibodies until last follow-up: one patient for RF50, two patients for ACPA and one for both antibodies. This resulted in 47 patients with RA who at last visit were RF50 and/or ACPA positive.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

High-titred IgM-rheumatoid factor (RF50) and anti-cyclic citrullinated peptide antibodies (ACPA) profiles of patients with rheumatoid arthritis (RA) at baseline (A) and last follow-up (B). The percentage of patients positive for RF50 (ie, IgM-RF>50 IU/ml) and/or ACPA remained constant over the study period: However, the percentage of patients with RA positive for RF50 but negative for ACPA decreased from 18% at baseline to 9% at last follow-up, whereas the prevalence of patients who were ACPA positive/RF50 negative increased from 13% to 19%.

Of note, at baseline 18% of the patients with RA were RF50 positive/ACPA negative, and 13% were ACPA positive/RF50 negative (fig 2A) while at last follow-up these proportions changed to 9% and 19%, respectively (fig 2B). The number of patients who were RF50 positive therefore declined under therapy, while the number of patients who were ACPA positive increased over time.

For anti-RA33, in 13 of the 28 patients initially positive for anti-RA33 antibodies no other antibody was present; of these patients 7 were still positive at last follow-up, 1 patient had additionally become ACPA positive, 2 patients had become completely negative and 3 were lost to follow-up. Four patients who had been negative for all antibodies at baseline were solely positive for anti-RA33 at last follow-up. Thus, although the total percentage of patients who were positive for anti-RA33 decreased (from 28% to 23%), the percentage of patients showing singular positivity for anti-RA33 remained constant (13%). Remarkably, since only 1 of the 35 patients without RA available for last follow-up testing was anti-RA33 positive, specificity of anti-RA33 at last follow-up was of the same order of magnitude (97%) as the specificities of the other antibodies investigated.

Autoantibody titres in patients with RA

Titres of IgM-RF decreased over time from 284 (437) U/ml at baseline to 243 (233) U/ml at last follow-up (p = not significant). Also, titres of patients positive for IgA-RF and IgG-RF did not change significantly over time, nor did ACPA titres, although a moderate increase from 51 (33) AU/ml to 61 (34) AU/ml was observed.

Correlation of RF subtypes and anti-RA33 with RF50 and ACPA

IgA-RF was significantly correlated with RF50 (r = 0.422; p<0.001) and ACPA (r = 0.383, p<0.001). Among patients with RA who were positive for IgA-RF, 69% were also RF50 positive and 73% were ACPA positive while only 8% were negative for both antibodies; this high overlap was maintained throughout the whole observation period. IgG-RF was detectable in only 14% of patients with RA and more than two-thirds were also positive for RF50 and/or ACPA. Furthermore, IgG-RF was correlated with IgA-RF (r = 0.366, p = 0.001) with 70% of the patients who were IgG-RF positive also being IgA-RF positive. In fact, only a single patient proved solely positive for IgG-RF. In contrast, anti-RA33, which showed a similar prevalence as IgA-RF, was not associated with RF50 or ACPA.

Prognostic value of autoantibodies at baseline and during follow-up

Complete x ray sets performed once yearly were available for 66 patients with RA. Table 3 shows the antibody profiles of patients with erosive and non-erosive disease, and the corresponding modified Larsen scores. RF50 was strongly linked to erosive disease showing high PPV at baseline and at last follow-up (78% and 82%, p = 0.002), even though the total number of patients who were positive for RF50 decreased. IgA-RF and IgG-RF both showed a PPV of 75% at baseline, which, however, did not reach the level of significance; nevertheless, at last follow-up the association of IgA-RF with erosive disease was significant (PPV = 80%, p = 0.030). ACPA, which had proven to be highly predictive for the development of erosions at baseline9 continued to be strongly associated with erosive disease showing a PPV of 82% (p<0.001) at last follow-up. Thus, while the number of patients with RA who were ACPA positive increased and the number of patients who were RF50 positive decreased, at last follow-up the proportion of erosive patients was identical (82%) in both groups.

Larsen scores of patients positive for RF50 and/or ACPA were significantly higher at last follow-up than in those patients negative for both antibodies (14.5 (15.8) vs 3.2 (6.7), p<0.001), whereas no significant differences were seen at baseline. Finally, as reported previously,9 anti-RA33 was not associated with erosive disease, neither at baseline nor at last follow-up resulting in a negative predictive value (NPV) of 71% (p = not significant) at baseline and of 78% (p = 0.036) at last follow-up. Thus, only two patients solely positive for anti-RA33 at baseline developed erosive disease; of note, none of the four patients who became anti-RA33 positive during follow-up developed erosive disease (table 3). Therefore, patients solely positive for anti-RA33 showed significantly lower Larsen scores than patients who were ACPA and/or RF50 positive (2.4 (7.3), p = 0.006).

Discussion

It has become evident that the early detection of autoantibodies such as ACPA or RF is of considerable diagnostic and prognostic value in patients with early arthritis. However, less is known about the usefulness of these autoantibodies at later time points when disease has progressed and diagnosis is established. To address this issue we have investigated changes of autoantibody profiles in follow-up sera of patients with very early arthritis over a period of up to 5 years from baseline. Since only patients with inflammatory joint disease of less than 3 months symptom duration were included we were able to investigate the autoantibody profiles within only a few weeks from onset of synovitis in patients with RA and in a control group, meeting the same inclusion criteria, but developing other forms of inflammatory arthritis. The data obtained are of particular interest because the first follow-up at a mean of 5 months from baseline and the last follow-up at less than 3 years mean are both relatively early in the disease course,18 19 20 and we were thus able to analyse the development of autoantibodies and changes in antibody profiles in the early stages of disease.

Apart from RF50 (ie, IgM-RF>50 IU/ml), ACPA and anti-RA33, we additionally studied IgA-RF and IgG-RF subclasses in order to evaluate their value for diagnosis and prognosis of early RA. At baseline IgA-RF and IgG-RF both showed similar specificity as RF50, while sensitivity was considerably lower with 29% and 14%, respectively, as compared to 45% of RF50. Moreover, the majority of patients positive for IgA-RF and/or IgG-RF were also positive for RF50 and/or ACPA and highly significant correlations of IgA-RF with RF50 and ACPA were seen. With respect to their prognostic value, IgA-RF was significantly correlated with the development of erosive disease at last follow-up visit, in contrast to IgG-RF, which turned out to be of neither diagnostic nor prognostic value. Our data are in good agreement with those of Lindquvist et al21 who also showed that ACPA had a high correlation with IgA-RF. However, ACPA predicted a more destructive form of RA after 5 and 10 years, whereas IgA-RF predicted only 5-year outcome. While Berglin et al reported that baseline values of ACPA and IgA-RF significantly predicted radiological progression after 2 years, Agrawal et al did not find a significant association of IgA-RF with erosions or deformities.22 23 In line with these findings Nikolaisen et al reported that combining assays for ACPA and/or IgM-RF with IgA-RF did not improve RA diagnosis.24 Taking all these observations into account, IgA-RF does not seem to provide additional diagnostic or prognostic value in patients with early arthritis and testing for this antibody in addition to IgM-RF and ACPA does not appear to be useful. In this context it should also be mentioned that in two other early arthritis studies high titre IgM-RF was found to be quite specific for RA.11 12 This may be explained by the fact, that the majority of patients without RA in early arthritis cohorts have diagnoses such as osteoarthritis, reactive arthritis, psoriatic arthritis or UDA which usually do not show high RF levels.

During follow-up the percentage of patients who were RF50 positive declined from 45% to 38%, which was most likely due to early DMARD therapy which is known to affect RF levels more than ACPA levels.25 26 Nevertheless, specificity of RF50 for erosive disease did not decrease confirming the high prognostic value of this antibody. In contrast, the number of patients with RA who were ACPA positive increased over time and a small but not significant increase in antibody levels was also seen.

Concerning anti-RA33, the data obtained during follow-up confirmed the previously made observation that in the absence of RF50 and ACPA this antibody is associated with a relatively mild non-erosive disease course.9 Thus, only 2 of 10 patients who were solely anti-RA33 positive at baseline developed erosive disease and, interestingly, all 4 patients who seroconverted to anti-RA33 during follow-up had non-erosive disease.

The question has been raised of whether it is time to abandon RF testing.27 We think that our data confirm our previous findings that there is a subpopulation of patients positive for RF50 and/or ACPA who have a significantly worse prognosis than patients negative for both antibodies. Interestingly, the percentage of patients who were RF50 and/or ACPA positive remained constant over time (59% at baseline vs 57% at last follow-up), although the antibody profiles changed: at baseline 30% of patients in this subgroup were solely RF50 positive as compared to only 15% at last follow-up. Thus, almost one-third of patients at high risk for erosive disease would have escaped detection if tested only for ACPA. Concerning the latter antibody, the proportion of patients solely ACPA positive increased from 23% at baseline to 34% at last follow-up. These data show that RF50 and ACPA are not interchangeable and one antibody cannot be substituted with the other. It is therefore crucial to use both antibodies as diagnostic and prognostic tools in the differential diagnosis of very early synovitis and also during follow-up at least in patients who are negative for these antibodies at baseline.