Patients and methods

All 558 adult patients who met ACR criteria for RA and had been regularly followed up at our institution since the first year of their disease (incipient RA) were included. Patients with RA of less than five years’ duration were excluded. Follow up radiographs and an immunogenetic study were obtained in most of these RA patients (469 of 558 (84%)). Radiographs of the hands, wrists, feet, and joints with clinical symptoms were obtained in all patients after a mean disease duration of eight years. Two independent observers who were blinded to the status of the patients scored each radiograph using Larsen’s method. The interobserver correlation coefficient determined after the observers received specific training in Larsen score determination was 0.94. Among the RA patients, two groups of white patients were differentiated based on severity of radiographic joint lesions, taking care to match the groups on sex distribution and disease duration (mean (SD) 8.95 (4.77) years in group I, 9.66 (7.20) years in group II). Group I patients (n = 43; 33 women and 10 men) had severe radiographic damage with a Larsen score ⩾2 for each of both wrists, both hands, both feet, and at least one other joint. Group II patients (n = 43; 37 women and six men) had a Larsen score < 2 for each joint studied. At symptom onset, mean age of the 86 patients in groups I and II combined was 43.29 years (range 18–69). Mean disease duration before the first visit to our clinic was 7.8 months (range 2–12).

In addition to a physical evaluation and to the usual laboratory tests, including ESR and CRP, the following immunological markers were looked for: RF by laser nephelometry (positive if ⩾ 40 IU/ml), IgG AKA by indirect immunofluorescence on cryostat sections of rat oesophagus (positive if ⩾1:20), IgGAM APF by indirect immunofluorescence on human buccal epithelial cells (positive if ⩾1:40), and IgG anti-RA 33 antibodies by western blotting with HeLa cell nuclear extracts (serum dilution 1:20) as described elsewhere.5 Antinuclear antibody determination was performed routinely using HEp-2 cells as the substrate.

A frozen serum sample collected during the first year of the disease was used to determine the serological markers for RA. Non-routine tests (AKA, APF, anti-RA33ab) were performed in duplicate. HLA-DR genetic oligotyping was performed using a non-radioactive reverse dot blot technique as described elsewhere.8

Statistical analysis was performed using BMDP statistical software. Fisher’s exact test or the χ² test and the non-parametric Kruskal-Wallis test were used for between group comparisons of qualitative and quantitative data, respectively. A stepwise multiple logistic regression model was constructed to look for relevant independent prognostic variables. The prognostic variables included in the multivariate model were selected based on the results of a univariate analysis. The significance level was 0.05.

Results

Rates of occurrence (sensitivity) of RF, APF, AKA, anti-RA33 antibodies, and ANA in RA patients with (group I) and without (group II) severe articular damage are indicated in table 1, as well as the specificity and the positive predictive value of these five autoantibodies for severe radiographic damage. For purposes of comparison, the sensitivity, specificity, and positive predictive value of a high ESR (> 28 mm 1st h) or a high CRP level (> 15 mg/l) are also indicated. Of the five autoantibodies studied, AKA had the highest positive predictive value (0.68) with the highest OR (3.95; 95% CI, 1.59, 9.76) for severe destructive RA.

Mean serum autoantibody titres were higher in group I than in group II, and the differences were significant for RF (p < 0.01) APF (p < 0.02), and AKA (p < 0.001) (fig 1). Mean (SD) initial ESR and CRP levels were significantly higher in group I than in group II: 48 (31) versus 21 (17) (p < 0.0001) for ESR and 51 (45) versus 12 (13) (p< 0.0001) for CRP.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Values of RF, APF, and AKA in patients with destructive (group 1) and non-destructive (group 2) RA. The horizontal line within boxes represents the median, boxes show the interquartile range (± 25% of the median), vertical bars show the 95% confidence intervals (values above or below these levels were plotted separately). x: p< 0.01; y: p< 0.02; z: p< 0.001.

To determine which autoantibodies had the best predictive value for severe radiographic damage, all variables were entered into a dichotomous stepwise logistic regression model. Table 2 shows the results of this analysis.

Subcutaneous nodules, HLA DRB1*04 or DRB1*01 susceptibility alleles, AKA, and CRP were the four best variables for predicting severe articular damage.

A subsequent analysis excluding clinical parameters from the stepwise logistic regression model showed that AKA was the autoantibody with the best predictive value for severe articular damage (OR, 3.23; 95%CI of OR, 1.16, 8.96).

Discussion

Recent research has concentrated on identifying clinical and laboratory abnormalities present in early RA, with the goal of predicting the long term outcome of the disease.6-8 As radiological articular damage can occur within two years of the first manifestations of RA,6 there is an urgent need for identifying independent predictive factors present at disease onset. Such factors would be of use for evaluating the risk-benefit ratio of potentially toxic second line drugs.

In addition to clinical, epidemiological, socioeconomic, and health assessment characteristics,10 several studies have found a strong correlation between a persistently high ESR or CRP level and less favourable disease outcomes.7 ,8 ,11

DR4 and DR1 alleles, which share an epitope on the third hypervariable region of the β chain, have also been found to be associated with a poor prognosis.8 ,12 ,13

Among autoantibodies, IgM (or IgA) RF has been the most extensively studied. Presence of IgM RF was strongly predictive of an unfavourable outcome.11 Very few published studies have evaluated the predictive value of the other autoantibodies, and most looked at only one antibody.

Westgeest et al 14 reported that APF indicated more severe disease in seronegative RA and also identified the most severe cases among RF positive patients. In a study by Paimela et al,15 RA patients who were initially positive for AKA had more active disease as assessed based on clinical, laboratory tests, and radiological variables, as compared with AKA negative patients. However, none of these studies used multiple regression analysis to determine which autoantibodies were the best independent predictive variables, and which combinations of autoantibodies predicted the outcome in early RA. In an attempt to obtain more sharply contrasted findings regarding the prognostic value of autoantibodies, we compared two groups of RA patients with and without severe radiographic damage. Although our study was not prospective, all patients had serum drawn within one year of RA onset and were followed up at our institution starting during the first year of their disease. All had initial radiographs with no articular damage, and in all a second series of radiographs was obtained after at least five years’ follow up. In our model, presence of AKA (60.5% of group I patients with severe radiographic damage versus 27.9 % of group II patients with a benign articular outcome) was associated with a relative risk of 3.1. These data are very similar to those obtained with inflammatory markers such as ESR or CRP, which have been established as important prognostic indices for severe articular damage in retrospective and prospective studies.7 ,8 ,11 Relative risks were 0.86 for RF, 0.66 for APF, 0.36 for ANA, and 0.76 for anti-RA 33 antibodies. Apparent discrepancies between our data and those from earlier studies may be ascribable to the fact that only patients with ‘definite’ RA after a mean follow up of eight years were included in our study; as a result, as many as 50% of patients in the favourable outcome group (group II) were positive for RF, contrasting with smaller proportions in most prospective studies in populations with short-term follow up periods.

The rate of occurrence of APF may seem low in our study compared with many others.1 ,16 ,17 However, we used a cut off titre of 1:40 because in an earlier study5 we found that a cut off of 1:10 had relatively poor specificity for RA. In several countries, the frequency of AKA18 or anti-RA 33 antibodies1 is considerably lower than in most of Europe.4 ,5

As the target of APF and AKA seems to be immunologically related to the same acidic form of profilaggrin,3 determination of one or the other of these two antibodies is sufficient for diagnostic or prognostic purposes in early RA. These autoantibodies have been found in early RA,1 ,2 and even before the onset of clinical symptoms.1 Consequently, we recommend adding AKA to the list of clinical, physical function, laboratory (ESR, CRP), and immunogenetic parameters (HLA DRB1*04/DRB1*01 genotype) currently used for early prediction of the outcome of RA.

Even when they are statistically significant, however, differences between groups of patients are not necessarily helpful for making therapeutic decisions in individual patients. In the near future, development of a composite index including three or more independent variables may improve the ability of clinicians to predict the outcome of early RA.