Rheumatoid arthritis (RA) is a chronic inflammatory disease characterised by progressive erosions and cartilage destruction. Radiographic progression is considered a key outcome variable, as it is thought to reflect the cumulative effect of inflammation on bone and cartilage. New effective treatments with particularly good effect on joint destruction have become available over the last years.1^–3 However, the disease course varies considerably between patients, and there is a need for reliable predictors of radiographic progression to select patients for early aggressive treatment. In addition, it is important not to expose patients with a good prognosis to expensive treatment that has potentially serious side effects.

Several studies have addressed prognostic factors in RA, but the results are conflicting, and the current lack of internationally accepted prognostic criteria for RA makes further investigations on this issue warranted.4^–6

A clinically useful prognostic marker must be readily available and easy to measure, as are both acute phase reactants and autoantibodies. Erythrocyte sedimentation rate (ESR), C-reactive protein (CRP) and rheumatoid factor (RF) have been found to predict radiographic progression in RA.4^–6 Recently, antibodies to cyclic citrullinated peptide (anti-CCP) have consolidated its role as an important diagnostic marker for RA.7 The more specific second-generation anti-CCP assays (anti-CCP 2)8 9 have also been suggested to predict radiographic progression.10^–14 It has not been clarified whether the level of anti-CCP adds prognostic information.

We have previously shown that radiographic progression is associated to loss of physical function.15 In the current study we used the same cohort to examine the combined predictive role of anti-CCP, IgM RF, IgA RF, ESR and CRP with regard to long-term radiographic progression, and to examine whether the level of anti-CCP adds prognostic information.

METHODS

Patients

A cohort of 238 (175 females, 63 men) patients with RA were enrolled consecutively and followed longitudinally for 10 years. Of these, 125 had x-rays of the hands available for scoring at baseline and 10 years, and were included in the present study (fig 1). All patients had RA16 with a short disease duration (maximum 4 years, mean 2.3 years) at baseline. The procedures for inclusion have been described in detail elsewhere.17 Patient disposition is depicted in fig 1 and described in detail elsewhere.15 The cohort was part of the European Research on Incapacitating Disease and Social Support (EURIDISS) project, which started in 1992.17 The regional ethics committee evaluated the study, and all enrolled patients gave informed consent. Patients were assessed with standard clinical examinations, collection of serum (stored at −70°C) and x-rays of hands at baseline, after 5 and 10 years. Physical function was measured by the Health Assessment Questionnaire (HAQ). The patients were treated according to clinical judgement by their rheumatologist. At baseline (10 years follow-up) 56% (48%) were using disease-modifying anti-rheumatic drugs (DMARDs), 52% (44%) non-steroidal anti-inflammatory drugs (NSAIDs), 25% (36%) prednisolone and 0% (12%) tumour necrosis factor-blocking agents; 14.1% did not use any DMARDs or tumour necrosis factor blocking agents during the study.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Patient disposition and distribution of available hand radiographs for scoring over the 10-year follow-up.

Radiographic evaluation

Baseline anteroposterior radiographs of the hands were available for scoring in 163 patients, 150 patients had available x-rays after 5 years, 147 after 10 years, and the 125 patients with x-rays available both at baseline and 10 years were included in the present analyses (fig 1). The hand radiographs were scored by one experienced reader by the van der Heijde modified Sharp score (SHS).18 The films were scored with a known time order, and the reader was blinded for clinical data and study purposes. The potential maximum total score for both hands is 280 (16 areas scored for erosions (score 0–5) and 15 areas for joint space narrowing (score 0–4) in each hand). A change in the SHS of hands from baseline to 10 years of more than 10 units was regarded as radiographic progression. The reliable change according to the method based on interobserver reliability is about 4–5 units (hands and feet) between two radiographs, and a smaller change is already considered meaningful by clinicians.19

Laboratory analyses

Anti-CCP was analysed by a second-generation ELISA (INOVA Diagnostics, San Diego, CA, USA) and considered positive above a cut-off value of 25 U/ml. Antibody concentrations were reported from 1 to 251 U/ml. For statistical purposes the value of 251 was assigned to all measurements >251. IgM and IgA RFs were measured by an “in house” ELISA described elsewhere.20 The range of measurement was 2–300 U/ml, with cut-off set at 25 U/ml. CRP was measured by phyCardioPhase hsCRP, nefelometri (Dade Behring Inc., Newark, NJ, USA) with the lowest detectable limit at 0.15 mg/l. ESR was measured by the Westergren method, with the range from 0 to 140 mm/h.

Statistical analyses

Statistical analyses were performed using SPSS 12 (SPSS Inc., Chicago, IL, USA). The radiographic progression (dependent variable) was calculated as the difference in SHS of hands between baseline and 10-year follow-up and dichotomised into the presence or absence of radiographic progression. The serum markers were dichotomised, with high levels according to the following cut-offs: anti-CCP >25 U/l, IgM RF and IgA RF >25 U/l, CRP >10 mg/l, ESR >20 mm/h.

Comparisons between groups were performed by Mann–Whitney test or χ² tests and correlations were examined by the Spearman rank correlation test. Multiple logistic regression analyses were performed to identify independent predictors and to calculate the probability of having radiographic progression over 10 years for individual patients with different combinations of risk factors. The independent variables were selected from univariate analyses if p<0.15. To examine the effect of anti-CCP level, the association between anti-CCP and radiographic progression was checked for linearity in the logistic regression model. Anti-CCP was divided in three categories: <25 U/ml (negative), 25–200 U/ml (low to moderate level) and >200 U/ml (high level), and this new category variable was entered into the model. Radiographic outcome in the three anti-CCP level categories was depicted as a cumulative probability plot.21

All tests were two-sided and conducted at the 0.05 significance level.

RESULTS

Demographic, clinical and biological characteristics of the cohort

The baseline characteristics of the whole cohort (n = 238), the 125 patients with complete x-ray sets and separate baseline values for patients with and without radiographic progression at 10 years are given in table 1. There were no significant differences between the patients with x-rays available (n = 125) and the whole cohort (n = 238).

Radiographic outcome over 10 years

At baseline 55.2% (n = 69) of the patients had erosive disease, and after 10 years 84.4% (n = 105). Mean (SD) SHS (hands) at baseline was 6.8 (11.8), at 10 years this was 36.2 (36.6). Radiographic progression developed in 59.2% (n = 74) of patients (>1 U/year). Mean (SD) yearly progression rate (SHS/disease duration) at baseline was 2.9 (4.6) and 2.8 (2.9) during follow up.

Associations between serological markers and radiographic progression

The patients who developed radiographic progression during 10-year follow-up had a significantly (p<0.05) higher proportion of anti-CCP, IgM RF and IgA RF positive tests and a significantly higher baseline CRP, ESR and anti-CCP level (table 1). Change in SHS over 10 years was moderately correlated to the baseline levels of anti-CCP (rs = 0.49, p<0.01), IgM RF (rs = 0.49, p<0.01), IgA RF (rs = 0.44, p<0.01), ESR (rs = 0.35, p<0.01) and CRP (rs = 0.31, p<0.01).

All the serum markers were significantly associated to radiographic progression in the univariate analyses: high CRP (OR 6.1; 95% CI 1.2 to 6.7) , anti-CCP (OR 5.7; 95% CI 2.6 to 12.5), IgM RF (OR 5.3; 95% CI 2.4 to 12.0), IgA RF (OR 4.0; 95% CI 1.7 to 9.3), high ESR (OR 3.9; 95% CI 1.8 to 8.5) as were female sex (OR 2.6; 95% CI 1.1 to 5.9) and radiographic progression rate at baseline (OR 1.39; 95% CI 1.2 to 1.7), but not age and baseline HAQ level.

Independent predictors of radiographic progression from multivariate analyses

The baseline variables with a p<0.15 were included in a multiple logistic regression model in which radiographic progression was the dependent variable. As shown in table 2 (left column), a positive test for anti-CCP (OR 4.00; 95% CI 1.60 to 10.00) appeared to be the strongest independent predictor of radiographic progression. Female gender (OR 3.32; 95% CI 1.34 to 7.57), high ESR (OR 3.18; 95% CI 1.19 to 7.57), and a positive test for IgM RF (OR 3.07; 95% CI 1.18 to 7.94) were also independent predictors of radiographic progression. The significant associations to IgA RF and high CRP from the univariate analyses were not maintained in the multivariate model.

The probability of developing radiographic progression was calculated as follows: log (p/1– p)  =  −2.28 + 1.38 anti-CCP + 1.20 female sex + 1.16 high ESR + 1.12 IgM RF; where p is the probability and each variable is entered as a categorical variable (0 or 1) modulated by the importance of each variable. From this algorithm, the probability of radiographic progression is 92.5% in females with a positive anti-CCP and IgM RF test and a high ESR, compared with 9.3% in males with a negative anti-CCP and IgM RF test and a low ESR. These results are depicted in fig 2. The ability of the model to correctly identify patients that have a probability of progression above 50% (the sensitivity) is 89.2% and the ability to correctly identify patients with a probability of progression less than 50% (the specificity) is 51%. The total accuracy of the model was 73.6%.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 The probability of radiographic progression (change in van der Heijde modified Sharp Score (SHS) hands >10 U/10 years) according to different combinations of the independent predictors from the logistic regression model. ESR, erythrocyte sedimentation rate; RF, rheumatoid factor; CCP, cyclic citrullinated peptide.

Because of the different disease duration (0–4 years) at baseline, baseline radiographic damage was defined as a progression rate (baseline SHS divided by disease duration at baseline). The adjustment for progression rate at baseline (table 2) only slightly affected the OR of the model, but the specificity increased to 76.5% with 81.1% sensitivity.

Effect of anti-cyclic citrullinated peptide level

Anti-CCP was checked for linearity in the logistic regression model with radiographic progression as the dependent variable. The variable showed linearity, supporting a relationship between anti-CCP level and radiographic progression. Accordingly, the accuracy of the multivariate model improved from 73.6% to 77.6% when anti-CCP was entered into the multivariate analyses as a continuous variable (table 2, right column). The OR for anti-CCP in this model was 1.008 (β 0.008), for example, an increase of 1 U/ml anti-CCP will increase the odds of progression by 0.8% and, for example, an increase in 50 U/ml gives a 49% increase in the odds of progression.

The patients were divided into three groups according to anti-CCP levels: <25 U/ml (negative, n = 51), 25–200 U/ml (low to moderate level, n = 41) and >200 U/ml (high level, n = 33). Mean progression (SD) differed significantly (p<0.05) between the anti-CCP-negative group (13.1 (20.3)), the low to moderate level group (31.4 (29.1)) and the high-level group (46.3 (28.6)). Compared with the anti-CCP-negative patients, patients with low to moderate levels (OR 3.5; 95% CI 1.5 to 8.4) and anti-CCP-positive patients with high levels (OR 13.3; 95% CI 4.0 to 43.8) were more likely to develop radiographic progression (table 3). The trend remained when adjusting for other known predictors of radiographic progression, though the OR for the low to moderate group was only borderline significant (table 3). The accuracy of this model was 78.4%. The patients with high levels of anti-CCP were more likely to progress, also when directly compared with patients with low to moderate levels of anti-CCP (OR 4.8; 95% CI 1.2 to 19.2) (table 3). When entering anti-CCP categorised into high versus low (cut-off 200 U/ml) the accuracy and sensitivity of the model improved and the specificity worsened (table 3 right column).

The change in SHS over 10 years in the three anti-CCP level groups is depicted in a probability plot (fig 3). Probability plots show data from each individual patient ranked from the lowest to the highest observed change score within each subgroup. On the x-axis an arbitrary percentage of patients can be selected and the corresponding change score on the y-axis can be recorded. This graphical approach enhances insight into the coherence of the data and offers the possibility to determine several cut-offs by the reader.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3 Change in van der Heijde modified Sharp Score (SHS) of hands from baseline to 10 years according to the level of anti-CCP depicted by a cumulative probability plot.

DISCUSSION

Joint damage accounts for a considerable amount of the disability in RA, both in established and in the earlier phases of disease.15 22 An optimal treatment strategy should include considerations on the presence or absence of predictors of joint damage, as prevention of such damage is a major treatment goal.

The presence of anti-CCP has been suggested to predict radiographic progression.10^–14 None of these studies has explored the effect of the anti-CCP level. Quinn et al and Kroot et al, however, did not find anti-CCP to be an independent predictor of radiographic progression.23 24 Conflicting results in longitudinal studies may be due to different study design. One negative study23 also used the less specific anti-CCP 1 test.8 Numerous studies have investigated the predictive ability of RF, and the issue has also been reviewed.5 6 In studies incorporating both anti-CCP and RF.10–14 23 24 the independent effect of RF over anti-CCP has been confirmed in some studies,10 12 13 23 whereas others have demonstrated contradictory results.11 14 24 It is yet to be clarified how different antibody profiles should influence treatment decisions in individual patients, but the recommendations for management of early arthritis included prognostic factors and the development of prediction algorithms on the research agenda.25

This study shows that anti-CCP, IgM RF, ESR and female gender all are independent predictors of radiographic progression and could be combined into an algorithm for better prediction. Anti-CCP is the strongest contributor to the overall prediction model. The algorithm predicted radiographic progression on the individual level with good accuracy. Prediction of mild disease (specificity), however, is more difficult than prediction of progressive disease (sensitivity). By also including baseline x-ray scores in the analyses, the accuracy improved, but as this information is not easily available to most clinicians, we preferred the most feasible model for use in clinical practice, for example, with only the serological markers.

It is a matter of ongoing research whether anti-CCP-positive and -negative RA are two different disease entities.26 This study confirms findings from other multiparameter longitudinal studies that have found anti-CCP to be the strongest independent predictor of radiographic progression.10^–14 This finding is of clinical importance, as anti-CCP has some of the characteristics of an ideal prognostic marker: it appears early in the disease course10 and is only moderately affected by treatment.27

Nell et al found that when RF-levels were dichotomised at a higher cut point (40 U/l) the additional prognostic effect of anti-CCP disappeared.28 In our study, we tested also higher cut-offs for IgM RF without improving the model and without weakening the prognostic effect of anti-CCP (data not shown). The additional benefit of anti-CCP must, however, be weighted against the cost of running an additional test.

Some previous studies have suggested that IgA RF predicts radiographic progression better than IgM RF.10 12 29 Our study did not support these findings. There are conflicting results regarding the predictive role of measures of disease activity.5 6 This study shows that ESR but not CRP, was an independent predictor of radiographic progression. However, we only modelled the influence of baseline variables, and we cannot exclude the possibility that a longitudinal analysis would have revealed an association between high CRP levels and radiographic progression.

This study indicates that female sex is an independent predictor of joint damage, and support the recent findings by Forslind et al.30 In agreement with other studies5 6 baseline physical function (HAQ) did not predict radiographic progression.

Importantly, this is the first study to show that the level of anti-CCP plays an independent role. In this study, patients with high levels of anti-CCP are 10 times more likely than anti-CCP-negative patients to develop radiographic progression, and about five times more likely than patients with low–moderate levels. This novel finding is supported by the findings of Kuhn and colleagues, which showed that anti-CCP enhanced tissue injury in an animal model of RA.31

The value 200 U/ml as a cut-off for high-level anti-CCP was chosen based on what is considered high in clinical practice. The categorisation was done to make the association between anti-CCP level and radiographic progression easier to interpret.

Our study has both its strengths and limitations. Radiographs of the feet were not available at baseline and could not be used in the present analyses. As the range of disease duration at the start of the study is quite large we adjusted the baseline x-ray score for disease duration. However, radiographic progression is not entirely linear in the early disease course. Missing cases are always a challenge in long-term observational studies. In this study the attrition rate among living patients was 27%.15 We cannot rule out the possibility that loss of patients can have affected the results, although the baseline characteristics of the patients with complete x-ray sets (n = 125) were similar as for the whole cohort (n = 238) (table 1).

Patients were treated according to clinical judgement before inclusion and during the study. Fifty-two per cent (n = 124) of the patients received DMARD at baseline. It can be questioned whether treatment might be a confounding factor in the relationship between the serological markers and radiographic progression. The results, however, showed that the patients with high antibody levels received more potential anti-erosive therapy and still had greater increase in joint damage. Further, we also adjusted for DMARD use at baseline without any alteration in the results of the regression analyses (data not shown).

The enrolment to this study started in 1992. This cohort was at that time considered “early RA”, which no longer holds true according to current understanding of early disease.13 Thus, the algorithm and the finding of the association between anti-CCP level and radiographic progression must be tested in cohorts of earlier RA or undifferentiated arthritis to prove clinical relevance.

An obvious strength of this study is the long observational period of 10 years. The results were similar when 5 years data were analysed, or when a linear regression approach was chosen (data not shown), supporting the robustness of the results.

In conclusion, we have shown that anti-CCP, ESR, IgM RF and female gender are independent predictors of 10-year radiographic progression, and that anti-CCP is the strongest contributor to the prediction model. These individual factors can be combined into an algorithm with good sensitivity to predict radiographic progression. However, prediction of mild disease (specificity) is still far from perfect and more research on this issue is warranted to protect patients with a predicted good prognosis from therapeutic regimens that have potentially serious side effects and high cost. The association of anti-CCP level to radiographic progression is a novel finding, which indicates that the anti-CCP level may add important prognostic information when evaluating patients with RA.