Rheumatoid arthritis (RA) is a potentially severe but heterogeneous disease. It can vary from mild to severe and in some cases can lead to severe joint damage and functional disability. Predicting RA outcomes is fundamental for optimal clinical management. Predictive factors of long-term outcome would help physicians determine the patients who will develop a severe form of the disease and treat them with appropriate aggressive therapy at an early stage. This ability is even more important with the availability of new treatments that can reduce or even stop the progression of RA. Radiographic damage is frequently used as a major assessment criterion for RA outcome. Numerous studies have identified possible initial individual factors associated with worse radiographic outcome, but there are many discrepancies between the studies and few were long-term (or) and prospective.

Joint damage increases slowly over the course of RA, and disability, decreasing during the first years, worsens with disease duration [1]. Disability in RA is influenced by parameters such as age, sex, social and psychological factors, muscle strength and co-morbidities. It is also associated with disease-related factors such as disease activity and joint destruction. The links between functional disability, joint damage and disease activity seem to vary with disease duration [2–4]. In early RA, functional impairment is believed to be mostly due to inflammatory processes as measured by disease activity [2, 4–6]. In established RA, disability may be due to joint damage [2–4]. Prospective studies of the links between joint damage and functional disability are scarce and discordant in part, so the association between damage and disability remains uncertain.

Several assessment tools are available for measuring functional capacity. The easiest and cheapest are self-administered questionnaires. The most widely used instrument for assessing functional capacity in RA is the Health Assessment Questionnaire Disability Index (HAQ-DI) [7]. Joint damage is commonly assessed with radiographic scores, such as the Sharp score, modified by van der Heijde [8].

The main objective of our study was to determine the predictive factors of long-term radiographic outcome in early RA. The secondary objective was to describe the long-term outcome of joint destruction and disability in RA and their interrelation over the course of the disease.

Evaluating the prognosis of RA is more than ever of high importance. Many studies have been published on this subject, but the results are often conflicting. Discrepancies are probably due to differences between study designs and length of follow-up. Only five prospective studies were conducted over more than seven years [17–21]. We performed a long-term study of 10 years to investigate predictive factors of radiographic outcome in RA and found the best independent predictive factor of the 10-year radiographic score to be baseline erosion score. After excluding radiographic scores from the entry parameters, the presence of ACPA and ESR were also predictive of the final total Sharp score.

The predictive value of the baseline radiographic score has been shown in many short-term studies [15, 22–24] and two long-term studies [17, 18]. For Kaarela and colleagues, who followed 200 patients for six to nine years (mean 7.6 years), the independent predictive factor of final radiographic score was baseline radiographic score [17]. In the study by Lindqvist, the baseline radiographic score, determined by the Larsen's method, was correlated with radiographic progression at five and 10 years by univariate analysis [18]. In the other long-term cohorts, the predictive value of the initial joint destruction was not studied.

In our study, the presence of anti-CCP antibodies at baseline was strongly associated with total radiographic score after 10 years, but this parameter was not selected as an independent predictive factor on multivariate analysis, because of the number of missing data (24 of 112) at baseline. With the contribution of data on anti-perinuclear or anti-keratin antibodies, the logistic regression model identified the presence of ACPA at baseline as an independent predictive factor of the total Sharp score after 10 years. The predictive value of the anti-CCP antibodies had already been suggested in short-term studies [22, 24–26]. Also, in a recent 10-year longitudinal study, Syversen and colleagues showed the presence and level of anti-CCP antibodies was predictive of radiographic progression [20].

A high ESR at baseline also predicted an elevated final total Sharp score in our study. This result is in agreement with that of most short-term [15, 23, 27, 28] and long-term studies [17, 20, 21]. In our study, CRP level was not predictive of the final Sharp score, and the predictive value of this parameter remains controversial in the literature, perhaps because of different dosage techniques.

Being positive for and the level of IgA RF present were strongly correlated to 10-year Sharp score by univariate analysis, but these parameters were not selected as independent predictive factors on logistic regression. Both IgA and IgM RF were also correlated with radiographic progression. Very few studies have distinguished the predictive value of these two isotypes of RF. Syversen and colleagues showed IgM RF was an independent predictive factor of 10-year radiographic progression. Lindqvist and colleagues obtained an equivalent result during the same length of follow-up but did not distinguish between IgA and IgM RF in one of their studies [18]. In another study, they showed a significant association between the presence of IgA RF and more severe joint damage after five years, but the presence of RF, whatever the isotype, did not predict the radiographic score after 10 years [19]. Two other long-term studies also isolated the presence of RF at baseline as an independent predictive factor of radiographic score after 7.6 and 8.6 years [17, 21].

We also noticed the potential interest of the MMP3 level, involved in degradation of cartilage proteoglycans, as a predictive factor of radiographic outcome. We found quite a strong correlation between baseline level of MMP3 and final radiographic score and with 10-year radiographic progression. To our knowledge, ours is the only long-term study to take this baseline parameter into account. Two short-term studies had shown such a correlation [29, 30].

The role of several demographic and clinical parameters, such as sex, age and number of tender or swollen joints, has been suggested by several short-term studies, with conflicting results. None of these factors were shown to be independent prognostic factors in our study. In long-term studies, only Syversen and colleagues found an influence of the female sex on prognosis [20]; Kaarela and colleagues found elevated age at diagnosis and a high number of swollen joints to be independent prognostic factors [17], but these findings were not confirmed in other studies.

The predictive value of the presence of the shared epitope, suggested in our study after three years of follow-up and in a few other short-term studies, was not confirmed after 10 years [15, 31–33]. These results are in agreement with Lindqvist and colleagues who found the presence of the shared epitope to be predictive of radiographic progression during the first five years but not after five years of follow-up [18]. Even if the decrease in the number of patients with time can partly explain these results, we can assume that the genetic data influence the radiographic outcome in the short but not long term.

As mentioned in the results, in this study it was unfortunately not possible to carefully analyse the potential effect of corticosteroids during the first years of the disease on the occurrence of erosions.

The mean HAQ score at baseline was higher in our cohort than that found in other prospective studies of early RA (1.29 vs. 0.63 to 1) [2, 34, 35]. This result may be due to our patients not having received DMARDs at baseline. The decrease in HAQ score we observed at three years confirmed the results of two previous studies showing an improvement in functional capacity during the first two years of RA [36, 37]. After 10 years, the mean HAQ score for our patients was similar to that from other cohorts.

The radiographic scores in our study were slightly lower than those for other prospective cohorts at baseline: the median total modified Sharp score reached only 2 in our cohort compared with 11 in the Welsing and colleagues study and 12 in the Drossaers-Bakker and colleagues study [2, 3]. However, the proportion of patients with radiographic erosion at baseline was similar to that found in the study by Lindqvist and colleagues [18]. Of interest, after 10 years, the radiographic score in our study was very low compared with that for other cohorts: the median reached only 18 in our study compared with 83 in the Welsing and colleagues study after nine years and 145 in the Drossaers-Bakker and colleagues study after 12 years [2, 3]. The proportion of patients with no erosion at 10 years was 16.9% in our study compared with 4% in the Lindqvist and colleagues cohort [18]. In our study, joint damage gradually, slightly worsened over the 10-year follow-up, without a higher progression rate during the first years of the disease as shown in previous reports [38, 39]. The most likely explanation is the difference in treatments received by patients, because in our study most patients were treated with DMARDs, such as methotrexate, which have demonstrated a structural effect, compared with older studies where patients received other DMARDs, frequently hydroxychloroquine.

We did not find any correlation between HAQ and Sharp scores throughout the study, but disease activity and radiographic scores remained strongly linked. Several previous studies had suggested that functional capacity was influenced largely by disease activity in early RA and by joint destruction in established RA [2–4]. However, the radiographic scores we found after 10 years were lower than those observed in these older studies. Our results could reflect the consequences of an adequate management of early RA, and perhaps the expected links between HAQ and Sharp scores appear much later in the evolution of the disease.

In this prospective study, baseline radiographic score, ESR and ACPA were the best predictive factors of 10-year radiographic outcome in early RA. HAQ disability was associated with disease activity throughout the 10-year follow-up but not with joint damage. This discrepancy with previous reports may be partly due to the early start of DMARD therapy.