CAN SDAI REMISSION BE PREDICTED IN PATIENTS WITH ESTABLISHED RHEUMATOID ARTHRITIS TREATED WITH ANTI-TNF AGENTS?

Objective. To identify predictors for Simplifi ed Disease Activity Index (SDAI) remission in established rheumatoid arthritis (RA) and to develop a predictive score for remission. Methods. Prospective 12-month observational study in ninety active RA receiving their fi rst TNF-α inhibitor. Standard assessments consisted of disease activity scores (DAS28-ESR, SDAI) and immune parameters (total rheumatoid factor, RF; IGA-RF; anti-cyclic citrullinated peptide antibodies, ACPA). The primary outcome measure was SDAI remission (≤ 3.3) at 12 months. Univariate and multivariate logistic regression models were used to estimate association between baseline variables and SDAI remission. Results. 39.7% RA achieved remission, while 56.8% low disease activity. Signifi cant association between SDAI remission and RA-onset before 50 (p = 0.000), history <5 years (p = 0.000), stage (p = 0.000), class I and II Steinbroker functional status (p = 0.022), HAQ-DI≤2 (p = 0.034), CRP ≤ 20mg/l (p = 0.041), IgA-RF ≤ 20 IU/ml (p = 0.002), ACPA ≤ 40 IU/ml (p = 0.047), concomitant DMARDs (p = 0.003) were identifi ed. Four parameters independently predicted 12-month remission (age at onset under 50, RA duration <5 years, ACPA≤40 IU/ml, IgA-RF ≤ 20 IU/ml) as demonstrated by multivariate logistic regression (p<0.05), making correct prediction in 84.4% patients. Furthermore, the remission score correctly classifi ed 90.6% RA, while the transformed simplifi ed version up to 89.4% cases. Gender, clinical parameters and ESR were not predictors for treatment response (p > 0.05). Conclusion. SDAI remission can be predicted in established RA using a score based on age at onset, disease duration, titers of ACPA and RF isotype A. Such a simplifi ed score may help clinicians to manage remission in RA patients according to the current treatment guidelines.


INTRODUCTION
Rheumatoid arthritis (RA) is a chronic systemic infl ammatory disease of a still unknown etiology, but a complex and dynamic pathophysiology, hypothesized to develop in genetically susceptible host (1).
It has been shown that early intervention with disease-modifying antirheumatic therapies (non-biologic and biologic DMARDs) represents the optimal care of patients with RA and gives the best opportunity for attempting to achieve disease remission (2).
However, predicting the course of a particular case of RA at the outset of a new treatment option remains challenging, although different predictors of an unfavorable prognosis in terms of joint damage and disability have already been recognized (3).
Remission is considered a disease status, not a simple change or transition, meaning generally the absence of disease activity and predicting the best clinical, functional and structural outcomes. Nevertheless, increasing numbers of patients reaching remission as well as abundance of Boolean and indexbased remission defi nitions (Disease Activity Score DAS28, Clinical Disease Activity Index CDAI, and Simplifi ed Disease Activity Index SDAI) have suggested the need for a uniform defi nition of RA remis-sion such as the new ACR/EULAR provisional definition of RA remission (4,5,6).
SDAI is a simple sum of fi ve outcome measures counting tender and swollen joints, patient and physician global RA assessment as well as C-reactive protein (CRP). Different cut-off levels are actually validated for SDAI as follows: 3.3 indicating remission (REM), 11 indicating low disease activity (LDA), 26 moderate disease activity (MDA), while a value higher than 26 is commonly used for defining high disease activity (HDA) (7,8).
Several research papers (9,10,11,12) have already indicated that SDAI may be successfully used in clinical practice instead of other validated indexes in order to defi ne patients achieving remission in various RA settings.
We performed a prospective study aiming to identify predictors for SDAI remission in patients with established RA treated with TNF inhibitors and to develop a prediction score for remission.

PATIENTS AND METHODS
Ninety consecutive patients fulfi lling the 1987 ACR classifi cation criteria for RA, with established severe active disease (DAS28 ≥ 5.1, SDAI ≥ 26) requiring biologics were enrolled in a prospective observational 12-months study. The inclusion and exclusion criteria were defi ned according to the recommendations of the Romanian Society of Rheumatology that calls for anti-TNFs in highly active disease with suboptimal response to previous therapy with at least two synthetic Disease Modifying Anti-Rheumatic Drugs (DMARDs) including methotrexate (13).
Patients were assigned to one of three treatment groups according to the decision of their treating rheumatologist, 33 RA further receiving adalimumab, 30 etanercept and 27 infl iximab. Concomitant non-biological DMARDs (methotrexate, lefl unomide, sulfasalazine, hydroxycloroquine) were allowed, while oral corticosteroids (≤ 10 mg/day prednisone or equivalent) only if maintained at a stable dose within 12 weeks prior to enrollment.
The main outcome was SDAI remission at 12 months. Treatment response was measured by EU-LAR-DAS28 criteria (14).
Local Ethical Committee approval and written informed consent were obtained prior the study.

STATISTICAL ANALYSIS
The baseline characteristics were analyzed by Mann-Whitney U test for continuous variables, whereas chi square was used for categorical variables.
Univariate and multivariate logistic regression expressed as odds ration (OR), with 95% confi dential interval (CI), and 2-tailed "p" were used to estimate the association between potential predictors and SDAI remission.
We created several models based on different (demographic, clinical, biological) variables aiming to investigate their infl uence on disease outcome (SDAI remission) after 12 months of biological therapy. We considered remission as a binary variable and, subsequently, we constructed binary logistic regression models using different parameters as predictors. We have applied the logistic regression Forward LR model; the innitial step without any predictor, further steps of the algorithm adding one by one different predictors.
Variables included in the logistic regression model have been identifi ed to be statistically signifi cant (p < 0.05) by initial individual logistic regression (univariate analysis) performed for each parameter potentially involved as a predictor for therapeutic response To obtaine a remission score we considered the mathematical formula (1 × V1) + (2 × V2) + (3 × V3) …+ α + e, where  is the regression coeffi cient of the variable, V the independent variable, alpha the constant and e the error. To evaluate the predictive power of the score we constructed a receiver operating characteristic (ROC) curve, the area under this curve measuring the concordance of predictive values with current outcome.
Finally, we developed a simplifi ed remission score based on factors obtained in the regression analysis. All statistical analysis was carried out with SPSS16, with "p" < 0.05.

Patients and remission rates
A cohort of long-standing RA, mainly female (81.1%), with an average age 55.56 ± 10.75 years, average disease duration 10.9 ± 6.2 years, mean DAS28 7.50 ± 0.40 and mean SDAI 51.38 ± 5.42 were enrolled in the study. Baseline characteristics ( Table 1) did not differ signifi cantly between study groups, except the ESR (p = 0.009).

Predictors
All variables were independently analyzed by univariate logistic regression; only nine parameters were statistically signifi cant (chi squared, p<0.05) and used further as predictors for remission: age at onset, RA duration and stage, functional class and HAQ-DI, CRP, IgA-RF, ACPA, and concomitant DMARDs.
Statistical analysis showed signifi cant association between SDAI remission and onset before 50 years (OR: 5 (15), we generated a multivariate remission score using the coeffi cients derived from multivariate logistic regression.
We calculated the score for each patient: higher values suggested higher probability that a specifi c patient will achieve SDAI remission after 12 months of anti-TNFs. The area under the ROC curve was 0.90 ( Figure 1A), meaning that the score correctly classifi ed 90.6% of patients, with high sensitivity (84.8%) but low specifi city (19.3%).

Simplifi ed remission score
As the proposed score for predictings remission is time consuming and possible not feasible in daily practice, we created a simplifi ed version in which we assigned either zero or one point for each of the above mentioned parameters: one point for age at onset before 50, one point for symptom duration under 5 years, one point for IgA-RF ≤ 20 IU/ ml and one point for ACPA ≤ 40 IU/ ml. Thus, our simplifi ed remission score offered values ranging from 0 to 4 for all above signifi cant remission predictors.
The transformed scores revealed a signifi cant high correlation between simplifi ed and multivariate remission scores (Spearman's test: r=0.99, p<0.05).
The area under ROC curve was 0.89 (fi gure 1B), the score correctly classifying up to 89.4% of patients with established RA, with a high sensitivity (87.9%) but poor specifi city (26.3%).
Furthermore, we evaluated the likelihood of achieving SDAI remission as well as LDA and MDA.
The results are shown in fi gure 2, 15.6% patients had a simplifi ed remission score of 3 and 10% a remission score of 4, with a high chance of remission at 12 months (78.6% and 100% respectively). On the other hand, low scores as defi ned by values of 0 and 1 were seen in 23.3% and 27.8% of RA, respectively, and were not associated with SDAI remission.

DISCUSSION
We have hypothezed that SDAI remission could be predictied in patients with active established RA treated with TNF inhibitors by using a simplifi ed validated index (SDAI) and we have, fi nally, demonstrated that several predictors are currenly relevant for patients matching the inclusion criteria. Moreover, we proposed a prediction model for SDAI remission in established RA based on four parameters (age of patient at RA onset, disease duration, IgA isotype of RF and ACPA levels) that effectively classifi ed up to 80% of cases as achieving remission. That means patients with a disease onnset before the age of 50, with a RA history lower than 5 years, low levels of IgA-RF (< 20 IU/ml) and ACPA (< 40 IU/ ml) at baseline, before starting the biologic DMARD were most likely to achieve SDAI remission over 12 months of a TNF blocking agent (infl iximab, etanercept, adalimumab).
Reaching remission is a desirable status even in patients with longstanding established RA (4,6,10,11). Moreover, LDA is considered a good alternative op- tion in patients in whom remission cannot be reached based on varia reasons ranging from comorbidities, contraindications to negative prognostic factors (10,11). In our study we have particularly focused on the SDAI remission as, to our knowledge, this topic was only occasionally directed especially in established RA. We considered SDAI remission at a single time point (12 months) as the individual condition to evaluate the response to biologic treatment and we defi ned four independent predive factors: age, symptoms duration, and IgA-RF and ACPA levels as well. Interestingly, we did not fi nd any clinical parameter or acute phase reactants involved in SDAI remission prediction.
Furthermore, as the multivariate remission score could be considered too laborious for routine clinic, we provided a simplifi ed score, intended to correctly identify future responders who can be given additional treatment with TNF inhibitors in daily practice.
As our predictive model of SDAI remission in established RA receiving anti-TNF therapy was the result of a small cohort of patients, the model should be validated in larger cohorts and the concept extended to other RA clinical settings and for other biologic and non-biologic DMARDs.

CONCLUSION
SDAI remission is predictible in biological-naive established RA using a score based on age, disease duration, baseline ACPA and IgA-RF. Such a simplifi ed score may help clinicians to manage remission in RA patients according to the current treatment guidelines.