Influence of the FCGR2A rs1801274 and FCGR3A rs396991 Polymorphisms on Response to Abatacept in Patients with Rheumatoid Arthritis

Abatacept (ABA) is an immunosuppressant indicated for treatment of rheumatoid arthritis (RA). Effectiveness might be influenced by clinical RA variants and single-nucleotide polymorphisms (SNPs) in genes encoding protein FCGR2A (His131Arg) and FCGR3A (Phe158Val) involved in pharmacokinetics of ABA. An observational cohort study was conducted in 120 RA Caucasian patients treated with ABA for 6 and 12 months. Patients with the FCGR2A rs1801274-AA genotype (FCGR2A-p.131His) showed a better EULAR response (OR = 2.43; 95% CI = 1.01–5.92) at 12 months and low disease activity (LDA) at 6 months (OR = 3.16; 95% CI = 1.19–8.66) and 12 months (OR = 6.62; 95% CI = 1.25–46.89) of treatment with ABA. A tendency was observed towards an association between the FCGR3A rs396991-A allele (FCGR3A-p.158Phe) and better therapeutic response to ABA after 12 months of treatment (p = 0.078). Moreover, we found a significant association between the low-affinity FCGR2A/FCGR3A haplotypes variable and LDA after 12 months of ABA treatment (OR = 1.59; 95% CI = 1.01–2.58). The clinical variables associated with better response to ABA were lower age at starting ABA (OR = 1.06; 95% CI = 1.02–1.11) and greater duration of ABA treatment (OR = 1.02; 95% CI = 1.01–1.04), lower duration of previous biological therapies (OR = 0.99; 95% CI = 0.98–0.99), non-administration of concomitant disease-modifying antirheumatic drugs (DMARDs) (OR = 24.53; 95% CI = 3.46–523.80), non-use of concomitant glucocorticoids (OR = 0.12; 95% CI = 0.02–0.47), monotherapy (OR = 19.22; 95% CI = 2.05–343.00), lower initial patient’s visual analogue scale (PVAS) value (OR = 0.95; 95% CI = 0.92–0.97), and lower baseline ESR (OR = 0.92; 95% CI = 0.87–0.97). This study showed that high-affinity FCGR2A-p.131His variant, low-affinity FCGR3A-p.158Phe variant, and combined use of FCGR2A/FCGR3A genetic variations could affect ABA effectiveness. Further studies will be required to confirm these results.


Introduction
Abatacept (ABA) is an immunosuppressant indicated for treatment of rheumatoid arthritis (RA), psoriatic arthritis, and juvenile idiopathic arthritis 1 . ABA comprises a fusion of the extracellular domain of human cytotoxic T-lymphocyte-associated protein 4 (CTLA4) with the fragment crystallizable (Fc) region of human immunoglobulin G1 (IgG1) [1]. The mechanism of action of ABA is based on interaction of CTLA4 with the CD80/CD86 complex, preventing the latter from binding with the CD28 membrane receptor of T lymphocytes [2]. Consequently, the co-stimulatory signal for T cell activation is blocked [2]. The function of the Fc region of IgG is to improve the pharmacokinetics of ABA, increasing its stability and prolonging the half-life of the drug [3,4]. In addition, the Fc region of IgG1 binds to the Fc-gamma receptors (FCGRs) given rise to a series of immune reactions such as apoptosis, cytokine release, antibody-dependent cellular cytotoxicity (ADCC), and macrophage-mediated immune complex elimination [5][6][7][8]. For this reason, the FCGRs play a part in innate and acquired immune activation and have been extensively investigated for their role in the pharmacogenetic of biological therapies (BTs) [4,[9][10][11][12]. Several subtypes of FCGRs have been described, the most extensively studied being FCGR2A and FCGR3A [13,14]. The protein FCGR2A is expressed on monocytes, macrophages, dendritic cells, neutrophils, and platelets, inducing immune reactions such as phagocytosis of opsonized IgGs, ADCC, reactive oxygen species production, and cytokine production [15]. Similarly, FCGR3A is expressed on monocytes, macrophages, neutrophils, and NK cells, promoting phagocytosis and ADCC mechanisms [15]. The affinity of FCGR2A and FCGR3A for the Fc region of IgG of ABA could vary due to structural changes in extracellular domain of the receptors, which could interfere with the therapeutic response to ABA [13,14]. The FCGR structural changes could be produced by single-nucleotide polymorphisms (SNPs) located in the genetic coding region of FCGR2A and FCGR3A [13,14]. The FCGR2A rs1801274 (A > G) polymorphism gives rise to a histidine (His) to arginine (Arg) substitution (His131Arg) [16,17]. As has been observed in previous studies, the FCGR2A with His instead of Arg at position 131 showed greater affinity for IgG1 [18]. The rs396991 (A > C) polymorphism located on the FCGR3A gene produces a phenylalanine (Phe) to valine (Val) substitution (Phe158Val) [19]. The Val158 variant showed a greater affinity for IgG1, IgG2, and IgG3, which was associated with greater immune response (ADCC, apoptosis) [19,20]. Studies conducted on tumor necrosis factor inhibitors (TNFis) have evaluated the relationship of these SNPs to the effectiveness and variation in clearance of the drug, obtaining contradictory results [4,9,11,[21][22][23][24][25]. The low affinity variants FCGR2A-p.131Arg and FCGR3A-p.158Phe could decrease the binding of ABA to the receptors, and thus its clearance would be lower, increasing the therapeutic effect [18]. Cañete et al. showed that patients undergoing treatment with infliximab (IFX) and carrying the FCGR2A rs1801274-GG and FCGR3A rs396991-AA genotypes, both low-affinity, exhibited lower clearance of the drug, and that its therapeutic effect, and consequently the response to therapy, therefore increased [18]. Similarly, these SNPs are related to the development of autoimmune diseases [26]. Low-affinity receptors could present less binding and clearance of autoimmune complexes, increasing tissue damage in patients carrying low-affinity variants and producing less response to treatment [15]. Another recent study, performed by Jiménez Morales et al., identified the high-affinity genotype FCGR2A rs1801274-TT and allele FCGR3A rs396991-G as predictors of greater therapeutic response to rituximab (RTX) and the low-affinity genotype FCGR3A rs396991-TT as a predictor of good response to tocilizumab (TCZ) [9]. Theories about the influence of FCGR2A and FCGR3A SNPs in response to BTs are multiple and the effect of the interaction of these receptors with biologic drugs is unknown. Nevertheless, these SNPs could determine the great interindividual variability in the response to ABA; the therapeutic failure rate is approximately 30% [27]. However, no previous study has evaluated the relationship between the FCGR2A rs1801274 and FCGR3A rs396991 SNPs and ABA response in patients diagnosed with RA. Recently, SNPs have been investigated in genes related to the mechanism of action of ABA, such as CTLA-4 (rs5742909, rs231775), which could be associated with ABA response [28].
Within this conceptual framework, the objective of this study was to evaluate the involvement of the FCGR2A rs1801274 and FCGR3A rs396991 SNPs and low-affinity FCGR2A/FCGR3A haplotypes as possible biomarkers of response to ABA (EULAR re-sponse, low disease activity (LDA), and remission) after 6 and 12 months of treatment in patients diagnosed with RA.

Study Design
We conducted a retrospective observational cohort study.

Ethics Statements
The study was carried out with the approval of the Ethics and Research Committee of the University Hospital Virgen de las Nieves (HUVN) in accordance with the Declaration of Helsinki. The subjects who participated in the study signed an informed consent for collection and genetic analysis of saliva samples and for their donation to the Andalusian Public Health System Biobank. The samples were identified by alphanumeric codes.

Study Population
The study was conducted at the HUVN in Granada, Spain. We recruited 120 patients over the age of 18 years diagnosed with RA in the Rheumatology Department of the HUVN, according to the American College of Rheumatology (ACR) classification criteria [29], treated with ABA for a period of at least 6 months. ABA was administered intravenously (IV), in doses of 500 mg (<60 kg), 750 mg (60-100 kg), or 1.000 mg (>100 kg), in weeks 0, 2, and 4, respectively, and every 4 weeks subsequently at the same doses in 52 patients, or subcutaneously (SC) at 125 mg/week in 68 patients.

Sociodemographic and Clinical Variables
The sociodemographic and clinical data were collected by reviewing clinical histories. The sociodemographic data collected were sex, smoking, age at RA diagnosis, number of years with disease, age at start and duration of ABA treatment, mode of administration of the drug (IV or SC), concomitant GCs, concomitant csDMARDs (methotrexate (MTX), leflunomide (LFN)), number and duration of previous BTs, and reason for suspension of ABA. In addition, the clinical data collected were Disease Activity Score in 28 joints (DAS28) [30][31][32], Health Assessment Questionnaire (HAQ) score, C-reactive protein (CRP) level, erythrocyte sedimentation rate (ESR), presence of rheumatoid factor (RF), anticitrullinated protein antibodies (ACPA), number of painful joints (NPJ), number of swollen joints (NSJ), and patient visual analogue scale (PVAS).

DNA Isolation
Following the patients' inclusion and signing of the informed consent, saliva samples were collected with buccal swabs (OCR-100). The DNA was extracted using the QlAamp DNA Mini Kit (Qiagen GmbH, Hilden, Germany), following the manufacturer's instructions for purifying DNA from saliva, and stored at −40 • C. The DNA concentration and purity were measured using a NanoDrop 2000 UV spectrophotometer with the absorbance ratio at 280/260 and 280/230.

Statistical Analysis
The descriptive analysis was performed using R 3.5.1 software. The quantitative variables were expressed as the mean (±standard deviation) for those that complied with normality and as the median and percentiles (25 and 75) for the variables that did not follow a normal distribution. Normality was confirmed by the Shapiro-Wilk test. The bivariate analysis between the response and the genetic variables was performed using Pearson's chi-squared test or applying Fisher's exact test for the qualitative variables. For the quantitative variables, Student's t-test was applied to the variables that complied with normality. The Mann-Whitney U test was applied for non-normal variables. The lowaffinity FCGR2A/FCGR3A haplotype variable was analyzed as a quantitative variable. The alleles considered as low-affinity for the polymorphisms studied were FCGR2A rs1801274-G (FCGR2A-p.131Arg) and FCGR3A rs396991-A (FCGR3A-p.158Phe), whereas FCGR2A rs1801274-A (FCGR2A-p.131His) and FCGR3A rs396991-C (FCGR3A-p.158Val) are highaffinity alleles [18][19][20]. Transformation from a categorical to a quantitative variable was performed by assigning a score to each allele. A value of 0 was assigned to each of the FCGR2A rs1801274-A and FCGR3A rs396991-C alleles, being high-affinity alleles, while the value 1 was assigned to the low-affinity alleles FCGR2A rs1801274-G and FCGR3A rs396991-A. As each patient had four alleles, two for each polymorphism, the score assigned was from 0 to 4, where 0 signified the absence of low-affinity alleles (AACC) and 4 signified the presence of four low-affinity alleles (GGAA).
Multivariate analysis (logistic or linear regression) was used to calculate the adjusted odds ratio (OR) and 95% confidence interval (95% CI) for potential prognostic factors for EULAR response, LDA, and remission. The goodness of fit for each model was analyzed with the Hosmer-Lemeshow test and the omnibus test of coefficients, and the Cox-Snell and Nagelkerke r 2 coefficients were also calculated. All tests were two-sided, with a probability of 0.05 or less being considered statistically significant, and were performed using R 3.5.1 or PLINK toolset free-access software for whole-genome association analysis [36][37][38].
We determined the Hardy-Weinberg equilibrium and the haplotype frequencies and calculated Lewontin's D-prime (D ) and the linkage disequilibrium coefficient (r 2 ). The linkage disequilibrium (LD) for each polymorphism was calculated with the PLINK genome association analysis program [37]. The haplotype frequencies and their association with the response variable were analyzed using the snpStats program, a web-based tool for analysis of association studies [39][40][41][42][43].

Patient Characteristics
A total of 120 patients receiving ABA as treatment were included in the study. The clinical and sociodemographic data are shown in Table 1. Of all the patients diagnosed with RA, the mean age at diagnosis was 45.15 ± 13.72 years; 74.17% (89/120) were women. All the patients were treated with other DMARDs for a median period of 36 (24-72) months. The median number of previous BTs was 2 (1)(2)(3). The percentage of ABA-naïve patients was 12.50% (15/120). The mean age at starting ABA was 56.63 ± 13.03 years, and the median duration of treatment with ABA was 24 (14.75-44.25) months. In our cohort, 5.83% (7/120) received ABA treatment as monotherapy during the study period. Concomitant MTX and GC were received by 35% (42/120) and 85% (102/120), respectively. The mean DAS28 score on starting ABA was 4.70 ± 1.43. The description of the clinical and sociodemographic parameters is shown in Table 1.

Distribution of the Genotypes Analyzed
All the gene polymorphism distributions were in agreement with those expected according to the Hardy-Weinberg equilibrium (HWE) model (Table S1). The D linkage disequilibrium (LD) and r 2 values are given in Table S2. All the polymorphisms showed minor allele frequencies higher than 1%, and none of them were excluded from the analysis (Table S3).  (Table S10).

Association between Low-Affinity FCGR2A/FCGR3A Haplotypes and ABA Response
Significant association was found in the bivariate analysis between the low-affinity FCGR2A/FCGR3A haplotype variable and LDA after 12 months of treatment with ABA (OR = 1.59; 95% CI = 1.01-2.58) (Table S11). Moreover, a tendency was found towards an association between the low-affinity FCGR2A/FCGR3A haplotype variable and satisfactory EULAR response at 12 months of ABA therapy (p = 0.088) (Table S10).

Discussion
The interindividual response of patients diagnosed with RA and treated with BTs is very variable [33,44,45]. The search for biomarkers of response to these treatments is the main objective of numerous research studies carried out in recent years [9,12,46]. For this purpose, we need to evaluate the effectiveness of the treatments in different populations and find the biomarkers that determine that effectiveness. In our study, bionaïve patients showed a greater EULAR response after 6 and 12 months of ABA treatment (53.33% and 78.57%, respectively) compared to non-bionaïve patients, who had a lesser EULAR response to treatment with ABA at 6 and 12 months (31.67% and 45.71%, respectively). Furthermore, remission of the disease was greater in the bionaïve group than in the non-bionaïve group (64.29% vs. 27.62%) after 12 months of ABA treatment. A study conducted by Cagnotto et al. in 2716 Caucasian patients (from Sweden) with RA found a greater EULAR response in bionaïve patients after 12 months of treatment with ABA (OR adjusted = 4.29, 95% CI = 2.77-6.65) [47]. Moreover, higher values in duration and number of previous BTs were identified as being responsible for a lower response to ABA treatment. The duration of the disease, and also age on starting treatment with ABA and duration of that treatment, have been identified in our study as predictors of ABA treatment response. Other studies have obtained the same results, highlighting the importance of early treatment of RA to achieve remission or, failing that, LDA [48][49][50]. In our patients, the administration of ABA in monotherapy, without concomitant GCs, was associated with a satisfactory EULAR response, as well as with greater remission and LDA, after 12 months of ABA treatment. However, previous studies have not found differences in effectiveness and safety between the use of ABA in monotherapy compared to ABA in combination with other DMARDs [51]. As for the clinical variables measuring the disease, the patients with lower baseline values for DAS28, NPJ, NSJ, PVAS, ESR, and HAQ had a better response to ABA treatment (greater EULAR response, LDA, and remission). These findings have been presented previously in other studies [52,53].
According to the results mentioned thus far, various biomarkers have been found as possible predictors of response to BTs in general and to ABA in particular. If we add to this the use of pharmacogenetics as a tool to achieve personalized medicine, treatments could be optimized so that the development of the disease could be delayed or arrested, avoiding the onset of irreversible disability [54]. The field of pharmacogenetics is one of those involved in this research process, as a number of SNPs have been implicated in the therapeutic response to various biological drugs [9,21,55]. Moreover, SNPs in the FCGR2A and FCGR3A genes have been studied as biomarkers of response to TNFis, RTX, or TCZ in various pathologies [4,9,11,22]. However, no study has investigated the involvement of these SNPs in the therapeutic response to ABA in patients with RA. Our results show that the FCGR2A rs1801274 polymorphism is associated with the clinical effectiveness of ABA after 6 and 12 months of treatment. The FCGR2A rs1801274-AA genotype, considered to be of high affinity, has been associated with satisfactory EULAR response after 6 months and with LDA after 6 and 12 months of ABA treatment. In line with our results, a study conducted by Jiménez Morales et al. in 55 Caucasian patients (from Spain) treated with RTX showed that individuals carrying the FCGR2A rs1801274-AA genotype had a higher rate of remission after 6 months of treatment and greater EULAR response after 6, 12, and 18 months of treatment with RTX (p = 0.035; OR = 1.53, 95% CI = 1.11-21.12) [9]. However, in this same study, no association was found between the FCGR2A rs1801274 polymorphism and therapeutic response to TCZ in 98 Caucasian patients (from Spain) with RA [9]. Another study in 429 Caucasian patients (from Spain) diagnosed with RA and undergoing treatment with IFX showed that individuals carrying the FCGR2A rs1801274-G allele had a lower therapeutic response to this BT (p = 0.04) [56]. Avila-Pedretti et al. presented a study carried out in 348 Caucasian patients (from Spain) diagnosed with RA and treated with adalimumab (ADA), in which individuals carrying the FCGR2A rs1801274-GG genotype did not respond to ADA treatment after 12 weeks of treatment (GG vs. AG/AA; p = 0.022; OR = 2.54; 95% CI = 1. 19-5.40) [46]. In contrast, the results of a meta-analysis performed in 3058 Caucasian patients diagnosed with RA undergoing TNFi therapy showed that Caucasian patients carrying the FCGR2A rs1801274-AA genotype had a lower EULAR response following treatment with ADA (p = 0.029; OR = 0.591, 95% CI = 0.37-0.95; I 2 = 0, p heterogeneity = 0.770) [22]. However, no significant results were obtained in patients treated with IFX and etanercept [22]. The association between the high-affinity FCGR2A-p.His131 variant and the better therapeutic response to ABA may be conditioned by an improved RA immune response [26]. Patients carrying the FCGR2A-p.His131 variant could present greater uptake and elimination of the autoimmune complexes produced by RA, decreasing tissue damage, and presenting a greater therapeutic response to ABA [26].
The FCGR3A rs396991 polymorphism has been investigated in numerous studies on RA and TNFi treatment [4,9,11,21,22]. No previous study has evaluated the association between the FCGR3A rs396991 polymorphism and ABA treatment in patients with RA. In our study, FCGR3A rs396991-A, a low-affinity allele, showed a tendency towards association with the EULAR response at 12 months of ABA therapy (A vs. CC; p = 0.078). Similarly, in a study conducted in 87 Caucasian patients (from Spain) diagnosed with RA, a significant association was found between the FCGR3A rs396991-AA genotype and greater EULAR response after 12 months of treatment with TCZ (AA vs. C; p = 0.027; OR = 5.08; 95% CI = 1.20-21.33) [9]. In this same study an association was found in 55 Caucasian patients (from Spain) diagnosed with RA and being treated with RTX between the FCGR3A rs396991-C allele and LDA after 18 months of treatment with RTX (C vs. AA; p = 0.077; OR = 4.90; 95% CI = 0.84-28.48) [9]. A meta-analysis of 1427 patients receiving TNFi, RTX, and IFX therapy showed that patients being treated with RTX who carried the FCGR3A rs396991-CC and FCGR3A rs396991-CA genotypes showed a lower therapeutic response (CC/CA vs. AA; p = 0.007; OR = 0.566; 95% CI = 0.37-0.86; I 2 = 45.2, p heterogeneity = 0.161) [57]. However, no association was found between the FCGR3A rs396991 polymorphism and effectiveness of TNFis (CC/CA vs. AA; p = 0.186) or IFX (CC/CA vs. AA; p = 0.065) [57]. Similarly, another meta-analysis performed by Montes et al. in 429 Caucasian patients diagnosed with RA and receiving TNFi therapy found no association between the FCGR3A rs396991 polymorphism and response to BTs (p = 0.5; OR = 1.11; 95% CI = 0.8-1.5; I 2 = 62) [4]. The greater response to ABA in patients carrying the low affinity variant FCGR3A-p.158Phe could be due to the lower binding affinity of FCGR3A towards ABA so that the drug remains longer in the blood circulation, increasing the therapeutic response to ABA [4,9].
In our study, we analyzed the joint influence of the FCGR2A rs1801274-G (FCGR2Ap.131Arg) and FCGR3A rs396991-A (FCGR3A-p.158Phe) low-affinity alleles on the effectiveness of ABA in patients diagnosed with RA, since BTs, as we have already described, can be affected by both these SNPs. According to our results, an association was found between a larger number of low-affinity alleles and LDA after 12 months of treatment with ABA (Table S11). The presence of low-affinity FCGR2A/FCGR3A haplotypes could give rise to lower drug plasma clearance, increasing the half-life of ABA and therefore its effectiveness [14,21]. No study has evaluated the association between the additive power of the FCGR2A rs1801274-G and FCGR3A rs396991-A low-affinity alleles and ABA response in RA patients. A previous study in Asian patients (from Japan) diagnosed with RA and undergoing treatment with IFX showed that patients carrying low-affinity haplotypes had lower drug clearance compared to carriers of high-affinity haplotypes [14]. Furthermore, another study carried out by Dávila-Fajardo et al. in Caucasian patients (from Spain) treated with ADA showed an association between low-affinity haplotypes and satisfactory EULAR response (p = 0.017; OR = 1.53; 95% CI = 1.08-2.17) [21]. A study conducted in Caucasian patients (from Spain) with psoriatic arthritis treated with TNFi therapy found a significant association between low-affinity haplotypes and better therapeutic response after 6 to 8 weeks of treatment (p = 0.04) [58]. In our study, the presence of low-affinity FCGR2A/FCGR3A haplotypes was associated with greater clinical effectiveness of biological treatment; however, plasma ABA levels were not measured in our patients, and therefore we cannot correlate the pharmacogenetic response with the pharmacogenetics of ABA. These contradictory results regarding the influence of highand low-affinity alleles in the FCGR2A rs1801274 and FCGR3A rs396991 SNPs could be determined by the different mechanisms of action of each of the BTs indicated for treatment of RA [4]. In addition, the differential expression of FCGR2A and FCGR3A in immune system cells could also influence the contradictory response results obtained between the two receptors [15]. A previous study demonstrated that ABA showed low affinity for the FCGR2 and FCGR3 receptors, and therefore its action through complement-dependent cytotoxicity and ADCC pathways is more limited than may be the case with other BTs that do act through these immunological pathways [9,59]. Furthermore, FCGR2A and FCGR3A effect on the pathophysiology of RA could also influence the response to therapies [26].
The limitation of our study was the sample size, which could be responsible for the loss of statistically significant association between the FCGR2A rs1801274 and FCGR3A rs396991 SNPs and remission after ABA treatment. Nevertheless, all the patients were recruited from the same hospital cohort following the same therapeutic protocols by the same team of rheumatologists, which ensured the homogeneity and reliability of the clinical variables collected. All the patients diagnosed during the study period were recruited, ensuring the representativeness of the sample. Despite the limited sample size, the effects observed in these patients were clear. Further studies will be required in larger cohorts to confirm the prognostic value of the FCGR2A rs1801274 and FCGR3A rs396991 polymorphisms and the response to ABA treatment in patients diagnosed with RA.

Conclusions
In conclusion, this study shows that patients with high-affinity FCGR2A-p.131His and low-affinity FCGR3A-p.158Phe could be associated with better therapeutic response to ABA in patients diagnosed with RA. The presence of low-affinity alleles of FCGR2A and FCGR3A was associated with greater clinical effectiveness and a lower rate of LDA after treatment with ABA. In addition, lower duration of previous BTs, the use of ABA in monotherapy, non-administration of concomitant GCs, greater duration of treatment with ABA, and early age of starting ABA therapy seem to be variables predictive of greater EULAR response, LDA, and remission in the individuals studied. As for clinical markers, lower baseline ESR and PVAS values are associated with better response to ABA therapy.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/10 .3390/jpm11060573/s1, Table S1: Hardy-Weinberg equilibrium, Table S2: Linkage disequilibrium, Table S3: Minor allele frequencies of SNPs, Table S4: Haplotype frequency estimation EULAR response at 6 months of ABA, Table S5: Haplotype frequency estimation LDA at 6 months of ABA, Table S6: Haplotype frequency estimation remission at 6 months of ABA, Table S7: Haplotype frequency estimation EULAR response at 12 months of ABA, Table S8: Haplotype frequency estimation LDA at 12 months of ABA, Table S9: Haplotype frequency estimation remission at 12 months of ABA, Table S10: Predictors of EULAR response at 6 and 12 months of treatment with abatacept in rheumatoid arthritis patients (bivariate analysis), Table S11: Predictors of LDA at 6 and 12 months of treatment with abatacept in rheumatoid arthritis patients (bivariate analysis), Table S12: Predictors of remission at 6 and 12 months of treatment with abatacept in rheumatoid arthritis patients (bivariate analysis). Funding: This work was partly supported by a contract for Noelia Márquez Pete from the University of Granada and the Fundación de Investigación Biosanitaria de Andalucía Oriental (FIBAO). The Virgen de las Nieves University Hospital Biobank was supported by grants co-funded by ERDF funds (EU) from the Instituto de Salud Carlos III (PT13/0010/0039).

Institutional Review Board Statement:
The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Ethics Committee of University Hospital Virgen de las Nieves (HUVN) (20 June 2014).

Informed Consent Statement:
Informed consent was obtained from all subjects involved in the study.

Acknowledgments:
The authors would like to acknowledge all the patients who participated in the study, as well as the professionals from the University Hospital Virgen de las Nieves that contributed to the management of the samples, especially the nursing team from the Rheumatology Department and the professionals of outpatient pharmacy service. The results of this research study are part of the doctoral thesis to be submitted by Noelia Márquez Pete at the University of Granada, carried out within the "Clinical Medicine and Public Health" doctoral program.

Conflicts of Interest:
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported, and that there are no competing financial interests in relation to the work described in this article.