The Status of Rheumatoid Factor and Anti-Cyclic Citrullinated Peptide Antibody Are Not Associated with the Effect of Anti-TNFα Agent Treatment in Patients with Rheumatoid Arthritis: A Meta-Analysis

Objectives This meta-analysis was conducted to investigate whether the status of rheumatoid factor (RF) and anti-cyclic citrullinated peptide (anti-CCP) antibody are associated with the clinical response to anti-tumor necrosis factor (TNF) alpha treatment in rheumatoid arthritis (RA). Methods A systemic literature review was performed using the MEDLINE, SCOPUS, Cochrane Library, ISI Web of Knowledge, and Clinical Trials Register databases, and Hayden's criteria of quality assessment for prognostic studies were used to evaluate all of the studies. The correlation between the RF and anti-CCP antibody status with the treatment effect of anti-TNFα agents was analyzed separately using the Mantel Haenszel method. A fixed-effects model was used when there was no significant heterogeneity; otherwise, a random-effects model was applied. Publication bias was assessed using Egger's linear regression and a funnel plot. Results A total of 14 studies involving 5561 RA patients meeting the inclusion criteria were included. The overall analysis showed that the pooled relative risk for the predictive effects of the RF and anti-CCP antibody status on patient response to anti-TNFα agents was 0.98 (95% CI: 0.91–1.05, p = 0.54) and 0.88 (95% CI: 0.76–1.03, p = 0.11), respectively, with I2 values of 43% (p = 0.05) and 67% (p<0.01), respectively. Subgroup analyses of different anti-TNFα treatments (infliximab vs. etanercept vs. adalimumab vs. golimumab), response criteria (DAS28 vs. ACR20 vs. EULAR response), follow-up period (≥6 vs. <6 months), and ethnic group did not reveal a significant association for the status of RF and anti-CCP. Conclusions Neither the RF nor anti-CCP antibody status in RA patients is associated with a clinical response to anti-TNFα treatment.


Introduction
Rheumatoid arthritis (RA) is a chronic inflammatory autoimmune disease that affects approximately 1% of the population worldwide [1]. OLE_LINK10Although the introduction of anti-TNFa agents has dramatically improved the outcome of RA, there unfortunately remains a proportion of RA patients who do not exhibit an adequate response to this treatment. Considering the high cost and potential side effects of anti-TNFa treatment, it is important to identify those RA patients who will be more likely to respond to these agents. Indeed, numerous studies have been conducted to investigate potential predictors for patient response to anti-TNFa therapy [2][3][4].
We performed a meta-analysis to investigate whether RF and anti-CCP have predictive value for a clinical response to anti-TNFa treatment. Suitable studies investigating an association of the status of RF or anti-CCP and response to anti-TNFa treatment were searched and included. We also performed subgroup analyses on different variables to explore potential sources of independent predictive factors for an effect of anti-TNFa treatment.

Search strategy
A literature search was performed for all studies evaluating an association between the status of RF or anti-CCP antibody and a response to anti-TNFa therapy in RA patients using the Medline, Cochrane Library, SCOPUS (including EMbase), ISI Web of Knowledge, and Clinical Trials Register (clinical trials.gov) databases. The following keywords were searched: rheumatoid arthritis, anti-TNFa, rheumatoid factor, anti-cyclic citrullinated peptide antibody, clinical trials, and systematic review. Synonyms and spelling variations were taken into account (Search strategy for Scopus was listed in Table S1 in File S1). There was a limitation with regard to language, i.e., we only considered English publications, but not the year of publication. We also contacted authors to request a full-text review or specific data from studies when there was no electronic version of the full text or sufficient data for the meta-analysis. Citations were reviewed to search   Table 2. The results of an assessment for bias in accordance with Hayden's criteria. Unsure: not enough information to evaluate. Low, low risk of bias; moderate, moderate risk of bias; high, high risk of bias. *One of the domains was assessed as ''unsure'' due to unavailable information even after the authors were contacted. doi:10.1371/journal.pone.0089442.t002 relevant original studies, and an electronic search alert was set to cover recent studies.

Study selection
There were 1649 references identified by the literature search. Three individual investigators (QL, YY, & XL) evaluated the references, and the decision of inclusion was made by consensus. A study was included based on the following criteria: 1) the patients were older than 16 years old, diagnosed with RA using ACR criteria, and treated with at least one anti-TNFa agent (adalimumab, infliximab, etanercept, certolizumab, or golimumab); 2) efficacy was measured using EULAR or ACR or DAS28 criteria after a minimum duration of 12 weeks; and 3) the status of RF or anti-CCP antibody at baseline and sufficient data to calculate the risk ratio (RR) were reported in the study. The following information was extracted from each study: the study design, patients' characteristics, baseline status of RF or anti-CCP antibody, interventions, outcomes, and study duration.

Study assessment
All of the studies included were evaluated for potential bias using Hayden's criteria of the quality assessment of prognostic studies [20] (The details of the modified criteria were provided in Table S2 in File S2). Six domains were considered in the bias assessment method: ''study participation'', ''study attrition'', ''prognostic factor measurement'', ''outcome measurement'', ''confounding measurement and account'', and ''analysis''. Scores of ''0'', ''1'', or ''2'' represented ''high bias'', ''partial bias'', and ''low bias'', respectively. If the available information was insufficient to obtain a decision after contacting the authors, ''unsure'' was used, and no score was calculated. The total scores were calculated to show the level of study quality; studies with a score of ''11-12'' were rated as having relatively high quality, with ''9-10'' as moderate, and ''less than 8'' as low quality. The evaluation was performed independently by three investigators (QL, YY, & XL) to improve the validity of the results.

Statistical analysis
We estimated the risk ratio of the RF or anti-CCP antibody status for a response to anti-TNFa treatment. The point estimate of the risk ratio (RR) and 95% confidence intervals (CI) were calculated for each study, and a Mantel Haenszel analysis was used to calculate the pooled RR and 95% CI. The heterogeneity of the effects across studies was assessed using the Chi square test and qualified by I 2 values, which represent the proportion of between-study variability that is attributable to heterogeneity rather than to chance; I 2 values of 25%, 50%, and 75% are referred to as low, moderate, and high heterogeneity effects, respectively. A fixed-effects model was used when there was no significant heterogeneity, whereas a random-effects model was used in other cases. Subgroup analyses were performed to investigate potential sources of heterogeneity by stratifying the different anti-TNFa agents, response criteria, duration of followup, and ethnic group. In addition, a sensitivity analysis was performed by the sequential omission of individual studies. Publication bias was estimated by funnel plots and an Egger's test, with p,0.05 considered as representing significant bias. All of the statistical analyses were performed using Review Manager 5.0 software, and the results of the publication bias were estimated using STATA 12.0 software.  Table S3 in File S1). As two described the same cohort [11,21], fourteen studies were ultimately included in this meta-analysis [7,9,10,[12][13][14][15][16][17][18][19][21][22][23]. These studies comprised a total of 5561 RA patients: 5374 patients had their status of RF detected, and 1283 patients had their status of anti-CCP antibody detected. All but one were prospective cohort studies. All of the participants in the 14 studies had active RA, as evaluated by the number of swollen and tender joints, or DAS28 value, and the patients in all but one_ENREF_7 study had exhibited failure for at least one type of Disease-modifying Anti-Rheumatic Drug (DMARD) treatment [13]. Almost all of the patients had established RA, with a mean disease duration from 5 to 14 years. The data extracted from the 14 studies are summarized in Table 1. With regard to the anti-TNFa agents utilized, 10 studies involved infliximab, 8 involved adalimumab, 6 involved etanercept, and 1 involved golimumab.

Characteristics of the included studies
There was no study that involved patients receiving certolizumab. Only four studies investigated the status of RF, with only two investigating anti-CCP; the other eight studies measured both RF Table 3. Subgroup meta-analysis of RF and RA patient response according to different anti-TNFa agents, follow-up periods, response criteria, and ethnic groups.

Test of association
Test of heterogeneity

Study quality
Our assessment was performed on the basis of Hayden's criteria for prognostic studies [20]. The results revealed that only one of the included studies was considered to be of low quality, with a score of 7; the other thirteen studies were considered as having moderate or high quality (11 moderate and 2 high) ( Table 2).

Association evaluation of RF/anti-CCP antibody status and patient response to anti-TNFa agents
Twelve of the included studies, with a total of 5374 patients, examined the predictive effect of the RF status for patient response Table 4. Subgroup meta-analysis of anti-CCP and RA patient response according to different anti-TNFa agents, follow-up periods, response criteria, and ethnic groups.

Test of association
Test of heterogeneity  to anti-TNFa agents. The overall analysis revealed that the pooled RR was 0.98 (95% CI: 0.91-1.05, p = 0.54) and that I 2 was 43% (Figure 2a), suggesting that the RF status was not associated with a patient's response to anti-TNFa treatment, with a moderate heterogeneity observed. Subgroup analyses on the different anti-TNFa agents, follow-up periods, response criteria, and ethnic groups were performed to investigate sources of this heterogeneity, though the results revealed no significant differences (Figure 2b, Table 3). Ten studies, with a total of 1283 RA patients, were analyzed for an association of anti-CCP antibody status and patient response to anti-TNFa treatment. The overall meta-analysis showed no association between the status of anti-CCP antibody and a patient's response to anti-TNFa treatment, with a pooled RR of 0.88 (95% CI: 0.76-1.03, p = 0.11) and an I 2 of 67% (Figure 3a). When we performed a subgroup analysis to explore potential heterogeneity, we did not identify any association by stratifying the studies using either different anti-TNFa agents, follow-up periods, response criteria, or ethnic groups (Figure 3b, Table 4).

Sensitivity and publication bias
Sensitivity analyses were performed based on the results of the bias assessments, Hayden's criteria of quality assessment for prognostic studies, by the sequential omission of individual studies, and the result revealed that the significance estimate of the overall pooled RR was not influenced by omitting any single study. The publication bias was evaluated by a funnel plot, which showed no significant evidence of asymmetry (Figures 4-5). We also performed an Egger's linear regression test to quantify the publication bias. The p values of the RF and anti-CCP antibody status analyses were 0.777 and 0.422 (Figures 4-5), and all other Egger's test p values in this meta-analysis were .0.05, suggesting no significant bias in the analysis.

Discussion
Previous single studies have suggested that the presence of RF and anti-CCP antibody is associated with a reduced response to anti-TNFa treatment [7][8][9][10][11][12][13][14];_ENREF_9 however, a number of studies have reported a contrasting conclusion [15][16][17][18][19]. Thus, to investigate whether the status of RF or anti-CCP antibody in RA patients has predictive value for a clinical response to anti-TNFa treatment, we performed a meta-analysis by a comprehensive literature search and inclusion of all of the available qualifying studies. The results of our meta-analysis demonstrated that there was no association between the status of RF and patient response to anti-TNFa treatment. Furthermore, the status of anti-CCP antibody was also not associated with patient response, with a pooled RR of 0.88 and 95% CI of 0.76-1.03 (p = 0.11), indicating a tendency of association between the absence of anti-CCP antibodies and a patient's response to anti-TNFa therapy. Therefore the cumulative results did not support or completely reject the association between the presence of RF and/or anti-CCP and the effect of treatment with TNFa inhibitors. Anti-CCP is an independent predictor of radiological damage and progression in RA patients [24,25], and anti-CCP-positive patients have a relatively worse prognosis. As a result, patients with a different anti-CCP antibody status may have a different baseline of the disease, which could influence their response to anti-TNFa agents. To explore this influence, analyses should be performed to investigate potential differences in the baseline characteristics between groups stratified by the status of anti-CCP antibody. Alternatively, only analyses based on studies with placebo groups are powered to definitively identify such an association. Unfortunately, such information was not available for the included studies.
This meta-analysis does have some limitations. First, only 14 studies were included in this meta-analysis, with 12 studies reporting the status of RF and 10 studies reporting the status of anti-CCP antibody. Nonetheless, after a comprehensive literature search covering five databases and the selection of eligible studies by three different investigator strictly according to the inclusion criteria, most of the included studies were assessed with moderateto-high quality. Therefore, it is reasonable to obtain a conclusion based on the studies included in this meta-analysis. However, as potential heterogeneity exists between studies after subgroup analyses were performed, the result of our analyses should be interpreted with caution, and our conclusions should be updated using new, larger studies in the future. Secondly, all of the patients in the studies included in our meta-analysis had long-standing disease, with a mean value of disease duration ranging from 5 to 14 years, and had previously shown failure with at least one type of DMARD. Indeed, patients with more severe chronic disease as a result of irreversible joint damage may be less likely to respond to anti-TNFa treatment [17]. Thirdly, most of the included studies reported a combined treatment of DMARDs in a proportion of the patients enrolled. Because studies have demonstrated improved outcomes among patients receiving combination therapy with MTX plus anti-TNFa [17], the heterogeneity of combined therapy among the studies may have introduced potential bias.
In summary, our meta-analysis indicated that the status of RF and anti-CCP are not associated with the clinical response to anti-TNFa treatment in RA patients.

Supporting Information
File S1 File includes Tables S1-S3. Table S1: Search strategy for Scopus. Table S2: Guidelines for assessing quality in prognostic studies.