Association of rheumatoid arthritis treatment response and disease duration with declines in serum levels of IgM rheumatoid factor and anti–cyclic citrullinated peptide antibody




To examine the association of treatment response and disease duration with changes in rheumatoid factor (RF) and anti–cyclic citrullinated peptide (anti-CCP) antibody levels among patients with rheumatoid arthritis (RA).


The study sample included 66 RA patients who completed double-blind, randomized clinical protocols and for whom baseline and followup serum samples were available. Anti-CCP and RF levels were measured using commercially available assay kits. Odds ratios (ORs) and 95% confidence intervals (95% CIs) were used to describe the association of response and disease duration with declines in antibody levels.


Patients had a mean ± SD age of 49.9 ± 12.0 years and were predominantly female (n = 51; 77%). The mean ± SD duration between the times at which the baseline and followup serum samples were obtained was 13.7 ± 8.6 months. Among the 64 subjects with positive antibody at baseline, 33 (52%) experienced a ≥25% reduction in the anti-CCP antibody level during the course of treatment, and 35 patients (55%) had a ≥25% reduction in RF. After adjustment for the baseline anti-CCP antibody level, only a shorter disease duration (≤12 months) was significantly associated with a decline in the level of anti-CCP antibody (OR 3.0, 95% CI 1.0–8.8), and no association with treatment response was observed. Conversely, treatment response was the only significant determinant of a decrease in RF levels (OR 3.6, 95% CI 1.2–10.4).


Shorter disease duration predicts greater declines in anti-CCP antibody levels with treatment in RA. Although treatment response is a robust determinant of a decrease in RF, it does not appear to be associated with declines in the anti-CCP antibody level.

Rheumatoid arthritis (RA) is a systemic inflammatory disease frequently characterized by circulating autoantibodies. Rheumatoid factor (RF), an antibody directed against the constant region of IgG, is elevated in ∼75% of patients with RA (1). In addition to RF, autoantibodies targeting cyclic citrullinated peptide (anti-CCP antibodies) are commonly observed in the sera of patients with RA. In contrast to RF, which has only modest disease specificity, anti-CCP antibodies appear to be observed almost exclusively in RA, with a disease specificity approaching 100% (2, 3). This high disease specificity of anti-CCP antibody, coupled with its frequent presence in early—even preclinical—disease (4, 5) suggests that anti-CCP antibody plays an important role in RA pathogenesis.

Several studies have documented that the level of IgM-RF decreases with the administration of effective disease-modifying therapies (6–11). In contrast, few studies have examined the impact of active RA treatments on the level of anti-CCP antibodies (6, 10, 11). In one such study, B cell–depleting therapy resulted in a significant decrease in anti-CCP antibody levels, independent of overall changes in serum immunoglobulins (6). Given the proposed role of anti-CCP antibody in both the pathogenesis and the early diagnosis of RA (4, 5, 12, 13), it is critically important to understand potential determinants of anti-CCP antibody change.

We examined serologic changes in both RF and anti-CCP antibodies among patients with RA who were enrolled in previously reported clinical trials (14–16). We hypothesized that the levels of both RF and anti-CCP antibody would decrease in patients receiving treatment intervention, and that these serologic decreases would be associated with treatment response and disease duration at the time of enrollment.



For the purpose of this investigation, we used a convenience sample of 71 patients with RA, all of whom had participated in 1 of 3 past randomized, double-blind clinical studies (n = 208 total trial participants) (14–16). Subjects who were negative for both anti-CCP antibody (<5 units/ml) and RF (<15 IU/ml) at baseline (n = 5) were eliminated from all analyses, leaving a total of 66 patients with evaluable data. All subjects had a diagnosis of RA according to the classification criteria of the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) (17) and provided written informed consent for post hoc serologic analyses. Patients provided a total of 2 serum samples, one at the time of study enrollment and the other during study followup. Followup samples were obtained at variable time points throughout the course of the investigations. To examine the association of clinical response with changes in autoantibody expression, approximately half of the samples were selected from clinical responders, and the other half were selected from nonresponders. Clinical response was assessed at the same time that followup serum samples were obtained and was defined as 50% clinical improvement using either the ACR criteria (18) (1 study [14]) or the Paulus criteria (19) (2 studies [15,16]) (Table 1).

Table 1. Summary of randomized, double-blind, placebo-controlled protocols for subjects included in post hoc serologic analyses*
Year (ref.)Total no. of subjects (no. of subjects in post hoc analysis)TreatmentInclusion criteriaLength of patient followupOutcome criteria
  • *

    MTX = methotrexate; HCQ = hydroxychloroquine; SSZ = sulfasalazine; DMARD = disease-modifying antirheumatic drug; Paulus 50 = Paulus criteria for 50% improvement (19); RF = rheumatoid factor; ACR50 = American College of Rheumatology core criteria for 50% improvement (18).

1996 (16)102 (24)MTX vs. HCQ–SSZ vs. MTX–HCQ–SSZSuboptimal response to at least 1 DMARD, disease duration >6 months2 yearsPaulus 50
1997 (15)46 (22)Minocycline vs. placeboRF positive, disease duration <1 year6 monthsPaulus 50
2001 (14)60 (20)Minocycline vs. HCQRF positive, disease duration <1 year2 yearsACR50

Following serum collection, samples were processed and stored in a −80°C freezer until used for this study. A description of the 3 study protocols is shown in Table 1. Briefly, one study examined the use of methotrexate (with or without combination disease-modifying antirheumatic agents [DMARDs]) in RA patients with well-established disease (16). The other 2 studies examined the use of minocycline (versus placebo [15] and versus hydroxychloroquine [14]) in RA patients with early disease. All patients in the latter 2 studies had disease duration of <1 year and were seropositive for RF at the time of study entry.

Anti-CCP and RF.

Anti-CCP (IgG) antibodies were measured using commercially available second-generation enzyme-linked immunosorbent assay kits (Diastat; Axis-Shield Diagnostics, Dundee, UK). The assay was performed according to the manufacturer's instructions. Anti-CCP antibodies were measured in arbitrary units per milliliter and were considered to be positive at a cutoff value of ≥5 units/ml. The interassay coefficients of variation for the anti-CCP assay in this laboratory were 12.7% and 17.6% for internal control samples, with means of 76 units/ml and 15 units/ml, respectively. RF (IgM) was measured in IU per milliliter and was assessed using standard nepholometry according to the manufacturer's specifications (Dade-Behring, Newark, DE). A positive result was defined as a level of ≥15 IU/ml (with a lower threshold of detection of 11 IU/ml). The interassay coefficients of variation for RF were 9.3%, 6.2%, and 6.6% for RF concentrations of 92 IU/ml, 155 IU/ml, and 306 IU/ml, respectively (Dade-Behring). For both assays, in cases in which the antibody level was sufficiently high that optical densities did not fall on a standard curve for the original dilution, samples were further diluted until reaching the acceptable range for measurement.

Statistical analysis.

Two patients were anti-CCP antibody negative and RF positive at baseline. Likewise, 2 patients were RF negative and anti-CCP antibody positive at baseline. This resulted in available data for 64 patients for the analysis of each outcome measure. Followup measures for values that fell below the detectable threshold were assumed to be 5 units/ml for anti-CCP (n = 2) and 11 IU/ml for RF (n = 3). Wilcoxon's signed rank test was used to determine whether the median change in the antibody levels differed significantly from zero. Correlations of baseline autoantibody levels and changes in antibody levels were examined using Spearman's rank correlation coefficients. Declines in RF and anti-CCP antibody levels were modeled as separate outcomes. Specifically, we modeled the odds of a ≥25% decline as a dichotomous outcome. Declines of ≥50% were modeled as secondary outcomes. Odds ratios (ORs) and 95% confidence intervals (95% CIs) were used to describe the association of clinical parameters (i.e., treatment response) with declines in autoantibody levels and were calculated using logistic regression. In addition to examining clinical response as a dichotomous predictor variable, we assessed the association of disease duration (≤12 months versus >12 months), age, sex, length of study followup, age of the baseline sample at the time of analysis, and specific DMARD treatments with changes in RF and anti-CCP antibody levels. Data on disease duration were not available for 2 patients. Baseline autoantibody levels served as a covariate in all analyses.

To adjust for confounders of serologic changes in autoantibody levels, multivariate models were developed. All variables with a P value of <0.25 at the univariate level were entered into the initial multivariate analyses. A P value of <0.1 was required to be maintained in the final model. Model goodness-of-fit was assessed using the Hosmer-Lemeshow goodness-of-fit test (20). In subsequent sensitivity analyses, we eliminated patients who appeared to have influential data based on dfbeta statistics or the presence of high residual values (20). To examine whether there was an interaction between disease duration and treatment response, we repeated our analyses, stratifying the sample by response. All analyses were performed using SAS software for Windows (release 8.02; SAS Institute, Cary, NC).


The sociodemographic and disease-related characteristics of the 66 study participants are shown in Table 2. The mean ± SD age of the patients was 49.9 ± 12.0 years, and most patients were female (n = 51; 77%). The most common treatments received included minocycline (29%) and hydroxychloroquine (HCQ; 24%), followed by methotrexate alone (MTX; 17%), the combination of sulfasalazine (SSZ) and HCQ (9%), triple therapy with MTX, HCQ, and SSZ (11%), and placebo (11%). The mean ± SD interval from the time of enrollment to followup was just over 1 year (13.7 ± 8.6 months).

Table 2. Sociodemographic and disease-related characteristics of 66 patients with rheumatoid arthritis*
  • *

    Except where indicated otherwise, values are the number (%).

Age, mean ± SD years49.9 ± 12.0
Female51 (77)
Caucasian ethnicity63 (95)
Disease duration, mean ± SD months52.4 ± 82.4
Followup duration, mean ± SD months13.7 ± 8.6
Age of baseline sample at analysis, mean ± SD months110.1 ± 25.7
Treatment responder35 (53)
Drug treatment 
 Methotrexate11 (17)
 Hydroxychloroquine16 (24)
 Minocycline19 (29)
 Hydroxychloroquine/sulfasalazine6 (9)
 Methotrexate/hydroxychloroquine/sulfasalazine7 (11)
 Placebo7 (11)

Anti-CCP antibody levels decreased significantly, by a median of 33.3 units/ml (25th to 75th percentile 179 units/ml decrease to 3 units/ml increase), over the followup period (P < 0.001). Thirty-three patients (52%) experienced a ≥25% reduction in anti-CCP antibodies, while 21 patients (33%) experienced a reduction of at least 50%. Anti-CCP antibody levels normalized (fell below 5 units/ml) in 2 patients during the course of followup. In 1 patient, anti-CCP antibodies declined from a baseline value of 1,671 units/ml to just 1.4 units/ml (this patient attained clinical remission with minocycline monotherapy).

RF values decreased significantly, by a median of 39.3 IU/ml (25th to 75th percentile 173 IU/ml decrease to 16 IU/ml increase) (P < 0.001). Thirty-five patients (55%) had a ≥25% reduction in RF, while 18 patients (28%) had at least a 50% decline. The RF level normalized (fell below 15 IU/ml) in 3 patients during the course of followup.

A marginally positive correlation between baseline levels of anti-CCP antibody and RF was observed (r = 0.19, P = 0.1). In contrast, individual baseline antibody levels showed strong negative correlations with subsequent declines (Figure 1), with correlation coefficients of −0.5 (P < 0.001) for both anti-CCP antibody and RF.

Figure 1.

Plot graphs showing baseline anti–cyclic citrullinated peptide (anti-CCP) antibody levels and baseline rheumatoid factor (RF) levels versus the change in anti-CCP antibody levels (n = 64) and RF levels (n = 64), respectively, with treatment intervention. There was a strong negative correlation between baseline levels and change (r = −0.5, P < 0.001 for both anti-CCP antibody and RF).

Results of our primary analyses are summarized in Table 3. After adjusting for baseline anti-CCP antibody levels, shorter disease duration (≤12 months) was significantly associated with a ≥25% decline in anti-CCP antibody (OR 3.0, 95% CI 1.0–8.8). There was no association of dichotomous treatment response with decreases in the anti-CCP antibody level, at either the 25% or 50% threshold. Similarly, there was no significant association of a decrease in the anti-CCP antibody level with duration of followup, sex, age, drug received, or age of the sample at baseline. Although not reaching statistical significance (P = 0.17), the association of shorter disease duration with anti-CCP antibody level change remained when we modeled the odds of a ≥50% decline (OR 2.3, 95% CI 0.7–7.2) (data not shown). These results were unchanged after adjusting for multiple potential confounding variables. Disease duration and baseline anti-CCP antibody level were the only variables retained in the final model.

Table 3. Determinants of a ≥25% decline in anti-CCP antibody and IgM-RF levels among RA patients enrolled in randomized, double-blind clinical trials*
VariableAnti-CCP antibody decline ≥25%, OR (95% CI)IgM RF decline ≥25%, OR (95% CI)
  • *

    All associations with rheumatoid factor (RF) change were adjusted for baseline RF level. All associations with anti–cyclic citrullinated peptide (anti-CCP) antibody change were adjusted for baseline anti-CCP antibody level (n = 64 for each antibody measure): P > 0.1 for all models shown, by Hosmer-Lemeshow goodness-of-fit, suggesting adequate model fit. In addition to baseline autoantibody levels, only disease duration (for anti-CCP antibody change) and treatment response (for RF change) were included in the final multivariable models. RA = rheumatoid arthritis; OR = odds ratio; 95% CI = 95% confidence interval; MTX = methotrexate; HCQ = hydroxychloroquine; SSZ = sulfasalazine.

Disease duration ≤12 months2.99 (1.02–8.78)0.71 (0.25–1.99)
Treatment response1.30 (0.47–3.61)3.55 (1.22–10.36)
Followup duration, months0.96 (0.91–1.03)1.04 (0.98–1.10)
Male sex1.08 (0.33–3.59)1.11 (0.33–3.79)
Age, years0.98 (0.93–1.02)1.00 (0.96–1.05)
Drug treatment (vs. placebo)  
 MTX0.41 (0.05–3.21)1.96 (0.25–15.24)
 HCQ2.37 (0.37–15.36)4.14 (0.59–29.16)
 Minocycline2.78 (0.46–16.88)3.46 (0.51–23.67)
 HCQ/SSZ0.37 (0.03–5.27)1.77 (0.16–19.50)
 MTX/HCQ/SSZ0.77 (0.08–7.55)6.14 (0.56–66.84)
Age of baseline sample, months0.98 (0.96–1.00)0.99 (0.97–1.01)

When we repeated the analyses, stratifying the sample by treatment response, we observed no evidence of an interaction with disease duration (data not shown). In subanalyses, these results were unchanged after removing patients with evidence of influential data points (n = 4 for each analysis, as determined using dfbeta statistics and examination of residual values; data not shown).

After adjusting for baseline RF level, only treatment response was associated with a ≥25% decrease in RF (Table 3). In contrast to its association with changes in anti-CCP antibody levels, shorter disease duration had no association with a decrease in RF levels. Additionally, changes in the RF level were not associated with the duration of followup, sex, age, drug received, or age of the sample. Treatment response and baseline RF level were the only variables retained in the final multivariable model. Compared with nonresponders, clinical responders were >3-fold as likely (OR 3.6, 95% CI 1.2–10.4) to experience a ≥25% decline in the RF level. The association of treatment response was even greater (OR 6.2, 95% CI 1.5–26.4) when we modeled a ≥50% decline in the level of RF (data not shown). When the analysis was repeated after stratifying patients by treatment response, we found no significant evidence of an interaction with disease duration (data not shown). The removal of patients with influential data points for RF (n = 2 for the 25% cut point and n = 3 for the 50% cut point) did not change our results (data not shown).


In this study, we observed declines in the serologic levels of circulating autoantibodies among RA patients, a majority of whom received active therapy with DMARDs. Although autoantibody levels only rarely normalized with treatment, one-half of patients experienced a 25% or greater reduction in the level of anti-CCP antibody and RF. Levels of both anti-CCP antibody and RF decreased, but the primary factors predicting these serologic changes were specific to each autoantibody. Although we observed no association of treatment response with changes in anti-CCP levels, we found that RA patients with disease duration of <1 year were substantially more likely than those with more established disease to experience a significant decrease in the anti-CCP antibody level. Conversely, treatment response (not disease duration) was the primary determinant of treatment-related declines in the RF level.

Two recent studies investigated the impact of infliximab therapy (tumor necrosis factor α inhibition) plus MTX on the levels of RF and anti-CCP antibodies in RA (10, 11). Although observing no significant overall effect on median anti-CCP antibody levels, de Rycke and colleagues described a subset of patients (23 of 38) who, with treatment, experienced a marked (>20%) decline in the anti-CCP antibody level (11). Also, consistent with our findings, Bobbio-Pallavicini and coworkers observed significant declines in the levels of both RF and anti-CCP antibody with treatment intervention, although the autoantibody levels normalized in just 7% of patients (10). Moreover, those investigators found that changes in RF, but not anti-CCP antibody, roughly paralleled measures of disease activity. Our data, coupled with other recent findings, suggest that RF and anti-CCP antibody represent 2 distinct autoantibody systems in RA.

An accumulating body of evidence suggests a pathogenic role for anti-CCP antibodies in RA. In contrast to anti-CCP antibody production, synovial expression of the CCP antigen does not appear to be specific to RA. Moreover, the appearance of anti-CCP antibody in the circulation (along with RF) often precedes disease onset in RA by several years. In a study involving serial measurements in blood donors, Nielen and associates found that approximately one-half of patients with RA had positive IgM-RF and/or anti-CCP antibody on at least 1 occasion, a median of almost 5 years prior to disease onset (4). Additionally, the presence of anti-CCP antibody in early disease is highly predictive of more rapid radiographic disease progression, a clinical hallmark of aggressive RA (21, 22).

Given the central role of anti-CCP antibody in RA, our findings may have important implications. The success of clinical interventions in lowering anti-CCP antibody levels (and thus likely minimizing the pathogenic effects of autoantibody binding) appears to be dependent on disease duration. The association of shorter disease duration with greater declines in anti-CCP antibody levels is highly consistent with the growing body of evidence that shows improved clinical outcomes with earlier disease intervention in RA (23–25).

Alternatively, it is possible that these findings may simply reflect the natural history of autoantibody production over the course of the disease. In well-established disease, anti-CCP antibody levels may peak early in the disease process, irrespective of therapeutic interventions, before declining and reaching a nadir. It is possible that this early disease period, characterized by the stimulation and activation of B cells and relatively high levels of circulating anti-CCP antibody, represents an important therapeutic window in RA. Because the therapies used in these clinical protocols are not known to affect B cell function, our observations suggest that selection of these autoantibodies may be diminished with the treatment of synovial inflammation.

Results of studies examining the association between a decline in RF levels and treatment response have been inconsistent. In a small study involving only 18 patients with RA, RF levels fell to 30% of pretreatment values after 1 year of treatment with D-penicillamine (9). However, decreases in the RF level showed no correlation with changes in either the erythrocyte sedimentation rate or the joint score (single composite score based on degree of joint swelling/tenderness). Because that study predated the availability of either the ACR or the Paulus criteria, no assessment of the association between decreases in the RF level and composite measures of disease improvement was performed. In a study involving the use of either intramuscular gold or D-penicillamine, the decline in the level of IgM-RF among patients with substantial clinical improvement (>55% using a modified articular index) was compared with RF decreases observed in those without evidence of improvement. Significant reductions in the level of IgM- RF (P < 0.001) were seen only in those treated with gold and experiencing significant clinical improvement (7). These observations suggest that a decrease in the RF level may be dependent not only on treatment response but also on the specific therapeutic regimen used. Limited by the small size of the individual treatment groups, we observed no differential impact on a decline in the RF level (or a decline in anti-CCP antibodies) based on the specific therapy received.

In addition to the small size of the individual treatment groups and resulting patient heterogeneity, our study has other limitations. Due to limited specimen availability, we examined followup serum samples that were obtained at variable time points during protocol followup. However, each sample was linked with comprehensive clinical data collected in a double-blind manner that allowed a dichotomous treatment response to be clearly defined at the same time point at which the samples were obtained. Although it would be desirable to examine treatment response as both a continuous and a dichotomous variable, there were insufficient data on all patients at the time points of interest to calculate a composite continuous outcome (i.e., a disease activity score). It is also possible that there are unmeasured confounders that might impact autoantibody production. For instance, cigarette smoking, a health behavior that may fluctuate during the course of a clinical protocol, is associated with RF status (26). However, comprehensive data on the smoking behaviors of the study participants were not routinely collected. Two of the 3 studies involved in this post hoc analysis included RA patients with early disease who were seropositive for RF at the time of enrollment. Therefore, these results may not be generalizable to patients with seronegative disease or to patients receiving other DMARD regimens.

It is unknown whether decrements in the levels of circulating autoantibodies translate directly into improvements in long-term outcomes in RA. The lack of an association between treatment response and a decline in the level of anti-CCP antibodies would, on the surface, suggest that no clinical benefit is attributable to decreases in anti-CCP antibody levels. It is possible, however, that anti-CCP antibody level declines may predict improvement only over longer periods of followup or improvements in radiographic or other outcome measures that were not assessed in the present study. As further studies contribute added insight into the pathogenic role of circulating autoantibodies in RA, there will be an increasing need for comprehensive investigations designed to assess and compare the impact of specific therapeutic interventions on anti-CCP antibody and RF production.


The authors greatly appreciate the work of the many Rheumatoid Arthritis Investigational Network (RAIN) investigators and wish to thank the RAIN patients who made this study possible.