Baseline autoantibody profile in rheumatoid arthritis is associated with early treatment response but not long-term outcomes

The Induction therapy with Methotrexate and Prednisone in Rheumatoid Or Very Early arthritic Disease (IMPROVED) study is a multicentre, randomized controlled trial that enrolled 610 patients with early (< 2 years) untreated RA or undifferentiated arthritis. It was aimed at change in the disease activity score-remission (DAS44 < 1.6), and for those achieving remission, aimed at drug-free remission (DFR), with treatment adjusted every 4 months according to whether treatment targets had been reached. Initial treatment comprised methotrexate (MTX) and high-dose prednisone, followed by either tapering of medication or randomization to one of two treatment arms: MTX, prednisone, hydroxychloroquine, and sulphasalazine combination (multi-DMARD arm) or MTX and adalimumab combination as described previously [2]. According to the protocol, patients tapered and discontinued methotrexate at 8 months if they achieved early remission, allowing them to become drug-free and remain so until the DAS increased to > 1.6.

All 479 patients fulfilling the 2010 American College of Rheumatology (ACR)/European League Against Rheumatism (EULAR) RA criteria were selected. Of these patients, those seropositive at baseline by routine clinical testing for anti-CCP2 IgG or RF IgM, or by our in-house assay for anti‐CarP IgG [10], were selected (n = 395; see Fig. 1 for detailed selection algorithm). If patients were fully seronegative at baseline, we measured anti-CCP2 IgG, RF IgM, and anti-CarP IgG in serum collected after 1 year of follow up to include any patients that seroconverted to positive, yielding 399 seropositive patients, of whom 356 had baseline, untreated serum available and 209 had 8-month treated serum available for further serological measurements as described subsequently [2].

Main outcomes we investigated were initial DAS change from baseline to 4 months (ΔDAS 0–4 months) and DFR. DAS change from baseline to 4 months occurred under treatment with MTX and high-dose prednisone. DFR was defined as the ability to discontinue medication and remain in remission for (at least) 1 year after achieving DAS44 < 1.6. We differentiated between initial DFR and long-term sustained DFR. Initial DFR was defined as DFR between 1 and 2 years of the study, which due to the protocol could only be achieved by patients that were in early DAS remission at 4 months after tapering prednisone and MTX. Long-term sustained DFR was defined as DFR of at least 1 year duration until the last follow up in all patients (including those who were randomized to the multi-DMARD or adalimumab treatment arm), which is the closest approximation of disease cure currently available for RA. Due to the protocol design, the group of patients that could achieve initial DFR was smaller than (i.e. a subgroup of) all patients who could achieve long-term sustained DFR.

Enzyme-linked immunosorbent assay (ELISA) was used essentially as described previously to measure anti-CCP2 IgG, IgM, and IgA [22], RF IgM and IgA [12], and anti-CarP IgG, IgM, and IgA [10, 23] in baseline serum from untreated patients. We also conducted fine-specificity ELISA for IgG directed against four citrullinated peptides: citrullinated-vimentin 59-74, citrullinated-fibrinogen β 36-52 and α 27-43, and citrullinated-enolase 5-20 [24]. Last, ELISA for anti-acetylated lysine and anti-acetylated ornithine IgG (Orgentec Diagnostika GmbH, Germany) was performed as previously described [11].

Absorbance was converted to arbitrary units per millilitre (aU/mL) using a standard curve of pooled, serially diluted highly positive patient serum. Samples were considered positive if they fell above the cutoff of the mean aU/mL value plus two standard deviations in serum samples from 76 healthy controls from the Leiden area, run in tandem with the samples for each ELISA. Because antibodies may be aspecifically directed to the unmodified variant of the peptide/protein of interest, we applied a specificity criterion to each ELISA. For anti-CCP2 IgA and IgM, the difference between the citrullinated and unmodified optical density (OD) had to be more than 0.1; for anti-CarP and anti-acetylated peptide ELISAs, the difference (aU/mL) between the modified and unmodified signal had to be above the cut-off. Since previous experiments revealed minimal aspecific signals for the citrullinated fine-specificity ELISA, no specificity criterion was applied.

The technical success rate of the ELISA was at least 96% (Fig. 1). There was good agreement in positivity between the original baseline measurement performed routinely for anti-CCP-antibodies and RF at inclusion and the in-house baseline re-measurement (Additional file 1: Figure S1).The first and second in-house measurement of anti-CarP IgG showed fair agreement (Additional file 1: Figure S1). Positivity for the various isotypes measured largely overlapped (Additional file 1: Figure S2).

We constructed categories reflecting the breadth of the antibody response that consisted of the sum of positive antibody tests. First, the number of isotypes present; both in total (anti-CCP2 IgG, IgM, IgA; RF IgM, IgA; anti-CarP IgG, IgM, IgA; range 1-8) and per family (anti-CCP2 and anti-CarP range 1-3; RF range 1–2). Second, the number of IgG anti-modified peptide responses, both in total (anti-CCP2 IgG, anti-CarP IgG, anti-citrullinated-vimentin 59-74 IgG, anti-citrullinated-fibrinogen β 36-52 IgG, and α 27-43 IgG, anti-citrullinated-enolase 5-20 IgG, anti-acetylated-lysine IgG, anti-acetylated-ornithine IgG; range 1–8) and per modification (citrullinated peptides range 1–4; acetylated peptides range 1–2). Differences between categories were calculated using analysis of variance (ANOVA) for continuous outcomes (DAS and ΔDAS 0–4 months), adjusted for age, gender, and smoking (ever/never), and baseline body mass index and Health Assessment Questionnaire (HAQ) score, which were independent predictors of early remission in the IMPROVED study [5]. Binary logistic regression was used for categorical outcomes, adjusted as above, with the analyses of initial DFR additionally adjusted for baseline DAS and the analyses of long-term sustained DFR additionally adjusted for baseline DAS and treatment arm. Holmes-Bonferroni methods were used to correct the alpha level for multiple testing, assuming the same number of hypotheses as pairwise comparisons made. All reported p values are derived from the analysis models following correction; only p values that remained significant after correction for multiple testing are reported in the figures.

At baseline in the full cohort, 68% (323/472) of patients were anti-CCP2 IgG positive, 70% (330/469) were RF IgM positive, and 39% (162/416) were anti-CarP IgG positive. Within the patients that were positive for at least one of these autoantibodies at baseline or at 1 year (n = 399), we (re)measured anti-CCP2, RF, and anti-CarP isotypes and anti-citrullinated and anti-acetylated peptide antibodies in baseline serum and in 8-month serum. Since we selected patients based on baseline seropositivity of anti-CCP2 IgG, RF IgM, or anti-CarP IgG, the high rates of positivity for these antibodies at baseline and 8 months are to be expected (Table 1). The lower rates of antibody positivity at 8 months compared to baseline are largely due to seroconversion from positive to negative in this time period.

We first analysed the association between the patients’ baseline autoantibody profiles and initial treatment response. As shown in Fig. 2a, seropositive patients (defined by the presence of anti-CCP2 IgG and/or RF IgM and/or anti-CarP IgG in the original baseline measurement) had a lower DAS at baseline than triple-negative patients. This was most likely due to the ACR/EULAR2010 RA criteria selection we used; seropositive patients require fewer other clinical items to fulfil the criteria and thus have a lower DAS at baseline than seronegative patients. Notably, despite these differences in absolute DAS, the initial change in DAS from baseline to 4 months was equal between seropositive and seronegative patients (Fig. 2b and Additional file 1: Figure S3A), also after correction for relevant covariates.

Strikingly, within the seropositive patients, initial DAS response in patients with many isotypes was more pronounced than in those with few isotypes (ΔDAS 0–4 months of 7–8 isotypes vs 1–2, 3–4, and 5–6 isotypes, respectively: -2.2 vs -1.5 (p < 0.001), -1.7 (p = 0.003), and -1.8 (p = 0.001)) (Fig. 2c, d and Additional file 1: Figure S3B). This pattern remained when analysing the number of isotypes present separately for each antibody family: those with more isotypes had better initial DAS response than those with fewer isotypes, and were statistically significant (after correction for multiple testing) for the RF and anti-CarP families (Fig. 2g).

There was the same dose-dependent association between breadth of the autoantibody profile and DAS decline when analysing the overall number of AMPAs present. Initial DAS response in seropositive patients with many AMPAs was better than in those with few AMPAs, although this was not always statistically significant after correction for multiple testing: ΔDAS 0–4 months of 7–8 AMPAs vs 1–2, 3–4, and 5–6 AMPAs, respectively: -2.1 vs -1.7 (p = 0.016), -1.5 (p < 0.001), and -1.9 (p =0.22) (Fig. 2e, f and Additional file 1: Figure S3C). This pattern was also present when analysing the number of antibodies present per post-translational modification, and was significant for citrullinated peptides: ΔDAS 0–4 months of 3–4 vs 1–2 citrullinated peptides was -2.0 vs -1.6 (p = 0.01) (Fig. 2g). No single isotype or antibody was disproportionally associated with a better initial DAS response (Additional file 1: Figure S4A).

To investigate whether the autoantibody profile at baseline or at the moment of tapering was also relevant for more long-term treatment outcomes, we next examined whether the autoantibody response is associated with ability to discontinue medication and remain in remission for one year after reaching early remission (initial DFR), independently of factors also associated with this outcome (see “Statistical analysis”). In line with previous findings [2], patients with RA who were positive for anti-CCP2 IgG and/or RF IgM were less likely than their negative counterparts to reach initial DFR, although this difference was not significant: 17% of anti-CCP2 IgG positive versus 20% of negative patients (p = 0.14) and 16% of RF IgM positive versus 19% of negative patients (p = 0.43). Anti-CarP IgG positive patients were also less likely to reach initial DFR than negative patients (11% vs 22%; p = 0.03). Since we selected patients on ACR/EULAR2010 RA criteria, thereby enriching for seropositivity, the differences between anti-CCP2 IgG, RF IgM, and anti-CarP positive and negative patients found here were less pronounced than previously reported in the entire IMPROVED study population because in the current study patients negative for one of these antibodies were by definition positive for another [2].

Interestingly, while a broad baseline autoantibody response was favourable for initial DAS response, it was unfavourable for the chance of achieving initial DFR (Fig. 3). Within patients seropositive for anti-CCP2 IgG, RF IgM, or anti-CarP IgG at baseline, there was a non-significant trend indicating that patients with more isotypes achieve less initial DFR (1–2, 3–4, and 5–6 isotypes vs 7–8 isotypes, respectively: 21% (p = 0.07), 20% (p = 0.13), and 20% (p = 0.10) vs 3%) (Fig. 3a). Patients with more AMPAs also achieved significantly less initial DFR (1–2 AMPAs vs 3–4, 5–6, and 7–8 AMPAs, respectively: 37% vs 13% (p = 0.004), 14% (p = 0.007), and 11% (p = 0.005) (Fig. 3b).

This trend remained when examining the number of isotypes present separately for each antibody family and the number of antibodies present against citrullinated/acetylated peptides (Fig. 3c). The presence of an anti-CCP2 IgM and/or IgA isotype within patients positive for anti-CCP2 IgG did not decrease the chance of initial DFR, nor did the presence of RF IgA in patients positive for conventional RF IgM (data not shown). Presence of a specific isotype or antibody did not confer increased or decreased chance of reaching initial DFR (Additional file 1: Figure S4B).

We investigated whether the autoantibody profile at the moment of tapering had a similar association with initial DFR as the baseline profile. Patients received tapered methotrexate at 8 months if they achieved early remission, allowing them to reach initial DFR between 12 months and 2 years. Seropositive patients with more isotypes at 8 months (i.e. the moment of tapering) tended to achieve less initial DFR than those with few isotypes, but this effect was not as clear as observed in relation to the baseline profile (Fig. 4a). Patients with more AMPAs at 8 months achieved slightly less initial DFR (1–2 AMPAs vs 3–4, 5–6, and 7–8 AMPAs, respectively: 38% vs 24% (p = 0.025), 10% (p = 0.004), and 13% (p = 0.023) (Fig. 4b)), but only the comparison of 1–2 AMPAs with 5–6 AMPAs remained significant after correction for multiple testing. When examining the antibody families separately, there was no clear pattern indicating that more isotypes or reactivity against citrullinated/acetylated peptides was associated with less initial DFR (Fig. 4c).

Finally, we wished to determine whether the baseline autoantibody profile was associated with the most favourable long-term outcome of long-term sustained DFR. Fifty-seven percent of patients that had initial DFR also achieved long-term sustained DFR, defined as at least 1 year of DFR lasting until the last follow up, an outcome approximating disease cure. For patients that were not in early remission at 4 months and therefore could not achieve initial DFR, it was still possible to taper medication at a later stage and reach long-term sustained DFR. In the full RA cohort, baseline anti-CCP2 IgG positive patients reached this outcome significantly less often than their negative counterparts (10% vs 26% (p < 0.001)); RF IgM and anti-CarP IgG positive patients followed a similar trend (14% vs 19% (p = 0.05).

In contrast to the previous results on initial DFR (broad baseline autoantibody response: decreased chance of initial DFR), there was no difference in rates of long-term sustained DFR between seropositive patients with many isotypes versus few isotypes, or between patients with many AMPAs versus few AMPAs (Fig. 5a, b). Furthermore, when separately assessing the number of isotypes in a family and the number of antibodies present against citrullinated/acetylated peptides, there were also no differences in long-term sustained DFR rates (Fig. 5c). Only anti-CarP isotypes showed a trend that was no longer present after correction for multiple testing: one isotype versus two isotypes and three isotypes, respectively: 20% vs 12% (p = 0.49) and 5% (p = 0.02). The presence of an anti-CCP2 IgM and/or IgA isotype within patients positive for anti-CCP2 IgG did not decrease the chance of long-term sustained DFR, nor did the presence of a RF IgA in patients positive for conventional RF IgM (data not shown). Last, positivity to a specific isotype or antibody did not confer increased or decreased chances of reaching long-term sustained DFR (Additional file 1: Figure S4C).

It is conceivable that patients who achieved early remission had different chances of reaching long-term sustained DFR than the full IMPROVED population examined here, and that results would have been different for the patients who achieved early remission. To investigate this, we performed sensitivity analysis of the association between baseline antibody profile and long-term sustained DFR in only the patients who achieved early remission. The results were the same in this group as in the whole cohort (Additional file 1: Figure S5).