Characteristics of Patients with RA Receiving TNF Inhibitors by RF Quartile

A total of 1010 patients with RA receiving TNFi recruited from the FIRST registry were included in this study. The quartiles of RF levels at the time of TNFi initiation were calculated, and the patients were divided into four groups: RF < 18.65 IU/mL (Q1), 18.65 ≤ RF < 54.85 IU/mL (Q2), 54.85 ≤ RF < 136.45 IU/mL (Q3), and RF ≥ 136.45 IU/mL (Q4) (Fig. 1). The patient demographics of all individuals and those treated with each TNFi by RF quartiles (Q1–Q4) are presented in Table 1 and Supplementary Table S1. More than half of the patients were bio-naïve, and the mean dose of concomitant MTX was 12–13 mg/week across all groups. Women of childbearing age (WoCBA, defined as female patients aged 18–45 years) constituted 16.6% (168/1010) of the overall patient population. The Q4 group, which had the highest RF levels, included a greater proportion of males and exhibited a longer disease duration compared with the other groups. Additionally, the Q4 group demonstrated higher CDAI, SDAI, CRP, and ESR values, as well as a higher number of patients with high disease activity relative to the other groups.

Fig. 1

Study diagram. ADA adalimumab, CTLA4-Ig cytotoxic T lymphocyte-associated antigen-4 immunoglobulin, CZP certolizumab pegol, ETN etanercept, GLM golimumab, IL-6Ri anti-interleukin-6 receptor antibody, IFX infliximab, JAKi Janus kinase inhibitor, MTX methotrexate, TNFi tumor necrosis factor inhibitor

Table 1 Baseline characteristics of all patients and each quartileComparison of Efficacy of TNF Inhibitors by RF Quartile

The efficacy of TNFi was compared across four RF quartiles.

The persistence rate at 26 weeks following the initiation of TNFi was observed to be lower in the Q4 group compared with the Q1–Q3 groups (the rates for Q1, Q2, Q3, and Q4 were 85.3%, 88.1%, 84.6%, and 79.8%, respectively; p = 0.0650; Supplementary Fig. S1A). The predominant reason for discontinuation of TNFi within 26 weeks of treatment was poor response across all groups (Q1–Q4), particularly in the Q4 group. The second most common reason for discontinuation was adverse events (Supplementary Table S2).

The SDAI remission rate at 26 weeks following the initiation of TNFi was the lowest in the Q4 group, with rates for Q1, Q2, Q3, and Q4 recorded at 32.5%, 41.5%, 34.4%, and 27.4%, respectively (Fig. 2a, left panel). Additionally, there was a trend indicating that the SDAI remission rate decreased as the RF level increased, as determined by the Cochran–Armitage trend test (p = 0.0450). A comparison of the SDAI remission rates for Q4 and the combined Q1–Q3 at 26 weeks revealed that the SDAI remission rate for the Q4 group was significantly lower than that for the combined Q1–Q3 group (27.4% vs. 36.1%, respectively; p = 0.0109) (Fig. 2a, right panel). In bio-naïve patients (n = 727), the SDAI remission rate at 26 weeks was also the lowest in the Q4 group, with rates for Q1, Q2, Q3, and Q4 reported as 36.5%, 43.2%, 37.0%, and 31.3%, respectively. When comparing the SDAI remission rates in the Q4 group and the combined Q1–Q3 group, the SDAI remission rate for the Q4 group was lower than that for the Q1–Q3 combined group, although this difference was marginally non-significant (31.3% vs. 38.9%, p = 0.0776) (Supplementary Fig. S2).

Fig. 2

Comparison of efficacy according to RF quartile in overall patients treated with TNFi. a SDAI remission rates at 26 weeks by RF quartile (left) and Q1–3 versus Q4 (right). b Multivariate logistic regression for TNF inhibitors showing associations with SDAI remission. Each TNF inhibitors are adjusted for sex, disease duration, Steinbrocker’s classification (stage II or below), MTX dose, glucocorticoid dose, Bio-naïve, SDAI, RF, and Anti-CCP. c SDAI remission rates at 26 weeks for CZP (right) and Fc-containing TNFi (left). ADA adalimumab, CCP cyclic citrullinated peptide, CZP certolizumab pegol, ETN etanercept, GLM golimumab, IFX infliximab, MTX methotrexate, RF rheumatoid factor, SDAI Simplified Disease Activity Index, TNFi tumor necrosis factor inhibitor

These results suggest that the efficacy of TNFi may be lower in the Q4 group compared with the other groups with lower RF levels. To identify the most effective TNFi in Q4, the odds ratio (OR) for achieving SDAI remission at 26 weeks for each TNFi was calculated using multivariate logistic regression analysis. The results of the univariate analysis are presented in Supplementary Table S3. In the multivariate analysis, results were adjusted for sex, disease duration, non-articular deformity (Steinbrocker’s classification of stage II or lower), dose of concomitant MTX, dose of concomitant GC, previously used bDMARDs, SDAI, RF level, and anti-CCP antibody level, as these factors are associated with clinical remission. Consequently, in Q4, only the introduction of CZP significantly contributed to achieving SDAI remission at 26 weeks (OR 2.03, 95% confidence interval [CI] 1.10–3.76, Fig. 2b).

On the basis of the aforementioned results, we compared the SDAI remission rates between the Q1–Q3 and Q4 groups in patients who received Fc-containing TNFi and those who received CZP, respectively. There was no significant difference in the proportion of WoCBA patients between the CZP and Fc-containing TNFi groups (18.8% [82/436] vs. 15.0% [86/574], p = 0.1059). Among these patients, the rate of MTX discontinuation at 26 weeks was significantly higher in the CZP group compared with the Fc-containing TNFi group (15.9% [13/82] vs. 1.2% [1/86], p = 0.0006). In the Fc-containing TNFi group, the SDAI remission rate for the Q4 cohort was significantly lower, with remission rates of 35.3% for combined Q1–Q3 and 21.7% for Q4 (p = 0.0025; refer to Fig. 2c, left panel). Conversely, in the CZP group, there was no significant difference in SDAI remission rates between the combined Q1–Q3 and Q4 cohorts (Fig. 2c, right panel). Furthermore, when we examined the time course of disease activity from the initiation of CZP to week 26 across the Q1–Q4 groups, there were no substantial changes in disease activity over this period (see Supplementary Fig. S3).

Comparison of Efficacy Between CZP and ADA by RF Quartile in Patients with RA

As described above, CZP may have consistent efficacy regardless of RF level at baseline.

In clinical trials involving patients with RA, ADA has been used as an active comparator [30,31,32]. In this study, it served as the control, and the efficacies of CZP and ADA were compared within each RF quartile.

Patients with RA treated with CZP (CZP group, n = 436) and those treated with ADA (ADA group, n = 285) were stratified into four groups (Q1–Q4) based on RF quartiles derived from the overall cohort (Supplementary Table S4). In both the CZP and ADA groups, higher RF levels were correlated with increased disease activity, elevated serological markers, and a greater number of patients exhibiting progressive joint destruction (stage III or higher according to the Steinbrocker’s classification).

The persistence rates at 26 weeks for Q1, Q2, Q3, and Q4 were 83.5%, 87.2%, 86.6%, and 78.1%, respectively, in the ADA group (p = 0.4526), and 83.9%, 88.6%, 85.5%, and 79.8% in the CZP group (p = 0.3562). These results indicate no significant differences between RF quartile in both the CZP and ADA groups (Supplementary Fig. S4A). Additionally, there were no differences in adverse events among RF quartile in either the ADA or CZP group (Supplementary Table S5). From baseline to 26 weeks, the MTX dose (mg/week) in each RF quartile was as follows: In the CZP group, a significant reduction was observed in all quartiles: Q1, 12.0 ± 3.5 to 10.6 ± 4.9 (p = 0.0007); Q2, 12.7 ± 3.4 to 10.5 ± 4.8 (p < 0.0001); Q3, 12.8 ± 3.4 to 11.4 ± 4.5 (p = 0.0004); Q4, 12.0 ± 3.5 to 10.4 ± 5.4 (p = 0.0003). In the ADA group, the MTX dosage significantly decreased in all quartiles except Q3: Q1, 13.7 ± 2.8 to 13.0 ± 3.3 (p = 0.0118); Q2, 13.8 ± 2.8 to 12.8 ± 3.8 (p = 0.0013); Q3, 12.7 ± 3.5 to 12.4 ± 3.7 (p = 0.0702); Q4, 12.9 ± 3.5 to 11.7 ± 4.3 (p = 0.0209). The SDAI remission rates at 26 weeks in the ADA group for Q1, Q2, Q3, and Q4 were 36.5%, 50.0%, 35.8%, and 20.0%, respectively (p = 0.0057), demonstrating that the remission rate was lower in patients with high RF (Q4). Conversely, in the CZP group, the rates for Q1, Q2, Q3, and Q4 were 33.0%, 44.8%, 34.6%, and 34.9%, respectively (p = 0.2655), indicating no significant differences among RF quartiles (Supplementary Fig. S4B).

Furthermore, because anti‑CCP antibody titers increased with higher RF levels (see Table 1), we performed a multivariate logistic regression analysis to examine the association between baseline anti‑CCP antibody titers and achieving SDAI remission at 26 weeks. The explanatory variables for the analysis were sex, disease duration, Steinbrocker’s classification, concomitant MTX dose, concomitant GC dose, prior use of biologics, SDAI, and RF level; no significant association was detected. We then divided the cohort into quartiles based on baseline anti‑CCP antibody titers: < 5.0 U/mL (Q1′), 5.0 to < 65.8 U/mL (Q2′), 65.8 to < 318.3 U/mL (Q3′), and ≥ 318.3 U/mL (Q4′), and compared SDAI remission rates at 26 weeks among these groups. The remission rates in Q1′ through Q4′ were 33.7%, 32.1%, 34.8%, and 35.2%, respectively, with no significant differences between quartiles.

Comparison of Efficacy Between CZP and ADA by RF Quartile After Adjustment by PS-Based IPTW

The efficacy of CZP in patients with high RF (Q4) was found to be non-inferior to that in patients with lower RF, indicating that its efficacy is not affected by RF levels. Conversely, the efficacy of ADA appears to be affected by RF levels. In other words, CZP may be more effective than ADA in patients with high RF. However, a direct comparison was not feasible owing to substantial differences in patient background characteristics between the two groups, which resulted from drug selection bias. Therefore, the comparison of efficacy between the two groups was conducted by minimizing selection bias through the application of PS-based IPTW.

Table 2 presents the background variables for the two groups after adjustment using PS-based IPTW. Following this adjustment, no significant differences were observed in any of the patient background factors between the two groups, and the standardized difference was < 0.1, indicating a good balance of the variables.

Table 2 Characteristics of patients in adjusted data using PS-based IPTW

Although there was no significant difference in persistence rates between the two groups, a tendency toward lower persistence rates was observed in Q4 for both the ADA and CZP groups (Supplementary Fig. S5). When comparing the MTX dosage at 26 weeks across RF quartile, the results were as follows: Q1, 10.9 ± 5.0 mg/week for the CZP vs. 12.3 ± 3.2 mg/week for the ADA group (p = 0.0007); Q2, 10.9 ± 4.9 vs. 12.1 ± 4.0 mg/week (p = 0.0096); Q3, 11.3 ± 4.4 vs. 12.0 ± 3.9 mg/week (p = 0.1075); and Q4, 10.5 ± 5.4 vs. 11.4 ± 4.3 mg/week (p = 0.1094). The dosage was significantly lower in the CZP group than in the ADA group in Q1 and Q2, while no significant differences were observed between the groups in Q3 and Q4. The primary reason for study discontinuation was a poor response, while the secondary reason was adverse events, applicable across all quartile groups (Q1–Q4). Furthermore, there were no differences in the incidence of discontinuation due to poor response or adverse events between the two groups across any of the quartile groups (Q1–Q4) (Table 3).

Table 3 Reason for discontinuation of TNF inhibitors in adjusted data using PS-based IPTW

In the Q2 group, the SDAI at 26 weeks was significantly lower in the ADA group compared with the CZP group (CZP vs. ADA = 6.6 ± 7.2 vs. 4.5 ± 5.6, p = 0.0032). Conversely, in the Q4 group, the SDAI was significantly lower in the CZP group than in the ADA group (CZP vs. ADA = 6.8 ± 8.2 vs. 9.1 ± 12.2, p = 0.0259; see Fig. 3a). In the Q1 and Q3 groups, there was no difference in the SDAI at 26 weeks between the CZP and ADA groups. In the Q4 group, the SDAI remission rate at 26 weeks after the initiation of TNFi was significantly higher in the CZP group than in the ADA group (CZP vs. ADA = 36.6% (59/161) vs. 24.5% (40/163), p = 0.0172; see Fig. 3b). Actually, when changes in disease activity were analyzed up to 26 weeks in both groups, disease activity tended to improve more rapidly in the CZP group than in the ADA group, indicating a higher efficacy of CZP (Fig. 3c). In contrast, there were no significant differences in the SDAI remission rate between these two groups in the Q1–Q3 groups.

Fig. 3

Comparison of efficacy between ADA and CZP after adjustment by PS-based IPTW. a SDAI over 26 weeks for ADA and CZP by RF quartile. b SDAI remission rates at 26 weeks by RF quartile. c SDAI categories over 26 weeks after the introduction of ADA and CZP in the Q4 group. (Left panel) ADA, (Right panel) CZP. Numbers represent the proportion of all patients. Non responder imputation is not applied to SDAI remission rates in c. ADA adalimumab, CZP certolizumab pegol, HDA high disease activity, LDA low disease activity, MDA moderate disease activity, PS-based IPTW propensity score-based inverse probability of treatment weighting, REM remission, RF rheumatoid factor, SDAI Simplified Disease Activity Index