Patient selection
Of the 107 patients, nine were excluded because of a lack of available data, and 98 patients were finally analyzed. During the observational period (median: 56.0 months, IQR: 36.0–76.0 months), 12 patients completely discontinued CZP treatment for the following reasons: patient preference, three; injection site reaction, two; angina pectoris, two; urinary tract infection, one; peritonsillar abscess, one; malignant lymphoma, one; depression, one; and drug rash, one. The median duration until withdrawal occurred was 17.5 (IQR: 5.3–45.5) months after CZP initiation. After excluding the 12 patients from the total analysis population, the effectiveness analysis population was identified (n = 86), of which 45 (52.3%) and 41 (47.7%) patients were divided into continuation and discontinuation groups, respectively. During the median follow-up duration of 56.0 (IQR: 36.0–76.0) months, 17 (41.5%) and 24 (58.5%) patients discontinued CZP because of primary and secondary failures, respectively.
Baseline demographics and clinical features
As shown in Table 1, majority of patients were females (87.8%). Median age at CZP initiation and disease duration were 56.0 (42.5–64.0) and 5.0 (2.0–8.2) years, respectively. Seventy-three (74.5%) patients had positive ACPA with median titers of 115.0 (44.6–500.0) U/mL, and 74 (75.5%) patients had positive RF with median titers of 137.7 (50.8–334.5) IU/mL. MTX was simultaneously administered to 60 (61.2%) patients at a median dose of 8.0 (8.0–10.0) mg/week. Frequency and median dose of concomitant PSL were 43.9% and 5.0 (2.6–7.8) of mg/day, respectively. Initial loading was administered in 73 (74.5%) patients. CZP was introduced as 1-, 2-, and ≥ 3-line biologics in 43 (43.9%), 31 (31.6%), and 24 (24.5%) patients, respectively. Distribution of b/tsDMARDs before CZP use was TNFi in 48 (50.0%) patients, interleukin-6 receptor inhibitor (IL-6Ri) in 18 (18.4%), T-cell co-stimulation modulator (abatacept) in 12 (12.2%), and Jak kinase inhibitor (JAKi) in 2 (2.0%). Median level of CRP was 1.2 (0.4–3.0) mg/dL. Median tender and swollen joint counts were 3.0 (2.0–6.0) and 3.0 (1.0–6.0), respectively. Additionally, patients who discontinued CZP for reasons other than ineffectiveness had a significantly higher proportion of b/tsDMARD failure than the effective analysis population (58.3% vs. 27.9%; P = 0.047); however, other baseline demographics and clinical features did not statistically differ between the two groups.
Table 1
Baseline demographics and clinical characteristics of 98 patients with rheumatoid arthritis who were administered CZP
| Total analysis population (n = 98) | Others¶ (n = 12) | Effective analysis population (n = 86) | p value† |
Continuation (n = 45) | Discontinuation (n = 41) |
Age at CZP initiation, years | 56.0 (42.5–64.0) | 62.0 (54.3–70.0) | 56.0 (41.0–61.5) | 55.0 (40.0–65.5) | 0.39 |
Female, n (%) | 86 (87.8%) | 11 (91.7%) | 39 (86.7%) | 36 (87.8%) | > 0.99 |
Disease duration, years | 5.0 (2.0–8.2) | 2.2 (1.8–6.9) | 4.0 (1.7–6.8) | 5.0 (2.0–8.8) | 0.34 |
ACPA positive, n (%) | 73 (74.5%) | 11 (91.7%) | 32 (71.1%) | 30 (73.2%) | > 0.99 |
ACPA titer, U/mL | 115.0 (44.6–500.0) | 78.7 (46.2–115.0) | 105.6 (35.7–428.9) | 241.2 (49.4–668.9) | 0.21 |
RF positive, n (%) | 74 (75.5%) | 9 (75.0%) | 33 (73.3%) | 32 (78.1%) | 0.14 |
RF titer, IU/mL | 137.7 (50.8-334.5) | 109.4 (30.0-608.0) | 84.0 (43.0–207.0) | 206.6 (95.3–439.5) | 0.004 |
MTX use, n (%) | 60 (61.2%) | 7 (58.3%) | 32 (71.1%) | 21 (51.2%) | 0.077 |
MTX dose, mg/week | 8.0 (8.0–10.0) | 8.0 (6.0–10.0) | 10.0 (8.0–12.0) | 8.0 (7.0–10.0) | 0.22 |
PSL use, n (%) | 43 (43.9%) | 4 (33.3%) | 19 (42.2%) | 20 (48.8%) | 0.67 |
PSL dose, mg/day | 5.0 (2.6–7.8) | 5.8 (2.1–15.4) | 5.0 (2.5–5.0) | 5.5 (3.1–8.0) | 0.17 |
Initial loading, n (%) | 73 (74.5%) | 9 (75.0%) | 37 (82.2%) | 27 (65.9%) | 0.091 |
Number of b/tsDMARDs used prior to CZP | | | | | |
0, n (%) | 43 (43.9%) | 3 (25.0%) | 28 (62.2%) | 12 (29.3%) | 0.003 |
1, n (%) | 31 (31.6%) | 7 (58.3%) | 11 (24.4%) | 13 (31.7%) | 0.48 |
≥ 2, n (%) | 24 (24.5%) | 2 (16.7%) | 6 (13.3%) | 16 (39.0%) | 0.012 |
Class of b/tsDMARDs used prior to CZP | | | | | |
TNFi, n (%) | 48 (50.0%) | 7 (58.3%) | 14 (31.1%) | 27 (65.9%) | 0.002 |
IL-6Ri, n (%) | 18 (18.4%) | 1 (8.3%) | 4 (8.9%) | 13 (31.7%) | 0.013 |
CTLA4-Ig, n (%) | 12 (12.2%) | 2 (16.7%) | 3 (6.7%) | 7 (17.1%) | 0.18 |
JAKi, n (%) | 2 (2.0%) | 0 (0.0%) | 0 (0.0%) | 2 (4.9%) | 0.22 |
CRP, mg/dL | 1.2 (0.4–3.0) | 2.7 (0.5–4.7) | 0.91 (0.4–3.2) | 1.4 (0.2–2.9) | 0.54 |
Tender joint count | 3.0 (2.0–6.0) | 2.0 (2.0–5.3) | 3.0 (1.0–5.8) | 4.0 (1.5–7.5) | 0.11 |
Swollen joint count | 3.0 (1.0–6.0) | 2.0 (1.0–5.3) | 3.0 (1.0–6.0) | 3.0 (2.0–8.0) | 0.12 |
Data are presented as median (interquartile range) or number (%). |
¶Others include patients who discontinued CZP due to reasons other than non-response to CZP. |
†The continuation group versus the discontinuation group by the Mann-Whitney U test, the chi-square test or the Fisher’s exact test, as appropriate. |
ACPA, anti-citrullinated protein antibody; CRP, C-reactive protein; CTLA4-Ig, cytotoxic T lymphocyte-associated antigen-4-Ig; CZP, certolizumab pegol; b/tsDMARDs, biological/targeted synthetic disease-modifying antirheumatic drugs; IL-6Ri, interleukin-6 receptor inhibitor; JAKi, Jak kinase inhibitor; MTX, methotrexate; PSL, prednisolone; RF, rheumatoid factor; TNFi, tumor necrosis factor inhibitor. |
Comparison between patients who continued or discontinued CZP due to ineffectiveness
We assessed differences in clinical parameters between patients continuing and discontinuing CZP due to lack of effectiveness (Table 1). The continuation group had significantly lower RF titer than the discontinuation group (84.0 [IQR: 43.0–207.0] vs. 206.6 [IQR: 95.3–439.5], P = 0.004), although no significant difference in the rate of RF positivity was observed between the two groups (73.3% vs. 78.1%, P = 0.14). The proportion of b/tsDMARDs-naïve population was significantly higher in the continuation group than that in the discontinuation group (62.2% vs. 29.3%, P = 0.003). Moreover, the continuation group was significantly lower in patients with history of ≥ 2 b/tsDMARDs, TNFi, and IL-6Ri than those in the withdrawal group (13.3% vs. 39.0%, P = 0.012; 31.1% vs. 65.9%, P = 0.002; 8.9% vs. 31.7%, P = 0.013, respectively). Although there was a trend towards a higher prevalence of MTX use (71.1% vs. 51.2%, P = .077) and initial loading (82.2% vs. 65.9%, P = .091) in the continuation group than that in the withdrawal group, the differences were not statistically significant.
Drug survivals in the overall population
Drug survival rates at 12, 24, 48, and 60 months in the total analysis population (n = 98) were 65.0%, 55.8%, 46.4%, and 40.2%, respectively, which were comparable to those in the effective analysis population (n = 86), excluding the 12 patients who withdrew CZP for reasons other than non-response (66.0% at 12, 59.3% at 24, 51.9% at 48, and 47.2% at 60 months) (Supplementary Fig. 2). Next, we analyzed the retention rates of CZP among groups divided according to different clinical features possibly associated with its effectiveness [12, 13, 25–27] in the effective analysis population. Drug survivals with and without MTX at 12, 24, and 60 months were 73.4% vs. 54.0%, 71.4% vs. 38.9%, and 55.6% vs. 34.6%, respectively (HR: 0.51, 95% CI: 0.27–0.98, P = 0.024) (Fig. 1a). Retention rates in patients with MTX of ≥ 8 mg/week were significantly longer than those in MTX of < 8 mg/week (73.5% vs. 58.8% at 12 months, 71.0% vs. 47.8% at 24 months, and 58.9% vs. 34.8% at 60 months, respectively) (HR: 0.53, 95% CI: 0.28–0.99, P = 0.047) (Fig. 1b). Patients receiving initial loading showed significantly lower discontinuation rates at 12, 24, and 60 months (71.5%, 64.4%, and 53.6%, respectively) than those who did not (50.0%, 44.4%, and 28.6%, respectively) (HR: 0.43, 95% CI: 0.20–0.92, P = 0.030) (Fig. 1c). Furthermore, retention rates in the b/tsDMARD-naïve groups were significantly higher than those in the b/tsDMARD-switch groups, with 79.1% and 54.3% at 12 months, 72.9% and 47.4% at 24 months, and 66.3% and 32.2% at 60 months, respectively, (HR: 0.39, 95% CI: 0.21–0.72, P = 0.003) (Fig. 1d).
Predictive factors for effectiveness associated with long-term continuation of CZP
The HR with 95% CI of the baseline variables when administering CZP was calculated using the Cox regression analysis to detect independent factors associated with CZP persistence attributable to its effectiveness in the effective analysis population. As shown in Table 2, the univariate analysis revealed that concomitant MTX use (HR: 0.50, 95% CI: 0.27–0.94, P = 0.031), MTX dose (HR: 0.92, 95% CI: 0.86–0.98, P = 0.013), initial loading (HR: 0.50, 95% CI: 0.26–0.98, P = 0.045), and no b/tsDMARDs exposure (HR: 0.38, 95% CI: 0.19–0.73, P = 0.003) were statistically significant variables associated with CZP retention. Because a negative correlation between the history and class of b/tsDMARDs was observed, the class of b/tsDMARDs was not applied to the multivariate analysis, considering collinearity. The multivariate model adjusted with ACPA and RF titer, concomitant MTX use, MTX dose, and initial loading demonstrated b/tsDMARD-naive as the most influential factor related to CZP discontinuation due to ineffectiveness (HR: 0.46, 95% CI: 0.22–0.90, P = 0.024). Next, the Cox regression analysis was performed to explore the predictive factors for effectiveness leading to CZP continuation in each group with or without a b/tsDMARD-naïve background. The multivariate analysis identified that IL-6Ri exposure was most associated with CZP discontinuation in the b/tsDMARD-switched population (HR: 2.46, 95% CI: 1.11–5.57, P = 0.027) (Supplementary Table 1). Furthermore, in the b/tsDMARD-naïve patients, multivariate analysis revealed that RF titer was the strongest predictor of CZP ineffectiveness (HR per 100 IU/mL increment: 1.25, 95% CI: 1.06–1.60, P = 0.009) (Supplementary Table 2).
Table 2
Predictive factors for ineffectiveness associated with CZP discontinuation in the effective analysis population
| Univariate | Multivariate |
| HR (95% CI) | p value | HR (95% CI) | p value |
Age at CZP initiation (years) | 1.01 (0.99–1.04) | 0.29 | | |
Female | 1.10 (0.47–3.21) | 0.83 | | |
Disease duration (years) | 1.01 (0.96–1.04) | 0.77 | | |
ACPA positivity | 1.09 (0.56–2.28) | 0.79 | | |
ACPA titer (×102 U/mL) | 1.06 (0.991–1.12) | 0.085 | 1.03 (0.96–1.09) | 0.39 |
RF positivity | 1.73 (0.82–4.26) | 0.16 | | |
RF titer (×102 IU/mL) | 1.06 (0.995–1.11) | 0.067 | 1.04 (0.97–1.10) | 0.27 |
MTX use | 0.50 (0.27–0.94) | 0.031 | 0.94 (0.23–3.46) | 0.93 |
MTX dose (mg/week) | 0.92 (0.86–0.98) | 0.013 | 0.95 (0.82–1.10) | 0.50 |
PSL use | 1.03 (0.56–1.92) | 0.90 | | |
PSL dose (mg/day) | 1.04 (0.96–1.11) | 0.33 | | |
Initial loading | 0.50 (0.26–0.98) | 0.045 | 0.77 (0.38–1.58) | 0.47 |
b/tsDMARDs naïve | 0.38 (0.19–0.73) | 0.003 | 0.46 (0.22–0.90) | 0.024 |
CRP (mg/dL) | 1.12 (0.97–1.27) | 0.11 | | |
Tender joint count | 1.06 (0.98–1.14) | 0.12 | | |
Swollen joint count | 1.06 (0.97–1.13) | 0.16 | | |
ACPA, anti-citrullinated protein antibody; b/tsDMARDs, biological/targeted synthetic disease-modifying antirheumatic drugs; CI, confidence interval; CRP, C-reactive protein; CZP, certolizumab pegol; HR, hazard ratio; MTX, methotrexate; PSL, prednisolone; RF, rheumatoid factor. |
Supplementary Table 1. Predictive factors for ineffectiveness leading to CZP discontinuation in patients with previous b/tsDMARD exposure |
Lower retention rate of CZP in b/tsDMARD-switch patients with previous IL-6Ri therapy
Because IL-6Ri exposure was the strongest contributor to CZP discontinuation in patients who failed first-line b/tsDMARDs, retention rates were stratified according to IL-6Ri to assess the extent to which it could influence CZP effectiveness. Patients with previous IL-6Ri treatment had significantly higher discontinuation rate compared to those without IL-6Ri (HR: 2.40, 95% CI: 1.04–5.57, P = 0.041). Median retention durations in the IL-6Ri and non-IL-6Ri groups were 5 and 39 months, respectively (Fig. 2a). Among the 46 patients treated with b/tsDMARDs, the proportion of history of ≥ 2 b/tsDMARDs was markedly significantly higher in the IL-6Ri group than that in the non-IL-6Ri group (15/17, 88.2% vs. 7/29, 24.1%; P < 0.0001) (Supplementary Table 3); however, among 22 patients with a history of ≥ 2 b/tsDMARDs, no significant differences in the discontinuation rates were observed between the IL-6Ri and non-IL-6Ri groups (11/15, 73.3% vs. 5/7, 71.4%; P > 0.99), indicating no correlation between increasing treatment line and CZP effectiveness in those populations.
Comparison of drug survivals by an optimal cut-off value of RF in the b/tsDMARDs-naive population
As the RF titer was an outstanding determinant of CZP effectiveness in patients taking b/tsDMARDs, an ROC curve analysis was performed to determine the optimal cut-off value of the RF titer demonstrating the highest sensitivity and specificity for predicting long-term CZP effectiveness. The cut-off value was determined to be 79.9 IU/mL (area under the curve, 0.76; P = 0.003). Finally, we calculated retention rates, stratified according to RF titer of 79.9 IU/mL. As shown in Fig. 2b, ≥ 79.9 IU/mL groups demonstrated significantly lower retention rate of CZP compared with < 79.9 IU/mL groups (HR: 6.09, 95% CI: 1.95–19.0, P = 0.007) in the b/tsDMARDs-naïve population. The retention rates in ≥ 79.9 vs. <79.9 IU/mL groups at 12, 24, and 60 months were 63.2% vs. 95.2%, 57.4% vs. 88.9%, and 44.7% vs. 88.9%, respectively. There were no statistical differences except for positivity and RF titer between patients with ≥ 79.9 and < 79.9 IU/mL (Supplementary Table 4).