Baseline characteristics
Baseline clinical characteristics of the bDMARDs-naïve cases are shown in Table 1. Overall, mean age was 57.0 years, 81.8% of participants were female, mean disease duration was 7.3 years, RF positivity was 78.6%, ACPA positivity was 81.4%, mean DAS28-ESR score was 4.4, mean CDAI was 17.8, and mean HAQ-DI score was 1.0. Mean doses and ratio of concomitant medications were PSL 6.3 mg/day (39.6%) and MTX 8.6 mg/week (65.4%).
Baseline clinical characteristics of the bDMARDs-switched cases are shown in Table 2. Overall, mean age was 58.1 years, 83.3% of participants were female, mean disease duration was 10.5 years, RF positivity was 78.1%, ACPA positivity was 83.4%, mean DAS28-ESR score was 4.2, mean CDAI was 15.7, and mean HAQ-DI score was 1.1. Mean doses and ratio of concomitant medications were PSL 5.7 mg/day (49.3%) and MTX 8.3 mg/week (57.1%). The bDMARDs were administered as the second agent in 59.5% of patients, and as the third or latter agent in 40.5% of patients.
Drug retention and causes for discontinuation
Cause-specific cumulative discontinuation rates were assessed using Gray’s test, and statistically compared using Fine-Gray hazard competing risk regression model at 36 months (Figs. 1-4 and Supplementary Fig. 1).
Drug discontinuation rates due to lack of effectiveness in the bDMARDs-naïve cases were as follows (Fig. 1a): ABT (13.7%), GLM (16.1%), TCZ (16.6%), ADA (20.6%), IFX (21.8%), ETN (22.4%), and CZP (26.9%) (P<0.001). These rates in the bDMARDs-switched cases were as follows (Fig. 1b): TCZ (18.9%), TOF (22.8%), ABT (28.7%), ETN (30.3%), GLM (33.3%), ADA (38.4%), IFX (39.4%), and CZP (46.1%) (P<0.001).
Drug discontinuation rates due to toxic adverse events in the bDMARDs-naïve cases were as follows (Fig. 2a): ABT (4.6%), CZP (5.5%), TCZ (6.8%), ADA (7.9%), GLM (9.3%), IFX (9.7%), and ETN (11.2%) (P<0.001). These rates in the bDMARDs-switched cases were as follows (Fig. 2b): ETN (5.0%), CZP (6.6%), ABT (9.2%), GLM (9.9%), IFX (12.2%), TCZ (13.2%), ADA (14.3%), and TOF (15.7%) (P=0.004).
Drug discontinuation rates due to remission in the bDMARDs-naïve cases were as follows (Fig. 3a): ETN (2.9%), ABT (4.0%), CZP (8.4%), GLM (9.0%), ADA (9.8%), TCZ (9.8%), and IFX (10.0%) (P<0.001). These rates in the bDMARDs-switched cases were as follows (Fig. 3b): CZP (1.1%), TCZ (1.2%), ABT (1.4%), ADA (2.1%), TOF (2.3%), ETN (2.5%), IFX (2.8%), and GLM (3.3%) (P=0.9).
Drug discontinuation rates due to non-toxic events in the bDMARDs-naïve cases were as follows (Supplementary Fig. 1a): CZP (3.8%), IFX (9.0%), ABT (10.8%), TCZ (11.4%), ADA (12.3%), ETN (13.5%), and GLM (17.2%) (P=0.07). These rates in the bDMARDs-switched cases were as follows (Supplementary Fig. 1b): CZP (4.1%), GLM (7.0%), ETN (7.4%), TOF (7.7%), IFX (8.5%), TCZ (9.4%), ABT (11.3%), and ADA (14.6%) (P=0.5).
Finally, drug discontinuation rates due to all adverse events (including lack of effectiveness and toxic adverse events) in the bDMARDs-naïve cases were as follows (Fig. 4a): ABT (18.3%), TCZ (23.5%), GLM (25.3%), ADA (28.4%), IFX (31.5%), CZP (32.4%), and ETN (33.6%) (P<0.001). These rates in the bDMARDs-switched cases were as follows (Fig. 4b): TCZ (32.1%), ETN (35.2%), ABT (37.9%), TOF (38.5%), GLM (43.2%), IFX (51.6%), ADA (52.7%), and CZP (52.7%) (P<0.001).
Hazard ratios (HRs) and 95% confidence intervals (CI) for discontinuation due to each specific cause were calculated using Fine-Gray hazard competing risk regression model adjusted for confounders (Tables 3 and 4).
In the bDMARDs-naïve cases (Table 3), HRs for discontinuation due to lack of effectiveness were significantly higher with CZP (HR=2.4, P<0.001), ETN (HR=1.7, P<0.01), and IFX (HR=1.5, P<0.05) compared with ABT (P<0.001 between agents). In terms of toxic adverse events, ADA (HR=2.8, P<0.001), ETN (HR=4.0, P<0.001), GLM (HR=2.5, P<0.01), IFX (HR=4.3, P<0.001), and TCZ (HR=2.2, P<0.05) showed a significantly higher rate compared with ABT (P<0.001 between agents). HR for discontinuation due to non-toxic reasons was significantly lower with CZP (HR=0.3, P<0.05) compared with ABT, although no significant difference was observed between agents (P=0.07). HRs for discontinuation due to remission were significantly higher with ADA (HR=2.9, P<0.001), GLM (HR=2.4, P<0.05), IFX (HR=3.1, P<0.001), and TCZ (HR=2.5, P<0.01) compared with ABT (P<0.001 between agents). Finally, HRs for all adverse events (including lack of effectiveness and toxic adverse events) were significantly higher with ADA (HR=1.8, P<0.001), CZP (HR=2.5, P<0.001), ETN (HR=2.3, P<0.001), GLM (HR=1.5, P<0.05), IFX (HR=2.1, P<0.001), and TCZ (HR=1.4, P<0.05) compared with ABT (P<0.001 between agents).
In the bDMARDs-switched cases (Table 4), HRs for discontinuation due to lack of effectiveness were significantly higher with CZP (HR=1.5, P<0.05), although significantly lower with TCZ (HR=0.6, P<0.001) compared with ABT (P<0.001 between agents). As for all toxic adverse events, ETN (HR=0.4, P<0.05) showed a significantly lower rate compared with ABT (P=0.004 between agents). There were no significant differences in HRs for discontinuation due to non-toxic reasons (P=0.5) and remission (P=0.9) between agents. Finally, HRs for all adverse events (including lack of effectiveness and toxic adverse events) were significantly higher with ADA (HR=2.7, P<0.001), CZP (HR=2.2, P<0.01), and IFX (HR=2.0, P<0.05) compared with ABT (P<0.001 between agents).