Clinical response among golimumab-treated Japanese patients with rheumatoid arthritis by number of previous biologic therapies: Real-world evidence from post-hoc analysis of post-marketing surveillance data

Abstract

Objectives

To assess the real-world effectiveness of golimumab in Japanese patients with rheumatoid arthritis who had previously received one or more biologic therapies.

Methods

A post-hoc analysis of post-marketing surveillance was performed. The clinical response to golimumab was analyzed in 1216 patients who had previously received one or more biologic agents including non-TNF inhibitors with stratification by the number of previous biologic agents. Logistic regression analyses were conducted to identify factors associated with DAS28-CRP response to golimumab.

Results

While treatment persistence is comparable, the response to golimumab declined with an increasing number of previous biologic therapies. When stratified by golimumab dose, patients receiving golimumab at 100 mg had higher disease activity at baseline with an increasing number of previous bDMARDs, but they still achieved comparable disease activity at 24 weeks regardless of how many bDMARDs had been previously used. Univariate and multivariate analyses both identified concomitant oral corticosteroid therapy as a factor negatively associated with the likelihood of achieving a DAS28-CRP response.

Conclusion

Switching to golimumab was effective regardless of how many biologic agents had been previously used, but the response declined with an increasing number of prior biologic agents. A golimumab dose of 100 mg was also effective for those who previously received three or more bDMARDs.

Keywords:

• Disease-modifying antirheumatic drugs
• golimumab
• post-marketing surveillance
• rheumatoid arthritis
• switching

Introduction

The advent of biologic disease-modifying antirheumatic drugs (bDMARDs) heralded a new treatment paradigm in rheumatoid arthritis (RA) [1–3]. In fact, the success in increasing the number of bDMARDs available to treat RA has provided expanding therapeutic options, which is particularly important for a longstanding disease like RA that is likely to involve a sequence of therapies. The use of bDMARDs is generally reserved for the treatment of patients who discontinued conventional synthetic disease-modifying antirheumatic drugs (csDMARDs) because of a lack of efficacy or intolerance [2]. However, regardless of the biological target, all bDMARDs currently available in clinical practice demonstrate comparable clinical efficacy, with a proportion of patients remaining refractory to treatment [4–6]. Thus, the question emerges as to which therapy should be given as a next step when a particular treatment must be discontinued due to inadequate response or intolerance, and how much effect is expected from the new treatment. With regard to the legitimate therapeutic options, the majority of recent studies have suggested that switching to another bDMARD with a different mode of action is superior to a bDMARD with the same mode of action [1]. Some RCTs have addressed the question of the expected effect when switching biologic agents. In the GO-AFTER trial, golimumab (GLM), a monoclonal antibody targeting tumor necrosis factor (TNF), showed clinical efficacy in RA patients who had discontinued prior anti-TNF therapies [7], and the clinical response to GLM decreased as the number of prior TNF inhibitors increased [8]. A similar trend was observed in the rituximab REFLEX trial [9], and the tocilizumab ARRIVE trial [10]. In the RADIATE trial, tocilizumab plus methotrexate was demonstrated to be effective in achieving sustained clinical improvement in RA patients with inadequate response to TNF inhibitors, and patients responded irrespective of the number of failed therapies [11]. More data derived from observational studies [12,13], meta-analysis studies [14] as well as systematic reviews [15] have suggested that the probability of achieving a clinical response in patients switching to a second or third biologic therapy is lower than that in patients receiving their first biologic therapy.

Despite these findings, no evidence evaluating the benefit of switching to GLM has been available from the real-world clinical setting. Accordingly, this study was performed to evaluate the clinical benefit of switching to GLM in RA patients who had previously received one or more bDMARDs. This study used a post-hoc analysis of data from post-marketing surveillance conducted in Japan, and we stratified the patient population based on the number of prior bDMARD therapies. We also conducted the same line of analysis with stratification by first-line bDMARD therapy, which has been reported in a companion paper (Shimizu et al.).

Methods

Data source

Detailed methods of the post-marketing surveillance (PMS) have been described previously [16]. Briefly, 5154 patients with RA who were treated with golimumab (GLM) in Japan between September 2011 and May 2013 were monitored for 24 weeks at selected medical facilities throughout the country. Enrolled patients received subcutaneous GLM (50 mg or 100 mg) once every 4 weeks in an open-label manner by the attending physician as part of routine clinical practice. The study protocol, including ethical considerations, were assessed by the institutional review board of each study center and endorsed by the Japanese PMDA. This PMS was registered with the University Hospital Medical Information Network Clinical Trials Registry (UMIN-CTR, Identifier: UMIN000015895), was conducted in accordance with Japanese regulations (Ministry of Health, Labor and Welfare Ministerial Ordinance No. 171) on Good Post-marketing Study Practice, and was carried out by Janssen Pharmaceutical K.K.

Study design and patients

This study included patients who had previously received one or more biologic therapies, including infliximab (IFX), adalimumab (ADA), etanercept (ETN), certolizumab pegol (CZP), tocilizumab (TCZ), and abatacept (ABT). Eligible patients had moderate or high disease activity at study entry based on the Disease Activity Score 28 based on C-reactive protein (DAS28-CRP) score, and the reason for discontinuing the most recent biologic therapy must have been documented as lack of effectiveness by the attending physician. Because patients could have received more than one biologic agent, they could have discontinued each one for a different reason. The analysis was conducted with stratification by the number of previous bDMARDs.

Evaluation of effectiveness

The effectiveness of GLM was assessed by comparing the DAS28-CRP, Disease Activity Score 28 based on erythrocyte sedimentation rate (DAS28-ESR) score, Simplified Disease Activity Index (SDAI), and Clinical Disease Activity Index (CDAI) between baseline and 24 weeks. The EULAR responses (defined by the absolute change in DAS28-CRP and DAS28-ESR) and the remission rates based on DAS28-CRP, DAS28-ESR, SDAI, and CDAI were also assessed at 24 weeks. Cut-off values for disease activity (high/moderate/low, respectively) were defined for the DAS28-CRP and DAS28-ESR scores (5.1/3.2/2.6), SDAI score (26/11/3.3), and CDAI score (22/10/2.8). Remission was defined as a DAS28-CRP score <2.6, DAS28-ESR score <2.6, SDAI score ≤3.3, and CDAI score ≤2.8 [17]. Subjects were considered to show a good EULAR response if the DAS28 score decreased by >1.2 and low disease activity (DAS28 ≤ 3.2) was achieved by 24 weeks. A moderate EULAR response was defined as a decrease in the DAS28 score of >0.6 and ≤1.2, and low disease activity (DAS28 ≤ 3.2) at 24 weeks, a decrease in the DAS28 score of >0.6 and moderate disease activity (3.2< DAS28 ≤ 5.1) at 24 weeks, or a decrease in the DAS28 score of >1.2 and high disease activity (5.1< DAS28) at 24 weeks [18].

Statistical analysis

Data are expressed as the mean ± standard deviation (SD) for continuous variables and as proportions (%) for categorical variables unless otherwise described. The paired t-test was employed to compare changes of clinical parameters from baseline to Week 24, with statistical significance being defined as p < .05 (two-tailed). Correction for multiplicity due to the multiple testing procedure was done using the Bonferroni method or Tukey–Kramer method to control the type I error. Kaplan–Meier analysis was conducted to estimate treatment persistence, which was defined as the period from initiation of GLM to its discontinuation for any reason. Survival curves were compared with the log-rank test, using the most persistent subgroup as the reference, and log-rank p values were corrected by the Bonferroni method for multiple comparisons.

To evaluate the association between baseline patient characteristics and the effectiveness of GLM, univariate logistic regression analysis was performed followed by multivariate logistic regression analysis. First, univariate regression analysis was done to evaluate the association of baseline demographic and clinical factors associated with the likelihood of achieving DAS28-CRP response (good/moderate) after 24 weeks of GLM treatment. Variables tested in the univariate logistic regression analysis were also evaluated in multivariate models. In all analyses, p < .05 was considered statistically significant. We also calculated odds ratios (OR) with the corresponding 95% confidence intervals (CI) to describe associations between a good or moderate EULAR response and baseline patient characteristics. The same analyses were also performed to examine factors associated with a good EULAR response.

A modified intent-to-treat (ITT) analysis was employed for effectiveness assessment, with the ITT population including patients who had at least baseline data. If data were missing at 24 weeks, baseline values were carried forward for imputation. In addition, we conducted an observed case (OC) analysis of patients who had data at both baseline and 24 weeks, as a sensitivity analysis of clinical parameters to confirm the robustness of findings in the ITT analysis. All statistical analyses were performed by SAS version 9.3 software (SAS Institute Inc, 100 SAS Campus Dr, Cary, NC 27513.).

Results

Patient disposition

Of the 5154 patients initially enrolled in the PMS study, a total of 1216 patients had either high or moderate disease activity according to the baseline DAS28-CRP score, despite using one or more biologics prior to study entry, and these patients formed the ITT/effectiveness population. Among the 1216 patients in this population, approximately 60% had previously used one bDMARD (n = 734), about 20% had received two bDMARDs (n = 267), and about 20% had received three or more bDMARDs (n = 215) (Figure 1). As expected for the current study population (i.e. biologic-experienced), most patients had longstanding RA, with a mean disease duration of 12.6 ± 10.1 years. The mean disease activity scores in the total study population were as follows: DAS28-ESR, 5.25 ± 1.08; DAS28-CRP, 4.59 ± 0.94; SDAI, 25.91 ± 11.47; and CDAI, 23.95 ± 11.08. Based on the DAS28-CRP score, 27.3% were classified as having high disease activity, and 72.7% had moderate disease activity (Table 1). These proportions differed per subgroup stratified by the number of prior biologic therapies; there was a clear trend showing that the proportion of patients with high disease activity increases with an increasing number of previous treatments with bDMARDs (one prior biologic therapy, 25.1%; two prior biologic therapies, 29.4%; three or more biologic therapies, 32.1%). Consistent with this observation, the DAS28-CRP, DAS28-ESR, CDAI, and SDAI scores also increased with a higher number of previous bDMARD treatments (Table 1). We also noted that patients previously exposed to two biologics appeared to less frequently receive methotrexate (MTX) as a concomitant medication; instead, this subgroup used csDMARDs other than MTX more frequently than the other two subgroups (Table 1). Glucocorticoid was concomitantly administered more frequently with an increasing number of previous treatments with bDMARDs (Table 1). Approximately 80% of patients initiated GLM at a dose of 50 mg while the rest received 100 mg at baseline, and these proportions were comparable across the subgroups of patients evaluated (Table 1). Among patients receiving only one bDMARD, 13.1% had received non-TNF inhibitors, such as tocilizumab and abatacept, before GLM, and this proportion was approximately two-fold greater among those who had previously received two bDMARDs. Furthermore, most of those having experience with three or more bDMARDs had previously received non-TNF inhibitors (Table 1).

Figure 1. Flow diagram showing patient disposition. ITT: intent-to-treat; GLM: golimumab. *Patients could be allocated to more than one group with regard to reasons for exclusion.

Table 1. Patient demographics and clinical baseline characteristics of the ITT/effectiveness analysis set.

	Total n = 1216	# of previous biologics
		1 n = 734	2 n = 267	≥3 n = 215
Sex
Male/Female, %	17.4/82.6	17.2/82.8	19.5/80.5	15.3/84.7
Age (years)	63.0 ± 12.7	63.8 ± 12.5	61.7 ± 12.9	61.8 ± 12.7
Duration of RA (years)	12.6 ± 10.1	12.0 ± 10.2	13.0 ± 9.6	13.9 ± 10.2
Tender joint count (0–68)	7.0 ± 6.0	6.6 ± 5.7	7.2 ± 6.0	8.2 ± 6.7
Swollen joint count (0–66)	5.7 ± 4.9	5.4 ± 4.9	6.0 ± 4.9	6.3 ± 5.1
ESR (mm/h)	45.5 ± 30.2	44.5 ± 29.9	44.7 ± 31.5	49.4 ± 29.3
CRP (mg/dL)	1.92 ± 2.25	1.87 ± 2.23	1.83 ± 2.02	2.22 ± 2.58
Patient VAS	58.7 ± 22.0	57.9 ± 21.9	59.4 ± 22.7	60.3 ± 21.5
Physician VAS	52.4 ± 20.9	51.8 ± 21.3	51.6 ± 20.5	55.4 ± 19.9
DAS28-CRP	4.59 ± 0.94	4.52 ± 0.94	4.61 ± 0.96	4.80 ± 0.91
DAS28-ESR	5.25 ± 1.08	5.16 ± 1.07	5.24 ± 1.11	5.55 ± 1.05
SDAI	25.91 ± 11.47	25.02 ± 11.35	26.33 ± 11.46	28.41 ± 11.56
CDAI	23.95 ± 11.08	23.20 ± 10.99	24.27 ± 11.08	26.12 ± 11.14
Disease activity (DAS28-CRP), %
High/moderate	27.3/72.7	25.1/74.9	29.4/70.6	32.1/67.9
Disease activity (DAS28-ESR), %
High/moderate/No	52.2/45.5/2.3	48.4/49.0/2.6	53.4/45.2/1.4	62.4/35.4/2.2
Steinbrocker’s stage, %
I/II/III/IV	6.2/24.9/31.1/37.8	7.6/25.7/32.8/33.9	4.6/26.1/29.5/39.8	3.7/20.9/27.0/48.4
Steinbrocker’s class, %
I/II/III/IV	9.6/57.9/29.3/3.2	11.3/58.4/26.6/3.7	7.3/59.8/31.4/1.5	6.5/54.0/35.8/3.7
Comorbidities, %	70.8	67.3	77.2	74.9
Hepatic dysfunction	11.5	10.4	12.0	14.9
Renal impairment	6.3	6.5	6.7	5.1
Previous medication, %
Glucocorticoid	60.3	54.5	66.7	72.3
MTX	77.0	79.6	69.1	77.7
csDMARDs other than MTX	33.0	32.3	35.4	32.7
Concomitant medication, %
Glucocorticoid	61.0	55.4	66.9	72.8
MTX	79.1	80.8	71.3	83.2
csDMARDs other than MTX	33.3	30.6	39.8	34.7
Initial dose of GLM, %
50 mg	80.1	80.7	76.4	82.8
100 mg	19.9	19.3	23.6	17.2
Patients who had previously received non-TNFi bDMARDs, %	30.7	13.1	27.7	94.4

CDAI: Clinical Disease Activity Index; CRP: C-reactive protein; csDMARD: conventional synthetic disease-modifying antirheumatic drugs; DAS28-CRP: Disease Activity Score 28 based on C-reactive protein; DAS28-ESR: Disease Activity Score 28 based on erythrocyte sedimentation rate; ESR: erythrocyte sedimentation rate; MTX: methotrexate; RA: rheumatoid arthritis; SDAI: Simplified Disease Activity Index; VAS: visual analog scale.

Persistence with GLM stratified by the number of prior biologic agents

According to the Kaplan–Meier analysis, patients who had previously used one bDMARD had a persistence rate of 77.2%. The rate was comparable for patients who had used three or more bDMARDs, being 81.6% for patients who had previously received two bDMARDs, and 75.8% for those who previously received three or more bDMARDs (Figure 2). The Kaplan–Meier analysis did not detect any significant differences when the one prior bDMARD subgroup was compared with each of the other two subgroups (p = .2264 vs. two prior bDMARDs and p = 1.0000 vs. three or more prior bDMARDs, log-rank test with p values being adjusted by Bonferroni correction). Persistence with GLM therapy (time to discontinuation) was also similar among these subgroups, being 20.26 weeks for one prior biologic, 21.21 weeks for two prior biologics, and 20.54 weeks for three or more prior biologics.

Figure 2. Persistence with GLM treatment stratified by the number of prior bDMARDs. A Kaplan–Meier analysis was conducted to assess persistence with GLM treatment during the 24-week post-marketing surveillance period in relation to the number of previous bDMARDs (1, 2, or ≥3). Kaplan–Meier curves were compared with the log-rank test, using the subgroup who had previously received one bDMARD as a reference. Log-rank p values were adjusted for multiplicity by using Bonferroni correction. Descriptive statistics are presented in the table. GLM: golimumab.

Clinical effectiveness of GLM stratified by the number of prior biologic agents

In the overall study population, improvement of disease activity was observed at 24 weeks as indicated by the significant reduction of the DAS28-CRP, DAS28-ESR, SDAI and CDAI scores from baseline (Figure 3(a,d) and Supplemental Figure 1(a,d)); p < .001). When stratified by the number of prior biologic therapies, the improvement of clinical outcomes at 24 weeks was found to be less marked among patients who had previously used three or more bDMARDs (Figure 3(a,d) and Supplemental Figure 1(a,d)). This trend is well represented by the changes in DAS28-CRP and DAS28-ESR scores from baseline, in which patients previously exposed to three or more bDMARDs demonstrated significantly less improvement than those previously exposed to only one biologic therapy, as suggested by pairwise comparisons (p = .016 and p = .017, respectively using the Tukey–Kramer method) (Supplemental Table 1) while the changes in SDAI and CDAI scores demonstrated only numerical differences between these two subgroups, without being statistically significant (p = .0656 and p = .0609, respectively using the Tukey–Kramer method) (Supplemental Table 1).

Figure 3. Change in DAS28-CRP and SDAI scores from baseline to week 24 of GLM treatment stratified by the number of prior bDMARDs. Disease activity at baseline and 24 weeks was evaluated in each subgroup stratified by the number of prior bDMARDs, using DAS28-CRP (a–c), and SDAI (d–f) scores. Total study population (a, d) and subgroup analysis according to the dose of GLM: 50 mg (b, e) or 100  mg (c, f). DAS28-CRP: Disease Activity Score 28 based on C-reactive protein; SDAI: Simplified Disease Activity Index. ***p<.001 versus baseline by the paired t-test. †p<.001 versus baseline by the paired t-test with Bonferroni correction for multiplicity.

Consistent results were obtained when the proportion of patients achieving good/moderate EULAR response based on DAS28-CRP was analyzed; however, we noted a more apparent trend that the rates of response to GLM decrease with prior exposure to an increasing number of biologic agents (Figure 4(a)). Also, similar patterns were observed for remission rates based on the DAS28-CRP, DAS28-ESR, SDAI, and CDAI scores, demonstrating that the likelihood of achieving remission after 24 weeks of GLM treatment declines with an increasing number of previous treatments with biologic agents (Figure 4(b)). Patients with three or more prior bDMARDs showed particularly low remission rates, which were confirmed by pairwise statistical comparison (Supplemental Table 2).

Figure 4. Percentage of patients achieving EULAR response and remission. (a) Percentage of patients achieving a EULAR response at 24 weeks stratified by the number of previous bDMARDs. The clinical response to GLM at 24 weeks was assessed by the EULAR response criteria based on DAS28-CRP. A good or moderate response was defined as improvement of the DAS28 score by more than 1.2 from any baseline score or improvement of 0.6–1.2 from a baseline score ≤ 5.1, respectively. (b) Percentage of patients achieving remission as defined by DA28-CRP (score <2.6), DA28-ESR (score <2.6), CDAI (score ≤2.8), and SDAI (score ≤3.3). CDAI: Clinical Disease Activity Index; DAS28-CRP: Disease Activity Score 28 based on C-reactive protein; DAS28-ESR: Disease Activity Score 28 based on erythrocyte sedimentation rate; SDAI: Simplified Disease Activity Index.

Table 2. Univariate and multivariate logistic regression analysis to identify patient characteristic variables associated with the likelihood of achieving a good or moderate EULAR response.

Variables	Univariate analysis		Multivariate analysis
	Odds Ratio (95% CI)	p value	Odds Ratio (95% CI)	p value
Demographics
Age, years	1.00 (0.99, 1.01)	.5011	1.00 (0.99, 1.01)	.7983
Male, n (%)	0.91 (0.68, 1.23)	.5515	0.92 (0.66, 1.28)	.6304
Disease duration, years	1.00 (0.98, 1.01)	.5058	1.00 (0.98, 1.01)	.6719
Disease characteristics
DAS28-CRP	1.05 (0.93, 1.19)	.4015	1.08 (0.94, 1.23)	.2684
Comorbidities, n (%)	0.86 (0.67, 1.10)	.2380	0.97 (0.73, 1.28)	.8251
Concomitant medications
MTX	1.23 (0.95, 1.61)	.1190	1.16 (0.80, 1.68)	.4480
Oral corticosteroid	0.71 (0.57, 0.90)	.0040	0.75 (0.56, 0.99)	.0450
MTX + Oral corticosteroid	1.16 (0.78, 1.71)	.4689	1.25 (0.70, 2.23)	.4567
Number of previous biologic therapies
1	1.22 (0.97, 1.54)	.0959	(Reference)
2	0.92 (0.70, 1.21)	.5372	0.93 (0.68, 1.27)	.6558
≥3	0.80 (0.59, 1.08)	.1442	0.74 (0.52, 1.03)	.0757

DAS28-CRP: Disease Activity Score 28 based on C-reactive protein; EULAR: European League Against Rheumatism; MTX: methotrexate.

When stratified by the dose of GLM (50 mg or 100 mg), the overall improvement of clinical outcomes at 24 weeks was demonstrated to be comparable regardless of which dose of GLM had been used. However, among the patients who had previously received one biologic therapy, those who received GLM at 100 mg showed less prominent responses in DAS28-CRP, DAS28-ESR, SDAI and CDAI than those received GLM at 50 mg. In contrast, among the patients who had previously received two biologic therapies, those who received GLM at 100 mg showed better clinical responses than those received GLM at 50 mg. This trend is even more salient among those who had previously received three or more biologic therapies (Figure 3(b,c,e,f), Supplemental Figure 1(b,c,e,f)). We noted, however, that patients receiving GLM at 100 mg had higher disease activity at baseline with an increasing number of previous biologic therapies, but they still achieved comparable disease activity at 24 weeks regardless of how many bDMARDs had been previously used (Figure 3(b,c,e,f), Supplemental Figure 1(b,c,e,f)).

Taken together, the data indicated that, among patients who had previously received one or more biologic therapies, clinical signs and symptoms of RA were significantly improved after 24 weeks GLM treatment and clinical response is likely to decline with an increasing number of previous biologic therapies. This conclusion was also supported by our sensitivity analysis performed in the OC population, which provided similar results for all clinical measures investigated (Supplemental Table 3), suggesting the robustness of our findings. In addition, a GLM dose of 100 mg is more likely to be effective than 50 mg in patients with an increasing number of prior bDMARD therapies.

Baseline factors predictive of clinical response

Logistic regression analysis was used to identify baseline demographic and clinical factors associated with the likelihood of achieving a EULAR response (good/moderate) based on DAS28-CRP after 24 weeks of GLM treatment. Univariate analyses identified concomitant oral corticosteroid use as a significant baseline clinical factor associated with lower likelihood of achieving an EULAR response (OR = 0.71, p = .0040). Consistently, this factor was also identified with multivariate analyses (OR = 0.75, p = .0450) (Table 2).

In addition, we extended the same line of analysis to identify factors associated with the likelihood of achieving a good EULAR response (Supplemental Table 4). Among baseline demographics, disease characteristics and concomitant medications, univariate analyses identified disease duration (OR = 0.98, p = .0061), DAS28-CRP (OR = 0.86, p = .0328), presence of comorbidities (OR = 0.72, p = .0183), and concomitant use of oral glucocorticoid (OR = 0.66, p = .0016) as significant factors. In contrast, multivariate analyses identified disease duration (OR = 0.98, p = .0194) and concomitant use of oral glucocorticoid (OR = 0.66, p = 0.0109) as predictors associated with a lower likelihood of achieving a good EULAR response (Supplemental Table 4). Among previous biologic therapies, experience with one biologic therapy was identified as a predictor (OR = 1.38, p = .0188) significantly associated with higher odds of achieving a good EULAR response, whereas experience with three or more biologic therapies was identified as a predictor associated with lower likelihood of achieving a good EULAR response (OR = 0.67, p = .0290) by univariate analyses. Multivariate analyses failed to identify any predictors regarding the number of previously used bDMARDs associated with a good EULAR response compared with previous experience with one biologic agent (Supplemental Table 4).

Discussion

In the present study, we investigated the effectiveness of GLM in RA patients who had previously received one or more bDMARD therapies through stratification by the number of prior bDMARDs. Approximately 40% of the patients had received multiple bDMARDs before study entry: 734 patients (60.4%) had received only one prior bDMARD, whereas 267 patients (22.0%) had received two prior bDMARDs, and 215 patients (17.7%) three or more prior bDMARDs. This population was somewhat different from that of the GO-AFTER study as a higher proportion of patients had previously received three or more bDMARDs (17.7% vs. 9.3%) [8]. More remarkably, the GO-AFTER study only included patients with previous experience with TNF inhibitors [8], whereas approximately 30% of patients included in the current study had previously received non-TNF inhibitors including tocilizumab and abatacept, presumably better reflecting real-world practice. Nevertheless, our data are consistent with the post-hoc analysis of the GO-AFTER study, which showed that the response to GLM decreased as the number of previous bDMARDs increased [8]. Importantly, the current study demonstrated that this trend is consistent across several clinical parameters examined, including DAS28-CRP, DAS28-ESR, SDAI, and CDAI, further validating our findings. Our data are also in good agreement with observational studies [12,13], meta-analyses [14] as well as systematic reviews [15], suggesting that the possibility of achieving a clinical response declines following the first biologic therapy.

GLM is approved at two doses (i.e. 50 mg and 100 mg) for the treatment of active RA in Japan. In the previous GO-AFTER trial, among patients who discontinued prior anti-TNF therapy, those who received GLM at 100 mg appeared to show a numerically higher response than those who received GLM at 50 mg [7]. Additionally, our previous post-hoc analysis of PMS data suggested that physicians are more likely to select 100 mg of GLM for patients who failed previous bDMARD therapy in real-world daily practice [19]. However, there has been no evidence as to the effectiveness of GLM at 100 mg compared with 50 mg in RA patients stratified by the number of previous biologic agents. In the present study, it was found that 100 mg of GLM appears to be more effective than 50 mg in patients with an increasing number of previous biologic therapies. This trend appears to be due to the fact that patients receiving GLM at 100 mg had higher baseline disease activity with an increasing number of prior bDMARD therapies, while disease activity at 24 weeks was comparable irrespective of how many bDMARDs had been previously used. Although the reason behind this observation remains enigmatic, one plausible explanation is that previous failure of an anti-TNF agent had been caused by the lack of suppression of TNF due to inadequate dosing [20,21], in which case patients might have had better chance to overcome high levels of TNF with the increased dose of GLM. Another possibility is that patients previously treated with TNF inhibitors had developed antibodies to the prior agents that hindered continued response to those TNF inhibitors [22,23], and these antibodies might have cross-reacted with the human GLM molecule, requiring an increased dose of GLM to overcome this state. Future studies should investigate the serum concentrations of TNF and the presence of antibodies to prior TNF inhibitors to better understand the underlying mechanism of the dose-dependent response to GLM.

During the pivotal RCTs of GLM, its efficacy was not investigated in relation to concomitant treatment other than MTX [7,24,25]. In the present study, concomitant use of oral corticosteroids was associated with a lower response to GLM at 24 weeks. Our previous analysis of this PMS dataset demonstrated that disease activity was higher in patients who were using concomitant oral corticosteroids [19]. Therefore, concomitant use of oral corticosteroids is more likely to be a consequence rather than a cause of failure to control disease activity with GLM. Nevertheless, DAS28-CRP at baseline was not identified as an independent predictor by the present analysis, suggesting that patients with hard-to-control disease activity might not necessarily have had high disease activity at baseline and other factors might have collectively contributed. Interestingly, disease duration was identified as a negative predictor of clinical response when the association with a good EULAR response was analyzed, suggesting that concomitant use of oral glucocorticoid and disease duration might be jointly associated with clinical response to GLM independent of disease activity at baseline.

The present study is subject to several limitations as a retrospective analysis in the real-world clinical setting. First, there is a risk of selection bias due to the lack of blinding and randomization. Also, decisions about the therapeutic strategies, such as the selection of csDMARDs and the dose of GLM (50 mg or 100 mg), were all at the discretion of the attending physician. These variables could have impacted on the outcomes but were not considered in the current analysis, which potentially leads to a risk of confounders. Moreover, the follow-up period of the PMS was only 24 weeks, which could be too short to evaluate the effectiveness of switching to GLM because the progression of joint destruction in RA is long-term [26]. Finally, there may have been discrepancies between the timing of discontinuing the previous bDMARD and initiating GLM treatment, which may have affected baseline patient characteristics (i.e. at initiation of GLM), albeit we selected patients with moderate to high disease activity. Due to these limitations, the data presented in this manuscript should be interpreted with caution.

In conclusion, we demonstrated the effectiveness of GLM in RA patients who had previously received one or more bDMARDs in the real-world clinical setting. Switching to GLM was effective irrespective of how many biologic agents had been previously used, but the response declined with an increasing number of prior biologic therapies. This real-world evidence validates previous findings that switching to GLM is an appropriate treatment strategy for RA patients with an inadequate response to prior bDMARDs.

Author contributions

HS, HK and YI contributed to the conception and design of the study; HS wrote the manuscript; HK, MK and YI critically reviewed and edited the manuscript; all authors approved the final version of this manuscript for submission.

Acknowledgements

The authors would like to thank all of the patients and physicians who participated in this PMS for their cooperation. The authors are also grateful to Toshiro Yano (Ikuyaku. Integrated Value Development Division, Mitsubishi Tanabe Pharma Corporation) for study conception and design. Support with statistical analysis was provided by Hiroki Nakane (EPS Corporation) and professional medical writing and editorial assistance was provided by Yamada Translation Bureau, Inc., both of which were jointly funded by Janssen Pharmaceutical K.K. (Tokyo, Japan) and Mitsubishi Tanabe Pharma Corporation (Osaka, Japan). Sponsorship for this study and article processing charges were jointly provided by Janssen Pharmaceutical K.K. and Mitsubishi Tanabe Pharma Corporation.

Conflict of interest

Hirohito Shimizu, Hisanori Kobayashi, Masayoshi Kanbori, and Yutaka Ishii are employees of Janssen Pharmaceutical K.K., a wholly owned subsidiary of Johnson & Johnson.