Switching to biological DMARDs versus cycling among JAK inhibitors in patients with rheumatoid arthritis and with inadequate response to JAK inhibitors: from FIRST registry

Abstract

Objectives

This study aimed to identify characteristics of patients with rheumatoid arthritis (RA) with an inadequate response to Janus kinase inhibitors (JAKi-IR) and evaluate the efficacy and safety of subsequent treatments.

Methods

This study included 434 patients with RA who started JAKi treatment. JAKi-IR patients were those who switched to another drug due to inadequate response or did not reach low disease activity within 26 weeks of beginning JAKi. The efficacy and safety of switched biological disease-modifying anti-rheumatic drugs (bDMARDs) or cycled targeted synthetic disease-modifying anti-rheumatic drugs were analysed 26 weeks after switching treatment in JAKi-IR patients.

Results

Patients with JAKi-IR RA accounted for 31.8% (n=138/434). Multiple logistic regression identified factors contributing to JAKi-IR, such as the prior use of multiple ineffective bDMARDs and suboptimal JAKi dosing. There were no differences in patient background when comparing patients with RA with JAKi-IR who cycled to another JAKi (n=31) versus those who switched to bDMARDs (n=45). Among those cycling to another JAKi, the Clinical Disease Activity Index (CDAI) scores improved by week 26, with higher remission rates, while retention and adverse events remained similar. Trajectory analysis identified three CDAI response patterns, with the ‘treatment response’ group showing rapid and sustained improvement when cycling to another JAKi. Multiple logistic regression in this group identified another JAKi cycle as the critical factor for the treatment response.

Conclusions

Cycling JAKis is more effective than switching to bDMARDs in JAKi-IR RA, with no differences in safety or retention. This study suggests that cycling to another JAKi may be appropriate for patients with RA with JAKi-IR.

Correspondence to Professor Yoshiya Tanaka; [email protected]

WHAT IS ALREADY KNOWN ON THIS TOPIC

• A minority of patients with rheumatoid arthritis (RA) exhibit an inadequate response to Janus kinase inhibitors (JAKi-IR), and it has been suggested that cycling to another JAKi in these patients may result in a higher drug retention rate compared with switching to biological disease-modifying anti-rheumatic drugs (bDMARDs).

• However, the characteristics of patients with JAKi-IR RA and the most effective targeted therapy for these patients remain unclear.

WHAT THIS STUDY ADDS

• We found that patients with multiple previous ineffective bDMARDs, those unable to use optimal doses of JAKi and those with a high Health Assessment Questionnaire Disability Index at the time of JAKi initiation are more likely to become patients with JAKi-IR RA.

• Additionally, we demonstrated that cycling to another JAKi, rather than switching to bDMARDs, most effectively improves disease activity at 26 weeks in patients with JAKi-IR RA.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The results of this study could help establish more appropriate therapeutic strategies for patients with JAKi-IR RA.

Introduction

Rheumatoid arthritis (RA) is a systemic inflammatory disease that causes progressive bone and joint destruction and irreversible functional impairment.^{1 2} Since the advent of biological disease-modifying anti-rheumatic drugs (bDMARDs), the treatment of RA has undergone a paradigm shift in the last 20 years. However, bDMARDs have some disadvantages. Their high molecular weight limits the route of administration to injection. Additionally, they can be associated with secondary failure due to the development of anti-drug antibodies.³ Therefore, Janus kinase inhibitors (JAKis), which are low-molecular-weight compounds that can be orally administered, were developed. Currently, five types of JAKis are available for use in patients with RA. These inhibitors significantly improved disease activity in clinical trials.³ However, a subset of patients with RA demonstrated an inadequate response to JAKis (JAKis-IR) and failed to exhibit significant improvements in disease activity.⁴ Although several studies have focused on patients with JAKi-IR RA, they compared subsequent molecular-targeted drugs^5–7 in terms of retention rate. It is unknown which molecular-targeted drug is more effective in patients with RA with JAKi-IR. One of the major clinical questions in real-world settings is which molecular-targeted drugs should be subsequently selected for patients with RA with JAKi-IR.

This study aimed to determine the characteristics of patients with JAKi-IR and identify molecular-targeted drugs appropriate for these patients. Thus, we analysed the characteristics of patients with RA with JAKi-IR and the efficacy and safety of individual molecular-targeted drugs after switching to them in patients with JAKi-IR. In particular, growth mixture modelling (GMM), an analytical method for identifying trajectory group categories for factors that change over time,⁸ enables the characterisation of each trajectory group and analysis of factors affecting trajectories. GMM was used to analyse the trajectories of changes in disease activity after switching to other molecular-targeted drugs in patients with RA with JAKi-IR and to assess the clinical characteristics in terms of response to the switched drugs.

Material and method

Patient and study design

Based on the FIRST registry,^{9 10} a multicentre observational study that primarily included patients with RA initiating treatment with molecular-targeted anti-rheumatic drugs at our institution and several affiliated centres, 434 patients who initiated JAKi treatment (tofacitinib, baricitinib, peficitinib, upadacitinib and filgotinib) and were followed-up for 6 months between August 2013 and April 2022 were extracted to compare the efficacy and safety of various types of JAKis. The first JAKi (tofacitinib) became available for patients with RA in Japan in July 2013. This study includes all patients with RA who began treatment with JAKis between August 2013 and April 2022 and were registered in the FIRST registry, a multicentre collaborative study led by our department. RA was diagnosed in patients who met the 2010 American College of Rheumatology/European League Against Rheumatology or the 1987 American College of Rheumatology classification criteria.^{11 12} In Japan, JAKis are covered by health insurance for RA. The observational period was 26 weeks.

Treatment with biological and targeted synthetic DMARDs

bDMARDs and targeted synthetic disease-modifying anti-rheumatic drugs (DMARDs) were administered to patients with RA in whom disease activity had not been controlled, even after treatment with conventional synthetic DMARDs, including methotrexate at standard doses, or to patients with RA who could not receive conventional synthetic DMARDs, including methotrexate. The study patients were orally administered JAKis at the maximum tolerable dose for hepatic and renal functions. For patients with hepatic or renal dysfunction, JAKi was administered at reduced doses. The types and dosages of each administered drug are shown in online supplemental table S1. All patients were admitted to the University of Occupational and Environmental Health Hospital, Japan, to initiate JAKis treatment. JAKi treatment was initiated after haematological and coagulation function tests, faecal occult blood tests and whole-body CT screening¹³ were performed to assess the risk factors.¹⁴ Online supplemental table S2 shows the risk factors for the patients participating in this study. Endocrinologists assessed and treated patients with diabetes mellitus, hypertension, obesity and dyslipidaemia. Patients with a history of malignancy were evaluated by surgeons, who verified the absence of recurrence or any associated complications and confirmed the safety of JAKi administration. After providing patients with comprehensive explanations regarding the risks associated with JAKi use and presenting alternative therapy options, treatment with JAKis was initiated on obtaining informed consent.

All patients participating in this study were admitted to the University of Occupational and Environmental Health Hospital, and treatment was initiated with bDMARDs or targeted synthetic DMARDs (tsDMARDs). All b/tsDMARDs are presented in online supplemental table S1, and no bias was observed with respect to the use of b/tsDMARDs.

Clinical efficacy and outcome

In accordance with the 2022 European Alliance of Associations for Rheumatology recommendations,¹⁴ the primary endpoint was the remission rate at week 26 in each group, as assessed using the Clinical Disease Activity Index (CDAI).^{15 16} CDAI remission was defined as a score of less than or equal to 2.8, and low disease activity (LDA) was defined as a score of less than or equal to 10.0. Additional secondary endpoints included the evaluation of disease activity, patient retention rates and safety measures at week 26.

Sample size

The required sample size was estimated based on a threshold remission rate of 28%, an expected remission rate of 35%, a power of 80% and an alpha value of 0.05 (two-sided) using a binomial test based on previous observational studies. Considering the 20% of ineligible patients, a target sample size of at least 431 cases was determined.

Definition of inadequate response to JAK inhibitors

Patients with JAKi-IR RA were defined as those who switched to other bDMARDs or cycled to another JAKi because of inadequate response or those who had never achieved LDA (CDAI<10) during the 26 weeks after initiating JAKi treatment. Cases where patients experienced joint destruction progression or severe joint symptoms causing activities of daily living (ADL) impairment despite achieving LDA were identified as inadequate responses, as determined by the attending physician. This led to a switch from the JAKi to another therapeutic approach.

Propensity score-based inverse probability of treatment weighting

Propensity scores were calculated by performing multivariable logistic regression analysis with JAKi use as a dependent variable and independent variables including sex, age, disease duration, previously used bDMARDs, concomitant methotrexate doses, concomitant glucocorticoid use, tender joint count, swollen joint count, Patient’s Global Assessment of Disease Activity Visual Analogue Scale, Evaluator Global Assessment of Disease Activity Visual Analogue Scale, Pain Global Assessment of Disease Activity Visual Analogue Scale, Health Assessment Questionnaire Disability Index (HAQ-DI), the European Quality of Life 5 Dimension questionnaire, C-reactive protein (CRP), rheumatoid factor (RF), anti-cyclic citrullinated peptide antibody and matrix metalloproteinase 3 (MMP-3). The calculated propensity scores were weighted using a ratio of ‘the percentage of patients receiving JAKs in all patients/propensity score’ in the JAKi group and the ratio of ‘the percentage of patients receiving bDMARD in all patients/1-propensity score’ in the bDMARD group as a stabilised weighting coefficient to adjust for the differences in baseline background factors between the two groups. Likewise, to compare and analyse the three groups of patients who cycled to other JAKis or switched to tumour necrosis factor (TNF) inhibitors or non-TNF inhibitors, generalised propensity scores were calculated using a dependent variable of the other JAKis, TNF inhibitors or non-TNF inhibitors, respectively.

Safety

Clinical laboratory tests and other safety assessments were performed during these visits. The incidences and severities of all adverse events were recorded. The National Institutes of Health Common Terminology Criteria for Adverse Events (V.5.0) were used to describe adverse events and laboratory abnormalities.

Growth mixture modelling

GMM⁸ was performed using Stata V.16.0 (StataCorp, College Station, Texas, USA). The Bayesian information criterion (BIC) and Akaike information criterion (AIC) were used to compare the fit of the linear models with quadratic-to-quintic functions, and the models with a better fit were applied. We used the combination of the lowest absolute BIC and AIC values as the optimal model. The BIC and AIC were calculated for cases divided into two to seven groups, and the model with the best fit was used.⁸ The following independent variables were used to construct the GMM model: baseline CDAI; CDAI at 2, 4, 12 and 26 weeks after switching therapies; and remission status at 26 weeks (CDAI≤2.8). For model selection, we first identified the best-fitting trajectory shape (from linear to fifth-order polynomial) using BIC and AIC values, selecting the model with the lowest values. The number of trajectory groups (ranging from 2 to 10) was subsequently determined using the same criteria.

Other statistical analyses

Characteristics of the patients were presented as the mean±SD, median (IQR), or the count (percentage) of individuals. The Kaplan-Meier approach was used to evaluate retention rates, and the log-rank test was used to determine statistical differences. Student’s t-test, Mann-Whitney U test, or Bonferroni method was used for between-group comparisons, and Fisher’s exact test was used for categorical variables.

All reported p values were two-sided and not adjusted for multiple testing. Multiple imputation was also used to account for missing follow-up CDAI.

Statistical significance was set at p<0.05. All analyses were conducted using JMP V.15.0 (SAS Institute, Cary, North Carolina, USA) and SPSS software (V.25.0; SPSS, Chicago, Illinois, USA).

Results

Baseline patients’ background, safety and efficacy in patients with RA who initiated JAKis treatment

Online supplemental table S2 shows the baseline characteristics of 434 patients with RA who were registered in the FIRST registry, initiated JAKi treatment between July 2013 and October 2022, and were observed for 26 weeks or longer. The patients mainly comprised those who failed to respond to an average of approximately three bDMARDs and had high disease activity. Approximately 10% of patients were unable to receive JAKis at their optimal dosages due to issues such as hepatic or renal dysfunction and other related events.

The retention rate of JAKis was 78.8% (342/434 patients) (figure 1A). Online supplemental table S3 shows the adverse events that resulted in the discontinuation of JAKis. Three patients were diagnosed with malignancy (0.7%) (one with signet-ring cell carcinoma and two with malignant lymphoma), one had a major adverse cardiovascular event (0.2%) and 17 had herpes zoster (3.9%). Of the 67 patients who had been immunised against herpes zoster (HZ), only one developed HZ.

Figure 1. Retention rate and efficacy of Janus kinase inhibitors (JAKi) in patients with rheumatoid arthritis who initiated JAKi treatment. (A) Retention rate of JAKi until week 26 after initiating JAKi treatment (Kaplan-Meier curve). (B) Changes in disease activity (stratified by the Clinical Disease Activity Index) until week 26 after JAKi treatment initiation. HDA, high disease axctivity. LDA, low disease activity. MDA, moderate disease activity. REM, remission.

At week 26, after initiating JAKi treatment, 32.5% of the patients (141/434) achieved remission and 69.8% (303/434) achieved LDA (figure 1B).

Characteristics of patients with JAKis-IR RA

In total, 138 patients had JAKi-IR RA, and 76 of these patients switched to other drugs because of inadequate response during 26 weeks after initiating JAKi treatment. The remaining 62 patients had not achieved LDA 26 weeks after starting JAKi treatment.

Table 1 shows a comparison between JAKi responders and JAKi-IR patients. We performed univariate logistic regression analysis with JAKi-IR as the dependent variable and all baseline factors collected as explanatory variables. For the multivariable analysis, we used JAKi-IR as the dependent variable. In addition, we included clinically relevant factors related to treatment resistance and those with p<0.05 in the univariate analysis as explanatory variables. EuroQol 5 Dimension (EQ-5D) and anti-cyclic citrullinated peptide (CCP) positivity were excluded due to collinearity with HAQ and actual anti-citrullinated protein antibody (ACPA) levels, respectively. Therefore, HAQ and actual ACPA levels were selected as explanatory variables. In the univariate logistic regression analysis, it was observed that patients with JAKi-resistant RA exhibited a lack of response to multiple bDMARDs before commencing JAKi treatment, not being able to receive JAKis at the recommended dosages and demonstrated elevated scores on the CDAI and HAQ-DI in comparison to patients with RA who responded positively to JAKis. Multivariate logistic regression analysis was performed with the following explanatory variables to determine the characteristics of patients with JAKi-IR RA: factors identified by univariate analyses, age, sex and items presumed to be clinically important. The factors associated with JAKi-IR were identified as follows in the multivariable logistic regression analysis: a larger number of ineffective bDMARDs before initiating JAKi treatment, inability to receive JAKi at optimal doses and higher HAQ-DI scores. These factors were associated with JAKi-IR even when HAQ was replaced with EQ-5D or actual ACPA levels with anti-CCP positivity as explanatory variables.

Table 1

Comparison of patient characteristics at the initiation of treatment with JAK inhibitors (JAKi) between responders and patients with an inadequate response to JAKi (JAKi-IR)

Variables	JAKi-responder n=296	JAKi-IR n=138	P value
			Univariable analysis OR (95% CI), p value	Multivariable analysis OR (95% CI), p value
Age (years)	58.6±13.1	57.3±14.7	0.99 (0.98 to 1.01), 0.3546	0.97 (0.95 to 1.01), 0.0784
Gender, n (% female)	234 (79.1)	110 (79.7)	1.04 (0.63 to 1.72), 0.8752	0.81 (0.46 to 1.44), 0.4797
Disease duration (month)	87 (27 to 168)	101 (36 to 219)	1.00 (0.99 to 1.00), 0.0627	1.00 (0.99 to 1.00), 0.9714
Steinbrocker stage	2.2±1.0	2.5±1.1	1.28 (1.05 to 1.57), 0.0135	1.06 (0.79 to 1.43), 0.6822
Treatment history
MTX use at baseline, n (%)	197 (66.6)	90 (65.2)	0.94 (0.62 to 1.44), 0.7841
Dose, mg/week	12 (10 to 16)	12 (8 to 16)	0.99 (0.99 to 1.02), 0.8376
Glucocorticoid use at baseline, n (%)	62 (21.0)	33 (23.9)	1.19 (0.73 to 1.92), 0.4863
Dose, mg/day	7.5 (5.0 to 17.9)	7.0 (5.0 to 10.0)	0.98 (0.95 to 1.01), 0.4616
Number of previous bDMARDS use	1.5±1.5	2.4±1.7	1.45 (1.27 to 1.65), <0.0001	1.47 (1.24 to 1.74), <0.0001
Patients who were unable to use a sufficient dose of JAKi, n (%)	21 (7.1)	21 (15.2)	0.43 (0.23 to 0.81), 0.0083	0.45 (0.22 to 0.92), 0.0305
Type of JAKi	Tofacitinib, n=106	Tofacitinib, n=56	0.1912
	Baricitinib, n=137	Baricitinib, n=48
	Peficitinib, n=4	Peficitinib, n=4
	Upadacitinib, n=41	Upadacitinib, n=25
	Filgotinib, n=8	Filgotinib, n=5
28-tender joint count	7.8±5.8	10.0±6.5	1.06 (1.02 to 1.09), 0.0005
28-swollen joint count	6.7±4.9	6.8±5.2	1.00 (0.96 to 1.04), 0.8945
GH, VAS 0–100 mm	50.1±25.8	58.7±21.8	1.01 (1.01 to 1.02), 0.0008
EGA, VAS 0–100 mm	43.4±21.7	49.4±21.0	1.01 (1.00 to 1.02), 0.0069
Pain VAS 0–100 mm	48.1±26.8	58.6±22.9	1.02 (1.01 to 1.02), <0.0001
SDAI	25.3±13.4	29.4±13.1	1.02 (1.01 to 1.04), 0.0033
CDAI	24.1±12.2	27.5±12.0	1.02 (1.01 to 1.04), 0.0060	1.02 (0.99 to 1.04), 0.0816
HAQ-DI	1.1±0.8	1.5±0.8	1.75 (1.35 to 2.27), <0.0001	1.50 (1.09 to 2.07), 0.0010
EQ-5D	0.6±0.2	0.5±0.2	0.15 (0.04 to 0.55), 0.0039
CRP (mg/L)	14.5±26.0	18.0±35.0	4.5 (0.97 to 11.14), 0.3620
ESR (mm/hour)	39.0±29.4	35.8±31.3	1.00 (0.99 to 1.00), 0.3017
Rheumatoid factor positive, n (%)	227 (76.7)	98 (71.0)	0.74 (0.47 to 1.17), 0.2043
Rheumatoid factor (U/mL)	64.5 (17.9 to 172.8)	55.1 (8.9 to 170.8)	1.00 (0.99 to 1.00), 0.4440
Anti-CCP antibody, n (%)	222 (75.0)	90 (65.2)	0.63 (0.40 to 0.97), 0.0348
Anti-CCP antibody (U/mL)	62.1 (4.4 to 379.6)	25.3 (1.3 to 261.1)	1.00 (0.99 to 1.00), 0.0667	0.99 (0.99 to 1.00), 0.2326
MMP-3 (ng/mL)	112.5 (53.9 to 254.8)	80.5 (49.1 to 215.2)	1.00 (0.99 to 1.00), 0.2623

Data are mean ± SD, median (IQR) or number (%) of patients.

Univariate analyses were conducted using logistic regression with JAKi-IR as the dependent variable and all baseline factors collected as explanatory variables. Multivariable logistic regression analyses were performed with JAKi-IR as the dependent variable; clinically relevant factors related to treatment resistance and variables with p<0.05 in the univariate analysis were used as explanatory variables.

bDMARDS, biological disease-modifying anti-rheumatic drugs; CCP, cyclic citrullinated peptide; CDAI, Clinical Disease Activity Index; CRP, C-reactive protein; DAS, Disease Activity Score; EQ-5D, EuroQol 5 Dimension; ESR, erythrocyte sedimentation rate; HAQ-DI, Health Assessment Questionnaire Disability Index; JAK, Janus kinase; MMP-3, matrix metalloproteinase 3; MTX, methotrexate; SDAI, Simplified Disease Activity Index; EGA VAS, Evaluator Global Assessment of Disease Activity Visual Analogue Scale; GH VAS, Patient’s Global Assessment of Disease Activity Visual Analogue Scale.

Comparison of efficacy and safety between JAKis and bDMARDs in patients with RA with JAKis-IR

Patients with RA with JAKi-IR were divided into those who cycled to another JAKi and those who switched to bDMARDs to compare efficacy and safety between the two groups.

Table 2 and online supplemental table S4 present patient characteristics during drug switching, differentiating between the group transitioning to another JAKi and the group switching to bDMARDs. There were no statistically significant differences in patient characteristics between the two groups. No clinically meaningful differences were observed in patient background factors.

Table 2

Characteristics at the time of drug switching in patients with rheumatoid arthritis with inadequate response to JAK inhibitors (JAKi)

Variables	JAKi→bDMARDs n=45	JAKi→JAKi n=31	Absolute difference	P value
Age (years)	60.3±15.1	57.4±15.0	2.9 (–4.1 to 9.9)	0.4126
Gender, n (% female)	38 (84.4)	26 (83.9)	0.6 (–15.8 to 18.1)	0.9463
Disease duration (months)	108 (49 to 187)	123 (81 to 300)	−15 (–189 to 17)	0.1070
Steinbrocker stage	2.4±1.0	2.7±1.1	0.2 (–0.7 to 0.2)	0.3338
Treatment history
MTX use at baseline, n (%)	26 (57.8)	17 (54.8)	2.9 (–19.2 to 25.0)	0.7995
Dose, mg/week	10.2±4.2	10.7±4.3	−0.5 (–2.6 to 3.1)	0.1225
Glucocorticoid use at baseline, n (%)	16 (35.6)	11 (35.5)	0.1 (–21.6 to 21.2)	0.9949
Dose, mg/day	10 (8.1 to 23.1)	7.5 (4.0 to 12.5)	2.5 (–1.7 to 4.8)	0.1613
Number of previous bDMARDS use	4.0±1.6	4.3±1.6	−0.3 (–1.1 to 0.4)	0.3922
Type of b/tsDMARDs	TNFi, n=16	Tofacitinib, n=1
	IL-6Ri, n=22	Baricitinib, n=13
	CTLA-4-Ig, n=7	Peficitinib, n=3
		Upadacitinib, n=11
		Filgotinib, n=3
28-tender joint count	9.8±7.2	7.5±6.4	2.3 (–0.9 to 5.4)	0.1647
28-swollen joint count	6.4±6.8	6.2±5.5	0.2 (–2.8 to 3.1)	0.9128
GH, VAS 0–100 mm	58.7±21.8	53.9±23.2	4.8 (–9.2 to 12.4)	0.7702
EGA, VAS 0–100 mm	44.4±21.4	40.5±21.4	3.8 (–6.1 to 13.8)	0.4463
Pain VAS 0–100 mm	54.4±25.4	59.4±20.5	−5.0 (–15.9 to 6.0)	0.3682
SDAI	22.1±11.0	21.7±10.0	0.4 (–4.5 to 5.3)	0.8766
CDAI	20.7±9.6	20.9±9.2	−0.2 (–4.9 to 4.2)	0.6529
HAQ-DI	1.2±0.8	1.4±0.8	−0.2 (–0.6 to 0.2)	0.2458
EQ-5D	0.6±0.2	0.6±0.2	0.0 (–0.1 to 0.1)	0.7993
CRP (mg/L)	17.9±31.6	10.7±23.1	7.2 (–6.0 to 20.5)	0.3887
ESR (mm/hour)	44.8±34.2	38.4±28.8	6.4 (–8.5 to 21.4)	0.3953
Rheumatoid factor positive, n (%)	34 (75.6)	20 (64.5)	11.0 (–9.8 to 31.4)	0.2970
Rheumatoid factor (U/mL)	82.0 (14.3 to 244.1)	31.6 (5.7 to 161.8)	50.4 (–19.9 to 110.2)	0.1195
Anti-CCP antibody, n (%)	29 (64.4)	21 (67.7)	−3.3 (–24.0 to 18.3)	0.7659
Anti-CCP antibody (U/mL)	82.1 (0.8 to 590.6)	58.3 (1.2 to 330.2)	23.8 (–30.2 to 202.4)	0.2216
MMP-3 (ng/mL)	84.1 (47.7 to 227.0)	63.5 (35.7 to 165.2)	20.6 (–37.9 to 66.1)	0.1945

Data are mean ± SD, median (IQR), or number (%) of patients. The absolute difference represents the numerical difference (95% CI) when using the group that switched to bDMARDs as the reference, compared with the group that cycled to another JAKi.

Univariate analyses were conducted using logistic regression with JAKi-IR as the dependent variable and all baseline factors collected as explanatory variables.

bDMARDS, biological disease-modifying anti-rheumatic drugs; CCP, cyclic citrullinated peptide; CDAI, Clinical Disease Activity Index; CRP, C-reactive protein; CTLA-4, Cytotoxic T-lymphocyte Antigen-4; DAS, Disease Activity Score; EGA VAS, Evaluator Global Assessment of Disease Activity Visual Analogue Scale; EQ-5D, EuroQol 5 Dimension; ESR, erythrocyte sedimentation rate; GH VAS, Patient's Global Assessment of Disease Activity Visual Analogue Scale; HAQ-DI, Health Assessment Questionnaire Disability Index; IL, interleukin; JAK, Janus kinase; JAKi-IR, inadequate response to JAKi; MMP-3, matrix metalloproteinase 3; MTX, methotrexate; SDAI, Simplified Disease Activity Index; TNF, tumour necrosis factor; TNFi, tumor necrosis factor inhibitor; tsDMARDs, targeted synthetic disease-modifying anti-rheumatic drugs.

The retention rate at week 26 after drug switching was 87.1% in the group that cycled to other JAKis and 75.6% in the group that switched to bDMARDs. It tended to be higher in the former than the latter group, although no significant difference was observed between the groups (p=0.1721, log-rank test) (figure 2A). Online supplemental table S5 shows the rate of adverse events in the two groups during the 26 weeks after drug switching. Although there was no statistically significant difference in the incidence of adverse events, the infection rate showed an absolute difference of 6%, with a tendency for higher rates in the group that cycled to another JAKi than in the bDMARD-treated group. The group cycling to another JAKi regimen did not include any patients who developed malignancy, major adverse cardiovascular events or thromboembolism. Figure 2B shows the changes in the CADI in the two groups. In the group that switched to bDMARDs, the CDAI significantly improved from baseline to week 12 after drug switching and remained improved until week 26. In the group that cycled to another JAKi, the CDAI significantly improved from baseline to week 2 after drug cycling and continued to improve until week 26. When the CDAI was compared between the two groups, in the group that cycled to another JAKi, it tended to be lower at weeks 2 and 4 after drug cycling and was significantly lower at week 26 (figure 2C). When the rate of CDAI-LDA-achievement and CDAI-remission at week 26 were compared, they were significantly higher in the group cycling to another JAKis than in the group switching to bDMARDs (CDAI-LDA-achievement rate: JAKis→JAKis: JAKis→bDMARD=64.5%:28.9%, p=0.0021; CDAI-remission rate: JAKis→JAKis: JAKis→bDMARD=38.7%:2.2%, p<0.0001) (figure 2D). The group that cycled to another JAKi had lower CDAI at 2, 4 and 12 weeks compared with the group that switched to bDMARDs. At 26 weeks, the CDAI showed a large absolute difference of −6.0 in favour of the group that cycled to another JAKi. Furthermore, remission and LDA achievement rates showed an absolute difference of >35%, suggesting higher efficacy in the group that cycled to another JAKi than in the group that switched to an alternative bDMARD (online supplemental table S6).

Figure 2. Comparison of efficacy and safety after cycling to another JAK inhibitor (JAKi) or switching to biological disease-modifying anti-rheumatic drugs (bDMARDs) in patients with rheumatoid arthritis (RA) with inadequate response to JAKi (JAKi-IR RA). (A) Comparison of the retention rates at week 26 after cycling to another JAKi or switching to bDMARDs in patients with JAKi-IR RA (Kaplan-Meier curve). (B) Changes in the CDAI until week 26 after drug switching in patients with JAKi-IR RA (left: patients with JAKi-IR RA who switched from JAKi to bDMARDs; right: patients with JAKi-IR RA who cycled from JAKi to another JAKi). Numbers represent mean values. *p<0.05, **p<0.01, paired t-test. (C) Comparison of changes in the CDAI until week 26 after drug switching in patients with JAKi-IR RA. Numbers represent mean+-SD, and p values are calculated by Student’s t-test. (D) Comparison of the CDAI-LDA-achievement rate (left) and the CDAI-remission rate (right) at week 26 after drug switching in patients with JAKi-IR RA. Pearson’s [chi] 2 test. Numbers represent percentages of all patients (%). bDMARDs, biological disease-modifying anti-rheumatic drugs; CDAI, Clinical Disease Activity Index; JAK, Janus kinase; LDA, low disease activity.

Similar comparisons were performed for sensitivity analysis after drug selection bias was minimised by the use of propensity score-based inverse probability of treatment weighting (PS-IPTW). After adjusting the PS-IPTW, there was no difference in patient characteristics at the time of switching to molecular-targeted drugs between the two groups of patients with JAKi-IR RA (online supplemental table S7). The SD were also <0.1 for all factors. No clinically relevant differences were observed in patient background factors. In patients with JAKi-IR RA, cycling to another JAKi was effective, even after adjustment with PS-IPTW (online supplemental figure S1 and table S8).

One type of bDMARD was ineffective in 26 patients. The patients were subsequently switched to a JAKi but experienced an incomplete response. Of these, nine patients were cycled to another JAKi, whereas 17 were switched to another bDMARD. At 26 weeks, higher remission (JAKi→JAKi vs JAKi→bDMARD=33.3% (3/9) vs 5.9% (1/17), p=0.0649) and LDA achievement (JAKi→JAKi vs JAKi→bDMARD=66.7% (6/9) vs 11.8% (2/17), p=0.00391) rates were observed in the group that cycled to another JAKi than in the group that switched to another bDMARD.

Logistic regression analysis revealed that higher RF levels at the time of the switch were associated with a higher likelihood of high disease activity (HDA) after 26 weeks in patients with JAKi-IR RA who switched to bDMARDs, with a cut-off value of 346 IU/mL. Conversely, no factors associated with HDA 26 weeks after the switch were identified for those who cycled to another JAKi. This suggests that cycling to another JAKi may be a more effective option than switching to a bDMARD for patients with high RF levels, particularly those with RF≥346 IU/mL.

Trajectories of the CDAI after drug switching in patients with RA with JAKis-IR by GMM

GMM was used to analyse the trajectories of changes in CDAI after drug switching in 76 patients with JAKi-IR RA. Among the linear models of the trajectories, the model with the cubic function exhibited the best fit (online supplemental table S9). As for the number of groups, the best fit was achieved by dividing into three groups (online supplemental table S10). Group 1 (patients who had high disease activity at baseline and showed no improvement in disease activity despite switching to molecular-targeted drugs), Group 2 (patients who had moderate disease activity at baseline and showed only partial improvement in disease activity despite switching to molecular-targeted drugs) and Group 3 (patients who had moderate-to-high disease activity at baseline but showed prompt improvement in disease activity after switching to molecular-targeted drugs) (figure 3A).

Figure 3. Analysis using growth mixture modelling for the trajectories of the CDAI. (A) The trajectories of CDAI were analysed to model the CDAI response after drug switching in patients with rheumatoid arthritis (RA) with inadequate responses to JAK inhibitors (JAKi-IR RA). (B) Changes in the CDAI of all patients with JAKi-IR RA after drug switching and the proportion of patients classified into each trajectory group. Group 1 black line, Group 2 blue line, Group 3 red line. bDMARDs, biological disease-modifying anti-rheumatic drugs; CDAI, Clinical Disease Activity Index; JAK, Janus kinase.

When the proportions of patients classified into each group were compared between those switching to bDMARDs and those cycling to another JAKis, the proportion of patients classified as Group 3 was higher in those cycling to another JAKis than in those switching to bDMARDs (JAKis→JAKis: JAKis→bDMARD=75.0%:11.1%, p=0.0003, Pearson’s χ² test) (figure 3B). Although no difference was observed in the proportion of patients classified as Group 1 (JAKis→bDMARD: JAKis→JAKis=13.3%:19.4%, p=0.4793, Pearson’s χ² test), the proportion of patients classified as Group 2, which included patients with partial improvement in disease activity, was higher in those switching to bDMARDs than in those cycling to another JAKis (JAKis→bDMARD: JAKis→JAKis=75.6%:32.3%, p=0.0001, Pearson’s χ² test).

Characteristics of patients with JAKis-IR RA who were subsequently included in the treatment response group

Group 3, defined by the GMM as a group of patients with JAKi-IR who showed prompt improvement in disease activity after switching to molecular-targeted drugs, was regarded as the treatment response group. The characteristics of the patients in this group were examined. Online supplemental table S11 shows patient characteristics at the time of drug switching according to the trajectory groups identified by the GMM. Only the proportion of patients classified into the three groups differed between those switching to bDMARDs and those cycling to other JAKis, and no clinically relevant differences were observed in the patient backgrounds across the trajectory groups. A multiple logistic regression analysis was conducted, including the following explanatory variables: factors identified with a significance level of p<0.05, age, gender and the number of previously administered bDMARDs, considered clinically relevant for identifying factors associated with treatment response within the study group (table 3). The CDAI was excluded from the explanatory variables because the grouping was based on CDAI trajectories. CRP was also excluded as an explanatory variable in the multivariate analysis because it showed multicollinearity with MMP-3 and HAQ-DI. Based on the results of multiple logistic regression analysis, only cycling from one JAKi to another JAKis contributed to the inclusion into the treatment response group (OR: 7.15, 95% CI: 2.21 to 23.10, p=0.0004) (table 3).

Table 3

Identification of factors contributing to treatment response

	Univariable analysis		Multivariable analysis
	OR (95% CI)	P value	OR (95% CI)	P value
Age (years)	0.99 (0.95 to 1.02)	0.4252	0.99 (0.95 to 1.03)	0.5066
Female	1.09 (0.26 to 4.48)	0.9102	1.12 (0.23 to 5.54)	0.5068
Disease duration (months)	1.00 (0.99 to 1.00)	0.1735
Steinbrocker stage	1.41 (0.85 to 2.36)	0.1811
MTX dose	0.93 (0.85 to 1.02)	0.1061
Number of previous bDMARDS use	1.15 (0.84 to 1.58)	0.3820	1.13 (0.79 to 1.63)	0.4984
GH, VAS 0–100 mm	0.99 (0.97 to 1.01)	0.3892
EGA, VAS 0–100 mm	0.97 (0.94 to 1.00)	0.0268	0.97 (0.94 to 1.01)	0.0563
Pain VAS 0–100 mm	0.99 (0.97 to 1.01)	0.3859
SDAI	0.96 (0.91 to 1.02)	0.1318
CDAI	0.99 (0.93 to 1.05)	0.6468
HAQ-DI	1.22 (0.64 to 2.34)	0.5360
CRP (mg/dL)	0.95 (0.78 to 1.16)	0.6141
ESR (mm/hour)	0.99 (0.97 to 1.01)	0.2603
Rheumatoid factor titre	0.99 (0.99 to 1.00)	0.1475
Anti-CCP antibody titre	0.99 (0.99 to 1.00)	0.1732
JAKi→TNF inhibitor	0.34 (0.09 to 1.32)	0.1733
JAKi→IL-6 receptor inhibitor	1.00 (0.99 to 1.00)	0.1199
JAKi→CTLA-4 Ig	1.00 (0.99 to 1.01)	0.9921
JAKi→JAKi	7.50 (2.34 to 24.08)	0.0007	7.15 (2.21 to 23.10)	0.0004

Univariate analyses were conducted using logistic regression with the inclusion of the treatment response group as the dependent variable and all baseline factors collected as explanatory variables. Multivariable logistic regression analyses were performed with inclusion into the treatment response group as the dependent variable; clinically relevant factors related to treatment responsiveness and variables with p<0.05 in the univariate analysis were used as explanatory variables.

bDMARDS, biological disease-modifying anti-rheumatic drugs; CCP, cyclic citrullinated peptide; CDAI, Clinical Disease Activity Index; CRP, C-reactive protein; CTLA-4, Cytotoxic T-lymphocyte Antigen-4; EGA VAS, Evaluator Global Assessment of Disease Activity Visual Analogue Scale; ESR, erythrocyte sedimentation rate; GH VAS, Patient’s Global Assessment of Disease Activity Visual Analogue Scale; HAQ-DI, Health Assessment Questionnaire Disability Index; IL, interleukin; JAK, Janus kinase; MMP-3, matrix metalloproteinase 3; MTX, methotrexate; SDAI, Simplified Disease Activity Index; TNF, tumour necrosis factor.

Comparison of the efficacy and safety of each molecular-targeted drug in patients with JAKis-IR RA

The efficacy and safety of each molecular-targeted drug were compared in 76 patients who inadequately responded to JAKis and were switched to TNF inhibitors, non-TNF inhibitors, or other JAKis for 26 weeks after switching treatment. Online supplemental table S12 shows patient characteristics at the time of drug switching in patients with RA with JAKi-IR who initiated treatment with molecular-targeted drugs. No significant differences were observed in the patient characteristics among the three groups of patients who switched to TNF inhibitors or non-TNF inhibitors or cycled to other JAKis. No clinically relevant differences were observed in patient background factors (online supplemental table S13). The retention rates were 75.0%, 75.9 % and 87.1% in the groups switching to TNF inhibitors, non-TNF inhibitors and 87.1% in the group cycling to another JAKis, respectively. The retention rate was higher in the group that switched to another JAKi than in the other two groups, with no difference among the three groups (p=0.0538, log-rank test) (online supplemental figure S2A). No statistically significant differences were observed. Furthermore, the group that switched to JAKi-non-TNF inhibitors had a lower incidence of adverse events than other groups. However, the small number of cases indicates that this difference is not clinically meaningful (online supplemental table S14). Online supplemental figure S2B shows the changes in CDAI in each group. In the group switching to TNF inhibitors, the CDAI did not significantly improve. In the group of patients who switched to non-TNF inhibitors or cycled to other JAKis, CDAI significantly improved from baseline to week 26 after drug switching. CDAI-LDA achievement and CDAI remission rates at week 26 in patients with JAKi-IR RA were compared using Pearson’s χ² test and residue analysis. Pearson’s χ² test revealed differences in CDAI-LDA achievement and CDAI remission rates. Residue analysis revealed that the CDAI remission rate was higher in the group cycling to another JAKi than in the patients switching to TNF or non-TNF inhibitors (online supplemental figure S2C). The group cycling to another JAKi showed significantly higher absolute differences in CDAI remission and LDA achievement rates than that switching to a bDMARD, suggesting greater efficacy (online supplemental table S15). In patients with JAKis-IR RA who were administered bDMARDs, TNF inhibitors were given to two patients with a history of secondary failure to TNF inhibitors and interleukin (IL)-6 receptor inhibitors were administered to four patients with a history of secondary failure to IL-6 receptor inhibitors. Among these patients, only one achieved LDA after 26 weeks.

Similar comparisons were performed for sensitivity analysis after drug selection bias was minimised by using PS-IPTW. After adjustment with PS-IPTW, there was no difference in patient characteristics at the time of switching to molecular-targeted drugs between the three groups of patients with JAKi-IR RA (online supplemental table S16). The SD were also <0.1 for all factors. No clinically meaningful differences were observed in patient background factors (online supplemental table S17). In patients with JAKi-IR RA, cycling to another JAKi was effective, even after adjustment with PS-IPTW (online supplemental figure S3 and table S18).

Discussion

In this study, patients with RA with JAKi-IR were defined as those who discontinued or changed treatment because of an inadequate response to JAKis or those who did not achieve LDA 26 weeks after initiating JAKi treatment. This study identified the characteristics of patients with RA with JAKi-IR. It showed that using another type of JAKi may be highly effective in patients with RA with JAKi-IR.

Previous studies have reported that patients with a low JAKi retention rate are refractory to multiple bDMARDs^{5 17} and that IL-6 receptor antibodies are ineffective in such patients.¹⁸ Our current study is the first to reveal the characteristics of patients with JAKis-IR in terms of drug efficacy. Patients with RA who received higher doses of JAKis were less likely to exhibit JAKi-IR. These doses may have affected the efficacy of JAKis because they are effective in a dose-dependent manner, as demonstrated in various clinical studies.^19–23

In this study, when the CDAI trajectories of patients with JAKi-IR RA were analysed using GMM, the patients were divided into three trajectory groups. Numerous patients who cycled to other JAKis were classified into Group 3 (treatment response group), which included patients who had moderate-to-high disease activity at baseline but showed prompt improvement in disease activity after drug switching and remained improved until week 26. Multivariable logistic regression analysis of the patient characteristics in the treatment response group identified cycling to another JAKi as the only factor contributing to classification into the treatment response group. The CDAI remission rate at week 26 after drug switching was significantly higher in the patients who cycled to another JAKi regimen than in those who switched to bDMARDs. These findings suggest that cycling to another JAKi may be appropriate in patients with RA with JAKi-IR. Prior research has indicated that the retention rate may be higher after cycling to other JAKis than after switching to bDMARDs in patients with RA who fail to respond to or cannot continue receiving their first JAKis.^5–7 Our study also raises the possibility that cycling to another JAKi may lead to superior drug efficacy in patients with RA and JAKi-IR. However, there were no differences in the retention rate or incidence of adverse events after switching between the drug-switching groups. Here, patients with RA with JAKi-IR were defined as those who failed to respond to JAKis; however, patients who discontinued treatment because of adverse events were not included in this study. Considering that patients with JAKi-IR can safely use JAKis, there may have been no differences in the rate of adverse events or retention rate in this study. Weng et al and Pope et al performed network meta-analyses of JAKis, suggesting their efficacy may vary.^{24 25} We also suggest that the efficacy of JAKis may vary from drug to drug based on data collected in real clinical practice.²⁶ Due to the differences in efficacy among JAKis, as reported in these studies, disease activity may have been improved by cycling to another JAKi in patients with RA with JAKi-IR in the current study.

Furthermore, TNF inhibitors have been reported to be effective in patients with RA showing poor response to JAKis.²⁷ In the current study, few patients with RA with JAKi-IR who switched to bDMARDs achieved remission at week 26; however, many of these patients were included in Group 2, which was identified by GMM as a group that included patients who achieved LDA at week 26. A possible reason for the limited effect of TNF inhibitors is that almost all the patients had used one or more TNF inhibitors and failed to respond to them. Multiple logistic regression analysis did not identify any clear factors involved in the failure to improve disease activity after drug switching, as observed in Group 1 (data not shown).

This study has some limitations. First, the number of patients with RA treated with JAKi-IR was small. Due to the small sample size, we might have been unable to accurately assess the efficacy of bDMARDs selected after patients exhibited JAKi-IR. Second, no conclusive basic investigations support differences in the efficacy of different JAKis. As the signal transduction analysis results were not necessarily consistent with the safety and efficacy profiles observed in clinical practice, this study may have contributed to the differences in efficacy among JAKis, as found when cycling to other JAKis; this needs to be investigated further in future studies. Moreover, a selection bias exists when molecular-targeted drugs are introduced to patients with JAKi-IR RA. This study found no difference in the characteristics of patients with JAKi-IR RA at the time of selecting JAKis or bDMARDs. In addition, all the drugs presented in online supplemental table S1 were available at the facilities participating in this study, and no bias in drug selection was observed between the facilities. Even when patients with JAKi-IR RA were compared after selection bias minimisation with PS-IPTW, selecting other JAKis was the most effective. Thus, although selection bias was evident, its influence was small. Finally, this study is a retrospective observational study. Thus, it is more prone to various biases, such as information and selection biases, than prospective observational studies. Additionally, differences in historical context may have influenced our results. To address these biases, data collection at each observation point was conducted by trained research assistants specialised in data processing, resulting in a data missing rate of approximately 10% at baseline and 26 weeks.

In conclusion, this study raises the possibility that cycling to another JAKi contributes to more significant improvement than switching to bDMARDs in patients with RA with JAKi-IR. Therefore, cycling to another JAKi may be the optimal option among molecular-targeted drugs for patients with RA and JAKi-IR.

The authors thank all medical staff at all participating institutions for providing the data, especially Ms Hiroko Yoshida, Ms Youko Saitou and Ms Machiko Mitsuiki for the excellent data management in the FIRST registry. We also thank Dr Kazuyoshi Saito at Tobata General Hospital, SF at Wakamatsu Hospital of the University of Occupational and Environmental Health, Dr Keisuke Nakatsuka at Fukuoka Yutaka Hospital and all staff members at Kitakyushu General Hospital and Shimonoseki Saiseikai Hospital for their engagement in data collection of the FIRST registry.

Data availability statement

Data are available upon reasonable request.

Ethics statements

Patient consent for publication

Not applicable.

Ethics approval

The FIRST registry includes patients with rheumatoid arthritis who initiated treatment with molecular-targeting rheumatic drugs in a registry study conducted by the University of Occupational and Environmental Health, Japan, and cooperated with its affiliated and multiple facilities. This study was approved by the ethics review board of the University of Occupational and Environmental Health, Japan (approval number #04-23), and informed consent was obtained from all patients of the FIRST registry. Participants gave informed consent to participate in the study before taking part.

Footnote

Contributors All authors were involved in the drafting and critical revision of the manuscript, and all authors approved the final version to be published. YM had full access to all of the data in the study, and YT unified the study and took responsibility for the integrity and accuracy of the data analysis. Study conception and design: YM, SN and YTa. Acquisition of data: YM and SK. Analysis and interpretation of data: YM, SN, HT, KH, SF, SK, AY, IM, YS-K, YTo, YI, MU and YTa. YTa is the guarantor and accepts full responsibility for the study and the accuracy of its content.

Funding This work was supported in part by a Grant-In-Aid for Scientific Research from the University of Occupational and Environmental Health, Japan, through the University of Occupational & Environmental Health, Japan (UOEH) for Advanced Research (19K17919).

Competing interests YM has received consulting fees, speaking fees, and/or honoraria from Eli Lilly and has received research grants from GlaxoSmithKline. SN has received consulting fees, lecture fees, and/or honoraria from Bristol-Myers, AstraZeneca, Pfizer, GlaxoSmithKline, AbbVie, Astellas, Asahi-kasei, Sanofi, Chugai, Eisai, Gilead Sciences, Eli Lilly, Boehringer Ingelheim. SK has received speaking fees from Bristol-Myers. YT has received consulting fees, speaking fees, and/or honoraria from AbbVie, Daiichi-Sankyo, Chugai, Takeda, Mitsubishi-Tanabe, Bristol-Myers, Astellas, Eisai, Janssen, Pfizer, Asahi-kasei, Eli Lilly, GlaxoSmithKline, UCB, Teijin, MSD and Santen, and received research grants from Mitsubishi-Tanabe, Takeda, Chugai, Astellas, Eisai, Taisho-Toyama, Kyowa-Kirin, AbbVie and Bristol-Myers. All other authors declare no conflict of interest.

Provenance and peer review Not commissioned; externally peer reviewed.

Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

References

¹ Smolen JS, Aletaha D, Barton A, et al. Rheumatoid arthritis. Nat Rev Dis Primers 2018; 4: 18001. doi:10.1038/nrdp.2018.1

² Tanaka Y. Rheumatoid arthritis. Inflamm Regen 2020; 40: 20. doi:10.1186/s41232-020-00133-8

³ Tanaka Y, Luo Y, O’Shea JJ, et al. Janus kinase-targeting therapies in rheumatology: a mechanisms-based approach. Nat Rev Rheumatol 2022; 18: 133–45. doi:10.1038/s41584-021-00726-8

⁴ Lauper K, Iudici M, Mongin D, et al. Effectiveness of TNF-inhibitors, abatacept, IL6-inhibitors and JAK-inhibitors in 31 846 patients with rheumatoid arthritis in 19 registers from the “JAK-pot” collaboration. Ann Rheum Dis 2022; 81: 1358–66. doi:10.1136/annrheumdis-2022-222586

⁵ Pombo-Suarez M, Sanchez-Piedra C, Gómez-Reino J, et al. After JAK inhibitor failure: to cycle or to switch, that is the question - data from the JAK-pot collaboration of registries. Ann Rheum Dis 2023; 82: 175–81. doi:10.1136/ard-2022-222835

⁶ Retuerto M, Trujillo E, Valero C, et al. Efficacy and safety of switching Jak inhibitors in rheumatoid arthritis: an observational study. Clin Exp Rheumatol 2021; 39: 453–5. doi:10.55563/clinexprheumatol/cbanza

⁷ Amstad A, Papagiannoulis E, Scherer A, et al. Comparison of drug retention of TNF inhibitors, other biologics and JAK inhibitors in RA patients who discontinued JAK inhibitor therapy. Rheumatology (Oxford) 2022; 62: 89–97. doi:10.1093/rheumatology/keac285

⁸ Ram N, Grimm KJ. Growth Mixture Modeling: A Method for Identifying Differences in Longitudinal Change Among Unobserved Groups. Int J Behav Dev 2009; 33: 565–76. doi:10.1177/0165025409343765

⁹ Sonomoto K, Tanaka H, Nguyen TM, et al. Prophylaxis against pneumocystis pneumonia in rheumatoid arthritis patients treated with b/tsDMARDs: insights from 3787 cases in the FIRST registry. Rheumatology (Oxford) 2022; 61: 1831–40. doi:10.1093/rheumatology/keab647

¹⁰ Kawabe A, Nakano K, Kubo S, et al. Differential long-term retention of biological disease-modifying antirheumatic drugs in patients with rheumatoid arthritis by age group from the FIRST registry. Arthritis Res Ther 2020; 22: 136. doi:10.1186/s13075-020-02233-9

¹¹ Arnett FC, Edworthy SM, Bloch DA, et al. The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum 1988; 31: 315–24. doi:10.1002/art.1780310302

¹² Aletaha D, Neogi T, Silman AJ, et al. 2010 Rheumatoid arthritis classification criteria: An American College of Rheumatology/European League Against Rheumatism collaborative initiative. Arthritis Rheum 2010; 62: 2569–81. doi:10.1002/art.27584

¹³ Miyata H, Sonomoto K, Fukuyo S, et al. Computed tomography for malignancy screening in patients with rheumatoid arthritis before initiation of disease modifying antirheumatic drug. Rheumatology (Sunnyvale) 2023; 62: 3339–49. doi:10.1093/rheumatology/kead075

¹⁴ Smolen JS, Landewé RBM, Bergstra SA, et al. EULAR recommendations for the management of rheumatoid arthritis with synthetic and biological disease-modifying antirheumatic drugs: 2022 update. Ann Rheum Dis 2023; 82: 3–18. doi:10.1136/ard-2022-223356

¹⁵ Felson DT, Smolen JS, Wells G, et al. American College of Rheumatology/European League Against Rheumatism provisional definition of remission in rheumatoid arthritis for clinical trials. Arthritis Rheum 2011; 63: 573–86. doi:10.1002/art.30129

¹⁶ Felson DT, Smolen JS, Wells G, et al. American College of Rheumatology/European League against Rheumatism provisional definition of remission in rheumatoid arthritis for clinical trials. Ann Rheum Dis 2011; 70: 404–13. doi:10.1136/ard.2011.149765

¹⁷ Fitton J, Melville AR, Emery P, et al. Real-world single centre use of JAK inhibitors across the rheumatoid arthritis pathway. Rheumatology (Oxford) 2021; 60: 4048–54. doi:10.1093/rheumatology/keaa858

¹⁸ Ebina K, Hirano T, Maeda Y, et al. Factors affecting drug retention of Janus kinase inhibitors in patients with rheumatoid arthritis: the ANSWER cohort study. Sci Rep 2022; 12: 134. doi:10.1038/s41598-021-04075-0

¹⁹ van Vollenhoven RF, Fleischmann R, Cohen S, et al. Tofacitinib or adalimumab versus placebo in rheumatoid arthritis. N Engl J Med 2012; 367: 508–19. doi:10.1056/NEJMoa1112072

²⁰ Keystone EC, Taylor PC, Drescher E, et al. Safety and efficacy of baricitinib at 24 weeks in patients with rheumatoid arthritis who have had an inadequate response to methotrexate. Ann Rheum Dis 2015; 74: 333–40. doi:10.1136/annrheumdis-2014-206478

²¹ Tanaka Y, Takeuchi T, Tanaka S, et al. Efficacy and safety of peficitinib (ASP015K) in patients with rheumatoid arthritis and an inadequate response to conventional DMARDs: a randomised, double-blind, placebo-controlled phase III trial (RAJ3). Ann Rheum Dis 2019; 78: 1320–32. doi:10.1136/annrheumdis-2019-215163

²² Tanaka Y. A review of upadacitinib in rheumatoid arthritis. Mod Rheumatol 2020; 30: 779–87. doi:10.1080/14397595.2020.1782049

²³ Combe B, Kivitz A, Tanaka Y, et al. Filgotinib versus placebo or adalimumab in patients with rheumatoid arthritis and inadequate response to methotrexate: a phase III randomised clinical trial. Ann Rheum Dis 2021; 80: 848–58. doi:10.1136/annrheumdis-2020-219214

²⁴ Weng C, Xue L, Wang Q, et al. Comparative efficacy and safety of Janus kinase inhibitors and biological disease-modifying antirheumatic drugs in rheumatoid arthritis: a systematic review and network meta-analysis. Ther Adv Musculoskelet Dis 2021; 13: 1759720X21999564. doi:10.1177/1759720X21999564

²⁵ Pope J, Sawant R, Tundia N, et al. Comparative Efficacy of JAK Inhibitors for Moderate-To-Severe Rheumatoid Arthritis: A Network Meta-Analysis. Adv Ther 2020; 37: 2356–72. doi:10.1007/s12325-020-01303-3

²⁶ Miyazaki Y, Nakano K, Nakayamada S, et al. Efficacy and safety of tofacitinib versus baricitinib in patients with rheumatoid arthritis in real clinical practice: analyses with propensity score-based inverse probability of treatment weighting. Ann Rheum Dis 2021; 80: 1130–6. doi:10.1136/annrheumdis-2020-219699

²⁷ Fleischmann RM, Blanco R, Hall S, et al. Switching between Janus kinase inhibitor upadacitinib and adalimumab following insufficient response: efficacy and safety in patients with rheumatoid arthritis. Ann Rheum Dis 2021; 80: 432–9. doi:10.1136/annrheumdis-2020-218412