Continuation, reduction, or withdrawal of tofacitinib in patients with rheumatoid arthritis achieving sustained disease control: a multicenter, open-label, randomized controlled trial

Abstract

Background: 

Rheumatoid arthritis (RA), a chronic systemic autoimmune disease, is characterized by synovitis and progressive damage to the bone and cartilage of the joints, leading to disability and reduced quality of life. This study was a randomized clinical trial comparing the outcomes between withdrawal and dose reduction of tofacitinib in patients with RA who achieved sustained disease control.

Methods: 

The study was designed as a multicenter, open-label, randomized controlled trial. Eligible patients who were taking tofacitinib (5 mg twice daily) and had achieved sustained RA remission or low disease activity (disease activity score in 28 joints [DAS28] ≤3.2) for at least 3 months were enrolled at six centers in Shanghai, China. Patients were randomly assigned (1:1:1) to one of three treatment groups: continuation of tofacitinib (5 mg twice daily); reduction in tofacitinib dose (5 mg daily); and withdrawal of tofacitinib. Efficacy and safety were assessed up to 6 months.

Results: 

Overall, 122 eligible patients were enrolled, with 41 in the continuation group, 42 in the dose-reduction group, and 39 in the withdrawal group. After 6 months, the percentage of patients with a DAS28-erythrocyte sedimentation rate (ESR) of <3.2 was significantly lower in the withdrawal group than that in the reduction and continuation groups (20.5%, 64.3%, and 95.1%, respectively; P < 0.0001 for both comparisons). The average flare-free time was 5.8 months for the continuation group, 4.7 months for the dose reduction group, and 2.4 months for the withdrawal group.

Conclusion: 

Withdrawal of tofacitinib in patients with RA with stable disease control resulted in a rapid and significant loss of efficacy, while standard or reduced doses of tofacitinib maintained a favorable state.

Trial Registration: 

Chictr.org, ChiCTR2000039799.

Introduction

Rheumatoid arthritis (RA), a chronic systemic autoimmune disease, is characterized by synovitis and progressive damage to the bone and cartilage of the joints, leading to disability and reduced quality of life. To avoid severe joint destruction and potential functional disability, disease progression must be controlled.^[1] Achieving clinical remission or at least maintaining low disease activity (LDA) is the main goal in the treatment of RA patients.^[2] Biologics, such as anti-tumor necrosis factor (TNF) agents, and targeted synthetic disease-modifying antirheumatic drugs (tsDMARDs), mainly Janus kinase (JAK) inhibitors, have helped many RA patients to control their disease activity, especially when conventional synthetic disease-modifying antirheumatic drugs (csDMARDs) have not been effective.^[3,4] On the other hand, it is feasible, given the high cost and latent adverse events (AEs) of these drugs, to reduce the drug dosage or even discontinue use when sustained disease control has been achieved. It is recommended that treatment with biological disease-modifying antirheumatic drugs (bDMARDs) or tsDMARDs be tapered when the patient is in persistent remission after tapering glucocorticoids (GCs), especially when these agents are combined with a csDMARD.^[4] Therefore, several high-quality studies in recent decades have focused on the tapering or withdrawal of bDMARDs and tsDMARDs in RA.^[5-11]

Tofacitinib, a selective JAK 1 and 3 inhibitor, was the first tsDMARD to be approved for the treatment of RA.^[12] Tofacitinib, used either as a monotherapy or in combination with methotrexate (MTX), works well in refractory or MTX-naïve RA.^[13-17] In China, tofacitinib was launched in March 2017 for the treatment of moderately to severely active RA, and in January 2020, it was included in the Chinese medical insurance directory to make it more affordable.

Only a few prospective studies have been conducted on the discontinuation or tapering of tofacitinib. A Japanese multicenter observational study enrolled 54 RA patients who discontinued tofacitinib after completing a phase III trial and a long-term extension study. At week 52, while 37% of the patients remained flare-free, disease activity in the discontinuation group was higher than that in the continuation group.^[18] The results of this study also suggested that low rheumatoid factor levels might predict a stable condition after tofacitinib discontinuation. In another Japanese investigation, 68 RA patients who had achieved remission/LDA were non-randomly assigned to withdrawal, reduction, or continuation groups. The results indicated that the reduction strategy was preferable to immediate withdrawal, and patients restarting tofacitinib regained control of their flare-ups within a month.^[19] A randomized sub-study of ORAL Sequel assessed the effects of tofacitinib re-establishment after a mandated interruption for 2 weeks; the study documented that the efficacy could be re-established but with a higher occurrence of AEs.^[20]

To date, there are no randomized clinical trial data on either reducing the tofacitinib dose to 5 mg daily or withdrawing the drug after achieving sustained disease control in RA. To address this issue, we performed a randomized, multicenter, prospective, parallel-group trial comparing the effects of the continuation, dose reduction, and withdrawal of tofacitinib in RA patients with remission/LDA.

Methods

Ethical approval

The study was done in accordance with the Harmonization Good Clinical Practice Guidelines and the Declaration of Helsinki. The study protocol and informed consent documents were reviewed and approved by the Institutional Research Ethics Committee at Ruijin Hospital (Nos. 2019-36 and 2019-36-2) in Shanghai, China, with continuing renewal during the trial. This study was registered on the Chinese Clinical Trial Registry, a primary registry of the International Clinical Trial Registry Platform, World Health Organization (No. ChiCTR2000039799). All patients provided written informed consent.

Study design

This study was a randomized, multicenter, open-label trial carried out at six top-class tertiary hospitals in Shanghai, China. After screening, eligible patients were randomly allocated to continuation, reduction, and withdrawal groups and were followed up for up to 6 months to assess the outcomes.

Participants

Patients aged 18 to 70 years diagnosed with RA based on the 2010 American College of Rheumatology (ACR)/European League Against Rheumatism (EULAR) classification criteria were screened (0–3 months before randomization) if they were taking an adequate dose of tofacitinib (5 mg twice daily) and had already achieved or were likely to achieve sustained remission/LDA. Sustained remission/LDA was defined as a disease activity score in 28 joints based on the erythrocyte sedimentation rate (DAS28-ESR) ≤3.2. Patients were allowed to take concomitant drugs in the screening phase only if they were receiving a stable dose of 10 to 20 mg/week MTX or GC (≤7.5 mg/day prednisone-equivalents). Patients would pass the screening and be assigned to randomization if they remained in remission/LDA for at least 3 months.

Patients were excluded if they appeared to have moderate to high disease activity during the screening. Patients with contraindications to tofacitinib were also excluded, including pregnant and breastfeeding women and patients with abnormal laboratory test results, cancer or cancer history, a history of autoimmune or inflammatory diseases other than Sjögren's syndrome, severe infection, or infection with herpes zoster virus, hepatitis B or C viruses, tuberculosis (TB), or human immunodeficiency virus. The detailed exclusion criteria are listed in [Supplementary Material, https://links.lww.com/CM9/B406].

Randomization and blinding

Eligible participants were randomly assigned at a 1:1:1 ratio to one of the three groups: continuation group (tofacitinib, 5 mg, twice daily); dose reduction group (tofacitinib, 5 mg, daily); and withdrawal group. Stratification was not performed. The allocation of patients was based on a computer-generated randomization sequence. Each patient was identified by a unique coded number linked to group identification. The study was an open-label trial, and so neither the investigators nor the patients were blinded.

Procedures

After randomization, the participants were required to take a stable dose of tofacitinib and concomitant drugs for up to 6 months; the dose could be adjusted only under the investigators’ guidance. If any patient in the reduction and withdrawal groups was unable to maintain LDA, tofacitinib was re-administered (5 mg twice daily) to control their disease, and the follow-up was deemed complete. Any participant that arbitrarily altered the dose during the 6 months was excluded from the study. Discontinuation due to AEs was recorded.

Participants were followed up for up to 6 months. Physical examinations and laboratory tests were conducted at every visit, and efficacy outcomes were assessed with a case report form at the baseline and after 1, 3, and 6 months. During the severe pandemic of Corona Virus Disease 2019 in China, regular follow-ups were also done by telephone or chat software, allowing the patients to report any discomfort. Patients unable to come to Shanghai were asked to consult rheumatologists in local hospitals for their assessments, including physical examination, which were documented in their medical records. Subsequently, the examination results were sent to the study investigators.

Outcomes

The primary endpoint was the proportion of patients maintaining LDA based on DAS28-ESR ≤3.2 in each group at 6 months. The secondary endpoints were the proportion of patients with LDA based on a simplified disease activity index (SDAI) ≤11, remission based on a DAS28 <2.6, and remission based on an SDAI ≤3.2. The proportion of patients with a normal score in the health assessment questionnaire disability index (HAQ-DI ≤0.5) was also assessed. Mean changes from baseline in the DAS28-ESR, DAS28-C-reactive protein (CRP), clinical disease activity index (CDAI), SDAI, swollen joint counts (SJCs), tender joint counts (TJCs) in 28 joints, level of CRP (g/L), ESR (0–100 mm/h), physician global assessments (PGAs, 0–10 scale), patient global assessments (PtGAs, 0–10 scale), patient-assessed pain visual analog scales (VAS, 10 mm), and HAQ-DI scores were all determined during follow-up. Time to loss of efficacy (ie, flare-up, defined as a repeat DAS28-ESR of >3.2) was assessed and compared using survival analyses. Patient-reported outcomes, including HAQ-DI and the EuroQol five dimensions total index, were evaluated. Predictors of patients’ maintenance of LDA after reduction or withdrawal of tofacitinib were assessed using Cox regression analysis.

Safety endpoints comprised all AEs, including common AEs associated with tofacitinib, AEs of special interest (such as herpes zoster and TB), serious AEs (defined as serious infection or other life-threatening condition that warranted hospital admission), any other recorded AEs, and discontinuations because of AEs.

Statistical analyses

It was calculated that a sample size of 105 patients would yield 90% power at the significance level of 0.05. This calculation was based on the conservative estimate that 90% of the participants in the continuation group would maintain LDA, with 65% maintaining LDA in both the reduction and withdrawal groups. These assumptions were based on a flare-free survival analysis performed in a previous study.^[19] Given the potential drop-out rate, the sample size was increased to 126, as indicated in the approved protocol.

Demographic and baseline disease characteristics were summarized with descriptive statistics. Patients who deviated early from the allocated treatment due to flare-ups were regarded as non-responders for all time points. The primary and secondary efficacy endpoints were assessed with χ² tests for categorical variables. The comparison of clinical outcomes at different follow-up time points was analyzed using the generalized estimating equations model. Time to loss of efficacy was also assessed. The Kaplan–Meier method was used to estimate flare-free survival, and the log-rank test was used to compare the differences. In the Kaplan-Meier analysis, data of patients who remained flare-free for 6 months, were lost to follow-up, or discontinued because of AEs or a protocol violation were censored. Univariate and multivariate Cox regression analyses were performed to identify the prognostic factors for the persistent discontinuation of tofacitinib. The analyses were carried out according to the intention-to-treat principle. All statistical calculations were done with GraphPad Prism, version 8.0.1 (GraphPad Software, San Diego, CA), or SPSS, version 20 (IBM Corp, Armonk, NY, USA).

Results

Patients and disposition

Figure 1 shows the study flow diagram. Between July 1, 2019 and January 18, 2022, 122 eligible patients were enrolled and subjected to randomization. Forty-one were assigned to the continuation group (tofacitinib, 5 mg twice daily), 42 were assigned to the reduction group (tofacitinib, 5 mg daily), and 39 were assigned to the withdrawal group. The sample size was comparable to that predicted in the study protocol. Demographic and disease characteristics were similar in all three groups at baseline [Table 1].

Figure 1: Trial profile of eligible rheumatoid arthritis patients who were taking tofacitinib.

Table 1 - Baseline characteristics of eligible RA patients who were taking tofacitinib.

Items	Total (n = 122)	Continuation group (n = 41)	Reduction group (n = 42)	Withdrawal group (n = 39)
Demographic characteristics
Age (years)	50.3 (12.9)	51.3 (10.5)	45.7 (13.8)	54.0 (13.0)
Female	104 (85.2)	36 (87.8)	36 (85.7)	32 (82.1)
Disease characteristics
Disease duration (years)	3.5 (3.9)	3.9 (4.7)	2.6 (2.7)	4.1 (3.9)
RF positive	101 (82.8)	36 (87.8)	36 (85.7)	29 (74.4)
Anti-CCP positive	108 (88.5)	37 (90.2)	38 (90.5)	33 (84.6)
Previous drug use
csDMARD excluding MTX	54 (44.3)	21 (51.2)	16 (38.1)	17 (43.6)
bDMARD	16 (13.1)	9 (22.0)	4 (9.5)	3 (7.7)
GCs	25 (20.5)	13 (31.7)	8 (19.0)	4 (10.3)
Concomitant drug use#
MTX use	98 (80.3)	32 (78.0)	32 (78.0)	33 (84.6)
MTX dose (mg/week)	11.5 (2.5)	11.3 (2.5)	11.7 (2.5)	11.7 (2.5)
GC use	30 (24.6)	11 (26.8)	8 (19.0)	11 (28.2)
GC dose (mg/day)	6.0 (1.9)	5.9 (1.9)	6.1 (1.8)	6.0 (2.1)
Clinical characteristics
DAS28-ESR	2.2 (0.7)	2.3 (0.6)	2.1 (0.6)	2.2 (0.7)
DAS28 LDA (≤3.2)	122 (100)	41 (100)	42 (100)	39 (100)
DAS28 remission (<2.6)	86 (70.5)	30 (73.2)	31 (73.8)	25 (64.1)
SDAI	8.3 (7.6)	5.9 (5.9)	10.3 (9.4)	8.8 (6.6)
SDAI LDA (≤11)	93 (76.2)	36 (87.8)	29 (69.0)	28 (71.8)
SDAI remission (<3.3)	35 (28.7)	16 (39.0)	11 (26.2)	8 (20.5)
DAS28-CRP	1.1 (0.7)	1.1 (0.7)	1.0 (0.6)	1.1 (0.8)
CDAI	2.2 (2.3)	2.6 (2.7)	1.9 (1.9)	2.2 (2.2)
TJC (28 joints)	0.5 (0.9)	0.7 (1.0)	0.4 (0.8)	0.5 (0.8)
SJC (28 joints)	0.3 (0.5)	0.3 (0.5)	0.2 (0.5)	0.3 (0.5)
ESR (mm/h)	15.0 (8.1)	14.1 (7.1)	15.4 (8.0)	15.6 (9.3)
CRP (mg/L)	2.5 (3.5)	2.5 (4.8)	2.5 (2.7)	2.4 (2.8)
PtGA	0.8 (0.8)	0.9 (1.0)	0.7 (0.6)	0.8 (0.8)
PGA	0.6 (0.7)	0.7 (0.9)	0.6 (0.6)	0.6 (0.7)
Pain (VAS)	0.7 (0.8)	0.7 (0.9)	0.6 (0.7)	0.8 (0.8)
Patient-reported characteristics
Total HAQ-DI (0–3 scale∗)	0.1 (0.2)	0.0 (0.1)	0.1 (0.2)	0.0 (0.1)
Normal HAQ-DI (≤0.5)	117 (95.9)	40 (97.6)	39 (92.9)	38 (97.4)
EQ-5D (0–1 scale†)	0.8 (0.0)	0.8 (0.1)	0.8 (0.0)	0.8 (0.0)

Data are mean (SD) or n (%).

^∗Lower score denotes less functional disability.

^†Higher score indicates better quality of life. bDMARD: Biological disease-modifying antirheumatic drug; CCP: Cyclic peptide containing citrulline.

^#Average dose of MTX and GC was calculated only among patients who were taking them. CDAI: Clinical disease activity index; CRP: C-reactive protein; csDMARD: Conventional synthetic disease-modifying antirheumatic drug; DAS28-ESR: Disease activity score in 28 joints based on erythrocyte sedimentation rate; EQ-5D: EuroQol five dimensions total index; ESR: Erythrocyte sedimentation rate; GC: Glucocorticoid; HAQ-DI: Health assessment questionnaire disability index; LDA: Low disease activity; MTX: Methotrexate; PGA: Physician global assessment; PtGA: Patient global assessment; RA: Rheumatoid arthritis; RF: Rheumatoid factor; SDAI: Simplified disease activity index; SJC: Swollen joint count; TJC: Tender joint count; VAS: Visual analog scales.

Withdrawal of tofacitinib leads to a more rapid and significant loss of efficacy

During follow-up after the randomization, the proportion of patients with DAS28-ESR remission/LDA decreased rapidly in the withdrawal group, while only a modest decrease was noted in the reduction group, and no change was seen in the continuation group [Figure 2]. A majority of patients in the withdrawal group experienced a loss of efficacy [Table 2 and Figure 2]. At month 6, only 8 of the 39 patients (20.5%) in the withdrawal group maintained DAS28-ESR LDA, much fewer than the 27 out of 42 (64.3%) patients in the reduction group and 39 out of 41 (95.1%) patients in the continuation group (P < 0.0001 for both comparisons) [Table 2 and Figure 2]. The proportion of patients with SDAI remission/LDA did not change significantly from baseline in the continuation and reduction groups but decreased in the withdrawal group [Table 2 and Supplementary Figure 1, https://links.lww.com/CM9/B406]. Fewer patients had normal HAQ-DI scores in the withdrawal group than in the continuation and reduction groups [Table 2]. A subgroup analysis in patients fulfilling the definition of remission (DAS28-ESR <2.6) at baseline suggested that the effect of a more stringent criterion for withdrawal/de-escalation (ie, remission) on the primary outcomes after 6 months was similar to that of a less stringent criterion (i.e., LDA) [Table 2].

Figure 2: Proportions of patients achieving LDA/remission based on DAS28-ESR at months 1, 3, and 6. All patients had DAS28-ESR ≤3.2 at baseline. Upper bars (lighter color) indicate LDA (DAS28-ESR ≤3.2); lower bars (darker color) indicate remission (DAS28-ESR <2.6). Numbers within the bars indicate the exact percentage. ^∗Withdrawal group vs. continuation group, P < 0.0001. ^†Withdrawal group vs. reduction group, P < 0.0001. ^‡Withdrawal group vs. reduction group, P = 0.0002. DAS28-ESR: Disease activity score in 28 joints based on erythrocyte sedimentation rate; LDA: Low disease activity.

Table 2 - Treatment efficacy at month 6 of RA patients.

				P values
Endpoints	Continuation group	Reduction group	Withdrawal group	Withdrawal vs. continuation	Withdrawal vs. reduction
For all patients	n = 41	n = 42	n = 39
DAS28-ESR LDA (≤3.2)	39 (95.1)	27 (64.3)	8 (20.5)	<0.0001	<0.0001
DAS28-ESR remission (<2.6)	34 (82.9)	25 (59.5)	6 (15.4)	<0.0001	<0.0001
SDAI LDA (≤11)	39 (95.1)	27 (64.3)	13 (33.3)	<0.0001	0.0054
SDAI remission (<3.3)	24 (58.5)	17 (40.5)	5 (12.8)	<0.0001	0.0052
Normal HAQ-DI (≤0.5)	40 (97.6)	34 (81.0)	24 (61.5)	<0.0001	0.0529
For patients fulfilling remission	n = 30	n = 31	n = 25
DAS28-ESR LDA (≤3.2)	28 (93.3)	24 (77.5)	6 (24.0)	<0.0001	0.0002
DAS28-ESR remission (<2.6)	26 (86.7)	23 (74.2)	5 (20.0)	<0.0001	0.0002
SDAI LDA (≤11)	28 (93.3)	23 (74.2)	16 (64.0)	0.0178	0.5944
SDAI remission (<3.3)	18 (60.0)	16 (51.6)	4 (16.0)	0.0024	0.0130
Normal HAQ-DI (≤0.5)	30 (100.0)	29 (93.6)	17 (68.0)	0.0009	0.0331

Data are n/N (%). DAS28-ESR: Disease activity score in 28 joints based on erythrocyte sedimentation rate; HAQ-DI: Health assessment questionnaire disability index; LDA: Low disease activity; SDAI: Simplified disease activity index.

The mean DAS28-ESR, DAS28-CRP, SDAI, and CDAI scores significantly deteriorated in the withdrawal group compared with the continuation and reduction groups from month 1 onward [Tables 2 and 3]. Patients assigned to the continuation and reduction groups continued to have sustained disease control, whereas those who ceased taking tofacitinib experienced moderate disease activity. Statistically significant increases in clinical outcomes, including TJC, SJC, ESR, CRP, PtGA, PGA, and VAS of pain, were also observed after tofacitinib withdrawal compared with the drug reduction or continuation [Table 3]. Patient-reported outcomes were also significantly worse in the withdrawal group [Tables 2 and 3].

Table 3 - GEE analysis (DAS28-ESR, DAS28-CRP, SDAI, CDAI, ESR, CRP, TJC, SJC, PtGA, PGA, pain-VAS, EQ-5D, and HAQ-DI during follow-up by treatment group).

		Month 1		Month 3		Month 6
Items		Mean difference (95% CI)	P values	Mean difference (95% CI)	P values	Mean difference (95% CI)	P values
DAS28-ESR	Withdrawal vs. continuation	1.72 (1.22, 2.22)	<0.0001	2.14 (1.31, 2.97)	<0.0001	2.35 (1.58, 3.13)	<0.0001
	Withdrawal vs. reduction	1.24 (0.64, 1.84)	<0.0001	1.79 (0.86, 2.71)	0.0001	1.62 (0.72, 2.52)	0.0004
DAS28-CRP	Withdrawal vs. continuation	1.72 (1.16, 2.28)	<0.0001	2.22 (1.29,3.15)	<0.0001	2.67 (1.86, 3.47)	<0.0001
	Withdrawal vs. reduction	1.28 (0.61, 1.94)	0.0002	1.85 (0.87, 2.82)	0.0002	1.93 (1.02, 2.84)	<0.0001
SDAI	Withdrawal vs. continuation	7.99 (5.23, 10.76)	<0.0001	7.34 (4.37, 10.32)	<0.0001	4.68 (2.83, 6.52)	<0.0001
	Withdrawal vs. reduction	5.61 (2.27, 8.95)	0.0010	4.89 (1.15, 8.63)	0.0104	2.49 (0.1, 4.87)	0.0409
CDAI	Withdrawal vs. continuation	17.20 (10.05, 24.35)	<0.0001	7.42 (0.12, 16.69)	0.1168	3.02 (0.68, 5.36)	0.0113
	Withdrawal vs. reduction	13.44 (5.53, 21.35)	0.0009	8.00 (0.07, 16.77)	0.0735	0.66 (–2.92, 4.24)	0.7172
ESR (mm/h)	Withdrawal vs. continuation	21.00 (13.36, 28.64)	<0.0001	18.93 (8.45, 29.41)	0.0004	26.64 (16.46, 36.82)	<0.0001
	Withdrawal vs. reduction	14.08 (5.41, 22.75)	0.0015	14.11 (2.78, 25.45)	0.0147	17.53 (6.09, 28.97)	0.0027
CRP (mg/L)	Withdrawal vs. continuation	9.29 (4.19, 14.40)	0.0003	6.26 (–2.66, 15.17)	0.1691	1.48 (–0.26, 3.21)	0.0954
	Withdrawal vs. reduction	7.87 (2.58, 13.16)	0.0036	7.13 (–1.23, 15.48)	0.0944	0.24 (–2.09, 2.58)	0.8381
TJC (28 joints)	Withdrawal vs. continuation	2.00 (1.22, 2.78)	<0.0001	3.48 (1.81, 5.15)	<0.0001	3.06 (1.83, 4.29)	<0.0001
	Withdrawal vs. reduction	1.45 (0.53, 2.36)	0.0019	2.89 (1.07, 4.71)	0.0018	2.10 (0.70, 3.51)	0.0033
SJC (28 joints)	Withdrawal vs. continuation	1.58 (0.86, 2.30)	<0.0001	2.26 (1.10, 3.42)	0.0001	1.49 (0.77, 2.21)	<0.0001
	Withdrawal vs. reduction	1.17 (0.33, 2.01)	0.0066	1.90 (0.68, 3.12)	0.0022	1.14 (0.35, 1.93)	0.0046
PtGA	Withdrawal vs. continuation	2.23 (1.52, 2.95)	<0.0001	2.53 (1.67, 3.39)	<0.0001	3.25 (2.26, 4.25)	<0.0001
	Withdrawal vs. reduction	1.56 (0.69, 2.42)	0.0004	1.99 (0.99,2.98)	<0.0001	2.22 (1.07, 3.38)	0.0002
PGA	Withdrawal vs. continuation	2.16 (1.37, 2.95)	<0.0001	2.60 (1.71, 3.49)	<0.0001	2.57 (1.80, 3.34)	<0.0001
	Withdrawal vs. reduction	1.43 (0.50, 2.37)	0.0026	1.97 (0.95, 2.99)	0.0002	1.52 (0.54, 2.51)	0.0024
Pain (VAS)	Withdrawal vs. continuation	2.46 (1.69, 3.23)	<0.0001	2.79 (1.94, 3.64)	<0.0001	3.40 (2.39, 4.41)	<0.0001
	Withdrawal vs. reduction	1.82 (0.88, 2.76)	0.0002	2.23 (1.27, 3.19)	<0.0001	2.37 (1.19, 3.55)	<0.0001
HAQ-DI	Withdrawal vs. continuation	0.29 (0.15, 0.44)	0.0591	0.43 (0.24, 0.62)	<0.0001	0.44 (0.27, 0.60)	<0.0001
	Withdrawal vs. reduction	0.17 (–0.01, 0.34)	<0.0001	0.35 (0.15, 0.55)	0.0006	0.28 (0.09, 0.48)	0.0036
EQ-5D	Withdrawal vs. continuation	–0.11 (–0.14, –0.07)	<0.0001	–0.13 (–0.18, –0.07)	<0.0001	–0.15 (–0.20, –0.11)	<0.0001
	Withdrawal vs. reduction	–0.06 (–0.10, –0.02)	0.0089	–0.10 (–0.15, –0.04)	0.0008	–0.11 (–0.16, –0.05)	0.0002

Matrix structure is non-structured. Model type is linear. Outcomes were selected as dependent variables, and group and time points were selected as predictors. The interaction between terms were constructed in the model. Multiple comparisons of marginal means were done (pairwise). Given that there exists interaction between group and time, the separate effects of grouping were further analyzed (through EM-MEANS clause). Pairwise comparisons between different groups at each time point were obtained. Mean differences are the means of the withdrawal groups minus that of the continuation/reduction groups. CDAI: Clinical disease activity index; CI: Confidence interval; CRP: C-reactive protein; DAS28-ESR: Disease activity score in 28 joints based on erythrocyte sedimentation rate; EQ-5D: EuroQol five dimensions total index; ESR: Erythrocyte sedimentation rate; GEE: Generalized estimating equations; HAQ-DI: Health assessment questionnaire disability index; PGA: Physician global assessment; PtGA: Patient global assessment; SDAI: Simplified disease activity index; SJC: Swollen joint count; TJC: Tender joint count; VAS: Visual analog scales.

Kaplan-Meier analysis showed significant differences in the survival curves among the treatment groups [Figure 3]. The withdrawal group exhibited the most acute loss of efficacy (log-rank, P < 0.0001, for withdrawal vs. continuation), while in the reduction group, the loss of efficacy was modest (log-rank, P = 0.0039, for reduction vs. continuation, and P < 0.0001 for reduction vs. withdrawal). The mean time of flare-free survival was 2.4 months (95% confidence interval [CI] 1.7–3.1 months) in the withdrawal group, 4.7 months (95% CI 4.1–5.4 months) in the reduction group, and 5.8 months (95% CI 5.5–6.1 months) in the continuation group, indicating that relapse occurred faster in the withdrawal group than in the reduction group [Table 4].

Figure 3: The proportion of patients who remained flare-free during follow-up was visualized with Kaplan–Meier curves. Crosses indicate censoring. Log-rank test was used for multiple comparisons of flare-free survival curves. P < 0.0001 for the comparison of the three treatment groups. ^∗ P < 0.0001 for withdrawal vs. continuation, and P < 0.0001 for reduction vs. withdrawal. ^† P = 0.0039 for reduction vs. continuation.

Table 4 - Time to loss of efficacy of RA patients.

Items	Continuation group (n = 41)	Reduction group (n = 42)	Withdrawal group (n = 39)
Flare-ups, n (%)	4 (9.8)	15 (35.7)	31 (79.5)
Mean time to loss of efficacy (months, 95% CI)	5.8 (5.5–6.1)	4.7 (4.1–5.4)	2.4 (1.7–3.1)
Median time to loss of efficacy (months, 95% CI)	NA	NA	1.0 (0.7–1.3)
Probability of 6 months flare-free survival (%, SE)	88.7 (5.4)	60.9 (8.0)	20.5 (6.5)

All results were estimated using Kaplan–Meier survival analysis. CI: Confidence interval; NA: Not available; SE: Standard error.

Predictors of loss of efficacy

The results of the univariate and multivariate Cox regression analyses are listed in Table 5. Univariate analysis implied that baseline characteristics, including DAS28-ESR ≥2.6, SDAI ≥3.3, SJCs (per one joint more), HAQ-DI >0, and, most importantly, treatment allocation (withdrawal vs. continuation and withdrawal vs. dose reduction), might be associated with the risk of disease flare-up. In the forward stepwise multivariate Cox regression analysis, only treatment allocation (hazard ratio [HR] = 15.752, P < 0.0001 for withdrawal vs. continuation; HR = 3.297, P = 0.0002 for withdrawal vs. dose reduction) and SJC (HR = 2.157, P = 0.0005) were identified as risk factors of disease flare-up.

Table 5 - Predictors for loss of efficacy of RA patients.

	Univariate analysis		Multivariate analysis
Items	HR (95% CI)	P values	HR (95% CI)	P values
Age per 1 year more	1.021 (0.996, 1.045)	0.104	–	–
Male vs. female	1.932 (0.989, 3.775)	0.054	–	–
RA duration per 1 year more	1.038 (0.997, 1.104)	0.228		–
RF positive	0.984 (0.478, 2.026)	0.966	–	–
Anti-CCP positive	0.666 (0.299, 1.480)	0.318	–	–
Previous biologics use	1.269 (0.596, 2.704)	0.536	–	–
DAS28-ESR ≥2.6	2.310 (1.320, 4.044)	0.003	–	–
CDAI ≥2.8	1.579 (0.906, 2.751)	0.107	–	–
SDAI ≥3.3	2.125 (1.032, 4.376)	0.041	–	–
TJC (28 joints) per one joint more	1.216 (0.918, 1.611)	0.173	–	–
SJC (28 joints) per one joint more	2.058 (1.300, 3.259)	0.002	2.157 (1.401, 3.321)	0.0005
PtGA per one point more	1.130 (0.832, 1.535)	0.433	–	–
PGA per 1 point more	1.217 (0.870, 1.702)	0.251	–	–
Pain (VAS) per 1 point more	1.278 (0.940, 1.738)	0.118	–	–
ESR >20	0.898 (0.422, 1.913)	0.781	–	–
CRP >0.5	1.269 (0.701, 2.300)	0.431	–	–
HAQ-DI >0	2.163 (1.203, 3.890)	0.010	–	–
Concomitant MTX per 2.5 mg more	1.082 (0.935, 1.252)	0.291	–	–
Concomitant GC per 2.5 mg more	1.152 (0.917, 1.446)	0.224	–	–
Withdrawal vs. continuation	14.390 (5.048, 41.020)	<0.0001	15.752 (5.493, 15.752)	<0.0001
Withdrawal vs. dose reduction	3.404 (1.824, 6.351)	0.0001	3.297 (1.755, 6.195)	0.0002

Variables with P values <0.05 in the univariate Cox regression analysis were entered into a forward stepwise multivariate Cox regression model with a cut-off significance level of 0.05, to predict the risk factors of relapse after reduction/withdrawal. The two variables for treatment allocation were included as mandatory variables. Dependent variable = relapse. Independent variables = disease and treatment factors. CCP: Cyclic peptide containing citrulline; CDAI: Clinical disease activity index; CI: Confidence interval; CRP: C-reactive protein; DAS28-ESR: Disease activity score in 28 joints based on erythrocyte sedimentation rate; ESR: Erythrocyte sedimentation rate; GC: Glucocorticoid; HAQ-DI: Health assessment questionnaire disability index; HR: Hazard ratio; MTX: Methotrexate; PGA: Physician global assessment; PtGA: Patient global assessment; RA: Rheumatoid arthritis; RF: Rheumatoid factor; SDAI: Simplified disease activity index; SJC: Swollen joint count; TJC: Tender joint count; VAS: Visual analog scales.

Safety

The safety information is summarized in Table 6. The incidence rate of AE in the follow-up period was low, and AEs were less frequent in the dose reduction group (7 of 42 [16.7%]) than in the continuation group (10 of 41 [24.4%]), although this difference was not statistically significant. The most frequent AEs were hyperlipidemia (5 of 41 [11.9%] in the continuation group, 6 of 42 [14.3%] in the reduction group, and 2 of 39 [5.1%] in the withdrawal group) and upper respiratory tract infection (2 of 41 [5.0%] in the continuation group and 2 of 42 [4.8%] in the reduction group). Serious AE, AE that led to discontinuation, and serious infection all referred to a single case of cholecystitis with pneumonia that occurred in the continuation group.

Table 6 - Safety summary of RA patients.

Items	Continuation group (n = 41)	Reduction group (n = 42)	Withdrawal group (n = 39)
Adverse events	17 (41.5)	13 (31.0)	3 (7.7)
Patients with adverse events	14 (34.1)	10 (23.8)	3 (7.7)
Patients with serious adverse events	1 (2.4)	0 (0)	0 (0)
Discontinuation because of adverse events	1 (2.4)	0 (0)	0 (0)
Common adverse events
Upper respiratory tract infection	2 (5.0)	2 (4.8)	0 (0)
Abnormal liver function (ALT or AST increase)	3 (7.3)	0 (0)	0 (0)
Temporary hemoglobin decrease	0 (0)	1 (2.4)	0 (0)
Temporary platelet decrease	0 (0)	1 (2.4)	0 (0)
Headache	1 (2.4)	0 (0)	0 (0)
Palpitation	1 (2.4)	0 (0)	0 (0)
Short of breath	1 (2.4)	0 (0)	0 (0)
Hyperlipidemia∗	5 (11.9)	6 (14.3)	2 (5.1)
Hypertension	0 (0)	1 (2.4)	0 (0)
Adverse events of special interest
Serious infections	1 (2.4)	0 (0)	0 (0)
Cholecystitis	1 (2.4)	0 (0)	0 (0)
Pneumonia	1 (2.4)	0 (0)	0 (0)
Herpes zoster†	1 (2.4)	1 (2.4)	0 (0)
TB	0 (0)	0 (0)	0 (0)
Malignancy	0 (0)	0 (0)	0 (0)
Opportunistic infections	1 (2.4)	0 (0)	0 (0)
Herpes simplex	1 (2.4)	0 (0)	0 (0)
Pulmonary embolism or deep vein thrombosis	0 (0)	0 (0)	0 (0)
Major adverse cardiovascular event	0 (0)	0 (0)	0 (0)
Gastrointestinal perforation or obstruction	0 (0)	0 (0)	0 (0)
Other adverse events recorded
Numbness of hands	0 (0)	1 (2.4)	0 (0)
Hyperuricemia	0 (0)	0 (0)	1 (2.6)

Data are n (%). Includes all events during the follow-up.

^∗Hyperlipidemia here means a mild increase in triglycerides or low-density lipoprotein with no effect on the cardiovascular system. Most patients did not receive drug treatment, and some patients received lipid-lowering drugs.

^†All herpes zoster events reported involved one or two adjacent dermatomes and were non-serious. ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; RA: rheumatic arthritis; TB: Tuberculosis.

Discussion

In this multicenter randomized trial, we demonstrated that in comparison with the reduction of tofacitinib dose to 5 mg daily, withdrawal of the drug in RA patients who achieved sustained LDA led to an increase in disease activity to a great extent. A larger fraction of patients in the withdrawal group (79.5%) than in the reduction group (35.7%) experienced flare-ups as defined by DAS28-ESR LDA. At all time points, the continued use of tofacitinib resulted in more favorable clinical outcomes than withdrawal. The mean time to the loss of efficacy due to withdrawal was significantly shorter than that due to dose reduction.

The loss of disease control in the majority of the patients who stopped taking tofacitinib indicated that sustained LDA/remission is a state that requires sufficient maintenance of JAK inhibitors. Inhibition of the JAK-signal transducer and activator of transcription (STAT) pathway by tofacitinib blocks the common downstream signal transduction pathways of a variety of inflammatory cytokines.^[21] Therefore, when the downstream block is removed, the upstream pro-inflammatory effects can quickly restart, which explains the onset of flare-ups after the reduction or withdrawal of tofacitinib.

Previous studies showed that flare-ups frequently occur after tapering or withdrawing synthetic, biologic, or targeted synthetic DMARDs in RA patients in LDA or remission. Systemic analysis showed that only about one-third of RA patients with LDA or in remission retained disease control within 1 year after drug tapering or withdrawal.^[11] For example, reducing the dose of anti-TNF biologics appears to result in lower flare-up rates than withdrawal, making it a preferable strategy.^[11] Guidelines developed in recent years also recommend DMARD tapering rather than cessation for patients who have achieved sustained disease control.^[4] As for tsDMARDs, the first rigorous, blinded, randomized clinical trial concerning tapering strategies that was conducted for baricitinib revealed that dose reduction resulted in a small increase in disease activity and more frequent instances of relapse. However, similar to the findings of our trial, the mean disease activity indices remained low.^[10] Before the present investigation was completed, a non-randomized observational study reported that dose reduction of tofacitinib was preferable to immediate withdrawal in RA patients, a finding that is also consistent with our results.^[19]

Compared to anti-TNF biologics, flare-ups after dose reduction or withdrawal of tofacitinib appeared to be more frequent and occurred sooner. This could be attributed to the much longer half-life of anti-TNF biologics than that of tofacitinib and, thus, a longer-lasting residual effect.^[22,23] On the other hand, previous studies showed that restoring tofacitinib to 5 mg twice daily helped to regain disease control within 1 month.^[19,20] In the present trial, we did not follow up all the participants after their dose rescue, as their follow-up were deemed complete in the protocol. However, we did perform some post hoc follow-up, which confirmed that returning to tofacitinib 5 mg twice daily rapidly returned LDA/remission, mostly within 1 or 2 weeks. Some patients had restarted a once-a-day dose of tofacitinib after their flare-ups, which acquired satisfactory outcomes.

As mentioned earlier, several studies had already reported the outcomes of reducing the tofacitinib dose or withdrawal strategies before our trial was completed.^[19,20,24] To some extent, our investigation partly remedied the limitations of these studies, such as a small sample size, lack of randomization, or unbalanced grouping. Our study is a randomized clinical trial concerning dose reduction and withdrawal of tofacitinib in patients with RA who have achieved remission or LDA.

The current work documented that the withdrawal of tofacitinib was a stronger predictor of flare-up than its continuation or dose-reduction. Besides, according to the forward stepwise multivariable Cox regression model, the SJC at baseline was a predictor of loss of efficacy after tapering or withdrawing tofacitinib. Concomitant use of GCs or MTX did not influence the incidence of flare-ups, a finding consistent with the reported outcomes of a non-inferiority study showing that MTX is not associated with the efficacy of tofacitinib in RA patients who achieved LDA.^[24]

Certain limitations of our trial have to be acknowledged. First, the study was not blinded; while the absence of masking allowed us to deal expeditiously with flare-ups after reduction/withdrawal, it may have led to bias. Second, radiographic assessments were precluded since the effects of tofacitinib on radiographic findings may not be obvious within 6 months.^[16] Thus, we cannot provide data on potential structural implications. Also, we did not perform a stratified randomization, which is quite common for multicenter trials. In addition, although all the patients had sustained LDA/remission for at least 3 months, each patient had been taking tofacitinib for a different length of time; we cannot exclude the influence of these differences on the outcomes. In fact, it will be very meaningful to establish how long patients have to take tofacitinib to avoid flare-ups upon drug withdrawal. Finally, since a step-down strategy is preferable to immediate withdrawal, an important question that remains to be answered is whether it is possible to stop the drug after reducing it for a certain period and how long that period should be. Future clinical trials may answer these questions.

In conclusion, this multicenter, randomized, open-label study demonstrated that, in a majority of RA patients with sustained disease control induced with tofacitinib 5 mg twice daily, withdrawal resulted in a rapid and dramatic increase in disease activity. In contrast, tapering the dose of tofacitinib to 5 mg daily led to more favorable clinical outcomes. Overall, the findings suggest that patients who respond well to tofacitinib could have their dose halved without a significant loss of response, which has substantial implications for the cost-effectiveness of tofacitinib therapy.

Acknowledgements

We sincerely thank the researchers in each sub-center for the enrollment of participants and data collection. We thank all the participants for taking part in this study and cooperating with the follow-up and assessment. We thank Jieyu Gu (formerly at Ruijin Hospital, Shanghai) for her contribution in application of the ethics review. We thank Jian Li (statistician from Clinical Research Center of Ruijin Hospital, Shanghai), Huihui Chi, Jianfen Meng, Xiaoqin He, and Zeping Qiu (all at Ruijin Hospital, Shanghai) for their advice on statistical methods.

Funding

This study was sponsored by Pfizer. Pfizer employees helped to collect some of the raw data, but the company was not involved in the study design, data analyses or interpretation, or the writing or publication of the report. The authors completed the research progress report provided by Pfizer at set intervals. The first and corresponding authors had full access to all data from the study and had final responsibility for the decision to submit the manuscript for publication.

Conflicts of interest

None.