Tofacitinib in rheumatoid arthritis: a German real-world study with focus on treatment changes, effectiveness and patient-reported outcomes (ESCALATE-RA)

Abstract

Objective

To identify predictors of treatment changes and to evaluate the effectiveness and patient-reported outcomes (PROs) in patients with rheumatoid arthritis (RA) initiating tofacitinib in a real-world setting.

Design

The non-interventional study ESCALATE-RA included 1518 patients with RA from Germany. RA treatment, including all changes in therapy, was documented for 24 months starting from the initial intake of tofacitinib.

Participants

All patients started with tofacitinib therapy, either as monotherapy or in combination with methotrexate (MTX).

Primary and secondary outcome measures

The impact of several factors of interest on the number and timing of treatment changes was assessed as primary outcome using Cox proportional hazards models. Further outcomes were tofacitinib drug survival and the use of follow-up disease-modifying antirheumatic drugs after first treatment change. We also assessed the effectiveness, concomitant glucocorticoid (GC) use, PROs (such as functional ability, patient satisfaction, pain and quality of life) and safety. Analyses were based on observed data.

Results

‘Lack of efficacy’ (HR 3.30) and ‘intolerance’ (HR 4.43) leading to termination of tofacitinib were key factors favouring therapy changes. Higher patient satisfaction was significantly associated with a reduced likelihood of treatment changes (HR 0.82). Increasing GC doses were associated with a higher probability of step-up/switch changes (HR 1.21). The estimated tofacitinib drug survival was 48% at the end of study. Proportions of patients achieving low disease activity (both Simplified Disease Activity Index (SDAI) and Clinical Disease Activity Index (CDAI) Δ62%) and remission (SDAI Δ25%, CDAI Δ28%) increased from baseline under tofacitinib and were comparable between monotherapy and combination therapy with MTX. Mean concomitant GC dose decreased (2 mg/day). PROs indicated reduced pain and fatigue, while functional ability and quality of life improved. 63.9% of the patients experienced a treatment-emergent adverse event (AE), 8.8% a treatment-emergent AE of special interest and deaths occurred in 0.5%.

Conclusion

Key factors for therapy changes in patients with RA treated with tofacitinib were lack of efficacy and intolerance. Higher patient satisfaction was associated with a reduced probability of treatment changes, while increased GC doses led to a higher likelihood of step-ups/switches. Patients demonstrated a marked reduction in disease activity for up to 24 months, along with improvements in functional ability, pain and quality of life. Observed AEs were consistent with the known safety profile of tofacitinib.

Trial registration number

NCT03387423.

Strengths and limitations of this study.

• Prospective observational study across multiple sites in Germany.

• Longitudinal assessment of predictors of treatment changes and drug survival over a period of 24 months.

• Reliance on real-world clinical data resulted in certain data gaps, potentially affecting the interpretation of the observed effects.

Introduction

Rheumatoid arthritis (RA) is a chronic inflammatory disease that continues to be associated with a significant disease burden and increased mortality despite advances in treatment.¹ According to current European Alliance of Associations for Rheumatology guidelines, a treat-to-target (T2T) approach is recommended, aiming for remission or low disease activity (LDA).^{2 3} This strategy involves regular monitoring of the disease activity and timely adjustment, when treatment targets are not met.

Even though disease activity is a key driver of treatment changes, real-world studies suggest that T2T recommendations are not consistently followed. For example, a study involving 1301 patients with early arthritis found that despite not having achieved Disease Activity Score 28 (DAS28) remission within 6 months, 54% of the patients did not have their treatment changed.⁴ Similarly, a real-world analysis based on data from the US real-world CorEvitas RA registry (n=2282) indicated that 61% of the patients continued their initial RA treatment at 12 months, despite inadequate response at month 6 based on Clinical Disease Activity Index (CDAI) score ≤10.⁵ These observations suggest that in a real-world setting, T2T recommendations are insufficiently followed, and treatment decisions may be influenced by additional factors such as patient satisfaction and preferences, comorbidities and tolerability.

Tofacitinib, a Janus kinase (JAK) inhibitor, has been well studied in randomised controlled trials (RCTs) demonstrating efficacy and safety both as monotherapy and in combination with methotrexate (MTX).6,11 However, RCTs are conducted in controlled environments with selected patient populations and may not fully reflect routine clinical care. Real-world evidence (RWE) aims to evaluate whether these data from RCTs reflect everyday clinical practice.¹² Furthermore, it provides valuable complementary findings by capturing treatment patterns and changes in a heterogeneous population.

Therefore, we conducted ESCALATE-RA, one of the largest prospective, non-interventional studies of tofacitinib in Germany. The primary objective of this study was to identify predictors for treatment changes in routine care for patients with RA initiating tofacitinib. Secondary objectives included treatment effectiveness, patient-reported outcomes (PROs) and safety over a period of 24 months for tofacitinib monotherapy and combination therapy with MTX.

Methods

Patients and study design

ESCALATE-RA was a prospective, non-interventional study conducted in patients with RA enrolled in multiple German sites. All treatment decisions were made prior to study inclusion. The target population comprised adult patients (≥18 years) with confirmed RA diagnosis, eligible for tofacitinib therapy according to the summary of product characteristics (SmPC¹³). Patients could be enrolled in this study when they signed an informed consent form and were permitted to have received any prior RA medications except from JAK inhibitors. Exclusion criteria included contraindications according to the SmPC, treatment with an investigational drug within the last 3 months and any personal involvement with the sponsor or study staff.

We observed drug treatment and treatment outcomes in routine medical care. Patient data were recorded from the date of first tofacitinib intake and followed up for 24 months (M24), also comprising data of patients who switched from the initial tofacitinib therapy to other therapies. Patients started with either tofacitinib monotherapy (without MTX) or combination therapy (tofacitinib+MTX). Data were collected during nine routine visits (baseline (BL) and every 3 months thereafter) over the course of the 24-month observation period and recorded on an electronic case report form and on patient questionnaires.

The study aimed to enrol 1500 patients in Germany over a recruitment period of 36 months. Data collection started on 2 November 2017 and was completed on 17 July 2023. This study was conducted in accordance with the ethical principles in the Declaration of Helsinki and was registered on ClinicalTrials.gov (NCT03387423).

Patient and public involvement

Patients were not involved in the design, conduct, reporting or dissemination plans of this study. In contrast, healthcare professionals in Germany have been involved in the design and discussion of interim and final study results, which have been presented at public congresses of major rheumatology professional organisations.

Primary and secondary outcomes

The primary outcomes were the impact of predictors and potential confounders (see the Statistical analyses section) on the number and timing of treatment changes in patients with RA who started tofacitinib in a real-world setting. In the context of this study, monotherapy was defined as tofacitinib monotherapy, and combination therapy was defined as tofacitinib in combination with MTX. A treatment change was defined as any change in disease-modifying antirheumatic drug (DMARD) therapy, when compared with the last visit, that is, a switch to another DMARD, combination of DMARDs or termination of DMARD therapy. The following types of treatment changes as primary outcome variables were distinguished at the first treatment change (table 1):

Table 1. Type of first treatment change with regard to DMARD class.

Type of treatment change	Change from	Change to
Step-up/switch	Monotherapy (tofa)	Monotherapy or combination therapy* with (different) ts/bDMARD
	Combination therapy (tofa+MTX)	Monotherapy or combination therapy* with different cs/tsDMARD or bDMARD
Step-down	Monotherapy (tofa)	Monotherapy csDMARD
		Combination therapy with two or more csDMARDs
	Combination therapy (tofa+MTX)	Monotherapy with tofa
		Monotherapy csDMARD
		Combination therapy with two or more csDMARDs
Treatment termination	All mentioned above	Treatment stop

^*In combination with csDMARD.

bDMARD, biological DMARD; csDMARD, conventional synthetic DMARD; DMARD, disease-modifying antirheumatic drug; MTX, methotrexate; tofa, tofacitinib; tsDMARD, targeted synthetic DMARD.

• Step-up/switch—for example, a switch from tofacitinib monotherapy to a combination therapy of biological DMARD+conventional synthetic DMARD (bDMARD+csDMARD).

• Step-down—a de-escalation from the current treatment regime, for example, the de-escalation from tofacitinib monotherapy to monotherapy with csDMARD or to a combination of two or more csDMARDs.

• Termination—the termination of a DMARD without starting a new DMARD therapy.

Further outcome parameters comprised the following:

• Tofacitinib drug survival—the probability of still being on tofacitinib at the end of the study.

• Follow-up DMARDs after first treatment change (post hoc).

• Effectiveness—LDA and remission as assessed by Simplified Disease Activity Index (SDAI; ≤11 and ≤3.3) and CDAI (≤10 and ≤2.8).

• Concomitant steroid use (post hoc).

• PROs:

  • Duration of morning stiffness.

  • Hannover Functional Ability Questionnaire (FFbH).

  • Quality of life (EuroQoL 5-Dimension).

  • Functional Assessment of Chronic Illness Therapy (FACIT) Fatigue Scale.

  • Patient satisfaction measured on a 5-point Likert scale (from 0=extremely dissatisfied to 4=extremely satisfied as answer to the question ‘How satisfied are you with the drugs that you have received for your arthritis since your last visit?’).

  • Arthritis pain measured on a visual analogue scale (VAS; 0–100 mm; from 0=no pain to 100=most severe pain as answer to the question ‘How severe is your pain today?’).

• Treatment-emergent adverse events (TEAEs) and treatment-emergent adverse events of special interest (TEAESIs; post hoc). Treatment emergent was defined as occurring during tofacitinib treatment including a period of 28 days of drug lag after tofacitinib discontinuation.

Statistical analyses

Approximately 1500 patients were expected to be enrolled. The estimation of the sample size was based on a simulation of recurrent event data using a Cox proportional hazards (CPH) model described by Jahn-Eimermacher et al¹⁴ and the distribution of five factors (below mentioned as predictors except satisfaction) obtained from ORAL Standard study.⁷ The simulation indicated that a sample size of 1000 patients would provide >80% power to detect an effect of 4/5 factors, while 1500 patients would provide sufficient power to detect the effects of all factors, including achieving DAS28 <3.2.

Categorical data were expressed by absolute and relative frequencies with 95% CIs, where appropriate. Percentages by categories were based on the number of patients with non-missing data. Numerical data were summarised as means (SDs) or medians (range, 25–75% quartiles). The analysis of descriptive variables was based on the available data only. Kaplan-Meier estimates were used to obtain survival estimates. Adverse events (AEs) were coded according to the Medical Dictionary for Regulatory Activities version 23.1.

The analysis populations were defined as ‘Safety Analysis Set’ (all patients who received at least one dose of tofacitinib), ‘Full Analysis Set’ (all patients who received at least one dose of tofacitinib and who had at least one post-BL visit) for primary analysis and ‘Secondary Full Analysis Set’ (data of patients while receiving tofacitinib) for secondary endpoints.

Incidence rates (IR) of patients with events in 100 patient-years and exact 95% Poisson CIs (95% CIs) were calculated based on cumulative patient-years of 1764. Cumulative patient-years were calculated as the sum of tofacitinib therapy periods for each patient by: (tofacitinib end date–tofacitinib start date)+1/365.25. If no tofacitinib end date was reported, the last visit was used instead.

A CPH gap time model¹⁵ with the adjustment for recurrent events (Prentice, Williams and Peterson gap time) was employed to account for the possibility of multiple events occurring per patient (ie, treatment changes). The model for primary endpoint analysis included the predictor variables on the rate at which treatment changes occur and potential confounders. Predictor variables were imputed using the temporarily ‘nearest observation’ to a change. If an assessment was missing for a given visit, the value from the nearest visit with a recorded assessment was used for imputation. To determine this, the difference between the visit with the missing assessment and both the preceding and subsequent visits was calculated; the visit with the smaller time difference was selected. If two differences were the same, the earlier assessment was chosen for imputation. Imputation was performed to avoid reduction of the dataset as the Cox model only uses observations with non-missing values across all predictor variables/confounders.

The selection of predictor and confounder variables for the CPH was based on statistical parameters. To avoid overfitting by the high number of 17 potential predictor variables/confounders, a stepwise model selection procedure with forward selection using conservative p value limits was employed, where only factors with p value ≤0.9 entered in the model and only factors with p value ≤0.08 remained in the final model.

HRs and 95% CIs were calculated. A parameter estimate >0 indicated an increased risk for treatment changes, whereas an estimate <0 indicated a decreased risk for treatment changes. The HR showed the effect size; the closer to 1, the smaller the effect. An increase of 1 unit on a VAS ranging from 0 to 100, with higher scores indicating worse outcomes, was associated with a 1% increase in the hazard of treatment change. A clinically relevant increase of 70 points (eg, from 10 to 80) would approximately double the hazard (1.01⁷⁰=2.01). In general, higher VAS scores are associated with a higher probability of treatment change; however, only large increases on the VAS result in a substantially higher probability.

The following predictor variables on the rate of treatment changes of tofacitinib patients were assessed in the primary endpoint analysis in the CPH model:

• Disease Activity Score in 28 joints derived from four measures (DAS28-4) with erythrocyte sedimentation rate (ESR) LDA.

• Change from BL of DAS28-4 (ESR) >1.8, indicating a meaningful change.

• Physician’s Global Assessment of Arthritis measured on a VAS (0–100 mm; from 0=no disease activity to 100=high disease activity).

• Patient’s Assessment of Arthritis Pain measured on a VAS (0–100 mm; from 0=no pain to 100=most severe pain as answer to the question ‘How severe is your pain today?’).

• Patient’s Global Assessment of Arthritis measured on a VAS (0–100 mm; from 0=very well to 100=very poorly as answer to the question ‘Considering all the ways your arthritis affects you, how are you feeling today?’).

• Patient satisfaction with drug treatment measured on a 5-point Likert scale.

Different potential confounders were employed in the CPH gap time model:

• BL characteristics such as age, sex, disease duration, poor prognostic factors, second or third-line treatment (medication history of RA treatment) and incidence of selected comorbidities.

• Indicator of tofacitinib dose (5 mg two times per day or once a day).

• Indicator of dose increase of combination partner.

• Indicator of dose increase of glucocorticoids (GC).

• Indicator of tofacitinib treatment termination due to lack of efficacy or intolerance (ie, efficacy and tolerability issues).

Results

Demography, BL and disease characteristics

Overall, 1518 patients were recruited at 88 study sites; of these, 29 did not start treatment and 30 were retrospectively considered as screening failures. A total of 1459 patients had at least one dose of tofacitinib (online supplemental file S1). Among the 1378 patients with at least one post-BL visit, the mean age was 59 years. The majority was female (76%). The mean disease duration was 10 years (table 2). At BL, the mean disease activity was moderate with 4.7 by DAS28-4 (ESR) and 4.4 by DAS28-4 (C-reactive protein) (table 2). Immediately prior to the start of tofacitinib therapy, 73% of patients had received csDMARDs (1059/1459), and 58% had MTX (846/1459). Overall, 51% of the patients had previously received bDMARDs (750/1459) and 66% (958/1459) GCs (table 2). At BL, 32% (436/1378) were receiving tofacitinib in combination with a csDMARD; 50% (691/1378) of the patients were having a comedication with GCs.

Table 2. Demographic data and baseline disease characteristics.

Parameter	Total (N=1378)
Sex
Female	1047 (76.0%)
Male	331 (24.0%)
Age
n (missing)	1378 (0)
Mean (SD)	59.4 (11.9)
Age category
18–44	156 (11.3%)
45–64	777 (56.4%)
≥65	445 (32.3%)
Weight
n (missing)	1299 (79)
Mean (SD)	79.0 (18.2)
Body mass index
n (missing)	1297 (81)
Mean (SD)	28.2 (6.0)
Body mass index categorised
n (missing)	1297 (81)
>30 kg/m2	396 (30.5%)
≤30 kg/m2	901 (69.5%)
Disease duration (years)
n (missing)	1374 (4)
Mean (SD)	9.9 (9.5)
Moderate or severe disease
n (missing)	1377 (1)
Yes	1305 (94.8%)
ACPA positive
n (missing)	1372 (6)
Yes	698 (50.9%)
Unknown	232 (16.9%)
RF positive
n (missing)	1377 (1)
Yes	835 (60.6%)
Unknown	44 (3.2%)
Presence of poor prognostic factors
Yes	1370 (99.4%)
Smoking status
n (missing)	1341 (37)
Smoker	320 (23.9%)
Ex-smoker	276 (20.6%)
Non-smoker	745 (55.6%)
Herpes zoster vaccination
n (missing)	1375 (3)
Yes	105 (7.6%)
Unknown	575 (41.8%)
Herpes zoster vaccination type
n (missing)	1377 (1)
Inactive live vaccine	1 (1.0%)
Dead vaccine	83 (79.8%)
Unknown	20 (19.2%)
DAS28-4 (ESR)
n (missing)	813 (565)
Mean (SD)	4.70 (1.41)
DAS28-4 (CRP)
n (missing)	836 (542)
Mean (SD)	4.4 (1.3)
	Total (N=1459)
Most common treatments immediately prior to the start of tofacitinib therapy*
csDMARD	1059/1459 (72.6%)
Methotrexate	846/1459 (58.0%)
Leflunomide	358/1459 (24.5%)
Glucocorticoids	958/1459 (65.7%)
Prednisolone	871/1459 (59.7%)
bDMARD	750/1459 (51.4%)
Etanercept (Enbrel)	154/1459 (10.6%)
Adalimumab (Humira)	155/1459 (10.6%)

n denotes sample size.

The denominator used for percentage calculation is the number of patients with a non-missing value.

Poor prognostic factors are severity, increased acute phase protein, high number of swollen joints, positive RF, ACPA positive, combination of factors, early erosions and failure of two or more csDMARDs. Data are based on observed cases in FAS (N=1378).

^*In >10% of the patients, based on SAS (N=1459).

ACPA, anti-citrullinated protein antibodies; bDMARD, biological disease-modifying antirheumatic drug; CRP, C-reactive protein; csDMARD, conventional synthetic disease-modifying antirheumatic drug; DAS28-4, Disease Activity Score in 28 joints derived from four measures; ESR, erythrocyte sedimentation rate; FAS, full analysis set; RF, rheumatoid factor; SAS, Safety Analysis Set.

Key predictors of treatment changes

Overall, 53% (731/1378) of the patients experienced any treatment change. The predominant type of treatment change was step-up/switch (38%, 528/1378), followed by step-down (15%, 210/1378) and DMARD treatment termination (10%, 144/1378; online supplemental file S3).

‘Lack of efficacy’ or ‘intolerance’ leading to termination of tofacitinib were the two factors with the largest effect sizes (HRs) and were considered as key factors favouring any type of treatment change (parameter estimates 1.19 (HR 3.30) and 1.49 (HR 4.43), respectively) (table 3). An increase in GC dose at any time during the study was associated with step-up/switch changes (parameter estimate 0.19 (HR 1.21)) (table 3), and a higher patient satisfaction was associated with a reduced risk of any treatment change (parameter estimate −0.20 (HR 0.82)), step-up/switch (parameter estimate 0.25 (HR 0.78)) or step-down (parameter estimate 0.23 (HR 0.79)) (table 3).

Table 3. Time to treatment change (days)—multivariate Cox proportional hazards model.

Predictor/confounder	Parameter estimate	SE	P value	HR	95% CI limits	n/observations
All changes
Termination due to lack of efficacy (no, yes)	1.19	0.08	<0.0001	3.30	2.82 to 3.86	1378/2210
Physician’s assessment of RA* (0–100)	0.01	0.00	<0.0001	1.01	1.01 to 1.01
Patient satisfaction with treatment* (0–4)	−0.20	0.03	<0.0001	0.82	0.77 to 0.87
Disease duration at BL (years)	−0.01	0.00	0.0242	0.99	0.99 to 1.00
Termination due to intolerance (no, yes)	1.49	0.09	<0.0001	4.43	3.68 to 5.32
Step-up/switch
Termination due to lack of efficacy (no, yes)	1.40	0.11	<0.0001	4.04	3.23 to 5.05	1378/1451
Glucocorticoid dose increase (no, yes)	0.19	0.09	0.0310	1.21	1.02 to 1.44
Physician’s assessment of RA* (0–100)	0.02	0.00	<0.0001	1.02	1.01 to 1.02
Patient satisfaction with treatment* (0–4)	−0.25	0.04	<0.0001	0.78	0.72 to 0.85
Medication history of RA treatment	0.13	0.05	0.0157	1.13	1.02 to 1.25
Termination due to intolerance (no, yes)	1.60	0.13	<0.0001	4.97	3.86 to 6.40
Step-down
Patient satisfaction with treatment* (0–4)	−0.23	0.06	<0.0001	0.79	0.71 to 0.89	1378/1550
Termination due to intolerance (no, yes)	0.97	0.17	<0.0001	2.65	1.89 to 3.70
Termination
Age	0.03	0.01	0.0002	1.03	1.01 to 1.05	1378/1374
Physician’s assessment of RA* (0–100)	0.02	0.00	<0.0001	1.03	1.02 to 1.03
Disease duration at BL (years)	−0.03	0.01	0.0173	0.97	0.95 to 1.00

Only significant predictors and confounders are shown (p<0.08).

Strata (change number) with <15 patients are excluded. Physician’s assessment: scale from 0 (no activity) to 100 (high activity). Patient satisfaction: scale from 0 (extremely dissatisfied) to 4 (extremely satisfied). Termination due to lack of efficacy/intolerance relates to termination of tofacitinib treatment due to an adverse event recorded by the investigator as related to lack of efficacy/intolerance. Based on FAS (n=1378).

^*Imputation with temporarily nearest observation.

BL, baseline; FAS, full analysis set; RA, rheumatoid arthritis.

For DMARD treatment termination, some factors indicated significance; however, these factors only had a minimal impact with an HR close to 1 (table 3). Disease activity by DAS28-4 (ESR) was not significantly correlated with treatment changes.

The median duration until the next treatment change decreased with an increasing number of changes (one, two, three and four changes: 207, 188, 182 and 165 days; online supplemental file S3). Interestingly, the median duration until treatment termination of all DMARDs was longer in the monotherapy group than in the combination therapy group (289 days (range 30–789) vs 138 days (range 92–183); online supplemental file S3).

Tofacitinib drug survival and follow-up DMARDs after first treatment change

The estimated drug survival rate of tofacitinib, that is, the probability of still being on tofacitinib, at the end of the study was 48% (95% CI 45% to 51%), with a mean treatment persistence of tofacitinib at 524 days (figure 1A, online supplemental file S2). Treatment persistence was similar for patients with tofacitinib monotherapy or combination therapy with MTX (figure 1A, online supplemental file S2).

Figure 1. Tofacitinib drug survival and follow-up DMARDs after first treatment change. (A) Tofacitinib drug survival at the end of the study. Kaplan-Meier plot censored at 800 days. Data are presented in online supplemental file S2. (B, C) Follow-up DMARDs after first treatment change in (B) monotherapy (tofacitinib) and (C) combination therapy (tofacitinib+MTX) displayed in ≥1% of patients in each subgroup. Data are based on observed cases in FAS (n=1378). bDMARD, biological DMARD; csDMARD, conventional synthetic DMARD; DMARD, disease-modifying antirheumatic drug; FAS, full analysis set; MTX, methotrexate; TNFi, tumour necrosis factor inhibitor; tofa, tofacitinib; tsDMARD, targeted synthetic DMARD other than tofacitinib.

More than one-third of the patients in monotherapy (35.9%, 255/711) group and combination therapy (37.7%, 153/406) group experienced a first DMARD treatment change. After the first treatment change, the most frequent follow-up DMARDs in the tofacitinib monotherapy group were a different targeted synthetic DMARD (tsDMARD) (33%, 84/255), tumour necrosis factor inhibitor (TNFi) (29%, 75/255) and non-TNFi bDMARD (21%, 53/255) (figure 1B). 12 patients (4.7%) changed from tofacitinib monotherapy to a combination therapy with tofacitinib+csDMARD.

In the combination therapy group, the most common DMARD treatment regimens after the first change were TNFi+csDMARD (26%, 40/153), csDMARD without bDMARD or tsDMARD (24%, 37/153), a different tsDMARD+csDMARD (22%, 34/153) and non-TNFi bDMARD+csDMARD (21%, 32/153) (figure 1C).

Overall, 45% (651/1459) of the patients terminated tofacitinib treatment during the study. The main reason for termination of tofacitinib therapy was ‘lack of efficacy’ (47.0%, 292/651; online supplemental file S2).

Effectiveness

The proportion of patients achieving LDA by SDAI (≤11) was 11% (93/828) at BL, 59% (539/920) at month 3 (M3) and 73% (307/419) at M24. An increase in the proportion of patients achieving LDA was also observed by CDAI (≤10), with 13% (126/1004) at BL, 58% (656/1134) at M3 and 75% (394/527) at M24 (figure 2A,B).

Figure 2. Effectiveness outcomes—LDA and remission. Percentage of patients who achieved (A, B) LDA and (C, D) remission by SDAI (A, C) and CDAI (B, D) for monotherapy (n=725), combination therapy (n=412) and total (n=1378). Data are based on observed cases in SFAS (n=1378), while they were receiving tofacitinib. *Sample size per visit based on total population. BL, baseline; CDAI, Clinical Disease Activity Index; Combi, combination therapy (tofacitinib+MTX); LDA, low disease activity; M3/M12/M18/M24, month 3/12/18/24; Mono, monotherapy (tofacitinib); MTX, methotrexate; SDAI, Simplified Disease Activity Index; SFAS, secondary full analysis set.

The proportion of patients achieving remission by SDAI (≤3.3) was 1% (10/828) at BL, 16% (144/920) at M3 and 26% (107/419) at M24. A similar increase in the proportion of patients achieving remission was also observed by CDAI (≤2.8), with 1% (14/1004) at BL, 17% (195/1134) at M3 and 29% (151/527) at M24 (figure 2C,D).

Effectiveness outcomes were largely comparable in monotherapy and combination therapy groups. The most pronounced increase in LDA and remission rate was observed between BL and M3 visit (figure 2).

Concomitant steroid use

The mean GC use (standardised to prednisone) in patients who were receiving tofacitinib decreased from BL to M24 by −2 (SD 8) mg/day (figure 3). The percentage of patients with concomitant GC use also decreased from BL to M24 (62% to 46%; figure 3). At the end of the study, 38% (217/570) of the patients who were still taking tofacitinib had a decreased mean GC dose, while only 8% (43/570) of the patients had an increased mean GC dose (online supplemental file S4).

Figure 3. GC use. Data are based on observed cases in SFAS (n=1378), data of patients while they were receiving tofacitinib. GC doses were standardised to prednisone. BL, baseline; GC, glucocorticoid; M12/M24, month 12/24; SFAS, secondary full analysis set.

Patient-reported outcomes

A reduction of morning stiffness was observed over the course of the study from BL to M24 visit (mean 60 (SD 74) min to 21 (SD 34) min) (figure 4A). Functional ability in terms of FFbH increased steadily. At BL, mean (SD) FFbH value in per cent of function was 62 (SD 23), indicating an impaired functional ability. FFbH increased to 73 (SD 23) approaching normal functional ability at M24 visit (figure 4B). Arthritis pain, measured by VAS, improved from a mean (SD) of 55 (25) mm at BL to 30 (25) mm at M24 (figure 4E).

Figure 4. Results of PROs. (A) Morning stiffness in minutes. (B) FFbH score as remaining percentage of function. (C) EQ-5D-3L VAS. (D) FACIT Fatigue Scale VAS. (E) Arthritis pain VAS. (F) Patient satisfaction. Data are based on observed cases in SFAS (n=1378), data of patients while they were receiving tofacitinib. BL, baseline; EQ-5D-3L, EuroQoL 5-Dimension 3-Level; FACIT, Functional Assessment of Chronic Illness Therapy; FFbH, Hannover Functional Ability Questionnaire; M12/M24, month 12/24; PROs, patient-reported outcomes; SFAS, secondary full analysis set; VAS, visual analogue scale.

Overall, the quality of life by EuroQoL 5-Dimension 3-Level VAS increased during the study. At BL, the mean (±SD) VAS was <50 mm (47 (SD 22) mm), suggesting a comparably poor overall health status. However, at the end of the study, the total patient population achieved a moderate health status (50–70 mm), with a mean VAS score of 67 (SD 22) mm (figure 4C). In parallel, FACIT total scores increased during the study (mean BL to M24: 30 (SD 11) to 37 (SD 11)), indicating decreased fatigue (figure 4D). The proportion of patients being satisfied or extremely satisfied with their current RA treatment since their last visit increased over the course of the study from 29% (290/993) at BL to 82% (399/486) at M24 (figure 4F).

Safety results

Over 24 months and a cumulative of 1764 patient-years, 64% (933/1459) of the patients experienced TEAEs (during tofacitinib treatment, including a period of 28 days of drug lag after tofacitinib discontinuation). The IR of patients with TEAEs per 100 patient-years was 52.9 (95% CI 49.6 to 56.4). Lack of effectiveness, including the preferred terms (ineffective drug, decreased therapeutic product effect and incomplete therapeutic product effect), was observed in 22% of the patients (325/1459; IR per 100 patient-years: 18.4, 95% CI 16.5 to 20.5). The most common TEAEs were influenza-like illness (3.2, 95% CI 2.4 to 4.2), herpes zoster (2.3, 95% CI 1.6 to 3.1) and nasopharyngitis (1.9, 95% CI 1.3 to 2.7) (table 4).

Table 4. Safety - Frequency and Incidence of TEAEs.

	SAS total (N=1459)
	n/N (%)	Incidence rate per 100 patient-years (95% CI)
Patients with TEAEs
TEAEs	933/1459 (63.9)	52.9 (49.6 to 56.4)
TESAEs	261/1459 (17.9)	14.8 (13.1 to 16.7)
Who discontinued the study drug due to AE but continued the study	323/1459 (22.1)	18.3 (16.4 to 20.4)
Deaths	7/1459 (0.5)	0.4 (0.2 to 0.8)
Most common TEAEs
Lack of effectiveness*	325/1459 (22.3)	18.4 (16.5 to 20.5)
Influenza-like illness	57/1459 (3.9)	3.2 (2.4 to 4.2)
Herpes zoster†	40/1459 (2.7)	2.3 (1.6 to 3.1)
Nasopharyngitis	34/1459 (2.3)	1.9 (1.3 to 2.7)
Most common TESAEs‡
Ineffective drug	31/1459 (2.1)	1.8 (1.2 to 2.5)
Incomplete therapeutic product effect	15/1459 (1.0)	0.9 (0.5 to 1.4)
Osteoarthritis	15/1459 (1.0)	0.9 (0.5 to 1.4)
Hypertensive crisis	11/1459 (0.8)	0.6 (0.3 to 1.1)
Decreased therapeutic product effect	9/1459 (0.6)	0.5 (0.2 to 1.0)
Pneumonia	8/1459 (0.5)	0.5 (0.2 to 0.9)
Herpes zoster	7/1459 (0.5)	0.4 (0.2 to 0.8)
Rheumatoid arthritis	7/1459 (0.5)	0.4 (0.2 to 0.8)
Fall	7/1459 (0.5)	0.4 (0.2 to 0.8)
TEAEs of special interest
Total	129/1459 (8.8)	7.3 (6.1 to 8.7)
Serious infections§	49/1459 (3.4)	2.8 (2.1 to 3.7)
Herpes zoster¶	45/1459 (3.1)	2.6 (1.9 to 3.4)
Neoplasia excluding NMSC	17/1459 (1.2)	1.0 (0.6 to 1.5)
Stroke/TIA**	5/1459 (0.3)	0.3 (0.1 to 0.7)
Myocardial ischaemia††	4/1459 (0.3)	0.2 (0.1 to 0.6)
NMSC‡‡	9/1459 (0.6)	0.5 (0.2 to 1.0)
Deep vein thrombosis§§	8/1459 (0.5)	0.5 (0.2 to 0.9)
Pulmonary embolism¶¶	5/1459 (0.3)	0.3 (0.1 to 0.7)
Tuberculosis	0/1459 (0.0)	–
Other relevant TEAEs
MACE***	10/1459 (0.7)	0.6 (0.3 to 1.0)

Data are based on observed patients in SAS (N=1459). All AEs were treatment emergent including a drug lag of 28 days. Incidence rates are based on 1763.973 patient-years.

N denotes the number of patients. n denotes the number of applicable patients.

^*Including the PTs: ineffective drug, decreased therapeutic product effect, incomplete therapeutic product effect.

^†PT: herpes zoster.

^‡All TESAEs that occurred in ≥0.5% of the total population.

^§TESAEs in system organ class‚ infections and infestations.

^¶PT: herpes zoster, ophthalmic herpes zoster.

^**PT: cerebrovascular accident, ischaemic cerebral infarction, transient ischaemic attack.

^††PT: myocardial infarction, myocardial ischaemia.

^‡‡PT: basal cell carcinoma, skin cancer, Bowen’s disease.

^§§PT: deep vein thrombosis, venous thrombosis limb, thrombosis.

^¶¶PT: pulmonary embolism.

^***MACE includes death from cardiovascular causes, non-fatal myocardial infarction or non-fatal stroke (stroke, transient ischaemic attack, myocardial ischaemia).

AE, adverse event; MACE, major adverse cardiovascular event; NMSC, non-melanoma skin cancer; PT, preferred term; SAS, Safety Analysis Set (all patients who received at least one dose of tofacitinib); TEAE, treatment-emergent adverse event; TESAE, treatment-emergent serious adverse event; TIA, transient ischaemic attack.

Overall, 18% (261/1459) of the patients experienced a treatment-emergent serious adverse event (TESAE), corresponding to an IR of 14.8 per 100 patient-years (95% CI 13.1 to 16.7). The most frequently reported TESAEs were ineffective drug (1.8, 95% CI 1.2 to 2.5), incomplete therapeutic product effect and osteoarthritis (both 0.9, 95% CI 0.5 to 1.4) (table 4).

In total, 9% of the patients (129/1459) had TEAESIs, with an IR of 7.3 (95% CI 6.1 to 8.7) per 100 patient-years. The most common TEAESIs were serious infections (2.8, 95% CI 2.1 to 3.7) and herpes zoster (2.6, 95% CI 1.9 to 3.4). The IR for patients with malignancies (neoplasia excluding non-melanoma skin cancer) was 1.0 in 100 patient-years (95% CI 0.6 to 1.5) and the IR for patients with an event defined as a major adverse cardiovascular event (MACE) (defined as death from cardiovascular causes, non-fatal myocardial infarction or non-fatal stroke) was 0.6 (95% CI 0.3 to 1.0; table 4).

Seven patients had a TESAE and died during tofacitinib treatment. One of these events was considered as related to tofacitinib, namely ‘pneumonia’ with onset about 1.5 months prior to death.

Discussion

This is the first German non-interventional study and one of the largest worldwide with tofacitinib in patients with RA. Our primary objective was to identify predictors of treatment changes in patients with RA initiating tofacitinib in a real-world setting. The analysis showed that higher patient satisfaction was correlated with a lower rate of treatment changes. An increase in GC dose might be an indicator for step-up/switch treatment changes. The estimated drug survival, that is, the probability of still being on tofacitinib at M24 visit, was 48% across the total population. Effectiveness outcomes of tofacitinib therapy showed increased LDA and remission rates and were similar for patients with tofacitinib monotherapy and for combination therapy with MTX.

PROs indicated less fatigue, reduced arthritis pain and morning stiffness, and improved functional ability and quality of life. Patient satisfaction improved over time. Patients were able to reduce concomitant GC doses. Safety results were consistent with the known safety profile of tofacitinib.

It is important to recognise the complex and multifaceted relationship between patient satisfaction and treatment changes. In this study, higher satisfaction was associated with lower rates of treatment changes. One of the more apparent reasons is that patients who experience favourable clinical outcomes are usually more satisfied and therefore may be less inclined to switch therapies. Consistent with this, data of the prospective German RABBIT registry cohort (n=10 646) revealed that satisfaction with RA treatment was positively correlated with improvements in disease activity (DAS28-4 (ESR)), pain and physical function.¹⁶ Additionally, the international non-interventional SENSE study (n=1624) identified high disease activity and lower patient satisfaction as predictors for a decision to switch to a different DMARD in patients with poorly controlled RA.¹⁷ Advanced treatment with a b/tsDMARD and better quality of life were found to be positive predictive factors of patient satisfaction.¹⁷

In contrast, patient-reported treatment satisfaction may also substantially diverge from clinical disease activity, with patients still unwilling to change treatments. Questionnaire data (n=6135) from the National Data Bank for Rheumatic Diseases in the USA showed that satisfaction with RA control was a strong predictor of unwillingness to change therapy (OR 7), even when patients had moderate or high arthritis activity, as assessed by the Patient Activity Score.¹⁸ Rather than seeking further improvement, patients prioritised maintaining their current status, potentially due to patient concerns such as fear of side effects or loss of control of RA following a switch. Conversely, patients whose expectations are not met and therefore report low patient satisfaction may pursue treatment changes despite measurable clinical improvements.

Taken together, these findings underscore the high level of complexity in the relationship between patient satisfaction and treatment changes. Satisfaction seems to be influenced by clinical outcomes and by individual expectations, perceptions and concerns. While our study provides valuable insights on predictors of treatment changes, it was not designed to assess the direct association between disease outcomes and patient satisfaction. Therefore, further research is needed to clarify this relationship.

In our study, tofacitinib drug survival was estimated to be 52% at M12 visit and 48% at the end of the study (M24). Interestingly, the median duration until treatment termination of any DMARD was longer in the monotherapy group than in the combination therapy group (289 days vs 138 days). A post hoc analysis¹⁹ evaluated tofacitinib persistence in 4967 patients with RA from two long-term extension studies (matched studies ORAL Sequel (NCT00413699) and Study A3921041 (NCT00661661)) for up to 9.5 years. Patients received 5 or 10 mg tofacitinib two times per day as monotherapy or with background csDMARDs. 1525 patients (30.7%) stayed on tofacitinib monotherapy and 2604 patients (52.4%) stayed on combination therapy with a background csDMARD.¹⁹ The estimated 2-year drug survival rate was 75.5%.¹⁹ A Canadian study²⁰ used real-world data (RWD) collected from the tofacitinib eXel programme (n=4276) to assess the treatment persistence of tofacitinib which was estimated as 62.7% after 1 year and 49.6% after 2 years. Results of our study were comparable to the Canadian RWD study,²⁰ but shorter than in the post hoc analysis which was based on RCTs with selected patients who had demonstrated acceptable efficacy and tolerability before being enrolled in the studies.¹⁹ The longer mean disease duration (9.9 years) in patients from our study and the high proportion of patients with prior bDMARD treatment (51%) indicate a more difficult-to-treat RA population which may contribute to the shorter drug survival of tofacitinib observed in this study compared with data from RCTs.¹⁹

The most frequent DMARD classes used after the first change in therapy were either another tsDMARD (33%) or a TNFi (29%) for patients switching from tofacitinib monotherapy. For those patients switching from tofacitinib combination therapy with MTX, the most common DMARDs were TNFi+csDMARD (26%), csDMARD monotherapy (24%) or tsDMARD+csDMARD (22%). Findings from the JAKpot collaboration which pooled data from 17 registries (n=2000) indicate that after failure of the first JAK inhibitor, both cycling of JAK inhibitors and switching to a bDMARD demonstrate comparable effectiveness.²¹

In our study, the proportion of patients achieving LDA and remission increased across different DAS (CDAI, SDAI) and was largely comparable between monotherapy and combination therapy. The remission rates by CDAI and SDAI at M18 visit were 25% and 22%, and thereby slightly lower than the respective CDAI and SDAI remission rates at 18 months (both 31%) from a retrospective, non-interventional cohort study²² that extracted data from the ‘Optimizing Patient outcomes in Australian Rheumatology’ dataset (n=1950). The comparison of effectiveness outcomes of our study results with data of a study²³ based on the US real-world CorEvitas RA registry (n=1298) indicated higher LDA and remission rates by CDAI at 12 months in our study (LDA: 70% vs 32% and remission: 24% vs 10%).

The mean concomitant GC dose standardised to prednisone decreased with tofacitinib therapy over the course of the study (BL to M24 −2 mg/day), and the fact that 38% of the patient population had decreased mean doses suggests a reduced reliance on GCs and improved disease control. These results suggest that long-term tofacitinib therapy leads to a reduced need for concomitant GC.

Tofacitinib was generally well tolerated, and the safety results in our study were consistent with the known safety profile of tofacitinib. The most frequent TEAESIs were serious infections (3% of patients, IR 2.8, 95% CI 2.1 to 3.7). The IR of herpes zoster was 2.6 events in 100 patient-years (95% CI 1.9 to 3.4). At the time of our study, the vaccination rate of herpes zoster was very low (7.6%), despite recommendations in the tofacitinib SmPC, as JAK inhibitors are known to be associated with a higher risk of herpes zoster infection.²⁴ However, this recommendation seems to be insufficiently implemented in German clinical care. Nevertheless, the IR of patients with herpes zoster per 100 patient-years in our study was only 2.6 (95% CI 1.9 to 3.4), which is lower than the IR of 4.4 (95% CI 3.8 to 4.9) reported in a study based on data from 13 studies in the tofacitinib development programme (n=5651).²⁵

Post hoc analyses of the ORAL Surveillance study have shown that patients with RA who were ≥65 years of age or long-time smokers (current or past) had an increased risk of AEs with tofacitinib versus TNFi. In patients without these risk factors, there was no detectable risk increase with tofacitinib versus TNFi.26,28 According to data of the German RABBIT registry, MACE did not occur more frequently in patients treated with JAK inhibitors than other DMARDs; the crude IRs for MACE with tofacitinib were 0.98 (95% CI 0.58 to 1.55) and 0.68 (95% CI 0.47 to 0.95) per 100 patient-years with any JAK inhibitor.²⁹ In our study, the IR of patients in 100 patient-years with ‘MACE’ with tofacitinib treatment was 0.6 (95% CI 0.3 to 1.0).

Limitations

Like in any observational research, there were a number of risks of bias. The potential major risks had been considered in the design and analysis of this study. Inherent limitations of non-interventional observational studies are the risk of selection/ascertainment bias and the inability to attribute causation. To limit potential selection bias, physicians were encouraged to include all eligible patients. Sensitivity analyses were carried out to investigate bias.

This is a non-interventional observational study with limited monitoring and data cleaning which may affect the quality of the data and lead to data discrepancies. Due to the non-interventional character of the study, only data that the investigator collected as part of their clinical routine were entered. None of the assessments were mandatory. This leads to missing data, particularly for PROs, which may bias the results in favour of patients with better treatment outcomes. Additionally, as patients discontinued from the study over time, the number of participants decreased, resulting in fewer patients being included in subsequent visits. This leads to reduced absolute values, including for some of the secondary effectiveness endpoints, while the percentage of these outcomes increased. Furthermore, patients who discontinued tofacitinib due to poor disease outcomes were no longer part of the analysis of secondary outcomes, leaving primarily those with better treatment responses. This selective dropout may lead to an overestimation of treatment effectiveness, as the remaining patients tend to have more favourable outcomes. This may further impact the interpretation of the effects.

Some confounders that were assessed in the primary analysis model, for example, the termination of tofacitinib, are inherently associated with a treatment change itself and therefore showed strong correlations with the response variable. In addition, in the primary analysis model, HRs are sensitive to the scale of the factors. Most of the confounders tested were dichotomous, whereas the predictor variables were mostly ordinal or numerical. Additionally, the effect of the imputation of predictor variables on the primary outcome cannot be assessed, as the sensitivity analysis was conducted using a different set of predictor variables.

Generalisability

A total of 1459 patients were documented in our study at 88 German sites, which can be regarded as representative for patients with moderate to severe RA in Europe. Patients in our study reflected the patients’ disease status and disease history well known for patients with RA.

Conclusions

The results of this study suggest that in a real-world setting in Germany, patient satisfaction and concomitant GC doses are key predictors of changes in RA therapy. Notably, continued tofacitinib therapy was associated with a reduced use of concomitant GCs, both in terms of lower doses and fewer patients requiring them.

This study significantly contributes to the body of RWE on tofacitinib’s effectiveness, demonstrating increased LDA and remission rates across various disease scores. Furthermore, the study highlights the favourable impact of tofacitinib on PROs, including improvements in fatigue, pain, functional ability and quality of life.

Pfizer Pharma GmbH was involved in the design and conduct of the study; collection, management, analysis and interpretation of data; and preparation, review and approval of the manuscript. The work reflects the views of its authors.

Data availability statement

Data are available upon reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.