Effectiveness, safety and persistence of filgotinib in rheumatoid arthritis with a specific focus on difficult-to-treat patients: results from a multicentre observational study

Abstract

Background

Filgotinib, a preferential JAK1 inhibitor, is among the most recently approved therapies for rheumatoid arthritis (RA). Real-world evidence remains limited, particularly for difficult-to-treat RA (D2T) patients, who are underrepresented in clinical trials. We therefore evaluated the efficacy, safety, and treatment persistence of filgotinib in a large cohort of patients with RA, with a specific focus on those with D2T RA.

Methods

This retrospective multicentre study included patients with RA (ACR/EULAR 2010 criteria) treated with filgotinib across nine Italian centres. Patients were followed every 3–6 months for up to 3 years, with collection of clinical variables, laboratory data, and adverse events (AEs).

Results

A total of 254 patients were included (mean age 58 ± 12 years; 85% female), of whom 47% had D2T disease. Disease activity indices (CDAI, DAS28-CRP, PGA, PhGA) improved significantly from baseline (p < 0.001), with remission rates reaching 41.3% (CDAI) and 70.5% (DAS28-CRP). Glucocorticoid (GC) use decreased from 64.4% to 32.6%. Improvements in disease activity and the GC-sparing effect were comparable between D2T RA and non-D2T RA patients. During follow-up, 153 AEs were reported, including 13 serious AEs and one infection-related death; non-serious infections were the most frequent events. Lipid profile remained stable throughout the study period, and two major cardiovascular events were reported. Treatment persistence was 88.4% at 6 months and 68% at 24 months, with no significant differences between D2T RA and non-D2T RA patients.

Conclusions

Filgotinib demonstrated effective disease control, good safety and high treatment persistence in a real-world RA population, including a high proportion of D2T patients.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13075-026-03881-z.

Background

Rheumatoid arthritis (RA) is a chronic autoimmune inflammatory disease, with a substantial impact on physical function and quality of life. RA affects approximately 17 million people worldwide, is 2–3 times more common in women, and its prevalence increases with age, with disease onset most frequently occurring between 60 and 70 years of age [1, 2]. Current therapeutic strategies, as outlined by 2025 EULAR recommendations, include the use of short-term glucocorticoids (GCs) and conventional synthetic (cs) disease-modifying antirheumatic drugs (DMARDs) as the first-line strategy. In patients with inadequate response after 3–6 months, treatment escalation with biological (b) DMARDs, or targeted synthetic (ts) DMARDs, is recommended irrespective of the presence of poor prognostic factors. In this context, tsDMARDs (i.e. JAKis) should be considered only after a careful individual risk assessment regarding cardiovascular events, malignancy, and thromboembolism; therefore bDMARDs are generally preferred [3]. Among the numerous therapeutic strategies developed for the treatment of RA, Janus kinase inhibitors (JAKis), belonging to the tsDMARDs class, represent the most recent pharmacological option. These drugs are small molecules that offer the significant benefit of being administered orally, in contrast to bDMARDs. Currently approved JAKis for the treatment of RA include: tofacitinib, baricitinib, upadacitinib, and filgotinib, as well as peficitinib (approved in Japan only). Filgotinib primarily inhibits JAK1-dependent pathways which play a central role in promoting synovial inflammation and bone erosion, two key pathological processes in RA [4]. The lower inhibition of JAK2- and JAK3-dependent pathways provides a mechanistic rationale for a potentially differentiated safety profile compared to non-selective JAKis. Genetic data associate JAK2 with erythropoiesis and thrombopoiesis [5], JAK3 with lymphocyte proliferation and immune homeostasis, and TYK2 with antiviral responses [6]. In clinical terms, the JAK1 selectivity of filgotinib has been associated with an improvement in anemia of chronic disease at a group level, likely due to reduced interference with erythropoietin signaling via JAK2 [7]. However, whether this selectivity translates into clinically meaningful differences in cardiovascular and infectious risk remains to be fully established. Despite robust evidence from randomised clinical trials demonstrating the efficacy and safety of filgotinib [8, 9], real-world data remain limited. This is particularly relevant for patients with difficult-to-treat (D2T) RA, a subgroup defined by persistent high disease activity despite exposure to multiple therapies with different mechanisms of action. D2T patients represent a major unmet clinical need, as they are typically associated with poorer outcomes and reduced likelihood of achieving treatment targets, while also being consistently underrepresented in randomised clinical trials. Although 47% of patients in FINCH 2 had failed at least two prior bDMARDs, previous biologic therapies were not stratified according to their mechanism of action, and other variables required for D2T RA classification were not reported, precluding a reliable distinction between D2T and non-D2T patients [10]. In this context, generating real-world data in cohorts with a substantial representation of D2T RA patients is essential to better clinical practice and optimize treatment strategies. Therefore, in this study we aimed to evaluate the efficacy, safety and treatment persistence of filgotinib in a large multicentre cohort of patients with RA, with a specific focus on patients with D2T RA.

Methods

Study design and data sources

This was a retrospective, multicentre, observational study including adult patients with RA, according to the 2010 ACR/EULAR classification criteria, who initiated treatment with filgotinib in routine clinical practice across nine Italian rheumatology centres from 2021 onwards. Participating centres were asked to include all consecutive patients treated with filgotinib during the study period. Although no external verification of patient capture was performed, this approach was intended to minimize selection bias. Data were recorded from the start of treatment with filgotinib (T0) and at subsequent follow-ups at 3- to 6-month intervals, in line with standard outpatient scheduling, up to a maximum duration of 36 months. Concomitant csDMARDs and GCs were allowed throughout the study. The collected data were divided into time points: T1 includes visits during the first 6 months, T2 up to 12 months, T3 up to 24 months, and T4 up to 36 months. At each visit, disease activity was assessed using validated composite indices employed in clinical practice, and laboratory data, concomitant therapies, and AEs were systematically collected. In our study, we considered minimal clinically significant change (MCID) to be a reduction of ≥ 12 in the CDAI and ≥ 1.2 in the DAS28-CRP [11].

D2T RA status was defined according to EULAR criteria [12] as a failure of at least two b/tsDMARDs with different mechanisms of action in the presence of moderate-to-high disease activity (DAS28-CRP or CDAI). The additional criterion of disease being perceived as problematic by patient and/or clinician was not explicitly recorded. However, in a treat-to-target strategy, treatment escalation is generally driven by the perception of insufficient disease control by the physician and/or patient; therefore, most patients requiring therapy changes are likely to meet this domain although this domain could not be formally assessed in our dataset.

Analyses of lipid profile were restricted to patients with measurements available at T0 and at least one follow-up visit. The Atherogenic Index of Plasma (AIP), an informative marker of atherogenic risk, was also calculated as the logarithm of the triglycerides-to-HDL cholesterol ratio.

AEs were recorded at each time point and classified according to severity and seriousness. Serious AEs were defined in accordance with European Medicines Agency (EMA) criteria. The severity was categorized as mild, moderate, or severe based on interference with daily activity. The relationship between reported AEs and filgotinib was assessed based on the treating physician’s clinical judgement. Major adverse cardiovascular events (MACE) were defined as a composite of myocardial infarction, stroke, and cardiovascular death. Severe aortic insufficiency requiring hospitalization was included in an expanded cardiovascular outcome set.

The study was conducted in accordance with the principles of the Declaration of Helsinki and Good Clinical Practice guidelines and received approval from the Ethics Committee of Padova University Hospital (5935/AO/24).

Statistical analysis

Continuous variables were summarized using mean ± standard deviation (SD) or median and interquartile range (IQR), as appropriate. Categorical variables were described using frequencies and percentages. Changes from baseline in continuous outcomes were initially evaluated using paired-sample t-tests, and effect sizes were calculated using Cohen’s d, interpreted according to conventional thresholds. Longitudinal changes were analyzed using repeated measures analysis of variance (ANOVA), applying Greenhouse–Geisser correction when required. Multivariate ANOVA results are reported using Pillai’s Trace, and linear and quadratic within-subject contrasts were examined to explore trends over time. Longitudinal changes in haemoglobin levels and lipid parameters were analyzed using a linear mixed-effects model, with patient as a random intercept and time point as a fixed effect. Between-group comparisons (D2T vs. non-D2T) were assessed using mixed-model ANOVA, including time × group interaction terms. Treatment persistence was assessed using Kaplan–Meier survival analysis, with the estimated probability of remaining on therapy at predefined time points reported. A sensitivity analysis was performed using a non-responder imputation (NRI) approach. Patients who discontinued treatment were conservatively classified as non-responders from the time of discontinuation or last available observation onwards. Predictors of treatment discontinuation were assessed using Cox proportional hazards regression adjusting for age, sex, disease duration, seropositivity, and D2T status. Potential predictors of infection were explored using contingency tables and a multivariable logistic regression model. Statistical significance was assessed with the chi-square (χ²) test, and univariate odds ratios were calculated to quantify the strength of associations. Exposure-adjusted incidence rates (EAIRs) were calculated as the number of patients experiencing the event divided by the total patient-years (PY) of exposure and expressed per 100 PY. A two-sided p-value < 0.05 was considered statistically significant. Missing data were not imputed. Analyses were conducted using all available data, according to the statistical method applied. All analyses were performed using SPSS (version 29, IBM Corp., Armonk, NY).

Results

Baseline characteristics

A total of 254 patients with RA were enrolled (mean age 58 ± 12 years; 85% female; 47% D2T). Patients had long-standing disease (median duration 13 years, IQR 7.1–20); 63.8% were RF-positive, 67.6% ACPA-positive, 57.1% RF- and ACPA-positive, 22.4% had extra-articular manifestations, of which 7.9% had pulmonary involvement. At baseline, most patients had moderate-to-high disease activity according to DAS28-CRP (85.3%) and CDAI (90.7%). Most patients started with a drug dose of 200 mg/day (94.9%) while the others started with half dose (Table 1). Main comorbidities are reported in Table 2.

Table 1.

Demographic and clinical characteristics of RA patients at baseline according to D2T status

	Non-D2T (n = 134)		D2T (n = 120)
Age (mean, SD)	56	13	60.6	11.1
Female sex (n, %)	115	85.8	102	85
Disease duration (median, IQR)	9.5	13.7	15.4	10.3
RF (n, %)	74	55.2	88	73.3
ACPA (n, %)	80	59.7	91	75.8
RF and ACPA (n, %)	66	49.2	79	65.8
Erosions (n, %)	68	50.7	80	66.7
Extra-articular disease (n, %)	24	17.9	33	27.5
Pulmonary involvement (n, %)	9	6.7	11	9.2
DAS-28 CRP (mean, SD)	4.21	1.17	4.85	1.08
CDAI (mean, SD)	20.3	9.4	25.7	9.56
PGA (mean, SD)	6.07	1.97	6.97	1.90
PhGA (mean, SD)	5.43	1.66	6.06	1.83
csDMARD prescribed (n, %)	68	50.7	39	32.5
Glucocorticoids prescribed (n, %)	86	64.2	77	64.2
PDN dose, mg/die (median, IQR)	5	2.5	5	5
Filgotinib 200 mg (n, %)	130	97	111	92.5
Previous csDMARDs	132	98.5	120	100
MTX	126	94	111	92.5
LFN	37	27.6	67	55.8
SSZ	28	20.9	28	23.3
HCQ	59	44	62	51.7
Others	15	11.2	17	14.2
Previous b/tsDMARDs (n, %)	91	67.9	120	100
No. of previous b/tsDMARDs (median, IQR)	1	2	3	3
Anti-TNF (n, %)	73	54.5	108	90
Anti-CTLA4 (n, %)	24	17.9	82	68.3
Anti-IL6 (n, %)	8	5.9	73	60.8
Other JAKis (n, %)	13	9.7	61	50.8
Anti-CD20 (n, %)	6	4.5	24	20
Anti-IL1 (n, %)	3	2.2	8	6.7

SD standard deviation, IQR interquartile range, D2T difficult-to-treat, RA Rheumatoid Arthritis, RF Rheumatoid Factor, ACPA Anti-citrullinated protein antibodies, DAS28-CRP Disease Activity Score 28 joints-C-reactive protein, CDAI Clinical Disease Activity Index, PGA Patient global Assessment, PhGA Physician global Assessment, PDN prednisone equivalent, csDMARDs conventional synthetic disease-modifying antirheumatic drugs, b/tsDMARDs biological/targeted synthetic DMARDs, CTLA4 cytotoxic T-lymphocyte-associated protein 4, IL interleukin, CD cluster of differentiation, JAKis Janus kinase inhibitors, TNF tumor necrosis factor

Table 2.

Main baseline comorbidities of the study population according to D2T status

	Non-D2T (n = 134)		D2T (n = 120)
BMI (mean, SD)	24.6	4.11	25.2	4.81
Smoking status (n, %)
Current	21	15.7	13	10.8
Former	19	14.2	10	8.3
Never	55	41	59	49.2
Unknown	39	29.1	38	31.7
AMI (n, %)	2	1.5	5	4.2
Peripheral vascular disease (n, %)	4	3	9	7.5
Cerebrovascular disease (n, %)	6	4.5	11	9.2
Chronic pulmonary disease (n, %)	14	10.4	13	10.8
Diabetes (without complications) (n, %)	3	2.2	11	9.2
Diabetes with end organ damage (n, %)	1	0.7	5	4.2
Systemic arterial hypertension (n, %)	39	29.1	44	36.7
Hypercholesterolemia/dyslipidaemia (n, %)	36	26.9	43	35.8
Osteoporosis or frailty fracture (n, %)	25	18.7	45	37.6

BMI Body mass index, AMI Acute Myocardial Infarction

Efficacy

In the overall population, all major disease activity indices (CDAI, DAS28-CRP, PGA and PhGA) significantly improved from baseline to subsequent follow-up time points (paired-sample t-test, all p < 0.001) (Table 3). Effect sizes were moderate to large (Cohen’s d 0.7–1.1). Repeated measures ANOVA performed for CDAI and DAS28-CRP confirmed a significant effect of time for both (CDAI: Pillai’s Trace = 0.776, F = 110.7, p < 0.001; DAS28-CRP: Pillai’s Trace = 0.722, F = 91.6, p < 0.001), with significant linear contrasts indicating a steady improvement over time. Improvements were already evident at the first follow-up and were maintained through the last available visit. Although D2T patients presented with higher disease activity levels at baseline, mixed-model ANOVA showed no significant Time × Group interaction (CDAI: F = 1.134, p = 0.335; DAS28-CRP: F = 1.292, p = 0.277), indicating that the magnitude and trajectory of improvement over time were comparable between D2T and non-D2T groups (Fig. 1). Remission rates progressively increased during follow-up, reaching 70.5% (DAS28-CRP) and 41.3% (CDAI) at the last time point. Rates of low disease activity (LDA; DAS28-CRP ≤ 3.2) also improved compared with baseline, although remission progressively became the predominant treatment target achieved over time. Throughout follow-up, remission and LDA rates remained consistently lower in D2T than in non-D2T patients (Table 4). Most patients achieved clinically meaningful improvement (≥ MCID) over time according to both DAS28-CRP (62–71%) and CDAI (54–66%), confirming consistent and clinically relevant improvement across both indices (Fig. 2). GC dosage and use progressively decreased during follow-up. Median GC dose ranged from 7.08 mg/day at T0 to 3.68 mg/day at T3, indicating a sustained steroid-sparing effect over time. At the last follow-up (T4), an apparent increase was observed; however, this estimate is based on a limited number of patients (Additional Table 1). At baseline, 64.4% of patients were receiving GC, while 35.6% were steroid-free. At last avaiable follow-up, 41.8% of patients were steroid-free; among patients achieving T4, only 32.6% were still on GC. A significant steroid-sparing effect compared with baseline was confirmed by the McNemar test (χ² = 41.3, df = 1, p < 0.001). Similar reductions in GC exposure were observed in both D2T and non-D2T groups.

Table 3.

Longitudinal assessment of disease activity and global assessments in RA patients

Time point	PGA	PhGA	DAS28-CRP	CDAI
T0	7.0 (3.0)	6.0 (2.0)	4.39 ± 1.13	22.0 (13.0)
T1	3.0 (3.0)	2.0 (3.0)	2.82 ± 1.16	8.0 (10.0)
T2	3.0 (4.0)	1.0 (2.0)	2.61 ± 1.06	6.0 (9.0)
T3	3.0 (4.0)	1.0 (2.5)	2.48 ± 1.04	5.0 (8.25)
T4	1.5 (4.0)	1 (3.75)	2.5 ± 1.2	4.0 (10.8)

Paired-sample t-test, p < 0.001 for all comparisons. Values are presented as median (IQR) for skewed variables and mean ± SD for symmetric variables

RA Rheumatoid Arthritis, PGA Patient global Assessment, PhGA Physician global Assessment, DAS28-CRP Disease Activity Score 28 joints-C reactive protein, CDAI Clinical Disease Activity Index

Fig. 1.

Temporal changes in mean DAS28-CRP and CDAI in D2T and non-D2T RA patients. Lines represent mean values, with 95% confidence intervals shown as error bars. D2T=difficult-to-treat; RA= Rheumatoid Arthritis; DAS28-CRP = Disease Activity Score 28 joints-C reactive protein; CDAI = Clinical Disease Activity Index

Table 4.

Remission, low disease activity, and moderate-to-high disease activity over time according to DAS28-CRP in D2T and non-D2T patients

Time point	Group	Remission n (%)	LDA n (%)	M to HDA n (%)
T0 (n = 246)	Non-D2T	11 (8.7)	22 (17.3)	94 (74)
	D2T	0	3 (2.5)	115 (97.5)
T1 (n = 244)	Non-D2T	73 (57)	22 (17.2)	33 (25.8)
	D2T	43 (37.1)	31 (26.7)	42 (36.2)
T2 (n = 183)	Non-D2T	64 (64)	14 (14)	22 (22)
	D2T	37 (44.6)	22 (26.5)	24 (28.9)
T3 (n = 119)	Non-D2T	44 (67.7)	11 (16.9)	10 (15.4)
	D2T	30 (56.6)	10 (18.9)	13 (24.5)
T4 (n = 44)	Non-D2T	20 (80)	1 (4)	4 (16)
	D2T	11 (57.9)	0	8 (42.1)

Disease activity distribution over time according to DAS28-CRP stratified by D2T status. Percentages are calculated within each group (non-D2T and D2T) at each timepoint. n refers to total number of patients with available assessment at each timepoint

Fig. 2.

Percentage of RA patients achieving MCID in DAS28 (CRP) or CDAI over time. RA= Rheumatoid Arthritis; MCID= Minimal Clinically Significant Change; DAS28-CRP= Disease Activity Score 8 joints-C reactive protein; CDAI= Clinical Disease Activity Index

No systematic assessment of radiographic progression was available in this study.

Safety

During follow-up, 153 AEs were reported in 132 patients. Those included 13 serious events and one death. Among AEs judged to be related to study treatment (n = 126), 7.2% were judged as definitely related to filgotinib, while no causal association was identified in 21.4%. Infections were the most frequently reported AEs (65 events), of which only five were classified as serious (7.7% of infections), one leading to death. There were 4 registered cases of varicella-zoster virus (VZV) reactivation and 6 of herpes simplex virus (HSV) reactivation. Among VZV cases, two were perceived as severe. Non-serious infections represented the most common adverse event overall. Pulmonary involvement emerged as the only factor associated with infections (OR 2.78, 95% CI 1.04–7.42, p = 0.035), whereas other major comorbidities and concomitant MTX or GC use were not associated. In the multivariable logistic regression model, RA-associated pulmonary involvement remained independently associated with infections during follow-up (OR 2.88, 95% CI 1.06–7.83, p = 0.039), while age, diabetes, and D2T status showed no significant association. Following infections, skin rash and gastrointestinal involvement (such as nausea and vomiting) were the most common reported AEs. Two MACE occurred, consisting of one non–ST-elevation myocardial infarction (NSTEMI) and one severe aortic insufficiency, both considered unlikely to be drug-related. A complete lipid profile at T0 and at least one subsequent time point was available for 116 patients. HDL levels increased significantly at T1 (+ 2.43, p = 0.041), triglycerides increased at T3 (+ 10.55, p = 0.012), while other parameters showed no significant changes up to 24 months (Table 5). Most of the variability was attributable to differences between patients (ICC 0.66–0.76), with minimal contribution of time (marginal R² 0.006–0.009). Concerning malignancies, a single case of non-melanoma skin cancer (basal cell carcinoma) was reported during follow-up (Table 6). Over 362.16 PY of exposure, the EAIRs were 42.3 (95% CI 35.8–49.5) per 100 PY for any AE, 17.95 (13.85–22.88) for total infections, 1.38 (0.45–3.22) for severe infections, 1.10 (0.30–2.83) for herpes zoster, 0.28 (0.01–1.54) for death, 0.55 (0.07–2.00) for major cardiovascular events (MACE), and 0.28 (0.01–1.54) for malignancies.

Table 5.

Lipid profile over time in RA patients treated with filgotinib

Time point	N	TC (mg/dL)	LDL (mg/dL)	HDL (mg/dL)	TG (mg/dL)	LDL/HDL	AIP
T0	116	208 ± 42.8	121 ± 36.9	68.5 ± 17.8	96.4 ± 36.3	1.90 ± 0.82	0.136 ± 0.208
T1	101	209 ± 41.5	119 ± 37	70.7 ± 17.6†	96.7 ± 36.9	1.81 ± 0.87	0.119 ± 0.221
T2	75	207 ± 40.5	118 ± 34.5	70.3 ± 17.5	96.1 ± 47.1	1.86 ± 1.09	0.116 ± 0.252
T3	53	214 ± 39.9	122 ± 35	70.9 ± 16.7	106 ± 43.1†	1.81 ± 0.64	0.155 ± 0.209
T4	17	197 ± 31.2	106 ± 26.7*	80.7 ± 10.8†	76.6 ± 26.9	1.30 ± 0.32*	−0.047 ± 0.186*

Data are presented as mean ± standard deviation

N = number of patients with available data at each time point

T1 = 0–6 months; T2; 6–12 months; T3 = 12–24 months; T4; 24–36 months

TC total cholesterol, LDL low-density lipoprotein, HDL high-density lipoprotein, TG triglycerides, AIP atherogenic index of plasma

“†” = significant increase from baseline (p < 0.05), “*” = significant decrease from baseline (p < 0.05)

Table 6.

Type and number of AEs reported during filgotinib treatment (n = 153)

Type of AE	N of AE
Mild infection	60
Among mild infection VZV/HSV	4/6
Serious infection	5
Skin reaction	9
GI involvement	9
Headache	8
MSK involvement	8
Hypercholesterolaemia	6
Changes in blood count	6
Vertigo	4
D-dimer elevation	3
Respiratory insufficiency	2
Arterial hypertension	2
MACE	2
NMSC	1
Other or unknown	28

AE adverse events, N number, VZV varicella-zoster virus, HSV herpes simplex virus, GI gastrointestinal, MSK musculoskeletal, MACE Major adverse cardiovascular events, NMSC non-melanoma skin cancer

Haemoglobin levels showed a statistically significant but clinically minimal increase over time (p = 0.017), with substantial inter-individual variability (ICC = 0.75).

Treatment persistence

Seventy treatment discontinuations occurred during the study period, with reasons including AEs, inefficacy and unknown causes. The proportion of D2T patients among discontinuers did not exceed that of non-D2T patients in most intervals; the apparent increase at the last follow-up should be interpreted cautiously due to the very low number of events (Fig. 3). The number of patients contributing data decreased over time, reflecting both treatment discontinuations and incomplete follow-up at later visits. The number of treatment discontinuations at each time point is reported in Fig. 3. The Kaplan–Meier survival curve did not drop below 50% during the observation period, so median drug survival was not estimable (Fig. 4). The probability of remaining on treatment was 88.4% at 6 months (95% CI: 84.5–92.6%), 81.4% at 12 months (95% CI: 76.5–86.6%), 77% at 18 months (95% CI: 71.5–83.0%), 68% at 24 months (95% CI: 61.6–75.5%), and 55.8% at 36 months (95% CI: 46.7–66.7%). Estimates beyond 24 months should be interpreted with caution due to the limited number of patients at risk. Likewise, the distribution of discontinuation reasons did not differ significantly between groups (Fisher’s exact test, p = 0.653) (Additional Table 2). Cox proportional hazards regression analysis showed no significant association between time to drug discontinuation and D2T status, seropositivity (RF + or ACPA+), age and sex, as all p-value were > 0.05. Disease duration and age were also not associated with earlier discontinuation. A sensitivity analysis using NRI has been provided in Additional Table 3.

Fig. 3.

Reasons for treatment discontinuation over time and proportion of D2T RA patients among all discontinuations. The left y-axis shows the number of patients who discontinued treatment. The right y-axis corresponds to the overlaid line indicating the percentage of difficult-to-treat (D2T) patients among the discontinuers. T1 = 0–6 months; T2; 6–12 months; T3 = 12–24 months; T4; 24–36 months

Fig. 4.

Kaplan–Meier estimates of filgotinib treatment persistence in difficult-to-treat (D2T) and non-D2T RA patients. The numbers displayed below the figure represent the number of D2T and non-D2T patients remaining in follow-up at each time point

Discussion

Randomised controlled trials (RCTs) are essential to establish the efficacy and safety of new therapies under controlled conditions; however, they may not fully capture the heterogeneity of patients seen in real-world clinical practice. In this context, our study confirms the efficacy of filgotinib in a multicentre real-world cohort of patients with RA, demonstrating clear improvements from the first assessment visit that progressed over time, as reflected by increasing remission rates reaching 70.5% (DAS28-CRP) and 41.3% (CDAI) at the last time point. Overall, these results are consistent with previous studies reporting significant improvements in disease activity as early as 3 months [13, 14] and with clinical trials [10, 15, 16], supporting the external validity of our findings. Particularly, comparable DAS28-CRP remission rates have been reported in real-world filgotinib cohorts, ranging from 65.9% to 66.2% after 6 months of treatment [17, 18]. The high remission rate observed at 36 months in our cohort should be interpreted in the context of the limited number of patients remaining under follow-up and the potential for survivor bias. Results were supported by a sensitivity analysis using non-responder imputation, although findings at later time points should be interpreted in light of the reduced number of patients remaining under follow-up. Moreover, the majority of patients reached clinically meaningful improvements (≥ MCID), emphasizing the clinical relevance of these observed changes. Notably, nearly half of our cohort (47.2%) fulfilled the EULAR definition of D2T RA [12]. Despite the challenging clinical scenario, filgotinib demonstrated substantial effectiveness. While D2T patients exhibited higher baseline disease activity, the extent and pattern of improvement paralleled those observed in non-D2T patients. To our knowledge, no other studies have included such a large cohort of D2T RA patients treated with filgotinib. Importantly, the comparable magnitude of response observed in D2T RA and non D2T RA patients is clinically meaningful, as D2T RA is typically associated with lower response rates and reduced likelihood of achieving treatment targets [19]. This finding is particularly relevant in clinical practice, where therapeutic options for D2T patients are often limited and treatment decisions are frequently empirical. Another key finding is the significant GC-sparing effect over time, observed consistently across D2T and non-D2T groups. This highlights the treatment strategy’s potential to diminish long-term GC dependency, a pivotal therapeutic objective in light of the well-documented long-term toxicity associated with chronic GC exposure [20, 21]. This finding aligned with the outcomes of another real-world study on filgotinib [13] and with other JAKis as a class [22, 23].

Our cohort included patients initiating filgotinib before and after the implementation of European Medicines Agency (EMA) recommendations, when restrictions on JAKi prescription were introduced following the ORAL Surveillance study [24] (in patients aged ≥65 years, current or long-term former smokers, or those with cardiovascular risk factors or history of malignancy). The mean age was 58 ± 12 years; 25.5% were current or former smokers, 32.8% had systemic arterial hypertension, 31.2% had hypercholesterolemia or dyslipidaemia, 7.9% had diabetes mellitus and mean body mass index was 24.84 ± 4.4 kg/m2. Despite the presence of multiple cardiovascular risk factors, only two MACEs were recorded during follow-up, with no apparent signal of increased short-term risk. Although the sample size and follow-up duration do not allow definitive conclusions regarding long-term cardiovascular safety, these findings are consistent with previous reports, which documented a low incidence of MACEs with filgotinib and other JAKis, despite heterogeneous cohorts [13, 18, 25, 26].

Overall, filgotinib treatment was associated with a largely stable lipid profile. The observed improvement at T4 should be interpreted cautiously due to the small number of patients available at this time point. Consistently, a real-world study reported no significant lipid changes during the first six months of treatment, in patients aged < 65 or ≥ 65 [27]. The stability of the AIP over time is reassuring, as higher AIP has been identified as an independent predictor of MACEs [28]. Infections were the most frequently reported AEs, predominantly non-serious. Serious AEs were uncommon, and only one death occurred during follow-up, without clear evidence of a causal relationship. In our cohort the overall incidence of Herpes zoster (HZ) remained low. HZ reactivation occurs in RA at an approximately 1.5-fold higher risk compared to the general population [29]. While this risk has been reported to be increased with tofacitinib [30] and baricitinib [31], available studies on filgotinib have not consistently shown an increased risk [9, 32]. Pulmonary involvement emerged as the only factor independently associated with an increased risk of infections, likely reflecting underlying patients’ vulnerability rather than a drug-specific effect [33]. The overall EAIRs for AEs were consistent with those reported in the DARWIN 3 long-term extension study [9]. In particular, rates of infections and serious infections were broadly comparable between cohorts (total infections: 17.95 vs. 23.3 per 100 patient-years; serious infections: 1.10 vs. 1.3 per 100 patient-years), while major adverse cardiovascular events and malignancies remained infrequent in both studies (< 1 per 100 patient-years), supporting a stable and reassuring safety profile of filgotinib in real-world clinical practice. Differences in study design, patient selection, and follow-up duration should however be considered.

Treatment persistence in our cohort was high. Persistence at 6 and 12 months was in line with another real-world study (88.4 vs. 84.5% and 81.4 vs. 75.8%) [34], while the 36-month estimate (55.8%) was broadly comparable to the interim findings from FINCH 4 (63% at 156 weeks), despite differences in study design and patient selection [8]. Notably, treatment persistence did not differ between D2T and non-D2T patients. This finding is particularly relevant, as D2T status is typically associated with higher rates of treatment discontinuation and cycling across therapies. The comparable drug survival observed in our cohort therefore represents a signal of sustained effectiveness and tolerability of filgotinib, even in a population traditionally considered at higher risk of therapeutic failure.

Strengths and limitations

The multicentre design enhances the external validity of the findings, reflecting heterogeneous clinical settings, though it may introduce variability in data collection. A key strength of this study is the high proportion of D2T patients, a population typically underrepresented in clinical trials. This expands the currently limited evidence on the effectiveness of filgotinib in this challenging subgroup. The relatively large sample provides a robust representation of patients with established RA and allows a direct comparison between D2T and non-D2T patients. Clinical response was assessed using both composite indices and MCID, capturing changes that are meaningful from the patient’s perspective. This is one of the first studies to demonstrate a GC-sparing effect of filgotinib in routine clinical practice, although this finding warrants further investigation. A further strength lies in the availability of real-world data on lipid profile over time, despite the absence of data on statin use and the limited availability of serial lipid measurements. Limitations include the retrospective design, the absence of a control group, and the follow-up duration which may limit detection of rare or long-term AEs and the possibility of residual confounding. Information on prior history of malignancy and venous thromboembolic events was not systematically collected and therefore could not be included in the baseline analysis. Radiographic progression was not systematically assessed. Importantly, the patient/physician perception criterion of the EULAR D2T definition could not be directly assessed due to the retrospective design of the study. This may have led to a limited overestimation of D2T prevalence; however, misclassification is expected to be non-differential and unlikely to affect between-group comparisons. Overall, the study provides valuable real-world insights into the effectiveness, safety, and persistence of filgotinib in a heterogeneous RA population, including a substantial proportion of D2T patients, and represents one of the longest real-world follow-ups reported to date.

Conclusions

In this real-world study, filgotinib demonstrated robust effectiveness and an acceptable safety profile in a heterogeneous RA population, including a substantial proportion of D2T patients. Significant improvements were observed across multiple composite indices, with many patients achieving a rapid and clinically meaningful response, alongside a steroid-sparing effect. AEs were generally manageable, with non-serious infections being the most frequently reported, and a low incidence of cardiovascular events, supporting a favourable safety profile in routine practice. High treatment persistence over 24 months further reflects sustained efficacy and good tolerability. Notably, comparable trajectories of clinical improvement and treatment persistence between D2T and non-D2T patients underscore the robustness of filgotinib across different clinical phenotypes, emerging as a potentially effective option even in patients with limited therapeutic alternatives.

Abbreviations

ACPA

Anti-citrullinated protein antibodies

ACR

American College of Rheumatology

AEs

Adverse Events

AIP

Atherogenic Index of Plasma

ANOVA

Analysis of Variance

b/bsDMARDs

Biological and biosimilar DMARDs

CDAI

Clinical Disease Activity Index

csDMARDs

Conventional synthetic DMARDs

D2T

Difficult-to-treat

DAS28-CRP

Disease Activity Score 28 joints C-reactive protein

DMARDs

Disease-Modifying Antirheumatic Drugs

EAIRs

Exposure-Adjusted Incidence Rates

EMA

European Medicines Agency

EULAR

European Alliance of Associations for Rheumatology

GC

Glucocorticoids

HDL

High-Density Lipoprotein

HSV

Herpes Simplex Virus

IQR

Interquartile Range

JAKis

Janus kinase inhibitors

MACEs

Major Adverse Cardiovascular Events

MCID

Minimal Clinically Significant Change

MTX

Methotrexate

NSTEMI

Non–ST-elevation myocardial infarction

PGA

Patient Global Assessment (of disease activity)

PhGA

Physician Global Assessment (of disease activity)

PY

Patient-Years

RA

Rheumatoid Arthritis

RCTs

Randomised Controlled Trials

RF

Rheumatoid Factor

SD

Standard Deviation

tsDMARDs

Targeted synthetic DMARDs

VZV

Varicella Zoster Virus

Authors’ contributions

Prof. MZ contributed to the conception and design of the work, the analysis and interpretation of data, and critically revised the manuscript for important intellectual content. AGO contributed to data analysis, graphical representation, and drafted the manuscript. MG contributed to the conception and design of the work and to data collection. AG critically revised the manuscript for important intellectual content. All other authors contributed to collection of data and critically revised the manuscript. All authors approved the final version of the manuscript and agree to be accountable for all aspects of the work, ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. Prof. MZ and AGO had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Funding

The present study did not receive any funding.

Data availability

The datasets analysed during the current study are available from the corresponding author on reasonable request. No additional unpublished data are available.

Declarations

Ethics approval and consent to participate

The study was approved by the University of Padua Ethics Committee (5935/AO/24). Patients and members of the public were not involved in the design, conduct, reporting, or dissemination plans of this research.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.