Study Design

FIRST-RA is a prospective, multicentre, non-interventional cohort study conducted at approximately 30 rheumatology centres in Germany and Austria. The study reflects routine clinical care; no treatment or diagnostic procedures are mandated beyond standard practice.

Filgotinib (Jyseleca®; Alfasigma S.p.A., Bologna, Italy) is prescribed independently by the treating physician, in accordance with the current product label. Only patients newly initiating filgotinib are eligible for enrolment. The study does not influence therapeutic decisions or enforce specific treatment algorithms. The first patient was enrolled on 6 May 2025, and the planned study end is 31 December 2026.

Sample Selection

Eligibility criteria are shown in Table 2. “Moderate to severe RA” is defined per the Jyseleca® label and not specified further by DAS28 or CDAI cut-off values. RA diagnosis follows German clinical practice, which applies the 2010 ACR/EULAR classification criteria [6]. A total of approximately 300 patients will be enrolled and stratified into three subgroups of roughly equal size:

AT-naïve: No prior use of bDMARDs or tsDMARDs

One prior AT: Previously treated with one bDMARD or tsDMARD

Multiple prior ATs: Previously treated with ≥ 2 bDMARDs or tsDMARDs

Table 2 Eligibility criteria

To ensure analytical robustness, recruitment caps may be introduced if one subgroup becomes disproportionately large.

Measurements

The overall study flow is illustrated in Fig. 1, with study procedures and assessments detailed in Table 3. Clinical assessments are scheduled at baseline and at weeks 4, 12 and 24, while patient-reported outcomes (PROs) are more frequently collected electronically (daily from day 0 to day 7; weekly through week 4; and monthly through week 24) to capture early and longitudinal treatment effects.

Fig. 1

Study flow. PRO: patient-reported outcome; RA: rheumatoid arthritis; RAID: Rheumatoid Arthritis Impact of Disease score.

Table 3 Variables and evaluation time points

Data are collected using electronic case report forms (eCRFs) and include:

Demographics and clinical history

Disease characteristics and serology

Prior and concomitant medications (including corticosteroids)

Clinical disease activity (DAS28-CRP, CDAI and joint counts)

Laboratory parameters (e.g. CRP)

Safety outcomes (ADRs, SADRs and pre-defined adverse events of interest)

PROs (RAID score, morning stiffness and treatment satisfaction)

No additional laboratory tests or procedures beyond routine clinical care are required.

Patient-Reported Outcomes

PROs are completed by patients on their own electronic devices using an online ePRO that writes directly to the study’s electronic data capture (EDC)/eCRF. At the scheduled site visits (baseline and weeks 4, 12 and 24), site staff verify the entries and, if needed, enter the PRO values into the eCRF. If electronic entry is not possible, patients use paper diaries as back-up; the diaries are collected and transcribed by site personnel into the database. Completion time per assessment is approximately 5–10 min.

RAID Score

The Rheumatoid Arthritis Impact of Disease (RAID) score assesses seven domains: pain, fatigue, physical function, sleep, physical and emotional well-being, and coping [17, 18]. Each domain is rated on a 0–10 numeric rating scale, where higher scores indicate greater disease impact.

Treatment Satisfaction

Treatment satisfaction is assessed at each time point using two RABBIT-style items [19]—satisfaction with effectiveness and with tolerability—and recorded in four ordered categories: very satisfied, rather satisfied, rather unsatisfied or very unsatisfied.

Morning Joint Stiffness

Assessed by a “yes/no” question. If present, patients report the duration in minutes.

Effectiveness criteriaTender Joint Count (TJC)

Tender joint counts are based on a standardised 28-joint assessment, including shoulders, elbows, wrists, metacarpophalangeal (MCP) joints, proximal interphalangeal (PIP) joints and knees [20]. Each joint is evaluated for tenderness upon palpation or movement. Artificial joints are excluded.

Swollen Joint Count (SJC)

The same 28 joints are assessed for swelling, with artificial joints excluded. The presence of synovial swelling is recorded based on clinical examination.

Patient Global Assessment (PtGA)

Patients rate their current overall disease impact by answering the PtGA, which captures the patient’s overall perception of their disease activity. Patients are asked to rate their current disease activity on a numerical rating scale (NRS) ranging from 0 to 10, where 0 indicates no disease activity and 10 indicates very high disease activity. Responses are to be given as whole numbers. If a response is not available, this is recorded accordingly.

Physician Global Assessment (PhGA)

Physicians assess overall disease activity using clinical judgment based on signs, symptoms and physical function. The following prompt is used: “How would you rate the disease activity?”

Responses are scored on NRS ranging from 0 to 10, where 0 indicates no disease activity and 10 indicates maximal disease activity.

DAS28-CRP

DAS28 with CRP (an acute phase reactant) is a validated composite index used to quantify inflammatory disease activity [21].

Clinical Disease Activity Index (CDAI)

The CDAI is a simplified composite score comprising TJC, SJC, PtGA and PhGA. It does not require laboratory parameters and allows disease activity assessment at any time point. CDAI correlates well with functional outcomes and joint damage, and it is often used in daily clinical practice [22, 23].

In FIRST-RA, physicians may choose between DAS28-CRP and CDAI based on clinical preference. The treating physician calculates the CDAI as part of routine care and manually enters the index value into the eCRF; the eCRF does not auto-calculate CDAI from its components. If the CDAI is not available at a visit, this is recorded as “value not available”.

Inflammation Biomarkers

CRP will be recorded as part of routine clinical assessments to monitor systemic inflammation. No additional study-specific laboratory testing is required.

Data AnalysisSample Size Calculation

The sample size was determined to detect a mean change of 1.0 point in the RAID pain score with 80% power (α = 0.05), assuming a standard deviation of 2.5 using a one-sided paired-sample t-test. A minimum of 41 evaluable patients would be required for this analysis. Accounting for an estimated 5% dropout rate and up to 10% of patients with incomplete primary endpoint data (i.e. missing pain score at baseline or week 4), approximately 49 patients per subgroup are needed. To ensure robustness across treatment strata and enable precise descriptive subgroup summaries, a total of ~ 300 patients will be enrolled. In case these assumptions are optimistic, the observed variability and rates of missing data and dropout will be prospectively monitored during enrolment. Should these parameters deviate materially from the assumptions, the recruitment target will be increased to maintain the planned precision overall and within treatment strata.

General Analytical Approach

Given the non-interventional nature of the study, analyses will be primarily descriptive. Results will be summarised for the total population and stratified by predefined subgroups (e.g. AT-naïve vs. AT-experienced) and relevant covariates such as age, sex, body mass index, disease duration and prior therapies. Continuous variables will be summarised using mean, standard deviation, median, interquartile range and range, while categorical variables will be presented as absolute and relative frequencies, with 95% confidence intervals (CIs) where appropriate. All analyses will be conducted using validated software (e.g. SAS, R or SPSS). Missing data for the primary endpoint will be handled primarily using complete case analysis, with sensitivity analyses applying multiple imputation under the assumption of missing at random. A detailed Statistical Analysis Plan (SAP) will be finalised before database lock.

Primary Endpoint Analysis

The change in RAID pain score from baseline to week 4 (or earlier) will be analysed using a one-sided paired t-test, as the study is designed to detect improvement in the expected direction based on consistent evidence from prior randomised trials and real-world studies of filgotinib. This directional testing approach increases statistical power for confirming benefit under the anticipated effect direction. In case of non-normal distribution, a non-parametric alternative (e.g. Wilcoxon signed-rank test) will be used in sensitivity analyses.

Secondary and Exploratory Endpoint Analysis

Changes in fatigue (RAID), disease activity (DAS28-CRP, CDAI) and patient satisfaction will be analysed analogously to the primary endpoint. Categorical outcomes (e.g. the proportion of patients achieving RAID < 2.0, DAS28-CRP remission [defined as DAS28-CRP < 2.6] or low disease activity [i.e. DAS28-CRP < 3.2]) will be reported descriptively.

Subgroup comparisons (e.g. treatment response by AT exposure or pre-/post-label update) will be descriptive; no formal hypothesis testing is planned.

Early-response analyses will be anchored to initiation of the index JAK inhibitor (Jyseleca®). In this short 6-month window, the effect of the newly initiated treatment is awaited first; hence, concomitant DMARD initiation or escalation before the first response assessment is expected to be infrequent. Any such changes will be recorded and handled via time-varying covariates or censoring, with robustness examined in sensitivity analyses excluding these cases and in the (expectedly small) subgroup with concurrent filgotinib initiation and DMARD start or escalation.

For non-naïve patients, the duration of each prior advanced therapy and the reason for discontinuation—categorised as primary failure (no meaningful initial response), secondary failure (loss of response after initial benefit), adverse events or other—will be recorded. These variables will be included as covariates in adjusted analyses and used to define prespecified subgroups for descriptive summaries and sensitivity analyses.

Persistence analyses will use Kaplan-Meier estimators.

Adverse events will be summarised as event rates per patient and per exposure time, stratified by seriousness and severity.

Control of Confounding

To address potential confounding (e.g. indication bias), baseline characteristics will be summarised across subgroups, and, where appropriate, exploratory multivariable regression models adjusted for key covariates (such as age, sex, disease duration, comorbidities and prior therapies) will be used to support interpretation.

External Reference Cohort

Where applicable, patient characteristics, prescribing patterns and safety outcomes from the study will be descriptively put into perspective with the FILOSOPHY study. For these comparisons, variables will be harmonised by aligning definitions, coding schemes and assessment time windows across both datasets, following the original data dictionaries of each study.

Data Management and Monitoring

Data are captured using a validated EDC system. Built-in plausibility checks, comprehensive site training and regular central data reviews will ensure data quality and consistency. Patient confidentiality and data protection are maintained in accordance with applicable data privacy laws.