Choosing the next option: a scenario-based roadmap for b/tsDMARD sequencing in rheumatoid arthritis

Abstract

Background

Biologic (b) and targeted synthetic (ts) DMARDs have expanded rheumatoid arthritis (RA) treatment options, but evidence guiding sequencing after b/tsDMARD failure remains heterogeneous and difficult to translate into clinical practice.

Objective

To provide a pragmatic, scenario-based synthesis to guide next-therapy choice in RA after discontinuation of a b/tsDMARD.

Methods

We performed a narrative review based on PubMed and Cochrane Library searches (2010–June 2025), including systematic reviews, meta-analyses, randomised controlled trials (RCTs), and real-world studies in adults with RA after ≥ 1 failed targeted therapy. Evidence was organised into predefined clinical scenarios by number and mechanism of prior b/tsDMARD failures and synthesised using a structured framework incorporating study design, consistency, and key limitations.

Results

Across scenarios, switching to a drug with other mechanism of action (OMA) or a JAK inhibitor (JAKi) generally showed comparable, and sometimes favourable, effectiveness and persistence versus within-class cycling, particularly after failure of multiple targeted therapies. After first TNF inhibitor (TNFi) failure, both strategies are effective, although most RCTs and observational data tend to favour switching, mainly for persistence. In more treatment-experienced RA, IL-6 inhibitors and JAKi often showed favourable effectiveness and retention, although results were inconsistent. After JAKi failure, limited and largely observational evidence suggests that cycling to a second JAKi may offer higher persistence than switching to a bDMARD.

Conclusions

This scenario-based synthesis may offer a useful complement to existing RA recommendations, potentially supporting clinicians in treatment sequencing decisions following b/tsDMARD failure, though further validation in real-world settings would help confirm its applicability.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10067-026-08172-1.

Background

Biologic (b) and targeted synthetic (ts) disease-modifying antirheumatic drugs (DMARDs) have transformed the management of rheumatoid arthritis (RA), yet practical guidance on how to sequence therapies after failure of one or more targeted agents remains limited. The 2025 EULAR update of the recommendations for treating RA allows within-class cycling or switching mechanism, after failure without specifying when to favour one approach over the other [1]. In particular, recommendation 10 provides an overarching ‘any other bDMARD or tsDMARD’ option after failure but does not explicitly differentiate sequencing choices by the number and class/mechanism of previously failed targeted therapies, which are common decision points in routine care [1]. Moreover, at the time of publication, direct comparative evidence specifically addressing JAK inhibitor (JAKi)-to-JAKi cycling after JAKi failure was scarce, limiting the ability of broad recommendations to offer scenario-level direction in this setting [1]. This flexibility is appropriate for broad recommendations, but it often leaves clinicians without scenario-level granularity when choosing the next b/tsDMARD in routine care. For practising rheumatologists, such guidance does not always provide answers in questions of everyday care, where scenario-based evidence becomes important. This article arose following the update of the Sociedad Española de Reumatología (SER) consensus [2] and provides a more clinically oriented synthesis. While the SER consensus offers a helpful step toward stratification (e.g. distinguishing failure of one vs ≥ 2 TNF inhibitor (TNFi) and addressing post-non-TNFi/JAKi failure), it still does not map the evidence in a granular, scenario-by-scenario manner across common post-failure pathways [2]. Building on the overarching principles of EULAR [1] and the national consensus framework [2], our manuscript therefore aims to add a structured and pragmatic scenario–based roadmap that organises comparative evidence by both the number and mechanism of prior b/tsDMARD failures, translating heterogeneous data into practical post-failure decision pathways aligned with real-world clinical questions. Although this roadmap focuses on comparative effectiveness and treatment persistence, real-world treatment choices may also be influenced by access restrictions, reimbursement policies, and drug costs, which vary across healthcare systems and can affect prescribing patterns and adherence. A structured literature search was conducted in PubMed and the Cochrane Library from January 2010 to June 2025. Keysearch terms and MeSH concepts included ‘rheumatoid arthritis’, ‘b/tsDMARDs’, ‘biologic DMARDs’, ‘targeted synthetic DMARDs’, ‘treatment failure’, and ‘switching strategies’, combined using Boolean operators and adapted to each database. The detailed search strategies as implemented in each database are provided in Supplementary Table 1. Studies were selected based on their relevance to predefined clinical scenarios. Predefined inclusion criteria were as follows: adult patients (≥ 18 years) with RA, prior exposure to at least one bDMARD or tsDMARD (TNFi, non-TNFi/other mechanism of action (OMA) bDMARD, or JAKi), and comparative studies evaluating at least two targeted options with outcomes including clinical effectiveness (e.g. DAS28, CDAI/SDAI, ACR responses, remission/EULAR response), functional status (HAQ), and/or treatment persistence (drug survival). We included only English-language publications. We excluded studies reporting cost-only outcomes, and case series with < 50 patients (with a predefined exception in the post-JAKi failure scenario, where we allowed smaller studies because of scarce evidence). A total of 564 records were identified, including 477 from database searches (469 from MEDLINE and 8 from Cochrane Library) and 85 additional studies identified through manual search. After removal of duplicates and screening, 60 studies were ultimately included based on their relevance to the predefined clinical scenarios and inclusion criteria. A detailed flow chart of the study selection process is provided in Fig. 1. Study selection and full-text assessment were performed independently by two reviewers (CP-R and AMO), with disagreements resolved by consensus. Data extraction was performed using a structured approach focusing on study characteristics, patient population, prior treatment exposure, and key outcomes relevant to the predefined clinical scenarios. This framework facilitates the critical appraisal and application of evidence to therapeutic decisions. Within it, we examine the evidence for selecting the subsequent treatment—TNFi, b with OMA, or JAKi—after discontinuation of previous b/tsDMARD therapy. We highlight evidence from meta-analyses (MA), systematic reviews (SR), randomised controlled trials (RCTs), and real-world registries/cohorts (Fig. 2) and summarise the relative support for each option within each scenario to provide a pragmatic, clinically oriented roadmap (Fig. 3), which was constructed by consensus among authors after reviewing the evidence considering the following predefined items: (1) hierarchy of study design (SR/MA > RCT > prospective registries or cohort studies > retrospective database studies), (2) consistency of findings across studies, (3) sample size and overall amount of available data, and (4) key methodological limitations influencing confidence in the evidence, such as confounding by indication, selection bias, missing data, imprecision, and short follow-up. To further improve interpretability and explicitly link each clinical scenario with its supporting evidence base, we have included a summary table (Supplementary Table 2) detailing not only the number and type of studies supporting each therapeutic strategy, but also the overall direction of evidence, the main limitations, and a qualitative confidence assessment for each scenario. This table complements the visual grading represented in Fig. 3, where treatment options are categorised as preferred, acceptable, or limited support according to the strength, consistency, and overall confidence of the available evidence.

Fig. 1.

Flow diagram of study selection

Fig. 2.

Evidence by study type across post-failure treatment scenarios. Counts denote the number of available studies that specifically address each clinical scenario. Study type is colour-coded (darkest: meta-analyses and systematic reviews; then randomised controlled trials; registries and databases; cohorts). ‘0’ indicates that no studies were identified (very light cell). Abbreviations: TNF, tumour necrosis factor; OMA, other mechanism of action; b, biologic; tsDMARD, targeted synthetic disease-modifying antirheumatic drug; JAKi, Janus kinase inhibitor; RCT, randomised controlled trial

Fig. 3.

Scenario-based post-failure treatment pathways (switch vs cycle). For each scenario, we display the treatment options considered most appropriate according to an overall assessment of the available evidence, integrating: (1) study design hierarchy (SR/MA > RCT > prospective registries/cohorts > retrospective studies), (2) consistency of findings across studies, (3) sample size and overall amount of available data, and (4) key limitations influencing confidence (e.g. confounding by indication, selection bias, missing data, imprecision, or short follow-up). Options are colour-coded within each scenario as follows: green, preferred, supported by the most consistent and methodologically robust body of evidence in that scenario; blue, acceptable, supported by some evidence, but with lower certainty or consistency than the preferred option; coral, limited support, supported only by limited, small, inconsistent, or lower-confidence evidence. Arrows indicate mechanism switching versus within-class cycling, as labelled, and the badge denotes the highest level of support available (SRs/MAs, RCTs, or RWE). Abbreviations: RA, rheumatoid arthritis; b/tsDMARD, biologic/targeted synthetic disease-modifying antirheumatic drug; TNFi, tumour necrosis factor inhibitor; OMA, other mechanism of action (non-TNFi; e.g. RTX, rituximab; ABT, abatacept; TCZ, tocilizumab); JAKi, Janus kinase inhibitor; bDMARD, biologic DMARD; SR, systematic review; MA, meta-analysis; RCT, randomised controlled trial; RWE, real-world evidence

Scenario 1—therapy after failure of a single b/tsDMARD

Scenario 1A—therapy after failure of a single TNFi

This is the most frequent scenario in clinical practice, as TNFi remain the most used first-line biologics for RA. After failure of a first TNFi, evidence from a network MA [3], RCTs [4–7], and registries and cohorts [8–19] indicates that switching to a b with OMA, such as rituximab (RTX), abatacept (ABT), tocilizumab (TCZ), or a JAKi, is generally comparable and often superior to cycling to another TNFi. The MA ranked TCZ, RTX, ABT, and tofacitinib (TOFA) above TNFi after previous TNFi failure [3]. The largest RCT showed nearly two-fold higher odds of good or moderate EULAR responses at weeks 24 and 52 with RTX, ABT, or TCZ than with a second TNFi [4]. Two smaller RCTs reported no short-term DAS28 advantage for RTX or ABT [5, 6], and the EXXELERATE trial showed that many early TNFi non-responders still achieve low disease activity with another TNFi [7]. Observational data generally support better responses and longer persistence with b with OMA or JAKi than with TNFi cycling [8–19], although findings vary. These discrepancies likely reflect differences in follow-up duration, comparators, outcome definitions, and confounding inherent to observational designs (Supplementary Table 3). In summary, while both TNFi cycling and switching to a b with OMA are effective after failure of a single TNFi, the balance of evidence tends to favour switching to a b with OMA or JAKi, mainly in terms of drug persistence and, in some studies, effectiveness, although these findings should be interpreted cautiously given the contribution of observational evidence.

Scenario 1B—therapy after failure of a single b with OMA (non‑TNFi)

This scenario is less common in practice; however, the first b might be an OMA owing to trial allocation or specific clinical considerations. Observational data generally favour switching to a b with OMA or a JAKi rather than switching to a TNFi, with higher retention and fewer discontinuations in the GISEA and ANSWER registries [20, 21]. Nonetheless, a Japanese retrospective cohort reported better outcomes with a switch to TNFi versus ABT after failure of TCZ [22]. These inconsistencies likely once again reflect heterogeneity across studies (index drug, comparators, follow-up, and outcome definitions). Study-specific findings are in Supplementary Table 4. In our opinion, the limited and heterogeneous evidence in this scenario does not reveal a clear preference.

Other studies have analysed the choice of a second b/tsDMARD; however, they are not readily applicable to our scenario-based decisions because they pool second-line therapies regardless of the first targeted agent (TNFi, b with OMA, or JAKi) (see Supplementary Table 5) [23–29]. In this context, two phase 3 extension clinical trials further explored this question by analysing subgroups of patients initially treated with a TNFi or a JAKi (upadacitinib [UPA]), evaluating outcomes after switching to the alternate mechanism of action, and showing sustained long-term efficacy and acceptable safety [23, 24]. In these two latter studies, fewer than 20% of these patients may have previously received a bDMARD but discontinued it due to intolerance or had prior exposure for less than 3 months; patients who had discontinued a prior bDMARD due to inefficacy were excluded.

Scenario 2—therapy after failure of two or more b/tsDMARD

Scenario 2A—therapy after failure of two or more TNFi

In this scenario, we examine the evidence for patients who have received ≥ 2 TNFi excluding those who have received b with OMA or JAKi. Most studies analyse mixed cohorts and, except for one [30] that restricts inclusion to third or later lines of therapy, also include patients for whom one TNFi has failed, thus complicating interpretation. This limitation could bias the findings toward continued TNFi use, although that is not the observed trend (see below). A SR of five RCTs in RA patients whose TNFi failed found that TOFA achieved outcomes comparable to ABT, golimumab (GLM), RTX, and TCZ, with similar safety and fewer withdrawals due to lack of efficacy [31]. A MA across non-TNFi biologics and JAKi also found no major efficacy or safety differences, although TNFi were not included as second-line comparators [32]. Observational data generally favour switching mechanisms: RTX outperformed TNFi [30, 33, 34] and TCZ achieved higher remission and EULAR responses than TNFi [35]. Evidence comparing ABT with TNFi is mixed: one study showed comparable results [36] and another suggested better long-term control with ABT in the third line [37]. Large databases and a prospective cohort study reported higher persistence when switching to b with OMA/JAKi rather than cycling TNFi [38, 39]. Study details are shown in Supplementary Table 6.

In summary, reported data suggest that, after failure of multiple TNFi, switching to a b with OMA or a JAKi can be more effective. Differences between b with OMA are inconsistent.

Scenario 2B—therapy after failure of two or more targeted agents (TNFi, b with OMA, or JAKi)

Here, the focus is on failure of ≥ 2 targeted therapies in pooled, heterogeneous cohorts involving multiple mechanisms (TNFi and/or b with OMA and/or JAKi). Importantly, studies in which patients have failed only TNFi are not included here and are addressed exclusively in Scenario 2A. Scenario 2B is reserved for analyses in which previous therapies are reported irrespective of mechanism of action and no stratified (TNFi-only) results are available, preventing identification of a pure TNFi-failure cohort. The evidence base is highly heterogeneous because most studies also included pooled patients for whom only one targeted therapy failed in proportions that vary widely across cohorts. This pooling of prior exposures complicates interpretation, and even studies providing data that can be extrapolated to second- or third-line settings [22, 28, 29, 40] remain heterogeneous. Therefore, translation of results into practice requires caution. A recent EULAR SR on difficult-to-treat RA [41] showed that efficacy declines with increasing prior bDMARD failures for GLM, TCZ 4 mg IV, TOFA 5 mg, and baricitinib (BARI) 2 mg but remains more stable for UPA, filgotinib, and standard doses of TCZ, TOFA, and BARI. The biopsy-driven R4RA RCT demonstrated that molecular characterization of synovial tissue may help guide treatment decisions in patients with prior inadequate response to TNFi [42]. Observational studies generally favour non-TNFi options: the TOCERRA study, two large US claims analyses and a Korean registry reported higher persistence for TCZ (often ≥ 70%) and advantages for RTX over TNFi [43–46]. Results after failure of TCZ showed no advantage for ABT over TNFi [22]. In a large real-world claims-based study, switching from a TNFi to a non-TNFi therapy, including JAKi (TOFA), was associated with greater treatment persistence and modestly improved effectiveness compared with cycling to another TNFi [47]. Other more recent cohorts including JAKi show consistent patterns, with TCZ and TOFA maintaining high retention and low discontinuation across multiple datasets [28, 29, 48–54], and the OPAL registry identifying UPA as a long-lasting JAKi with no new safety signals [40]. Even in the absence of TNFi comparators, TCZ and TOFA matched or exceeded the efficacy of other non-TNF agents [55–57] (see Supplementary Table 7). These data suggest that after two or more prior b/tsDMARD failures in RA, switching out of class ensures higher persistence and effectiveness that is comparable or superior to TNFi use. IL-6 inhibitors or JAKi may be more efficacious than other b with OMA, although this observation should be interpreted with caution.

Scenario 3—therapy after failure of a JAKi

Studies address a key practical question: after failure of a JAKi, is a second JAKi preferable to changing mechanism? In most available datasets, this scenario includes patients treated with a JAKi in later lines and therefore often comprises mixed prior exposures to other targeted mechanisms (TNFi and/or b with OMA), although prior treatment history is not consistently stratified across studies. A 17-registry collaboration (~ 2000 patients) found similar effectiveness but higher retention with a second JAKi versus a bDMARD [58]. Another cohort showed longer survival and lower discontinuation with JAKi than with TNFi and intermediate results for b with OMA [59]. Smaller studies also support greater effectiveness of switching between JAKi [60, 61]. Data from the OPAL registry showed that after discontinuation of JAKi, switching to another JAKi was frequent, particularly in later treatment lines. Persistence across JAKi was comparable, with longer retention earlier in the sequence [62]. Observational data suggest that JAKi-to-JAKi cycling may be associated with higher persistence than switching to a bDMARD in some cohorts; however, these findings should be interpreted cautiously given heterogeneity, potential confounding, and incomplete reporting of prior b/tsDMARD exposure and reasons for discontinuation (see Supplementary Table 8).

Key messages

Current evidence supports flexibility across clinical decisions. The strongest data come from patients whose first TNFi failed, where both cycling and switching are valid, although evidence, largely derived from observational studies, tends to favour switching to a b with OMA in terms of treatment persistence and, in several studies, effectiveness.

Elsewhere, interpretation is more difficult, because studies often include patients with different prior therapies or treatment lines, limiting their applicability to routine practice and increasing the risk of scenario misclassification when evidence is mapped to specific post-failure pathways. In these settings, the available evidence generally suggests a possible advantage for switching to a b with OMA or JAKi, particularly after failure of multiple targeted therapies, but this pattern is based largely on observational data and should therefore be interpreted cautiously and should not be considered supported by evidence. In addition, the lack of stratified data according to the exact number and mechanism of prior treatment failures further limits scenario-specific inference. Of note, our analysis relies mainly on efficacy, effectiveness, and drug persistence. Much of the comparative evidence is observational and therefore remains susceptible to confounding by indication, channeling bias, and residual confounding. Heterogeneity in prior exposures, comparators, and treatment lines also limits the precision of the scenario-based model. As indicated in current recommendations [1, 2], treatment choice in RA should be individualized, integrating patient and treatment-related factors, such as comorbidities, safety issues, concomitant treatments, type and number of previously failed targeted therapies, or costs. In particular, patient-related factors such as cardiovascular disease, infection risk, history of malignancy, interstitial lung disease, age, frailty, concomitant glucocorticoid exposure, reproductive considerations, route of administration, and patient preferences may substantially influence the risk–benefit balance of a given sequencing strategy [63–67]. This work is not intended to amend existing recommendations that consider all b/tsDMARDs as adequate options in the different scenarios analysed here [1, 2] but may offer some additional considerations that could help inform clinical decision-making.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contribution

CPR and AMO reviewed the literature and drafted the manuscript. MGR performed the literature searches and reviewed the manuscript. LC and PDC supervised the methodology and reviewed the manuscript. JAG reviewed the manuscript.

Funding

No external funding was received for this work.

Declarations

Ethical approval

Not applicable.

Patient and public involvement

Neither the patients nor the public were involved in the design, conduct, reporting, or dissemination plans of this research.

Disclaimer

The views expressed in this article are those of the authors and do not necessarily reflect the official positions of their affiliated institutions.

Competing interests

CP-R has received research grants/honoraria from AbbVie, Pfizer, Novartis, Lilly, UCB, and Biogen. AMO has received research grants/honoraria from AbbVie, Alfasigma, Asacpharma, Gebro Pharma, Pfizer, Stada, UCB, and Sanofi. LC has not received fees or personal grants from any pharmaceutical company; however, her institution works by contract for pharmaceutical companies and other entities, such as BIOHOPE Scientific Solutions for Human Health S.L, BMS, Fresenius Kabi, Alfasigma, GSK, Lilly, Novo Nordisk Pharma SA, Nordic Pharma, Novartis Pfizer, Sandoz, and Sanofi. In addition, LC has participated on but did not receive personal remuneration from drug safety monitoring boards: Lilly and Hospital Clínico San Carlos. LC’s institution also receives EU funding via HORIZON (SQUEEZE, grant no. 101095052; SPIDeRR, grant no. 101080711; MDR-RA, grant no. 101155807) and IHI programmes (AutoPiX, grant no. 101194766). MGR has not received fees or personal grants from any pharmaceutical company. PDC has no conflicts of interest to declare in relation to this manuscript. JAG has received honoraria for lecturing, advisories, and Congress attendance from AbbVie, AstraZeneca, Galapagos, Gilead, Pfizer, Novartis, GSK, Lilly, MSD, and UCB.