Clinical and epidemiological researchers may want to identify or determine groups with similar characteristics that exist within a dataset or population. For this purpose, clustering techniques can be applied. However, clustering can be complex, and results can be misleading if the methods are not applied with fidelity.The aim of this paper is to provide practical guidance to (clinical) researchers in rheumatology and other fields who wish to explore or confirm (sub) groups within their study population, and are considering to apply clustering techniques to (patient) data. We discuss when clustering is useful for addressing your research question (and when it is not), the need to define the cluster concept, the choice of distance measure between 2 observations, the preprocessing of the data, the assessment of clustering tendency, the types of clustering methods, the determination of the number of clusters and end with the evaluation of the derived clustering solutions using visualizations and statistical measures.The following methods are presented in this paper: K-means, partitioning around medoids, hierarchical clustering, Density-Based Spatial Clustering of Applications with Noise, spectral clustering, fuzzy clustering, and latent class analysis. To illuminate the different considerations involved in clustering, we use a case example: secondary data from 895 people with rheumatoid arthritis, spondyloarthritis, and gout who completed the Health Literacy Questionnaire (HLQ). In this case example, we compute, appraise and evaluate clustering solutions using different methods in 6 steps, including statistical measures and expert opinion.The 6 steps proposed in this paper describe a systematic approach to cluster analysis to ensure all important aspects are considered. In the case example, different clustering methods led to different clustering solutions and insights. Qualitative interpretation by content experts was insightful here. Where possible, researchers should apply more than one clustering method fitting to the objective and reflect on eventual differences in solutions.

Secukinumab is an interleukin (IL)-17 inhibitor approved to treat patients with axial spondyloarthritis (axSpA), including radiographic (r)-axSpA and non-radiographic (nr)-axSpA. Treatment recommendations highlight the importance of individualized treatment of patients with axSpA according to symptoms, extra-musculoskeletal manifestations, and comorbidities. The objective of this post hoc analysis was to compare clinical responses to secukinumab treatment through Week 16 between younger and older patients with r-axSpA and nr-axSpA from six placebo-controlled phase 3 clinical studies. This post hoc analysis evaluated data pooled from 1427 patients with axSpA from the MEASURE 1–5 trials (NCT01358175, NCT01649375, NCT02008916, NCT02159053, and NCT02896127) and the PREVENT trial (NCT02696031). All patients were randomized to secukinumab 150 mg, 300 mg, or placebo through Week 16. Patient subgroups were defined based on median age at baseline (18 to < 40 years vs. ≥ 40 years). Treatment effect (secukinumab 150/300 mg vs. placebo) within each age subgroup and the difference in treatment effect between younger vs. older age subgroups were assessed. Secukinumab treatment demonstrated clinical efficacy and improvements in patient-reported outcomes vs. placebo at Week 16 in patients with r-axSpA or nr-axSpA irrespective of age. No difference in treatment effect of ASAS20, ASAS40, or ASDAS response was observed between younger vs. older patients with r-axSpA or nr-axSpA (all P>.05). Among patients with r-axSpA, younger patients experienced a larger treatment effect of secukinumab on SF-36 than older patients (difference in least-squares mean [LSM]: 2.2; P < .05). Among patients with nr-axSpA, younger patients experienced a larger treatment effect of secukinumab on the Berlin SIJ total edema score (difference in LSM: − 1.2; P < .05) and Total ASspi-MRI-a score (LSM, − 0.4; P<.05). Clinical efficacy of secukinumab vs. placebo at Week 16 was generally similar between older and younger patients with either r-axSpA or nr-axSpA.

OBJECTIVE:To compare frailty prevalence and factors driving frailty between older adults with RA and controls, using two conceptually distinct frailty instruments. METHODS:Cross-sectional data from the Studying Ageing in Rheumatoid Arthritis (STAR) study were used, including 207 patients with RA and 214 population controls aged 55-85 years. Frailty was assessed using the Groningen Frailty Indicator (GFI; 15 yes/no items; physical, cognitive, social, psychological domains; frail ≥ 4/15) in all participants and the Fried criteria (five yes/no items; physical domain; prefrail = 1-2, frail ≥ 3, and combined frail/prefrail ≥ 1 deficit) in a clinically assessed subgroup (RA: n = 88, controls: n = 96). Prevalence and overlap between instruments were described by group. Uni- and multivariable logistic (GFI: frailty vs robust) and Poisson (Fried: frail/prefrail vs robust) regressions assessed effects of age, group, the age*group interaction and additional covariates. RESULTS:Frailty prevalence was higher in RA than controls for GFI-frailty (34% vs 18%) and Fried-frailty/prefrailty (72% vs 50%). Among GFI-frail persons, 88% of patients with RA compared with 69% of controls were also frail or prefrail by Fried. Age was not associated with frailty in uni- or multivariable analyses for either instrument, in RA or controls (pinteraction age*group >0.10). The association with RA became insignificant after multivariable adjustment. Living alone, comorbidity score ≥1, higher fatigue and anxiety were associated with GFI-frailty, and higher BMI and poorer physical function with Fried-frailty/prefrailty. CONCLUSION:Older adults with RA show higher (pre)frailty prevalence than controls, which is associated with disease-related consequences and vulnerability rather than age, questioning the added value of frailty assessment beyond routine clinical evaluation.

BACKGROUND:The Outcome Measures in Rheumatology (OMERACT) group recommends pain as a core domain in clinical trials for shoulder disorders. This study evaluated the suitability of the Numerical Pain Rating Scale (NPRS) for measuring pain using the OMERACT Filter 2.2. METHODS:Following the OMERACT Handbook for instrument selection, a systematic review assessed the NPRS's construct validity, reliability, longitudinal construct validity, clinical trial discrimination, and thresholds of meaning. Articles were independently screened, appraised with the COSMIN-OMERACT Good Methods Checklist, and rated as green (good), amber (caution), red (poor), or white (no evidence). Only green and amber studies were included. Findings were summarized in a Summary of Measurement Properties table and discussed at the 2025 OMERACT Shoulder Special Interest Group workshop. RESULTS:Twelve studies, including 18 pieces of evidence met eligibility criteria. Three studies examined construct validity, two test-retest reliability, four longitudinal construct validity, six clinical trials discrimination and three thresholds of meaning. Six (33%) components of evidence had good methods and were rated green and 12 were rated amber. There was significant heterogeneity in NPRS versions and formats evaluated across studies, questioning the validity of synthesizing the data together. Therefore, 75% of respondents at the 2025 OMERACT conference agreed that synthesizing results from studies using different NPRSs was inappropriate. CONCLUSIONS:The NPRS cannot progress to the endorsement stage for the core domain of pain for shoulder conditions due to insufficient evidence for any single NPRS version. Current evidence is insufficient to support its use to measure pain in clinical trials of shoulder disorders.

Objectives The objective of this study was to evaluate and update the evidence on pharmacologic and interventional treatments for major organ involvement in Behçet syndrome (BS), in order to inform the 2025 update of the European Alliance of Associations for Rheumatology recommendations. Methods A systematic literature review was conducted using 2 distinct search strategies for pharmacologic and surgical/interventional therapies, covering major databases from October 2015 to November 2024. Controlled studies comparing active interventions with placebo or another active treatment were included. If no controlled trials were available for a specific research question, uncontrolled studies or case series with at least 10 patients with BS were included. Results Of 7128 citations, 69 studies on major organ involvement and tapering/withdrawal of immunosuppressives were included. Only 5 were randomised controlled trials (RCTs), 4 of them included patients with eye involvement, and 1 included patients with vascular or nervous system involvement. RCTs and comparative observational studies for eye involvement supported the use of monoclonal tumour necrosis factor alfa inhibitors and interferon alfa. The RCT that included patients with vascular or nervous system involvement favoured infliximab over cyclophosphamide based on complete response rates and safety. Observational data additionally supported the use of interferon alfa for venous thrombosis, whereas the role of adjunctive anticoagulation remains unclear. Noncomparative observational studies showed benefit with monoclonal tumour necrosis factor alfa inhibitors in patients with gastrointestinal involvement. Conclusions This systematic literature review provided evidence on the management of major organ involvement to inform the task force for developing the 2025 update of the European Alliance of Associations for Rheumatology recommendations for the management of BS.

Bimekizumab, a monoclonal IgG1 antibody that selectively inhibits IL-17F in addition to IL-17A, showed efficacy to 2 years in patients with axial spondyloarthritis (axSpA). In this post hoc analysis, we compare the impact of shorter versus longer symptom duration on the efficacy of bimekizumab to Week 104. Efficacy outcomes by symptom duration (≤ 2 [ASAS early axSpA definition] versus > 2 years; ≤ 5 versus > 5 years) were assessed across patients from BE MOBILE 1 and 2 (non-radiographic [NCT03928704]/radiographic axSpA [NCT03928743]) and the combined open-label extension (NCT04436640). (Relative) odds ratios and (relative) differences were calculated to compare 16-week bimekizumab versus placebo treatment effect and 104-week outcomes, and infer the significance of differences, between symptom duration subgroups. Analyses were neither powered for these comparisons nor multiplicity adjusted, and should be interpreted accordingly. Improved disease activity, physical function, fatigue, health-related quality of life and objective signs of inflammation were seen with bimekizumab versus placebo at Week 16 regardless of symptom duration. Outcomes were then sustained or improved with bimekizumab to Week 104 across all subgroups. 16-week bimekizumab versus placebo treatment effect was comparable between subgroups (e.g., ≤ 2-year versus > 2-year symptom duration relative odds ratio [95

Objectives To adjust the BASMI for measurement bias related to age, sex and height, and to compare scores and construct validity of a corrected metrology index (CASMI) against BASMI. Methods Spinal mobility data from non-axSpA individuals and severe axSpA patients were used to develop CASMI. For each BASMI component, the anchors for normal (score=0) and severe (score=10) mobility were redefined based on the 50th percentiles of non-axSpA individuals and the 95th percentile of patients with severe axSpA respectively. These anchors were individually adjusted for age, sex, and height using the regression coefficients from the MOBILITY study of non-axSpA individuals. Score ranges, floor/ceiling and construct validity (correlations and know-group discrimination) of BASMI and CASMI were assessed in an r-axSpA-population (OASIS cohort). It was hypothesized that CASMI would reduce correlations with age/height, while maintaining construct validity. Results Applying these corrections resulted in lower CASMI scores compared with BASMI (3.2 (2.0) vs 3.8 (1.6)), particularly in older and shorter individuals. Floor effects increased as more subjects were correctly classified within the normal mobility range, while ceiling effects remained unchanged. CASMI showed markedly reduced correlations with age and height, confirming a lower influence of these non-disease factors. Construct validity was maintained, as correlations with functional, disease activity, and structural outcomes were comparable to BASMI. Known-group discrimination also remained good, with standardized mean differences consistently above 0.80. Conclusions An age-, sex- and height-corrected mobility score, the CASMI, has been developed as a more personalised and valid measure of mobility in axSpA, compared to BASMI.

OBJECTIVES:To identify the most appropriate instrument to assess disease activity related to peripheral arthritis in axial spondyloarthritis (axSpA) and peripheral spondyloarthritis (pSpA) in clinical research. METHODS:This Assessment of Spondyloarthritis International Society (ASAS) project followed the Outcome Measures in Rheumatology (OMERACT) framework. Candidate instruments were identified through a systematic literature review (SLR). Domain match and feasibility were assessed by a working group (WG). A second SLR evaluated measurement properties (construct validity, test-retest reliability, responsiveness, clinical trial discrimination, and thresholds of meaning). Additional analyses were performed using individual patient data from randomised controlled trials (RCTs). Evidence was synthesised in summary of measurement properties tables, followed by a WG discussion and ASAS consensus voting. RESULTS:Composite disease activity instruments consistently outperformed single-item measures. Evidence from the second SLR was limited, particularly for axSpA with peripheral arthritis, but analyses of 13 RCTs strengthened the evidence. In both axSpA and pSpA, composite instruments, particularly the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), Axial Spondyloarthritis Disease Activity Score (ASDAS), and Disease Activity index for Psoriatic Arthritis based on 44 joints (DAPSA44), showed better construct validity, responsiveness, and clinical trial discrimination than joint counts alone. Clinical trial discrimination was particularly better for DAPSA44, making it the best-performing instrument across all measurement properties. Following ASAS voting, DAPSA44 was the preferred instrument, receiving 92% agreement for axSpA and 97% for pSpA. CONCLUSIONS:DAPSA44 demonstrated the most adequate measurement performance across axSpA and pSpA and was endorsed by ASAS as the recommended instrument for assessing disease activity related to peripheral arthritis in pSpA clinical research. Future clinical trials and research studies should report DAPSA44 when analysing disease activity related to peripheral arthritis in SpA.

Objective : To identify the most important and relevant items to consider when developing criteria to measure response to treatment in giant cell arteritis (GCA). Methods : As part of an ongoing project to develop response criteria for GCA, a 4-round web-based Delphi exercise was conducted. Participants included patients with GCA and health professionals with expertise in GCA. Participants rated the importance (1=lowest, 9=highest) of 51 items selected based on a systematic literature review, grouped into six domains, and could suggest additional items. Items scored 7-9 by at least 70% of health professionals and 70% of patients were considered critically important and to have reached consensus. In the last round of the Delphi, participants ranked the top 10 most relevant items. A task force (n=32 members) meeting followed to review the Delphi results, discuss rankings, and finalize the selection of items. Results : One hundred eighty-seven physicians and 85 patients, from 38 countries, participated in the Delphi exercise. Twenty-four (75%) task force members participated in the virtual meeting. Thirteen new items were proposed in round 1. Twenty-four items were rated as critically important. No items were excluded during any round, and consensus was not reached for 40 items. The top 3 highest rated items by both patients and physicians were headache (ranked highest by 52%), amaurosis fugax (highest=10%), and jaw claudication (highest=7%). Twelve items were selected for the next phase of the project. Conclusion : Experts and patients identified 12 items considered important for measuring response to treatment in GCA.

Objective To characterise patients with ‘very early’ axial spondyloarthritis (axSpA), defined as a duration ≤1 year of back pain and to determine the effectiveness of a first tumour necrosis factor inhibitor (TNFi) in very early versus established axSpA in a large observational registry.Methods We included a total of 3324 patients with axSpA from the Swiss Clinical Quality Management in Rheumatic Diseases registry with available data on duration of back pain (≤1 year=very early axSpA, n=441; >1 year and ≤2 years=early axSpA, n=218; >2 years=established axSpA, n=2575). A first TNFi was started in 31%, 38% and 36% of patients with very early, early and established axSpA. Adjusted logistic regression models were used to compare the probability of achieving low disease activity status according to the Axial Spondyloarthritis Disease Activity Score (ASDAS <2.1) at 1 year. Drug survival was analysed with multiple-adjusted Cox proportional hazards models. Missing data were handled using multiple imputation by chained equations.Results Objective signs of inflammation were more prevalent in very early disease. No difference was found regarding the achievement of ASDAS <2.1 after adjustment for age, sex, human leucocyte antigen-B27 status, education, body mass index, smoking, ASDAS and sacroiliac inflammation on MRI (OR 1.08, 95% CI 0.70 to 1.68 in very early vs established axSpA). Similarly, no significant difference in TNFi retention was found in very early versus established axSpA (HR for drug discontinuation 1.05, 95% CI 0.84 to 1.31).Conclusion We found no evidence for a better effectiveness of TNFi in patients with very early versus established axSpA.

Objectives: There are limited data on how NSAID use and dosage influence hypertension risk in axial spondyloarthritis (axSpA). We evaluated the association between NSAID use and incident hypertension in a longitudinal axSpA inception cohort.Methods: We analyzed data from the DEvenir des Spondylarthopathies Indifférenciées Récentes (DESIR) cohort, with visits every six months for two years, then annually up to 10 years. Hypertension was defined by self-report, antihypertensive medication use, and/or systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90 mmHg on ≥2 visits. NSAID use over the prior six months was quantified using the ASAS NSAID index at each visit, categorized as high-dose (≥50) or low/no use. Cox proportional hazards models with time-varying NSAID exposure and confounder adjustment were used to study the relationship between NSAID use and incident hypertension.Results: We included 631 patients with recent onset axSpA without baseline hypertension (mean age 33 years, 54% female, 247 [39.1%] patients with high-dose NSAID use). Over 10-year follow-up, 88 developed incident hypertension. Multivariable models showed no significant association between NSAID use and incident hypertension (hazard ratio 1.01, 95% confidence interval 1.00-1.02, main analysis) across multiple exposure definitions after accounting for baseline and time-varying demographic and clinical covariates including disease activity and other medication use.Conclusion: In this axSpA inception cohort, we found no significant association between NSAID use and the development of hypertension. These findings, if confirmed in other studies, can provide reassurance to prescribe NSAIDs, particularly in younger patients with axSpA.

Imaging, particularly magnetic resonance imaging (MRI) scans of the sacroiliac joints, has become central in the evaluation of axial spondyloarthritis (axSpA), partly due to the absence of reliable biomarkers and variable human leukocyte antigen B27 (HLA-B27) prevalence. However, its increasing use as a standalone diagnostic tool represents a conceptual shift that risks undermining clinical reasoning. This viewpoint critically examined the specificity and limitations of imaging findings and highlighted that lesions such as bone marrow oedema and structural changes may also occur in non-SpA populations, including healthy individuals and those with mechanical back pain. Misinterpretation, rather than intrinsic imaging limitations, is a major driver of diagnostic error.In this viewpoint it is argued that imaging should serve a supportive and prognostic role, integrated within a broader clinical framework that includes patient history, clinical features, and laboratory data. Overreliance on imaging contributes to diagnostic inflation, particularly in non-adiographic and atypical populations, with implications for patient care and healthcare systems. Emerging tools such as artificial intelligence may amplify these challenges if not carefully validated.Recentring clinical reasoning, promoting multidisciplinary collaboration, and ensuring imaging is requested and interpreted within an appropriate clinical context are essential steps forward. Ultimately, imaging should inform but not replace clinical judgement in the diagnosis of axSpA.

Axial spondyloarthritis is an immune-mediated inflammatory disease that primarily affects the sacroiliac joints, the spine, and the peripheral entheses and joints. The disease is associated with reduced functional ability and quality of life due to pain, fatigue, stiffness, sleep disruption, and inflammation-induced structural damage. The pathogenesis of axial spondyloarthritis involves, among other etiologies, the activation of immune-mediated proinflammatory cytokines, including tumor necrosis factor alpha, which is inhibited by golimumab, an approved treatment for the disease. Golimumab is produced using a recombinant cell line, and the antibody was derived from genetically modified mice immunized with human tumor necrosis factor alpha. This narrative review presents the primary efficacy and safety endpoints from the pivotal randomized controlled trials of golimumab. Additionally, it presents data from the post hoc analyses and real-world studies, which demonstrated benefits including decreased back pain, sleep disruption, fatigue, and disease activity as well as improved quality of life and work productivity and prolonged drug survival. Not only was golimumab associated with improvements in quality of life, general health status, and daily productivity, but a decrease in disease activity was significantly associated with these outcomes. Furthermore, golimumab was found to have a favorable safety profile, and the adverse event pattern observed was consistent with other tumor necrosis factor alpha inhibitors. Collectively, these studies showed that golimumab is an effective and safe treatment for axial spondyloarthritis.

OBJECTIVE:This study aimed to investigate the relationship between spinal axial spondyloarthritis (axSpA)-related lesions and degenerative lesions (DLs) over 10 years (10Y). METHODS:Whole spine MRI and cervical/lumbar spine radiographs at baseline/5Y/10Y from patients with axSpA from the DESIR cohort were assessed for axSpA-related lesions and DLs by three independent readers, different teams for the two lesion types. We used multilevel (patient and vertebra, considering consensus across readers), standard and time-lagged autoregressive generalized estimating equation (GEE) models. The relationship between syndesmophytes and the subsequent development of osteophytes/syndesmophytes in adjacent vertebrae on radiographs was analysed using a time-lagged autoregressive GEE model, after excluding vertebrae with both lesions. All models were adjusted for age, sex, HLA-B27 status, BMI, smoking and job type, and bDMARDs during follow-up. RESULTS:Data from 326 patients (35 [S.D. = 9] years; 46% men) showed a significant association between axSpA-related lesions on MRI and the total number of DLs on MRI, though the effect sizes were small (β-coefficients: 0.07-0.17). On radiographs, paravertebral syndesmophytes were significantly associated with the total number of DLs (β-coefficient: 0.37; 95%CI: 0.26-0.48). However, these associations were not found in time-lagged autoregressive models. Syndesmophytes increased the risk of adjacent syndesmophyte [odds ratio (OR):6.92; 95%CI: 2.44-19.61], but not of osteophytes (OR: 1.05; 95%CI: 0.25-4.36). CONCLUSION:Although significant associations were found between axSpA-related lesions and DLs at the same time point, no temporal relationship was observed. On radiographs, syndesmophytes increased the risk of syndesmophytes at adjacent levels, but there was no association with osteophyte development. AxSpA-related lesions and DLs coexist, but progress independently of each other.

Objectives This study aimed to determine the effectiveness of a first tumour necrosis factor inhibitor (TNFi) in patients with short and long symptom duration in axial spondyloarthritis (axSpA), using alternative cut-offs for back pain duration: ≤1 year as very early, >1 year, and ≤5 years as intermediate axSpA, and >5 years as late axSpA. Methods A total of 1161 patients with axSpA from the Swiss Clinical Quality Management registry, who had data on the duration of back pain and initiated their first TNFi, were included. Patients were categorised as follows: very early axSpA (N = 138), intermediate axSpA (N = 329), and late axSpA (N = 694). Adjusted logistic regression models were used to compare the likelihood of achieving a low disease activity status (Axial Spondyloarthritis Disease Activity Score [ASDAS] <2.1) at 1 year, adjusting for age, sex, human leucocyte antigen-B27, educational level, body mass index, ASDAS, current smoking, and presence of inflammation on magnetic resonance imaging of the sacroiliac joints. Drug survival was analysed using multiple-adjusted Cox proportional hazards models. Missing data were addressed through multiple imputation by chained equations. Results Higher disease activity parameters were found at baseline in very early axSpA. However, no significant differences were observed in treatment response or TNFi retention between very early and late axSpA (odds ratio for ASDAS < 2.1 response 0.99, 95% CI: 0.62-1.59 and hazard ratio for drug discontinuation 1.09, 95% CI: 0.87-1.37). Conclusions Based on this real-world observational data, no evidence was found to suggest a superior effectiveness of TNFi in patients with very early axSpA compared with those with late axSpA.

OBJECTIVES:To assess sex-based differences in response to bimekizumab in patients with axial spondyloarthritis (axSpA) from two phase 3 trials. METHODS:Improvements to Week 52 in measures of disease activity (ASAS40, ASDAS <2.1, BASDAI) and function (BASFI), pain, fatigue, health-related quality of life (HRQoL) and objective signs of inflammation (MRI, hs-CRP) were assessed post hoc in patients with non-radiographic (nr-) and radiographic (r-)axSpA from BE MOBILE 1 (NCT03928704) and BE MOBILE 2 (NCT03928743), respectively. Comparisons were made between sexes at Week 16 (bimekizumab versus placebo treatment effect) and Week 52 (treatment response in bimekizumab-randomised patients only) using odds ratios for dichotomous outcomes and difference values for continuous outcomes. For hs-CRP, ratio to baseline was calculated due to skewed distribution. RESULTS:While sex distribution was balanced in nr-axSpA (54.3% male), 72.3% of patients with r-axSpA were male. At Week 16, bimekizumab-randomised male patients typically had better treatment responses versus placebo compared with female patients across ASAS40, ASDAS <2.1 and improvements in BASDAI, BASFI and measures of HRQoL. At Week 52, female patients had longer-term improvements across these outcomes, though male patients continued to have numerically higher treatment responses. In contrast, active inflammation (MRI, hs-CRP) was reduced to similarly low levels in both sexes at Week 16 and maintained to Week 52, despite baseline differences. CONCLUSIONS:Although male patients had earlier clinical responses to bimekizumab treatment, female patients demonstrated longer-term improvements in composite and patient-reported outcomes. Both sexes had substantial improvements in objective inflammation at Week 16 and 52, despite baseline differences. TRIAL REGISTRATION:Clinicaltrials.gov; NCT03928704 and NCT03928743.

OBJECTIVES:To assess whether symptom duration modifies flare risk after biologic/targeted synthetic DMARD (b/ts DMARDs) withdrawal in axial spondyloarthritis (axSpA) following remission/inactive disease. METHODS:We systematically identified randomized placebo-controlled trials evaluating withdrawal or tapering of b/tsDMARDs in axSpA through a previous systematic literature review. Eligible studies included adults with axSpA who achieved inactive disease or remission by ASDAS and were randomized to continuation or withdrawal/tapering, with available flare data. Patient-level data were obtained from Vivli and stratified by symptom duration thresholds of≤2, 3, 4, or 5years. Relative risks (RRs) of flare for continuation versus withdrawal were calculated within each subgroup, and relative risk ratios (RRRs) were estimated. Random-effects meta-analysis was performed. A subgroup analysis included only patients with nr-axSpA. RESULTS:Three RCTs involving 773 patients were included, evaluating adalimumab, certolizumab pegol, and ixekizumab versus placebo; no tsDMARD or tapering studies were eligible. Continuation of bDMARDs was consistently associated with fewer flares than withdrawal. Symptom duration did not significantly modify flare risk overall. In the overall axSpA population, pooled RRRs showed no statistically significant effect modification: 0.61 (95% CI: 0.29-1.28) for≤2years, 0.77 (0.54-1.12) for≤3years, 0.83 (0.45-1.54) for≤4years and 0.82 (0.49-1.37) for≤5years. In nr-axSpA (n=508), pooled RRRs favoured shorter symptom duration at≤4years (0.25 [0.07-0.96]) but not at≤2,≤3 or≤5years. CONCLUSION:Symptom duration did not consistently modify flare risk after bDMARD withdrawal in axSpA overall, although exploratory findings suggest a potential effect in early nr-axSpA.