Prevalence and clinical factors associated with difficult-to-manage axial spondyloarthritis: a multicentre Italian cohort study with meta-analysis Authors: Giulio Forte, Gilberto Cincinelli, Daniele Mauro, Ilenia Pantano, Matteo Ferrito, Saviana Gandolfo, Andrea Amati, Antonio Ciancio, Miriam Longo, Maura Raimondi, Maria Manara, Enrico Tirri, Roberto Caporali, Francesco Ciccia Published: 2026 ABSTRACT AbstractBackground:The concept of difficult-to-manage axial spondyloarthritis (D2M-axSpA) has recently been introduced to describe patients with a persistent disease burden despite appropriate treatment. However, its real-world prevalence and associated clinical features remain insufficiently characterised.Objective:To estimate the prevalence of D2M and treatment-refractory (TR) axSpA in routine clinical practice, identify factors associated with D2M, and assess the consistency of selected D2M-associated features across different cohorts.Design and methods:A cross-sectional study was conducted in 252 patients with axSpA from two Italian tertiary centres. The 2024 Assessment of Spondyloarthritis International Society consensus-based definition was applied to identify D2M and TR axSpA. Clinical, laboratory and imaging characteristics were compared between D2M and non-D2M patients. Factors independently associated with D2M were evaluated using multivariable logistic regression. To explore consistency across populations, an exploratory meta-analysis was performed using harmonised data from the published Argentinian Reuma-Check cohort.Results:Among 252 patients with axSpA, 16 (6.3%) fulfilled the definition of D2M, and 8 (3.2%) met criteria for TR axSpA. Patients with D2M exhibited higher disease activity (Axial Spondyloarthritis Disease Activity Score 2.50 ± 0.62 vs 1.24 ± 0.65; p < 0.001), greater axial pain (2.50 ± 2.75 vs 0.80 ± 1.63; p = 0.017) and worse global assessments (Patient Global Assessment 3.20 vs 1.16; p < 0.001). In multivariable analysis, fibromyalgia (odds ratios (OR) 5.5, 95% confidence interval (CI): 1.2–25.6) and a history of uveitis (OR 3.3, 95% CI: 1.0–11.1) were independently associated with D2M status. Current nonsteroidal anti-inflammatory drug (NSAID) use (OR 3.8, 95% CI: 1.2–11.9) was also associated with D2M, most likely reflecting higher symptom burden rather than a causal relationship. Meta-analytic findings identified psoriasis (risk ratios (RR) 2.17; p = 0.008), smoking (RR 1.53; p = 0.045), uveitis (RR 2.45; p = 0.044) and radiographic structural damage (RR 1.84; p = 0.004) as consistent features associated with D2M axSpA.Conclusion:D2M axSpA was identified in approximately 6.3% of patients, while 3.2% fulfilled criteria for TR axSpA. Fibromyalgia, NSAID use and uveitis were independently associated with the D2M phenotype, whereas meta-analytic evidence highlighted additional associations with psoriasis, smoking and structural damage. These findings support the multifactorial nature of D2M axSpA, in which both inflammatory and non-inflammatory mechanisms contribute to disease complexity.Keywords: clinical phenotypes, difficult-to-manage axSpA, fibromyalgia, psoriasis, uveitis Introduction Axial spondyloarthritis (axSpA) is a chronic inflammatory disease that primarily affects the axial skeleton (sacroiliac joints and spine), leading to pain, stiffness, functional impairment and thus reduced quality of life (QoL). 1 Over the past two decades, the introduction of biologic and targeted synthetic disease-modifying antirheumatic drugs (b/tsDMARDs) has significantly improved the management of axSpA, providing considerable control of symptoms and inflammation for most patients. 2 However, up to 40% of patients fail to achieve sustained low disease activity or remission despite the use of targeted therapies, with a response rate even lower for subsequent therapies, highlighting the dimension of refractoriness in axSpA. 3 In this scenario, the concept of difficult-to-treat (D2T) was first introduced in rheumatoid arthritis (RA) 4 to identify patients with persistent disease activity and disability despite exposure to multiple therapies, often in the context of comorbidities, treatment intolerance or non-inflammatory pain. 5 This framework has subsequently been adapted to psoriatic arthritis6,7 and, more recently, to axSpA. 3 Given the complexity of axSpA, characterised by heterogeneous clinical domains, variable response to treatment and the frequent coexistence of extra-musculoskeletal manifestations, the term difficult-to-manage (D2M) has been proposed to better capture the multidimensional challenges in patient care, extending beyond pharmacological refractoriness alone.8,9 The first analysis applying this concept to axSpA cohorts has recently emerged, underscoring both the prevalence and features of such patients and the need for further validation in real-world settings.10,11 Identifying and characterising D2M-axSpA patients holds direct implications for treatment decision-making, healthcare resource allocation and patient outcomes. These patients often experience high symptom burden, frequent treatment switches and impaired QoL, yet remain poorly defined within clinical pathways. 12 Thus, operationalising the D2M construct in axSpA may help inform more tailored and multidisciplinary care strategies. Nevertheless, data on the prevalence, clinical correlates and practical implications of the D2M-axSpA definition remain limited, and its utility in routine clinical practice is yet to be clarified. The primary objective of this study was to apply the recently proposed definition of D2M-axSpA to a real-world cohort of adult patients with axSpA, to characterise its prevalence and associated clinical features. Specifically, we aimed to: (i) assess and compare baseline characteristics of patients with D2M-axSpA versus those without these phenotypes, describing demographic, clinical, laboratory and imaging characteristics; (ii) to explore potential baseline variables independently associated with the D2M phenotypes and (iii) for D2M patients, a meta-analysis was conducted combining the MINA cohort (Milan–Naples) and Reuma-Check cohorts to improve the power of the analysis and to explore the consistency of selected clinical predictors across independent datasets. Methods Study design This multicentre, cross-sectional study included a large real-world Italian cohort of patients with axSpA recruited from two tertiary Rheumatology centres (Milan and Naples), forming the MINA Cohort. The MINA cohort is an investigator-initiated, multicentre observational cohort that has not been previously published. The study population consisted of 252 adult patients (⩾18 years) fulfilling the 2022 Assessment of Spondyloarthritis International Society (ASAS) classification criteria for axSpA. 13 Patients were consecutively enrolled between 1 February 2025 and 1 April 2025, representing the predefined data-collection window for the cross-sectional assessment. D2M status was adjudicated at the enrolment visit according to the 2024 ASAS consensus-based definition. Clinical variables and comorbidities (e.g. psoriasis, uveitis, inflammatory bowel disease (IBD) and other diagnoses) were recorded as cumulative ‘history of’ events occurring up to the time of inclusion, whereas Patient Reported Outcomes (PROs) and disease activity measures (including Axial Spondyloarthritis Disease Activity Score (ASDAS) and pain/global assessments) referred to the enrolment visit. Before inclusion, alternative diagnoses were excluded by expert rheumatologists based on clinical, laboratory and imaging evaluations. All patients had a confirmed diagnosis of axSpA and had been under clinical follow-up before study inclusion. To ensure applicability of the ASAS D2M definition, only patients with a disease duration of at least 12 months at the time of inclusion were considered eligible for analysis. No formal sample size calculation was performed, as this was a descriptive and exploratory study including all consecutive eligible patients during the study period. Additionally, selected aggregate data from the published Argentinian Reuma-Check cohort were included for an exploratory secondary meta-analysis. 11 As individual-level data were not accessible, only summary-level estimates for harmonised variables were pooled to assess cross-cohort consistency of selected associations; differences in study design and time anchoring (baseline at diagnosis with longitudinal follow-up in Reuma-Check vs cross-sectional assessment at enrolment in MINA) were considered in interpretation. The primary outcome was to assess the prevalence of the D2M-axSpA phenotype in the MINA cohort and to describe associated clinical, demographic and radiological characteristics. Secondary, exploratory outcomes included identifying factors independently associated with D2M status and evaluating cross-cohort consistency of selected associations through pooled meta-analytic estimates using publicly available summary-level data. The reporting of this study conforms to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement. Definition of D2M and treatment-refractory axSpA D2M-axSpA was defined according to the 2024 ASAS definition. 8 To ensure an appropriate assessment of the D2M construct, only patients with a disease duration of at least 1 year at enrolment were considered eligible for D2M classification, allowing sufficient therapeutic exposure and follow-up. According to the ASAS definition, D2M axSpA required the coexistence of three elements during follow-up: (1) failure of at least two b/tsDMARDs with different mechanisms of action; (2) persistence of active disease, documented by at least one of the following: (a) high or very high disease activity (ASDAS ⩾2.1), (b) signs or symptoms suggestive of active disease, including musculoskeletal or extra-musculoskeletal manifestations, active inflammation on magnetic resonance imaging (MRI) and/or elevated C-reactive protein (CRP) levels, (c) rapid radiographic spinal progression, defined as the appearance of two or more new syndesmophytes and/or bony bridges within 2 years or (d) patient-reported outcomes indicating persistent axSpA-related symptoms associated with impaired QoL despite otherwise controlled disease and (3) recognition by the patient and/or the treating physician of problematic disease management. Criteria related to impaired QoL and perceived problematic disease management were recorded as binary (yes/no) variables at the time of evaluation, based on clinical judgement by the treating rheumatologist and patient input. In line with the ASAS consensus, 8 no validated instrument or formal scale is currently available for these criteria, and therefore no predefined patient-reported outcome measure or cut-off value was applied. Treatment refractory status was defined as patients meeting D2M definition, additional objective signs of inflammation, including elevated CRP and/or active sacroiliitis on MRI, and clear evidence of high disease activity (ASDAS ⩾2.1). Data collection Comprehensive demographic, clinical, laboratory and imaging data were collected. Baseline variables included sex, age, age at symptom onset, disease duration, diagnostic delay, body mass index (BMI) and smoking history, as part of routine clinical assessment. Clinical features were recorded as history of disease manifestations (ever) and comprised peripheral arthritis, enthesitis, dactylitis, uveitis, psoriasis, IBD, fibromyalgia and osteoporosis. Fibromyalgia was defined according to the 2016 revised classification criteria based on the 2010/2011 American College of Rheumatology criteria and was systematically assessed in the cohort. 14 Osteoporosis was defined by a history of prior fragility fracture and/or low bone mineral density, with low bone mineral density defined as a T-score ⩽ −2.5. Laboratory parameters included erythrocyte sedimentation rate (ESR), CRP and human leukocyte antigen B 27 (HLA-B27) status. Imaging assessments considered the presence of radiographic sacroiliitis, spinal structural damage (the presence of syndesmophytes in different spinal tracts) and MRI features of active inflammation (in the previous 6 months). Patient- and physician-reported outcomes were recorded using the visual analogue scale (VAS) for axial and peripheral pain, Patient Global Assessment (PGA), Physician Global Assessment (PhGA) and the Axial Spondyloarthritis Disease Activity Score based on CRP (ASDAS-CRP). Additional data included perceived active disease, impaired QoL and current treatments, including nonsteroidal anti-inflammatory drugs (NSAIDs), conventional synthetic DMARDs (csDMARDs), b/tsDMARDs and time to first bDMARD initiation. Statistical analyses Descriptive statistics were used to summarise baseline characteristics of the entire cohort and of the D2M cohort (Table 1). Continuous variables were reported as mean and standard deviation (SD). Categorical variables were presented as absolute counts and percentages. Group comparisons were conducted for D2M groups. Continuous variables were compared using Welch’s t test or the Mann–Whitney U test, as appropriate. Categorical variables were analysed using the Chi-square or Fisher’s exact test. For each candidate predictor, we fitted univariable logistic regression models to estimate odds ratios (ORs) with 95% confidence intervals (CIs). Variables with p < 0.10 or established clinical relevance were considered for multivariable modelling. The multivariable analysis employed a parsimonious logistic regression with elasticnet penalisation to account for collinearity, including age, sex, BMI, HLA-B27 status and significant univariable associated factors; results are reported in a forest plot as adjusted ORs with 95% CIs (Figure 2). Variables included in the ASAS definition of D2M were excluded from the predictive modelling process (Supplemental Table 1). Table 1.Baseline demographic, clinical, laboratory, imaging and treatment characteristics of the overall MINA cohort.VariableN = 252Male sex, n (%)182/252 (72%)Age, mean ± SD (years)50.67 ± 13.96Radiographic axSpA, n (%)187/231 (81%)HLA B-27 positive, n (%)108/167 (65%)IBD (ever), n (%)15/252 (6.0%)Peripheral arthritis (ever), n (%)64/251 (25%)Enthesitis (ever), n (%)24/250 (9.6%)Uveitis (ever), n (%)37/251 (15%)Dactylitis (ever), n (%)9/251 (3.6%)Psoriasis (ever), n (%)26/250 (10%)Age at symptoms onset, mean ± SD (years)33.02 ± 13.32Disease duration, mean ± SD (years)13.69 ± 9.93Diagnostic delay, mean ± SD (years)5.19 ± 5.84Smoking (ever), n (%)50/252 (20%)BMI, mean ± SD (kg/m2)25.55 ± 4.95Osteoporosis, n (%)16/252 (6.3%)Fibromyalgia, n (%)13/252 (5.2%)ESR (mm/h), mean ± SD12.61 ± 12.17CRP (mg/L), mean ± SD3.58 ± 5.39PhGA, mean ± SD0.32 ± 0.91PGA, mean ± SD1.31 ± 1.88VAS axial pain, mean ± SD0.92 ± 1.78VAS peripheral pain, mean ± SD0.43 ± 1.17ASDAS, mean ± SD1.32 ± 0.72Radiographic sacroiliitis, n (%)156/165 (95%)Spinal radiographic changes, n (%)62/252 (25%)MRI sacroiliitis, n (%)33/252 (13%)Structural damage on X-ray, n (%)166/181 (92%)TNFi use, n (%)179/252 (71%)IL17i use, n (%)54/252 (21%)JAKi use, n (%)19/252 (7.5%)Current NSAIDs use, n (%)32/252 (13%)Current csDMARD use, n (%)9/252 (3.6%)Descriptive statistics of the MINA Cohort at baseline. Categorical variables are reported as numbers and percentages (n/N, %), and continuous variables as mean ± SD.ASDAS, Axial Spondyloarthritis Disease Activity Score; axSpA, axial spondyloarthritis; BMI, body mass index; CRP, C-reactive protein; csDMARDs, conventional synthetic DMARD; ESR, erythrocyte sedimentation rate; HLA, human leukocyte antigen; IBD, inflammatory bowel disease; IL17i, interleukin-17 inhibitor; JAKi, JAK inhibitor; MRI, magnetic resonance imaging; NSAIDs, non-steroidal anti-inflammatory drugs; PGA, Patient Global Assessment; PhGA, Physician Global Assessment; SD, standard deviation; TNFi, tumour necrosis factor inhibitor; TR-axSpA, treatment-resistant axial spondyloarthritis; VAS, visual analogue scale. Baseline demographic, clinical, laboratory, imaging and treatment characteristics of the overall MINA cohort. Descriptive statistics of the MINA Cohort at baseline. Categorical variables are reported as numbers and percentages (n/N, %), and continuous variables as mean ± SD. ASDAS, Axial Spondyloarthritis Disease Activity Score; axSpA, axial spondyloarthritis; BMI, body mass index; CRP, C-reactive protein; csDMARDs, conventional synthetic DMARD; ESR, erythrocyte sedimentation rate; HLA, human leukocyte antigen; IBD, inflammatory bowel disease; IL17i, interleukin-17 inhibitor; JAKi, JAK inhibitor; MRI, magnetic resonance imaging; NSAIDs, non-steroidal anti-inflammatory drugs; PGA, Patient Global Assessment; PhGA, Physician Global Assessment; SD, standard deviation; TNFi, tumour necrosis factor inhibitor; TR-axSpA, treatment-resistant axial spondyloarthritis; VAS, visual analogue scale. Forest plots of multivariate logistic regression models for D2M- and D2T-axSpA-associated factors. Forest plots showing the OR with 95% confidence intervals for variables independently associated with D2M-axSpA. For D2M, uveitis, fibromyalgia and NSAIDS use were significantly associated with the phenotype. The vertical dashed red line indicates OR = 1 (null effect). Only retained variables are shown in Figure 2. axSpA, axial spondyloarthritis; D2M, difficult-to-manage; D2T, difficult-to-treat; NSAIDS, nonsteroidal anti-inflammatory drugs; OR, odds ratios. To assess the consistency of phenotypic features across cohorts, we first descriptively compared baseline characteristics of D2M-axSpA patients from the MINA and Reuma-Check datasets. For meta-analytical purposes, albeit exploratory, a fixed-effect model was applied to clinically relevant and common variables available in both cohorts. This model was chosen for its interpretability and simplicity, considering the small number of included cohorts and the limited between-study variance. Pooled risk ratios (RRs) with 95% CIs were calculated, and heterogeneity was assessed via the I2 statistic (Table 3). Descriptive and meta-analyses were performed using R software (version 4.3.2, R Foundation for Statistical Computing). The following R packages were used: dplyr for data manipulation, ggplot2 and cowplot for figure generation, meta for meta-analysis officer and flextable for table rendering and magick for image processing. Univariable and multivariable logistic regression analysis were performed with Python 3.11 (pandas, scikit-learn, statsmodels). Table 3.Meta-analysis of clinical predictors of D2M-axSpA in the merged MINA and Reuma-Check cohorts.VariablePooled RR (95% CI) p I 2 Elevated CRP 1.64 (1.15–2.33) 0.006 94.1% Enthesitis1.26 (0.68–2.32)0.4600.0%HLA-B27 positive1.15 (0.67–1.97)0.6110.0%MRI sacroiliac lesions1.14 (0.84–1.53)0.39251.8%Male sex0.65 (0.30–1.40)0.27350.9%Peripheral arthritis1.58 (0.83–3.00)0.1630.0% Psoriasis 2.23 (1.22–4.07) 0.009 0.0% Smoking (ever) 1.79 (1.08–2.95) 0.023 11.2% Structural damage 2.33 (1.24–4.38) 0.009 0.0% Uveitis 2.45 (1.02–5.85) 0.044 0.0% Meta-analysis of clinical predictors of D2M-axSpA in the merged MINA and Reuma-Check cohorts. The table shows pooled RR with 95% CI, p-values and heterogeneity estimates (I2). Variables with I2 > 50% indicate moderate-to-high heterogeneity.Statistically significant variables are shown in bold.axSpA, axial spondyloarthritis; CI, confidence interval; CRP, C-reactive protein; D2M, difficult-to-manage; I2, heterogeneity index; RR, risk ratio. Meta-analysis of clinical predictors of D2M-axSpA in the merged MINA and Reuma-Check cohorts. Meta-analysis of clinical predictors of D2M-axSpA in the merged MINA and Reuma-Check cohorts. The table shows pooled RR with 95% CI, p-values and heterogeneity estimates (I2). Variables with I2 > 50% indicate moderate-to-high heterogeneity.Statistically significant variables are shown in bold. axSpA, axial spondyloarthritis; CI, confidence interval; CRP, C-reactive protein; D2M, difficult-to-manage; I2, heterogeneity index; RR, risk ratio. Ethical considerations The study protocol was approved by the Ethics Committees of the participating centres (Milan Ethics Committee 138_1999, Naples Ethics Committee 2021.001336). All procedures complied with the Declaration of Helsinki. Results Descriptive analysis A total of 252 patients with axSpA were analysed; most were male (182/252, 72%) and had radiographic axSpA (187/231, 81%), with a mean age of 50.67 ± 13.96 years. The cohort largely consisted of patients with established disease, with a mean disease duration of 13.69 ± 9.93 years and a mean diagnostic delay of 5.19 ± 5.84 years. HLA-B27 was positive in 108/167 (65%) patients. Extra-musculoskeletal manifestations were frequently reported, including uveitis (37/251, 15%), psoriasis (26/250, 10%) and IBD (15/252, 6.0%), while peripheral arthritis was present in 64/251 (25%). Objective inflammatory markers were overall low (CRP 3.58 ± 5.39 mg/L; ESR 12.61 ± 12.17 mm/h), and the average disease activity was in the low-to-moderate range (ASDAS 1.32 ± 0.72). Most patients showed structural damage on conventional radiographs (166/181, 92%) and were receiving advanced therapies, including tumour necrosis factor inhibitors (TNFi; 179/252, 71%), IL-17 inhibitors (54/252, 21%) and JAK inhibitors (JAKi; 19/252, 7.5%). Overall characteristics of the MINA cohort are summarised in Table 1. Sixteen patients (16/252, 6.3%; 95% CI: 3.6%–9.5%) fulfilled the ASAS definition for the D2M phenotype, and 8 (8/252, 3.2%) fulfilled the definition for treatment-refractory (TR)-axSpA (Figure 1; Supplemental Table 1). Study summary. (a) Venn diagram showing the distribution of patients with D2M axSpA and TR axSpA within the total MINA cohort (N = 252). A total of 16 patients (6.3%) met the ASAS definition for D2M axSpA, of whom 8 (3.2% of the total) also fulfilled the definition for TR-axSpA, representing a subset with objective signs of persistent inflammation despite treatment. Percentages are shown in parentheses to facilitate interpretation. (b) Flowchart representing the proportion of patients fulfilling the definition for the D2M phenotype and the TR phenotype within the MINA cohort (N = 252). A total of 16 patients (6.3%) met the ASAS 2024 criteria for D2M, of whom 8 (3.2%) also fulfilled the definition for TR. No TR cases were observed outside the D2M group. ASAS, Assessment of Spondyloarthritis International Society; axSpA, axial spondyloarthritis; D2M, difficult-to-manage; TR, treatment-refractory. Compared with the 236 non-D2M patients, those with D2M axSpA had a similar age (48.44 ± 12.77 vs 50.82 ± 14.05 years) and a comparable sex distribution (12/16, 75% male vs 170/236, 72% male). The prevalence of radiographic axSpA (15/16, 94% vs 172/215, 80%), disease duration (14.47 ± 9.32 vs 13.63 ± 9.99 years), diagnostic delay (6.14 ± 5.16 vs 5.11 ± 5.89 years) and HLA-B27 positivity (8/13, 62% vs 100/154, 65%) were also comparable between groups. Similarly, no relevant differences were observed in BMI, age at symptom onset, smoking history or most extra-musculoskeletal manifestations, including IBD, peripheral arthritis, enthesitis, uveitis and psoriasis. By contrast, fibromyalgia was more frequent among D2M patients (3/16, 19% vs 10/236, 4.2%; p = 0.041). As expected, D2M patients reported worse physician and patient global assessments (PhGA: 1.11 ± 1.54 vs 0.27 ± 0.84; p = 0.012; PGA: 3.20 ± 2.37 vs 1.16 ± 1.76; p < 0.001), higher axial pain (VAS: 2.50 ± 2.75 vs 0.80 ± 1.63; p = 0.017) and more frequently reported low QoL symptoms (7/16, 44% vs 10/236, 4.2%; p < 0.001). Disease activity was higher in the D2M group (ASDAS: 2.50 ± 0.62 vs 1.24 ± 0.65; p < 0.001), and D2M patients were more likely to be receiving NSAIDs (6/16, 38% vs 26/236, 11%; p = 0.008) and JAKi (7/16, 44% vs 12/236, 5.1%; p < 0.001), while TNFi use was less frequent (6/16, 38% vs 173/236, 73%; p = 0.004). Overall characteristics of D2M-axSpA patients in the MINA cohort are shown in Table 2. Table 2.Baseline characteristics of D2M-axSpA patients compared with non-D2M patients in the MINA cohort.VariableNon-D2M(N = 236)D2M(N = 16) p Male sex, n (%)170/236 (72%)12/16 (75%)>0.9Age, mean ± SD (years)50.82 ± 14.0548.44 ± 12.770.6Radiographic axSpA, n (%)172/215 (80%)15/16 (94%)0.3Disease duration, mean ± SD (years)13.63 ± 9.9914.47 ± 9.320.6Diagnostic delay, mean ± SD (years)5.11 ± 5.896.14 ± 5.160.2HLA B-27 positive, n (%)100/154 (65%)8/13 (62%)0.8BMI, mean ± SD (kg/m2)25.51 ± 4.9925.96 ± 4.590.8Age at symptoms onset, mean ± SD (years)33.23 ± 13.5630.00 ± 9.110.6Smoking (ever), n (%)47/236 (20%)3/16 (19%)>0.9Psoriasis (ever), n (%)23/234 (9.8%)3/16 (19%)0.2IBD (ever), n (%)14/236 (5.9%)1/16 (6.3%)>0.9Peripheral arthritis (ever), n (%)59/235 (25%)5/16 (31%)0.6Uveitis (ever), n (%)32/235 (14%)5/16 (31%)0.068Enthesitis (ever), n (%)22/234 (9.4%)2/16 (13%)0.7Dactylitis (ever), n (%)9/235 (3.8%)0/16 (0%)>0.9Fibromyalgia, n (%)10/236 (4.2%)3/16 (19%) 0.041 Osteoporosis, n (%)15/236 (6.4%)1/16 (6.3%)>0.9ESR (mm/h), mean ± SD12.60 ± 12.3012.81 ± 10.450.7CRP (mg/L), mean ± SD3.47 ± 5.215.21 ± 7.650.4Low QoL symptoms, n (%)10/236 (4.2%)7/16 (44%) <0.001 PhGA, mean ± SD0.27 ± 0.841.11 ± 1.54 0.012 PGA, mean ± SD1.16 ± 1.763.20 ± 2.37 <0.001 VAS axial pain, mean ± SD0.80 ± 1.632.50 ± 2.75 0.017 VAS peripheral pain, mean ± SD0.45 ± 1.200.17 ± 0.580.5ASDAS, mean ± SD1.24 ± 0.652.50 ± 0.62 <0.001 Radiographic sacroiliitis, n (%)142 (94%)14 (100%)>0.9Spinal radiographic changes, n (%)57/236 (24%)5/16 (31%)0.6MRI sacroiliitis, n (%)29/236 (12%)4/16 (25%)0.2Structural damage on X-ray, n (%)152/167 (91%)14/14 (100%)0.6Time to first bDMARD (months), mean ± SD78.99 ± 87.7355.40 ± 65.410.5Current csDMARD use, n (%)8/236 (3.4%)1/16 (6.3%)0.5Current NSAIDs use, n (%)26/236 (11%)6/16 (38%) 0.008 TNFi use, n (%)173/236 (73%)6/16 (38%) 0.004 IL17i use, n (%)51/236 (22%)3/16 (19%)>0.9JAKi use, n (%)12/236 (5.1%)7/16 (44%) <0.001 Comparison of baseline characteristics of D2M-axSpA patients compared with non-D2M patients in the MINA cohort.Statistically significant variables are shown in bold.ASDAS, Axial Spondyloarthritis Disease Activity Score; axSpA, axial spondyloarthritis; bDMARD, biologic DMARD; BMI, body mass index; CRP, C-reactive protein; csDMARD, conventional synthetic disease-modifying antirheumatic drug; D2M, difficult-to-manage; ESR, erythrocyte sedimentation rate; HLA, human leukocyte antigen; IBD, inflammatory bowel disease; IL-17i, interleukin-17 inhibitor; JAKi, Janus kinase inhibitor; MRI, magnetic resonance imaging; NSAIDs, non-steroidal anti-inflammatory drugs; PGA, patient global assessment; PhGA, physician global assessment; QoL, quality of life; SD, standard deviation; TNFi, tumour necrosis factor inhibitor; TR-axSpA, treatment-resistant axSpA; VAS, visual analogue scale. Baseline characteristics of D2M-axSpA patients compared with non-D2M patients in the MINA cohort. Comparison of baseline characteristics of D2M-axSpA patients compared with non-D2M patients in the MINA cohort. Statistically significant variables are shown in bold. ASDAS, Axial Spondyloarthritis Disease Activity Score; axSpA, axial spondyloarthritis; bDMARD, biologic DMARD; BMI, body mass index; CRP, C-reactive protein; csDMARD, conventional synthetic disease-modifying antirheumatic drug; D2M, difficult-to-manage; ESR, erythrocyte sedimentation rate; HLA, human leukocyte antigen; IBD, inflammatory bowel disease; IL-17i, interleukin-17 inhibitor; JAKi, Janus kinase inhibitor; MRI, magnetic resonance imaging; NSAIDs, non-steroidal anti-inflammatory drugs; PGA, patient global assessment; PhGA, physician global assessment; QoL, quality of life; SD, standard deviation; TNFi, tumour necrosis factor inhibitor; TR-axSpA, treatment-resistant axSpA; VAS, visual analogue scale. Univariable and multivariable analysis In univariable logistic regression analyses, D2M status was associated with NSAIDs use (OR 4.8, 95% CI: 1.6–14.4), comorbid fibromyalgia (OR 5.2, 95% CI: 1.08–19.7.3) and a history of uveitis (OR 2.9, 95% CI: 0.9–8.84). Moreover, each one-point increase in axial pain VAS was associated with a 40% increase in the odds of D2M (OR 1.4, 95% CI: 1.1–1.8; p = 0.004). Other clinical factors, including sex, age, HLA-B27 status, smoking, psoriasis, peripheral arthritis, ESR and diagnostic delay, were not significantly associated with D2M. A parsimonious multivariable logistic regression model adjusted for age, sex, BMI and HLA-B27 status identified that NSAIDs use (adjusted OR 3.81, 95% CI: 1.22–11.89; p = 0.021), fibromyalgia (adjusted OR 5.52, 95% CI: 1.19–25.62; p = 0.029) and uveitis (adjusted OR 3.34, 95% CI: 1.00–11.15; p = 0.050) were associated factors of the D2M phenotype (Figure 2). Permutation feature importance analysis identified these three variables as the most influential contributors to model performance; inclusion of higher-order interaction terms did not improve model fit. Cross-cohort comparison A cross-cohort comparison between D2M patients from the MINA and the Argentinian Reuma-Check cohorts revealed important phenotypic differences (Supplemental Table 2). D2M patients from the Reuma-Check cohort were predominantly female and had a higher prevalence of smoking (80% vs 19%, p < 0.001) and active sacroiliitis on MRI (90% vs 25%, p < 0.001). They also reported greater pain and disease activity (mean VAS 6.3 ± 2.5; ASDAS 4.0 ± 1.0). In contrast, all Italian D2M patients exhibited structural damage on radiographs, compared with only 54% of Argentinian patients. Age at diagnosis, ESR, CRP levels and other clinical features were largely comparable. These findings suggest cross-population variability, with the Argentinian cohort representing a younger, inflammation-driven phenotype and the Italian cohort reflecting greater cumulative structural damage. Meta-analysis of predictors Using harmonised variables available in both cohorts, the pooled analysis identified multiple factors significantly associated with D2M (Table 3; Figure 3). Meta-analysis of clinical predictors of the D2M axSpA phenotype. Forest plot of RR with 95% confidence intervals for clinical and laboratory predictors of D2M axSpA. Data are shown for both the MINA Cohort and the Reuma-Check Cohort. Pooled estimates were obtained using a fixed-effect model. Each square represents the RR for the individual study cohort, with the size proportional to the study weight; horizontal lines indicate the 95% CI. Diamonds represent the pooled effect estimate for each variable. An RR >1 indicates a higher risk of D2M associated with the presence of the corresponding variable. axSpA, axial spondyloarthritis; D2M, difficult-to-manage; RR, risk ratios. Psoriasis emerged as a consistent predictor of D2M (pooled RR 2.23, 95% CI: 1.22–4.07; p = 0.009), despite its relatively low prevalence. Smoking was also associated with increased risk (pooled RR 1.79, 95% CI: 1.08–2.95; p = 0.023), as was radiographic structural damage (pooled RR 2.33, 95% CI: 1.24–4.38; p = 0.009). Importantly, uveitis was identified as an additional significant predictor (pooled RR 2.45, 95% CI: 1.02–5.85; p = 0.044). Elevated CRP levels (>5 mg/L) were strongly associated with D2M (RR 1.64, 95% CI: 1.15–2.33; p = 0.006), although high heterogeneity (I2 ≈ 94.1%) limited interpretability, and this variable was considered a definitional rather than independent predictor. Other variables, including enthesitis, HLA-B27 positivity, male sex, peripheral arthritis and MRI-detected sacroiliitis, were not significantly associated with D2M status in the pooled analysis. Because the meta-analysis was restricted to variables available in both the Italian and Argentinian cohorts, several clinically relevant factors from our regression model, most notably fibromyalgia and current NSAID use, could not be evaluated. Discussion In this cross-sectional study, we applied the 2024 ASAS definition of D2M to a large real-world Italian cohort of patients with axSpA recruited from two tertiary Rheumatology centres. Using this definition, we identified a D2M phenotype in 6.3% of cases. Among these, eight patients (3.2% of the total cohort) also fulfilled the definition for TR-axSpA, indicating that half of the D2M cases were associated with objective signs of ongoing inflammation, indicated by elevated CRP levels and active MRI lesions. Compared to previous studies, the prevalence of D2M in our cohort was lower. Philippoteaux et al. 15 reported a prevalence of 28.3% in a French cohort, and Fakih et al. 16 found a prevalence of approximately 20%. These discrepancies can be largely attributed to differences in case definitions. While those studies relied on the ‘D2T’ concept developed in the context of RA, focusing primarily on treatment failure, we applied the recently standardised ASAS definition of ‘D2M’ axSpA, which incorporates both treatment response and objective and subjective measures of disease activity and patient burden. 8 As a result, our estimate aligns more closely with studies that adopted similar criteria, such as those by Smits et al. 17 (9.7%), and Garcia-Salinas et al. 11 (8.5%). More recently, data from the German RABBIT SpA registry identified approximately 9% of patients as D2M in a large cohort 18 suggesting that, based on currently available evidence, the prevalence of D2M axSpA generally remains below 10%. Regarding TR-axSpA, we found a prevalence of 3.2%, which is identical to that reported by Garcia-Salinas et al. 11 in the Argentinian cohort and German-RABBIT SpA registry. 18 A lower prevalence was reported in the study by Smits et al. 17 (1.7%). When interpreting these findings, methodological differences across studies should be carefully considered. Variations in recruitment strategies and the timing of data collection may have influenced reported D2M prevalence. In our study, clinical variables refer to data collected up to enrolment; in Smits et al., the assessment timepoint for the reported clinical variables was not specified, which limits direct comparability. By contrast, Proft et al. collected data at the time of initiation of the first b/tsDMARD, reflecting an earlier and more treatment-anchored assessment. Differences in treatment access, timing and cohort composition should therefore be taken into account when comparing prevalence estimates across studies. Additional heterogeneity arises from differences in the operationalisation of the ASAS 2024 D2M criteria. Currently, validated instruments and universally accepted cut-off values are lacking for certain components of the definition, particularly criterion 2d (low QoL symptoms) and criterion 3 (disease perceived as problematic by the patient and/or physician). Consequently, heterogeneous proxy measures have been applied. For instance, Smits et al. defined impaired QoL using the SF-36 Physical Component Summary score (⩽40/100), whereas Proft et al. relied on a PGA threshold of ⩾4. Similar variability was observed for criterion 3, with different PGA and/or PGA thresholds applied across studies. In our study, criteria 2d and 3 were assessed through a binary patient-reported question at the most recent follow-up visit, rather than predefined scores or specific patient-reported outcome measures. In the absence of validated tools and recommended cut-off values, all currently used operational definitions, including those adopted in previous studies, should therefore be considered, at least in part, arbitrary, and this methodological heterogeneity should be considered when comparing D2M prevalence across cohorts. From a methodological perspective, these choices may have influenced our estimates and comparisons. Capturing criterion 2(d) and criterion 3 as binary items may increase measurement variability and could contribute to between-study differences in D2M prevalence when compared with cohorts using validated PRO-based thresholds. However, directly recording criterion 3 (‘problematic disease management’) as a dedicated item – rather than using disease activity proxies such as PtGA/PhGA – may reduce conceptual overlap with criterion 2 and represents a pragmatic strength of our real-world approach. Overall, while D2M prevalence varies according to definitions and cohort characteristics, the prevalence of TR axSpA appears more consistent, as it relies on objective measures of inflammation. This consistency supports the robustness of inflammation-based criteria in identifying truly refractory disease. The available data in literature suggest that persistent objective inflammatory activity despite treatment is relatively uncommon.11,18 Conversely, D2M may also encompass patients with high symptom burden and impaired QoL despite apparently adequate control of objective inflammation, in whom non-inflammatory mechanisms such as poor treatment adherence, comorbidities, chronic pain syndromes, physical inactivity and psychological distress should be considered. At the same time, the proportion of D2M patients with ongoing inflammation may be underestimated due to the limited sensitivity of current biomarkers and imaging techniques, and future advances, including AI-integrated imaging, may improve identification of subclinical inflammatory disease. 19 Our analysis, although exploratory, also sheds light on clinical features independently associated with the D2M phenotype. NSAIDs use, comorbid fibromyalgia and a history of uveitis emerged as significant associated factors in multivariable modelling. NSAIDs use may reflect attempts to manage persistent symptoms, but it could also signal therapeutic inertia or inadequate response to biologics. The observed association is more likely to reflect underlying symptom severity than to indicate a distinct component of the D2M phenotype. Therefore, despite a statistically robust association with D2M, this finding should be interpreted cautiously due to potential confounding by indication. Fibromyalgia likely amplifies symptom burden and impairs treatment responsiveness, consistent with prior research linking fibromyalgia to poorer outcomes and reduced treatment satisfaction in axSpA.15,21,22 Although uveitis emerged as an independent associated factor, the CI barely excluded unity, indicating a borderline statistical significance. Further validation in larger cohorts is needed. However, the low number of D2M cases, while consistent with prevalence estimates, limits statistical power for multivariable modelling and increases the risk of model overfitting. Results should therefore be interpreted with caution, but should be considered as preliminary and warrant future prospective investigations. The pooled meta-analysis across cohorts confirmed several associated factors of D2M, including psoriasis, smoking and radiographic structural damage. Elevated CRP levels were also strongly associated with D2M status but were excluded from regression modelling due to definitional overlap and high heterogeneity. These findings suggest that specific comorbidities and lifestyle factors may predispose individuals to a more challenging disease course across geographic and healthcare contexts. Notably, several features that were not independently significant in the Italian multivariable model, such as psoriasis and smoking, emerged as consistent predictors in the meta-analysis. Although the definition of D2M was consistent across cohorts, heterogeneity in pooled estimates may reflect not only differences in the assessment and recording of clinical covariates (e.g. fibromyalgia, uveitis, smoking) but also demographic and case-mix differences between the Italian and Argentinian populations, including variation in disease phenotype distribution and treatment pathways within distinct healthcare settings. These factors may influence both the prevalence of specific features and their apparent association with D2M, and should be considered when interpreting the meta-analytic results as exploratory. Limitations of the study This study has several limitations. First, its cross-sectional design precludes any inference on causality and does not allow assessment of disease trajectories or longitudinal treatment outcomes. Longitudinal studies are therefore needed to clarify whether the D2M phenotype is associated with poorer clinical outcomes over time, reduced treatment response or increased healthcare utilisation, key aspects that remain largely unexplored in the current literature. Second, the external comparison with the Reuma-Check cohort was based on published aggregate data rather than individual-level information, limiting the granularity of the analyses and warranting cautious interpretation of cross-cohort comparisons. Moreover, the lack of a D2M definition anchored to a specific disease timepoint limits direct comparability with our cross-sectional assessment. An additional limitation relates to the operationalisation of some components of the ASAS 2024 D2M definition. In our study, unlike in some previously published cohorts, validated patient-reported outcome measures or predefined instruments were not systematically used to define certain criteria, particularly impaired QoL and perceived problematic disease management. Instead, these aspects were assessed using a binary patient-reported question during the most recent follow-up visit. While this approach reflects routine clinical practice and mirrors real-world decision-making, it may reduce comparability with studies adopting validated PRO-based thresholds. Finally, data on other potentially relevant determinants of D2M axSpA, such as adherence to pharmacological therapy and engagement in complementary interventions (e.g. physical activity or physiotherapy), were not available. These factors are increasingly recognised as integral components of axSpA management, as also emphasised in recent EULAR recommendations. Nevertheless, this study benefits from a well-phenotyped, multicentre cohort, a standardised application of the ASAS D2M and TR definitions, and rigorous analytical strategies. To our knowledge, this is the first Italian multicentre study to apply the 2024 ASAS definition of D2M axSpA in a real-world setting. By integrating clinical and imaging data with meta-analytic evidence, our findings contribute to a more nuanced understanding of the clinical features underlying the D2M phenotype. Conclusion In this study, D2M-axSpA was identified in 6.3% of patients, and TR-axSpA in 3.2 Fibromyalgia, NSAID use and a history of uveitis were independently associated with the D2M phenotype, while meta-analytic data highlighted additional associations with psoriasis, smoking and radiographic damage. These findings underscore the multifactorial nature of D2M axSpA, driven by both inflammatory and non-inflammatory determinants. Despite its relatively low prevalence, D2M axSpA represents a clinically relevant subgroup that requires tailored management approaches. Future research should focus on integrating psychosocial factors, treatment adherence and non-pharmacological interventions, in line with current EULAR recommendations. 20 In TR axSpA, further progress will likely depend on the development of novel therapeutic strategies and targeted combination approaches. Supplemental Material sj-docx-1-tab-10.1177_1759720X261434275 – Supplemental material for Prevalence and clinical factors associated with difficult-to-manage axial spondyloarthritis: a multicentre Italian cohort study with meta-analysis Supplemental material, sj-docx-1-tab-10.1177_1759720X261434275 for Prevalence and clinical factors associated with difficult-to-manage axial spondyloarthritis: a multicentre Italian cohort study with meta-analysis by Giulio Forte, Gilberto Cincinelli, Daniele Mauro, Ilenia Pantano, Matteo Ferrito, Saviana Gandolfo, Andrea Amati, Antonio Ciancio, Miriam Longo, Maura Raimondi, Maria Manara, Enrico Tirri, Roberto Caporali and Francesco Ciccia in Therapeutic Advances in Musculoskeletal Disease sj-docx-2-tab-10.1177_1759720X261434275 – Supplemental material for Prevalence and clinical factors associated with difficult-to-manage axial spondyloarthritis: a multicentre Italian cohort study with meta-analysis Supplemental material, sj-docx-2-tab-10.1177_1759720X261434275 for Prevalence and clinical factors associated with difficult-to-manage axial spondyloarthritis: a multicentre Italian cohort study with meta-analysis by Giulio Forte, Gilberto Cincinelli, Daniele Mauro, Ilenia Pantano, Matteo Ferrito, Saviana Gandolfo, Andrea Amati, Antonio Ciancio, Miriam Longo, Maura Raimondi, Maria Manara, Enrico Tirri, Roberto Caporali and Francesco Ciccia in Therapeutic Advances in Musculoskeletal Disease Acknowledgments We thank all the patients who participated in this study and the clinical staff at the Rheumatology Units of the University of Campania ‘Luigi Vanvitelli’ and the University of Milan. The authors are responsible for the content of this manuscript. Footnotes ORCID iD: Francesco Ciccia https://orcid.org/0000-0002-9352-1264 Supplemental material: Supplemental material for this article is available online. Artificial intelligence disclosure: Generative artificial intelligence tools were used to assist with language editing and revision of the manuscript. The authors reviewed and edited the output and take full responsibility for the content, accuracy and integrity of the manuscript. Contributor Information Giulio Forte, Department of Precision Medicine, University of Campania ‘Luigi Vanvitelli’, Naples, Italy. Gilberto Cincinelli, University of Milan and ASST PiniCTO – Presidio Gaetano Pini, Milan, Italy. Daniele Mauro, Department of Precision Medicine, University of Campania ‘Luigi Vanvitelli’, Naples, Italy. Ilenia Pantano, Department of Precision Medicine, University of Campania ‘Luigi Vanvitelli’, Naples, Italy. Matteo Ferrito, University of Milan and ASST PiniCTO – Presidio Gaetano Pini, Milan, Italy. Saviana Gandolfo, Department of Precision Medicine, University of Campania ‘Luigi Vanvitelli’, Naples, Italy; Rheumatology Unit, Ospedale del Mare, ASL Napoli 1 Centro, Naples, Italy. Andrea Amati, University of Milan and ASST PiniCTO – Presidio Gaetano Pini, Milan, Italy. Antonio Ciancio, Department of Precision Medicine, University of Campania ‘Luigi Vanvitelli’, Naples, Italy. Miriam Longo, Department of Precision Medicine, University of Campania ‘Luigi Vanvitelli’, Naples, Italy. Maura Raimondi, Department of Precision Medicine, University of Campania ‘Luigi Vanvitelli’, Naples, Italy. Maria Manara, University of Milan and ASST PiniCTO – Presidio Gaetano Pini, Milan, Italy. Enrico Tirri, Rheumatology Unit, Ospedale del Mare, ASL Napoli 1 Centro, Naples, Italy. Roberto Caporali, University of Milan and ASST PiniCTO – Presidio Gaetano Pini, Milan, Italy. Francesco Ciccia, Department of Precision Medicine, University of Campania ‘Luigi Vanvitelli’, Via Pansini, 5, Naples 80131, Italy. Declarations Ethics approval and consent to participate: The study protocol was approved by the Ethics Committees of the participating centres (Milan Ethics Committee 138_1999, Naples Ethics Committee 2021.001336). All procedures complied with the Declaration of Helsinki. Oral informed consent was obtained from all participants prior to inclusion, in accordance with local regulations. Consent for publication: Not applicable. Author contributions: Giulio Forte: Data curation; Formal analysis; Methodology; Software; Writing – original draft. Gilberto Cincinelli: Data curation; Methodology; Writing – original draft. Daniele Mauro: Formal analysis; Methodology; Writing – review & editing; Writing – review & editing. Ilenia Pantano: Investigation; Writing – review & editing. Matteo Ferrito: Data curation; Investigation; Writing – review & editing. Saviana Gandolfo: Conceptualisation; Data curation; Visualisation. Andrea Amati: Data curation; Formal analysis. Antonio Ciancio: Data curation; Writing – review & editing. Miriam Longo: Data curation; Investigation. Maura Raimondi: Investigation. Maria Manara: Data curation; Investigation. Enrico Tirri: Project administration; Writing – review & editing. Roberto Caporali: Conceptualisation; Supervision; Validation; Writing – review & editing. Francesco Ciccia: Conceptualisation; Project administration; Supervision; Validation; Visualisation; Writing – review & editing. Funding: The authors received no financial support for the research, authorship and/or publication of this article. The authors declare that there is no conflict of interest. Availability of data and materials: Data can be made available upon request to the corresponding author. References Associated Data Supplementary Materials sj-docx-1-tab-10.1177_1759720X261434275 – Supplemental material for Prevalence and clinical factors associated with difficult-to-manage axial spondyloarthritis: a multicentre Italian cohort study with meta-analysis Supplemental material, sj-docx-1-tab-10.1177_1759720X261434275 for Prevalence and clinical factors associated with difficult-to-manage axial spondyloarthritis: a multicentre Italian cohort study with meta-analysis by Giulio Forte, Gilberto Cincinelli, Daniele Mauro, Ilenia Pantano, Matteo Ferrito, Saviana Gandolfo, Andrea Amati, Antonio Ciancio, Miriam Longo, Maura Raimondi, Maria Manara, Enrico Tirri, Roberto Caporali and Francesco Ciccia in Therapeutic Advances in Musculoskeletal Disease sj-docx-2-tab-10.1177_1759720X261434275 – Supplemental material for Prevalence and clinical factors associated with difficult-to-manage axial spondyloarthritis: a multicentre Italian cohort study with meta-analysis Supplemental material, sj-docx-2-tab-10.1177_1759720X261434275 for Prevalence and clinical factors associated with difficult-to-manage axial spondyloarthritis: a multicentre Italian cohort study with meta-analysis by Giulio Forte, Gilberto Cincinelli, Daniele Mauro, Ilenia Pantano, Matteo Ferrito, Saviana Gandolfo, Andrea Amati, Antonio Ciancio, Miriam Longo, Maura Raimondi, Maria Manara, Enrico Tirri, Roberto Caporali and Francesco Ciccia in Therapeutic Advances in Musculoskeletal Disease