Evaluating clinical heterogeneity in relapsing polychondritis through unsupervised cluster analysis.

Introduction

Relapsing polychondritis (RP) is a rare systemic autoimmune disease characterized by relapsing episodes of inflammation and degeneration of cartilaginous tissues [ 1 ]. Its incidence ranges from 0.71 to 3.5 per million person-years [ 2 – 4 ].Typical clinical features include inflammation and destruction of nasal bridge, auricular cartilage, proteoglycan rich structures of the eyes and respiratory tract.Due to its rarity and the non-specific clinical manifestations, patients often experience significant diagnostic delay following symptom onset.Currently, no specific biomarkers are routinely used in clinical practice. The etiology and pathogenesis of RP remains unclear.Various factors have been implicated, including genetic susceptibility, autoimmune responses targeting cartilaginous structures, and dysregulation of cytokine and chemokine profiles that alter immune cell activation [ 5 ]. Thus, treatment is currently based on symptom control typically using steroids and immuno- suppressive drugs.A standard treatment strategy is lacking and management mainly relies on the clinician’s experience.The disease follows a progressive course, potentially leading to permanent cartilage destruction. Overall, survival outcomes for RP have improved considerably ‌over the past 10 years due to the availability of interventional treatments.However, the mortality rate remains high, with a standardized mortality ratio of 2.16 compared to the general population [ 2 ].

RP patients exhibit marked interindividual heterogeneity in clinical manifestation, disease progression and treatment response [ 1 ]. Moreover, growing evidence highlights significant variations in clinical phenotypes and prognostic outcomes across different ethnicities and geographic regions [ 6 – 8 ].These observations underscore the critical need for personalized management strategies. Refining the phenotypic classification of RP could yield substantial clinical benefits, yet there remains a notable lack of standardized frameworks for subtype stratification based on core clinical variables.

Clustering analysis, a powerful and widely employed data-mining tool in biological research, plays an important role in identifying patterns in complex diseases.It is also used to investigate novel or unexpected relationships that are not initially obvious.Furthermore, it can be used to identify subgroups of patients with similar trajectories in heterogeneous medical conditions, as it offers valuable insights into the mechanisms underlying disease heterogeneity.Therefore, in this study, we employ clustering analysis to investigate the existence of clinically significant phenotypes within a relatively large cohort of patients with RP.This analytical framework can be used to inform treatment decisions and improve clinical outcomes for RP patients, and thus contribute to the development of more personalized and effective care strategies.

Methods

Study population and data collection

A total of 135 patients with relapsing polychondritis (RP) were enrolled in this retrospective cohort study. These patients were recruited from two tertiary referral centers—the First Affiliated Hospital of Shandong First Medical University and the Affiliated Provincial Hospital of Shandong First Medical University—between January 2005 and December 2024. All patients met either McAdams or Damiani’s diagnostic criteria for RP [ 8 ], and written informed consent was obtained from each patient prior to study initiation. The clinical information of patients were collected.

Demographic and clinical characteristics of the 135 RP patients were systematically recorded.Clinical signs and symptoms related to RP were confirmed through a combination of examinations and imaging studies. In addition to routine laboratory test results, we collected data on three emerging inflammatory markers: the neutrophil-to-lymphocyte ratio (NLR), C-reactive protein-to-albumin ratio (CAR), and platelet-to-lymphocyte ratio (PLR) [ 9 , 10 ]. At the time of initial treatment, all patients were evaluated by a specialized research team with expertise in RP management. Each patient underwent a standardized comprehensive assessment, which included laboratory investigations, laryngeal imaging, and chest imaging. Disease severity at presentation was quantified using the Relapsing Polychondritis Disease Activity Index (RPDAI ) [ 11 ].

Cluster analysis

To maximize the sample size and ensure the robustness of the analysis, missing data were imputed using the mean imputation method. Variables with zero variance were excluded. To eliminate the influence of dimensional differences across various indicators (e.g., age and C-reactive protein [CRP]), all continuous variables were subjected Z-score normalization.

The K-means clustering algorithm was adopted as the core classification tool. Patients with similar clinical characteristics were grouped into the same cluster by calculating the Euclidean distance between samples. The optimal number of clusters, K = 3, was determined through a comprehensive evaluation combining the Elbow Method and Silhouette Analysis. To validate the clustering performance, Principal Component Analysis (PCA) was utilized to reduce the dimensionality of the multi-dimensional data and visualize the results.

Result

Characteristics of cohort

This study included 71 male and 64 female patients with an average age of 51.21 ± 14.65 years at disease onset(IQR 43–63). The average time from initial symptom onset to diagnosis was 9.68 + 16.25 months (IQR 2–10 ). About thirteen clinical variables were collected at enrolment, including demographic characteristics (age, sex), initial organ involvement (ear, nose, respiratory tract, joints, etc.), and laboratory markers (ESR, CRP, NLR, CAR, etc.).Detailed clinical characteristics are shown in Tables 1 and 2 .

Cluster analysis

By combining ANOVA variance analysis and random forest feature importance ranking, several clinical features were identified as the most significant determinants of cluster membership.A 3 cluster model was found to best fit the overall data set, with no significant gender differences observed among the three clusters.

Cluster 1 describes a cohort with respiratory involvement and hyperinflammation phenotype ( n = 45,33.33%). These patients exhibited a high incidence rate of laryngeal and primary bronchial involvement, affecting 59.26% and 53.85% of the cluster population respectively.As shown in Table 1 ,Cluster 1 had the highest percentage of patients who underwent tracheotomy (11.11%). Fever was more common in this cluster, with a significantly higher incidence rate compared to the other two clusters. Additionally, three parameters—ESR, CRP, and PLR showed statistically significant differences across clusters.The core characteristic of these patients is an active systemic inflammatory response, as indicated by the highest CAR (2.97 ± 1.97),a composite index reflecting both inflammation and nutritional indicators. Cluster 1 patients are in critical condition, characterized by a high risk of airway collapse and severe systemic inflammatory storm.

Cluster 2 is defined as a severe phenotype with sensory and multi-system involvement ( n = 11,8.15%). With an average age of 60.18 years, this cluster included the oldest patients among the three groups. Nerve damage was present in 27.27% of patients, and all patients in this cluster had hearing impairment.Renal impairment was observed in 18.18% of cases, while ocular and nasal involvement rates were the highest across all clusters. Anemia of chronic disease was detected in 45.45% of Cluster 2 patients. Although this cluster had the smallest sample size (11 patients, 8.15% of the total cohort), it was characterized by the highest Disease Activity Score (RPDAI) and a high mortality risk.

Cluster 3 included the largest proportion of patients, with 79 cases accounting for nearly 58.52% of the cohort. We describe this cluster as auricular involvement and mild phenotype.Symptoms in these patients are localized, primarily presenting with classic auricular onset (54.43%), which facilitates early diagnosis.The involvement of internal organs (heart, liver, kidneys, nerves) and the respiratory tract was relatively rare. Overall, these patients had a milder clinical course, with the lowest RPDAI score (18.49) and ESR level (29.25 mm/h). Cluster 3 represents the most typical clinical manifestation of RP, with mild systemic damage and a relatively favorable prognosis.

Clinical variables of the three clusters

Key features are crucial for clustering.Clinical variables were ranked according to their importance.(Fig. 1 ) Heat maps of cluster feature variables and radar charts illustrate the distinct performance of each variable across the identified clusters (Figs. 2 and 3 ).Notably, inflammatory indicators such as CRP and CAR, appear to play a more important role in clustering than clinical symptoms. The Relapsing Polychondritis Disease Activity Index (RPDAI) is a valuable tool for assessing disease activity, as it comprehensively captures all aspects of disease activity through a combination of typical clinical symptoms and laboratory parameters.Previous studies have reported that novel inflammatory markers, including C-reactive protein to albumin ratio (CAR)and the platelet-to-lymphocyte ratio (PLR), are associated with RPDAI scores [ 9 , 10 ]. These two emerging inflammatory indexes, along with other relevant features, constitute key factors influencing clustering performance. A significant positive correlation was found between PLR, NLR, CAR and RPDAI( R = 0.301,0.227and 0.278). In contrast, a significant negative correlation was found between airway involvement and auricular chondritis.( R = -0.487, p < 0.001).

Continuous variables were compared among three clusters using the Kruskal -Wallis test.The results revealed statistically significant differences in age, ESR, CRP, CAR, NLR, and RPDAI among the three clusters (Fig. 4 A) Additionally, one-way analysis of variance (ANOVA) showed statistically significant differences in albumin levels and PLR across the three clusters. (Fig. 4 B)

Principal component analysis of the clusters

The core idea of Principal Component Analysis(PCA) is to transform multiple original clinical variables into a small set of new composite indicators via linear combinations, termed “Principal Components”. Firstly, 29 clinical variables are selected.Then, 2 zero-variance variables are removed.Finally, 27 variables are used for PCA analysis.The first principal component (PC1) captures the direction of of maximum variance in the original data, while the second principal component (PC2) accounts for the next largest source of variation, under the constraint of orthogonality to PC1 (i.e., statistical independence from PC1).We employed PCA to identify which clinical variables drove the observed clustering patterns.(Fig. 5 ).

PC1 accounts for 17.3% of the total variance, and the variables with high absolute loading are variables related to systemic inflammation and disease activity. PC2 accounts for 10.8% of the total variance. Within the PCA framework, shifts along the PC1 axis correspond to changes in the overall inflammatory intensity of patients. In contrast, PC2 primarily captures differences in organ involvement patterns. Variable loading for PC1 and PC2 were shown in Fig. 6 .

Treatments of three Cluster

Systemic glucocorticoids are the first line treatment to RP.At the initiation of treatment, most patients received adequate(1 mg/kg/day prednisone equivalent)or moderate( 0.5 mg/kg/day prednisone equivalent) doses of glucocorticoids.(47.41% and 45.93% respectively)Only 9 patients received an excessively high glucocorticoids (> 1 mg/kg ≦ 200 mg/day prednisone equivalent). Small numbers of patients(9) did not administer immunosuppressants.The group received the following immuno- suppressants in sequence: methotrexate(38.52%), cyclophosphamide (20.74%), mycophenolate mofetil(20.74%), tripterygium glycosides (8.89%),leflunomide (7.41%), azathioprine(3.70%) and Janus kinase inhibitor(3.70%).‌ Biologic therapeutic agents such as etanercept(tumor necrosis factor α inhibitors) and tocilizumab (IL-6 inhibitors) had been introduced in patients who failed to the routine treatment. High-dose glucocorticoids and intensive immunosuppressive therapy were predominantly administered Cluster 1 and Cluster 2.Details are shown in Table 2 .

Discussion

Cluster analysis is an essential tool for gaining insight into disease heterogeneity and has been used to identify distinct clinical phenotypes with similar traits. This method provides a feasible approach to stratify conditions with high clinical variability, such as asthma, diabetes mellitus, and interstitial lung disease [ 12 – 14 ].The identification of disease phenotypes based on clinical presentation helps guide precise management strategies.To the best of our knowledge, only two main clinical patterns have been described in RP so far.In 2016, French researchers [ 15 ]identified 3 clinical phenotypes: the “hematologic”, “respiratory” and “mild” phenotypes based on the data from a cohort of 142 patients.The largest group, comprising 65% of patients, corresponded to the mild form, which was associated with an absence of severe complications and a low recurrence risk.The respiratory phenotype, accounted for 26% of cases, was characterized by younger patients with predominant airway(tracheobronchial or laryngeal)involvement.This phenotype associated with restricted respiratory function and a high infection rate.The hematologic phenotype, observed in 9% of patients, was characterized by older age and a diagnosis of myelodysplastic syndrome.It was the most severe phenotype, with a mortality rate of 58%.However, in our patient cohort, only one patient had a hematological comorbidity (multiple myeloma).Indeed, previous series [ 15 – 18 ] have demonstrated that hematological diseases are uncommon in RP.The hematologic subgroup in Dion’s [ 15 ] research exhibited characteristics resembling those of patients with the newly identified disease, VEXAS syndrome.The discovery of VEXAS syndrome in 2020 could explain the association between relapsing polychondritis and myelodysplastic syndromes in some patients [ 19 ]. VEXAS syndrome should be regarded as a disease mimicker, rather than as a distinct subtype of relapsing polychondritis. Detection of UBA1 mutations should be strongly considered for elderly (> 50 years of age)patients who had haematological abnormalities [ 20 ].Nevertheless,12.5% of patients in the present study had anemia of chronic inflammation.It was more common in Cluster 1 patients. Ferrada et al. identified three similar phenotypes [ 21 ]. A minority of patients, had a classic pattern of disease characterized by ear chondritis and extensive cartilage damage to nose and upper airway cartilages. A Majority of patients were defined by lower-airway predominant disease. A few patients presented with limited cartilaginous involvement and hematologic abnormalities .

Interestingly, researchers observed a negative correlation between airway involvement and auricular chondritis [ 16 , 21 ]. Thereby, Shimizu [ 22 ] and Liu [ 16 ] developed: auricular pattern(A), respiratory pattern (R group) and Overlap pattern(O group) based on their different prominently tissue involvement and disease progression.The majority of patients belonged to A pattern and R pattern at disease onset, while O pattern was less common.A pattern was associated with the best prognosis and a relatively lower inflammatory state. Notably, their studies emphasized that clinical patterns may transform during follow-up.In our cohort, only 4.76% of patients exhibited both features at disease onset, corresponding to the Overlap pattern. We also observed a negative correlation between airway and auricula involvement, which were the two major initial manifestations of RP.

It is well known that RP is a systemic autoimmune disease, that can affect multiple organs.Relying solely on classifications based on airway and auricula involvement may not provide a comprehensive view of disease heterogeneity.Through unsupervised machine learning clustering analysis, we identified three distinct clinical phenotypes: respiratory-high inflammation type, sensory organ-involved multisystem severe type, and auricular-dominant mild type. It Provides a unique perspective on the classification of disease. Principal component analysis revealed that these clusters exhibited statistical differences in multidimensional clinical characteristics, validating them as non-arbitrary categorizations.Random forest analysis further revealed that systemic inflammation levels (CRP, CAR, ESR) and sensory organ involvement (hearing/inner ear dysfunction) are critical discriminators for these subtypes. Patients assigned to cluster 1 characterized by its extremely high airway involvement and heavy inflammatory burden.Despite having the highest CRP, PLR and CAR, the RPDAI was not very high due to fewer multiorgan involvement in this cluster.Given its high risk of airway obstruction, clinicians should recognize it as an early warning signal of “critical and emergency conditions”. We recommend early dynamic airway assessment (e.g. bronchoscopy) and consideration of more aggressive anti- inflammatory strategies (such as early biologic agents combination therapy) in treatment to prevent life-threatening airway collapse. Patients in Cluster 2 (multiple system/high-score type), though comprising a small patient subset, presented with the most complex clinical manifestations. Cluster 2 had the highest RPDAI because of multiple organs involvement. Atypical initial symptoms and widespread multiple organs involvement often lead to misdiagnosis and diagnostic delay. Cluster 2 exhibited the longest misdiagnosis period due to its non-specific clinical presentation.It is therefore crucial for clinicians, especially non-rheumatologist, to maintain high vigilance for this rare disease. We strongly recommend implementing a Multidisciplinary Team approach to address this challenge, with management priorities focused on preventing irreversible sensory organ damage and controlling severe disease activity.Cluster 3 (mild type) is the most common subtype, characterized by mild disease condition. These patients had the lowest RPDAI and presented excellent long-term survival.The treatment goal should primarily focus on reducing recurrences. Despite having mild disease, regular follow-ups are still necessary to monitor disease progression.

Patients in Cluster 1 and 2 received more intensive treatment, including high dose of glucocorticoid and strong immunosuppressants, which may increase the risk of infections.None of our patients died from airway collapse or obstruction due to RP.Infectious complications is the leading cause of mortality.Inappropriate corticosteroids and strong immunosuppressants are risk factors for severe infections.Administration of biologics, such as interleukin-6 inhibitors, could help reduce the dose of glucocorticoids and immunosuppressants.For elderly patients in Cluster 3, tripterygium glycosides offer as safer and milder alternative to conventional immunosuppressants. Overall, the classification system established in this study provides an evidence-based foundation for precision and stratified management of RP.

Predictive biomarkers are pivotal for individualized disease monitoring. Unfortunately, The rarity of relapsing polychondritis (RP) has hindered the development of the application of biomarkers in relapsing polychondritis (RP).In recent years, the value of composite indicators such as the Systemic Immune Inflammation Index (SII), Neutrophil-to-Lymphocyte Ratio ‌(NLR), platelet- to- lymphocyte ratio ‌(PLR), C-Reactive Protein to Albumin(ALB) Ratio (CAR) have gained increasing attention for their value in disease early warning and severity assessment [ 9 , 10 ].During the acute-phase response, CRP levels are elevated by inflammatory mechanisms, whereas serum albumin levels are reduced.Cluster 1 and Cluster 2 in a high inflammatory state had lower serum albumin levels.CAR is one of sensitive composite biomarkers that integrate both nutritional condition and inflammation of the body. Both ALB and CAR existed statistical differences among three clusters.By integrating multiple blood biomarkers parameters such as platelets, neutrophils, and lymphocytes, these composite indicators provide a comprehensive reflection of the body’s inflammatory and immune status. Our study confirms that three new inflammatory markers(PLR, NLR and CAR) are important indicators for disease classification and exhibited strong correlations with RPDAI.Notably, these parameters are simple to measure and widely available in routine clinical practice, making them highly practical for assisting clinicians in patient stratification and the formulation of tailored treatment strategies.

Limitations

Our study has several limitations. First, as a retrospective analysis of hospitalized patients, inherent selection bias may exist, as the cohort primarily includes individuals with relatively severe disease manifestations, potentially limiting the generalizability of our findings to milder or non-hospitalized cases. Second, recall bias cannot be excluded, as participants may have inaccurately reported past symptoms, disease onset timing, or exposure history, which could affect the reliability of data related to disease progression and risk factor associations.Third, the cross-sectional design of this study precludes the ability to establish temporal relationships between variables and assess long-term outcomes. Without longitudinal follow-up, we may have underestimated the frequency of organ involvement, particularly in patients with short disease courses, as certain complications or progressive organ damage may emerge over time rather than at the initial presentation. Future prospective, multicenter studies with larger sample sizes and extended follow-up periods are warranted to validate our findings and address these limitations.In addition, due to the small number of Cluster 2, the statistical efficiency is limited. The results of this study need to be validated in future multicenter, large sample cohorts.

Conclusions

In summary, our study found that several biomarkers, including CAR, PLR and CRP are significantly associated with activity index in patients with relapsing polychondritis. These biomarkers play a critical role in patient stratification. Through unsupervised clustering analysis, we categorized RP patients into three distinct clinical phenotypes, each characterized by unique organ involvement patterns and prognostic trajectories. This classification system provides clinicians with a practical tool to rapidly assess individual patients’ prognostic risks, tailor treatment strategies, and ultimately optimize long-term quality of life for patients with RP.

Acknowledgements

We sincerely thank all the designers and participants who contributed to the study.

Abbreviations

Albumin

Analysis of Variance

Azathioprine

C-reactive protein to albumin ratio

C-reactive protein

Cyclophosphamide

Erythrocyte sedimentation rate‌

Interquartile Range

Janus kinase inhibitor‌

Leflunomide

Mycophenolate Mofetil

Methotrexate

Neutrophil to lymphocyte ratio

Principal Component Analysis

Platelet to lymphocyte ratio

Relapsing polychondritis

Relapsing Polychondritis Disease Activity Index

Tripterygium glycosides

Tumor Necrosis Factor-alpha inhibitor‌

Interleukin-6 inhibitor

Authors’ contributions

Wenping Pan and Liguo Yin: co-drafter of manuscript and substantively revised manuscript, are equal contribution to the article; Lantao Wang: Methodology, Statistical analysis; Qing Wang, Tingting Zhai, Jin Cao, Xia Wang: Data collection; Hongsheng Sun: Conceptualization.

Funding

All authors declare that no funding was received for this study.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

The studies involving humans were approved by Studies on material of human origin were approved by the ethics committee of the Ethics Committee of First Affiliated Hospital of Shandong First Medical University & Shandong Provincial Qianfoshan Hospital and the ethics committee of the Ethics Committee of Shandong Provincial Hospital Affiliated to Shandong First Medical University. Informed consent was obtained from all patients/participants. The studies were conducted in accordance with the local legislation and institutional requirements.

Consent for publication

All authors have agreed to the publication of this manuscript.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Lahmer T, Treiber M, von Werder A, Foerger F, Knopf A, Heemann Uwe, et al. Relapsing polychondritis: an autoimmune disease with many faces.Autoimmun Rev. 2010;9(8):540–6.

2. Hazra N, Dregan A, Charlton J, Gulliford MC, D’Cruz DP. Incidence and mortality of relapsing polychondritis in the UK: a population-based cohort study. Rheumatology (Oxford). 2015;54(12):2181–7.

3. Kent PD, Michet CJ Jr. Luthra HS.Relapsing polychondritis. Curr Opin Rheumatol. 2004;16(1):56–61.

4. Horváth A, Páll N, Molnár K, Kováts T, Surján G, Vicsek T. Pollner P.A nationwide study of the epidemiology of relapsing polychondritis. Clin Epidemiol. 2016;23(8):211–30.

5. Vitale A, Sota J, Rigante D, Lopalco G, Molinaro F, Messina M, et al. Relapsing Polychondritis: an Update on Pathogenesis, Clinical Features, Diagnostic Tools, and Therapeutic Perspectives. Curr Rheumatol Rep. 2016;18(1):3.

6. Ferrada M, Rimland CA, Quinn K, Sikora K, Kim J, Allen C, et al. Defining clinical subgroups in relapsing polychondritis:a prospective observational cohort study. Arthritis Rheumatol. 2020;72(8):1396–402.

7. Lin DF, Yang WQ, Zhang PP, Lv Q, Jin O, Gu JR. Clinical andprognostic characteristics of 158 cases of relapsing polychondritis inChina and review of the literature. Rheumatol Int. 2016;36(7):1003–9.

8. McAdam LP, O’Hanlan MA, Bluestone R, Pearson CM. Relapsing polychondritis: prospective study of 23 patients and a review of the literature. Med (Baltim). 1976;55(3):193–215.

9. Cao XY, Zhao MZ, Li HJ, Xu D, Li MT, Zhang X, et al. Three new inflammatory markers C reactive protein to albumin ratio, neutrophil to lymphocyte ratio, and platelet to lymphocyte ratio correlated with relapsing polychondritis disease activity index. Clin Rheumatol. 2021;40(11):4685–91.

10. Liu YM, Cheng LL, Zhao MZ, Zhan HT, Li XM, Huang Y, et al. Development and validation of diagnostic and activity-assessing models for relapsing polychondritis based on laboratory parameters. Front Immunol. 2023;14:1274677.

11. Arnaud L, Devilliers H, Peng SL, Mathian A, Costedoat-Chalumeau N, Buckner J, et al. The Relapsing Polychondritis Disease Activity Index: development of a disease activity score for relapsing polychondritis. Autoimmun Rev. 2012;12(2):204–9.

12. Salvatori B, Wegener S, Kotzaeridi G, Herding A, Eppel F, Dressler-Steinbach I, et al. Identification and validation of gestational diabetes subgroups by data. -driven cluster analysis. Diabetologia. 2024;67(8):1552–66.

13. Park SY, Jung HW, Lee JM, Shin B, Kim HJ, Kim MH, et al. Novel Trajectories for Identifying Asthma Phenotypes: A Longitudinal Study in Korean Asthma Cohort, COREA. J Allergy Clin Immunol Pract. 2019;7(6):1850–7.

14. Adegunsoye A, Oldham JM, Chung JH, Montner SM, Lee C, Witt LJ, et al. Phenotypic Clusters Predict Outcomes in a Longitudinal Interstitial Lung Disease Cohort. Chest. 2018;153(2):349–60.

15. Jérémie D, Nathalie Costedoat-Chalumeau, Damien Sène, Judith Cohen-Bittan,Gaëlle Leroux, Charlotte Dion, et al. Relapsing polychondritis can be characterized by 3 different clinical phenotypes:Analysis of a recent series of 142 patients.Arthritis &. Rheumatol. 2016;68(12):2992–3001

16. Zhang L, Yun S, Wu TG, He YJ, Guo JY, Han LS, et al. Clinical patterns and the evolution of relapsing polychondritis based on organ involvement: a Chinese retrospective cohort study.Orphanet. J Rare Dis. 2021;16(1):225.

17. Zeuner M, Straub RH, Rauh G, Albert ED, Scholmerich J, Lang B. Relapsing polychondritis: clinical and immunogenetic analysis of 62 patients. J Rheumatol. 1997;24:96–101.

18. Oka H, Yamano Y, Shimizu J, Yudoh K, Suzuki N. A large-scale survey of patients with relapsing polychondritis in japan. Inflamm regen. 2014;34:149–56.

19. Beck DB, Ferrada MA, Sikora KA, Ombrello AK, Collins JC, Pei W, et al. Somatic mutations in UBA1 and severe adult-onset autoinflammatory disease.N. Engl J Med. 2020;383(27):2628–38.

20. Mertz P, Costedoat-Chalumeau N, Ferrada MA, Moulis G, Mekinian A, Grayson PC, et al. Relapsing polychondritis: clinical updates and new differential diagnoses. Nat Rev Rheumatol. 2024;20(6):347–60.

21. Ferrada M, Rimland CA, Quinn K, Sikora K, Kim J, Allen C, et al. Defining clinical subgroups in relapsing polychondritis: a prospective observational cohort study. Arthritis Rheumatol. 2020;72(8):1396–402.

22. Shimizu J, Yamano Y, Kawahata K, Suzuki N. Relapsing polychondritis patients were divided into three subgroups: patients with respiratory involvement (R subgroup), patients with auricular involvement (A subgroup), and overlapping patients with both involvements (O subgroup), and each group had distinctive clinical characteristics. Med (Baltim). 2018;97(42):e12837.

Associated Data

Data Availability Statement

No datasets were generated or analysed during the current study.