PO:16:226 | Clinical Phenotypes of Systemic Lupus Erythematosus: An Exploratory Cluster Analysis in a Monocentric Cohort

Richard Borrelli¹, Stefania Nicola¹, Federica Corradi¹, Iuliana Badiu¹, Luca Lo Sardo¹, Marzia Boem¹, Valentina Marmora¹, Simone Negrini¹, Luisa Brussino¹ | ¹SCDU Immunologia e Allergologia, AO Mauriziano, Dipartimento di Scienze Mediche, Università degli Studi di Torino, Italy

Background/Aim. Systemic Lupus Erythematosus (SLE) is a clinically heterogeneous autoimmune disease characterized by a wide spectrum of organ involvement and variable disease severity. The identification of distinct clinical phenotypes may aid in personalizing disease management, stratifying patients for prognosis, and optimizing therapeutic strategies. Cluster analysis is a powerful, data-driven technique that can group patients based on shared features without relying on pre-defined categories. The aim of this study was to identify clinical subgroups of SLE patients through hierarchical clustering based on organ involvement at diagnosis.

Materials and Methods. We retrospectively analyzed clinical data from a monocentric cohort of SLE patients followed at a tertiary care center, enrolled as per the 2019 EULAR/ACR classification criteria. Binary clinical variables were extracted, representing major organ/system involvement at diagnosis: cutaneous, articular, neuropsychiatric (NPSLE), renal, pulmonary, cardiac, and hematological (including AIHA, thrombocytopenia, and leukopenia). Only patients with complete data for all selected variables were included in the analysis. Hierarchical clustering was performed using Ward's method on standardized data. The optimal number of clusters was determined via visual inspection of a type-oriented dendrogram. Principal Component Analysis (PCA) was applied for two-dimensional visualization of the clusters.

Results. A total of 119 SLE patients were included in the analysis. Three distinct clinical clusters were identified: - Cluster 1, Cutaneous-Articular (n = 62): included patients with high prevalence of articular (98%) and cutaneous (61%) involvement, with minimal major organ involvement. - Cluster 2, Renal- Hematologic (n = 22): was characterized by predominant hematologic manifestations (100% with autoimmune hemolytic anemia), frequently associated with renal (47%) and pulmonary involvement (10%). - Cluster 3, Neuro-Hematologic (n = 35): comprised patients with a combination of neuropsychiatric manifestations (21%), hematologic abnormalities (81%), and renal involvement (15%), suggesting a more complex and potentially severe disease profile. Specific results can be seen in the PCA biplot (figure 1). While Clusters 2 and 3 showed partial overlap in PCA visualization, they maintained distinct clinical features, particularly regarding the presence of NPSLE.

Conclusions. This exploratory cluster analysis identified three clinically distinct phenotypes of SLE based on organ involvement. While the cutaneous-articular cluster appeared less severe, the other two clusters showed broader systemic manifestations. These findings suggest that unsupervised learning techniques may aid in clinical classification of SLE and could inform future research on prognosis and therapy. Although limited by sample size and lack of formal validation, this study provides a rationale for further investigation in larger, multi-center cohorts.