Prognostic insights from symptom clustering analysis in systemic lupus erythematosus

Systemic lupus erythematosus (SLE), a systemic autoimmune disease with heterogeneous severity, poses a challenge in classifying patients due to its dynamic phenotype. We therefore aimed to identify clinical clusters at diagnosis and at the last available visit to define subgroups with different disease trajectories.

This retrospective study included 278 SLE patients fulfilling the 2019 EULAR/American College of Rheumatology classification criteria. Data were extracted from the patient's medical record, encompassing demographic, clinical and immunological features. Two hierarchical clustering analyses were performed: one at diagnosis and the other at the last available visit based on 23 clinical manifestations. We assessed the distribution of autoantibodies among clusters, and survival analyses compared prognosis using the Cox regression model.

Three symptom clusters were identified at diagnosis and confirmed at the last available visit, with consistent clinical profiles. The largest, clusters 1/1', exhibited early articular (95.3%; p=0.0043) and mucocutaneous (58.7%; p=0.0031) symptoms with the best survival rate. Clusters 2/2' displayed the most severe phenotype, including renal involvement (62.2%; p<0.0001), positivity for anti-DNA (90.0%; p=0.0151) and anti-Sm (34.4%; p<0.0001) autoantibodies, and worse prognosis. Clusters 3/3' exhibited a high proportion of SLE patients fulfilling the definition of mixed connective tissue disease (MCTD) (50.0%; p<0.0001) with anti-U1-RNP 70 kDa autoantibodies (87.5%; p<0.0001). Clusters 1 and 2 remained stable during follow-up, cluster 2 even expanding over time, while over half of cluster 3 patients transitioned to a different subgroup.

We identified a distinct MCTD phenotype within SLE patients, a severe SLE phenotype with poor prognosis, and a third group of SLE patients with lower visceral involvement. Clusters remained stable over time, providing insights into disease progression.