Prognostic value of baseline clinical features in SLE: a cluster analysis
Betül Dikkanoğlu Demirok, Sibel Balci, Andaç Komaç, Duygu Temiz Karadağ, Ayşe Çefle, Ayten YaziciObjective
SLE is a heterogeneous systemic autoimmune disease with diverse clinical manifestations. We aimed to identify clinical subgroups of SLE patients using cluster analysis at diagnosis and at the last follow-up and to evaluate their prognostic implications for disease activity, organ damage and mortality.
Methods
In this single-centre retrospective cohort study, 199 patients with SLE who fulfilled the 1997 American College of Rheumatology (ACR) classification criteria and were regularly followed at the Rheumatology Clinic of Kocaeli University Hospital were included. Patients were divided into clusters based on their clinical involvement at diagnosis and the last visit using the Gower distance and the Partitioning Around Medoids algorithm. Differences between clusters were assessed using one-way analysis of variance, Kruskal-Wallis and χ 2 analyses. Analyses were performed using SPSS V.29.0 (IBM Corp, Armonk, New York, USA) and R V.4.3.0.
Results
At diagnosis, three clusters were identified: Cluster 1 (n=108) with predominant arthritis; Cluster 2 (n=49) with renal and haematological involvement and highest disease activity (SLE Disease Activity Index 2000, p=0.004); Cluster 3 (n=42) with mucocutaneous manifestations, often accompanied by haematological involvement. After a median follow-up of 168 months, four clusters were identified. Cluster 4′, a heterogeneous group with mucocutaneous, articular and renal involvement, had significantly higher Systemic Lupus International Collaborating Clinics (SLICC)/ACR Damage Index (SDI) scores (median 4.0 (0–8), p<0.001), indicating poor prognosis. During follow-up, 29.6% of Cluster 1, 22.4% of Cluster 2 and 42.9% of Cluster 3 transitioned to Cluster 4′. Patients with mucocutaneous-predominant disease at diagnosis had a twofold increased risk of transitioning to the poor-prognosis cluster (OR 2.05, 95% CI 1.01 to 4.16, p=0.046). Mortality did not differ significantly between clusters.
Conclusion
Cluster analysis based on clinical involvement in SLE can identify homogeneous patient subgroups and predict long-term outcomes. Patients with mucocutaneous-predominant onset, often receiving less intensive treatment, may have a worse prognosis, highlighting the importance of individualised monitoring and management strategies.