Abstract: Objective: Patients with SLE have a well-known increased risk of major comorbidities, although they are also very heterogeneous in terms of the prevalence of comorbid conditions. The relationships of such comorbidities with the outcomes and the severity of index diseases are less known. We aimed to evaluate the interactions between comorbid conditions, in a large multicentre SLE cohort, and their impact on severity and outcomes, using a cluster analysis.
Methods: Data on 14 cumulative comorbidities were derived from patients with SLE (American College of Rheumatology (ACR)-97 criteria) who had been included in the retrospective phase of the RELESSER (Spanish Society of Rheumatology National Register of SLE). The Severity Katz Index and the SLICC/ACR Damage Index were calculated. Unsupervised cluster analysis was performed to better characterise the relationships between comorbidities in a large multicentre cohort of patients with SLE. For intercluster differences testing, analysis of variance and Tukey tests were used to compare continuous numerical variables; a Kruskal-Wallis test to discrete variables and the ?² (or Fisher's exact test) were used for categorical ones.
Results: A total of 3658 patients with SLE were included. Men accounted for 9.6% of patients. The mean (SD) age was 45.9 years, and 93% were Caucasian. Four clusters, with markedly different comorbidity profiles and outcomes, were identified: in cluster 2 (n=516), patients were grouped around depression (100% of the cases); in cluster 3 (n=418) around serious infections (100%); and in cluster 4 (n=388) around cardiovascular events (also 100%). However, in cluster 1, the largest one (n=2336), no patient had any of the three defining comorbidities of the other clusters, and this cluster was associated with the best outcomes.
Conclusions: Cluster analysis identifies well-differentiated subsets of patients with SLE in terms of their comorbidities. The most relevant comorbidities in SLE tend to aggregate in the most severe patient subsets.
