Clusters based on demography, disease phenotype, and autoantibody status predicts mortality in lupus: data from Indian lupus cohort (INSPIRE)

Objectives SLE is associated with significant mortality, and data from South Asia is limited. Thus, we analysed the causes and predictors of mortality and hierarchical cluster-based survival in the Indian SLE Inception cohort for Research (INSPIRE). Methods Data for patients with SLE was extracted from the INSPIRE database. Univariate analyses of associations between mortality and a number of disease variables were conducted. Agglomerative unsupervised hierarchical cluster analysis was undertaken using 25 variables defining the SLE phenotype. Survival rates across clusters were assessed using non-adjusted and adjusted Cox proportional-hazards models. Results Among 2072 patients (with a median follow-up of 18 months), there were 170 deaths (49.2 deaths per 1000 patient-years) of which cause could be determined in 155 patients. 47.1% occurred in the first 6 months. Most of the mortality (n = 87) were due to SLE disease activity followed by coexisting disease activity and infection (n = 24), infections (n = 23), and 21 to other causes. Among the deaths in which infection played a role, 24 had pneumonia. Clustering identified four clusters, and the mean survival estimates were 39.26, 39.78, 37.69 and 35.86 months in clusters 1, 2, 3 and 4, respectively (P < 0.001). The adjusted hazard ratios (HRs) (95% CI) were significant for cluster 4 [2.19 (1.44, 3.31)], low socio-economic-status [1.69 (1.22, 2.35)], number of BILAG-A [1.5 (1.29, 1.73)] and BILAG-B [1.15 (1.01, 1.3)], and need for haemodialysis [4.63 (1.87,11.48)]. Conclusion SLE in India has high early mortality, and the majority of deaths occur outside the health-care setting. Clustering using the clinically relevant variables at baseline may help identify individuals at high risk of mortality in SLE, even after adjusting for high disease activity.