Predictors of refractory risk in systemic lupus erythematosus-related thrombocytopenia: a dual-centre retrospective study

Objectives Based on clinical and laboratory indicators, this study aimed to establish a multiparametric nomogram to assess the risk of refractory cases of SLE-related thrombocytopenia (SLE-related TP) before systematic treatment. Methods From June 2012 to July 2021, a dual-centre retrospective cohort study of prospectively collected data of patients with SLE-related TP was conducted. The cohort data were divided into a developing set, internal validation set and external validation set. Refractory thrombocytopenia (RTP) was defined as failed to prednisone at 1 mg/kg per day with a platelet count cannot achieve or maintain higher than 50x10(9)/L. In the developing set, a nomogram were established to predict RTP risk based on clinical characteristics and laboratory indicators by multivariable logistic regression, and its performance was assessed by receiver operating characteristic (ROC) curves, calibration curves, decision curve analysis (DCA) and clinical impact curve (CIC). Results A total of 1778 patients with SLE were included, and 413 eligible patients were involved in the final analysis with 121 RTPs. The RTP risk assessment (RRA) model was composed of five significant risk variables: pregnancy, severity of TP, complement 3, anticardiolipin antibody-immunoglobulin G and autoimmune haemolytic anaemia. In three datasets, the AUCs were 0.887 (95% CI 0.830 to 0.945), 0.880 (95% CI 0.785 to 0.975) and 0.871 (95% CI 0.793 to 0.949), respectively. The calibration curve, DCA and CIC all showed good performance of the RRA model. Conclusion The RRA model demonstrated good capability for assessing the refractory risk in SLE-related TP, which may be helpful for early identification and intervention.