LASSO-derived nomogram predicting new-onset diabetes mellitus in patients with kidney disease receiving immunosuppressive drugs.

Abstract

Patients with kidney disease receiving immunosuppressive drugs (ISDs) (tacrolimus, cyclosporine and glucocorticoids) have a high risk of developing new-onset diabetes mellitus (NODM). We aimed to establish a precise and convenient model for predicting NODM in patients receiving immunosuppressive drugs. This retrospective study recruited 1883 patients receiving ISDs between January 2010 and October 2018. The occurrence of NODM was the primary endpoint. The patients were randomly divided into training (n=1318) and validation cohorts (n=565) at a 7:3 ratio. A nomogram was established based on a least absolute shrinkage and selection operator (LASSO)-derived logistic regression model. The nomogram's discrimination and calibration abilities were evaluated in both cohorts using the Hosmer-Lemeshow test and calibration curves. Decision curve analysis (DCA) was used to evaluate the net benefit of the predictive efficacy. Amongst the 1883 patients with kidney disease receiving immunosuppressive drugs, 375 (28.5%) and 169 (29.9%) developed NODM in the training (n=1318) and validation cohorts (n=565), respectively. Nine clinic predictors were included in this LASSO-derived nomogram, which is easy to be operated clinically. The discriminative ability, determined by the area under the receiver operating characteristic curve (AUC), was 0.816 (95% confidence interval [CI] 0.790-0.841) and 0.831 (95%CI 0.796-0.867) in the training and validation cohorts, respectively. Calibration was confirmed with the Hosmer-Lemeshow test in the training and validation cohorts (p=0.238, p=0.751, respectively). Nearly one-third of patients with kidney disease receiving immunosuppressive drugs developed NODM. The nomogram established in this study may aid in predicting the occurrence of NODM in patients with kidney disease receiving immunosuppressive drugs.