An area under curve-based nomogram to predicts vancomycin-associated nephrotoxicity in critically ill patients: a retrospective cohort study

Abstract

Objective: To develop an area under curve (AUC)-based nomogram to predict vancomycin-associated nephrotoxicity in critically ill patients. Methods: This retrospective cohort study included adult patients treated with vancomycin in the intensive care unit at a tertiary teaching hospital from January 2015 to December 2017. Baseline clinical characteristics before vancomycin treatment and pharmacokinetic parameters were collected to establish a prediction model of nephrotoxicity. Univariate analysis was used to screen variables, and multivariate logistic regression analysis was used to establish the prediction model and nomogram. Results: A total of 159 patients met the inclusion criteria, sixty-four were included in the final analysis. Sixteen patients (25%, 16/64) developed vancomycin-associated nephrotoxicity. The following variables were incorporated into the prediction model: vancomycin AUC, estimated glomerular filtration rate (GFR), and combined nephrotoxic drugs. The following equation was established to calculate the probability of nephrotoxicity: logit (P)=-4.83+0.009×AUC-2.87×1 (if GFR>60 ml/min)+2.53×1 (if number of combined nephrotoxic drugs≥2). A nomogram was generated based on the equation. The receiver-operating characteristic curve demonstrated that the AUC of the prediction model was 0.927 (95%CI 0.851-1.000). The cut-off value of the probability of nephrotoxicity was 26.48%. The sensitivity and specificity were 87.5% and 87.5% respectively. Conclusion: The incidence of vancomycin-associated nephrotoxicity is high. The AUC-based nomogram can effectively predict vancomycin-associated nephrotoxicity in critically ill patients.