Predicting vancomycin trough serum concentration in augmented renal clearance patients through an artificial neural network model

Abstract

PurposeArtificial neural network (ANN) model has been developed in prediction of serum drug concentration. The aim of this study was verifying the accuracy of ANN model in predicting vancomycin trough serum concentration in ARC adult patients. MethodsA total of 162 ARC patients (258 observations) who received vancomycin treatment (500–2000 ​mg) at the First Affiliated Hospital of China Medical University between Jan 2017 and Nov 2020 were included. The performance and accuracy of the ANN model was evaluated by mean absolute deviation (MAD), mean absolute percent error (MAPE), mean square error (MSE), and root mean squared error (RMSE). In addition, the vancomycin calculator and multivariable linear regressions were compared with the ANN model in predicting vancomycin concentration. ResultsFrom results of linear regression analyses, dosage, dosage interval, age, creatinine clearance, sex, weight, total bilirubin (T-Bil), haemoglobin (HGB) and total protein (TP) were used as input variables to develop an ANN model for predicting vancomycin concentration in ARC patients. The MAPE, MSE, RMSE, MAD and R2 were 14.81%, 2.47%, 0.44%, 1.40% and 0.88%, respectively.The ANN model might be an useful tool to predict vancomycin concentration and help modify the dosing regimen accurately and in a timely manner to improve the clinical efficacy in ARC patients. In the future, ANN model could be developed to predict serum concentration and modify dosing regimens for more types of therapeutic drugs and patients with more complex diseases.