Bayesian Network to Optimize the First Dose of Antibiotics: Application to Amikacin

Abstract

Objective: To construct and validate a network to predict the first dose of amikacin. Methods: Anthropometric and therapeutic data were recorded for 120 patients. Bayesian network (BN) was built to predict the dose to achieve a fixed target peak concentration of 64 mg/l. In 40 subjects, doses predicted with the BN (BND) and based on body weight (BWD) were compared with adjusted doses calculated using a pharmacokinetic software (MM-USCPACK; BID). Results: The calculated dose differed by <20% from the ideal dose in 62.5% of the patients with the BN and in 43.8% of the patients with the BW. Conclusion: BN is a promising approach to optimize the prediction of the first dose.