Evaluation on the performance of a bayesian adaptive control system to forecast the chloramphenicol serum concentrations in pediatric patients with sepsis and malnutrition

Objective. To validate the population pharmacokinetic parameters of chloram phenicol in pediatric patients with sepsis and malnutrition (PPSM) using a bayesian forecasting program. Design. Retrospective evaluation of predictiv e performance of a bayesian program in PPSM. Setting. Tertiary care center. Patients. Fifteen MPSP and ten NMPSP that receiving treatment with chloram phenicol. Methods and main results. In the first part of the study, the med ical records of 10 MPSP and 10 NMPSP who had received treatment with chlora mphenicol were reviewed. The population pharmacokinetic parameter values fo r each group were estimated using a nonparametric expectation maximization algorithm (NPEM). in the second part, data gathered from five other MPSP re ceiving chloramphenicol were entered into a bayesian program. Chloramphenic ol pharmacokinetic values For each of these five patients were estimated, f irst using the values of NMPSP as a priori distribution and then repeating the analysis using the MPSP values. The bayesian serum chloramphenicol conc entrations predicted for each population model were compared with the actua l peaks and troughs. The specific model for MPSP permitted forecasting the peak and trough serum chloramphenicol concentrations with less bias and a b etter precision compared with the NMPSP population model. Conclusions. Thes e data indicate that chloramphenicol pharmacokinetics in PPSM can be predic ted with minimal bias and good precision using a bayesian forecasting progr am, allowing a better control of the chloramphenicol serum concentrations. In addition, the limited number of samples required by the bayesian method may represent an important economical benefit for the patient.