Résumé
Critically ill patients with sepsis are predisposed to physiological changes that can reduce the probability of achieving target antibiotic exposures. Precision dosing software programs may be used to improve probability of obtaining these target exposures.
To quantify the accuracy of a precision dosing software program for predicting antibiotic concentrations as well as to assess the impact of using software predictions on actual dosing adjustments.
The software program ID-ODS was used to predict concentrations for piperacillin, meropenem and vancomycin using patient covariate data with and without the use of therapeutic drug monitoring (TDM) data. The impact of these predictions on actual dosage adjustments was determined by using software predicted concentrations versus measured concentrations.
Software predictions for piperacillin and meropenem exhibited large bias that improved with the addition of TDM data (bias improved from -28.8 to -2.0 mg/L for piperacillin and -3.0 to -0.1 mg/L for meropenem). Dosing changes using predicted concentrations of piperacillin and meropenem with TDM data versus measured concentrations were matched on 89.2% (107/120) and 71% (9/69) occasions, respectively. Although vancomycin predictions demonstrated good accuracy with and without TDM, these findings were limited by our small sample size.
These data demonstrate that precision dosing software programs may have scope to reasonably predict antibiotic concentrations in critically ill patients with sepsis. The addition of TDM data improves the predictive performance of the software for all three antibiotics and the ability to anticipate the correct dose change required.