In this episode, I’ll discuss bayesian vancomycin monitoring in the critically ill.
In order to use AUC to monitor vancomycin, the 2020 Vancomycin guidelines recommend using either 2 vancomycin levels with first-order kinetic calculations or Bayesian software programs to estimate AUC.
The guidelines also state:
As part of their output, Bayesian dosing programs provide innovative treatment schemes, such as front-loading doses with subsequent transition to a lower maintenance dosing regimen, to rapidly achieve target concentrations within the first 24 to 48 hours among critically ill patients.
Researchers recently published in the journal Critical Care Medicine a retrospective cohort study looking at the predictive performance of Bayesian monitoring of vancomycin in critically ill patients.
The authors evaluated 2 commercially available Bayesian software programs as well as an additional recently published model.
Predicted vancomycin concentrations were compared against actual, and bias and precision of the models were evaluated.
Data from almost 200 critically ill patients were analyzed. The authors found that not only were bias and precision low, but that the models underpredicted at higher observed vancomycin concentrations.
This means that there was a significant divergence between the observed and predicted values.
While the performance of these models was poor in this patient population, there was no comparison made to dosing by trough levels, and so this study does not directly refute the recommendation to use Bayesian models made by the guidelines. However, the authors suggest:
Until this precision is improved, calculated AUC24 using two concentrations via kinetic equations, or contin- uous vancomycin infusion with a single concentration at steady state, may be preferable in this population.
To access my free download area with 20 different resources to help you in your practice, go to pharmacyjoe.com/free.
If you like this post, check out my book – A Pharmacist’s Guide to Inpatient Medical Emergencies: How to respond to code blue, rapid response calls, and other medical emergencies.
Baysaian stats can be powerful, but after reading about Bayesian dosing, I came away with the impression that it is just fancy empiric dosing. After we have two levels we can know the patient specific Vd and kel, and we no longer need an estimate based on population averages. I still consider myself fairly inexperienced, but it seems to me that vanc doesn’t like to acknowledge any of our predictive kinetics formulas. I’m not surprised to see that these bayesian programs don’t help us dose vanc any better, despite bayesian awesomeness in other areas.