Based on the principle that vancomycin therapy requires sustained ther
apeutic concentrations while avoiding high peaks, some authors reporte
d that optimal vancomycin levels could be ensured by measuring trough
levels alone (Cmin), The aim of this work, was to assess the performan
ce of a one-compartment Bayesian forecasting method for estimating van
comycin 2 hours after Infusion (C2h) and mean vancomycin concentration
in steady state (Cavgss) on the basis of a single trough sample (Cmin
), in different conditions (steady slate, patient renal function, and
age), and according to clinical significance. Vancomycin serum concent
rations (n = 108) were analyzed by fluorescence polarization immunoass
ay, from 79 adult patients. The predictive performance of the Bayesian
method was determined by calculating the mean prediction error (ME),
the mean absolute error (MAE) and the root squared prediction error (R
MSE). A linear regression analysis was carried out between estimated a
nd observed concentrations. The predicted C2h were not significantly d
ifferent from the observed, and the least biased (ME = -1.08) and most
precise (MAE = 3.81) predictions were from patients with normal renal
function and steady state conditions. In this population, the concord
ance in dosage recommendations viith the data pair results was 75% of
patients. The best correlation between observed and predicted concentr
ations was found for Cavgss (r = 0.94;p < 0.00005). Predictions of the
Cavgss were more precise (ME = -0.54) and accurate (MAE = 1.74) than
the C2h predictions. Vancomycin can be monitored by determining one le
vel in steady state for most patients with normal renal function.