Amikacin population pharmacokinetics in critically ill Kuwaiti patients

Abstract

Amikacin pharmacokinetic data in Kuwaiti (Arab) intensive care unit (ICU) patients are lacking. Fairly sparse serum amikacin peak and trough concentrations data were obtained from adult Kuwaiti ICU patients. The data were analysed using a nonparametric adaptive grid (NPAG) maximum likelihood algorithm. The estimations of the developed model were assessed using mean error (ME) as a measure of bias and mean squared error (MSE) as a measure of precision. A total of 331 serum amikacin concentrations were obtained from 56 patients. The mean (± SD) model parameter values found were Vc = 0.2302 ± 0.0866 L/kg, kslope = 0.004045 ± 0.00705 min per unit of creatinine clearance, k12 = 2.2121 ± 5.506 h(-1), and k21 = 1.431 ± 2.796 h(-1). The serum concentration data were estimated with little bias (ME = -0.88) and good precision (MSE = 13.08). The present study suggests that amikacin pharmacokinetics in adult Kuwaiti ICU patients are generally rather similar to those found in other patients. This population model would provide useful guidance in developing initial amikacin dosage regimens for such patients, especially using multiple model (MM) dosage design, followed by appropriate Bayesian adaptive control, to optimize amikacin dosage regimens for each individual patient.

Figure 1

Marginal density plot of the parameters k _s (h⁻¹ per unit of creatinine clearance); renal component of elimination rate constant (a) and V _c (L/kg); apparent volume of distribution of the central compartment (b) generated by NPAG program for adult ICU patients (n = 56) who received amikacin.

Figure 2

Scatter plot of predicted (x-axis) versus measured (y-axis) serum amikacin concentrations (μg/mL) based on means (a) and medians (b) of population parameter distributions. Pooled data from all patients (n = 56).

Figure 3

Scatter plot of predicted (x-axis) versus measured (y-axis) serum amikacin concentrations (μg/mL) based on means (a) and medians (b) of each individual patient's Bayesian posterior parameter distributions. Pooled data from all patients (n = 56).

Figure 4

Bland-Altman plot of the mean serum amikacin (x-axis) trough concentrations (a) and peak concentrations (b) versus difference (y-axis) between predicted and observed amikacin concentrations. The solid line represents the mean difference; the dashed lines represent the limits of agreement (mean difference ±2 SD difference).

Figure 5

Three-dimensional plot of apparent volume of distribution of the central compartment (V _c; L/kg) versus renal component of elimination rate constant (k _s; h⁻¹ per unit of creatinine clearance) for Kuwaiti ICU patients (n = 56) using NPAG program.