Estimation of Population Pharmacokinetic Parameters of Tobramycin in Intensive Care Unit Patients Using Different Programs

M.T. Mateu; J.J. Perez-Ruixo; V.G. Casabo; N.V. Jimenez

Department of Pharmacy and Pharmaceutics. University of Valencia, Pharmacy Service. Hospital Dr Peset. 46017 Valencia Spain

Aminoglycoside (AMG) antibiotics are used in prophylaxis and treatment of serious infections in the intensive care units (ICU). Many physicians continue to use standard dosing regimens of AMG in all ICU patients, what often result in suboptimal serum concentrations. Several factors, including changing fluid and haemodynamic status of the ICU patients may help to explain-the increase of the distribution volume (Vd) in these patients. A balance between efficacy and toxicity of AMG can be achieved optimizing the dose via interpretation of serum drug concentrations by using population specific parameter values with the bayesian forecasting method. Many approaches to estimate these parameters have been employed: standard two stage (STS), nonlinear mixed-effects models and non parametric maximum likelihood models. The ability to estimate population pharmacokinetic parameters from sparse routine clinical data is, probably now, the most interesting task in the framework of clinical population pharmacokinetics. The predictive and relative performances of the population pharmacokinetic parameters has been evaluated by the bias, precision and the newest “design number” approach.

Objective: To evaluate the efficiency with which different approaches estimate Tobramycin population pharmacokinetic parameters in intensive care unit patients.

Methods: Patients: twenty-five adults ICU patients, with unstable renal function, were given standard doses of Tobramycin 30 min IV infusion.
Blood samples: 84 serum concentrations (Cp)(3.36 ± 1.75 / patient) were collected 30 min after the end of infusion (peak) and just before the dose (trough).
Pharmacokinetic analysis: a monocompartimental model was assumed. Calculations of clearance (CL) and Vd were performed, in steady-state, using STS methods with non-linear least square regression (Abbott PKS, USC*PACK) and nonlinear mixed-effects models (MULTI(ELS), NONMEM). In this last case, CL, Vd, and Cp were affected by an additive error.
Simulations: five sets of data (5×25 patients) with identical features of our ICU patients were generated. Individual Vd values were sampled from a normal population distribution (29.5 ± 8.75 L), using a random number generator. CL was linearly related to the individual creatinine clearance (CL=0.814*CLCr). An additive random error (0 ± 1.1 L/h) was added to the total drug clearance. Covariance between parameters was assumed to be zero. A proportional random error of 20 per cent was added to serum concentrations to produce final observations. The bias and precision for all parameters obtained from the approaches used were calculated through overall design number (Φr).

Results: Population pharmacokinetic parameter estimates for Tobramycin in ICU patients and overall design number for simulation study.

Approaches

Vd(L)

SD VD (L)

CL slope

SD CL (L/h)

Φr

Abbott PKSa

28.70

9.03

1.110

2.14

0.217

USC*PACKa

29.28

15.10

0.772

2.19

0.249

MULTI(ELS)

26.22

7.62

0.329

0.99

0.224

NONMEM

27.90

5.92

0.865

0.69

0.199

aCalculations were realized with 12 patients

Conclusion: All methods were able to estimate similar measures of Tobramycin Vd for a population of ICU patients. However, the lowest interindividual variability of population parameters analysed were achieved with NONMEM method.

Reference: PAGE 3 () Abstr 864 [www.page-meeting.org/?abstract=864]

Poster: poster