 |
 |
 |
 |
 |
 |
 |
|
|
|
|||||||||
|
|||||||||||
PhD Candidate, Pharmacy Resident, Service Pharmacie-ADCAPT and UMR CNRS 5558, Université Lyon 1, Hospices Civils de Lyon, Hôpital Antoine Charial, 40 avenue de la Table de Pierre, 69340 Francheville, France, fax 33-4-72-32-39-08, sylvain.goutelle{at}chu-lyon.fr
PhD Candidate, Assistant Pharmacist, Department of Pharmacy, ADCAPT, and UMR CNRS 5558, University Hospitals of Lyon, Pierre Garraud Hospital, Lyon, France
Associate Pharmacist, Department of Pharmacy, ADCAPT, University Hospitals of Lyon, Antoine Charial Hospital
Senior Pharmacist, Department of Pharmacy, ADCAPT, and UMR CNRS 5558, University Hospitals of Lyon, Antoine Charial Hospital
Published Online, June 2, 2009. www.theannals.com, DOI 10.1345/aph.1L144a
First, estimated pharmacokinetic parameters were calculated using 2 equations: ke = 0.0024 x CrCl + 0.01, where ke is the elimination rate constant and CrCl is the estimated creatinine clearance (or the estimated glomerular filtration rate), and Vd = 0.3 x weight, where Vd is the volume of distribution. It seems that the use of those coefficient values was supported by a study conducted by Sawchuk and Zaske.2 However, the coefficients reported by Sawchuk and Zaske were different: ke = 0.0042 x CrCl + 7.1 x 10-6, and Vd = 0.25 ± 0.086 L/kg (mean ± SD).
In addition, such equations with nonvariable coefficients are likely to provide poor parameter estimates because aminoglycoside pharmacokinetic variability is inevitably underestimated. In a meta-analysis on netilmicin pharmacokinetics, Keller et al.3 found marked variability in the regression equation for ke: the intercept parameter varied from -0.013 to 0.169 h-1 and the slope parameter ranged from 0.0018 to 0.0031 min·mL-1·h-1. In 1640 patients treated with gentamicin, Vd values reported by Zaske et al.4 ranged from 0.04 to 0.74 L/kg. Moreover, body weight explained only 16% of the variance of Vd, and CrCl explained only about 30% of the variance of ke.
The article also gives very little information about the way that patient-specific parameter values were estimated. It appears that only 2 aminoglycoside concentrations were used to estimate ke, Vd, and aminoglycoside clearance (AGCl). One may suppose that the authors used linear least-squares regression over log-transformed data. When a pharmacokinetic model is fitted to drug concentration data, various factors, including the data fitting method, may affect the estimation of pharmacokinetic parameters. Jelliffe et al.5 have shown that a Bayesian method gave better results in the prediction of gentamicin concentrations than nonlinear and linear least-squares regression methods.
Finally, we have many reservations with the authors' conclusion that the MDRD formula predicts AGCl better than the CGm formula. We believe that the mathematical result is correct but that the interpretation of the MDRD formula as a better predictor of aminoglycoside elimination is misleading. No statistical difference was found between the MDRD and CGm formulas in predicting ke; patient-specific values were overestimated with both formulas. As the equation for clearance is a function of ke and Vd (ie, AGCl = ke x Vd), the estimation of AGCl was influenced by the second factor, Vd, which was arbitrarily fixed at 0.3 L/kg. If the authors had used Vd factors of 0.2 or 0.4 L/kg, which are acceptable values according to the literature, results for clearance might have been totally different, while renal function estimation would have remained unchanged.
Further research is required to quantify relationships between aminoglycoside pharmacokinetic parameters and renal function. To explore this issue, we recommend the use of efficient methods for pharmacokinetic parameter estimation such as Bayesian techniques.
Footnotes
We gratefully acknowledge Professor Roger W Jelliffe MD, Director, Laboratory of Applied Pharmacokinetics, USC School of Medicine, Los Angeles, for his thoughtful comments on the topic.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
 |
 |
 |
 |
 |
 |
