Experience with a computer-based technique for estimating pharmacokinetic constants from limited data.

Abstract

A computer-based approach for estimating pharmacokinetic data and predicting plasma drug concentrations based on two measurements of plasma drug concentrations was compared with a standard two-point method for estimating aminoglycoside (AG) pharmacokinetic parameters. The computer-based technique was also used to predict a theophylline steady-state nadir from two preceding ones. The computer-based and standard two-point methods provided similar estimates for mean (+/- SD) values of elimination rate constant (lambda) (0.172 +/- 0.103 h-1 and 0.179 +/- 0.010 h-1, respectively) and apparent volume of distribution (V) (22.7 +/- 9.9 L and 23.3 +/- 11.5 L, respectively). The mean (+/- SD) of predicted AG troughs was 1.3 +/- 0.5 mg/L for the standard method and 1.3 +/- 0.6 mg/L for the computer-based method. Neither was significantly different from the actual AG trough mean of 1.3 +/- 0.7 mg/L. Both methods exhibited a slightly negative bias in their predictions of actual AG troughs. When applied to theophylline data from patients or subjects receiving multiple-dose theophylline, the computer-based method effectively predicted trough theophylline concentrations for three formulations of theophylline. Mean prediction errors ranged from 0.1 to 0.5 mg/L, and root mean squared error ranged from 1.1 to 1.2 mg/L. This computer-based technique may be useful in dose individualization since relatively few constraints are placed on the type of data required for its application.