Limited Sampling Strategies to Estimate the Area under the Concentration-time Curve

Abstract

Over 100 limited sampling strategies (LSSs) have been proposed to reduce the number of blood samples necessary to estimate the area under the concentration-time curve (AUC). The conditions under which these strategies succeed or fail remain to be clarified. We investigated the accuracy of existing LSSs both theoretically and numerically by Monte Carlo simulation. We also proposed two new methods for more accurate AUC estimations. We evaluated the following existing methods theoretically: i) nonlinear curve fitting algorithm (NLF), ii) the trapezium rule with exponential curve approximation (TZE), and iii) multiple linear regression (MLR). Taking busulfan (BU) as a test drug, we generated a set of theoretical concentration-time curves based on the identified distribution of pharmacokinetic parameters of BU and re-evaluated the existing LSSs using these virtual validation profiles. Based on the evaluation results, we improved the TZE so that unrealistic parameter values were not used. We also proposed a new estimation method in which the most likely curve was selected from a set of pre-generated theoretical concentration-time curves. Our evaluation, based on clinical profiles and a virtual validation set, revealed: i) NLF sometimes overestimated the absorption rate constant Ka, ii) TZE overestimated AUC over 280% when Ka is small, and iii) MLR underestimated AUC over 30% when the elimination rate constant Ke is small. These results were consistent with our mathematical evaluations for these methods. In contrast, our two new methods had little bias and good precision. Our investigation revealed that existing LSSs induce different but specific biases in the estimation of AUC. Our two new LSSs, a modified TZE and one using model concentration-time curves, provided accurate and precise estimations of AUC.