Kernel Function Smoothing of Insulin Absorption Kinetics

Abstract

Drug absorption kinetics are usually described by simple descriptive statistics such as the time until 50% is absorbed or by parametric analysis, e.g., compartmental modelling. This paper describes how the absorption alternatively can be estimated nonparametrically by means of kernel function smoothing. This avoids the need for modelling the lag phase and it is still possible to estimate the absorption rate. The methods are applied to analyse the disappearance of radioactively labelled insulin from a subcutaneous depot. The size of the depot is measured by external counting. One application is the absorption of the fast-acting human insulin Actrapid, where the amount of insulin in blood is also predicted. Another application is the study of variation in absorption of a long-acting insulin analogue, Insulin 174, which was found to have significantly less variation than a comparable crystalline insulin suspension preparation.