In vitro versus canine data for predicting input profiles of isosorbide-5-mononitrate from oral extended release products on a confidence interval basis

Abstract

The objective of this study was to assess the relative usefulness of in vitro and canine data sets in the prediction of input rates of isosorbide-5-mononitrate (ISMN) from extended release (ER) products on a confidence interval basis. ISMN is a highly soluble compound with high permeability in the upper GI tract but not in the colon. Two ISMN ER tablet formulations were studied: Imdur™ and an osmotic pump. Differences between amount of drug absorbed versus time predicted from in vitro and canine data and results observed in humans were assessed using the estimated median value of the difference factor, f₁, and its associated 90% bootstrap confidence intervals. For both products, median f₁ values and their associated confidence intervals for the comparisons between in vitro and human profiles and between canine and human profiles were similar. Although the median difference between the human input profiles of the two ISMN ER products was equally well predicted by in vitro and canine data, the 90% confidence intervals for this difference were better predicted by the canine data.

Introduction

For orally administered drugs, optimization of the absorption profile is frequently crucial to therapeutic effectiveness. For example, the hemodynamic response to nifedipine is heavily dependent on the rate of increase of its plasma concentration (Kleinbloesem et al., 1987). Similarly, during long-term treatment with oral nitrates, avoidance of tolerance with simultaneous reduced risk of a rebound increase of myocardial ischemia can be achieved only with long acting nitrate formulations (Waller, 1999).

Optimization of input rates is typically achieved by extending the period of drug release from the administered product using various technological approaches (Mathiowitz, 1999). Selection of the most appropriate extended release (ER) product, or the effect of scaling-up or post-approval formulation changes on in vivo performance must usually be justified by in vivo studies, making development expensive and time consuming.

The possibility of predicting the in vivo performance of an ER product and, therefore, eliminating a number of in vivo studies during the development phase is not a new issue in the scientific literature. Prediction of in vivo performance of ER products is based on in vitro release data (Young et al., 1997, USP 24, 2000) and in few cases on data collected in animals with similar to human gastrointestinal (GI) physiology (Dressman and Yamada, 1991).

Based on numerous relevant studies published during the last four decades, prediction of in vivo performance of ER products from their in vitro release characteristics has been possible only on a serendipitous basis for two major reasons. First, release may not be the exclusive absorption limiting process. Second, intra-luminal release is usually not adequately simulated in vitro. A further limitation is that in vitro–in vivo correlations (IVIVCs) are usually developed on an average basis (i.e. without considering variability issues), presumably due to the high variability of the correlated parameters and, therefore, due to the inappropriateness of common parametric procedures for the construction of confidence intervals. However, during the last decade there has been considerable progress in all these subjects. The Biopharmaceutics Classification Scheme provided the theoretical basis for defining the conditions under which an IVIVC is to be expected (Amidon et al., 1995). Based on this concept and using appropriate in vitro test conditions, it became possible to predict the input profile of various lipophilic compounds (Nicolaides et al., 2001). Further, recently proposed distribution-free procedures for comparing highly variable cumulative or non-cumulative data versus time sets (Vertzoni et al., 2003, Vertzoni et al., in press) can also be used in developing IVIVCs on a confidence interval basis.

Correlations of animal with human data have also been largely serendipitous for similar reasons. First, absorption may not be limited to same degree by the release process in both species. Second, the in vivo model conditions are frequently far from being similar to human. For example, pigs are known to have unpredictable gastric emptying kinetics and substantially longer small intestine than humans (Kararli, 1995). On the other hand, beagles, the most frequently used canine model, have substantially shorter colon and, therefore, they are not appropriate for assessing ER products of drugs that are absorbed to a significant extent by the colonic mucosa (Dressman and Yamada, 1991).

The aim of this manuscript was to compare the usefulness of biorelevant in vitro release data and canine data collected in large-sized dogs for the prediction of input rates of isosorbide-5-mononitrate (ISMN) on a confidence interval basis. ISMN is a small (MW 191), non-ionizable, highly soluble compound (Martindale, 1999) with oral absolute bioavailability in humans of at least 95% when administered in immediate release formulations (Abshagen et al., 1981, Wood et al., 1984, Hutt et al., 1995). In an extensive study by Fischer et al. (1987) in which plasma levels were correlated with gastrointestinal transit times in humans, it was shown that ISMN is efficiently absorbed for up to 6–8 h post-dosing from both stomach and small intestine when administered in ER pellets in the fed state, whereas absorption from the colon is low.