Potent and selective inhibitors of thrombin were sought based on the (R)-Phe-Pro-Arg sequence. The objective was to generate similar binding interactions to those achieved by potent competitive inhibitors of the argatroban type, so eliminating the need for covalent interaction with the catalytic serine function, as utilized by aldehyde and boronic acid type inhibitors. Improving the S(1) subsite interaction by substitution of arginine with a 4-alkoxybenzamidine residue provided potent lead 2 (K(i) = 0.37 nM). Though an amide bond, which H-bonds to the active site, is lost, modeling indicated that a new H-bond is generated between the alkoxy oxygen atom and the catalytic Ser-195 hydroxyl group. Substitution of the benzamidine system by 1-amidinopiperidine then gave compound 4, which provided a further gain in selectivity over trypsin. However, previous work had shown that these compounds were likely to be too lipophilic (Log D +0.4 and +0.2, respectively) and to suffer rapid hepatic extraction, presumably via biliary elimination. Accordingly, both proved short-acting when administered intravenously to rats and showed poor activity when given intraduodenally. The aim was then to reduce lipophilicity below a log D of -1.2, which in a previously reported series had been effective in preventing rapid clearance. It was anticipated that compounds of this type would rely on the cation selective paracellular route of absorption from the gastrointestinal tract. Potent polar analogues with selectivity >1000 over trypsin were obtained. The best in vivo activity was shown by compound 12. However, in the final analysis, its oral bioavilability proved poor, relative to analogues with similar physicochemical properties derived from argatroban, consistent with the hypothesis that molecular shape is an additional important determinant of paracellular absorption.