Prediction of the metabolic sites for new compounds, synthesized or virtual, is important in the rational design of compounds with increased resistance to metabolism. The aim of the present investigation was to use rational design together with MetaSite, an in silico tool for predicting metabolic soft spots, to synthesize compounds that retain their pharmacological effects but are metabolically more stable in the presence of cytochrome P450 (P450) enzymes. The model compound for these studies was the phenethyl amide (1) derivative of the nonsteroidal anti-inflammatory drug (NSAID) indomethacin. Unlike the parent NSAID, 1 is a potent and selective cyclooxygenase-2 (COX-2) inhibitor and nonulcerogenic anti-inflammatory agent in the rat. This pharmacological benefit is offset by the finding that 1 is very unstable in rat and human microsomes because of extensive P4503 A4/2D6-mediated metabolism on the phenethyl group, experimental observations that were accurately predicted by MetaSite. The information was used to design analogs with polar (glycinyl) and/or electron-deficient (fluorophenyl, fluoropyridinyl) amide substituents to reduce metabolism in 1. MetaSite correctly predicted the metabolic shift from oxidation on the amide substituent to O-demethylation for these compounds, whereas rat and human microsomal stability studies and pharmacokinetic assessments in the rat confirmed that the design tactics for improving pharmacokinetic attributes of 1 had worked in our favor. In addition, the fluorophenyl and pyridinyl amide derivatives retained the potent and selective COX-2 inhibition demonstrated with 1. Overall, the predictions from MetaSite gave useful information leading to the design of new compounds with improved metabolic properties.