Non-steroidal anti-inflammatory drugs (NSAIDs) achieve their anti-inflammatory actions through an inhibitory effect on cyclooxygenase (COX). Two COX subtypes, COX-1 and COX-2, are responsible for the majority of COX activity at the gastrointestinal mucosa and in tissues with inflammation, respectively. We previously suggested that both gastric mucosal cell death due to the membrane permeabilization activity of NSAIDs and COX-inhibition at the gastric mucosa are involved in NSAID-induced gastric lesions. We have also reported that loxoprofen has the lowest membrane permeabilization activity among the NSAIDs we tested. In this study, we synthesized a series of loxoprofen derivatives and examined their membrane permeabilization activities and inhibitory effects on COX-1 and COX-2. Among these derivatives, 2-{4'-hydroxy-5-[(2-oxocyclopentyl)methyl]biphenyl-2-yl}propanoate 31 has a specificity for COX-2 over COX-1. Compared to loxoprofen, oral administration of 31 to rats produced fewer gastric lesions but showed an equivalent anti-inflammatory effect. These results suggest that 31 is likely to be a therapeutically beneficial and safer NSAID.