Recent data have suggested that regular consumption of nonsteroid anti-inflammatory drugs (NSAIDs), particularly selective inhibitors of cyclo-oxygenase-2 (COX-2), is associated with an increased risk of thrombotic events. It has been suggested that this is due to NSAIDs reducing the release from the endothelium of the antithrombotic mediator prostaglandin I2 as a result of inhibition of endothelial COX-2. Here, however, we show that despite normal human vessels and endothelial cells containing cyclo-oxygenase-1 (COX-1) without any detectable COX-2, COX-1 in vessels or endothelial cells is more readily inhibited by NSAIDs and COX-2-selective drugs than COX-1 in platelets (e.g., log IC50+/-SEM values for endothelial cells vs. platelets: naproxen -5.59+/-0.07 vs. -4.81+/-0.04; rofecoxib -4.93+/-0.04 vs. -3.75+/-0.03; n=7). In broken cell preparations, the selectivities of the tested drugs toward endothelial cell over platelet COX-1 were lost. These observations suggest that variations in cellular conditions, such as endogenous peroxide tone and substrate supply, and not the isoform of cyclo-oxygenase present, dictate the effects of NSAIDs on endothelial cells vs. platelets. This may well be because the platelet is not a good representative of COX-1 activity within the body as it produces prostanoids in an explosive burst that does not reflect tonic release from other cells. The results reported here can offer an explanation for the apparent ability of NSAIDs and COX-2-selective inhibitors to increase the risk of myocardial infarction and stroke.