Delayed gastric ulcer healing is a well recognized problem associated with the use of cyclooxygenase (COX) inhibitors. In contrast, NO-releasing COX inhibitors do not interfere with ulcer healing. These divergent effects may in part be due to differences in their effects on platelets, which are known to influence ulcer healing. Therefore, we compared the effects of a nonselective COX inhibitor (flurbiprofen), a nitric oxide-releasing COX inhibitor (HCT-1026), and a selective COX-2 inhibitor (celecoxib) on gastric ulcer healing, angiogenesis, and platelet/serum levels of vascular endothelial growth factor (VEGF) and endostatin. Gastric ulcers were induced in rats by serosal application of acetic acid. Daily treatment with the test drugs was started 3 days later and continued for 1 week. Celecoxib and flurbiprofen impaired angiogenesis and delayed ulcer healing, as well as increasing serum endostatin levels relative to those of VEGF. HCT-1026 did not delay ulcer healing nor impair angiogenesis, and also did not change the ratio of serum endostatin to VEGF. Incubation of human umbilical vein endothelial cells with serum from celecoxib- or flurbiprofen-treated rats resulted in suppressed proliferation and increased apoptosis, effects that were reversed by an antiendostatin antibody. These results demonstrate a previously unrecognized mechanism through which nonsteroidal antiinflammatory drugs can delay ulcer healing, namely, through altering the balance of anti- and proangiogenic factors in the serum. The absence of a delaying effect of HCT-1026 on ulcer healing may be related to the maintenance of a more favorable balance in serum levels of pro- and antiangiogenic growth factors.