Two prostanoids have major physiologic roles in the vasculature: TxA2 and PGI2 (prostacyclin) (Figure 7).^[78] TxA2 promotes platelet activation, vasoconstriction, and smooth muscle proliferation. TxA2 is mainly produced in platelets and its formation is increased upon platelet activation. Increased excretion of TxA2 metabolites has been reported in patients with unstable angina during episodes of chest pain and has been associated with higher risk of major vascular events in patients with peripheral arterial obstructive disease.^[79,80] The importance of TxA2 in development of acute arterial thrombosis has been established by the efficacy of aspirin in clinical trials of secondary prevention of stroke and MI.^[81] Prostacyclin is a vasodilator and a potent inhibitor of platelet aggregation. It is produced by macrovascular endothelial cells and is the major product of arachidonic acid metabolism when these cells are cultured in vitro.^[82] Pharmacologic studies of the physiologic roles of prostacyclin are limited because there is no specific antagonist of the prostacyclin receptor that can be used in vivo. Studies in mice lacking prostacyclin receptors have indicated the importance of prostacyclin for limiting occlusive vascular events. These mice did not develop spontaneous thrombosis, but were more susceptible to thrombogenic stimuli in vivo compared with normal animals.^[83] In patients with unstable angina, prostacyclin synthesis is increased during episodes of chest pain and parallels increases in the synthesis of TxA2.84 Therefore, animal model and clinical data support the notion that prostacyclin mediates a local compensatory response, limiting consequences of platelet activation. It should be noted that even low doses of aspirin depress synthesis of prostacyclin, but the overall cardioprotective effect of such therapy demonstrates that any negative effects are smaller than the benefits derived from TxA2 inhibition.^[8]

Of the 2 COX isoforms, only COX-1 is expressed in the platelets. 2,85 Therefore, only those COX inhibitors that inhibit this isoform can inhibit synthesis of TxA2 and have cardioprotective effects. Apart from aspirin, all NSAIDs inhibit COX competitively, and the effects on platelet aggregation depend on the pharmacokinetic profiles of the agents. As thrombotic events can occur at any time and for a prolonged period after rupture of a vulnerable plaque, sustained inhibition of platelet activity is needed in order to provide cardioprotection. Even short periods of reduced inhibition during trough periods between dosings may be permissive to occlusive events in a thrombogenic environment. Aspirin is unique among COX inhibitors because it irreversibly inactivates COX enzymes by covalent modification.^[86] Platelets do not have a nucleus and cannot synthesize new COX molecules to replace those that have been inactivated. After administration of a single dose of aspirin, platelet aggregation is impaired for up to 4 days, until new platelets enter the circulation in sufficient numbers. This mechanism of action accounts for the effectiveness of aspirin in prevention of vascular events in patients at risk.^[81]

Both COX isoforms can be expressed in vascular endothelium. In unstimulated cultured endothelial cells, only COX-1 is detectable and TxA2 is the predominant arachidonic acid metabolite produced.^[87] Activation of endothelial cells by inflammatory stimuli or shear stress leads to expression of COX-2.87,88 As the overall COX activity of the endothelial cells increases, all prostanoids are produced at a higher rate. Kinetic activities of Tx synthase, PGI synthase, and PGE synthase, the enzymes involved in metabolism of arachidonic acid downstream from the COX, have different kinetic properties and are saturated at different amounts of the substrate.^[82] Formation of TxA2 in endothelial cells plateaus early, whereas prostacyclin and PGE2 synthesis are increased to a greater degree. Studies in healthy volunteers show that treatment with COX-2 inhibitors decreases systemic production of prostacyclin with no effects on platelet-derived TxA2 synthesis.^[61,89] Therefore, under normal physiologic conditions, COX-2 seems to be induced in the vascular endothelium and seems to be the major determinant of systemic prostacyclin synthesis. The difference between in vitro and physiologic data may be explained by partial activation of COX-2 expression in the vasculature under conditions of laminar blood flow. In addition, the expression of COX-2 is increased in atherosclerotic plaques.^[90]

In an animal model, myocardial ischemia led to increased COX-2 activity and production of prostacyclin and PGE2.^[91] The production of prostacyclin is increased during episodes of chest pain in patients with unstable angina, indicating that COX-2 is activated during ischemia in humans as well.^[84] If upregulation of endothelial COX-2 represents a protective mechanism against vascular injury, inhibition of COX-2 without concomitant inhibition of platelet TxA2 production may influence the overall prothrombotic/antithrombotic equilibrium and exacerbate the potential for thrombotic complications in patients at risk. However, the relative role of prostacyclin in limiting thrombosis compared with nitric oxide and other protective factors is not known. Additionally, inflammation is an important component of atherosclerosis, and the expression of COX-2 is increased in the foam cells in the atherosclerotic plaques.^[90] Therapy with COX-2 inhibitors may therefore suppress development of atherosclerosis and/or the processes leading to plaque rupture. Traditional NSAIDs inhibit platelet synthesis of TxA2, but in most cases the inhibition probably does not last throughout the dosing period.^[76] An epidemiologic study found no evidence of cardioprotective or prothrombotic effects of traditional NSAIDs, 92 but no randomized trials with prespecified cardiovascular end points have been performed with these agents. The clinical experience with coxibs is accumulating, but the questions of their cardiovascular safety in certain patient populations and of the need for concomitant aspirin therapy are still open.^[1,93]

In the 2 large trials with coxibs, the NSAID comparators were naproxen in VIGOR and diclofenac and ibuprofen in CLASS. Neither study included a placebo arm. Although all traditional NSAIDs inhibit COX-1, different NSAIDs may have different effects on platelet aggregation and thus varying potential for cardioprotection. This is not surprising, as the virtually complete inhibition of TxA2 synthesis in platelets needs to be achieved in order to inhibit Tx-dependent platelet activation. Evidence suggests that the antiplatelet effects of naproxen are significant and comparable to those of aspirin, while they are smaller with ibuprofen and diclofenac.^[1,76] Pharmacodynamic studies indicate that inhibition of platelet aggregation with naproxen 500 mg twice daily is consistent throughout the dosing period. Ibuprofen, which has a shorter half-life, achieves adequate platelet inhibition during periods of peak level, but the inhibition is not sustained throughout the dosing interval (Figure 8).^[76] The efficacy of naproxen for prevention of arterial thrombosis has never been examined in a clinical trial, but pharmacologic properties of this agent indicate that it may have greater cardioprotective effects than some of the other NSAIDs. This inhibition of platelet activation may be particularly relevant in the absence of low-dose aspirin and in patients at higher risk for adverse cardiovascular events.

A relatively high percentage of patients (21%) in the CLASS trial were receiving therapy with cardioprotective doses of aspirin.^[36] In the VIGOR trial, on the other hand,

concomitant use of aspirin was one of the exclusion criteria. A retrospective analysis of the VIGOR study examined the rates of cardiovascular events among patients with risk factors associated with recommendations for prophylactic use of low-dose aspirin (history of stroke, MI, unstable angina, angina pectoris, or surgical or percutaneous coronary revascularization).^[35,76] A total of 321 patients were identified in this group, representing 3.9% of the total study population. Almost half of the total MIs (47%) occurred in this group of patients. Among patients who did not fulfill criteria for aspirin indication, the difference in the rates of MIs was not statistically significant (0.2% with rofecoxib and 0.1% with naproxen). Therefore, it is possible that all patients who could have derived benefit from low-dose aspirin therapy received such treatment in the CLASS trial, whereas use of aspirin could have prevented some, if not most, of the excess events in a select group of patients enrolled in the VIGOR trial.

Finally, the percentage of patients with RA was 27% in the CLASS study and 100% in the VIGOR study. Several epidemiologic studies documented increased cardiovascular morbidity and mortality among patients with RA compared with the general population or with OA patients.^[94-96] A recent analysis of thromboembolic events among patients with RA used data collected in the General Practice Research Database (GPRD).^[97] More than 6.5 million patients who were older than 40 years of age and did not have a prior history of MI or stroke were included in the study. In this analysis, patients with RA had a 55% greater risk of an MI and 45% greater cardiovascular mortality than patients who did not have arthritis. Compared with OA patients, those with RA had a 32% greater risk of an MI and 41% greater risk of cardiovascular mortality. The reasons for increased cardiovascular risk in this patient population seem to include inflammation as the underlying pathophysiology for both RA and atherosclerosis.^[96,98] The greater risk for cardiovascular-events among patients with RA may have played a role in the outcome of the VIGOR trial. Based on results from animal models, absence of COX-2 activity does not lead to spontaneous thrombosis, but may increase response to thrombotic stimuli.^[83] In patients with RA, the presence of mediators of inflammation and increased plasma levels of fibrinogen and other thrombogenic factors may provide an environment conducive to adverse cardiovascular events.^[96] Cardioprotection with a platelet inhibitor is likely to be of greater importance for RA patients with additional risk factors, compared with matched patients without arthritis or those with OA.

In summary, coxibs interact with vascular physiology by inhibiting synthesis of prostacyclin by vascular endothelial cells. This effect is seen with traditional NSAIDs as well, but nonspecific COX inhibitors additionally suppress COX-1-mediated platelet production of TxA2. The degree and duration of inhibition of platelet aggregation varies among different NSAIDs and its clinical significance has not been determined with certainty. None of the smaller placebo- or active comparator-controlled trials, individually or in aggregate, indicated that therapy with coxibs is associated with an increase in cardiovascular events. Only 1 trial to date, VIGOR, has suggested such an association. This trial enrolled a population of patients that is at greater risk for vascular events, therapy with low-dose aspirin was not used, and patients in the comparator arm were treated with an NSAID that may have significant cardioprotective activity. The total number of vascular events is small and the possibility that the difference in events is the result of a play of chance cannot be discounted. If the difference is real, there are 2 possible explanations for the higher rate of events in the rofecoxib arm. First, in a subset of patients with RA who have additional cardiovascular risk factors, selective COX-2 inhibition may shift the balance between prothrombotic and antithrombotic prostanoids and thus suppress a homeostatic mechanism that decreases sensitivity to thrombotic stimuli. The second possibility is that patients treated with rofecoxib in the VIGOR trial experienced the same number of events as they would have without therapy, while those in the control group had the risk for vascular events reduced through cardioprotective effects of therapy with naproxen. As the VIGOR trial did not include a placebo arm, it is impossible to distinguish between these 2 possibilities with certainty, and both are consistent with known pharmacologic activities of rofecoxib and naproxen. A recent analysis by Mukherjee et al attempted to determine whether the rates of cardiovascular events in the studies with coxibs were the same or greater than those that would be expected in the placebo arms.^[93] The authors compared the annualized rate of MIs among 23,407 patients receiving placebo in 4 aspirin primary prevention trials with those seen among patients treated with coxibs in the CLASS and VIGOR trials. In the aspirin trials meta-analysis, the annualized MI rate in the placebo group was 0.52%. The annualized MI rates were higher in large trials with coxibs: 0.74% in the rofecoxib arm of VIGOR (P = .04 vs placebo of the meta-analysis) and 0.80% in the celecoxib arm of CLASS (P = .04 vs placebo of the meta-analysis). As the authors of this analysis acknowledge, intertrial comparisons of data are always problematic. Confounding factors include differences in patient inclusion criteria, concomitant therapy, and event reporting/adjudication procedures. As noted above, patients who had manifested vascular disease were included in VIGOR and accounted for a large percentage of events, while the comparison group involved primary prevention. It should be noted that patients enrolled in the VIGOR trial had RA and were therefore probably at higher risk for cardiovascular events than the population enrolled in placebo-controlled primary prevention trials with aspirin. Interestingly, the difference between the annualized rates of MIs in the rofecoxib arm of VIGOR and the placebo arms from the prevention trials meta-analysis is within the range of reported excess of MIs in patients with RA compared with patients who do not have arthritis.

The implications of findings from existing studies are not clear, and large trials with prospectively defined cardiovascular end points would be needed to define the effects of specific COX-2 inhibition on these clinical outcomes. At this time, coxibs should be used with caution in patients with evidence of vascular disease and should be administered with low-dose aspirin therapy in those patients. Similar recommendations for concomitant aspirin use exist for traditional NSAIDs.