Complete author affiliations and disclosures are at the end of this activity.
Release Date: February 27, 2006
The balance between thrombosis and bleeding has always been a central part of the risk-benefit equation in cardiovascular (CV) therapeutics. As the cardiology community develops more potent means to reduce the risk of thrombosis, the number of patients exposed to a substantial bleeding risk increases. Emphasizing the precariousness of this balance is the recent evidence that agents developed to lower gastrointestinal (GI) bleeding risks (ie, cyclooxygenase-2 [COX-2] inhibitors) may carry an increased risk of thrombosis. These issues weigh heavily on physicians when treating patients to minimize either thrombosis or bleeding. What is the optimal balance to maximize benefits and minimize risks when treating patients?
In this monograph, antithrombotic treatment strategies are reevaluated from a broader perspective, focusing on the potential for serious adverse effects on the GI tract. Aspirin is commonly recommended by cardiologists for the primary and secondary prevention of CV disease, based upon evidence that it significantly reduces the risk of thrombotic CV events. However, aspirin, alone or in combination with nonsteroidal anti-inflammatory drugs (NSAIDs), increases the risk of life-threatening GI complications. Guidelines recommend clopidogrel for hospitalized cardiac patients who cannot take aspirin due to GI intolerance and for patients who need additional cardioprotection from the combination of aspirin with clopidogrel after percutaneous coronary intervention (PCI). This recommendation should be viewed with caution as there is increased recognition that clopidogrel has been found to increase the risk of GI bleeding.
Cardiologists may not routinely evaluate patients for GI risk before recommending antithrombotic/antiplatelet therapy. This will have to change, given the high incidence of NSAID- and antiplatelet-related GI bleeding and the availability of strategies to significantly reduce GI risk. Common treatment strategies used to lower the risk of NSAID-associated GI events are COX-2 inhibitor monotherapy and NSAID/proton-pump inhibitor (PPI) co-therapy. Gastroprotective agents, such as PPIs, are being used more widely because of their well-established efficacy and safety profile and ability to reduce not only dyspeptic side effects, but also the incidence of gastric and duodenal ulcers and associated adverse outcomes.
Recent clinical issues and the safest options for patients with CV and GI risks and/or symptoms associated with NSAID treatment will be covered, including rational strategies for patients requiring anti-inflammatory drugs, taking into consideration the risks and benefits of NSAIDs, COX-2 inhibitors, and PPIs.
As use of over-the-counter (OTC) and prescription NSAID tablets to treat arthritis and pain conditions became widespread, it was recognized that the associated GI side effects were associated with considerable morbidity and mortality. According to data from the Arthritis, Rheumatism and Aging Medical Information System (ARAMIS), it was estimated that every year in the United States, NSAIDs were responsible for as many as 16,500 deaths and more than 100,000 hospitalizations for adverse events, mostly due to GI side effects.
NSAIDs achieve their anti-inflammatory effects by blocking the production of proinflammatory prostanoids by prostaglandin endoperoxide synthase (cyclooxygenase, or COX). The discovery in the 1990s of 2 isoforms of cyclooxygenase, COX-1 and COX-2, led to characterization of their functions. It was determined that COX-1 is constitutively expressed in many cells and is the only isozyme expressed in platelets. COX-2 is induced by inflammatory cytokines. These findings were taken to mean that COX-2 was responsible for the proinflammatory prostanoids and that selectively blocking COX-2 could provide an anti-inflammatory effect without increasing bleeding risk from GI or other sites. Thus, COX-2 inhibitors were developed as a safer alternative to traditional NSAIDs?offering equivalent analgesic and anti-inflammatory efficacy with lower incidence of GI side effects. Consequently, there was a rapid rise in the number of prescriptions that have been written for this class of drugs from 1999 through 2001. In addition to the potential GI benefits of COX-2 inhibitors, basic research has suggested other potential benefits for these drugs. One of the most compelling potential uses of COX-2 inhibitors has been based on the observation that they have been shown to inhibit colonic tumorigenesis. These preclinical observations have led to a number of randomized controlled trials of COX-2 inhibitors to attenuate progression of colonic adenomas.
As is well known, these therapeutic benefits of COX-2 inhibitors have been eclipsed by accumulating evidence of an increased incidence of CV adverse events associated with the use of this class of drugs. The evidence came from parallel sources: (1) basic research, elucidating the potential for COX-2 inhibitors to promote and stimulate a prothrombotic state; (2) observational data; and (3) randomized trials, demonstrating that COX-2 inhibitors, particularly rofecoxib, are associated with an increased risk of adverse cardiac events.
The proposed mechanism of action of COX-2 inhibitors involves membrane phospholipids being converted to arachidonic acid by the action of phospholipase A2. Arachidonic acid is oxidized by COX and then converted to a variety of prostanoids (eg, prostaglandin E2 [PGE2], thromboxane, prostacyclin) by tissue-specific isomerases. The 2 COX isoforms have a specific tissue distribution. COX-1 is found in the GI tract, where it is involved in the production of the gastroprotective PGE2, and in platelets, where it produces the proaggregatory vasoconstricting thromboxane. COX-2 is inducible at sites of inflammation such as the arthritic joint.
The rationale for selective COX-2 inhibitors was based on this concept that inhibition of COX-2 in inflamed joints would spare the gastric housekeeping functions of COX-1. This theory was an oversimplification because subsequent studies showed that COX-1 and COX-2 are more widely distributed than originally thought. Both COX-1 and COX-2 are present at sites of inflammation such as arthritic joints and, importantly, also in atherosclerotic plaques. COX-2 may be found in the vascular endothelium, where it generates prostacyclin, which inhibits platelet aggregation, and has vasodilator properties.[5-7] Prostacyclin and thromboxane have a reciprocal relationship. Thromboxane is critical for platelet aggregation and vasoconstriction; prostacyclin is antiaggregatory and vasodilatory. Thus, a traditional NSAID?by blocking both the COX-1 and COX-2 enzymes?inhibits the generation of both thromboxane and prostacyclin. A COX-2 inhibitor, on the other hand, only inhibits the COX-2?mediated production of prostacyclin (Figure 1).[2,6,9] This tips the balance so that the prothrombotic effects of thromboxane are unopposed, which raises the potential for CV adverse events.
Figure 1. Mechanism-Based FitzGerald Hypothesis
Clinical evidence of the effects of NSAIDs and COX-2 inhibitors on CV risk comes from observational data and randomized trials.
COX-2 inhibitors were shown to increase blood pressure in a meta-analysis of 19 randomized controlled trials (N = 45,451) conducted to assess the influence of COX-2 inhibitors on blood pressure measurements. Based on calculations of weighted mean difference for systolic and diastolic blood pressure, the data suggest that virtually every COX-2 inhibitor is associated with an increase in blood pressure, relative to placebo and nonselective NSAIDs. Of all COX-2 inhibitors, rofecoxib was associated with the greatest level of blood pressure elevation (Figure 2). In addition, rofecoxib appeared to confer a greater risk of developing hypertension in comparison to celecoxib.
Figure 2. Effect of Various NSAIDs on Blood Pressure
Observational data suggested an association between rofecoxib and increased CV events. A nested case-control study using data from Kaiser Permanente in California (N = 1,394,764) was performed to determine whether the risk of serious coronary heart disease was enhanced with rofecoxib at either high (> 25 mg) or standard (</= 25 mg) doses compared with traditional NSAIDs or celecoxib. A cohort of NSAID-treated patients who had acute myocardial infarction (MI) and sudden cardiac death was identified (1999-2001). For every case, 4 controls matched for age, sex, and health plan region were selected. Consistent with data from the aforementioned study, these observational data suggest that rofecoxib increases the risk of serious coronary heart disease, especially at doses > 25 mg. The risk with rofecoxib was greater than the risk with celecoxib. When all current users of rofecoxib were compared with those taking celecoxib, the CV risk was increased 1.59-fold (P = .015).
Prior to this large observational study, a signal that rofecoxib might be associated with increased CV risk was observed in the Vioxx Gastrointestinal Outcomes Research (VIGOR) trial, which was a phase 3, parallel-group, double-blind study of patients with rheumatoid arthritis. VIGOR was conducted: (1) to determine the relative risk of confirmed perforation, ulcers, and bleeding in patients taking rofecoxib, 50 mg/d (Vioxx; Merck & Co, Inc), with those taking naproxen, 1000 mg/d; and (2) to evaluate the safety and tolerability of rofecoxib. The study met the GI end point and demonstrated that during a follow-up of 9 months there were 2.1 confirmed GI events per 100 patient-years with rofecoxib compared with 4.5 per 100 patient-years with naproxen (relative risk [RR], 0.5; 95% confidence interval [CI], 0.3 to 0.6, P < .001). However, the published paper in the New England Journal of Medicine reported that MIs were less common in the naproxen group than in the rofecoxib group (0.1% vs 0.4%; RR, 0.2; 95% CI, 0.1 to 0.7). An analysis of confirmed adjudicated CV events presented to the US Food and Drug Administration (FDA) showed the following comparative rates per 100 patient-years?1.67 vs 0.70 for rofecoxib and naproxen, respectively. Because this was not a placebo-controlled trial, it was unclear whether the increased risk was in fact attributable to rofecoxib or to a cardioprotective effect of naproxen. This question was debated in the scientific community. Based on the results of VIGOR, the FDA requested that information about the increased risk of CV events with rofecoxib, 50 mg, be added to the product labeling. Whether the increased risk might be specific to the 50-mg dose of rofecoxib, and not the 12.5-mg dose or the 25-mg dose, was open to question.
The question of the CV safety vs placebo and safety across lower doses of rofecoxib was settled by the Adenomatous Polyp Prevention on Vioxx (APPROVe) trial. This was a randomized controlled trial (N = 2586) designed to evaluate the hypothesis that 3 years of treatment with rofecoxib (25 mg/d) would reduce the risk of recurrent adenomatous polyps among patients with a history of colorectal adenomas. During the first 18 months of treatment, the CV event rates were similar in both the rofecoxib and placebo groups. However, after 18 months of treatment, there was a 2-fold RR for MI and strokes associated with rofecoxib (Figure 3). The results of this trial led to the withdrawal of Vioxx from the market.
Figure 3. Rofecoxib: Cardiovascular Data APPROVe APTC Events
Oral valdecoxib and its parenteral prodrug form, parecoxib (never approved in the United States), were studied in 2 randomized controlled studies of patients undergoing coronary artery bypass graft (CABG) surgery.[16,17] The first of these trials was undertaken to examine the efficacy of this drug combination as analgesia for incisional pain after CABG. One group of patients was treated with parenteral parecoxib (the prodrug of valdecoxib; Dynastat; Pfizer Inc) followed by oral valdecoxib (Bextra; Pfizer Inc), while the other group was treated with either standard care or placebo. In this first trial (N = 462) the treatment with parecoxib/valdecoxib was associated with a higher incidence of serious adverse events and sternal wound infection in comparison to placebo. Serious adverse events included MIs in 5 patients taking parecoxib/valdecoxib vs 1 of the control patients, and cerebrovascular events in 9 patients taking parecoxib/valdecoxib vs 1 patient in the control group. The second study was undertaken to evaluate the CV safety of parecoxib/valdecoxib in a large population of CABG patients (N = 1671). It showed that patients who received parecoxib/valdecoxib had an almost 4-fold increased risk of MI, cardiac arrest, stroke, and pulmonary embolism compared with placebo (P = .03). In 2004, a warning was added to the label for valdecoxib contraindicating its use in patients undergoing CABG surgery. In April 2005, Bextra was withdrawn from the market because the overall risk-vs-benefit profile for the drug was unfavorable.
Celecoxib was studied for its effect in preventing GI events in the large Celecoxib Long-term Arthritis Safety Study (CLASS); an analysis of this trial for CV events in aspirin nonusers found that there was no increased risk of CV events associated with celecoxib. Subsequently, a signal for increased CV events associated with celecoxib was uncovered in the Adenoma Prevention with Celecoxib (APC) study (Figure 4). This trial was conducted with the primary objective of evaluating the potential of celecoxib (200 mg or 400 mg twice daily) to reduce the risk of recurrent adenomatous polyps in 2035 patients with a history of colonic polyps. After 2.8 to 3.1 years of follow-up, an independent data and safety monitoring board concluded that continued exposure to celecoxib placed patients at increased risk of CV events: 4 (0.6%) of the patients who received placebo died or had a nonfatal MI, whereas 27 (2.0%) of the patients who received celecoxib died or had a nonfatal MI. A post hoc analysis of the APC data revealed that concomitant aspirin use did not eliminate the increased CV risk attributable to celecoxib.
A sister study, the Prevention of Colorectal Sporadic Adenomatous Polyps (PreSAP) trial, used celecoxib, 400 mg once daily, and placebo. Unlike the APC, this study did not show a statistically significant CV risk associated with celecoxib. This is consistent with findings from observational studies of celecoxib (Figure 5).[23-27] However, there are caveats for observational studies?eg, confounders, biases (such as recall bias). Given the wide CIs obtained from the PreSAP study, the data cannot refute a hazard association with celecoxib. The differences in results obtained from the APC and PreSAP trials may be attributed to the differences in doses used in the study and in the patient population. PreSAP had a broader international representation than APC. In April 2005, the FDA asked Pfizer to put a black box warning on the celecoxib label.
Figure 4. APC Cancer Trial
Figure 5. Summary of Observational Studies of Celecoxib and CV Disease
In summary, randomized placebo-controlled trials indicate that the COX-2 inhibitors rofecoxib and valdecoxib are associated with an increased CV risk. While this is likely to be a class effect, there may be real differences in the degree of CV risk. There is likely a dose-response relationship. It is noteworthy that approximately 75% of patients take lower doses of celecoxib (200 mg/d) than what was used in the chemoprevention trials (400 mg). The critical duration of exposure to a COX-2 inhibitor associated with the development of CV risk is unclear because of the small sample size of clinical trials to date. In the meantime, as with all drugs, the potential risks of COX-2 inhibitors need to be weighed against the benefits.
Some authors have recommended low-dose aspirin (81 mg) to modulate CV risk in patients who must take NSAIDs or COX-2 inhibitors. Aspirin is 1 of a panoply of antithrombotic agents used to reduce CV risk that may be taken by cardiac patients who are also taking NSAIDs or COX-2 inhibitors for arthritis or pain. In addition to aspirin, patients may be taking clopidogrel or warfarin after stenting, after CABG, or for stroke prevention in atrial fibrillation. Safety issues related to the use of these drugs should be a major consideration.[28,29] All of these agents are associated with a degree of bleeding risk. Some principles for minimizing bleeding risk will be reviewed, followed by a closer look at the risk of GI bleeding and its prevention.
Warfarin is a commonly used anticoagulant whose antithrombotic effect is a consequence of the reduction of all 4 vitamin K?dependent coagulation factors (ie, a vitamin K antagonist). Avoidance of either undertreatment or overtreatment with an anticoagulant, resulting in ischemic stroke or hemorrhagic complications, respectively, is an important therapeutic goal.
The degree of anticoagulation, monitored within narrow therapeutic margins, is influenced by prothrombin time, expressed as the international normalized ratio (INR). The following safety challenges are associated with this anticoagulant: (1) a delayed onset and offset of action; (2) unpredictable dose response; (3) a narrow therapeutic range; (4) multiple drug-drug and drug-food interactions; (5) problematic INR monitoring; (6) a high bleeding rate; and (7) the slow reversibility. Regarding the very narrow therapeutic range?when the INR of 3.0 is exceeded, the rate of intracranial hemorrhage increases exponentially. When the INR decreases to below 2.0, the stroke risk rises.
The best data on anticoagulation come from Sweden, where every person taking warfarin is assigned to 1 of the 46 anticoagulation clinics. A large study, based on computerized records (1990-1997) in a Swedish registry of 42,451 patients, was undertaken to assess how mortality from all causes and intracranial hemorrhage was influenced by varying degrees of anticoagulation, as reflected by INR. A total of 1,250,000 INR measurements were analyzed; there were 3533 deaths. Mortality was strongly related to the level of INR. Excess mortality was associated with a high INR, with the lowest risk of death at an INR of 2.2 to 2.3. All-cause mortality is very high when the INR is less than 2.0 and rises exponentially when the INR rises above 3.0. The risk of death doubles with every 1-unit increase of an INR that is greater than 2.5. Hence, targeting lower values of INR and titrating within a narrower therapeutic window (relative to the current norm) will improve patient outcomes.
The following recommendations are for minimizing the bleeding risk associated with warfarin: (1) beware of acetaminophen; a highly significant dose-response relationship between acetaminophen and warfarin has been shown to alter the response to anticoagulation; (2) initiate with a 5-mg dose of warfarin vs 10-mg dose, despite an eagerness to get patients rapidly anticoagulated; the smaller dose avoids the development of a potential hypercoagulable state; (3) beware of a genetic mutation (variation in CYP2C9 enzyme) that may lead to a significant difference in the metabolism of the S-enantiomer of warfarin. This means that patients who are slow metabolizers (those born with this genetic mutation) will rapidly attain an INR of 2.0 and 3.0 when given a dose of warfarin that is only 1 mg/d or 1.5 mg/d; and finally, (4) consider patient self-management; similar to patients with insulin-dependent diabetes who test their blood glucose, self-management for anticoagulation treatment has been shown to be as effective as clinical management. Empowering patients to manage their condition is motivational and increases patient satisfaction.
Admissions data from the Brigham and Women's Hospital indicate that there has been a time-related increase in warfarin-related bleeding from 0.97 per 1000 admissions to 1.19 per 1000 admissions from 1995 to 2001. The main reason was that patients were receiving more concomitant medications that increased the risk of bleeding events, including other anticoagulants, antiplatelet agents, NSAIDs, or warfarin potentiators. Patients who bled maintained a steady therapeutic INR until a few days before the bleeding episode, which suggests that an uncharacteristic increase in the INR within the therapeutic range?eg, between 2.0 and 3.0?should not be ignored. This means that if a patient's INR rises relatively rapidly from 2.2 to 2.8, an intervention may be necessary at an INR of 2.8, even though this is still within the therapeutic range. This intervention may prevent an exponential increase in INR (to 4.0 or 5.0), which inevitably leads to hemorrhage.
Data show that bleeding risk can be lessened by patients' attendance at an anticoagulation clinic. A study comparing outcomes of patients from an anticoagulation clinic (n = 183) vs patients obtaining usual medical care (n = 145) with regards to anticoagulation control, patient outcomes, and healthcare costs indicated significantly better overall control of bleeding events in an anticoagulation clinic. This was attributed to a systematic approach to care that reduced complications that may or may not be related to warfarin treatment. Clinician-patient communication can be a key component in determining factors not directly related to warfarin treatment but that may nonetheless affect the treatment outcome. For example, nurses, pharmacists, or physician assistants, while interacting with the patient, can obtain information suggesting that the patient is experiencing other medical problems that, if treated, may improve the outcome of anticoagulation therapy and may even prevent emergencies from developing.
The use of antiplatelet agents such as aspirin and clopidogrel is widespread. Clopidogrel had been recommended by the American College of Cardiology/American Heart Association guidelines for patients who cannot tolerate aspirin. As Chan and others have reported, however, clopidogrel treatment also increases the risk of recurrent GI bleeding in patients who have an ulcer. Cardiac and stroke patients who had a recent history of bleeding ulcer were enrolled in a 12-month, prospective, double-blind trial (N = 320) where patients were randomized to either clopidogrel, 75 mg, or aspirin, 80 mg, plus the PPI esomeprazole, 20 mg twice daily (for the prevention of recurrent bleeding from ulcers). The cumulative incidence of recurrent upper GI tract bleeding during the 12-month period was 8.6% in the clopidogrel group vs 0.7% in the aspirin plus esomeprazole treatment group (P = .001). The longer the patients were followed up, the more dramatic the differences were between the 2 groups, showing a significant protective effect of aspirin plus the PPI over clopidogrel monotherapy.
A combination of 2 antiplatelet agents with independent risks for bleeding events compounds the chances of hemorrhage. Results from Clopidogrel in Unstable angina to prevent Recurrent Events (CURE), a study of 12,562 patients with acute coronary syndrome, were consistent with this premise. Patients using aspirin (75 mg/d to 325 mg/d) were randomized to clopidogrel or placebo for up to 1 year; the 3 aspirin dose groups were </= 100 mg, 101 to 199 mg, and >/= 200 mg. A dose-response relationship between aspirin and bleeding was demonstrated. The incidence of major bleeding rose with increasing aspirin dose both in the placebo (1.9%, 2.8%, and 3.7%, respectively; P = .0001) and clopidogrel groups (3.0%, 3.4%, and 4.9%, respectively; P = .0009). By lowering the dose of aspirin, the risk of major bleeding may be substantially attenuated without loss of efficacy. Indeed, the major bleeding rates for combined use of low-dose aspirin plus clopidogrel are lower than for using aspirin alone at a higher dose.
To minimize bleeding risk and maximize antithrombotic efficacy, physicians should use low doses of aspirin for CV prevention, whether alone or in combination with clopidogrel. For PCI, the American College of Chest Physicians (ACCP) recommends pretreating patients with aspirin, between 75 mg and 325 mg. However, for long-term treatment after PCI, the ACCP recommends lower-dose aspirin, between 75 and 100 mg/d, especially for patients taking other antithrombotic agents such as clopidogrel or warfarin.
At-risk patients taking aspirin for cardioprotection may also need to take NSAIDs or COX-2 inhibitors for the treatment of arthritis or pain. NSAIDs (including aspirin) have been shown to increase the risk of upper GI tract bleeding and perforation.[39-41]
There are 3 types of GI symptoms associated with NSAID use: (1) dyspepsia, (2) peptic ulcer, and (3) upper GI tract complications such as bleeding and perforations. NSAID users may be twice as likely to experience dyspepsia (compared with controls), especially at high doses. Gastric or duodenal ulcers occur in 15% to 30% of individuals who take NSAIDs regularly. However, upper GI tract symptoms correlate poorly with endoscopic ulcers. Any patient taking NSAIDs has a significant risk of upper GI tract complication; there is a 4-fold increase in risk for upper GI tract bleeding. A history of ulcer and older age are the high risk factors for upper GI tract complications.[44,45]
NSAIDs interfere with 3 lines of the gastric mucosal defense: preepithelial, epithelial, and postepithelial. The mucus layer is part of the preepithelial defense?protecting the gut from acid. The surface epithelial cells can repair acid-related damage through cell proliferation, a process that is dependent on prostaglandins. The postepithelial barrier protects against deep mucosal damage through the secretion of bicarbonate, which neutralizes protons that breach the more superficial lines of defense. NSAIDs impair all lines of mucosal defense, increase acid secretion, and induce microvascular ischemia. Aspirin, for instance, can cause superficial injury to the epithelium and render the stomach vulnerable to the physiologic amounts of acid that are normally present.
Several strategies may be used to reduce the vulnerability of the gut to NSAID-related GI complications: (1) eradicate Helicobacter pylori; (2) use alternatives to NSAIDs?eg, acetaminophen, safer "traditional" NSAIDs, and COX-2 inhibitors; and (3) co-therapy with a gastroprotective agent.
An NSAID-induced ulcer is not typically associated with inflammation. On the contrary, the presence of H pylori causes inflammation. H pylori and NSAID use are independent risk factors for ulcer formation that have an additive effect. Hence, if an H pylori?positive patient is taking an NSAID (including low-dose aspirin), H pylori eradication is recommended, especially if the patient has a history of ulcer, with or without complications. Would it make sense to test every NSAID or aspirin user for H pylori? It would make sense for patients with at least a 2-fold risk of developing ulcers.
What about safer anti-inflammatory drugs? Consider the "tweener" NSAIDs. They are neither COX-2 inhibitors nor potent COX-1 inhibitors. There are a number of them, including etodolac and diclofenac. At low doses, these agents inhibit only COX-2, but at higher doses, they inhibit COX-1.
The safety of etodolac relative to a traditional NSAID was demonstrated in a historical cohort study conducted at the Dallas Veterans Affairs Medical Center (N = 16,286). The objective of this study was to assess clinically significant upper GI tract event rates associated with either etodolac or naproxen (with low-dose aspirin) during a 3-year period. Etodolac was associated with a significantly decreased incidence of clinically significant upper GI tract events (similar to a COX-2 inhibitor)?about a 50% reduction relative to naproxen. Concomitant use of low-dose aspirin increased clinically significant upper GI tract event rates 2-fold with naproxen and 9-fold with etodolac. Hence, co-therapy with aspirin negated the GI benefits of etodolac.
Evidence that COX-2 inhibitors facilitate a larger reduction in the incidence of GI side effects compared with NSAIDs or aspirin was gleaned from the VIGOR trial, which showed approximately a 50% difference in the GI events favoring rofecoxib over naproxen monotherapy. This difference is not maintained if aspirin is used concurrently. Animal data suggest that the dual inhibition of COX-1 and COX-2 isoforms underlie an NSAID-induced ulcer. The considerable increase in the frequency of ulcers seen with combination treatment of aspirin and a COX-2?specific inhibitor, compared with COX-2 inhibitor monotherapy, is consistent with this finding.
Because of the relatively lower rates of upper GI tract adverse events associated with COX-2 inhibitors, this class of drugs was widely used even among patients whose GI risk was poorly identified. However, COX-2 inhibitors have been associated with ulcer rates that are higher than placebo. A population-based cross-sectional time series analysis was performed in Ontario, Canada, using administrative healthcare databases of more than 1.3 million residents who were at least 66 years of age. The time frame of the study was divided into 15 six-month intervals from September 1994 to February 2002. Hospitalization rates for upper GI tract hemorrhage were examined. A 41% rise in COX-2 inhibitor use was associated with a 10% increase in hospitalization rates for GI bleeding.
Given the CV adverse effects that limit the therapeutic utility of COX-2 inhibitors, the co-therapy of NSAIDs with gastroprotective agents has emerged as a preferred option for the prevention of NSAID-induced GI events?especially since no study has shown that COX-2 inhibitors are more efficacious than NSAIDs.[13,53,54] Four pharmacotherapeutic agents are currently being used for co-medication with NSAIDs: misoprostol, sucralfate, histamine2 receptor antagonists (H2RAs), and PPIs. Misoprostol, an oral synthetic prostaglandin analog, is FDA approved for reducing the risk of developing NSAID-related gastric ulcerations and complications from gastric ulcers. However, this agent is not commonly used for these indications because few patients can tolerate the 200-μg 4 times daily dose that is needed to effectively reduce GI risk. Moreover, cramps and diarrhea are typical side effects of the drug. Sucralfate is indicated for the treatment of duodenal ulcer, although there are no compelling data on the efficacy of this agent for NSAID-related GI complications and it should not be used for this purpose. H2RAs such as cimetidine, famotidine, and ranitidine, on the other hand, have been shown to reduce NSAID-related duodenal ulcers. However, the majority of ulcers caused by aspirin and NSAIDs occur in the stomach, for which H2RAs demonstrate suboptimal efficacy. Comparative efficacy trials evaluating gastric acid inhibition have shown that PPIs are more effective than H2RAs in reducing the incidence of gastric or duodenal ulcers. Two PPIs, esomeprazole and lansoprazole, are specifically indicated for risk reduction of NSAID-associated gastric ulcer.[57,58]
The efficacy of gastroprotective therapy with PPIs was demonstrated in a trial of 1429 patients taking traditional NSAIDs or COX-2 inhibitors for osteoarthritis or rheumatoid arthritis who were randomized to esomeprazole, 20 mg or 40 mg, or placebo to evaluate the efficacy for the prevention of gastric or duodenal ulcers. Patients were required to have at least 1 of the following risk factors: (1) age of 60 years or older or (2) a history of gastric or duodenal ulcers within the past 5 years. Approximately 95% of patients remained ulcer free with active PPI therapy. Esomeprazole, both 20 mg and 40 mg, significantly reduced the incidence of gastric and duodenal ulcers, relative to placebo, in both traditional NSAID and COX-2 inhibitor users. Notably, the rate of ulcers among the patients taking placebo were comparable between those who were taking nonselective NSAIDs and those taking COX-2 inhibitors (Figure 6), suggesting that patients who are at risk may need multiple risk-reducing strategies for gastric mucosa protection.
Figure 6. Ulcer Incidence by NSAID Type
Both misoprostol and PPIs are effective for preventing NSAID-related ulcers. Their comparative efficacy for ulcer prevention in long-term users of NSAIDs was evaluated in a prospective, double-blind, placebo-controlled trial of patients with documented gastric ulcers (N = 537). In this 12-week study, patients were randomized to receive either 200 μg of misoprostol 4 times daily, 15 mg or 30 mg of lansoprazole once daily, or placebo. All of the active treatments were significantly better than placebo. By week 12, the percentages of gastric ulcer-free patients were as follows: placebo, 51%; misoprostol, 93%; lansoprazole, 15 mg, 80%; and lansoprazole, 30 mg, 82%. Discontinuations were significantly higher in the misoprostol treatment group. Taking both efficacy and safety/tolerability into consideration, therapy was successful for 69% in each of the active treatment groups and 35% for the placebo group. More trials are needed to assess the efficacy of PPIs for NSAID-induced ulcer complications.
Whether patients at high risk due to prior bleeding should receive a COX-2 inhibitor or a combination of NSAID plus PPI has been investigated. In a study by Chan et al, patients (N = 287) who used NSAIDs for arthritis and who presented with ulcer bleeding were randomized to receive either celecoxib, 200 mg twice daily, plus placebo daily or diclofenac, 75 mg twice daily, plus omeprazole, 20 mg daily, for 6 months. Data from the intent-to-treat analysis showed that the probability of recurrent bleeding during the 6-month period was 4.9% for patients who received celecoxib and 6.4% for those who received diclofenac plus PPI co-therapy. This suggests that, among patients with a recent history of ulcer bleeding, treatment with a COX-2 inhibitor is as effective as an NSAID plus PPI combination. Hence, if an anti-inflammatory needs to be used in a patient with a very high GI risk, combining a COX-2 inhibitor with a PPI may be the best option, provided that the patient does not have a CV risk and/or is not on an aspirin regimen.
In H pylori?positive patients who regularly take aspirin, the combined preventive strategies of eradicating H pylori and embarking on a maintenance regimen using a PPI may significantly reduce the recurrence of ulcer complications. In a prospective, 12-month, controlled trial, patients (N = 123) with aspirin-induced ulcer complications were randomized to either lansoprazole, 30 mg/d, or placebo, following H pylori eradication. Treatment with aspirin, 100 mg/d, was continued. After 12 months, patients in the lansoprazole group were significantly less likely to have a recurrence of ulcer complications compared with patients taking placebo (1.6% vs 14.8%, respectively; P = .008).
Data from standardized longitudinal information (1981-2000) on patients with rheumatoid arthritis (N = 5598) were used to examine whether recent preventive strategies have been associated with changes in the incidence of NSAID gastropathy. The annual rate of hospitalization due to bleeding, obstruction, or perforation was the major outcome measure used in the assessment. The risk of serious NSAID-induced GI events declined 67% since 1992. The rate reduction in GI-related hospitalizations was closely associated with lower doses of ibuprofen and aspirin, a decline in the use of more toxic NSAIDs and an increase in the use of "safer" NSAIDs. Moreover, it was estimated that 18% of this decline was associated with the use of PPIs.
In summary, GI toxicity is an important clinical consideration in patients who take NSAIDs and aspirin on a regular basis, particularly those with high GI risk. PPI co-therapy with NSAIDs or COX-2 inhibitors significantly decreases the risk of recurrence of ulcers. For patients with the highest GI risk, including those who use multiple NSAIDs and aspirin, more than one risk-reducing strategy may be necessary.
Until 2004, risk assessment in the process of choosing an NSAID (including aspirin) centered on GI issues. With the development and subsequent marketing of the COX-2 inhibitors, clinicians pondered whether the promise of substantial GI risk reduction associated with these agents justified the relatively high cost of this class of pharmacotherapy. As a result of the barrage of reports about COX-2 inhibitor?related thrombotic events, CV risk took over as the primary consideration in choosing NSAIDs. However, the shift in emphasis of risk assessment toward CV risk drew attention away from the NSAID-related GI issues. Clearly, a balanced assessment of CV and GI risks is necessary for long-term NSAIDs users.
Figure 7 shows the risk factors for NSAID-related GI events. A history of ulcer is a strong risk factor, as is the use of multiple NSAIDs, which oftentimes involves aspirin and another NSAID. In this context of NSAID co-therapy, aspirin is a "forgotten risk factor."[39,40,65,66]
Figure 7. Aspirin and NSAIDs: When Two Rights Make a Wrong
Aspirin is cardioprotective and NSAIDs are effective analgesics. However, the concomitant use of aspirin and another NSAID increases GI risk, as shown in a large cohort study from Denmark (N = 27,694) with low-dose aspirin (100-150 mg daily). The incidence rate ratio for GI events with low-dose aspirin monotherapy was 2.6, while that of a combination of low-dose aspirin and a traditional NSAID was 5.6. The risk was similar among users of noncoated and coated low-dose aspirin.
Three randomized trials have shown that aspirin, when taken concurrently with a COX-2 inhibitor, mitigates the gastroprotective effects of the COX-2 inhibitor.[50,68,69] In a 12-week, double-blind study (N = 1615) comparing placebo, low-dose aspirin (81 mg/d), rofecoxib (25 mg/d) plus low-dose aspirin, and ibuprofen (800 mg 3 times daily), the risk of endoscopic ulcers with the combined aspirin plus COX-2 inhibitor regimen was comparable to that of NSAID monotherapy. The cumulative incidence of ulcers at the end of the study was 16.1% for rofecoxib plus aspirin and 17.1% for ibuprofen. In CLASS (N = 8059), aspirin use (</= 325 mg) was also found to significantly increase the incidence of upper GI tract ulcer complications in celecoxib-treated patients (400 mg twice daily) who were treated for 6 months. Among low-dose aspirin users, those who took celecoxib had a relative risk of 2.01 for an upper GI tract ulcer complication, which was comparable to the relative risk of 2.12 in counterparts who took NSAIDs (P = .92). And lastly, the Therapeutic Arthritis Research and Gastrointestinal Event Trial (TARGET), a 52-week trial (N = 18,325), randomized patients to treatment with lumiracoxib, 400 mg daily, naproxen, 500 mg twice daily, or ibuprofen, 800 mg 3 times daily?stratifying for low-dose aspirin use (75-100 mg/d) and age (>/= 50 years old)?to assess GI and CV safety. Among those taking low-dose aspirin, the difference in the percentage of patients who experienced GI events was not significant between those taking lumiracoxib and those taking NSAIDs (0.69% vs 0.88%; P = .49).
The GI risks may be compounded even more by taking more than 2 NSAIDs concurrently (eg, adding an OTC NSAID such as ibuprofen or naproxen). In fact, a telephone survey of long-term COX-2 inhibitor users older than 55 years revealed that roughly half had used aspirin. It is also important to emphasize that the GI complications of aspirin may be compounded by the addition of another antiplatelet agent, especially when taken with another NSAID.
The format of the information in the Table allows practitioners to consider therapeutic options for individual patients in terms of their GI and CV risks. For those who have neither GI nor CV risk, it is acceptable to use traditional NSAIDs. Based on what we now know regarding COX-2 inhibitors, these agents should be limited strictly to patients who could benefit from their GI safety advantage but have no CV concerns?at least until it is known, for sure, that low-dose celecoxib is in fact safe for individuals with known CV risk factors. For those with a high GI risk and without cardiac risk, acetaminophen is an appropriate option. If acetaminophen is ineffective, COX-2 inhibitor monotherapy may be appropriate; alternatively, consider co-therapy using a traditional NSAID and a PPI.
Table. A Clinician's Guide to Balancing GI and CV Risk in 2005
However, the clinician has to be a bit more deliberative when dealing with a patient with known CV risk. There have been no studies to date showing that celecoxib, the lone COX-2 inhibitor available on the market, is proven to be safe in patients with CV risk factors. Aspirin may be used for cardioprotection, but it markedly diminishes the GI safety advantage of COX-2 inhibitors. Moreover, there is no evidence aspirin abrogates the CV risk associated with COX-2 inhibitor therapy. If a patient has a high risk of GI events and is taking aspirin, or if the patient has a moderate GI risk and is taking aspirin and another NSAID, most experts would recommend considering gastroprotection with either a PPI or misoprostol or considering non-NSAID therapy.
The patient with both CV and GI risk is the most problematic, but is typical in cardiology practice. These patients are typically older, have a GI risk factor, and are taking aspirin and perhaps another antiplatelet agent. Moreover, many of them are taking NSAIDs?OTC or prescription, alone or in combination?products that may contain ibuprofen or naproxen. Unreported OTC NSAID use is common. In patients with high CV risk who are taking aspirin alone, those with moderate CV risk taking aspirin and another antiplatelet agent, or patients with known CV risk factors who are taking aspirin and an NSAID, the standard of care is to add gastroprotective therapy to their medication regimen, especially if they must take drugs that would increase the likelihood of GI adverse events.
In summary, when choosing an NSAID, (1) assess cardiac risk factors first because of the evidence that COX-2 inhibitors may potentiate CV risk; (2) evaluate whether aspirin use is truly beneficial for the individual patient; and (3) remember that there is no GI safety advantage to COX-2 inhibitor treatment when given concomitantly with aspirin. For patients with GI risk factors, consider non-NSAID therapies. However, for those patients who require NSAID treatment, consider co-therapy with either a PPI or misoprostol.
A. Mark Fendrick, MD
Professor of Internal Medicine, School of Medicine; Professor of Health Management & Policy, School of Public Health, University of Michigan, Ann Arbor, Michigan
Disclosure: Consultant: Amgen, AstraZeneca Pharmaceuticals LP, Aventis Pharmaceuticals Inc., Eli Lilly and Company, GlaxoSmithKline, Merck & Co., Inc., Merck-Medco, Pfizer Inc., Procter & Gamble, TAP Pharmaceutical Products Inc.
Samuel Z. Goldhaber, MD
Professor of Medicine, Harvard Medical School; Staff Cardiologist; Director of the Anticoagulation Service; Director of the Venous Thromboembolism Research Group, Brigham and Women's Hospital, Boston, Massachusetts
Disclosure: No financial relationships to disclose with regard to commercial support
James M. Scheiman, MD
Professor of Internal Medicine, Division of Gastroenterology, University of Michigan Medical School, Ann Arbor, Michigan
Disclosure: Grant/Research Support: AstraZeneca Pharmaceuticals LP, Merck & Co., Inc., Pfizer Inc.; Consultant: AstraZeneca Pharmaceuticals LP, The GI Company, McNeil Consumer & Specialty Pharmaceuticals, Merck & Co., Inc., NitroMed Inc., Novartis Pharmaceuticals Corporation, Pfizer Inc., TAP Pharmaceutical Products Inc.; Speakers' Bureau: AstraZeneca Pharmaceuticals LP, Boehringer Ingelheim, TAP Pharmaceutical Products Inc., Wyeth Pharmaceuticals
Scott D. Solomon, MD
Director, Noninvasive Cardiology, Brigham and Women's Hospital; Associate Professor of Medicine, Harvard Medical School, Boston, Massachusetts
Disclosure: No financial relationships to disclose with regard to commercial support