Cardiovascular and Gastrointestinal Toxicity of Selective Cyclo-oxygenase-2 Inhibitors in Man

Financial Disclosures: The authors have no financial relationships or conflict of interest relevant to this publication. The authors bear all responsibilities for the information and their interpretation herein.

Abstract:

It is well established that the use of traditional nonsteroidal anti-inflammatory drugs (NSAIDs) increases the vulnerability of the gastrointestinal (GI) mucosa for the development of peptic lesions and serious ulcer complications. In addition, selective and traditional NSAIDs have also been associated with increased frequency of cardiovascular toxicity, especially in susceptible patients. The objective of this communication is to provide an overview of the salient GI and cardiovascular (CV) toxicity for these drugs. Traditional NSAIDs inhibit the constitutional cyclooxygenase-1 (COX-1) enzyme responsible for eicosanoids biosynthesis not only in joints, a beneficial effect, but also in the stomach, a detrimental effect. Selective NSAIDs were specifically designed to preferentially inhibit the cyclooxygenase-2 (COX-2), an inducible enzyme mediating the production of inflammatory eicosanoids in the joints but sparing the endogenous protective eicosanoids in the stomach. Selective COX-2 inhibitors (COXIBs) have been shown to possess much improved GI tolerability and reduced GI related adverse events when compared with nonselective COX-1inhibitors.

An unexpected CV toxicity had emerged during the COXIBs post marketing outcome studies. Many subsequent studies were carried out to define the CV risks associated with COXIBs and NSAIDs. All COX inhibitors had shown this CV toxicity. In many clinical studies, rofecoxib use was associated with significantly more elevated CV risk when compared with celecoxib and non selective NSAIDs. The COX inhibitors associated CV toxicity has multiple manifestations, which include the induction of myocardial infarction (MI), edema, thrombosis, blood pressure destabilization and death. Patients at risk of CV disease or with a history of CV disease were the most significant determinants of CV events after receiving COX inhibitors. This CV toxicity not only led to the marketing withdrawal of rofecoxib and valdecoxib but also resulted in more restricted, but essentially identical, product labels in the United States for celecoxib and traditional NSAIDS. This CV toxicity is dose and treatment duration dependent and appears to be compound specific rather than COX specific. Additional comprehensive, long-term, prospective investigations comparing the CV and GI safety profile of marketed NSAIDs against each other and against selective inhibitors are needed to address the controversy of COX inhibitors.

Nonsteroidal anti-inflammatory drugs are one the most widely used class of drugs throughout the world. It is also well established that the use of NSAIDs increases the vulnerability of the GI mucosa for the development of peptic lesions (erosions, inflammation and ulceration) and serious ulcer complications (bleeding and perforation). In fact, in the United States alone, a total of 16,500 patients with rheumatoid arthritis or osteoarthritis died during 1997 from the GI toxicity of NSAIDs [1].

The efficacy and toxicity of NSAIDs is a consequence of the inhibition of the COX enzymes [2]. Recent research has disclosed at least two types of COX enzymes exits: COX-1 is a constitutive enzyme responsible for housekeeping functions in organs such as the stomach, kidney, intestine and platelets, while COX-2 is an inducible enzyme exerting its action at inflammatory sites of the joints and muscles [3,4]. Such important findings led to the development and subsequent introduction of the selective COX-2 inhibitors celecoxib, valdecoxib and rofecoxib, which have considerably reduced GI ulcerogenicity potential when compared with the older, non-selective NSAIDs [5]. Unexpectedly, during the post marketing studies, it was learned that COXIBs exhibited CV toxicity which was not found during their initial clinical development. The purpose of this communication is to review salient aspects of the GI and CV toxicity of NSAIDs and COXIBs which are relevant to their use as anti-rheumatic and analgesic drugs.

Overview of Gastrointestinal Toxicity of NSAIDs

The mechanisms of GI injury by traditional NSAIDs include topical and systemic components [6-8]. Topical mucosal injury usually occurs after the ingestion of aspirin and other weakly acidic NSAIDs. Weak acids are not ionized in the acid environment of the stomach, freely penetrate the gastric barrier and increase the back diffusion of gastric acid across the mucosa causing further aggravation of the injury. The chronic administration of a prodrug, such as Sulindac, produces ulcers, even though its acute topical damaging action on the gastric mucosa is at minimum. Furthermore, enteric-coated preparations of aspirin and other NSAIDs are also ulcerogenic when administered chronically even though such preparations have minimum topical injury following acute administration. Such lines of evidence suggest that the topical injurious effect of NSAIDs is probably of limited pathophysiologic importance.

The inhibition of the COX-1 enzyme induced by traditional NSAIDs mediates their GI toxicity [6]. Traditional NSAIDs inhibit duodenal mucosal bicarbonate and gastric mucus secretion and reduce gastric mucosal blood flow as a consequence of their inhibitory effects on the biosynthesis of protective endogenous prostaglandins [5,6]. Of interest is the observation that NSAIDs are associated with a higher frequency of injury to the stomach than the duodenum and this observation has major therapeutic implications [7,8]. NSAIDs also prevent the increase in cell replication at the ulcer margins, an action that has obvious effects on mucosal repair and ulcer healing [6]. The introduction of COXIBs with high selectively for the inhibition of COX-2 enzyme afforded drugs which are considerably less ulcerogenic than traditional NSAIDs possessing COX-1 inhibitory activity [5,9,10]. In the celecoxib and the rofecoxib major efficacy studies, the incidence of symptomatic ulcers and ulcer complications were significantly lower with these COXIBs than the comparative NSAIDs indicating that COXIBs have much less GI toxicity than traditional NSAIDs, while possessing similar anti-inflammatory and analgesic actions [9,10].

Historical Overview of COXIBs Cardiovascular Toxicity

The initial concern about potential CV toxicity of COXIBs was prompted by the disclosure of the rofecoxib outcome study referred to "Vioxx and Gastrointestinal Outcomes (VIGOR) trial" [10]. The study compared the safety of rofecoxib (Vioxx) with naproxen in patients with rheumatoid arthritis who did not regularly take aspirin. In this trial, it was observed that there was a 5-fold increase in the risk of acute myocardial infarction (MI) with rofecoxib as compared with naproxen. In contrast, a similar outcome clinical study with another COXIB referred to as "Celecoxib Long-Term Arthritis Safety Study (CLASS)" compared celecoxib with ibuprofen and or diclofenac, but included the use of aspirin, did not show increased CV events between celecoxib and traditional NSAIDs, irrespective of aspirin use [9]. Although, the basis for the increased CV toxicity of rofecoxib over naproxen was not fully known at the completion of the VIGOR study, it was thought that it could have been a consequence of the absence of the use of aspirin, an antithrombotic agent in the trial design, since aspirin might induce ulcer and this could negate the rofecoxib benefits of decreased ulcerogenicity potential. However, given the absence of an increased CV toxicity noted with celecoxib in the CLASS trial when compared to the comparator drugs diclofenac and ibuprofen indicates that celecoxib does not exhibit more cardiovascular toxicity than the comparator NSAIDs ibuprofen and diclofenac. Additional independent reanalysis of the CLASS trial examining thromboembolic events found no increased MI risk associated with celecoxib over diclofenac and ibuprofen [11].

Subsequently, interim analyses of two large chemoprevention trials with rofecoxib [Polyp Prevention on Vioxx (APPROVEe)] and celecoxib [Adenoma Prevention with Celecoxib (APC)] showed increased cardiovascular risk among patients receiving the COXIBs in these studies and both trials had to be discontinued [12,13]. It is to be noted, however, that the dosages of the COXIBs used in these chemoprevention trials were much higher than the recommended anti-rheumatic/analgesic dosages used for treatment of osteoarthritis and rheumatoid arthritis.

Of more importance is the observation that naproxen (220 mg bid) also appeared to possess greater CV toxicity than celecoxib (200 mg bid) in the Alzheimer's Disease Anti-inflammatory Prevention Trial (ADAPT) and the study was therefore terminated because of this toxicity concern [14]. For celecoxib, the ADAPT data do not show the same level of risk as those of the APC trial, even though both studies used elderly patients. Nevertheless such observations derived from the COXIBs chemoprevention studies and the ADAPT trial clearly indicate that there is an inherent intrinsic risk for the induction of the CV toxicity not only in COXIBs but also for NSAIDs and this risk needs to be further assessed, especially with regards to other COX inhibitors, patient at risk, dosages, duration of treatment and other factors as discussed in this communication.

Retrospective and Prospective Controlled Clinical Studies Assessing Cardiovascular Risk for COXIBs and NSAIDS

Given the intrinsic CV toxicity shown for rofecoxib and celecoxib, there were several investigations and numerous review papers aimed at defining the CV risk for both COX-2 selective and COX-1 inhibitors. Several epidemiological studies assessing the CV risk of NSAIDs have been reported [15-21]. An analysis of retrospective observational studies, meta-analysis and systematic reviews found an increased CV risk of COX-2 inhibitors and NSAIDs [22-24]. It is important therefore to review some of the major clinical studies that were published with specific experimental design, with large patient population and well defined objectives in order to better define the CV risk associated with selective versus non-selective COX inhibitors. A brief narrative description of some of the trials and the main findings are detailed below:

(a) Solomon et al. [25] conducted a matched case-control study of 54,475 patients, aged 65 years or older, and studied the relative risk of acute MI among users of celecoxib, rofecoxib, and NSAIDs in Medicare beneficiaries with a comprehensive drug benefits over the course of 12 months period in the United States. The investigators constructed matched logistic regression models including indicators for patient demographics, healthcare use, medications use and CV risk factors. As is evident from Table 1, current use of rofecoxib was associated with significantly elevated MI risk compared with celecoxib. In contrast, celecoxib was not associated with an increased relative risk of MI in any of the comparisons examined (Table 1). Dosages of rofecoxib >25 mg were associated with higher risk than dosage less than or equal to 25 mg. Furthermore, the risk was elevated in the first 90 days of use. Although not shown in Table 1, the MI risk was significantly elevated in patients with co-morbid conditions such as diabetes, hypertension, previous MI, angina, previous coronary revascularization, congestive heart failure, cerebrovascular attack, use of hormone replacement therapy and rheumatoid arthritis. This study clearly suggests that patients with a history of diabetes, CV disease and other comorbidity are at much increased risk for the induction of MI and CV adverse events.

(b) Johnsen et al. [26] conducted a population-based, case-control Danish study of 10,280 cases of first time hospitalization for MI and 102,797 sex- and age-matched non MI controls. Relative risk estimates for MI were adjusted for a history of CV disease, hypertension, diabetes, chronic bronchitis or emphysema, alcoholism, liver cirrhosis, upper GI bleeding, rheumatoid arthritis and systemic lupus erythematosis. The investigators found that COX-1 and COX-2 inhibitors (rofecoxib, celecoxib, etodolac, meloxicam and nabumetone) were associated with increased risk of MI (Table 2). However, the MI risk estimate for celecoxib had the lowest value when compared with the other drugs studied in this report.

(c) Huang et al. [27] studied the risk of acute MI, angina, stroke and transient ischemic attack (TIA) in long-term users of rofecoxib and celecoxib in Taiwan and in comparison with meloxicam. The data were taken from National Health Insurance database for the period of 2001 to 2003. Patients included in this study had used one of these three drugs for at least 180 days. The main outcome measurements were the occurrence of MI, angina, stroke or TIA after the initiation of long-term continuous use of these drugs. Person-time exposures and hazard ratios (HRs) were calculated based on data obtained from 9602 patients. The investigators found that in patients without a history of CV events within the year before drug treatment was started; the overall rates for MI, angina, stroke and TIA were 1.1%, 0.6%, 2.0 % and 0.6%, respectively. In those with history of CV events in the year before treatment began, the overall rates for MI, angina, stroke, and TIA were 5.0%, 4.8%, 6.6% and 5.8%, respectively. Of major interest is the finding that celecoxib users had a statistically lower HRs for the development of acute MI (HR 0.78; P = 0.02) and stroke (HR 0.78; P < 0.001) when compared with meloxicam users (Table 3). In contrast, rofecoxib users were at no higher risk of CV events than those receiving meloxicam, indicating that celecoxib has better CV tolerability than either rofecoxib or meloxicam. This result clearly indicates that rheumatic patients with positive history of CV disease will exhibit increased CV events by some 5 to 10 times than patients with no history of CV disease.

In a related study reported by the same investigators who conducted a follow-up, population-based, analysis using data from the Taiwanese Bureau of National Health Insurance (Taipei, Taiwan) database [Huang et al., 28]. Briefly, the study examined eligible patients, 18 years or older, had received etodolac, nabumetone, ibuprofen, naproxen or celecoxib for period equal to or greater than 180 days. The primary outcomes measure was the prevalence of serious CV events, MI, angina, cerebrovascular attack and/or TIA. Analyses were performed on data from all eligible patients and hazard ratios (HRs) were calculated to determine the risk for CV events with long-term use of the drugs. It was found that there were no significant differences in the risk of treatment-related CV events between groups prescribed 1 of 4 NSAIDs (etodolac, nabumetone, ibuprofen, or naproxen) or celecoxib. Furthermore, history of CV disease was the most significant determinant of CV events risk. Patients with preexisting medical conditions appeared to have a significantly higher risk of CV events associated with the use of either NSAIDs or celecoxib compared with patients without these conditions.

(d) Rahme and Nedjar [29] conducted a retrospective cohort study using administrative healthcare records of patients aged 65 or greater and who filled a prescription for NSAID or acetaminophen during 1999-2000. The study compared the risks of hospitalization for acute MI and GI bleeding among these patients. Outcomes were compared using Cox regression models with time dependent exposures. Among non-users of aspirin, the adjusted hazard ratios (95% confidence interval) of hospitalization for the combined outcome for MI/GI vs. acetaminophen groups are shown in Table 4.

For non-aspirin users, naproxen had the highest risk of GI bleeding (Table 4) among all drugs tested. In contrast, the combined outcome MI/GI toxicity of celecoxib was the lowest among all drugs tested and was essentially similar to acetaminophen. Of interest is the observation that rofecoxib had a higher adjusted hazard ratio values for both MI and GI endpoints than celecoxib.

For patients who used aspirin, the adjusted hazard ratios for the combined MI/GI outcome are listed in Table 5. Rofecoxib had the highest MI/GI risk when compared to COX-1 NSAIDs and celecoxib. Although not shown in the Table 5, naproxen had the highest GI risk, but the lowest MI risk. The drugs acetaminophen, celecoxib and naproxen had the lowest combined MI/GI hazard ratios of 1.29, 1.34 and 1.35, respectively.

It is to be noted that although acetaminophen is not a NSAID and has no anti-inflammatory value, GI toxicity or antiplatelet activity, it is very frequently used as an analgesic in arthritic patients due to its GI safety. However, it has recently been reported that acetaminophen increases the risk of hypertension in older women and such hypertension might increase the CV risk [29].

(e) McGettigan and Henry [22] conducted a large meta-analysis of observational studies which included 17 case-control analyses (N = 614,193) and 6 cohort studies involving very large number of patients (N = 1,045,859) and recorded CV events (thrombotic events, MI and stroke, which can be fetal) relative to nonuse/remote exposure. The relative risks (95 % confidence intervals) for COX-1 and COX-2 inhibitors are shown in Table 6. The results clearly indicate that the CV risk appears to be compound specific rather than being specific to the type COX being inhibited. The highest CV events risk was associated with diclofenac and rofecoxib. On the other hand, the lowest CV events risk was associated with naproxen and celecoxib. This analysis further indicate that there was no statistically significant increase in CV events risk with the use of celecoxib compared with non use/remote NSAID exposure of any type.

Duration of Drug Treatment and Risk of CV Events

Many of the epidemiological studies concerning the CV risk with COXIBs and NSAIDs were not specifically designed to provide information on the duration of the treatment with the COX inhibitors and the occurrence of CV events. To address this issue, Motsko et al [31] conducted a retrospective analysis of the Veterans Integrated Service Network (VA) database. Medicare data and Texas Department of Health mortality data were incorporated to capture events occurring outside the VA healthcare network. The CV events included were MI, stroke, and MI-related death. Patients aged 35 or greater who received celecoxib, rofecoxib, ibuprofen, etodolac and naproxen for three years (1999-2001) were included. A short-term period was defined as 180 days or less, while long-term was defined to include exposure longer than 180 days. Multivariate Cox proportional hazard models and adjusted hazard ratio (95% CI) were determined. The investigators identified 12,188 exposure periods (11,930 persons) and 146 CV events over the entire study period. Compared with short-term ibuprofen, the short-term use of celecoxib (adjusted HR 0.75; 95% CI 0.42, 1.35) and rofecoxib (adjusted HR 0.85; 95% CI 0.39, 1.86) was not associated with any significant CV risk. However, compared with long-term ibuprofen use, long-term use of celecoxib (adjusted HR 3.64; 95% CI 1.36, 9.70) and rofecoxib (adjusted HR 6.64; 95% CI 2.17, 20.28) was associated with significant increase in CV risk. When restricted to patients aged 65 years or older, the CV risks associated with long-term celecoxib (adjusted HR 7.36; 95% CI 1,62, 33.48) and rofecoxib (adjusted HR 13.24; 95% CI 2.59, 67.68) use was increased. Neither long- nor short-term exposure to naproxen and etodolac was associated with cardionegative or cardioprotective effects when compared with ibuprofen. The findings of this VA observational study contradicted a similarly designed Taiwanese study with long-term (greater or equal to 180 days) exposure to NSAIDs and celecoxib as reported by Huang et al. [28]. The Huang et al study found no differences in CV risk between ibuprofen, etodolac, naproxen, nabumetone and celecoxib, when administered for a similar long-term period [28]. The basis for the different outcomes between these two studies is unknown.

Blood Pressure and Renal Effects

Hypertension and arthritis are among the most common chronic conditions in the United States and the prevalence of hypertension among arthritic patients has been estimated to be 32% [32]. Concomitant use of NSAIDs or COX-2 inhibitors with antihypertensive drugs is therefore common among many rheumatic patients. The fact that COX inhibitors can inhibit the production of protective prostaglandins in vascular beds, endothelium and kidneys could influence not only the response of anti-hypertensive drugs and could also trigger a deleterious vassopressor response. Two meta-analysis clearly showed that NSAIDs increased blood pressure (BP) when used at effective anti-rheumatic dosages and the effect was most pronounced in patients with hypertension [33,34]. With the introduction of selective COX-2 inhibitors, it was important to know whether such drugs possess the same pharmacological action on BP as nonselective NSAIDs. Several prospective randomized studies examined the effect of COXIBs and NSAIDs on BP in rheumatic patients with the following conclusions [35-38]:

(a) All COX inhibitors have the potential of increasing BP to variable degrees.

(b) The use of rofecoxib, but not celecoxib or COX-1 inhibitors, was associated with an increased risk of edema and BP increases compared to nonusers of NSAIDs [38].

(c) Head-to-head trial showed significantly lower incidences of destabilized BP and edema with celecoxib compared with rofecoxib in persons aged 65 and older with osteoarthritis and hypertension [36].

(d) Rofecoxib treated patients were approximately 2 to 4 times more likely to report edema and BP increases as side effects compared with celecoxib treated patients [38].

(e) Rofecoxib caused the greatest increase in systolic BP in patients receiving angiotensin converting enzyme inhibitors or beta blockers, whereas those on calcium channel antagonists or diuretic monotherapy receiving either rofecoxib or celecoxib showed no significant increase in BP [35].

Clearly, given these observations, it appears that the vasopressors/edema response induced by NSAIDs or COXIBs, administered at ordinary anti-rheumatic dosages, is compound specific rather than class specific toxicity and the highest risk was shown for rofecoxib. None of the studies reviewed showed that celecoxib had higher vaspressor/edema inducing activity than traditional NSAIDs.

Mechanism of CV Adverse Effects

. . .Continue to read rest of article (PDF).