NSAIDs, gastroprotection and cyclo-oxygenase-II-selective inhibitors*


  • *

    This article was originally published by the National Prescribing Centre (NPC) as a MeReC Briefing (Issue 20, October 2002). MeReC Publications are funded jointly by the Department of Health and the National Institute for Clinical Excellence (NICE), London, UK, and are available on the NPC website: http://www.npc.co.uk/merec.htm.

Correspondence to: R. Micklewright, The Infirmary, 70 Pembroke Place, Liverpool L69 3GF, UK. E-mail: Ruth.Micklewright@npc.nhs.uk


In patients at high risk of NSAID-associated serious upper gastrointestinal complications, gastroprotection with misoprostol or a proton pump inhibitor should be considered. Only misoprostol, 800 µg/day, has been shown to reduce serious upper gastrointestinal complications in a large clinical outcome trial. The benefit of Helicobacter pylori eradication in reducing NSAID-associated gastrointestinal toxicity is controversial, and routine testing for and eradication of H. pylori in NSAID users are not currently advised.

The gastrointestinal safety of rofecoxib and celecoxib has been assessed in large clinical outcome trials which, on first analysis, show benefits over non-selective NSAIDs in the incidence of serious upper gastrointestinal complications. However, longer term gastrointestinal data from the celecoxib study (CLASS) and cardiovascular adverse event data from the rofecoxib study (VIGOR) have questioned the risk–benefit profile of these new drugs and, until they are better understood, it seems sensible not to use them routinely in large numbers of individuals. The gastrointestinal safety of meloxicam and etodolac has not been adequately assessed in such trials. Therefore, evidence for their use instead of non-selective NSAIDs, or instead of celecoxib or rofecoxib, is not robust.

Non-steroidal anti-inflammatory drugs (nsaids) and gastroprotection


NSAIDs cause numerous side-effects, in particular gastrointestinal and renal toxicity. Renal side-effects include fluid retention and hypertension, and NSAIDs may precipitate congestive heart failure and renal failure in some patients. The gastrointestinal toxicity of NSAIDs can be broadly characterized into three groups:1 (i) symptoms such as dyspepsia, heartburn, nausea, vomiting and abdominal pain; (ii) mucosal lesions seen on endoscopy; (iii) serious gastrointestinal complications, such as perforated ulcers or bleeding requiring hospitalization.

It is estimated that 15–40% of patients taking NSAIDs experience dyspepsia.2 However, this common side-effect is not a reliable predictor of serious complications — 50–60% of patients who develop an ulcer or a life-threatening complication will have no prior warning signs or symptoms.1 The link between endoscopic lesions, which occur in up to 40% of NSAID users, and serious complications, which are much less common (incidence of about 1.5% per year), is also poorly understood.3

Table 1 illustrates the incidence of NSAID-associated upper gastrointestinal bleeds, acute renal failure and congestive heart failure that could be expected in an older primary care population. 4

Table 1.  Non-steroidal anti-inflammatory drug (NSAID)-associated adverse effects in an older population (over 65 years) of an average primary care organization of 100 000 patients (where 3800 patients over 65 years of age take NSAIDs) 4
EventCases per year
  1. Reproduced with permission of Bandolier.

Upper gastrointestinal bleed18
Acute renal failure10
Congestive heart failure22

The level of gastrointestinal risk associated with NSAIDs is affected by patient factors, such as advancing age and a history of peptic ulcer disease, and by factors relating to NSAIDs themselves, such as the dose.1, 5–7 All NSAIDs, including low-dose aspirin, increase the risk of serious upper gastrointestinal complications, but to different degrees.7–9 In a meta-analysis of 18 studies, the pooled relative risk (RR) of serious upper gastrointestinal complications associated with NSAIDs was 3.8 [95% confidence interval (CI), 3.6–4.1].7 For low-dose aspirin (50–162.5 mg/day), pooled results from eight studies have shown that the risk of gastrointestinal bleeding is almost doubled [odds ratio (OR), 1.59; 95% CI, 1.40–1.81].9 The Committee on Safety of Medicines has recently issued the following advice about the safe use of NSAIDs:10 (i) drugs such as ibuprofen, which are associated with a lower risk of gastrointestinal ulcer or serious complications, should generally be preferred; (ii) NSAIDs should be started at the lowest recommended dose; (iii) no more than one oral NSAID should be used at a time; (iv) combinations of an NSAID and low-dose aspirin may be associated with increased risk and should only be used if absolutely necessary.

Patients should be reviewed regularly to ensure that NSAIDs are still required and, for certain patients, gastroprotection should be considered. Cyclo-oxygenase-II (Cox-II)-selective inhibitors are discussed later.

Who should be considered for gastroprotection?

It is not feasible to provide gastroprotection for every NSAID user. Serious gastrointestinal complications are rare, and large numbers of patients would have to be treated to prevent one event. However, in populations at high risk of gastrointestinal toxicity, fewer people will need to be treated to prevent one event, and gastroprotection should be considered if an NSAID must be prescribed. A cost-effectiveness analysis of misoprostol,11 based on the MUCOSA study (see next section),5 found that the cost-effectiveness ratio was markedly reduced when high-risk patients were targeted. The cost per averted serious gastrointestinal complication was lowest for patients over 75 years of age who also had a history of peptic ulcer disease.11

It is generally agreed that gastroprotection should be offered to high-risk patients, but it is difficult to define clearly who these are. Suggestions have been based on observational data or the MUCOSA study,5 which, although large, based risk factors on only small numbers of events in each group. A similar assessment of risk factors has recently been carried out with data from the rofecoxib study, VIGOR.12 In its appraisal of Cox-II-selective inhibitors, the National Institute for Clinical Excellence (NICE) suggested that the following factors define patients as high risk:13 (i) age of 65 years or over; (ii) past history of peptic ulcer disease or serious gastrointestinal complications; (iii) concomitant oral steroids or anticoagulants; (iv) presence of serious co-morbidity, such as cardiovascular disease, renal or hepatic impairment, diabetes and hypertension; (v) requirement for prolonged use of maximal doses of NSAIDs.

A meta-analysis of 18 epidemiological studies found that patients with advanced age or a history of peptic ulcer disease (particularly if it was complicated) had the highest absolute risks of NSAID-associated serious gastrointestinal complications.7 Compared with patients aged 25–49 years, those aged 60–69 years had 2.4 times the risk, those aged 70–80 years had 4.5 times the risk, and those over 80 years had 9.2 times the risk. A history of peptic ulcer disease increased the risk 5.9 times, 15.4 times if it was complicated, compared with no history.7 In the MUCOSA study, four factors were found to be significant indicators of risk: age > 75 years, a history of peptic ulcer disease, a history of gastrointestinal bleeding and a history of cardiovascular disease.5 Combinations of these risk factors were additive.5 In the VIGOR study, a history of gastrointestinal events and advancing age, as well as severe disability caused by rheumatoid arthritis, were also found to be significant risk factors, but cardiovascular disease was not.12 An age of > 75 years increased the risk 3.7 times and an age of 65–74 years increased the risk 2.4 times.12

Gastroprotective agents

The safety and efficacy of various gastroprotective agents have been assessed in a Cochrane review of 40 randomized controlled trials.3 This found that misoprostol, proton pump inhibitors and double-dose H2-receptor antagonists were all effective at preventing NSAID-associated endoscopic gastric and duodenal ulcers. However, only misoprostol, 800 µg/day, has been shown to reduce the risk of ulcer complications in a large clinical outcome study.5

Misoprostol. The Cochrane review found that, compared with placebo, misoprostol significantly reduced endoscopic NSAID-associated gastric and duodenal ulcers.3 For gastric ulcers, a dose–response relationship was seen.3 Overall, pooled results from six randomized controlled trials (n = 1791) found endoscopic ulcer rates of 5.4% with misoprostol vs. 17.0% with placebo (OR, 0.29; 95% CI, 0.21–0.39).3 These studies suggest that misoprostol reduces the incidence of endoscopic ulcers by about 70%.3 However, the large clinical outcome study, MUCOSA,5 found that misoprostol reduced serious upper gastrointestinal complications (e.g. perforations, obstruction or bleeding) by 40%. In the MUCOSA study, 8843 patients, prescribed continuous NSAID treatment for rheumatoid arthritis, were randomized to receive either misoprostol, 800 µg/day (if tolerated), or placebo for 6 months. Serious upper gastrointestinal complications occurred in 0.57% of the misoprostol group vs. 0.95% of the placebo group [OR, 0.60; 95% CI, 0.36–0.98; number needed to treat (NNT), 263]. For patients with risk factors (age > 75 years or a history of peptic ulcer disease, gastrointestinal bleeding or cardiovascular disease), serious gastrointestinal complications were more likely. Therefore, although the relative risk reduction was still 40%, the absolute risk reduction was greater and the number needed to treat was lower.14

As outlined above, many people who take NSAIDs experience symptoms such as dyspepsia and abdominal pain. Whilst these correlate poorly with the occurrence of ulcers or serious gastrointestinal complications, they are a common reason for patients to return to their general practitioner and for doctors to prescribe a gastroprotective agent. Unfortunately, misoprostol has not been shown to prevent such symptoms, and it can itself induce gastrointestinal side-effects, particularly diarrhoea.6 In the MUCOSA study, withdrawals due to diarrhoea, abdominal pain and flatulence were more common in patients receiving misoprostol compared with placebo (P < 0.001), but withdrawals due to dyspepsia or nausea were not.5 Pooled data from the Cochrane review showed that misoprostol, both 800 µg/day and 400 µg/day, was associated with significantly more diarrhoea than was placebo, and, as the effectiveness of 400 µg/day on clinical outcomes has not been well studied, the review questioned the practice of using lower doses to avoid adverse effects.3

H2-receptor antagonists. Standard-dose H2-receptor antagonists should not be used for the prophylaxis of NSAID-associated ulcers. Pooled results from the Cochrane review of three randomized controlled trials of standard-dose ranitidine, nizatidine or famotidine (n = 981) showed that these agents reduced the risk of endoscopic duodenal ulcers, but not gastric ulcers, compared with placebo (duodenal ulcer rates, 2.0% vs. 5.5%; RR, 0.36; 95% CI, 0.18–0.74).3 There is evidence from three randomized controlled trials (n = 298) that double-dose H2-receptor antagonists (e.g. famotidine, 40 mg twice daily, or ranitidine, 300 mg twice daily) reduce the incidence of both endoscopic duodenal and gastric ulcers compared with placebo (duodenal ulcer rates, 3.3% vs. 13.6%; RR, 0.26; 95% CI, 0.11–0.65; gastric ulcer rates, 11.3% vs. 25.9%; RR, 0.44; 95% CI, 0.26–0.74).3 However, there is no clinical outcome study to show whether high doses prevent ulcer complications. The evidence that H2-receptor antagonists reduce NSAID-associated symptoms, such as dyspepsia, is weak,3 and the correlation between symptoms and ulcer complications is poor.1,3

Proton pump inhibitors. Compared with placebo, the Cochrane review found that proton pump inhibitors significantly reduced both endoscopic NSAID-associated gastric and duodenal ulcers.3 Overall, pooled results from five randomized controlled trials with omeprazole, lansoprazole or pantoprazole (n = 1216) found endoscopic ulcer rates of 14.5% with proton pump inhibitors vs. 35.6% with placebo (OR, 0.23; 95% CI, 0.18–0.31).3 It is generally viewed that these endoscopic results will translate into clinical benefits for the prevention of serious upper gastrointestinal complications,15, 16 but there is no large clinical outcome study to prove this.3, 16 Omeprazole and lansoprazole have shown reductions in serious upper gastrointestinal complications associated with NSAIDs or low-dose aspirin, but only in small secondary prevention studies of Helicobacter pylori-positive patients with a history of gastrointestinal bleeding.17, 18 There is some evidence that proton pump inhibitors reduce NSAID-associated dyspeptic symptoms.3, 15 However, as discussed previously, symptoms and ulcer complications correlate poorly.1, 3, 15

Direct comparisons of gastroprotective agents. It is difficult to compare the relative efficacy of different gastroprotective agents, as head-to-head studies have generally used doses of comparator drugs that are known to be less effective (e.g. ranitidine, 150 mg twice daily, or misoprostol, 400 µg/day).3 However, misoprostol, 800 µg/day, and omeprazole, 20 mg/day, have both shown superiority over ranitidine, 150 mg twice daily, in preventing NSAID-associated gastrointestinal toxicity.3, 19, 20

Studies have compared omeprazole and lansoprazole with misoprostol. Omeprazole, 20 mg/day, was compared with misoprostol, 400 µg/day, and placebo in a randomized controlled trial (n = 732) of 6 months of maintenance therapy in patients taking NSAIDs who had previously been treated to heal an NSAID-associated ulcer.21 Omeprazole was superior to misoprostol for preventing endoscopic duodenal ulcers (2.6% vs. 10.1%; RR, 0.25; 95% CI, 0.11–0.56; placebo incidence, 12.3%), but the incidence of gastric ulcers was not significantly different.21 However, misoprostol was given at a dose of 400 µg/day, and a higher dose might have been more effective. Also, the H. pylori status of the patients could have affected the results, as omeprazole was more effective at inducing remission in H. pylori-positive patients. Withdrawals due to side-effects were less common in the omeprazole group than in the misoprostol group (3.9% vs. 7.7%; placebo, 1.9%), and, when these withdrawals and dyspeptic symptoms were included, together with erosions or ulcers, in the combined end-point of ‘treatment failure’, more patients were in remission with omeprazole than with misoprostol (61% vs. 48%; P = 0.001; placebo, 27%).21

A recent randomized controlled trial (n = 537) compared lansoprazole (15 or 30 mg/day) with misoprostol, 800 µg/day, and placebo in long-term NSAID users with a history of gastric ulcer who were H. pylori negative.22 At 12 weeks, 93% of the misoprostol group were free from endoscopic gastric ulcers, significantly more than in the 15-mg lansoprazole group (80%, P = 0.01), the 30-mg lansoprazole group (82%, P = 0.04) or the placebo group (51%, P < 0.001). This study confirms that lansoprazole is superior to placebo in preventing NSAID-induced endoscopic gastric ulcers, and that there is no dose–response relationship. However, lansoprazole was inferior to misoprostol when the latter was used at its full therapeutic dose. Again, side-effects were more common in patients taking misoprostol (31%) compared with those taking placebo, lansoprazole (15 mg) or lansoprazole (30 mg) (10%, 7% and 16%, respectively; P < 0.006 for all comparisons with misoprostol). When early withdrawals were classified together with gastric ulcers as ‘treatment failures’, proton pump inhibitors and full-dose misoprostol were viewed as clinically equivalent (treatment was successful in 69% of patients in the active groups compared with 35% in the placebo group).22

Which gastroprotective agents should be used?

It is not possible to make an evidence-based decision on which is the best gastroprotective agent for all patients: a proton pump inhibitor or misoprostol? An individualized choice will be required, taking into account the clinical outcome data for misoprostol, the mainly endoscopic data for proton pump inhibitors, tolerability and cost. There have, however, been consensus-based suggestions in 2000 that:23 (i) proton pump inhibitors are the treatment of choice for H. pylori-positive patients who present with a bleeding ulcer and need to continue NSAIDs; (ii) full-dose misoprostol may be more effective for patients who are H. pylori-negative (but possibly not in patients with recent gastric bleeding); (iii) misoprostol has been shown to reduce NSAID-associated serious upper gastrointestinal complications in a large, relatively unselected group of patients (H. pylori status unknown).

Gastroprotection in patients taking low-dose aspirin

Studies with aspirin have shown that the risk of upper gastrointestinal bleeding is not reduced by changing to either modified-release or enteric-coated preparations.9, 24, 25 Minimizing the dose of standard soluble aspirin to 75 mg/day is likely to be the most effective way of reducing the risk associated with this drug.23, 26 On a population basis, the risk of serious upper gastrointestinal complications associated with low-dose aspirin is relatively low, and widespread gastroprotection does not seem appropriate. A sensible option may be to consider gastroprotection for patients taking low-dose aspirin if they are also taking an NSAID (as this is associated with increased risk10), or if they have a history of peptic ulcer disease or serious upper gastrointestinal complications (consensus suggests previous upper gastrointestinal bleeding is probably the most important risk factor23). Other risk factors, such as advancing age or co-morbidity, are best assessed on an individual patient basis. It could be argued that, as low-dose aspirin is widely used in patients with cardiovascular disease (a high-risk group according to the MUCOSA study5), many will require gastroprotection.27 However, cardiovascular disease was not found to be a risk factor in the VIGOR study,12 and prophylaxis for all of these patients would be a huge undertaking. Unfortunately, studies to guide the choice of gastroprotective agent in patients taking low-dose aspirin are limited. Based on studies of NSAIDs, it is reasonable to assume that misoprostol and proton pump inhibitors will be effective in this setting,27 and small randomized controlled trials with omeprazole and lansoprazole have shown promise.17, 18

What about H. pylori eradication?

Most peptic ulcers are caused by either NSAIDs or H. pylori, but whether there is any interaction between these two factors (which often co-exist) remains a matter for debate. A recent meta-analysis has suggested that there is synergy between these two risk factors.28 However, this finding has been questioned because of concerns over selection bias and heterogeneity between the included trials.29 Most studies involved only small numbers of patients and included even smaller numbers of end-point events (e.g. ulcers). There are data from individual trials to suggest that H. pylori increases, has no effect on, or decreases peptic ulcer risk in NSAID users.30

A recent endoscopic randomized controlled trial in 660 H. pylori-positive patients with no history of peptic ulcer disease found that eradication at the start of a 5-week course of diclofenac significantly reduced peptic ulcers compared with no eradication (2/161 patients with ulcers taking eradication therapy vs. 10/171 with ulcers taking placebo; P = 0.037).31 However, gastroprotection with omeprazole, 20 mg/day, was equally effective (0/155 patients with ulcers taking omeprazole). In contrast, other trials in patients with a history of upper gastrointestinal bleeding or peptic ulcer disease have questioned the benefits of H. pylori eradication.17, 32

So, how should patients taking NSAIDs be managed with regard to H. pylori status? For any patient who develops an ulcer whilst taking an NSAID, in addition to stopping the NSAID wherever possible, it is reasonable to eradicate H. pylori if present, as it will be impossible to tell which is the causative factor.16 It could be argued that there is no evidence of benefit for this approach, and there is a concern that the eradication of H. pylori may retard gastric ulcer healing by proton pump inhibitors; however, in practice, this problem only affects a minority of patients (i.e. those who develop gastric ulcers and continue taking NSAIDs).23 It may also be reasonable to eradicate H. pylori before starting NSAID treatment in any patient with a past history of peptic ulcer disease, as eradication is already recommended in these patients if they become symptomatic or take long-term antisecretory therapy.33 For all other patients who are already taking, or are about to start taking, an NSAID, testing for and eradicating H. pylori are not currently advised. This would be a huge undertaking for primary and secondary care, and the evidence of benefit is still conflicting.

Cox-ii-selective inhibitors

Four selective inhibitors of Cox-II — rofecoxib (Vioxx), celecoxib (Celebrex), parecoxib (Dynastat injection) and etoricoxib (Arcoxia) — have recently been launched in the UK, and valdecoxib is due to be launched at the end of 2002 or the start of 2003. In addition, NSAIDs with a high Cox-II selectivity (e.g. meloxicam and etodolac) are available. The classification of such drugs remains controversial, and NICE has used the term Cox-II-selective inhibitors to refer to all four drugs included in their technology appraisal — celecoxib, rofecoxib, meloxicam and etodolac. All the Cox-II-selective inhibitors are being marketed as drugs that are as effective as traditional non-selective NSAIDs, but with a safer gastrointestinal profile.

Current guidance on Cox-II-selective inhibitors

The NICE guidance on Cox-II-selective inhibitors for osteoarthritis and rheumatoid arthritis is outlined in Table 2.13

Table 2.  National Institute for Clinical Excellence (NICE) guidance on cyclo-oxygenase-II (Cox-II)-selective inhibitors for osteoarthritis and rheumatoid arthritis 13
Cox-II-selective inhibitors are not recommended for routine use  in patients with rheumatoid arthritis or osteoarthritis.  They should be used, in preference to standard NSAIDs,  when clearly indicated as part of the management of  rheumatoid arthritis or osteoarthritis only in patients  who may be at high risk of developing serious  gastrointestinal adverse effects
The risk of NSAID-induced complications is particularly  increased in patients with a previous clinical history  of gastroduodenal ulcer, gastrointestinal bleeding or  gastroduodenal perforation. The use of even a Cox-II-selective  agent should therefore be considered especially carefully  in this situation
In all patients with cardiovascular disease, there remains  uncertainty over the use of Cox-II-selective inhibitors and they  should not therefore be prescribed routinely in preference to  standard NSAIDs where these are indicated in this group.  Furthermore, many patients with cardiovascular disease  receive low-dose aspirin and this carries an increased  risk of gastrointestinal events
In patients who are taking low-dose aspirin, the benefit of using  Cox-II-selective agents is reduced. Prescribing Cox-II-selective  agents preferentially over standard NSAIDs in this situation is  therefore not justified on current evidence

This guidance was based on an assessment report prepared by the NICE appraisal team,34 plus submissions from manufacturers, professional/specialist groups, external experts and patient groups/advocates. NICE concluded that Cox-II-selective inhibitors have equivalent efficacy to non-selective NSAIDs, and that there is evidence that all four drugs (celecoxib, rofecoxib, meloxicam and etodolac) are associated with fewer gastrointestinal adverse events than NSAIDs.34 From indirect comparisons, there was no evidence to suggest that any one drug was clinically superior to any other. However, such comparisons should be viewed with caution because of the heterogeneity between trials. Only celecoxib and rofecoxib have been investigated in large, long-term trials specifically designed to assess their effects on serious upper gastrointestinal complications. Small or short-term tolerability studies were used to estimate the effect of etodolac and meloxicam on such outcomes.

When the NICE guidance was published, few economic evaluations of Cox-II-selective inhibitors were available in the public domain. However, based on evidence submitted to the appraisal team, it was concluded that the use of such drugs was most justified in high-risk patients, as the absolute number of gastrointestinal events avoided would be greater in this population. (This was despite a lack of clinical evidence specifically in this population.) This view has been confirmed in a recently published Canadian Technology Report35 on the cost-effectiveness of rofecoxib and celecoxib in osteoarthritis and rheumatoid arthritis, based on the large clinical outcome trials VIGOR36 and CLASS.37 This report has some limitations with regard to UK practice, as the cost of drugs and other resources were based on Canadian estimates. In addition, the authors made several assumptions which could be questioned. These included: that the gastrointestinal toxicity of Cox-II-selective inhibitors at licensed doses would be the same as that seen with the higher doses used in the VIGOR and CLASS trials; and that no gastrointestinal benefit would be seen with Cox-II-selective inhibitors in patients who received low-dose aspirin (based on the CLASS trial). However, the report provided more specific detail than the NICE guidance about exactly which high-risk patients Cox-II-selective inhibitors are cost-effective for. It concluded that rofecoxib and celecoxib:35 (i) are not cost-effective in patients at average risk of upper gastrointestinal events or in a population with a typical mix of average- and high-risk patients; (ii) are cost-effective for patients who are at high risk because they have a history of upper gastrointestinal events; (iii) become less cost-effective in high-risk patients as the rate of co-prescription of proton pump inhibitors increases; the drugs could lose their cost-effectiveness advantage altogether if the price of proton pump inhibitors decreases (as a regular NSAID combined with a low-priced proton pump inhibitor could become cost-effective); (iv) become cost-effective for patients without additional risk factors over the age of 76 years for rofecoxib and 81 years for celecoxib. (However, age thresholds should be viewed with caution, as precise estimates for the relationship between age and upper gastrointestinal event rates were not available.)

Evidence for the gastrointestinal safety of meloxicam and etodolac

Two large, short-term, randomized controlled trials have compared the tolerability of meloxicam with that of diclofenac and piroxicam. In the MELISSA study, patients with osteoarthritis were randomized to receive either meloxicam, 7.5 mg (n = 4635), or diclofenac, 100 mg modified-release (n = 4688), once daily for 28 days.38 Significantly fewer gastrointestinal adverse events (i.e. dyspepsia, nausea and vomiting, abdominal pain and diarrhoea) occurred in the meloxicam group (13% vs. 19%; P < 0.001), and withdrawals due to adverse events were also lower in this group (5.5% vs. 8%; P < 0.001). It has been suggested that the differences in tolerability between meloxicam and diclofenac could have occurred because the doses used were not comparable — the efficacy ratings in the MELISSA study constantly favoured diclofenac.38, 39 There was no significant difference between the groups with regard to the occurrence of perforations, ulcers or bleeds (five patients with meloxicam and seven with diclofenac), but the study was only short term and was not powered to show this.38

In the SELECT study, osteoarthritis patients were randomized to receive either 7.5 mg meloxicam (n = 4320) or 20 mg piroxicam (n = 4336) once daily for 28 days.40 Gastrointestinal adverse events were significantly lower with meloxicam (10.3% vs. 15.4%; P < 0.001), but there was no significant difference in perforations, ulcers or bleeds (seven patients with meloxicam and 16 with piroxicam), although, again, the study was not powered to show this. Meloxicam was as effective as piroxicam in this study.40 A meta-analysis of six studies comparing meloxicam with non-selective NSAIDs (n = 19 331) suggested that the risk of perforations, ulcers or bleeds was reduced by about one-half in patients taking meloxicam (OR, 0.52; 95% CI, 0.28–0.96).41 However, most data were taken from the two short-term studies outlined above, in which low-dose meloxicam was used.

The largest trial providing gastrointestinal safety data for etodolac is a long-term efficacy and safety study in which 1446 patients with rheumatoid arthritis were randomized to receive etodolac, 150 mg twice daily, etodolac, 500 mg twice daily (higher than the licensed dose), or ibuprofen, 600 mg four times daily.42 The study enrolled patients for 4.5 years, but only about 50% of patients completed 1 year of follow-up, 30% completed 2 years and 20% completed 3 years. Adverse events were similar across all treatment groups, with gastrointestinal side-effects being the most common. The only statistically significant differences in side-effects were for dyspepsia and rash, which occurred less frequently with the lower dose of etodolac than with ibuprofen (dyspepsia incidence, 26% vs. 35%; P < 0.05). Over the 3 years, patients in both etodolac groups had a lower incidence of ulcers and bleeds than patients in the ibuprofen group — two patients in each etodolac arm compared with nine patients taking ibuprofen (P = 0.005). However, the study was not designed or powered to show this.

Nabumetone is another NSAID for which claims of a safer gastrointestinal profile have been made. However, as the active metabolite 6-methoxy-2-naphthylacetic acid (6-MNA) has Cox-I and (preferential) Cox-II activity, this drug was not included in the NICE appraisal. A meta-analysis of eight studies comparing nabumetone with other NSAIDs for up to 6 months found that perforations, ulcers or bleeds were significantly less common with nabumetone — three of 4847 patients compared with 24 of 2621 patients taking NSAIDs (P < 0.0001).43 However, this result should be interpreted with caution, as post-marketing and open-label studies were included in the meta-analysis and the results are based on very small numbers of events.

Evidence for the gastrointestinal safety of rofecoxib and celecoxib

Unlike the drugs discussed above, rofecoxib and celecoxib have been the subject of large randomized controlled trials investigating clinically significant upper gastrointestinal events. The ‘second-generation’ Cox-II-selective inhibitors parecoxib, etoricoxib and valdecoxib have not been assessed in such trials, and a discussion of these products is beyond the scope of this article.

The VIGOR trial. In the VIGOR trial, 8076 patients with rheumatoid arthritis were randomized to receive either 50 mg rofecoxib daily (twice the recommended dose for rheumatoid arthritis) or 500 mg naproxen twice daily.36 Both drugs had a similar efficacy, but over a 9-month period, rofecoxib was associated with fewer clinical gastrointestinal events (perforation, obstruction, bleeding or symptomatic ulcers) than naproxen (1.38% of patients in the rofecoxib group vs. 3.00% of the naproxen group; RR, 0.46; 95% CI, 0.34–0.63; NNT, 62). Complications (perforation, obstruction or bleeding) were also lower with rofecoxib (0.40% vs. 0.92%; RR, 0.43; 95% CI, 0.24–0.77; NNT, 191). The reduction in clinical gastrointestinal events with rofecoxib was similar in both high- and low-risk subgroups, but as would be expected, the NNTs were lower in high-risk groups.12 Patients taking low-dose aspirin were excluded from the VIGOR trial.36

The CLASS trial. In the CLASS trial, 7968 patients with osteoarthritis or rheumatoid arthritis (27% of the total) randomly received celecoxib, 400 mg twice daily (twice the recommended dose for rheumatoid arthritis, two to four times the recommended dose for osteoarthritis), ibuprofen, 800 mg three times daily, or diclofenac, 75 mg twice daily.37 The results of 6 months of treatment were presented, but these figures were then extrapolated to 12 months (i.e. annualized). Upper gastrointestinal complications (bleeding, perforation or obstruction) plus symptomatic ulcers were significantly lower in the celecoxib group compared with the combined NSAID group (annualized incidence, 2.08% vs. 3.54%; RR, 0.59; 95% CI, 0.38–0.94; NNT, 68). However, gastrointestinal complications alone (the primary end-point) were not significantly reduced with celecoxib (annualized incidence, 0.76% vs. 1.45%; RR, 0.53; 95% CI, 0.26–1.11).37

Unlike the VIGOR trial, patients in the CLASS trial were allowed to take aspirin for cardiovascular benefits, and approximately 20% took up to 325 mg/day. Subgroup analysis found that this had a dramatic effect on the results. Patients not taking aspirin had significantly fewer gastrointestinal complications with celecoxib than with NSAIDs (annualized incidence, 0.44% vs. 1.27%; RR, 0.35; 95% CI, 0.14–0.98; NNT, 120). However, in aspirin users, these benefits were completely negated (annualized incidence, 2.01% vs. 2.12%; P = 0.92).37

Recently, it has become apparent that the CLASS trial was, in fact, two trials, both of which continued for longer than the 6 months initially presented. One trial (vs. diclofenac) continued for 12 months, and the other (vs. ibuprofen) for 15 months.44 The longer term data suggest that, by 12 months, much of celecoxib's gastrointestinal benefits were lost, with similar levels of serious gastrointestinal complications seen in the celecoxib and NSAID groups (see Table 3).44, 45 Almost all the ulcer complications that occurred in the second half of the trials were in patients taking celecoxib, and there is a concern that Cox-II-selective inhibitors could interfere with ulcer healing45 (as the expression of Cox-II is increased in the margin of healing ulcers47). The manufacturers and authors of the CLASS trial have argued that the 6-month data were presented initially, as they were the most clinically and scientifically valid. After 6 months, confounding factors in the NSAID groups, such as a higher withdrawal rate (possibly resulting in fewer high-risk patients remaining) and decreasing complication rates, introduced bias.45, 48

Table 3.  Results of the CLASS trial based on actual number of events presented in the original paper 37 (6-month data) and the Food and Drug Administration (FDA) reports 46 (12- and 15-month data)
 6-month data from original paperEntire study period from FDA
reports (12–15 months)
Entire study period from FDA
reports (12–15 months)
(% of patients)
(% of patients)
(% of patients)
(% of patients)
(% of patients)
(% of patients)
  • GI, gastrointestinal; RRR, relative risk reduction.

  • a  

    This difference was significant in patients not taking aspirin.

  • b  

    Based on time-to-event analysis, as in the FDA report, this difference was significant ( P  = 0.04).

  • c  

    Based on time-to-event analysis, as in the FDA report, this difference was significant in patients not taking aspirin ( P  = 0.037).

Serious upper GI  complications plus  symptomatic ulcers0.751.231.081.301.081.81
RRR 39% (95% CI 4–61%)RRR not significantRRR 41% (95% CI 4–63%)
  Celecoxib vs. both NSAIDs: RRR not significantb
Serious upper GI  complications0.280.500.430.500.430.55
RRR not significantaRRR not significantRRR not significantc
  Celecoxib vs. both NSAIDs: RRR not significant

A systematic review has recently concluded that celecoxib significantly improves gastrointestinal safety when compared with non-selective NSAIDs.49 However, as five of the six trials included that assessed gastrointestinal safety were endoscopic studies and the sixth trial included the 6-month data from the CLASS trial, this review adds little to the celecoxib evidence base. Pooled endoscopic data from this review suggest that celecoxib does have benefits over NSAIDs in patients taking low-dose aspirin (ulcer rates of 12.0% with celecoxib vs. 26.4% with NSAIDs; RR, 0.49; 95% CI, 0.28–0.86). However, the absolute benefit of celecoxib is greater in patients not taking aspirin (ulcer rates of 6.3% with celecoxib vs. 24.9% with NSAIDs; RR, 0.27; 95% CI, 0.16–0.48).49 Also, these findings were based on 55 endoscopically detected ulcers in 290 individuals, and should be compared with the results of the CLASS trial, where celecoxib provided no benefit over NSAIDs in terms of serious upper gastrointestinal complications in 1645 aspirin users.37, 49

Are Cox-II-selective inhibitors superior to non-selective NSAIDs?

Gastrointestinal safety and tolerability.  On first analysis, the results from the VIGOR trial and 6 months of the CLASS trial seem positive. Compared with non-selective NSAIDs, celecoxib and rofecoxib are associated with about 50% fewer serious upper gastrointestinal complications and symptomatic ulcers in patients not taking concomitant low-dose aspirin. However, the longer term data from the CLASS trial have raised concerns and, in Europe, the Committee for Proprietary Medicinal Products has initiated an investigation into Cox-II-selective inhibitors, focusing particularly on gastrointestinal and cardiovascular safety (see next section).50

In the USA, the Food and Drug Administration (FDA) has reviewed the complete data from the CLASS trial and has concluded that they do not show a gastrointestinal safety advantage for celecoxib over ibuprofen or diclofenac. (However, the FDA acknowledged that the use of aspirin may have confounded some of the findings.) As a result, the US label for celecoxib continues to include the standard NSAID warning about the risk of ulcers, bleeds and perforations.51 (Based on the VIGOR trial, the standard gastrointestinal warning has been modified, but not removed, on the US rofecoxib label.52)

In terms of tolerability, it is important to note that gastrointestinal effects, such as dyspepsia and abdominal pain, are still the most common side-effects seen with Cox-II-selective inhibitors. In the VIGOR and CLASS trials, withdrawals due to gastrointestinal events were lower with rofecoxib and celecoxib compared with non-selective NSAIDs (3.5% vs. 4.9%36 and 8.7% vs. 10.7%,37 respectively). However, such events are still common. For example, 14.4% of patients taking celecoxib had dyspepsia in the CLASS study, compared with 16.1% of patients taking either ibuprofen or diclofenac (P < 0.05).37

Cardiovascular and renal safety.  Cox-II-selective inhibitors also cause the well-recognized NSAID-associated side-effects of fluid retention, oedema and hypertension. In the CLASS trial, celecoxib caused less hypertension than non-selective NSAIDs (1.7% vs. 2.3%; P < 0.05), but peripheral oedema rates were not significantly different (2.8% vs. 3.5%).37 Limited information about renal adverse effects was provided in the VIGOR trial, but the incidence was low and not statistically different between the groups (1.2% with rofecoxib and 0.9% with naproxen).36 However, fluid retention, oedema and hypertension appear to be dose-related with rofecoxib, and occur with increased frequency with chronic use and at higher doses.53 The US rofecoxib label states that, in trials of rheumatoid arthritis patients, rofecoxib, 25 mg/day, was associated with a higher incidence of hypertension than naproxen, 1000 mg/day.52

In a small (n = 810) 6-week trial of older osteoarthritis patients taking antihypertensive agents, both rofecoxib, 25 mg/day, and celecoxib, 200 mg/day, were shown to increase blood pressure, but the latter to a lesser extent.54 Rofecoxib significantly increased systolic blood pressure in 17% of patients, compared with 11% of those taking celecoxib (P = 0.032 for the difference), and oedema was also more common with rofecoxib (9.5% vs. 4.9%; P = 0.014). However, the doses used in this trial may not have been equivalent.

Particular concerns about the cardiovascular safety of Cox-II-selective inhibitors have arisen from the VIGOR trial, as myocardial infarctions were more common in the rofecoxib group than in the naproxen group (0.4% vs. 0.1%; 95% CI for the difference, 0.1–0.6%).36 Patients taking low-dose aspirin were excluded from this study, but according to FDA criteria, 4% of patients should have been taking it because they had a history of myocardial infarction, angina, stroke, etc. The results from the VIGOR trial are in contrast with those for celecoxib in the CLASS trial, in which 21% of patients were taking low-dose aspirin. In the CLASS trial, there was no significant difference between the celecoxib and NSAID groups in the incidence of major cardiovascular events.37, 47

The reason why more myocardial infarctions were seen with rofecoxib than with naproxen in the VIGOR trial is not fully understood, but two main theories have been put forward.47, 55 First, and of most concern, is the theory that Cox-II-selective inhibitors could have prothrombotic effects (a class effect, because they do not block thromboxane, which causes platelet aggregation, but selectively inhibit prostacyclin, which has anti-aggregatory and vasodilatory properties). This could have been masked in the CLASS trial, either because of the use of low-dose aspirin, or because this trial involved mainly osteoarthritis patients who are at a lower thrombotic risk than rheumatoid arthritis patients. Alternatively, celecoxib, 400 mg twice daily, might not be associated with the same cardiovascular risk as rofecoxib, 50 mg/day. A recent observational study, which should be interpreted with caution due to the inherent potential for bias, found that high-dose rofecoxib (> 25 mg/day) could be associated with an increased risk of coronary heart disease, whereas rofecoxib, < 25 mg/day, celecoxib, naproxen and ibuprofen are not.56 Second, naproxen could have provided cardiovascular benefits (antiplatelet effects) in the VIGOR trial, in a manner similar to that of low-dose aspirin. However, the suggestion that non-selective NSAIDs are cardioprotective is controversial.47


Rofecoxib and celecoxib are new ‘black triangle’ drugs and their pharmacology is not fully understood. Much more information is required about their potentially important cardiovascular, renal and gastrointestinal effects, and their interaction with aspirin. One of the main drawbacks with the evidence base for Cox-II-selective inhibitors is the lack of trial data comparing their gastrointestinal safety with that of standard, non-selective NSAIDs plus a gastroprotective agent, particularly in high-risk individuals. Minor gastrointestinal side-effects, such as dyspepsia and abdominal pain, are the most common side-effects seen with Cox-II-selective inhibitors. Therefore, tolerability is still likely to be an issue, as it is with NSAIDs. Until there is a better understanding of the risk–benefit profile of these new drugs, it seems sensible to be cautious when prescribing them, and not to use them routinely in large numbers of individuals. The NICE guidance on Cox-II-selective inhibitors is due for review in May 2004, and there are currently no plans to bring forward this date. Despite the inclusion of meloxicam and etodolac in the NICE guidance, their place in therapy is unclear. Their gastrointestinal safety has not been adequately assessed in clinical outcome trials. Therefore, the evidence for their use instead of non-selective NSAIDs, or instead of celecoxib or rofecoxib, is not robust.