Proton pump inhibitors (PPIs) are effective at reducing the risk of nonsteroidal anti-inflammatory drug (NSAID)–induced endoscopic gastric and duodenal ulcers, and are well tolerated. Standard doses of histamine₂-receptor antagonist (H₂RA) are not very effective in reducing the risk of NSAID-induced endoscopic ulcers. Misoprostol (800 micrograms daily) is the only prophylactic drug that has been shown to reduce the occurrence of clinically important ulcer complications, but its use is associated with significant adverse effects and poor compliance, particularly at high doses. CKS found several systematic reviews that investigated gastroprotection, but only three reviews were found to comprehensively pool data from included trials. The key findings from these three reviews are reported in this section. Trials included in the systematic reviews on gastroprotection were mainly of people with osteoarthritis or rheumatoid arthritis. More recent trials have investigated the effect of a coxib with a PPI, and the encouraging results have persuaded the National Institute for Health and Clinical Excellence (NICE) guidance development group to recommend this as an option in people at increased risk of gastrointestinal adverse effects.

A systematic review compared the cost-effectiveness of five strategies for the prevention of NSAID-induced gastrointestinal (GI) toxicity [Brown et al, 2006]:

• The Cochrane Library, Medline, Embase, Current Controlled Trials, and SIGLE were searched to May 2002; 118 randomized controlled trials (RCTs) met the inclusion criteria (n = 76,322).
• Although a large body of evidence compared coxib and standard NSAIDs, this was not matched by studies with gastroprotection or studies directly comparing gastroprotective strategies. This lack of direct comparisons led to the use of indirect comparisons of the relative efficacy of these strategies.
• Table 1 summarizes the key findings from combined direct and indirect evidence on the five gastroprotective strategies.

A Cochrane review investigated the effectiveness of common interventions for the prevention of NSAID-induced upper GI toxicity [Rostom et al, 2002]:

• RCTs of prostaglandin analogues, histamine₂–receptor antagonist (H₂RA), or PPIs for the prevention of chronic NSAID-induced upper GI toxicity were included. A total of 40 RCTs met the inclusion criteria (n = 17,510 completed the studies). Updated searches in August 2003 and August 2004 did not reveal any further new studies.
• Authors' conclusions:
• Misoprostol, PPIs, and double-dose H₂RAs are effective at preventing chronic NSAID-related endoscopic gastric and duodenal ulcers. Standard doses of H₂RA are not effective at reducing the risk of NSAID-induced ulcers.
• To date, only misoprostol has been evaluated in a true clinical outcome trial; it was found to directly reduce the risk of ulcer complications, such as perforation, haemorrhage, or obstruction, at a dose of 800 micrograms daily. At this high dose, misoprostol is associated with significant adverse effects; at lower doses, misoprostol is associated with diarrhoea and is less effective.
• Table 2 shows the key findings from the two systematic reviews.

A systematic review compared dyspepsia rates with a coxib versus either a standard NSAID or a standard NSAID plus a PPI in high-risk people with arthritis [Spiegel et al, 2006]:

• A structured search for RCTs on Medline and subspeciality journals was performed from January 1990 to March 2005, and bibliographies of key review articles were hand-searched. A total of 32 trials (n = 42,079) met the inclusion criteria.
• Meta-analyses found:
• Compared with standard NSAIDs, coxibs had a 12% relative risk reduction in dyspepsia rate (RR 0.88, 95% CI 0.85 to 0.91), representing an absolute risk reduction of 3.7 percentage points (26 studies).
• Compared with a standard NSAID alone, a standard NSAID plus PPI had a 66% relative risk reduction in dyspepsia rate (RR 0.34, 95% CI 0.22 to 0.54), representing an absolute risk reduction of 9 percentage points (four studies).
• Number needed to treat (NNT) to prevent dyspepsia (compared with a standard NSAID):
• Coxibs, NNT = 27.
• A standard NSAID plus a PPI, NNT = 11.
• These data were considered to provide the best indirect evidence that an NSAID plus a PPI appears to be superior to coxibs in minimizing incident dyspepsia for people with chronic arthritis, because direct head-to-head trials are lacking. Despite analyses relying on indirect comparison of these drugs through the common baseline of NSAIDs, the incidence of dyspeptic symptoms did not significantly differ between the NSAID arms from both groups of included studies.

Comparison of a non-selective NSAID plus PPI versus celecoxib, and of a non-selective NSAID plus PPI versus celecoxib plus PPI:

• An RCT compared celecoxib alone (400 mg/day) with a non-selective NSAID plus a gastroprotective drug (diclofenac 150 mg/day plus omeprazole 20 mg/day) in people (n = 287) who had had recent GI haemorrhage on NSAIDs [Chan et al, 2002]:
• The probability of recurrent bleeding was not significantly different between the two groups at 6 months: 4.9% (95% CI 3.1 to 6.7) and 6.4% (95% CI 4.3 to 8.4), respectively. The absolute difference between these probabilities is 1.5 percentage points (95% CI –6.8 to +3.8).

Comparison of celecoxib plus PPI versus celecoxib plus placebo:

• An RCT investigated the prevention of recurrent ulcer bleeding in 441 consecutively presenting people who were taking non-selective NSAIDs for arthritis after admission to hospital with upper GI bleeding [Chan et al, 2007]. People were given 200 mg celecoxib twice daily, and half were randomly assigned to receive 20 mg esomeprazole twice daily and 136 to receive a placebo for 12 months.
• Combination treatment was more effective than celecoxib alone for prevention of ulcer bleeding in people at high risk.
• The 13-month cumulative incidence of the primary endpoint was 0% in the combined treatment group and 12 (8.9%) in the control group (95% CI difference, 4.1 to 13.7; p = 0.0004). The median follow up was 13 months (range 0.4–13.0). Discontinuation of treatment and the incidence of adverse events were similar in the two treatment groups.
• People at very high risk for recurrent ulcer bleeding who need an NSAID were found to benefit from combination of a coxib and a PPI.

Comparison of celecoxib versus non-selective NSAID plus PPI:

• An RCT investigated the effects of celecoxib 200 mg twice a day compared with diclofenac 75 mg twice a day plus omeprazole 20 mg once a day on the small bowel, colon, stomach, and duodenum over 6 months [Chan et al, 2010]. This was a double blind RCT that included 4484 people over 60 years of age with osteoarthritis or rheumatoid arthritis or people 18 years of age and older with previous gastrointestinal ulceration or GI haemorrhage. The primary outcome was a composite of gastroduodenal, small bowel or large bowel haemorrhage or perforation; gastric outlet obstruction; clinically significant anaemia of defined GI or presumed occult GI origin; and acute GI haemorrhage of unknown origin. Clinically significant anaemia was defined as a decrease in haemoglobin of 2 g/dL or more, or a decrease in haematocrit of at least 10 percentage points. The main results were:
• Composite primary outcome — occurred in 3.8% of patients receiving diclofenac plus omeprazole and 0.9% of patients receiving celecoxib (HR 4.3, 95% CI 2.6 to 7.0; p < 0.0001, NNH approximately 34). However this was driven entirely by differences in rates of anaemia.
• Anaemia of defined GI origin — 0.22% for celecoxib compared with 1.07% for diclofenac plus omeprazole (RR 4.9, NNH 118).
• Anaemia of presumed occult GI origin — 0.45% for celecoxib compared with 2.36% for diclofenac plus omeprazole (RR 5.24, NNH 52).
• Rates of gastroduodenal and large bowel haemorrhage were identical in the two study groups (0.13% for celecoxib compared with 0.04% for diclofenac plus omeprazole).
• There were no episodes of gastroduodenal, small-bowel or large-bowel perforation; small-bowel haemorrhage; or gastric-outlet obstruction.

Comparison of a PPI plus either a non-selective NSAID or COX-2 inhibitor versus placebo plus either a non-selective NSAID or COX-2 inhibitor:

• A report on two identically designed RCTs that investigated the effect of esomeprazole 20 mg or 40 mg versus placebo in adults with osteoarthritis or rheumatoid arthritis currently using either COX-2 inhibitors or non-selective NSAIDs over 26 weeks [Scheiman et al, 2006]. The two RCTs were the Verification of Esomeprazole for NSAID Ulcers and Symptoms (VENUS) trial (n = 844), and the Prevention of Latent Ulceration Treatment Options (PLUTO) trial (n = 585). Esomeprazole reduced the occurrence of both types of ulcer and upper GI adverse events over 6 months in participants receiving either COX-2 inhibitors or non-selective NSAIDs compared with users of anti-inflammatory drugs who received placebo instead of a PPI. From pooled data, outcomes reported included:
• The occurrence of gastric or duodenal ulcers over 6 months in people who received:
• COX-2 inhibitors and placebo: 16.5% (95% CI 9.7 to 23.4).
• Non-selective NSAID and placebo: 17.1% (95% CI 12.6 to 21.6).
• COX-2 inhibitors and esomeprazole 20 mg: 0.9% (95% CI 0 to 2.6). This was significantly (p < 0.001) better than placebo.
• Non-selective NSAID and esomeprazole 20 mg: 6.8% (95% CI 3.9 to 9.7). This was significantly (p < 0.001) better than placebo.
• COX-2 inhibitors and esomeprazole 40 mg: 4.1% (95% CI 0.6 to 7.6). This was significantly (p < 0.002) better than placebo.
• Non-selective NSAID and esomeprazole 40 mg: 4.8% (95% CI 2.3 to 7.2). This was significantly (p < 0.001) better than placebo.
• The occurrence of withdrawals due to adverse effects:
• Placebo group: 45 (16.7%) in VENUS; 24 (13%) in PLUTO.
• Esomeprazole 20 mg group: 22 (8.1%) in VENUS; 11 (5.7%) in PLUTO.
• Esomeprazole 40 mg group: 35 (12.7%) in VENUS; 11 (5.6%) in PLUTO.