Celecoxib Compared with Nonselective NSAIDs

Eleven of 12 randomized controlled trials of arthritis patients^22-30 found no significant difference in efficacy between celecoxib and an NSAID. The single study finding a difference was a randomized controlled trial of 249 randomized patients with severe osteoarthritis of the hip requiring joint replacement surgery. A significantly greater reduction in pain on walking was found for diclofenac 50 mg 3 times daily compared with celecoxib 200 mg once daily, as measured using an 100 mm visual analog scale, both in the primary 6-week assessment (difference, 12.1 mm; 95% CI, 5.8 to 18.4) and in the secondary 12-week assessment (difference 10.0 mm; 95% CI, 2.8 to 17.3) in the modified intention-to-treat population (N=235).²³ However, insufficient information was provided to determine if an adequate method was used to conceal the allocation sequence or whether the approach produced treatment groups that were comparable at baseline in terms of important prognostic factors. Baseline characteristics were only provided for the evaluable population (N=141), which only accounted for 60% of the modified intention-to-treat population (N=235). Consequently, this randomized controlled trial was rated poor quality and its results should be interpreted with caution.

The Agency for Healthcare Research and Quality Effective Health Care Program Comparative Effectiveness Review³¹ found no clear differences in efficacy between celecoxib and nonselective NSAIDs based on results from published trials^{22, 24, 26, 29} and meta-analyses^{32, 33} of published and unpublished trials. Celecoxib and nonselective NSAIDs were associated with similar pain reduction effects (Western Ontario and McMaster Universities Osteoarthritis Index, visual analogue scale, Patient Global Assessment) in published trials of patients with osteoarthritis, ^{22, 24, 26, 29} soft tissue pain, ^{34, 35} ankylosing spondylitis, ^36-38 or rheumatoid arthritis.^{29, 32, 39, 40} In the largest (13 274 patients) trial of patients with osteoarthritis of the hip, knee, or hand (SUCCESS-1), celecoxib 200-400 mg daily and diclofenac or naproxen were also associated with similar pain reduction effects (visual analogue scale, Western Ontario and McMaster Universities Osteoarthritis Index).²⁸

Celecoxib 200-400 mg was associated with slightly higher rate of withdrawals than other NSAIDs due to lack of efficacy (relative risk, 1.1; 95% CI, 1.02 to 1.23) in a recent meta-analysis based on analyses of company-held clinical trial reports from 31 primarily short-term trials.³³ This estimate of comparative efficacy may be the most precise available, but the validity of the findings cannot be verified as the data used in this analysis is not fully available to the public.³³ On the other hand, ibuprofen 2400 mg/day and diclofenac 150 mg/day were associated with higher rates of withdrawal due to lack of efficacy than celecoxib 800 mg/day after 52 weeks (14.8% compared with 12.6%; P=0.005) in the pivotal trial of patients with osteoarthritis or rheumatoid arthritis (Celecoxib Long-term Arthritis Safety Study [CLASS]).²⁷

Partially Selective NSAIDs Compared with Nonselective NSAIDs

Partially selective NSAIDs (meloxicam, nabumetone, and etodolac) were associated with similar pain reduction effects relative to nonselective NSAIDs in short-term randomized controlled trials. In double-blinded trials of meloxicam 7.5 mg, 15 mg, and 25 mg compared with other NSAIDs there were generally no differences in efficacy.^41-49 In 2 of the trials, however, patients taking nonselective NSAIDs were significantly less likely to withdraw due to lack of efficacy than patients taking meloxicam.^{44, 49} A systematic review of 3 short-term randomized controlled trials of nabumetone for soft tissue pain found no difference in efficacy when compared with ibuprofen or naproxen.⁵⁰ However, based on physician assessment, the same systematic review also found placebo to be as efficacious as nabumetone in reducing pain at 7 days. Etodolac and nonselective NSAIDs were generally associated with similar rates of withdrawals due to efficacy⁵¹ or improvements in pain⁵² in short-term randomized controlled trials of patients with osteoarthritis of the knee and/or hip. A sustained-release form of etodolac was also associated with similar rates of pain reduction relative to diclofenac in a small trial (N=64) of patients with osteoarthritis of the knee.⁵³

Comparisons Among Nonselective NSAIDs

Several recent good-quality systematic reviews by the Cochrane Collaboration found no clear differences among nonselective NSAIDs in efficacy for treating osteoarthritis of the knee, ⁵¹ hip, ⁵⁴ or low-back pain.⁵⁵ Results from 3 fair-quality randomized controlled trials published subsequent to the Cochrane reviews also consistently found no significant differences in efficacy among nonselective NSAIDs when used in patients with osteoarthritis.^56-58

Limited evidence from 2 trials found no difference in efficacy when salsalate 3 g daily was compared with indomethacin 75 mg daily⁵⁹ or diclofenac 75 mg daily.⁶⁰ No studies comparing salsalate to other NSAIDs were identified, and salsalate was not included in any of the systematic reviews included in this report.

Tenoxicam 20 mg and 40 mg, diclofenac, and indomethacin were associated with similar effects on pain in a good-quality systematic review of 18 randomized controlled trials.⁶¹ Tenoxicam was also associated with slightly greater improvements in pain management outcomes than piroxicam according to physician global assessment (odds ratio, 1.46; 95% CI, 1.08 to 2.03).

An older (1985) review of tiaprofenic acid 600 mg found no difference in efficacy when compared with aspirin 3600 mg, diclofenac 150 mg, ibuprofen 1200 mg, indomethacin 75-105 mg, naproxen 500 mg, piroxicam 20 mg, or sulindac 300 mg.⁶² A more recent randomized controlled trial confirmed the short-term comparative efficacy of tiaprofenic acid 600 mg and indomethacin 75 mg (at 4 wks, 43% and 45% of patients showed improvement respectively).⁶³ However, the same study found both drugs less efficacious in the long term (at 1 year, 39% for tiaprofenic acid and 36% for indomethacin).

Celecoxib Compared with Nonselective NSAIDs (with and without Antiulcer Medication)

Celecoxib is currently the only COX-2 inhibitor available in the United States. The Agency for Healthcare Research and Quality Effective Health Care Comparative Effectiveness Review is the most comprehensive review to date of the comparative safety of celecoxib relative to other NSAIDs, placebo, or nonuse.³¹ Conclusions of the review were based on numerous meta-analyses of primarily short-term randomized controlled trials (7 months or less)^{27, 32, 33, 79, 81-91} and population based observational studies.^92-102

With regard to upper gastrointestinal adverse events celecoxib seemed to offer a short-term advantage over nonselective NSAIDs, when neither were taken with antiulcer medication, but this has not been conclusively demonstrated in longer-term (>6 months) studies. CLASS remains the longest-term trial to date of patients with osteoarthritis/rheumatoid arthritis.²⁷ Results from an interim, 6-month analysis from the CLASS trial (32/3987 compared with 51/3981, annualized incidence rates 2.08% compared with 3.54%; P=0.02) ²⁷ and from meta-analyses of published and unpublished short-term trials^{33, 83} consistently suggested that celecoxib is associated with fewer serious gastrointestinal complications (bleeding, perforations, stricture) than nonselective NSAIDs. In a meta-analysis of 14 randomized controlled trials from 2000, annual rates of upper gastrointestinal ulcer complications were 2 per 1000 yearly for celecoxib and about 17 per 1000 yearly for NSAIDs (P=0.002).⁸³ Celecoxib was also associated with lower rates of clinical ulcers and bleeds relative to nonselective NSAIDs in a recent meta-analysis of data from Pfizer records of 18 primarily short-term randomized controlled trials.³³ Observational studies evaluating exposure to celecoxib of unknown¹⁰³ or short-term^{98, 104} duration are consistent with the randomized controlled trial results. Regarding longer-term gastrointestinal safety, however, celecoxib, diclofenac, and ibuprofen were associated with similar rates of complicated or symptomatic ulcers after 12 months in the CLASS trials, as reported by US Food and Drug Administration documents, ^{84, 90} and gastrointestinal safety outcomes associated with long-term use were not clearly reported in any observational study.

Additionally, 3 short-term randomized controlled trials found celecoxib was as effective as co-therapy with a nonselective NSAID and an antiulcer medication in preventing ulcer complications in high-risk patients.^105-107 In very high-risk patients with a recent gastrointestinal bleed, there were no statistically significant differences between either celecoxib 400 mg and diclofenac 150 mg plus omeprazole 20 mg¹⁰⁵ or celecoxib 200 mg and naproxen 750 mg plus lansoprazole 30 mg in recurrent ulcer bleeding after 6 months (mean rate: 4.3% for celecoxib compared with 6.3% for both diclofenac plus omeprazole and naproxen plus lansoprazole) or withdrawal rates due to adverse events (mean rate: 11.7% for celecoxib compared with 9.7% for diclofenac plus omeprazole and naproxen plus lansoprazole).^{105, 107} Likewise, in patients receiving aspirin (81 mg in 89% of the patients and 325 mg in 11% of patients ) and who required ongoing NSAID therapy for osteoarthritis (N=1045), rates of endoscopically confirmed gastroduodenal ulcers at 12 weeks were similar in patients given celecoxib 200 mg and those given naproxen 100 mg plus lansoprazole 30 mg (20.3% compared with 18.0%; difference 2.4%, 95% CI, –2.4% to 7.2%).¹⁰⁶

However, the most recent evidence suggested that the best protection of the upper gastrointestinal tract in higher-risk patients may come from taking celecoxib in combination with a proton pump inhibitor.^{108, 109} In a good-quality randomized controlled trial of very high risk patients with a recent gastrointestinal bleed (N=273), the 13-month cumulative incidence of recurrent ulcer bleeding was significantly lower for celecoxib 200 mg plus esomeprazole 20 mg (0%) compared with celecoxib 200 mg alone (8.9%; 95% CI, 4.1 to 13.7; P=0.0004), whereas there were no significant differences in withdrawals due to adverse events (6% compared with 7%) or in improvement in arthritis pain as measured using a 100 mm visual analog scale (–27% compared with –28%).¹⁰⁸ Additionally, in a subgroup analysis from a fair-quality, population-based retrospective cohort study in elderly patients which used data from the government of Quebec health services administrative databases, there were significantly fewer gastrointestinal hospitalizations when a proton pump inhibitor was added to celecoxib compared with celecoxib alone when age was above 75 years (adjusted hazard ratio, 0.56; 95% CI, 0.38 to 0.81), but not when age was 66 to 74 years (adjusted hazard ratio, 0.98; 95% CI, 0.63 to 1.52).¹⁰⁹

With regard to comparative risk of clinically significant adverse events throughout the gastrointestinal tract (upper and lower), a good-quality trial of 4484 patients with osteoarthritis and rheumatoid arthritis found a short-term advantage for celecoxib 400 mg/day over diclofenac slow release 150/day plus omeprazole 20 mg/day.¹¹⁰ At 6 months, significantly fewer patients receiving celecoxib met criteria for the composite primary endpoint of clinically significant event (gastroduodenal, small-bowel, or large-bowel hemorrhage; gastric-outlet obstruction; gastroduodenal, small-bowel, or large-bowel perforation; clinically significant anemia of defined gastrointestinal or presumed occult gastrointestinal origin; acute gastrointestinal hemorrhage of unknown origin) compared with those receiving diclofenac slow release plus omeprazole (0.9% compared with 3.8%; hazard ratio 4.3, 95% CI, 2.6 to 7.0). When the individual components of the composite outcome were evaluated separately, the difference was found to be primarily due to a significantly lower risk in the celecoxib group of having hemoglobin decrease of 20 g/L or more (0.7% compared with 3%, P value not reported). Among those, 0.4% of patients receiving celecoxib and 2% of patients receiving diclofenac slow release plus esomeprazole had hemoglobin decreases that were presumed to be of occult gastrointestinal origin, including possible blood loss from the small bowel. Because it was unclear whether the advantage for celecoxib would persist over the longer-term and because the difference was largely based on asymptomatic gastrointestinal disease characteristics, these findings should be interpreted with caution.

Based on findings from 3^{33, 111, 112} of 4^{33, 111-113} meta-analyses of randomized controlled trials that were primarily 12 weeks in duration, risk of myocardial infarction for celecoxib was not significantly different compared with NSAIDs (Table 5). Among the 3 meta-analyses that found no significant differences between celecoxib and NSAIDs, data were combined from up to 41 published and unpublished trials of primarily patients with osteoarthritis or rheumatoid arthritis and NSAID comparator groups consisting of diclofenac, naproxen, or ibuprofen.^{33, 111, 112} In contrast, the only meta-analysis of randomized controlled trials to find a significant increase in risk of myocardial infarction with celecoxib combined data from only 5 published trials of at least 6 weeks in duration in any population, including patients receiving celecoxib for colon polyp prevention and Alzheimer’s disease, and the pooled comparator group included placebo, diclofenac, ibuprofen, and paracetamol.¹¹³

Table 5

Risk of Myocardial Infarction: Celecoxib Compared with NSAID.

Risk of myocardial infarction was also assessed as an individual endpoint in 1 large case-control study of 54 475 patients 65 years of age or older, which also found no significant difference between celecoxib as compared with naproxen (adjusted odds ratio, 0.95; 95% CI, 0.74 to 1.21), ibuprofen (adjusted odds ratio, 0.98; 95% CI, 0.76 to 1.26), or other NSAIDs (adjusted odds ratio, 1.21; 95% CI, 0.82 to 1.10).¹⁰²

Additionally, no significant increase in risk of other cardiovascular events or cerebrovascular events was found for celecoxib as compared with nonselective NSAIDs in 6 meta-analyses of randomized controlled trials^{33, 112-116} and 5 observational studies.^{92, 95, 101, 117, 118}

With regard to cardiorenal harms, results from the longest-term CLASS trial¹¹⁹ and meta-analyses of shorter-term trials^{33, 112, 120} found no increased risk of hypertension or heart failure with celecoxib compared with nonselective NSAIDs. In CLASS, incidence of new-onset or aggravated hypertension was 2.7% for celecoxib, which was similar to the incidence with diclofenac (2.6%) and significantly lower than with ibuprofen (4.2%; P<0.05), whereas there was no significant difference between celecoxib, diclofenac, or ibuprofen in incidence of heart failure (0.3% compared with 0.2% or 0.5%).¹¹⁹ In the largest meta-analysis of primarily shorter-term published and unpublished trials, compared with various nonselective NSAIDs, celecoxib was associated with a significantly lower incidence of hypertension (1.5% compared with 2.0%; P=0.002) and a similar incidence of heart failure (0.1% compared with 0.2%; P=0.056).¹¹²

In a fair-quality, large-scale, population-based case-control study in which the Danish National Hospital Discharge Register was used to identify all subjects who sustained a fracture in the year 2000 (cases, N=124 655; controls, N=373 962), celecoxib was not associated with an increased risk regardless of dosage (adjusted odds ratios ranged from 0.84 to 0.97).¹²¹

Partially Selective NSAIDs

Among the partially selective NSAIDs (meloxicam, nabumetone, and etodolac), none were associated with any clear safety advantages relative to nonselective NSAIDs.

Nonselective NSAIDs (with and without Antiulcer Medications)

There was strong evidence from numerous randomized controlled trials^{122, 123} and observational studies^{103, 104, 133-135} that all nonselective NSAIDs are associated with relatively similar risks of serious gastrointestinal events relative to nonuse. Further study is needed to determine the potential comparative safety benefits of concomitant use of various gastroprotective agents in preventing clinical gastrointestinal events. Currently, misoprostol is the only gastroprotective agent proven to decrease risk of clinical gastrointestinal events (MUCOSA), but this was at the expense of significant increases in nausea, diarrhea, and abdominal pain.¹³⁶ Otherwise, misoprostol, double-dose H2 blockers, and proton pump inhibitors were all associated with significant reductions in risks of endoscopic gastric and duodenal ulcers when added to nonselective NSAIDs relative to nonselective NSAID use alone in short-term randomized controlled trials.^{137, 138}

Results from a fair-quality systematic review of 138 primarily short-term randomized controlled trials (≥ 4 weeks) suggested that nonselective NSAIDs other than naproxen are associated with similar risks of clinically important cardiovascular events (primarily myocardial infarction) compared with COX-2 inhibitors (data primarily on high-dose ibuprofen and diclofenac). On the other hand, naproxen 500 mg twice daily was associated with a lower risk of myocardial infarction compared with COX-2 inhibitors (relative risk, 2.04; 95% CI, 1.41 to 2.96; P=0.0002). In indirect analyses, naproxen was risk-neutral for cardiovascular events relative to placebo (relative risk, 0.92; 95% CI, 0.67 to 1.26), but other nonselective NSAIDs were associated with higher risks (rate ratio, 1.51; 95% CI, 0.96 to 2.37 and rate ratio, 1.63; 95% CI, 1.12 to 2.37 respectively for ibuprofen and diclofenac).⁷⁹ A recent, good-quality meta-analysis of observational studies found diclofenac associated with the highest risk, followed by indomethacin and meloxicam. Celecoxib, naproxen, piroxicam, and ibuprofen were not associated with increased risks. However, assessments of increased risk were modest (relative risk <2.0), and all of the main analyses were associated with substantial between-study heterogeneity.⁹⁹ Differences between the meta-analyses of randomized controlled trials and observational studies could be related to lower doses or patterns of use (such as intermittent use), differential use of co-medications, differences in populations, or other factors. For example, a meta-analysis of 11 observational studies found naproxen associated with a modest cardioprotective effect relative to other nonselective NSAIDs (odds ratio, 0.86; 95% CI, 0.75 to 0.99).¹³⁹ However, studies in this meta-analysis that were sponsored by the manufacturer of the competing drug rofecoxib found significantly greater cardioprotective effects compared with other studies. Findings from other observational studies suggested that naproxen is associated with similar^{92-94, 96} or lower ^140-143 cardiovascular risk relative to nonuse. However, protective cardiovascular effects of naproxen relative to nonuse observed in some observational studies usually appear explainable by issues related to study design or analysis.¹⁴⁴ More recent, high-quality observational studies are mostly consistent with a neutral cardiovascular effect of naproxen relative to nonuse.

Evidence of the comparative safety of nonselective NSAIDs regarding all-cause mortality, blood pressure, congestive heart failure, edema, renal function, hepatotoxicity, and fracture risk was limited, and no strong conclusions could be reached regarding differential safety. For hypertension outcomes, 2 meta-analyses of placebo-controlled trials suggested modestly differential effects for piroxicam, ibuprofen, indomethacin, and naproxen relative to other nonselective NSAIDs, though estimates for individual drugs were inconsistent between the 2 meta-analyses.^{5, 145} In addition, differential effects were not found in direct comparisons from a meta-analysis of head-to-head trials of these same nonselective NSAIDs.⁵ Publication bias was also an important concern because most trials did not report hypertension outcomes.

The only other limited evidence of differential safety pertained to hepatotoxicity. In 1 systematic review of published and unpublished short-term randomized controlled trials, diclofenac was associated with the highest rates of aminotransferase elevations >3 times the upper limit of normal (3.55%; 95% CI, 3.12 to 4.03) compared with ibuprofen (0.43%; 95% CI, 0.26 to 0.70) and naproxen (0.43%; 95% CI, 0.30 to 0.63).¹²³ No liver-related hospitalizations or deaths occurred with either diclofenac or ibuprofen and were very rare with naproxen (0.01%). Incidence of aminotransferase elevations >3 times the upper limit of normal (3.1%), liver-related hospitalizations (0.023%), and liver failure/transplant/death (0%) were similar in 17 289 patients with rheumatoid arthritis or osteoarthritis who received diclofenac 150 mg for a mean of 18 months during their enrollment in the Multinational Etoricoxib and Diclofenac Arthritis Long-Term Program (MEDAL), which was prospectively designed to pool data from 3 randomized controlled trials that compared diclofenac to etoricoxib.¹⁴⁶

Additionally, incidence of hepatic injury was 5-10 times higher for sulindac relative to other NSAIDs in a recent systematic review of 7 population-based epidemiological studies.¹⁴⁷ However, in all analyses the rates of hepatotoxicity were extremely low.

According to evidence from a fair-quality case-control study which used data from the Danish National Hospital Discharge Register, increased fracture risk was associated with recent use (within 1 year) of the majority of nonselective NSAIDs, except for diflunisal, sulindac, and tiaprofenic acid.¹²¹ With an adjusted relative risk of 1.76 (95% CI, 1.72 to 1.81), ibuprofen was distinguished as the nonselective NSAID that had the highest overall risk of fracture. An observed inverse dose-response relationship did not clearly suggest a direct correlation with the COX system. There was no control for potential over-the-counter use of NSAIDs in the control group and these findings have not yet been replicated in any other observational studies or randomized controlled trials.

Concomitant Anticoagulants

Randomized controlled trial evidence was limited to 1 small, fair-quality, crossover trial that evaluated how celecoxib and codeine compared in their potentiation of the anticoagulant effects of warfarin in 15 patients with osteoarthritis.¹⁶¹ Results from this trial found no significant changes in the mean international normalized ratio values after 5 weeks of either celecoxib or codeine therapy. Only 1 patient experienced an episode of excessive anticoagulation (international normalized ratio, 4.9), which occurred during treatment with celecoxib.

The only evidence regarding the comparative safety of nonselective NSAIDs relative to celecoxib or partially selective NSAIDs when used concomitantly with anticoagulants was available from 2 small observational studies and was inconclusive due to flaws in design.^{162, 163} Nonselective NSAIDs were associated with a risk of bleeding similar to celecoxib in 1 study, ¹⁶² but the risk was significantly greater than partially selective NSAIDs in another study¹⁶³ in patients using anticoagulants concomitantly.

Concomitant Aspirin

In patients receiving aspirin and who required ongoing NSAID therapy for osteoarthritis, we only found 1 trial (N=1045) designed to compare celecoxib alone to a nonselective NSAID plus a proton pump inhibitor to reduce endoscopic ulcer rates.¹⁰⁶ The daily dosage of aspirin was 81 mg in 89% of the patients and 325 mg in 11% of patients. After 12 weeks, the use of celecoxib 200 mg or naproxen 500 mg twice daily plus lansoprazole 30 mg daily resulted in similar rates of endoscopically confirmed gastroduodenal ulcers (20.3% compared with 18.0%; difference 2.4%, 95% CI, –2.4% to 7.2%). The only other evidence of the comparative safety of NSAIDs when used in combination with aspirin came from a systematic review that conducted subgroup analyses according to the use of low-dose aspirin (325 mg or less) and found that both celecoxib and nonselective NSAIDs were associated with significant increases in endoscopic ulcer rates.³³

In a 2003 fair-quality case-control study of patients with known cardiovascular disease, risk of overall mortality (adjusted hazard ratio, 1.93; 95% CI, 1.30 to 2.87) and cardiovascular mortality (adjusted hazard ratio, 1.73, 95% CI, 1.05 to 2.84) was significantly greater among users of ibuprofen plus aspirin (N=206) compared with users of aspirin alone (N=6285).¹⁶⁴ However, this study was small and did not control for potentially important confounders. Two subsequent fair-quality observational studies, using more broadly defined populations from the UK GPRD¹⁶⁵ and QRESEARCH⁹³ databases, found that the risk of myocardial infarction was not significantly different in users of aspirin, with or without NSAIDs.