Clinical use and pharmacological properties of selective COX-2 inhibitors
- First Online:
- Cite this article as:
- Shi, S. & Klotz, U. Eur J Clin Pharmacol (2008) 64: 233. doi:10.1007/s00228-007-0400-7
Selective COX-2 inhibitors (coxibs) are approved for the relief of acute pain and symptoms of chronic inflammatory conditions such as osteoarthritis (OA) and rheumatoid arthritis (RA). They have similar pharmacological properties but a slightly improved gastrointestinal (GI) safety profile if compared to traditional nonsteroidal anti-inflammatory drugs (tNSAIDs). However, long-term use of coxibs can be associated with an increased risk for cardiovascular (CV) adverse events (AEs). For this reason, two coxibs were withdrawn from the market. Currently celecoxib, etoricoxib, and lumiracoxib are used. These three coxibs differ in their chemical structure and selectivity for COX-2, which might explain some of their pharmacological features. Following oral administration, the less lipophilic celecoxib has a lower bioavailability (20–40%) than the other two coxibs (74–100%). All are eliminated by hepatic metabolism involving mainly CYP2C9 (celecoxib, lumiracoxib) and CYP3A4 (etoricoxib). Elimination half-life varies from 5 to 8 h (lumiracoxib), 11 to 16 h (celecoxib) and 19 to 32 h (etoricoxib). In patients with liver disease, plasma levels of celecoxib and etoricoxib are increased about two-fold. Clinical efficacies of the coxibs are comparable to tNSAIDs. There is an ongoing discussion about whether the slightly better GI tolerability (which is lost if acetylsalicylic acid is coadministered) of the coxibs is offset by their elevated risks for CV AEs (also seen with tNSAIDs other than naproxen), which apparently increase with dose and duration of exposure. In addition, the higher costs for coxibs (if compared to tNSAIDs, even when a “gastroprotective” proton pump inhibitor is coadministered) should be taken into consideration, if a coxib will be selected for certain patients with a high risk for GI complications. For such treatment, the lowest effective dose should be used for a limited time. Monitoring of kidney function and blood pressure appears advisable. It is hoped that further controlled studies can better define the therapeutic place of the coxibs.
KeywordsCOX-2 inhibitorsPharmacokineticsPharmacodynamicsClinical efficacySafety
Classification and basic differences of coxibs
Selectivity  (COX-2/COX-1)
Withdrawn in September 2004
Withdrawn in April 2005
Available (not approved by the FDA*)
Phenylacetic acid derivative
Available (withdrawn from the Canadian and Australian market; not approved by the FDA*)
Recently, concern regarding the cardiovascular (CV) safety of coxibs has been raised, particularly the increased risk of thrombotic events [e.g., myocardial infarction (MI), unstable angina, sudden and unexplained death, and stroke] [8–12]. Because of this increased risk, the manufacturers of rofecoxib and valdecoxib withdrew their products. Today, epidemiological data suggest that coxibs and tNSAIDs as a class all carry some variable potential CV risks, particularly when taken at high doses for prolonged periods of time. Therefore, the risks (CV) and benefits (GI) of coxibs must be carefully weighed before making therapeutic decisions, taking co-morbidity and comedicaton of the individual patient into account also.
As efficacy and safety of coxibs might be linked to their specific pharmacokinetic and pharmacodynamic properties , we have also summarized the current knowledge in these areas.
Clinical pharmacokinetics of coxibs
Pharmacokinetic characteristics of coxibs
Mainly by CYP2C9
Mainly by CYP3A4
Mainly by CYP2C9
Excretion of metabolites
Renal (27%) + fecal (58%)
Renal (70%) + fecal (20%)
Renal (54%) + fecal (43%)
Due to the poor aqueous solubility of celecoxib, its oral bioavailability is low (20–40%) [17, 18]. When celecoxib is taken with a high-fat meal, Cmax levels are delayed for approximately 1–2 h. The high-fat meal has a small effect on the extent of absorption of celecoxib, with an increase in area under the plasma concentration-time curve (AUC) and Cmax of 10.7 and 25%, respectively, which will hardly affect its safety or efficacy. Hence, it can be given with or without food . Within the therapeutic dose range, both AUC and Cmax of celecoxib increase in proportion to the dose (linear pharmacokinetics). The large apparent volume of distribution indicates extensive distribution. Unlike other coxibs that distribute almost equally throughout the body, celecoxib might be sequestered in body fat because of its extremely high lipophilicity . The milk-to-plasma concentration ratio of celecoxib is very low (approximately 0.18), suggesting that breastfeeding would pose a minimal risk [20, 21].
Celecoxib is extensively metabolized, primarily by CYP2C9 (80%); CYP3A4 has a minor role [22, 23]. The metabolism involves hydroxylation at the methyl moiety followed by further oxidation of the hydroxyl group to form a carboxylic acid, which is the major metabolite of celecoxib. In healthy adults, the elimination half-life (t1/2) is approximately 12 h and the apparent plasma clearance (CL/F) averages 450 ml/min [17, 24].
The impact of CYP2C9 genotype on the elimination of celecoxib is of some importance [25–28]. The heterozygous CYP2C9*1/*2 genotype has a minimal effect on celecoxib’s pharmacokinetics [22, 29, 30]. In two studies, the AUC in homozygous CYP2C9*3/*3 mutants was approximately two-fold higher (versus wildtype CYP2C9*1/*1 subjects) [22, 29]. Recently, it was reported that the AUC of celecoxib is approximately 10-fold higher in pediatric patients genotyped CYP2C9*3/*3 (versus two CYP2C9*1/*1 patients) . This fits with a recent observation in which a seven-fold higher AUC in three CYP2C9*3/*3 subjects was reported . The CYP2C9*3/*3 variant has a decreased metabolic capacity indicating that carriers of the CYP2C9*3 allele have an impaired hepatic elimination (clearance) of celecoxib . As a result, homozygous carriers of the CYP2C9*3 allele, representing only approximately 0.5% of Caucasians, will have a greatly increased exposure to celecoxib. This can result in an elevated risk for concentration-dependent side effects as exemplified by a significantly higher risk of NSAID-induced bleeding in this subgroup .
Compared to a 400-mg oral adult dose of celecoxib b.i.d., eight children (6.9–16.2 years) were dosed with 250 mg/m2. After the single dose, comparable Cmax levels were observed (1.2 versus 1.4 μg/ml) at about the same time (approximately 3 h), suggesting similarity in the rate and extent of absorption. However, when CL/F and t1/2 were compared, striking differences emerged: celecoxib was cleared approximately twice as fast in children (CL/F: 1.4 ± 1.0 l h−1 kg−1) and t1/2 (3.7 ± 1.1 h) was approximately half as long . Consequently, children have a drug exposure that is approximately 50% lower, which probably is due to their higher metabolic activity/capacity . Furthermore, the effect of food on the pharmacokinetics in children differed from that in adults. High–fat food can result in a significant increase in AUC and Cmax, both after a single dose (60% increase for AUC and 82% for Cmax) and at steady state (increase of 75% for AUC and 99% for Cmax). In addition, high-fat food tended to increase trough plasma levels so that patients spent a smaller proportion of the dosing interval below the targeted concentration .
Apparently, celecoxib does not interact with warfarin , methotrexate , or ketoconazole . Co-administration of the CYP2C9 inhibitor fluconazole resulted in a 134% increase of AUC of celecoxib . Celecoxib may diminish the antihypertensive effect of angiotensin-converting enzyme inhibitors, can reduce the natriuretic effect of diuretics, and will increase (56–99%) lithium levels . Celecoxib inhibited the metabolism of metoprolol with a 64% increase in its AUC . Furthermore, rifampicin increased the clearance of celecoxib by 185% with a decrease in AUC by 64% .
At steady state, elderly subjects (over 65 years old) had a 40% increase in Cmax and a 50% increase in AUC compared to young subjects. AUC of celecoxib was approximately 43% lower in patients with chronic renal insufficiency (glomerular filtration rate 35–60 ml/min), which can be explained by a 47% increase in CL/F. In patients with mild to moderate hepatic impairment, plasma concentrations were increased by 40–180% [37, 41].
The pharmacokinetics of etoricoxib is linear over the clinically relevant dose range [42, 43]. Etoricoxib is rapidly absorbed and the oral bioavailability approaches 100% . A high-fat meal delayed the rate (increase of tmax by 2 h) and slightly affected the extent (36% decrease in Cmax) of absorption of etoricoxib. These differences are not considered clinically significant. Steady-state conditions are achieved within 7 days of once-daily dosing with an accumulation ratio of approximately 2.0. Since t1/2 is approximately 25 h, once-daily dosing can be accomplished .
Etoricoxib is metabolized extensively, and the metabolites are largely excreted into the urine (70%) . Etoricoxib is metabolized via 6′-methyl-hydroxylation (major pathway) and 1′-N-oxidation, which is catalyzed by multiple P450 forms. CYP3A4 accounts for the majority of the activity (approximately 60%); the remaining part is divided over several CYPs such as CYP2C9, CYP2D6, CYP1A2, and possibly CYP2C19 . So, the metabolic pattern of etoricoxib differs from that of the other two coxibs that are primarily (≥80%) metabolized by a single CYP (CYP2C9) .
In healthy subjects, co-administration of ketoconazole (400 mg) resulted in a 43% increase in the AUC of etoricoxib, and rifampicin decreased AUC by 65% . Both interactions could be of clinical significance. Etoricoxib had no effect on the pharmacokinetics of S(−) warfarin but slightly increased the AUC of R(+) warfarin (approximately 10%), which is not clinically relevant. However, because both drugs can cause bleeding complications, caution should be exercised when used concomitantly .
The pharmacokinetics of etoricoxib was also studied in patients with renal or hepatic impairment [48, 49]. It was demonstrated that even severe renal impairment (creatinine clearance below 30 ml min−1 per 1.73 m2) had little effect on etoricoxib’s pharmacokinetics . In contrast, a trend of decreasing CL/F with increasing hepatic impairment was observed. In patients with mild hepatic insufficiency (Child-Pugh score 5–6), the AUC of etoricoxib was increased slightly (approximately 16%). Patients with moderate hepatic insufficiency (Child-Pugh score 7–9) receiving 60 mg etoricoxib every other day had similar AUC to healthy subjects administered etoricoxib 60 mg once daily .
In contrast to other coxibs, lumiracoxib possesses a carboxylic acid group that makes it weakly acidic. Its structure is almost identical to diclofenac. Lumiracoxib is characterized by a relatively slow absorption and is subject to a modest first-pass effect . Lumiracoxib has demonstrated linear and time-independent pharmacokinetics in healthy subjects and patients with OA [51, 52]. The oral bioavailability of lumiracoxib is high (approximately 74–90%), and it has a t1/2 of about 5–8 h. Because of its acid nature it is distributed to inflamed tissues and synovial fluid and can be retained there for up to 24 h . Lumiracoxib protein binding is similar in plasma and synovial fluid (approximately 98%). Lumiracoxib has a distribution volume of only approximately 13 l, which is almost identical to that of the structurally very similar diclofenac .
The steady-state pharmacokinetics of lumiracoxib (400 mg once daily for 1 week) was studied in plasma and synovial fluid of patients with RA . The peak concentrations in synovial fluid occurred 3–4 h later than in plasma and exceeded those in plasma starting 5 h postdosing. Substantially higher concentrations remained in synovial fluid until the end of the monitored 28-h period (AUC12–24 h in synovial fluid was 2.6 times higher than that in plasma). This unique property suggests that lumiracoxib, while having reduced systemic exposure, can still persist at sites for drug action [54, 55]. Consequently, the therapeutic effect of lumiracoxib will be longer than would have been expected from its plasma pharmacokinetics.
Lumiracoxib is metabolized extensively via oxidative metabolism of the 5-methyl group and by hydroxylation of its dihaloaromatic ring. Major metabolites of lumiracoxib in plasma are the 5-carboxy, 4′-hydroxy, and 4′-hydroxy-5-carboxy derivatives, of which only the 4′-hydroxy derivative is active and COX-2 selective. The CL/F of lumiracoxib averaged 140 ml/min .
Lumiracoxib does not exhibit any meaningful interaction with a range of commonly used drugs including aspirin , fluconazole , omeprazole , aluminium hydroxide/magnesium hydroxide , methotrexate , and warfarin . Thus, dose adjustments are not required when co-administering these agents with lumiracoxib [54, 60]. Mild and moderate hepatic or renal impairment does not appear to influence lumiracoxib exposure [54, 61].
Pharmacodynamic profile of coxibs
According to the COX-2 hypothesis, inflammatory prostaglandins are primarily derived from COX-2, while prostaglandins formed by COX-1 have in general a more homeostatic role. Theoretically, selective inhibition of COX-2 would provide anti-inflammatory effects without disrupting gastric cytoprotection and platelet function. Thus, the hypothesis that selective inhibition of COX-2 will have therapeutic actions similar to tNSAIDs without the GI side effects that are attributed to inhibition of COX-1 was the rationale for the development of coxibs .
TXA2 synthesized by COX-1 in platelets promotes vasoconstriction, smooth muscle proliferation, and platelet aggregation. Increased TXA2 production was reported in patients who had increased CV risks due to coxibs [69, 70]. In contrast, prostacyclin (PGI2), a product of AA from COX-2 in vessel walls, plays an important role in the homeostatic defense mechanism that promotes vasodilatation and inhibition of platelet aggregation . tNSAIDs block both COX isoforms and therefore inhibit TXA2 and PGI2 production to a similar extent, although there is some variation in the biochemical selectivity for each isoenzyme among different tNSAIDs . In contrast, coxibs exhibit differing inhibitory effects on PGI2 production. Thus, coxibs may create an imbalance between TXA2 and PGI2 and shift away from the protective effects attributable to PGI2. This might be the dominant mechanism that can lead to an increased risk of thrombosis [71, 72].
Coxibs are indicated in acute pain states caused, e.g., by dental surgery, orthopedic surgery, primary dysmenorrhea, and in patients with OA or RA. In addition, etoricoxib has been approved for chronic low back pain (CLBP; see below) and acute gouty arthritis (see below). Celecoxib is the only coxib approved for the reduction of colon and rectum polyps in patients with familial adenomatous polyposis (FAP; see below).
The efficacy of celecoxib (400 mg/day) for the treatment of acute pain (acute ankle sprain and acute shoulder pain) was evaluated in two double-blind, randomized controlled trials (RCTs) using naproxen (1,000 mg/day) as comparator [75, 76]. Primary efficacy assessment included Visual Analogue Scales (VAS) and/or Patient’s Global Assessment. Secondary efficacy measures included Physician’s Global Assessment, Patient’s Return to Normal Function/Activity, Patients’ and Physicians’ Satisfaction Assessments, or pain and functional assessments. It was found that celecoxib was as effective as naproxen in managing acute pain.
Two RCTs with 550 patients compared celecoxib (200–400 mg/day) and ibuprofen (600–2,400 mg/day) for the treatment of acute ankle sprain. Both drugs were equally efficacious and more effective than placebo (P < 0.05) [77, 78]. Celecoxib (400 mg) seemed to be more effective than ketoprofen (200 mg) for acute pain management in patients with tonsillectomy after their discharge .
Three studies indicated that perioperative administration of celecoxib (400 mg) could improve analgesia and decrease opioid use [80–82]. A recent review revealed an overall reduction in postoperative opioid use and significant patient satisfaction with perioperatively administered coxibs .
Four RCTs addressed the efficacy of etoricoxib for the treatment of acute dental pain (n = 1,126) [84–87]. Daily doses of etoricoxib (120 mg) were compared to ibuprofen (400 mg), oxycodone/acetaminophen (10 mg/650 mg), codeine/acetaminophen (60 mg/600 mg), or naproxen (550 mg). The primary endpoint was total pain relief over 6 or 8 h. All active treatments had significantly greater analgesic effects compared to placebo (P < 0.001). However, etoricoxib demonstrated significantly greater overall analgesic efficacy versus oxycodone/acetaminophen (P < 0.001) and codeine/acetaminophen (P < 0.001) [85–87]. Etoricoxib and naproxen had similar analgesic efficacy . Single oral doses of etoricoxib 120 and 180 mg were superior to etoricoxib 60 mg (P ≤ 0.001) and ibuprofen 400 mg (P < 0.05). A linear relationship between plasma concentrations of etoricoxib and pain-relief scores up to the maximum of the plasma levels was observed . Comparable results were seen in two other trials evaluating the analgesic efficacy of etoricoxib and naproxen in 228 patients after knee or hip replacement surgery and in 73 women with primary dysmenorrhea. Etoricoxib 120 mg provided rapid and sustained analgesia similar to that of naproxen 550 or 1,100 mg [88, 89].
In addition, three RCTs (n = 1,090; duration 4–12 weeks) evaluated the efficacy of etoricoxib for CLBP [90–92]. Primary endpoint included a low back pain VAS score. Etoricoxib 60 or 90 mg daily was superior to placebo (P < 0.001); this result was observed as early as 1 week after initiating therapy, was maximal at 4 weeks, and was maintained over 3 months.
In a double-blind multicenter study in 180 patients with moderate to severe pain, lumiracoxib (400 mg daily) and naproxen (500 mg b.i.d.) were shown to be of similar efficacy in reducing pain and more effective than placebo (P < 0.05) within 48 h of unilateral total knee or total hip arthroplasty . In 557 patients with postoperative dental pain, the efficacy of lumiracoib with ibuprofen (400 mg), rofecoxib (50 mg), and celecoxib (200 mg) was studied. Lumiracoxib provided pain relief comparable to ibuprofen and rofecoxib, but was superior to celecoxib and placebo (P < 0.05) [94, 95].
Two randomized, multicenter studies (n = 84; n = 99) evaluated the efficacy of lumiracoxib in the treatment of primary dysmenorrhea . Patients received lumiracoxib 400 mg daily, rofecoxib 50 mg daily, or naproxen 500 mg b.i.d. All active treatments were significantly (P < 0.001) superior to placebo in reducing pain intensity. The analgesic efficacy of lumiracoxib was similar to naproxen and rofecoxib.
Seventeen RCTs compared the efficacy of celecoxib (100–400 mg daily) with placebo for the treatment of OA of knee or hip [97–112]. Main outcome measures included pain intensity in the target knee or hip, patient’s assessment of disease activity, Western Ontario McMaster’s Osteoarthritis Index (WOMAC) total score, Outcomes Measures in Rheumatology Clinical Trials-Osteoarthritis Research Society International responder rates and WOMAC subscale scores. In all studies, celecoxib was significantly superior to placebo (P < 0.05).
Five RCTs (n = 16,029; duration up to 13 weeks) evaluated the efficacy of celecoxib with tNSAIDs for the treatment of OA of knee or hip [97–99, 113, 114]. B.i.d. doses of celecoxib of 50, 100, and 200 mg were compared to naproxen 500 mg b.i.d., diclofenac 50 mg b.i.d. or t.i.d., or dexibuprofen 400 mg b.i.d. Efficacy measures were the same as above, and celecoxib was as effective as naproxen, diclofenac, and dexibuprofen.
In addition, in four double-blind RCTs (n = 3,040, duration up to 12 weeks), celecoxib 200 mg daily was compared to acetaminophen (paracetamol) 4,000 mg daily, which was recommended as the initial treatment for knee or hip OA [103, 115, 116]. Patients were evaluated by Patient Global Assessment of Response to Therapy, WOMAC index, and VAS. In three studies, celecoxib provided superior efficacy to acetaminophen [103, 116]. In one study, celecoxib was not significantly more effective than acetaminophen .
Recently, a multicenter pilot study compared continuous and intermittent treatment with celecoxib 200 mg daily in patients with OA of knee or hip. The percentage of days with intake of the flare drug was significantly lower (P = 0.031) in the group receiving continuous versus intermittent celecoxib .
Four double-blind RCTs (n = 2,162; duration 6–52 weeks) evaluated the efficacy of etoricoxib for the treatment of OA of knee or hip [118–121]. In two studies [118, 119], the dose of etoricoxib was 60 mg (n = 1,017), one study  used 30 mg once daily (n = 528), another study  used doses up to 90 mg daily (n = 617). Naproxen 500 mg b.i.d., diclofenac 50 mg t.i.d., or ibuprofen 800 mg t.i.d. were applied as comparators. Etoricoxib was comparable to all three tNSAIDs, and it was significantly more efficacious than placebo (P < 0.05).
Two multicenter RCTs indicated that etoricoxib 30 mg once daily was at least as effective as celecoxib 200 mg once daily in the treatment of knee and hip OA; both coxibs were superior to placebo . More recently, the long-term efficacy (138 weeks) of etoricoxib for the treatment of OA was evaluated by a randomized, double-blind, parallel group study using naproxen as active comparator. In 997 patients with OA of hip or knee, daily doses of etoricoxib 50 mg were compared to naproxen 500 mg b.i.d. Both etoricoxib and naproxen demonstrated long-term efficacy based on WOMAC and VAS scores .
Six RCTs evaluated lumiracoxib at a dose of 50–400 mg daily for the treatment of OA of knee, hip or hand [102, 105−107, 124, 125]. All lumiracoxib doses provided significant improvement in all outcome measures compared to placebo (P < 0.05). Pain relief was similar for all lumiracoxib doses and to diclofenac (75 mg b.i.d.). Four 13-week, multicenter studies (n = 6,545) demonstrated similar efficacy between lumiracoxib 100–400 mg daily and celecoxib 200 mg daily [102, 105–107]. In a 7-day, double-blind study, lumiracoxib 400 mg was compared with celecoxib 200 mg b.i.d. in 364 patients with knee OA. Lumiracoxib provided similar analgesic effects as celecoxib 3–5 h after the first dose (P = 0.185). However, at the study end, 13.9% of lumiracoxib-treated patients experienced complete pain relief versus 5.5% of celecoxib recipients .
Rheumatoid arthritis (RA)
The treatment with celecoxib in OA and RA trials differed mainly in respect to the applied doses. In 1,149 patients with RA, b.i.d. doses of celecoxib 100, 200, and 400 mg were tested versus naproxen 500 mg b.i.d. Main outcome measures included the American College of Rheumatology (ACR) responder index, the ACR-20. Significantly greater improvement in sign and symptoms of RA was noted for all dosages of celecoxib and the naproxen group if compared to placebo . In two other RCTs (n = 1,116; duration 6 months), celecoxib 200 mg q.d. to b.i.d. was evaluated in comparison with daily doses of diclofenac 75–100 mg, diclofenac slow release 150 mg, meloxicam 15 mg, or nabumetone 1,000 mg for the long-term treatment for RA. Celecoxib showed sustained anti-inflammatory and analgesic activity similar to the other four drugs [127, 128].
In 1,987 patients with RA the efficacy of etoricoxib (90 mg/day) versus naproxen (500 mg b.i.d.) was compared in two 12-week RCTs [129, 130]. Primary efficacy measures included direct assessment of arthritis by counts of tender and swollen joints, and patient and investigator global assessments of disease activity. In both studies, etoricoxib showed significant improvements in all efficacy endpoints in relation to placebo (P < 0.05). One study indicated that etoricoxib was more effective than naproxen ; in the other study, etoricoxib was similar in efficacy to naproxen .
A 26-week RCT using naproxen (500 mg b.i.d.) as comparator evaluated the efficacy of 200–400 mg lumiracoxib in 1,124 patients with RA . The primary efficacy variable was response to treatment according to ACR-20 criteria at week 13. Lumiracoxib was superior to placebo (P < 0.05), but there was no significant difference between lumiracoxib and naproxen, although lumiracoxib was numerically superior at all doses.
Over-expression of COX-2 has been found in a number of cancers such as colorectal adenocarcinomas, gastric, esophageal, lung, breast, prostate, and pancreatic cancers. COX-2 appears to control many cellular processes [132–138] that have been associated with a more aggressive behavior and a poor prognosis of malignancies [135–139]. Because of its role in carcinogenesis, apoptosis, and angiogenesis, COX-2 has become an appealing target for prevention of human cancers.
Celecoxib has been evaluated in patients with FAP, which represents a precancerous condition. In patients with FAP, celecoxib 400 mg b.i.d. for 6 months decreased the size and number of colorectal and duodenal polyps by 28% (P = 0.003 for comparison with placebo) leading to approval for the treatment of this condition [140, 141]. Currently, celecoxib is the only approved coxib for the treatment of patients with FAP.
Several clinical trials have shown the effectiveness of celecoxib in the chemoprevention of colorectal, lung, and breast cancer. The Adenoma Prevention with Celecoxib (APC) and the Prevention of Spontaneous Adenomatous Polyps (PreSAP) trials were randomized, placebo-controlled trials evaluating for 3 years the efficacy of celecoxib for prevention of colorectal adenomas [142, 143]. The cumulative incidence of adenomas detected through year 3 of the APC trial was 43.2% for patients taking celecoxib 200 mg b.i.d. [relative risk (RR) 0.67, 95% CI 0.59–0.77, P < 0.001] and 37.5% for those receiving 400 mg b.i.d. (RR 0.55, 95% CI 0.48–0.64, P < 0.001) compared to 60.7% for those receiving placebo . In the PreSAP trial, the cumulative rate of adenomas was 33.6% in the celecoxib group and 49.3% in the placebo group (RR 0.64, 95% CI 0.56–0.75, P < 0.001). The cumulative rate of advanced adenomas was 5.3% in the celecoxib group and 10.4% in the placebo group (RR 0.49, 95% CI 0.33–0.73, P < 0.001) . Comparable results were calculated in a meta-analysis (RR 0.72, 95% CI 0.68–0.77) . Meanwhile, all these studies found significant increased CV events with celecoxib [142–144]. Thus, the balance of benefits to risks does not favor chemoprevention with coxibs in the general population.
In a recent case control study, the use of celecoxib or rofecoxib for more than 1 year resulted in a significant reduction (60%) in the risk of lung cancer . Another study demonstrated a significant reduction (71%) in the risk of breast cancer by coxibs . These results indicated that coxibs have the potential for chemoprevention of lung or breast cancer. In addition, celecoxib has shown some efficacy in the treatment of prostate and pancreatic cancer [147, 148]. Because the long-term use of coxibs is associated with CV risks, the role of COX-2 inhibitors for inhibition of cancer remains an open question .
Acute gouty arthritis
In 239 patients, the efficacy of etoricoxib 120 mg daily and indomethacin 50 mg t.i.d. was tested in the treatment of acute gouty arthritis [150, 151]. The primary efficacy endpoint was the patient’s assessment of pain in the study joint (0–4 point Likert scale). Both treatment groups experienced comparable pain relief over the entire treatment period of 8 days with significant pain relief starting 4 h after the first dose. It was reported that disease history and clinical features could influence treatment outcome with etoricoxib and indomethacin. A significantly greater response of acute gouty arthritis to either etoricoxib or indomethacin was demonstrated among those patients with monoarticular disease, severe/extreme baseline pain, and non-use of colchicine and/or allopurinol .
A 1-week RCT assessed the efficacy of lumiracoxib 400 mg once daily and indomethacin 50 mg t.i.d. for the treatment of acute gouty arthritis (n = 235). Lumiracoxib was comparable to indomethacin for all efficacy variables .
Gastrointestinal (GI) tract
The most commonly reported AEs were related to the GI tract and included dyspepsia, diarrhea, nausea, abdominal pain, and flatulence. Upper GI complications [perforation, ulcers and bleedings (PUBs)] have also occurred in patients treated with coxibs. Many large-scale RCTs have demonstrated that coxibs caused fewer GI AEs compared to tNSAIDs [113, 158–162, 165]. However, most, if not all, of the GI benefit will be lost in patients who take low-dose aspirin [113, 159, 165, 168].
The first large-scale trial to provide evidence that coxibs are associated with a lower incidence of GI events was the Vioxx Gastrointestinal Outcome Research (VIGOR). The incidence of confirmed upper GI events in the rofecoxib group was less than half of that observed in the naproxen group (P < 0.001) . In the Celecoxib Long-term Arthritis Safety Study (CLASS), the annualized incidence rates of upper GI ulcer complications [perforation, obstruction and bleeding (POBs)] alone and combined with symptomatic ulcers for celecoxib versus tNSAIDs (diclofenac and ibuprofen) were 0.76 versus 1.45% (P = 0.09) and 2.08 versus 3.54% (P = 0.02) after 6 months of treatment. The annualized incidence rates were 0.44 versus 1.27% (P = 0.04) and 1.40 versus 2.91% (P = 0.02) for patients not taking aspirin. In contrast, the protective GI benefit of celecoxib was not found in patients who were concomitantly taking low-dose aspirin . However, CLASS data failed to show a significant reduction in upper GI complications for celecoxib .
Similar results were observed in the Successive Celecoxib Efficacy and Safety Study I (SUCCESS-I) comparing the upper GI safety of celecoxib with tNSAIDs in 13,274 patients with OA. Celecoxib 100 or 200 mg b.i.d. was tested versus diclofenac 50 mg b.i.d. or naproxen 500 mg b.i.d. Significantly more ulcer complications occurred within the tNSAIDs group (0.8/100 patient years) compared to the celecoxib group (0.1/100 patient years) (odds ratio 7.02, 95% CI 1.46–33.80, P = 0.008). In the subgroup of patients taking aspirin concomitantly, the difference did not reach statistical significance .
In the Etoricoxib Versus Diclofenac Sodium Gastrointestinal Tolerability and Effectiveness (EDGE) trial, etoricoxib 90 mg once daily was compared for 1 year with diclofenac 50 mg t.i.d. in 7,111 patients with OA . The primary endpoint was the cumulative rate of discontinuations due to clinical and laboratory GI AE. The cumulative discontinuation rate was significantly lower with etoricoxib than diclofenac [9.4 versus 19.2 events per 100 patient years, respectively; hazard ratio (HR) 0.50, 95% CI 0.43–0.58, P < 0.001].
The Multinational Etoricoxib and Diclofenac Arthritis Long-term (MEDAL) programme—the largest randomized trial, performed in 34,701 patients with OA or RA−assessed upper GI safety of etoricoxib (60 or 90 mg daily) and diclofenac (150 mg daily) . Overall upper GI clinical events (POBs or ulcer) were significantly less common with etoricoxib than with diclofenac (HR 0.69, 95% CI 0.57–0.83, P = 0.0001). There were significantly fewer uncomplicated GI events with etoricoxib than with diclofenac (HR 0.57, 95% CI 0.45–0.74, P < 0.0001). The benefit was maintained in patients taking a proton pump inhibitor (PPI) or low-dose aspirin, but there was no difference in complicated GI events. Because the MEDAL programme represents industry-driven research and the FDA-recommended active comparator naproxen was not used, the gastroprotective benefit by etoricoxib (compared to CV risks) is still being discussed [163, 164].
In the Therapeutic Arthritis Research and Gastrointestinal Event Trial (TARGET), 18,325 patients with OA were randomly assigned to receive lumiracoxib 400 mg once daily, naproxen 500 mg b.i.d., or ibuprofen 800 mg t.i.d. for 52 weeks . Lumiracoxib showed a three- to four-fold reduction in ulcer complications compared to the tNSAIDs group (HR 0.34, 95% CI 0.22–0.52, P < 0.0001). Similar to other COX-2 inhibitors, the benefit of lumiracoxib was restricted to patients not taking aspirin (HR 0.21, 95% CI 0.12–0.37, P < 0.0001). The use of aspirin negated the benefit of lumiracoxib over tNSAIDs in reducing ulcer complications.
A pooled analysis of 10 multinational trials indicated that treatment with etoricoxib was associated with a significantly lower incidence of upper GI perforations and PUBs than treatment with tNSAIDs . Similarly, a recent meta-analysis found that coxibs produced significantly fewer gastroduodenal ulcers (RR 0.26, 95% CI 0.23–0.30), fewer clinically important ulcer complications (RR 0.39, 95% CI 0.31–0.50) and fewer treatment withdrawals caused by GI symptoms compared to tNSAIDs . In a nested case control study, the Health Improvement Network Database showed that coxibs provided better upper GI safety than tNSAIDs (RR 0.80, 95% CI 0.6–1.1). However, concomitant use of aspirin negated the superior GI safety of coxibs over tNSAIDs . In addition, coxibs were also associated with lower rates of lower GI injury than tNSAIDs . Interestingly, two RCTs found that healthy subjects treated with celecoxib developed significantly less small bowel damage compared to those subjects receiving naproxen/omeprazole or ibuprofen/omeprazole [170, 171].
In high-risk patients with a history of tNSAID-related complicated peptic ulcers, celecoxib 200 mg q.d. or b.i.d. was as effective as tNSAID plus co-therapy with a PPI in the prevention of recurrences of upper GI events. However, a significant proportion of subjects still had recurrent ulcer complications over a period of 24 weeks [172, 173]. It was also shown that celecoxib plus PPI was more effective than celecoxib alone for prevention of ulcer bleeding in patients at very high risk. The 13-month cumulative incidence of recurrent ulcer bleeding was 0% in the combined treatment group and 8.9% in the celecoxib group (95% CI 4.1–13.7, P = 0.0004) . In a population-based retrospective cohort study, the addition of a PPI to celecoxib conferred extra GI protection for patients aged 75 years or older, but it did not seem to be necessary for patients aged 66–74 years . However, two other studies demonstrated that coxibs plus PPI did not show more effective GI safety than the use of tNSAIDs plus PPI in high GI risk patients [176, 177]. In addition, tNSAIDs plus PPI may be superior to coxibs at minimizing the incidence of dyspepsia [173, 178].
From all these data, it can be concluded that patients with risk factors for GI complications (e.g., age > 70 years, past ulcerations, multiple NSAIDs or aspirin taken especially in high doses, anticoagulant or steroid use, and those positive for Helicobacter pylori) are in need of a “gastroprotective” PPI irrespective of the COX-2 selectivity of the applied NSAID.
Cardiovascular (CV) system
The first evidence that coxibs might increase the CV risk emerged from the VIGOR study . A five-fold increase in thromboembolic events (primarily acute MI) was observed in the rofecoxib group. The increased CV risk associated with rofecoxib was confirmed by the Adenomatous Polyp Prevention on Vioxx (APPROVe) study . Over a period of 3 years, rofecoxib 25 mg daily was compared to placebo in 2,586 patients with a history of colorectal adenomas. The rofecoxib patients had a greater risk of developing thrombotic events compared to the patients taking placebo (RR 1.92, 95% CI 1.19–3.11, P = 0.008). These results prompted withdrawal of rofecoxib. In high-risk patients with coronary artery bypass graft, a three-fold increased risk of CV events was seen in patients who initially received parecoxib intravenously followed by oral valdecoxib compared to patients on placebo [180, 181]. In addition, after reviewing a number of case reports of severe skin reactions, valdecoxib was withdrawn from the market. Subsequently, the cardiovascular safety of other coxibs attracted substantial attention.
In contrast, the CLASS study (which also allowed intake of low-dose aspirin) indicated that the incidence of serious CV events was not significantly different between celecoxib and tNSAIDs (ibuprofen and diclofenac) . This was consistent with two other long-term studies. In the placebo-controlled PreSAP trial in 1,561 patients with prior adenomatous polyps, no significant difference in CV risk was observed between celecoxib (400 mg/day) and placebo . In the Alzheimer’s Disease and Prevention Study (ADAPT), the potential of celecoxib 200 mg b.i.d. and naproxen 220 mg b.i.d. to delay or prevent Alzheimer’s disease was evaluated. This study was stopped after an average follow-up of 3 years because of an apparent increase in CV and cerebrovascular AE in the naproxen group compared to placebo. Interestingly, these rates were similar with celecoxib and placebo .
In contrast, the APC trial showed an increase in CV events in patients taking celecoxib . A total of 2,035 patients with prior adenomatous polyps were randomized to celecoxib 200 or 400 mg b.i.d. or placebo. The data revealed a dose-dependent toxicity: patients taking low-dose celecoxib had a 2.3-fold greater incidence and those taking high-dose celecoxib had a 3.4-fold greater incidence of a CV event compared to placebo. These results led to early termination of the trial. Subsequently, a pooled analysis of all doses of the PreSAP and APC trials demonstrated that celecoxib 200 or 400 mg b.i.d. or 400 mg once daily had a nearly two-fold increased CV risk (HR 1.9, 95% CI 1.1–3.1). There was a trend toward a dose-related increase in CV events and blood pressure (BP) .
The Celecoxib Rofecoxib Efficacy and Safety in Comorbidities Evaluation Trial (CRESCENT) evaluated the effects of celecoxib, rofecoxib, and naproxen on 24-h BP in patients with type 2 diabetes, hypertension and OA. Following 6 weeks of therapy, systolic BP was increased significantly by rofecoxib (to 134.5 from 130.3 mmHg; P < 0.001) but not by celecoxib (to 131.9 from 132.0 mmHg; P = 0.54) or naproxen (to 133.0 from 133.7 mmHg; P = 0.74). This suggested that use of the three drugs may result in different rates of CV events .
In the EDGE study, no significant differences were observed between etoricoxib and diclofenac in the risk of CV events. The same was true in patients taking aspirin, but in the patients on etoricoxib, the nonaspirin subgroup had more MIs . The MEDAL program compared etoricoxib and diclofenac in patients with OA. For the first time, the primary endpoint was a composite of major vascular events. After a treatment period of 18 months, thrombotic CV events were similar with etoricoxib and diclofenac (HR 0.95, 95% CI 0.81–1.11) . However, because the comparator diclofenac also has an increased risk (relative RR 1.4) in contrast to naproxen (relative RR 0.97) , etoricoxib is likely to be associated with a higher rate of CV events than naproxen, which should be the preferred active comparator [163, 164].
The TARGET study showed no significant difference in CV events between lumiracoxib and either ibuprofen or naproxen, irrespective of aspirin use . In patients with high CV risks taking aspirin, a post-hoc analysis demonstrated that patients on ibuprofen had more CV events than the lumiracoxib subgroup (2.14 versus 0.25%, P = 0.038), whereas in the naproxen subgroup, CV rates were similar to lumiracoxib. However, in patients not taking aspirin, CV events were lower in the naproxen group as compared to the lumiracoxib group (0 versus 1.57%, P = 0.027), but not for ibuprofen versus lumiracoxib (0.92 versus 0.80%, P = 0.920) .
A nested case-control study in 486,378 subjects supported the hypothesis that the elevated risk of acute MI was a class effect of coxibs. The increase in risk appeared to be dose dependent . Similar results were observed in other meta-analyses indicating that all coxibs were associated with higher MI risks when compared to placebo or tNSAIDs [191,192]. A population-based cohort study in 122,079 elderly people with and without previous MI found that only celecoxib was associated with an increased risk in people with previous MI . Health-register data suggested that coxibs in all dosages and tNSAIDs in high doses increase mortality in patients with previous MI .
Furthermore, two case-control studies revealed that current use of all coxibs and tNSAIDs resulted in an elevated relative risk for MI [195, 196]. It was found that coxibs and high-dose regimens of ibuprofen and diclofenac, but not naproxen, were associated with a moderate increase in the risk of CV events . Recently, a retrospective cohort study showed no difference between acute MI occurrences in elderly patients taking rofecoxib or celecoxib and those taking ibuprofen/diclofenac. However, the risk of acute MI was elevated among patients using doses above 25 mg/day of rofecoxib compared to patients using higher doses of celecoxib (>200 mg/day) .
A nested case-control analysis of the Kaiser-Permanente database involved more than 1.3 million patients and 2.3 person years of follow-up . Rofecoxib at doses above 25 mg daily exhibited a three-fold higher incidence of CV events. Naproxen did not protect against serious coronary heart disease. Interestingly, in this analysis, celecoxib was not associated with an increased CV risk. Another meta-analysis of 17 case-control and 6 cohort studies calculated the following relative risk factors: rofecoxib (≤25 mg/day) 1.33, rofecoxib (>25 mg/day) 2.19, celecoxib 1.06, diclofenac 1.40, naproxen 0.97, piroxicam 1.06, ibuprofen 1.07 . Similarly, a meta-analysis demonstrated a comparable rate of CV events for celecoxib and tNSAIDs .
A pooled analysis indicated that there was no discernible difference in the incidence of CV events in patients treated with etoricoxib or tNSAIDs . A meta-analysis of 34,668 patients found no evidence that lumiracoxib was associated with a significant increase in CV risk compared to all comparators (placebo, diclofenac, ibuprofen, celecoxib, rofecoxib, and naproxen) . In addition, two analyses of 147,837 patients supported the hypothesis that the elevated CV risk was not a class effect of the coxibs [203, 204].
Taken together, data from large-scale clinical trials and epidemiologic studies suggest the coxibs and tNSAIDs (except naproxen) as a group can potentially increase the risk of CV events and apparently there is a dose-dependent gradient among the various coxibs and tNSAIDs.
As already mentioned, COX-2 is also constitutively expressed in the kidney, and it is regulated in response to alterations in intravascular volume . COX-2-dependent prostaglandin formation is necessary for normal renal development. COX-2 inhibition may transiently decrease urinary sodium excretion and can induce mild to moderate elevation of BP [205, 206].
With respect to kidney damage, coxibs and tNSAIDs have similar effects. They can produce renal insufficiency, sodium retention with hypertension and peripheral edema, hyperkalemia, and papillary necrosis . In healthy elderly subjects on a controlled sodium diet, rofecoxib and celecoxib did not differ from naproxen in influencing renal function as measured by urinary sodium excretion, systolic and diastolic BP, creatinine clearance, or weight changes . When comparing the renal and hemodynamic effects of celecoxib or rofecoxib with indomethacin no differences were found between indomethacin and the two coxibs with respect to proteinuria and renal functions in 11 patients with renal amyloidosis secondary to rheumatological diseases . Etoricoxib also displayed dose-dependent renal AE, such as hypertension, lower-extremity edema, elevated serum creatinine concentrations, and congestive heart failure, similar to tNSAIDs .
An observational study showed that rofecoxib but not celecoxib or tNSAIDs was associated with an increased risk of edema and BP elevation compared to nonusers of any drug . In young and elderly (mean age 68 years) normotensive subjects on celecoxib (200 mg b.i.d. for 2 weeks) no significant effects on parameters of the renin-angiotensin-aldosterone system, insulin clearance, urinary marker proteins, and BP have been observed .
More recently, in a large meta-analysis of 114 clinical trials involving 116,094 patients, rofecoxib was associated with an increased risk of peripheral edema (RR 1.43, 95% CI 1.23–1.66), hypertension (RR 1.55, 95% CI 1.29–1.85), and renal dysfunction (RR 2.31, 95% CI 1.05–5.07). In contrast, celecoxib was associated with a risk comparable to controls .
Other adverse events
Other common side effects of coxibs include dizziness, headache, flu-like symptoms, fatigue, anxiety, and insomnia [68, 211, 212]. As a sulfonamide, celecoxib can cause cutaneous adverse reactions such as rash, urticaria, and photoallergic dermatitis as well as serious and potentially fatal dermatologic AEs such as exfoliative dermatitis, Steven Johnson syndrome, and toxic epidermal necrolysis [213–215]. These can occur without warning also in patients with no history of sulfonamide allergy. Rare cases of hypersensitivity reactions (anaphylaxis and angioedema) have been reported in patients receiving celecoxib or etoricoxib [216–218].
In the clinical trials with etoricoxib 30–90 mg daily for up to 1 year, the most frequently reported laboratory AEs were increased levels of alanine aminotransferase and aspartate aminotransferase, which accounted for 1.0–2.1% of patients ; also chronic use of lumiracoxib could increase these levels [107, 165] and according to the EMEA hepatic dysfunction represents a contraindication. Because of several cases of severe liver damage, lumiracoxib was withdrawn from the Australian and Canadian market and the FDA recently denied the approval of this drug . Thus, during long-term treatment with coxibs, liver function should be regularly monitored.
Two coxibs have been removed from the market, rofecoxib because of its increased risk of CV events and valdecoxib as a result of a combination of CV events and serious dermatological side effects. Therefore all agents of this class have to be carefully assessed including their pharmacokinetic and pharmacodynamic profiles, clinical use, and therapeutic safety.
Because of structural differences, it is understandable that coxibs are characterized by distinct pharmacological features that might also affect their actions. All coxibs are eliminated by the hepatic route and, with the exception of etoricoxib, the polymorphic CYP2C9 is mainly involved. When comparing the disposition of the three marketed coxibs, lumiracoxib might offer some minor advantages (no drug interactions, no effects of liver and kidney dysfunction, long persistence in compartments of action); the clinical relevance of these differences remains to be shown. Likewise, the agents differ in their COX-2 selectivity (see Fig. 2). Whether this has any impact on their efficacy and/or safety has yet to be clarified.
Coxibs are as effective as tNSAIDs for the relief from acute pain and symptoms of OA and RA. Several large studies have demonstrated that coxibs are associated with a reduced risk mainly of uncomplicated upper GI toxicity in the majority of patients who do not receive aspirin. However, most if not all improved GI tolerability of the coxibs is lost when they are given together with low-dose aspirin. Patients taking aspirin, tNSAIDs or COX-2 inhibitors will have a dose-dependent elevated risk for GI toxicity that can be minimized by concomitant intake of a PPI. In this context, the much higher costs of the coxibs should be also taken into account. For patients at risk of tNSAIDs-induced GI tract complications, treatment with tNSAIDs plus a PPI is less costly, as effective as and probably safer than coxibs [176–178]. Coxibs might have a place for patients with several GI risk factors and for patients who do not respond to multiple tNSAIDs.
Currently, there is growing evidence to assume that the increased risk of CV events is a class effect of both coxibs and tNSAIDs (except naproxen). Therefore, an appropriate black-box warning label for all these agents is required . This raises the question of what is the optimum choice of comparator in all these RCTs. Because naproxen provides the best available active treatment, in February 2005 the FDA recommended this tNSAID as the preferred comparator . Furthermore, the CV risks of coxibs apparently increase with dose and duration of exposure. Therefore coxibs should not be prescribed for patients with ischemic heart disease, cerebrovascular disorders (stroke), or peripheral arterial disease. Based on available data, coxibs should be restricted to second-line use as agents for carefully selected patients. If coxibs are indicated, the lowest effective dose should be used for a limited time. In addition, monitoring of kidney and liver function as well as BP appears advisable.
In clinical practice, many patients who require coxibs or tNSAIDs are likely to be elderly with a relatively high risk for GI and CV events who are also taking low-dose aspirin prophylactically. Hence, the potential GI benefits of the coxibs are limited by their CV risks, and both have to be weighed in the individual patient. Whether the benefits are offset by the risks is still a matter of discussion.
For short-term use, coxibs could offer an alternative to tNSAIDs in certain patients. Concerning this treatment, poor metabolizers of CYP2C9 taking celecoxib can have up to a 10-fold elevated drug exposure, which could represent a higher risk for this subpopulation. Such a large variability in drug disposition is not seen with the two other coxibs.
In conclusion, the clinical value of the new coxibs is limited, and they do not represent the progress that was initially promised and claimed [222, 223]. According to recent recommendations from the American Heart Association for the management of musculoskeletal symptoms COX-2-selective NSAIDs represent drugs of third choice and are contraindicated in patients with cardiovascular hazards . More research is required to understand better the differences among all the coxibs . In addition, further RCT and epidemiological studies, preferably sponsored by non-profit organisations or public health systems, can hopefully contribute to a better definition of the therapeutic place of the coxibs.
This work was supported by the Robert Bosch Foundation, Stuttgart, Germany.
Conflict of interest statement
The authors declare that they have no conflicts of interest.