| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
The William Harvey Research Institute, Barts and the London, Queen Mary’s School of Medicine and Dentistry, Charterhouse Square, London, U.K. and
* Unit of Critical Care Medicine, Royal Brompton Hospital, Imperial College School of Medicine, London, U.K.
2Correspondence: The William Harvey Research Institute, Barts & the London, Charterhouse Square, London EC1M 6BQ, U.K. E-mail: t.d.warner{at}qmul.ac.uk
 |
ABSTRACT |
|---|
 TOP ABSTRACT BACKGROUNDCOX ENZYMESCOX SUBSTRATE, PRODUCTS, AND...Nonsteroidal anti-inflammatory...COX-2 AS A THERAPEUTIC...SUMMARY AND CONCLUSIONSREFERENCES |
|---|
Key Words: cyclooxygenase-2 • nonsteroidal anti-inflammatory drugs • rofecoxib • celecoxib • thrombosis
|
BACKGROUND |
|---|
|
TOPABSTRACTBACKGROUNDCOX ENZYMESCOX SUBSTRATE, PRODUCTS, AND...Nonsteroidal anti-inflammatory...COX-2 AS A THERAPEUTIC...SUMMARY AND CONCLUSIONSREFERENCES |
|---|
|
COX ENZYMES |
|---|
|
TOPABSTRACTBACKGROUNDCOX ENZYMESCOX SUBSTRATE, PRODUCTS, AND...Nonsteroidal anti-inflammatory...COX-2 AS A THERAPEUTIC...SUMMARY AND CONCLUSIONSREFERENCES |
|---|
COX genes
The genes for COX-1 and COX-2 are located on chromosomes 9 and 1. The human COX-2 gene is 8.3 kilobases (kb) whereas the COX-1 gene is much larger at 22 kb (1
2
3
4
5
6
7)
. In general terms, the COX-1 gene exhibits the features of a housekeeping gene whereas the gene for COX-2 is a primary response gene with many regulatory sites. In vivo local increases in COX-2 expression have been associated with inflammation, rheumatoid arthritis, seizures and ischemia. The intracellular pathways regulating these events appear numerous and complicated, varying between cell types and cellular stimulus, with nuclear receptors such as peroxisome proliferator-activated receptor-
(PPAR) (8
9
10)
attracting more recent attention.
There is also regulation of COX-2 expression at the post-translational level. For example, Ras, which is implicated in many cancers (see below) and a key regulator of COX-2 expression, exerts this influence at least partly through a post-transcriptional process via protein kinase B (11)
leading to the stabilization of COX-2 mRNA (12)
.
COX proteins: presence of COX-3?
COX-1 and COX-2 are membrane-bound proteins that reside, after synthesis and transport, primarily in the endoplasmic reticulum. Although the genes for COX-1 and COX-2 are clearly different (see above), the proteins actually share 
60% homology at the amino acid level; both catalyze from arachidonic acid the formation of prostaglandin (PG) G2 followed by PGH2 via a peroxidase function, have a similar molecular mass of 70 kDa, and are identical in length. Studies of the tertiary structures of COX-1 and COX-2 have demonstrated that the amino acid conformation for the substrate binding sites and catalytic regions are almost identical. However, there are important differences in these regions, particularly the exchanges of Ile in COX-1 for Val in COX-2 at positions 434 and 523. These substitutions result in a larger and more flexible substrate channel in COX-2 than in COX-1 and in the inhibitor binding site in COX-2, being 25% larger than that in COX-1. Although COX-1 and COX-2 have nearly identical kinetic properties, COX-1 shows negative allosterism at low concentrations of arachidonic acid. This suggests that COX-2 may be a better competitor than COX-1 for arachidonic acid released within the cell (1
2
3
4
5
6
7)
.
It has been suggested recently that there is another COX enzyme formed as a splice variant of COX-1 (13)
. In the initial report of this enzyme it was named COX-3, although it may more appropriately have been named COX-1b. COX-3 is made from the COX-1 gene but retains intron 1 in its mRNA; it was initially reported to be expressed in canine cerebral cortex and in lesser amounts in other tissues analyzed. In humans, COX-3 mRNA was found to be expressed as an 5.2 kb transcript that was most abundant in cerebral cortex and heart. The difference at the protein level between COX-3 and COX-1 is the insertion of 30-34 aa, depending on the mammalian species, into the hydrophobic signal peptide. In COX-3 this signal peptide is not cleaved; the protein is glycosylated and displays COX activity. Again, the initial report showed that comparative assays of canine COX-3, murine COX-1, and murine COX-2 expressed by transfected insect cells demonstrated COX-3 to be selectively inhibited by analgesic/antipyretic drugs such as acetaminophen, phenacetin, antipyrine, and dipyrone as well as nonsteroidal antiinflammatory drugs (NSAIDs). Inhibition of COX-3, it was suggested, could represent a primary central mechanism by which these drugs decrease pain and possibly fever. It may well be, however, that COX-3 is not relevant to humans, as it appears we may be unlikely to express COX-3 (14)
. As noted in the first publication (13)
, there is a one nucleotide difference in intron 1 between human and canine that results in a shift in the reading frame. This would make it impossible for a full-length, catalytically active form of COX-3 to exist in humans. Similarly although COX-3 message may be found in the rat, this could not directly give rise to COX-3 protein as there is a similar shift in reading frame (15)
. There have been previous reports of other splice variants of the COX enzymes. For example, pharmacological studies showed different sensitivities of COX-2 to acetaminophen, although these did not provide sufficient data to conclude that different COX-2 isoforms were present (16)
. So although there are only two genes for COX enzymes (COX-1 and COX-2), multiple COX isoforms underlying the production of prostanoids may exist across a range of tissues.
|
COX SUBSTRATE, PRODUCTS, AND RECEPTORS |
|---|
|
TOPABSTRACTBACKGROUNDCOX ENZYMESCOX SUBSTRATE, PRODUCTS, AND...Nonsteroidal anti-inflammatory...COX-2 AS A THERAPEUTIC...SUMMARY AND CONCLUSIONSREFERENCES |
|---|
). Phospholipase A2 is expressed in numerous different isoforms, both constitutive and inducible, but it is the cytosolic 85 kDa phospholipase A2 that most commonly supplies the arachidonic acid for prostaglandin production. This form of phospholipase A2 requires calcium and calmodulin for activation, and so it has long been considered that phospholipase and not COX is the rate-limiting step in prostaglandin production (4)
. There do appear to be particular associations between the phospholipase and COX isoforms. For example, in rat peritoneal macrophages when COX-2 is not expressed, prostanoid production is driven by cytosolic phospholipase A2 and COX-1. After exposure to lipopolysaccharide, prostanoid production becomes associated with cytosolic phospholipase A2 and COX-2 (17)
. Such changes may be associated with the ability of COX-2 to out-compete COX-1 for arachidonic acid when this substrate is present in low concentrations (5)
.
|
Prostanoid products
Once arachidonic acid has been supplied, both isoforms of COX form PGG2 and PGH2 via identical enzymatic processes. Following these steps, PGH2 can be metabolized by different enzyme pathways to a range of products with potent biological effects (Fig. 1)
. The profile of products made by cells expressing COX-1 or COX-2 is therefore determined by the presence of different downstream enzymes. Earlier it appeared that when cells expressed large amounts of COX-2, the PGH2 formed could be saturating for the PG synthase enzymes resulting in the formation of proportionately larger amounts of PGE2 (4)
, possibly by nonenzymatic conversion. More recently is has become clear that in addition to the cytosolic PGE synthase, there is an inducible microsomal or membrane-associated perinuclear PGE synthase (mPGES) regulated, for instance, by proinflammatory cytokines and glucocorticoids. While the cytosolic PGE synthase is principally coupled with COX-1, this inducible mPGES appears coupled with COX-2 (18)
.
PGD synthase (PGDS) exists in both lipocalin-type and hematopoietic forms. The lipocalin type is a major constituent of the human cerebrospinal fluid (CSF), representing 3% of total CSF protein (19)
. The hematopoietic form produces PGD2 in cells such as mast cells, basophils, and Th2 cells. PGD2 is further dehydrated to produce PGJ2, delta12-PGJ2, and 15-deoxy-delta(12,14)-PGJ2, the last of which is a ligand for the nuclear receptor PPAR- (20
) (see below).
Prostaglandin F2 is synthesized via three pathways from PGE2, PGD2, or PGH2 by PGE 9-ketoreductase, PGD 11-ketoreductase, or PGH 9-,11-endoperoxide reductase, respectively (21)
. Although there are no reports of PGF synthase being inducible, its levels in the uterus may increase during pregnancy associated with the peak of PGF2
production that accompanies parturition (22)
.
PGI2 is produced from PGH2 by the action of PGI synthase, a member of the P450 superfamily (23)
, and there are some reports of this being inducible. For example, the elevated levels of PGI2 metabolites seen during pregnancy, especially in labor, may be associated with stretch induced up-regulation of prostacyclin synthase in the myometrium (24)
.
Thromboxane synthase, which forms thromboxane A2 from PGH2, was originally identified in platelets and cloned and sequenced more than 10 years ago. The important regulation of thromboxane synthase gene transcription appears to be via NF-E2. Other cis elements of the thromboxane synthase gene may also play important regulatory roles (25)
.
Receptors
At any site, the effects of prostanoids released as a consequence of the processes outlined above will be mediated by prostanoid receptors on target tissues and cells. These are cell membrane spanning G-protein-coupled receptors and five major subdivisions have been defined pharmacologically (Fig. 1)
. These correspond to each of the COX metabolites; DP for PGD2, EP for PGE2, FP for PGF2, IP for PGI2, and TP for TXA2. There are in fact four characterized receptors that PGE2 preferentially activates—EP1-4—with each EP receptor being linked to a different transduction pathway giving rise to activation or inhibition of cellular responses. Recent attention has turned to producing antagonists for prostanoid receptors, and these have helped in further defining their roles. For example, an antagonist of the EP1 receptor reduces visceral pain in humans (26)
while EP4 antagonists help to precipitate colitis (27)
and reduce osteoclastogenesis (28)
. Recently much interest has been paid to the role of PPAR- in the biological actions of PGJ2 and related metabolites. PGJ2 in particular has been found to activate PPAR-, leading to changes in cell proliferation, such as promoting adipogenesis.
|
Nonsteroidal anti-inflammatory drugs (NSAIDs) and COX isoforms |
|---|
|
TOPABSTRACTBACKGROUNDCOX ENZYMESCOX SUBSTRATE, PRODUCTS, AND...Nonsteroidal anti-inflammatory...COX-2 AS A THERAPEUTIC...SUMMARY AND CONCLUSIONSREFERENCES |
|---|
2
3
4
5
6
7)
. This appears to be associated with gastrointestinal toxicity: the more COX-1-selective drugs (Fig. 1)
appear to have the tendency to cause more gastrointestinal damage. This has provided the rationale for the development of selective inhibitors of COX-2.
The first two compounds to enter the market place following deliberate development as COX-2-selective agents were rofecoxib (VioxxTM) and celecoxib (CelebrexTM); these joined some existing NSAIDs, most notably etodolac (LodineTM), meloxicam (MobicTM, MobicoxTM), and nimesulide (AulinTM, MesulidTM, NimedTM, and others), that display some level of COX-2 selectivity. Recently the number of therapeutically available COX-2-selective agents has been increased by the addition of valdecoxib (BextraTM; ref 29
) and etoricoxib (ArcoxiaTM; ref 30
). This year another COX-2-selective agent, lumiracoxib (PrexigeTM), should enter the marketplace. This range of agents illustrates the important point that there are a number of chemically different structural classes of COX-2-selective inhibitors (31)
(Table 1
). These COX-2-selective drugs together with the NSAIDs cover a wide range of selectivities toward COX-1 and COX-2 (Fig. 2
).
|
|
|
COX-2 AS A THERAPEUTIC TARGET |
|---|
|
TOPABSTRACTBACKGROUNDCOX ENZYMESCOX SUBSTRATE, PRODUCTS, AND...Nonsteroidal anti-inflammatory...COX-2 AS A THERAPEUTIC...SUMMARY AND CONCLUSIONSREFERENCES |
|---|
and naproxen (33)
, celecoxib vs. naproxen (34)
, valdecoxib vs. naproxen (35)
, and etoricoxib vs. naproxen (36)
and there are many others. Outside of these controlled trials, postmarketing surveys have found, for instance, that patient and physician satisfaction with rofecoxib was high, as most respondents found the drug to be effective, easy to use, and a well-tolerated medication for treatment of osteoarthritis (37)
. As at the doses used these agents have little or no effect on COX-1 (see below), we must conclude that it is inhibition of COX-2 that accounts for the efficacy of NSAIDs in the treatment of osteoarthritis. It may well be that this is due to local inhibition of COX-2 within arthritic tissue, but it could also be due to inhibition of COX-2 within pain pathways in the spinal cord and brain (see below; Fig. 3
).
|
In the case of pain, PGEs appear to sensitize peripheral sensory nerve endings located at the site of inflammation and this could explain the actions of NSAIDs and COX-2-selective compounds. COX products have been shown to facilitate the transmission of pain responses within the spinal cord, and some studies have strongly implicated a role for COX-2 in inflammatory pain. Both COX-1 and COX-2 are expressed constitutively in dorsal root ganglia and spinal dorsal and ventral gray matter (38)
, and COX-1 and COX-2 are present in neurons and in non-neuronal cells such as astrocytes. In the spinal cord, COX-2 immunoreactivity is present in neurons of all lamina, particularly in the superficial layers. This gives a rationale for why the acute thermal hyperalgesia evoked by the spinal delivery of substance P or NMDA and the thermal hyperalgesia induced by the injection of carrageenan into the paw are suppressed by intrathecal and systemic COX-2-selective inhibitors. A COX-1-selective inhibitor is effective when applied systemically, but not spinally, against carrageenan-evoked thermal hyperalgesia and ineffective against spinal hyperalgesia (39)
. It is known that peripheral inflammation generates pain hypersensitivity in neighboring uninjured tissue (secondary hyperalgesia) because of increased neuronal excitability in the spinal cord (central sensitization). This is associated with widespread induction of COX-2 expression in spinal cord neurons and in other regions of the CNS, elevating prostaglandin E2 levels in the CSF. The widespread induction of COX-2 expression throughout endothelial cells within the CNS may underlie much of the PGE2 production associated with hyperalgesia (40)
. Intraspinal administration of a COX-2 inhibitor decreases inflammation-induced central PGE2 levels and mechanical hyperalgesia (41)
. So constitutive COX-2 within the spinal cord appears to be important in pain perception and COX-2 can be induced within the spinal cord, reinforcing pain pathways.
The selective inhibitors of COX-2 rofecoxib (42
, 43)
, celecoxib (43
, 44)
, and valdecoxib (45)
are all similarly analgesic to comparator NSAIDs when used in humans for postdental surgery pain, illustrating that in humans, as in animals, it is COX-2 that is the target for the analgesic effects of NSAIDs even in acute pain conditions. Likewise in primary dysmenorrhea COX-2-selective inhibitors such as valdecoxib and etoricoxib provide analgesia superior to placebo and similar to that of comparator NSAIDs (46
, 47)
, as does rofecoxib when used perioperatively for relief from pain following knee replacement (48)
. Clinically these effects may be explained either by effects at central sites or by inhibition of local production of COX-2 metabolites (44)
(Fig. 3)
.
Kidney
Like the CNS, the kidney contains constitutively expressed COX-2. In animals COX-1 and COX-2 are colocalized in the macula densa, although in humans COX-2 appears to be associated more with renal vascular tissues and podocytes. In normotensive subjects neither blood pressure nor renal function (determined by criteria such as urinary sodium excretion, creatinine clearance, and weight change) is significantly affected by standard doses of celecoxib, rofecoxib, diclofenac, or naproxen (49
, 50)
. However, in salt-depleted healthy subjects, selective inhibition of COX-2 causes sodium and potassium retention, demonstrating that an increased selectivity for COX-2 does not spare the kidney under these conditions (51
52
53)
(Fig. 4
). There have been reports that in elderly patients, who may have compromised renal function, celecoxib, rofecoxib, and naproxen cause reductions in glomerular filtration rate and a reduction in urinary sodium excretion, urinary PGE2, and 6-keto-PGF1 excretion (54
, 55)
. In hypertensive elderly patients, COX-2-selective inhibitors may also promote edema formation and elevations in blood pressure (56
, 57)
. It has recently been suggested that even small elevations in blood pressure induced by NSAIDs could have a major effect on cardiovascular risk profile, with potential for an additional 21,700 events within the U.S. alone (58)
.
|
Cancer
During the 1990s an association was made between regular consumption of NSAIDs (particularly aspirin) and a reduction in the incidence of colon cancer (6)
. Earlier trials showed that sulindac reduces the number and size of colorectal adenomas in patients with familial adenomatous polyposis (who have nearly a 100% risk of colorectal cancer), although this is not sufficient to make sulindac intervention a replacement for colectomy as primary therapy.
It is not entirely clear how this protective effect of NSAIDs is exerted, but it may be via inhibition of COX-2 (Fig. 4)
. Adenocarcinomas in human subjects are associated with marked increases in COX-2 expression, and evidence from studies with isolated cells in culture or animal models suggests that prostaglandins produced by COX-2 slow down the rate of apoptosis in cancerous cells (6)
. PGE2 trans-activates epidermal growth factor receptor and so promotes gastric and intestinal hypertrophy as well as growth of colonic polyps and cancers (59)
. In studies of subjects with familial adenomatous polyposis, celecoxib (60
, 61)
and rofecoxib (62)
have been found to produce significant benefits as scored by endoscopy. Despite these positive results in familial adenomatous polyposis, in colorectal cancer rofecoxib has been reported not to increase the anti-tumor activity of 5-fluorouracil and leucovorin, where it actually increased gastrointestinal toxicity (63)
, and to have little short-term effect on secondary liver metastases (64)
.
As much of the impetus for the research in this area came from epidemiological studies concerning regular aspirin consumption, some controlled studies have been made of the effects of aspirin in subjects with familial polyposis. In one study over 12 months, low-dose aspirin (81 mg), but not a standard dose (325 mg), reduced the adenomas occurring in colon cancer (65)
; in another study, the standard dose was found to reduce adenoma development (66)
. What is interesting about the anti-cancer effects of aspirin is that across studies the maximum anti-cancer effects appear at doses of 1 tablet (325 mg) per day or less, a dose an order of magnitude or more below that required to produce anti-inflammatory inhibition of COX-2 (67)
. Conversely, the anti-platelet actions of aspirin are produced at doses similar to the anti-cancer effects, with no additional effect at higher dose levels. A wealth of evidence also links platelets to the spread of cancer (68)
. On the one hand, could it therefore be that our understanding of the in vivo pharmacology of aspirin links its anti-cancer effects much more convincingly to platelets (the COX-1 of which is completely inhibited at the doses reported) than to COX-2 (which is almost completely untouched)? On the other hand, once-a-day aspirin (82 mg) reduces by >50% the colonic production of prostanoids (69)
whereas 650 mg aspirin produces no further inhibition (70)
. There may also be nonprostanoid-related effects of NSAIDs. For instance, in lung cancer cells in culture the effects of sulindac sulfide are mimicked by the PPAR- agonist ciglitazone and appear independent of effects on COX systems (71)
. Sulindac also acts through K-ras to inhibit the expression of COX-2 via a mechanism independent of inhibition of COX-2 (72)
. This fits with the observation that there is a strong correlation between COX-2 expression and K-ras gene mutation in human colon cancer (73)
and that K-ras and ß-catenin increase COX-2 mRNA and protein expression (74)
. This is consistent with the report that in vitro and in vivo sulindac targets the APC/ß-catenin/TCF pathway (Wnt signaling pathway) (75)
.
As touched on above, COX-2 expression has been reported to be associated with gastric cancers and precarcinogenic (dysplastic) lesions (76)
. Inhibitors of COX-2 reduce the proliferation of gastric cancer cell lines and the formation of aberrant crypt foci in the rat. As for the colon, this effect may be directly associated with the inhibition of COX-2, although some studies have implicated other targets such as protein kinase C-ß(77)
.
As might be expected, COX-2 has been implicated in esophageal cancer. COX-2 is up-regulated in these tissues and inhibition of COX-2 is associated with a reduction in proliferation and an increase in apoptosis of esophageal cancer cell lines. In a rat model of Barrett’s esophagus, COX-2 inhibitors inhibit inflammation, COX-2 activity, and development of adenocarcinoma (78)
. Even more significantly, rofecoxib inhibits cellular proliferation in humans with this condition (79)
.
COX-2 expression is greatly increased in pancreatic cancer (80)
and COX-2 inhibition reduces the proliferation of pancreatic carcinoma cell lines, whereas ibuprofen reduces pancreatic cancer development in hamsters (81)
. Celecoxib may enhance the anti-tumor efficacy of chemoradiation, although this appears associated with increased toxicity (82)
. For more than 20 years NSAIDs have been known to reduce the proliferation of hepatoma cell lines, and this effect is duplicated by selective COX-2 inhibitors (83)
. In vivo, COX-2 is up-regulated in liver cancer; in animal studies, COX-2 inhibition is associated with a reduction in liver metastases from colon cancer (84)
, although this has not as yet been translated into positive outcomes in humans (see above).
In human lung COX-2 expression is strongly associated with adenocarcinomas; inhibition of COX-2 reduces the development of lung cancer in murine models and prevents the marked increase in levels of PGE2 detected in human primary tumors after treatment with paclitaxel and carboplatin (85)
. In animal models, COX-2 inhibitors reduce bone metastases and associated events (86)
and the development of breast cancer (87)
, where COX-2-selective inhibitors may show some therapeutic utility (88)
. COX-2 also appears to be involved in the development of teratocarcinomas (89)
.
Alzheimer’s disease
In addition to colon cancer, epidemiology has pointed to NSAIDs reducing the risk of developing Alzheimer’s disease (90)
. This may be explained by the anti-inflammatory effects of NSAIDs, as paracetamol is without effect and COX-2 containing inflammatory cells are found around amyloid-ß plaques (91)
. A relationship between the neuronal expression of COX-2 and cell cycle markers cyclin D1 and cyclin E has been found in the human temporal cortex, suggesting that COX-2 may be involved early in Alzheimer’s disease pathology (92)
. COX-2 up-regulation is also found in transgenic mouse models of Alzheimer’s disease (93
, 94)
. The anti-platelet properties of NSAIDs may relieve or prevent Alzheimer’s by reducing the chances of ischemic damage caused by blockade of brain capillaries. Finally, it has been suggested that some NSAIDs—ibuprofen, indomethacin, and sulindac sulfide—preferentially decrease the 42 amino acid form of amyloid ß protein (95
, 96)
via an effect that is independent of COX. Disappointingly, the most recent trial of rofecoxib and naproxen found no slowing of cognitive decline in patients with mild-to-moderate Alzheimer’s disease over 12 months (97)
.
Reproductive tract
It has been well established over a number of years that prostanoids are involved in ovulation, fertilization, and blastocyst implantation (98)
. Targeted disruption of COX-2, but not COX-1, in mice produces multiple failures in female reproductive processes including ovulation, fertilization, implantation, and decidualization. For the process of parturition it may well be that COX-1 is required for at least normal entry into parturition since its absence leads to peripartum fetal death. As the COX-1 gene is not strongly inducible, it may be that this isoform mediates the initial stages of labor onset, in which uterine contractions and early phases of cervical dilatation begin, and that the COX-2 isoform induced by the action of proinflammatory cytokines coming from the decidua, trophoblast, or fetal membranes might then generate the prostanoids that sustain myometrial contractility and cervical ripening leading to expulsion of the fetus. Indeed, COX-2 is induced during human labor whereas COX-1 is not. In human studies, the COX-2-selective inhibitor nimesulide has been shown to reduce preterm labor and rofecoxib has been shown to have a negative local effect on human ovulation, resulting in delayed follicular rupture without affecting peripheral hormonal cyclicity (99)
.
Rat testes have substantial COX activity, producing mainly PGF2 with some PGE2 and PGD2. In the adult male rat reproductive system, COX-2 is the predominant isoform and is heavily localized to the epithelium of the distal vas deferens. In the mouse COX-1 is localized in epithelial cells of the caput, corpus, and cauda epididymis and of the vas deferens, and COX-2 in epithelial cells of the distal cauda epididymis and vas deferens (100)
.
Pancreas
More than 20 years ago it was shown that indomethacin lowers basal, glucose-, and glucagon-stimulated acute insulin response. Although in humans and hamsters, COX-2 and not COX-1 gene expression is dominant in pancreatic islet tissue, there are no reports of effects of COX-2-selective inhibitors in human disease.
Gastrointestinal tract
As discussed earlier, gastric damage is the main side effect associated with inhibition of COX-1/2 (Fig. 4)
. Thus, COX-2 inhibitors are expected to cause fewer gastric side effects. Although there have been small studies of gastrointestinal safety of COX-2-selective inhibitors, it is the large trials—VIGOR for rofecoxib (Vioxx) and CLASS for celecoxib (Celebrex)—that have received the most attention. In the VIGOR study rofecoxib 50 mg daily was compared with naproxen 500 mg twice daily in a population with rheumatoid arthritis over a median study time of 9 months (101)
. Rofecoxib and naproxen were found to be similarly efficacious, but those taking rofecoxib had less than half the confirmed gastrointestinal ears of those taking naproxen. In the CLASS study of celecoxib, patients received celecoxib 400 mg twice per day, ibuprofen 800 mg three times per day, or diclofenac 75 mg twice a day (102)
. Aspirin use was also permitted. The trial demonstrated that when the use of aspirin was considered or the definition of the upper gastrointestinal endpoints was expanded to include ulcer events not deemed to be clinically significant upper gastrointestinal events, celecoxib demonstrated statistical superiority to pooled NSAIDs and to ibuprofen. It must be noted, however, that to some extent this conclusion is controversial, as others have interpreted the data from CLASS as showing that celecoxib has no gastrointestinal safety advantage over traditional NSAIDs (103)
. To counter this suggestion, a recent study has underlined the potential gastrointestinal safety of COX-2-selective inhibitors by showing that in patients with a history of gastric ulcer bleeds, celecoxib is as safe as coadministered diclofenac and omeprazole (104)
. In smaller studies, valdecoxib has been shown to cause substantially fewer gastroduodenal ulcers than either ibuprofen or diclofenac while having comparable efficacy (105)
and, indeed, even when used at supratherapeutic doses to produce an ulcer rate significantly lower than naproxen and similar to placebo in healthy elderly subjects (106)
. Long-term etoricoxib use is associated with a reduced incidence of adverse upper GI events compared with nonselective NSAIDs (107)
whereas short-term use produces markedly fewer ulcers than naproxen and, unlike ibuprofen, no increase in daily fecal red blood cell loss (108)
. Similarly, in an 8 day study lumiracoxib use was associated with a gastroduodenal tolerability similar to placebo and superior to naproxen (109)
.
Despite these clinical findings, there is evidence that COX-2 is protective within the gastrointestinal tract, perhaps explaining why in some studies use of COX-2-selective inhibitors can be associated with ulceration rates higher than those of placebo. Both COX-1 and COX-2 are expressed in human gastric mucosa, and COX-2-selective inhibitors suppress the formation of prostanoids from healthy samples of human gastric (110)
and colonic tissue (111)
. Thus, various lines of evidence suggest that COX-2 is expressed throughout the human gastrointestinal tract, although it must be noted that in healthy subjects it appears that COX-2-selective drugs have relatively little effect on in vivo gastric prostanoid production (112)
. In animal models of ulceration and gastrointestinal damage, COX-2 products actually promote gastrointestinal healing, and inhibition of both COX-1 and COX-2 is required to produce acute gastrointestinal damage (113)
. However, in the presence of noxious stimuli it may well be that the action of COX-1 is more important in protecting gastric integrity (114)
. Inhibition of COX-2 may also delay esophageal ulcer healing by reducing ulceration-induced esophageal epithelial cell proliferation (115)
.
Cardiovascular system
The activity of COX in endothelial cells is thought to be beneficial, contributing to the normal functioning of the cardiovascular system via the release of PGI2 (Fig. 4)
. PGI2 promotes vasodilatation, inhibits platelet aggregation and adhesion, and is an endogenous anti-lipidemic agent. COX-2 is induced in animal and human blood vessels after physical damage or exposure to proinflammatory cytokines and appears to be associated with protective functions, inhibiting events such as cell proliferation, cytokine and endothelin-1 release, and adhesion receptor expression.
In healthy normal individuals, both celecoxib and rofecoxib reduce PGI2 production, as determined by the measurement of urinary metabolites (7)
. As PGI2 is known to be produced within blood vessels, this has been interpreted as suggesting that COX-2 is a feature of the human cardiovascular system under physiological conditions (although in endothelial cells, PGI2 synthase colocalizes with COX-1 and not with COX-2) and has led to suggestions that inhibition of COX-2 could be prothrombotic, as the production of anti-aggregatory PGI2 within the circulation would be reduced. These observations appeared to synchronize with outcomes from the VIGOR study in which it was noted there were significantly more myocardial infarctions, although not cerebral infarctions, in those taking rofecoxib than those taking naproxen (101). However, much controversy has surrounded these data—first as to whether VIGOR revealed a true effect of rofecoxib, and second, if true, whether this would be a class effect of COX-2-selective inhibitors. An early analysis of thrombotic events from 23 studies of patients treated with rofecoxib compared with those treated with placebo or nonselective NSAIDs (naproxen, diclofenac, ibuprofen, and nabumetone) suggested that the differences seen in VIGOR could be explained as an anti-thrombotic effect of naproxen on COX-1 in platelets (as VIGOR included no placebo arm rofecoxib could only be compared with naproxen) (116)
. Another study looked at outcomes in patients receiving rofecoxib, nonselective NSAIDs, and placebo in 8 osteoarthritis trials with a median treatment exposure of 3.5 months (117)
. Similar rates of thrombotic cardiovascular adverse events were found with rofecoxib, placebo, and comparator NSAIDs (ibuprofen, diclofenac, or nabumetone). One weakness of these analyses may be that for rofecoxib the larger part of the data concerning the high 50 mg dose comes from the VIGOR study. One analysis of the annualized myocardial infarction rates for COX-2 inhibitors in VIGOR and CLASS compared with that in the placebo group of a recent meta-analysis of 23,407 patients in primary prevention trials (placebo, 0.52%; rofecoxib, 0.74%; celecoxib 0.80%) produced the conclusion that there could be a risk of cardiovascular events in those taking COX-2 inhibitors (118)
. However, readers should be aware that this report was considered by many to be seriously flawed because of incorrect statistical analyses (119)
. Direct analyses of the data from CLASS for cardiovascular events (120)
showed the incidences of serious cardiovascular thrombolic events and of other adverse cardiovascular events were not different between celecoxib and the comparators, ibuprofen, and diclofenac for all patients as well as the subgroup of patients not taking ASA.
Data from observational and real world studies are somewhat contradictory. In a recent study of NSAID use in the Tennessee Medicaid program, users of high-dose rofecoxib (50 mg) were found to be 1.70-fold more likely than non-users to have serious coronary heart disease, while among new users this rate increased to 1.93 (121)
. In contrast, there was no evidence of raised risk among those using rofecoxib at doses of 25 mg or less or among users of other NSAIDs. On the other hand, a more recent analysis has been made of rates of acute myocardial infarction among patients 66 years and older administered selective COX-2 inhibitors, naproxen, and other NSAIDs in Ontario, Canada (122)
. In this analysis using a multivariate model, no significant differences in acute myocardial infarction risk were found for new users of celecoxib, rofecoxib, naproxen, or non-naproxen NSAIDs. Others have also reported that naproxen protects against acute myocardial infarction compared with other NSAIDs (123
, 124)
. COX-2 inhibition, can in certain cases actually improve endothelial cell function in cardiovascular disease (125)
and reduce subsequent cardiovascular events in patients with coronary syndromes without ST segment elevation admitted to a coronary care unit (126)
.
An alternate suggestion has been that the events seen in VIGOR were explained by the lack of effect of rofecoxib compared to naproxen on platelet function. For example, even in humans at supratherapeutic doses, rofecoxib does not affect thromboxane B2 formation as assayed ex vivo in whole blood. Similar studies have shown that ex vivo platelet aggregation and bleeding time are unaffected by supratherapeutic doses of celecoxib or valdecoxib (127)
. Celecoxib may, however, have weak effects on thromboxane B2 formation at high doses (127
, 128)
. Meloxicam and nimesulide, which are less selective than the newest generation of drugs, do not cause significant increases in bleeding time or inhibition of platelet aggregation, even in the case of meloxicam when used at double the maximum therapeutic dose (129)
. Using prescription event monitoring, direct comparisons have seemed to show that celecoxib may cause more cerebrovascular thrombotic events than meloxicam but not more cardiovascular or peripheral venous thrombotic events (130)
. For rofecoxib, a similar analysis showed a relative increase in the rate of cerebrovascular thrombotic events and relative reduc-tion in peripheral venous thrombotic events compared with meloxicam (131)
. There was no difference in the rate of cardiovascular thromboembolic events. For both these comparisons, however, the incidences of the events were low, leaving questions as to the relevance of the findings. A further layer of interest has been added by studies showing that NSAIDs may interfere with the anti-platelet effects of aspirin (132
, 133)
. This is explained by the instability of aspirin within whole blood and the circulation; if agents stop it from entering the COX active site, then aspirin may be quickly metabolized before it can act. Ex vivo studies have shown that concomitant administration of ibuprofen but not rofecoxib, celecoxib, acetaminophen, or diclofenac antagonizes the irreversible platelet inhibition induced by aspirin (134
135
136)
. We may conclude from these studies that COX-2-selective compounds have less ability to antagonize the anti-platelet effect of aspirin than do traditional NSAIDs.
Finally, VIGOR used a population of rheumatoid arthritic patients, and it appears that rheumatoid arthritis is associated with a doubling in the risk of a myocardial infarction, with no increase in the chances of stroke (137)
. Naproxen reduces the risk of acute thromboembolic cardiovascular events (myocardial infarction, sudden death, and stroke) within the rheumatoid population (138)
for whom regular use of NSAIDs or paracetamol (but not aspirin) is associated with an increased incidence of hypertension (139)
.
In conclusion, the mass of data and analyses available appear to show that at standard recommended doses, COX-2-selective inhibitors do not increase the risks of thrombotic events and do not interfere with the anti-thrombotic effects of aspirin. They may even improve endothelium-dependent vasodilation and reduce low-grade chronic inflammation and oxidative stress in coronary artery disease and hypertension, and so could be beneficial in some patients with cardiovascular disease (105
, 140)
.
COX-2 in the respiratory tract
COX-2 is induced by cytokines in a number of airway cells, including the epithelium and underlying smooth muscle. Asthma and related diseases are characterized by excessive proliferation of airway cells, which contributes to airway narrowing in some patients. COX-2 induction inhibits proliferation of human airway smooth muscle cells, suggesting a protective role of this enzyme in diseases such as asthma.
A subset of asthmatic patients experience symptoms after taking aspirin and related drugs, so-called aspirin-sensitive asthma (Fig. 4)
. Although the biochemical mechanisms underlying aspirin-sensitive asthma are still being debated, there is a general level of acceptance that COX activity suppresses leukotriene production. Thus, when COX is blocked by aspirin in sensitive patients, leukotriene production increases and asthma symptoms ensue. Although aspirin-sensitive asthmatics express COX-2 in their airways, clinical studies suggest that inhibition of this COX-2 is not involved in the aspirin-sensitive asthmatic response. Two studies of 60 patients who demonstrated typical naso-ocular and asthmatic reactions to aspirin showed no changes in nasal examination findings or declines in FEV1 after challenges with either rofecoxib (141)
or celecoxib (142)
. The statistical power of these large sample sizes led the authors to conclude that cross-reactivity between aspirin and rofecoxib or celecoxib does not occur in patients with aspirin-sensitive respiratory disease. A smaller study looked at 12 asthmatic patients for whom oral aspirin challenge precipitated symptoms of bronchial obstruction with a fall in FEV1 and a rise in urinary leukotriene E4 and showed that rofecoxib did not cause dyspnea or a fall in FEV1, or any changes in urinary leukotriene E4 (143)
; celecoxib has been shown not to cause asthmatic attacks in 33 subjects who had previous responses to two or more NSAIDs (144)
. Nimesulide was reported to be better tolerated by aspirin- and NSAID-sensitive asthmatics (144)
.
In 15 patients with challenge-proven, NSAID-induced cutaneous reactions, aspirin, nimesulide, and diclofenac (145)
caused urticaria, angioedema, and nonurticarial rash whereas rofecoxib produced no reactions (146)
. As above, this clearly suggests that it is inhibition of COX-1 that underlies the cutaneous reactions to aspirin and the other NSAIDs in these patients, possibly due to the release of a prostanoid "brake" on leukotriene production.
Healing and inflammatory resolution
In the rat, selective inhibition of COX-2 has been reported to impair ligament and fracture healing (147
, 148)
, and selective COX-2 inhibitors impair glomerular capillary repair following from glomerulonephritis (149)
. Slowing of fracture healing has been reported in COX-2 knockout mice (148)
, possibly due to a requirement for COX-2 in endochondral ossification. In rabbits, bone healing after spinal fusion has been reported to be more dependent on the activity of COX-1 than of COX-2 (150)
, as has healing of the rat femur (151)
, although others have shown that inhibition of COX-2 reduces bone ingrowth in rabbits (152)
. Currently there are few good data from studies of humans in whom NSAID use appears associated with, if anything at all, a slight increase in bone density. Although one study has reported that ketorolac reduces spinal fusion there appears little evidence to suggest that NSAIDs have inhibitory effects on fracture healing in humans (153)
.
It is unclear whether properly functioning COX-1 and COX-2 is (154)
or is not (155)
required for dermal wound healing. In mice, COX-2 inhibition may decrease the early inflammatory phase of epidermal wound healing and reduce scar tissue formation without disrupting reepithelialization or decreasing tensile strength (156)
.
In carrageenin-induced pleurisy in rats, a late burst of COX-2 protein expression has been reported during the resolution period and COX-2 inhibition was associated with significantly exacerbated inflammation at 48 h. These observations led the authors to conclude that COX-2 is important in the processes underlying inflammatory resolution in this model (157)
.
|
SUMMARY AND CONCLUSIONS |
|---|
|
TOPABSTRACTBACKGROUNDCOX ENZYMESCOX SUBSTRATE, PRODUCTS, AND...Nonsteroidal anti-inflammatory...COX-2 AS A THERAPEUTIC...SUMMARY AND CONCLUSIONSREFERENCES |
|---|
). However, it has come to be understood that COX-2 is expressed constitutively in certain tissues and is not only involved in inflammatory responses. So NSAID effects such as analgesia and fluid retention are now understood to be largely explained by inhibition of COX-2. Similarly, it now appears most likely that it is inhibition of both COX-1 and COX-2 that underlies the severe gastrointestinal side effects associated with NSAID use. Selective inhibitors of COX-2 spare COX-1, and so this dual inhibition is avoided even though COX-2 within the gastrointestinal tract is inhibited. This supports our understanding that the advantage of COX-2-selective drugs over traditional NSAIDs is that they retain efficacy while clearly producing less adverse side effects within the gastrointestinal tract and airways. Whether at supra-therapeutic doses of COX-2 selectives this advantage is offset by an increase in thrombotic events is a subject of hot debate. The availability and widespread use of COX-2-selective inhibitors has in the last few years provided a fuller understanding of the roles of COX-1 and COX-2 in health and disease. This understanding will grow as the different classes of COX-2-selective agents are used in increasingly varied patient groups.
|
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Received for publication June 26, 2003. Accepted for publication January 21, 2004.
|
REFERENCES |
|---|
|
TOPABSTRACTBACKGROUNDCOX ENZYMESCOX SUBSTRATE, PRODUCTS, AND...Nonsteroidal anti-inflammatory...COX-2 AS A THERAPEUTIC...SUMMARY AND CONCLUSIONSREFERENCES |
|---|
