NONSTEROIDAL antiinflammatory drugs (NSAIDs) have been shown to reduce pain and opioid consumption and often accelerate recovery after surgery. However, perioperative inhibition of prostaglandin synthesis by NSAIDs may cause complications, including renal injury, gastric ulceration, and bleeding. Recent molecular studies distinguishing between constitutive cyclooxygenase-1 (COX-1) and inflammation-inducible cyclooxygenase-2 (COX-2) enzymes have led to the exciting hypothesis that the therapeutic and adverse effects of NSAIDs could be uncoupled. The purpose of this article is to review the mechanistic differences between nonselective NSAIDs and selective COX-2 inhibitors (COX-2Is) and to examine currently available COX-2I clinical trials to consider the role of these drugs in postoperative pain management.
Safety and Analgesic Efficacy of NSAIDs
The administration of NSAIDs is one of the most common nonopioid analgesic techniques currently used for postoperative pain management. 1The efficacy of NSAIDs for postoperative pain has been repeatedly demonstrated in many analgesic clinical trials. 2,3The efficacy of traditional NSAIDs can be summarized by results from recent meta-analyses of postoperative single-dose trials showing numbers needed to treat (to obtain one patient with at least 50% pain relief) of 2.6 for 10 mg oral ketorolac, 42.4 for 1,200 mg oral acetylsalicylic acid, 5and 2.4 for 400 mg oral ibuprofen. 6Unlike opioids, which preferentially reduce spontaneous postoperative pain, 7,8NSAIDs have comparable efficacy for both spontaneous and movement-evoked pain, 9–11the latter of which may be more important in causing postoperative physiologic impairment. 12,13Furthermore, NSAIDs have been shown to reduce postoperative opioid consumption 14,15and accelerate postoperative recovery 16,17after certain types of surgery and are thus thought to be an important component of balanced postoperative analgesic regimens. 18
The majority of data about adverse effects of NSAIDs come from the setting of chronic use for arthritis. 19,20However, perioperative inhibition of cyclooxygenase (also called prostaglandin H synthase) by NSAIDs may also cause serious complications, including renal injury, gastric ulceration, and excessive bleeding. 21Brief perioperative NSAID use in healthy adults does not seem to cause important renal dysfunction, 22but clinicians continue to be cautioned by occasional but recurring reports of perioperative NSAID-related renal failure. 23–28Similarly, cases of gastrointestinal ulceration or bleeding have been reported after brief NSAID use, 29–33making this an important risk to consider when using NSAIDs for postoperative pain. Finally, the potential for excessive, and infrequently catastrophic, perioperative blood loss due to NSAID use has been well documented as yet another hazard of these drugs. 34–38Careful patient screening for renal dysfunction, gastritis, gastric ulcers, or bleeding diathesis and judicious administration of NSAIDs may largely prevent these major complications. Rare NSAID-related problems, which are also thought to be due to cyclooxygenase inhibition, include hepatocellular injury, 39asthma exacerbation, anaphylactoid reactions, tinnitus, and urticaria. 40
Mechanisms of Analgesia and NSAID-related Adverse Effects
Traditional NSAIDs comprise a chemically diverse 41group of compounds (e.g. , salicylates, benzothiazines, and indoleacetic, pyrrolacetic, and propionic acids) which, among other actions, inhibit prostaglandin synthesis 42by competing with arachidonic acid for binding to the cyclooxygenase active site. 43Until recently, NSAIDs have been thought mainly to suppress the peripheral nociceptive manifestations of postinjury inflammation. 44After the conversion of membrane phospholipids to arachidonic acid by phospholipase A2in the periphery, cyclooxygenase converts arachidonic acid to the cyclic endoperoxide prostaglandin G2(fig. 1) and then acts as a peroxidase to reduce prostaglandin G2to the cyclic endoperoxide prostaglandin H2. 41Several synthases then convert prostaglandin H2to other prostaglandins (e.g. , prostaglandin D2, prostaglandin E2, prostaglandin F2-alpha, prostaglandin I2) and to thromboxane A2. 45It has been observed that cyclooxygenase inhibition results in shunting of arachidonic acid to lipoxygenase pathways, resulting in increased leukotriene synthesis, a putative mechanism of NSAID-induced bronchospasm. 41NSAIDs are thought to reduce postoperative pain by suppressing cyclooxygenase-mediated production of prostaglandin E2, which is thought to be the primary inflammatory prostaglandin that directly activates and also up-regulates the sensitivity of peripheral nociceptors to cause pain. 41Prostaglandins have also been shown to play a role in spinal nociception, 46–48thus contributing to a growing body of evidence supporting a spinal analgesic mechanism of NSAIDs. 49NSAID-mediated suppression of prostaglandins and thromboxanes, which play a homeostatic role in the stomach (prostaglandin E2and prostaglandin I2), 50kidney (prostaglandin E2), 51and platelets (prostaglandin I2and thromboxane A2), 52is also thought to be the primary mechanism by which NSAIDs cause some of the adverse effects described above. In addition to these three major complications, inhibition of prostaglandin synthesis by NSAIDs is also thought to be the primary mechanism underlying NSAID-induced asthma 53and the suppression of heterotopic bone formation. 54
COX-1 and COX-2 Isoforms of Cyclooxygenase
Subsequent to cloning the gene that encodes for cyclooxygenase in 1988, 55several studies yielded the discovery of a second form of cyclooxygenase and distinguished between the constitutive COX-1 and the inducible COX-2 isoforms of cyclooxygenase. 56,57The advent of new selective COX-2Is has allowed the investigation of differential inhibition of COX-1 versus COX-2 58such that NSAIDs, new and old, can be evaluated 59–61with respect to their COX-1/COX-2 inhibitory profile (table 1). The data shown in table 1indicate that all NSAIDs have at least some effect on both COX-1 and COX-2 isoenzymes and that there are, as yet, no specific values that define a drug as a purely selective COX-2 inhibitor. COX-1 is active and present at a constant concentration in most tissues, particularly in the kidney, stomach, and platelets, where it plays a homeostatic and protective role through the production of prostaglandin E2and prostaglandin I2. 62COX-2, however, is normally present in only very low concentrations but is induced peripherally under conditions of inflammation. 63This functional distinction has led to the exciting hypothesis that selective COX-2Is could uncouple the therapeutic and adverse effects of traditional NSAIDs. However, it is important to note that some exceptions do exist, i.e. , COX-2 plays a homeostatic role in the renal medulla, and COX-1 may produce some prostaglandins that contribute to inflammation. 41Also of great interest in pain management, recent work has shown that COX-2 is constitutively expressed in brain and spinal cord and is further up-regulated after persistent noxious inputs such that spinal COX-2 inhibition may be an important mechanism for reducing postinjury hyperalgesia. 49Finally, COX-2 inhibition results in selective suppression of prostaglandin I2without affecting thromboxane A2, 41and this imbalance may explain the potential for cardiovascular toxicity discussed in the section entitled “Safety of Selective COX-2 Inhibitors in the Treatment of Chronic Arthritis.”
Evidence Suggesting Potential Advantages of COX-2 Inhibitors
Administration of aspirin to arthritis patients resulted in decreased platelet aggregation, whereas the COX-2I celecoxib failed to inhibit platelet aggregation. 64Consistent with animal studies showing that COX-1 inhibition but not COX-2 inhibition leads to gastric ulceration, 65,66multicenter arthritis trials have reported decreased incidences of gastrointestinal ulceration with COX-2Is in comparison with nonselective NSAIDs. 67,68Although these data do not come from the postoperative setting, they do provide further support for the theoretical advantages of COX-2Is.
Safety of Selective COX-2 Inhibitors in the Treatment of Chronic Arthritis
The majority of postmarketing data about COX-2Is comes from experience with celecoxib and rofecoxib, which were approved in the United States in 1998 and 1999, respectively. 69Other COX-2Is available in Europe include meloxicam and nimesulide. 41The COX-2Is nimesulide and meloxicam were marketed in Europe long before the discovery of COX-2 and have since been used as molecular precursors for the development of newer COX-2Is. 62Currently, the major indication of chronic COX-2I use is for the treatment of arthritic pain, although early studies may suggest promise for the prevention of colorectal cancer 70and Alzheimer disease. 71Evidence gathered to date suggests that COX-2Is are safer than traditional NSAIDs with respect to gastrointestinal ulceration and bleeding but not renal dysfunction, and furthermore, COX-2Is may confer increased risk for cardiovascular events (e.g. , cerebrovascular accident, angina, or myocardial infarction). 69Preclinical studies demonstrating the role of COX-2 in the kidney have been echoed by human data indicating that COX-2Is can cause sodium retention and decreased glomerular filtration rate and thus warrant similar precautions that are followed for traditional NSAIDs. 72Gastrointestinal safety data comes largely from two studies, the Vioxx Gastrointestinal Outcomes Research trial (VIGOR) 67and the Celecoxib Long-term Arthritis Safety Study (CLASS). 68In the VIGOR trial, rofecoxib was shown to cause a significantly lower incidence of upper gastrointestinal perforation, ulceration and bleeding as compared to naproxen. 67In the CLASS study, there was no difference in gastrointestinal toxicity between celecoxib and traditional NSAIDs across patients who were also taking low-dose aspirin; however, in patients not taking aspirin, celecoxib did demonstrate a lower incidence of symptomatic ulcers and ulcer complications compared to traditional NSAIDs. 68It was suggested that aspirin's gastrointestinal risks eliminated celecoxib's benefits. 68Important recent reports have suggested that COX-2Is cause an increased risk of thrombotic cardiovascular events. 67,69It has been postulated that COX-2Is may unfavorably alter the thromboxane-prostacyclin balance by inhibiting the vasoprotective prostacyclin (prostaglandin I2) but not the procoagulant thromboxane (thromboxane A2). 69In the VIGOR trial, rofecoxib caused a fourfold increase in the incidence of myocardial infarction compared to naproxen, 67whereas no increase in risk was observed for celecoxib in the CLASS trial. 68However, in the CLASS study, 22% of patients were taking low-dose aspirin for cardioprotection, and this trial did not include patients with rheumatoid arthritis, who have an increased risk of cardiovascular complications. 68This remains a critical issue that requires further investigation, and until resolved, the potential for cardiovascular toxicity should be considered when using COX-2Is in patients at risk for coronary artery disease. Using the example that even brief perioperative β blockade may significantly reduce mortality, 73the potential for postoperative COX-2I administration, however brief, to cause cardiovascular complications must be addressed. Further concerns regarding potential cardiovascular effects of COX-2Is are raised by a recent study in hypertensive osteoarthritis patients demonstrating that the COX-2I rofecoxib but not the NSAID namebutone increased nocturnal blood pressure. 74
Selective COX-2 Inhibitors and Postoperative Pain
In contrast to chronic treatment of arthritis, routine perioperative pain management generally occurs over a period of less than 4 weeks. However, surgery is associated with a set of special situations and problems, including blood loss, fluid shifts, risks of infection and thrombosis, and concomitant administration of anesthetic, analgesic, anticoagulant, and antibiotic drugs. For these reasons, the study and implementation of COX-2Is in the setting of perioperative pain require a unique perspective.
Perioperative Clinical Trials of COX-2Is
Literature searches of perioperative analgesic clinical trials of COX-2Is were conducted using the Cochrane Controlled Trials Register (third quarter 2002) and MEDLINE Database (1966 to February 2003). The database search strategy involved a Boolean search of [celecoxib OR etoricoxib OR flosulide OR meloxicam OR nimesulide OR parecoxib OR rofecoxib OR valdecoxib] AND [postoperative pain OR surgery OR surgical] AND [randomized controlled trials]. Trials reported in abstract form at recent scientific congresses were not included, given their preliminary nature and sometimes limited peer review. It has been well recognized that the use of a placebo control in analgesic trials serves to minimize the risk of false-positive and false-negative results. 75,76Only double-blind, randomized, placebo-controlled trials were evaluated in this review for these reasons. For differing measures of analgesic efficacy and side effects across these trials, statistically significant differences (P < 0.05) between treatments (e.g. , COX-2I, NSAID comparator, placebo) were reported in this review. Most studies use multiple analgesic efficacy measures (e.g. , analgesic use, pain intensity, pain relief). Only the outcome measure that demonstrated a difference was reported on in studies showing significant differences between treatment groups.
The above database search yielded a total of 27 publications of COX-2I trials, one of which described 6 trials, for a total of 32 controlled trials reported (table 2). These included 25 single-dose and 7 multidose trials (number of trials/drug) of rofecoxib (19), celecoxib (6), parecoxib (5), valdecoxib (3), nimesulide (1), and meloxicam (1). Some trials included more than one COX-2I among their treatment arms. Surgical procedures studied in these trials included minor oral surgery, gynecologic surgery, prostatectomy, lumbar discectomy, spinal fusion, and major joint arthroplasty. Reported efficacy measures also varied across studies and included pain intensity, pain relief, and consumption of other analgesics (table 3).
Analgesic Efficacy
Of the 19 rofecoxib trials, 17 demonstrated superior efficacy of rofecoxib to placebo, 77–88whereas two trials showed no difference. 89,90Five of the six celecoxib trials showed superiority to placebo, 81,85,91–93and one showed no difference. 94Parecoxib (the parenteral prodrug of valdecoxib), 95–99valdecoxib, 80,100,101nimesulide, 102and meloxicam 103were found to be superior to placebo in all reported trials. A recent meta-analysis of five rofecoxib trials that investigated 1,118 patients (of whom 211 received placebo and 464 received 50 mg rofecoxib) reported a number needed to treat of 2.3. 104Of 23 trial comparisons with nonselective NSAIDs (17), acetaminophen (3), or opioids (3), 13 NSAID 81–83,92,97,99,102and 1 opioid 91comparator were no different than the studied COX-2I (table 4). The studied COX-2I was observed to be more efficacious than the comparator NSAID 79or opioid 78,91,97in four comparative trials and less efficacious in two trials. 81,93It should be noted that the reported comparative studies are mostly single-dose trials that do not necessarily address relative potency of the drugs being compared. Thus, although one drug may be more potent than another, that drug can only be said to be more efficacious if optimal doses of each drug are being compared. Three trials compared COX-2Is to each other, two of which showed that rofecoxib is more efficacious than celecoxib, 81,85and the third of which demonstrated that valdecoxib is more efficacious than rofecoxib. 80One orthopedic trial by Reuben et al. 86showed that 50 mg rofecoxib given 1 h preoperatively was more effective at reducing postoperative pain than the same dose given 15 min postoperatively, suggesting that, as with traditional NSAIDs, COX-2Is may have preemptive analgesic effects.
Postoperative Analgesic Dose–Response Studies
The analgesic dose–response relation of COX-2Is has been studied in trials of rofecoxib, 83,84parecoxib, 95–99valdecoxib, 100,101and nimesulide 102(table 5). Rofecoxib was studied at 7.5, 12.5, 25, 50, 100, and 200 mg orally in the six controlled trials reported by Morrison et al. 83, and, whereas 50 mg was significantly more efficacious than 7.5, 12.5, and 25 mg, no differences were noted between 50 mg and 100 or 200 mg, suggesting an analgesic ceiling at approximately 50 mg. During the multidose segment (postoperative days 2–5) of the orthopedic rofecoxib trial by Reicin et al. , 84daily doses of 50 mg rofecoxib but not 25 mg resulted in significantly less consumption of supplemental analgesic medication (hydrocodone–acetaminophen). In the parecoxib trial by Desjardins et al. , 9640 mg intravenously was more efficacious than 20 mg but indistinguishable from 80 mg. Rasmussen et al. 97also observed that 40 mg parecoxib was more effective than 20 mg after knee surgery, but higher doses were not studied. Postoperative differences between 20 and 40 mg intravenous parecoxib were not as pronounced in the oral surgery study by Daniels et al. 95Camu et al. 100and Tang et al. 98showed no difference in pain scores or analgesic consumption between 20 and 40 mg oral valdecoxib or between 20 and 40 mg intravenous parecoxib in two other postoperative trials. A recent study of valdecoxib by Desjardins et al. 101demonstrated dose-dependent analgesia between 10 and 40 mg but no difference between 40 and 80 mg, suggesting an analgesic ceiling also for valdecoxib. Finally, the oral surgery study by Ragot et al. 102showed no difference between 100 and 200 mg nimesulide. In summary, these data suggest that COX-2Is, at least in the case of rofecoxib, parecoxib, and valdecoxib, have a postoperative analgesic dosage ceiling similar to that of traditional NSAIDs 41(table 5).
Safety of COX-2Is in the Postoperative Setting
Evaluation and reporting of adverse effects varied considerably across studies from no measures at all to spontaneous patient reporting to specific measures of nausea, vomiting, or blood loss (table 3). All but six trials reported no difference between the studied COX-2I and placebo or active comparator in the overall incidence of adverse effects. However, it should be noted that all COX-2I trials included here were designed and statistically powered with analgesia, not adverse effects, as the primary outcome. One trial did not report adverse effects, 86and in two trials, a significantly greater incidence of postdental extraction alveolitis (“dry socket”) was observed with 50 mg oral rofecoxib as compared to placebo. 80,83Four trials reported significantly fewer adverse effects with the studied COX-2I in comparison with placebo or the active comparator. 78,81,91,94Only three perioperative studies incorporated specific measures of blood loss in the trial design (table 3), and none of these three reported any difference in blood loss between the studied COX-2I and placebo. 85,87,90In addition to adverse effects reported in the postoperative trials cited in this review, single isolated cases of celecoxib-induced oliguria 105and rofecoxib-induced aseptic meningitis 106after brief postoperative use have been recently reported.
Side Effect Profiles from Postoperative COX-2I Trials
Common (5–28%) treatment-emergent signs and symptoms associated with COX-2Is (rofecoxib, parecoxib, and valdecoxib) from postoperative clinical trials that tabulated adverse effects 79,80,84,95–97,100include headache, nausea, vomiting, dizziness, and postdental extraction alveolitis. However, only one of these, postdental extraction alveolitis, occurred more frequently with rofecoxib than with placebo, 80which was also observed in one of the trials reported by Morrison et al. 83
Summary
Postoperative pain management has gone through revolutionary innovations over the past century with the widespread clinical introduction of systemic and neuraxial opioids, regional local anesthetic techniques, patient-controlled analgesia, and coanalgesic therapies such as NSAIDs. 107Current needs for improvement in postoperative pain management include (1) more effective relief of pain and suffering for all postoperative patients 108,109; (2) preventing and/or treating other postoperative symptoms (which may or may not be related to analgesic therapies) such as nausea, pruritus, sedation, and cognitive dysfunction 110; and (3) promoting recovery from surgery by preventing and/or treating postoperative physiologic dysfunction such as atelectasis and ileus. 111,112Thus, therapeutic improvements in postoperative pain management should advance at least one of these goals without impeding the others. In the interest of relieving postoperative pain for all patients, further attention needs to be given to special populations such as patients undergoing tonsillectomy, ocular procedures, spinal fusion, and other surgeries for which nonselective NSAIDs have a relative contraindication.
Current evidence published to date does not suggest that COX-2Is provide a major advantage over traditional NSAIDs. However, it is possible that their development will lead to specific drugs with a superior therapeutic profile. For example, after oral surgery, valdecoxib was recently shown to be significantly more effective than rofecoxib, 80which in turn was shown to be more effective than codeine–acetaminophen 78or diclofenac. 79It remains to be determined whether these differences in analgesic efficacy can be replicated using multidose trials with equipotent dose comparisons and after other, more painful procedures. However, such observations lead to the anticipation that future advances in drug development may result in COX-2Is with clinically important advantages over traditional NSAIDs.
Several COX-2I trials have demonstrated an opioid-sparing effect after surgery, 85,100and comparisons with opioids have reported fewer postoperative side effects. 78,91Thus, COX-2Is are at least as effective as nonselective NSAIDs in reducing opioid requirements and/or opioid-related adverse effects after surgery. Provided that recent evidence of fewer gastrointestinal complications with COX-2Is from arthritis studies 67,68holds true in the postoperative setting, it is hoped that patients with gastrointestinal risk factors (e.g. , previous gastritis, ulcers), in whom NSAIDs are contraindicated, may safely benefit from the addition of a COX-2I to their postoperative analgesic regimen. Both experimental and clinical evidence suggest that NSAIDs impair bone healing. 113,114Thus, spinal fusion surgery patients present another group who may be denied the benefits of NSAIDs because of fear of postoperative deleterious effects on bone graft healing. Early evidence from a rabbit model 115and a small spinal fusion clinical trial 85suggesting that COX-2Is do not interfere with bone healing has led to the optimistic proposal that COX-2Is may be a useful alternative for these patients. 116More recent data does in fact support a role for COX-2 in bone healing, 117and further clinical investigation is needed to address this problem. 118
Issoui et al. 94were unable to demonstrate any difference in postoperative recovery times across postoperative patients receiving acetaminophen, celecoxib, their combination, or placebo. No study has been reported to date that compares COX-2Is to nonselective NSAIDs with respect to postoperative recovery or postoperative physiologic impairment. Such investigations as have been previously conducted with nonselective NSAIDs 119are needed to identify whether COX-2Is have any advantage.
Cardiovascular risks of COX-2Is discussed above remain controversial, and more recent evidence suggests that COX-2Is may not confer greater cardiovascular danger than nonselective NSAIDs. 120,121,122However, comparative postoperative studies that carefully track cardiovascular outcomes are needed to resolve this controversy.
Discovery of the COX-2 enzyme and subsequent development of selective COX-2Is has contributed to a resurgence of therapeutic research in postoperative pain. However, whether these developments have resulted in any tangible improvements in patient care requires further study. Comparative COX-2I trials published to date generally suggest similar analgesic efficacy to nonselective NSAIDs in postoperative pain. Also, these mostly single-dose studies suggest similar safety and tolerability as compared to currently used NSAIDs. Additional data from larger, multicenter, multidose comparative trials could determine whether individual COX-2Is are more efficacious, cost-effective, and/or safe versus nonselective NSAIDs with respect to gastric, renal, and coagulation problems and whether COX-2Is confer greater cardiovascular risk in the postoperative setting. Multiple unresolved questions (table 6) remain to be answered. Until then, cost–benefit considerations 123will likely guide therapeutic choices in the absence of strong evidence supporting any major advantage of COX-2Is.
The authors thank Allan Bell, B.Sc. (Medical Student, University of Toronto, Toronto, Ontario, Canada), for his technical assistance and Dr. Tanveer Towheed, M.D., M.Sc., F.R.C.P.C. (Assistant Professor, Division of Rheumatology, Queen's University, Kingston, Ontario, Canada) for thoughtful comments made on previous versions of this manuscript.