Myths and facts in the use of anti-inflammatory drugs

Abstract

Background. Because of the prominence of pain-related conditions and the growing complexities of clinical management we aimed to explore and attempt to dispel the several myths that surround these serious therapeutic issues.

Aims. We aimed to provide a careful analysis of the evidence and draw factually based guidance for physicians who manage the broad range of patients with pain.

Methods. Current myths were identified based on the authors’ clinical, scientific, and academic experience. Each contributor addressed specific topics and made his own selection of primary references and systematic reviews by searching in MEDLINE, EMBASE, and CINAHL databases (1990–2008) as well as in the proceedings of the major digestive and rheumatology meetings. The writing and references provided by each contributor were collectively analyzed and discussed by all authors during several meetings until the final manuscript was prepared and approved.

Results. Seven major ‘historical’ myths that may perpetuate habits and beliefs in clinical practice were identified. Each of them was thoroughly examined and dispelled, drawing conclusions that should help guide physicians to better manage patients with pain.

Conclusions. Pain relief must be considered a human right, and patients with osteoarthritis pain should be treated appropriately with analgesic or/and anti-inflammatory drugs. The risk of gastrointestinal (GI) complications with traditional non-steroidal anti-inflammatory drugs (t-NSAIDs) is present from the first dose (with both short-term and long-term use), and strategies to prevent GI complications should be considered regardless of the duration of therapy. Compared with t-NSAIDs, coxib use is associated with a small but significant reduction of dyspepsia. While protecting the stomach, proton pump inhibitors do not prevent NSAID-induced intestinal damage. To this end, coxib therapy could be the preferred option, although further randomized studies are needed. A substantial number of patients who need NSAIDs are also taking low-dose aspirin for cardiovascular prophylaxis. From a GI perspective, the combination of aspirin plus a coxib provides a preferred option compared with aspirin plus a t-NSAID, for patients at high GI risk. As the incidence of renovascular adverse effects with t-NSAIDs and coxibs is similar, blood pressure should be monitored and managed appropriately in patients taking these drugs, although they should be avoided in those with severe congestive heart failure. Due to increased cardiovascular risk, which is dependent on the dose, duration of therapy, and base-line cardiovascular risk, both t-NSAIDs and coxibs should be used with caution in patients with underlying prothrombotic states and/or concomitant cardiovascular risk factors.

• Anti-inflammatory therapy
• arthritis
• aspirin
• cardiovascular system
• coxibs
• GI tract
• kidney
• myths
• NSAIDs
• side-effects

Introduction

There is currently no established treatment to reverse or slow the progression of osteoarthritis (OA). Thus, the primary goal of treatment is optimal pain control and preservation of joint function, thereby breaking the vicious circle of pain and loss of motion, leading to instability, frailty, and total disability. To achieve this goal, pharmacological, non-pharmacological, and, if necessary, surgical treatment regimes must form an integrated approach [1].

Pain is a common reason for patients to visit their family physician [2], [3], and the numbers seeking treatment for pain are anticipated to rise as the population ages and chronic conditions such as OA increase. In the UK, annually, more than 17 million prescriptions are written for anti-inflammatory and analgesic drugs [4].

Paracetamol (acetaminophen) is the oral analgesic of first choice. Although it is less effective than non-steroidal anti-inflammatory drugs (NSAIDs), it is safe when taken within the recommended dose range [5]. However, its hepatotoxicity must be considered, and paracetamol hepatotoxicity far exceeds other causes of acute liver failure in the USA [6].

Key messages

• Pain relief must be considered a human right, and patients with osteoarthritis pain should be treated appropriately. Analgesic or/and anti-inflammatory drugs are an essential and effective part of the treatment concept of this condition.

• Although non-steroidal anti-inflammatory drugs (NSAIDs) represent a very effective class of drugs, their use is associated with a broad spectrum of untoward reactions in the upper and lower gastrointestinal tract, kidney, and cardiovascular system, some of which are shared also by the recently introduced cyclo-oxygenase (COX)-2-selective agents.

• Only a rational use of these drugs, alone or in combination with other compounds, can assure an effective and safe analgesic and anti-inflammatory therapy.

Opioid analgesics show efficacy comparable with paracetamol but are less effective than NSAIDs alone [7]. Opioids often cause adverse effects, particularly in the elderly, including constipation, central nervous system (CNS) reactions, increased risk of falls, and even death [8].

In patients who respond inadequately to paracetamol or opioids, NSAIDs or coxibs at the lowest effective dose for the shortest duration should be prescribed. Other pharmacological treatments include the symptomatic slow-acting drugs for OA (SYSADOA) (e.g. glucosamine sulphate, chondroitin sulphate, diacerhein), topical NSAIDs, capsaicin, and intra-articular glucocorticoid injections [1]. The data supporting these other agents are variable, but they may be successful in some patients. However, most of them will require a NSAID at some time during the clinical course of their OA. It is therefore evident that these drugs are an essential and effective part of disease management.

Abbreviations

Table

ACE	angiotensin-converting enzyme
AMI	acute myocardial infarction
ASA	aspirin (acetylsalicylic acid)
AT1	angiotensin type 1
BP	blood pressure
CHD	coronary heart disease
CHF	congestive heart failure
CNS	central nervous system
CV	cardiovascular
COX-2	cyclo-oxygenase 2
GI	gastrointestinal
OA	osteoarthritis
PG	prostaglandin
PPIs	proton pump inhibitors
RA	rheumatoid arthritis
RCTs	randomized clinical trials
SSRIs	selective serotonin reuptake inhibitors
t-NSAIDs	traditional (non-selective) non-steroidal anti-inflammatory drugs
UGIB	upper GI bleeding

About 25% of Canadians were prescribed short-term NSAIDs, and ~4% long-term (≥6 months) [9], equating to a prevalence of ~6.2 million short-term and ~1.0 million long-term NSAID users in Canada. However, this does not include over-the-counter (OTC) NSAID use, and a US cohort study reported point prevalence of NSAID use as 8.7% [10]. Low-dose aspirin (ASA) is also widely used for cardiovascular risk reduction and for chemoprevention. [11]

Gastrointestinal (GI) adverse events, including dyspepsia [12] or ulcer and ulcer complications [13], are common in patients taking NSAIDs. One systematic review of patients taking NSAIDs for at least 2 months found that 1 in 5 have an endoscopic ulcer, 1 in 70 a symptomatic ulcer, and 1 in 150 have GI bleeding [14].

Cyclo-oxygenase 2 (COX-2)-selective inhibitors (often referred to as coxibs [15]) were introduced as a safer alternative to traditional NSAIDs (t-NSAIDs) based on an improved GI safety profile but were subsequently found to be associated with an increase in cardiovascular (CV) adverse events [16], [17]. The withdrawal of rofecoxib and valdecoxib and the subsequent evidence that most if not all t-NSAIDs also carry a CV risk [18] have led to confusion, and physicians are questioning how best to manage pain, especially in older patients at increased GI and CV risk.

Low-dose aspirin is widely used for secondary prevention of myocardial infarction and thrombotic stroke [19] and is increasingly taken by people with no or only a few CV risk factors (primary prevention) where the risk of a serious GI complication or hemorrhagic stroke [20] is likely equal to or greater than any potential benefit [21].

Prescribing appropriate analgesics is further complicated by the need for careful assessment of concomitant medications and any underlying medical conditions. Clopidogrel is also used commonly in association with ASA in patients at high CV risk, and such combination therapy increases the risk of GI bleeding [19]. Selective serotonin reuptake inhibitors (SSRIs) may also cause GI bleeding, and this risk increases with concomitant ASA or NSAID use [22]. Common co-morbidities such as cardiovascular disease, diabetes, metabolic syndrome, and GI problems further confound appropriate management.

Because of the prominence of pain-related conditions and the growing complexities of clinical management we aimed to explore and attempt to dispel the several myths that surround these serious therapeutic issues. We aimed to provide a careful analysis of the evidence and draw factually based guidance for physicians who manage the broad range of patients with pain.

Methods

Current myths were identified based on the authors’ clinical, scientific, and academic experience. Each contributor addressed specific topics and made their own selection of primary references and systematic reviews by searching in MEDLINE, EMBASE, and CINAHL databases (1990–2008) as well as in the proceedings of the major digestive and rheumatology meetings. The writing and references provided by each contributor were collectively analyzed and discussed by all authors during several meetings until the final manuscript was agreed.

Results

Seven major ‘historical’ myths that may perpetuate habits and beliefs in clinical practice were identified. Each was thoroughly examined, drawing conclusions relevant to clinical practice.

Myth 1: NSAIDS when given for short periods (days) have a lower GI risk than when given in the longer term (weeks/months).

It is widely believed that treating patients with t-NSAIDs for short periods carries no, or very low, GI risk. However, current evidence suggests that this is not the case and that the risk may even be higher during the initial treatment period. This evidence comes both from randomized controlled trials and observational studies.

Recent outcome trials [23–27] have shown that upper GI complications and symptomatic ulcers occur within the first month of therapy. Although these trials have shown, in general, that COX-2-selective inhibitors are associated with a clear risk reduction for upper GI events compared with t-NSAIDs, the question is whether severe upper GI events can also occur in patients taking COX-2-selective inhibitors early in the treatment period. The data reported show a lower incidence of these events in the initial treatment period and a less rapid increase in the incidence of these events over time compared with t-NSAIDs [23], [24], [27]. The APPROVe trial, which compared the efficacy of rofecoxib with placebo in preventing colon polyps, showed that patients taking rofecoxib did have events early, particularly within the first month of treatment [28].

Observational studies have either found a constant GI risk over time for patients taking t-NSAIDs or an increased relative risk during the first month (short-term) of therapy [29]. A well performed cohort study [30] showed that admission to hospital for upper gastrointestinal bleeding (UGIB) and perforation remained constant over months of NSAID exposure, i.e. long-term therapy (Figure 1) and suggested a carry-over effect at the end of prescribing. In a more recent investigation it was found that the relative risk (RR) of ulcer bleeding for coxib use was lower (neutral for celecoxib) compared to t-NSAIDs. However, the RR was also higher during the first month of t-NSAID or coxib use compared to continuous use [31]. A meta-analysis of 18 observational studies reported a constant RR of upper GI complications over time, and that RR dropped quickly once treatment was stopped. Traditional-NSAID use increased the risk of UGIB among new users and among those already on therapy for several months, at least during the first year of treatment. However, the pooled RR was greater for new users (RR 5.7; 95% confidence interval (CI) 4.9–6.6). On average, 2 months after the end of therapy the risk had returned to base-line incidence among subjects not using t-NSAIDs [29].

Figure 1.  Relative risk (95% confidence intervals) for admission for complicated upper gastrointestinal events during each time-period of continuous exposure (mean value shown as upper dotted line) and during each time period of subsequent continuous non-exposure to non-steroidal anti-inflammatory drugs (NSAIDs) [30].

A careful analysis of available studies suggests that the reported higher risk of GI complications in short-term t-NSAID users compared to those taking long-term t-NSAIDs is likely due to the inclusion of patients prone to develop complications (e.g. at-risk patients carrying asymptomatic lesions or ulcers), whereas estimates of RR in long-term users are based on studies depleted of such susceptible populations.

In summary, the evidence from both randomized clinical trials (RCTs) and observational studies indicates that the risk of GI complications is present with both short-term and long-term use of t-NSAIDs from the first dose. Thus, even a short course of NSAID therapy (e.g. for postoperative pain or acute musculoskeletal injury) carries a risk equivalent to that of long-term treatment. As a consequence, strategies to prevent GI complications should be implemented regardless of the duration of therapy, especially in patients with high GI risk.

Myth 2: Coxibs and t-NSAIDs are associated with similar rates of dyspepsia

Upper GI symptoms (including dyspepsia, heartburn, abdominal bloating or cramping), associated with t-NSAID use occur in up to 25% of patients taking t-NSAIDs [32]. These symptoms, which have little relationship to erosions or ulceration seen endoscopically, are sufficient to prompt a change in therapy in 10% of patients. Dyspeptic symptoms are the most prevalent and resource-consuming adverse effects of t-NSAIDs and have important cost-effectiveness implications when selecting between COX-2-selective and non-selective t-NSAIDs [33].

Although dyspepsia has been estimated to occur in 15%–60% of t-NSAID users [34], data from a meta-analysis indicate that only 8%–12% of patients receiving chronic high-dose t-NSAID therapy develop recurrent dyspeptic symptoms [12]. The incidence of dyspepsia with coxib therapy has been extensively studied. In the diclofenac protocol of the CLASS study [23] dyspepsia was assessed by the Severity of Dyspepsia Assessment (SODA) questionnaire [35]. SODA revealed that celecoxib, at 2 to 4 times the recommended dose, had a superior dyspepsia-related tolerability and patient satisfaction compared with standard dose diclofenac [35], and this held true for both younger and older patients. A similar reduction in dyspeptic symptoms was reported in two pooled analyses of rofecoxib [36] and valdecoxib [37] trials, although in aspirin users celecoxib was associated to a significantly lower incidence of GI symptoms than rofecoxib [38]. Moreover, in an elderly population, upper GI symptoms and prescription of gastroprotective drugs were lower in COX-2-selective than t-NSAID users [39]. Furthermore, compared with those taking t-NSAIDs, coxib users were less likely to switch medications [40], and the ‘survival time’ on a coxib was significantly greater than on t-NSAIDs, with a median survival time of celecoxib (15 months) significantly longer than that of rofecoxib (13 months) [41].

The impact of dyspepsia in t-NSAID and COX-2-selective users was examined in a meta-analysis of 26 RCTs [42] (Figure 2). The study revealed a 12% relative risk reduction in dyspeptic symptoms for coxibs versus t-NSAIDs. Compared with a t-NSAID, the number needed to treat (NNT) with a coxib to prevent 1 patient from developing dyspeptic symptoms was 27.

Figure 2.  Meta-analysis using the fixed-effects model of RCTs reporting dyspeptic symptoms in patients who received COX-2-selective versus t-NSAIDs. Summary estimate is the relative risk with 95% confidence intervals [42].

Dyspepsia reduction with coxib therapy may not appear substantial, and patients on COX-2-selective NSAIDs may need co-therapy with proton pump inhibitors (PPIs) to obtain the same symptom improvement observed in those taking t-NSAIDs [43]. Experience, however, indicates that—in the individual patient—the lower dyspepsia rate of coxibs compared with t-NSAIDs, while not large, could be clinically relevant.

Although head-to-head data are limited, a t-NSAID + PPI affords greater risk reduction for dyspepsia than selective drugs when compared with the common base-line of t-NSAIDs [42]. Regardless of the treatment pursued, patient reassurance is important, as more than 50% of patients with NSAID-associated dyspepsia may improve without any intervention whatsoever [43].

Myth 3: A proton pump inhibitor and a traditional NSAID achieves full GI protection

Although the upper GI toxicity of t-NSAIDs is well documented, the appreciation that t-NSAID damage extends beyond the duodenum is less well recognized [44]. Traditional-NSAID-associated toxicity to the small and large bowel has several different manifestations, including subclinical and clinically evident damage. The prevalence of t-NSAID-associated lower GI adverse effects may exceed that detected in the upper GI tract, and the frequency of life-threatening complications in the lower GI tract represents about one-third of all GI complications associated with the use of these drugs [45].

Although inhibition of mucosal prostaglandin (PG) synthesis during t-NSAID treatment occurs along the entire digestive tract, there are significant differences between the intestine and the stomach and duodenum with regard to other pathogenic factors that may add to damage. The absence of acid (which plays a pivotal role in the pathogenesis of upper GI damage) and the presence of bacteria and bile in the intestine [46] (which may trigger specific t-NSAID-related pathogenic mechanisms in the distal GI tract) are the most important.

Co-administration of misoprostol (a synthetic PGE₁ analog) or proton pump inhibitors (PPIs) currently represents the best available strategy to prevent t-NSAID-associated upper GI symptoms and mucosal damage in patients with GI risk factors [32], [47]. However, t-NSAID-associated intestinal damage is not a pH-dependent phenomenon, and co-administration of antisecretory drugs is unable to either prevent or treat lower GI mucosal lesions beyond the proximal duodenum. Although recent experimental evidence [48] suggests a protective activity of PPIs on indomethacin-induced intestinal damage acting through antioxidant and anti-inflammatory properties, video capsule studies have shown that the combination of a PPI with a t-NSAID is not capable of preventing the lower intestinal damage associated with short-term administration of naproxen or ibuprofen [49], [50].

COX-2-selective inhibitors may induce less or no damage to the intact distal gastrointestinal tract compared to t-NSAIDs. The lack of intestinal damage with this class of drugs in animal experiments has been confirmed in clinical studies [45]. Most patients taking either meloxicam [51] or nimesulide [52], two preferential COX-2 inhibitors, had normal intestinal permeability and no increase in intestinal inflammation in comparison to control patients not taking the drug. Conversely from t-NSAIDs, neither rofecoxib [53] nor lumiracoxib [54] affects intestinal permeability in humans. In studies performed in healthy volunteers rofecoxib [55] and etoricoxib [56] compared with ibuprofen did not increase fecal blood loss. Similarly, the incidence of anemia with celecoxib is significantly lower than with t-NSAIDs [23], [57]. Although some case reports of acute colitis or lower intestinal complications have been reported with coxibs, data from the VIGOR trial have shown that the benefits of rofecoxib 50 mg/day compared with naproxen (500 mg b.i.d.) were seen in both the upper and lower GI tract, with risk reductions of 50% and 60%, respectively [24], [58]. Similarly, fewer lower GI adverse events were reported in a large, long-term RCT (MEDAL study) with etoricoxib (60 or 90 mg daily), where a 24% risk reduction over diclofenac in all investigator-reported events was observed [59].

Myth 4: Aspirin negates the GI benefits of a coxib

It is widely believed that co-prescription of aspirin negates the benefit of a coxib and that this drug combination should not be used. This has led to the clinical dilemma of how to manage patients who require an anti-inflammatory drug for arthritis but are at GI risk, and who also need to take aspirin for cardiovascular prophylaxis.

Since up to 45% of coxib users could be taking low-dose aspirin [60], the question of COX-2-selective inhibitor GI safety in the presence of aspirin is of substantial clinical importance.

Overwhelming evidence from large randomized trials (34,460 patients), meta-analyses of randomized trials (52,474 patients), and large observational studies (3,093 bleeding events) shows a better GI safety of coxibs, with consistent reductions in serious GI events of about 50% in comparison to t-NSAIDs [61]. However, subanalyses of aspirin users from two large-scale outcome trials (CLASS [26] and TARGET [27]) have demonstrated that, in the presence of aspirin use, there was no significant difference in the incidence of upper GI ulcer complications between t-NSAID and the COX-2-selective inhibitor users (i.e. celecoxib or lumiracoxib, respectively). The more recent MEDAL study also showed that the decrease in overall and uncomplicated GI events in etoricoxib versus diclofenac users was smaller in patients taking low-dose aspirin [62].

Thus, while coxibs present a better upper GI safety profile than t-NSAIDs, aspirin use is a strong effect modifier [63] so that the differences between them tend to be reduced when combined with low-dose aspirin [31]. It is, however, worth mentioning that neither of the above trials was specifically powered to address the issue of ulcer complications in aspirin users. As such, the generalizability of these results is open to question. The clinical importance of this issue needs to be addressed in adequately powered long-term safety studies evaluating ulcer complications. Until such time, the possibility remains of a relative benefit of COX-2-selective inhibitors compared with t-NSAIDs, also in aspirin users. Interestingly enough, in the subpopulation of aspirin users in the SUCCESS-I study the incidence of ulcer complications was numerically lower in the celecoxib group compared with the t-NSAIDs group, although this difference was not statistically significant (odds ratio 2.0; 95% CI 0.12–31.7) [64].

In contrast to the findings from these large trials, endoscopic studies offer mixed results. Two separate retrospectivepost hoc analyses of pooled endoscopic trials, one with celecoxib and the second with valdecoxib, demonstrated that the rate of endoscopic ulcers was approximately 50% lower in aspirin users (81–325 mg/day) receiving a COX-2-selective inhibitor compared with those randomized to a non-selective NSAID plus aspirin [65], [66]. However, in contrast to these results, a prospective analysis by Laine et al.[67] demonstrated that the use of rofecoxib (25 mg o.d.) with low-dose aspirin (81 mg o.d.) increased the incidence of endoscopic ulcers (≥3 mm in diameter) to a rate not significantly different from that of ibuprofen alone. The cumulative endoscopic ulcer incidence rate was 16% in the group receiving aspirin plus rofecoxib compared with 17% in the subjects receiving ibuprofen (P = 0.62) [67]. However, this analysis did not include a t-NSAID plus aspirin treatment arm.

Two meta-analyses of celecoxib plus ASA found 50% fewer endoscopic ulcers than in those taking t-NSAIDs and ASA [68], [69]. The study by Moore and colleagues [68] showed that in patients not taking ASA, the relative risk (RR) of celecoxib at any dose versus a t-NSAID was 0.28 (95% CI 0.22–0.36); with ASA, the RR for any dose of celecoxib versus a t-NSAID was 0.48 (95% CI 0.28–0.83), suggesting the protective advantage of the coxib was preserved but was smaller when ASA was also used [69]. In line with these findings, a recent endoscopic study [70] found, in subjects receiving celecoxib plus aspirin, an incidence of gastroduodenal ulcers significantly lower than for naproxen plus aspirin (7% versus 25.3%; RR 0.28 (95% CI 0.17–0.45); P<0.001), but significantly higher than placebo plus aspirin (7% versus 1.6%; RR 4.78 (95% CI 1.12–20.32); P=0.016).

While these endoscopic findings are important, they do not provide definitive guidance for making therapeutic decisions in patients requiring anti-inflammatory therapy for pain and low-dose aspirin for cardiovascular prophylaxis. Endoscopic ulcer rates are indeed surrogate markers for ulcer complications. However, data concerning hospital admissions for upper GI bleeding do confirm an advantage of coxibs over t-NSAIDs in low-dose aspirin users. A cohort study [71] using a Quebec health insurance database showed that, in elderly patients, coxibs (namely celecoxib and rofecoxib) and ASA were less likely to be associated with GI hospitalization than a t-NSAID and ASA (Table I) (HR 0.53, 95% CI 0.34–0.83). All these data, taken together, suggest that in patients receiving cardiovascular protection with ASA and pain control with anti-inflammatory drugs, celecoxib (and likely other coxibs) may be safer with regards to GI toxicity compared to t-NSAIDs.

Table I.  Risk for GI hospitalization amongst a cohort of elderly patients taking t-NSAIDs or coxibs and requiring CV protection with low-dose aspirin [71].

Treatment	Hazard ratio (95% CI)
t-NSAIDs only (reference)	1
t-NSAIDs + aspirin	1.61 (1.02–2.56)
Coxibs	0.62 (0.45–0.81)
Coxibs + aspirin	0.86 (0.63–1.17)
Coxibs + aspirin versus t-NSAIDs + aspirin	0.53 (0.34–0.83)

New data further show that concomitant t-NSAID use can attenuate the benefits of ASA therapy. Ibuprofen associated with the COX-1 binding site blocks access of ASA to the serine binding site in the apex of the COX-1 pocket, which has important clinical implications [72], [73] for the cardiovascular protective effect of ASA. Compared with those who used aspirin alone, it has been shown that patients taking aspirin plus ibuprofen had an increased risk of all-cause mortality (adjusted hazard ratio 1.93, 95% CI 1.30–2.87; P = 0.0011) and cardiovascular mortality (1.73, 1.05–2.84; P = 0.0305) [73]. Co-administration of placebo or celecoxib, however, did not adversely influence aspirin-related inhibition of platelet COX-1 activity [74].

In summary, in all available large coxib GI outcome studies, when aspirin was permitted, a coxib plus aspirin is associated with a non-significantly lower ulcer rate and lower ulcer complication rate when compared with t-NSAIDs alone. Moreover, a coxib plus low-dose aspirin is associated with lower GI risk than a t-NSAID plus low-dose aspirin. Thus, combination of aspirin plus a coxib is the preferred option compared with aspirin plus a t-NSAID, for patients at high GI risk who require aspirin for CV prophylaxis. Future RCTs specifically designed to compare the GI and CV safety of a coxib plus aspirin and a t-NSAID plus aspirin are needed to confirm these conclusions.

Myth 5: Coxibs increase the risk of renovascular events compared to traditional NSAIDs

In the human kidney prostaglandin E₂ and prostacyclin are the predominant prostanoids, which play an important role in maintenance of renal blood flow and glomerular filtration rate (GFR), inhibition of tubular sodium reabsorption, antagonism of antidiuretic hormone, and regulation of renin secretion [75]. As regards the involvement of COX isoenzymes in kidney physiology, clinical studies with selective COX-2 inhibitors found that COX-2 is the critical enzyme for sodium excretion and renin release [76]. For renal hemodynamics the evidence points to COX-1 as the predominant enzyme [75]. Thus, from a pharmacological standpoint there is no special risk of selective COX-2 inhibition as compared with inhibition of both COX isoenzymes by t-NSAIDs [75]. All t-NSAIDs and coxibs may affect renal function leading to edema, hypertension, hyperkalemia, water intoxication, decreased GFR, and acute renal failure [75]. In patients without renal risk factors (Table II) the renal adverse effects of t-NSAIDs are rarely seen [77]. However, the incidence of t-NSAID-induced renal side-effects can be as high as 20% in at-risk patients [78].

Table II.  Risk factors for renal side-effects of NSAIDs [75].

Risk factor
Pre-existing renal impairment
Dehydration
Sodium depletion by diet or diuretics
Congestive heart failure
Liver cirrhosis with ascites
Lupus nephritis
Postsurgical status
Co-medication with nephrotoxic drugs (e.g. aminoglycosides, cyclosporine)
Co-medication with ACE inhibitors, angiotensin type 1 (AT1) receptor antagonists
Co-medication with potassium-sparing diuretics

The renovascular risk of t-NSAIDs was seldom described until clinical trials with coxibs. In the study by Chan et al.[79] diclofenac 150 mg/d caused hypertension, edema, and acute renal failure in 18.9%, 5.6%, and 6.3% of patients, respectively, in a group of 143 patients with at least one renal risk factor; the corresponding figures for celecoxib 200 mg b.i.d. were 13.9%, 4.9%, and 5.6%, respectively. A meta-analysis of 19 RCTs with a total of 45,451 patients found a relative risk (RR) of developing hypertension with rofecoxib compared to t-NSAIDs of 1.8 (95% CI 1.2–2.7), while with celecoxib versus t-NSAIDs a trend for a lower RR was observed (RR 0.8; 95% CI 0.7–1.0) [80]. Similar findings were reported in a retrospective case-control study involving 17,844 patients aged over 65 years [81]. The prospective, randomized MEDAL program, which included 34,701 patients with OA or rheumatoid arthritis, showed that etoricoxib 90 mg/d was associated with discontinuations due to edema more frequently than diclofenac 150 mg/d; the same held true for hypertension [62]. The higher risk for edema and hypertension with rofecoxib and etoricoxib in comparison to celecoxib and t-NSAIDs is currently not well understood. The most plausible explanation is the long plasma half-life and sustained inhibition of COX-2 by rofecoxib and etoricoxib [15], [77] as well as their greater renal excretion compared to celecoxib [15].

The new-onset or exacerbation of edema and hypertension can precipitate congestive heart failure (CHF) [82]. Recent studies have focused on the risk of CHF associated with COX-2-selective inhibitors or t-NSAIDs. The population-based cohort study reported by Hudson and colleagues assessed the risk for recurrent CHF or death in elderly patients prescribed celecoxib, rofecoxib, or t-NSAIDs [83]. This risk, whether combined or separate, was higher in patients prescribed t-NSAIDs or rofecoxib than in those prescribed celecoxib (Figure 3). A similar conclusion can be derived from the epidemiological study by Mamdani and colleagues [84], which compared rates of admission for CHF in elderly patients who were newly dispensed COX-2 inhibitors (rofecoxib n=14,583; celecoxib n=18,908), or t-NSAIDs (n=5,391), and randomly selected non-t-NSAID users as controls (n=100,000). Relative to non-t-NSAID users, patients taking rofecoxib and t-NSAIDs had an increased risk for admission for CHF (adjusted rate ratio 1.8, 95% CI 1.5–2.2; and 1.4, 1.0–1.9, respectively), but not celecoxib (1.0, 0.8–1.3). In the MEDAL study, a higher rate of congestive heart failure was seen with etoricoxib 90 mg than with diclofenac, but this difference fell short of statistical significance; no difference was seen with etoricoxib 60 mg [62].

Figure 3.  Adjusted hazard ratios (95% confidence intervals) for death and recurrent congestive heart failure (combined and alone) in patients taking NSAIDs or coxibs, according to exposure group [83].

In summary, the renovascular effects of t-NSAIDs and coxibs are similar with molecule-specific quantitative differences between the various drugs. Blood pressure should be monitored and managed appropriately in patients on t-NSAIDs and coxibs. In patients with severe congestive heart failure, however, these drugs should be avoided.

Myth 6: Non-selective NSAIDs are not associated with increased CV risk

Several publications have raised concerns that coxibs may be prothrombotic and increase the risk of acute myocardial infarction (AMI). This has arisen because of a theoretical possibility that selective COX-2 inhibitors may affect the balance between prothrombotic and antithrombotic prostanoids [72], [85]. The possibility of a cardiovascular (CV) hazard of COX-2 inhibitors was first hypothesized during late-stage clinical development, when a marked depression of prostacyclin I₂ (PGI₂) biosynthesis (similar to that achieved by t-NSAIDs) was observed in healthy volunteers. Unlike ibuprofen and indomethacin, neither celecoxib nor rofecoxib inhibited COX-1-derived thromboxane A₂ (TXA₂) and thus did not affect platelet aggregation. Because of this, it was suggested that COX-2-selective inhibition could be prothrombotic in patients with vascular disease [86]. The patient population most likely to use chronic anti-inflammatory therapy, the elderly, is also at highest risk for atherosclerotic disease. In these patients, COX-2-specific inhibitors might lead to loss of the homeostatic platelet inhibition normally induced by PGI₂ and consequently promote or enhance platelet activation. Conventional t-NSAIDs, which inhibit the synthesis of both PGI₂ and TXA_2, should not pose this theoretical risk.

The first clinical signal of a CV hazard arose from a subanalysis of the VIGOR trial [24], which demonstrated a significant increase in the risk of AMI for rofecoxib relative to naproxen. The absence of a placebo group in this trial and the low event rate in this subgroup analysis make interpretation of these findings difficult. Moreover, rofecoxib was given at twice the recommended dose, and the population included rheumatoid arthritis (RA) patients (none of whom took low-dose aspirin for CV protection). Possible explanations include an increased risk of AMI for rofecoxib, a cardioprotective effect of naproxen, or a combination of both. Alternatively, the findings of the VIGOR trial with respect to AMI may have simply occurred by chance and neither rofecoxib nor naproxen truly affects the risk of‘ AMI. A thoughtful review to which the reader is referred discusses these issues in detail [87].

Subsequent to the publication of VIGOR results, Mukherjee and colleagues [88] extended the CV safety concerns to celecoxib and potentially all selective COX-2 inhibitors. After this report, several analyses or meta-analyses (for review see [89], [90]) provided evidence for or against an increased CV risk during selective COX-2 inhibitor therapy. The results of three different studies [10], [11], [91] documenting an increase in AMI risk after rofecoxib, celecoxib, and parecoxib/valdecoxib confirmed that CV adverse events with these drugs is a class- rather than a molecule-dependent effect, and that the risks may be dose-related. Meanwhile the National Institutes of Health decided to stop the ADAPT trial because of an increased incidence of AMI in patients treated with naproxen [92]. These results, unexpected in the light of the platelet antiaggregant effect of naproxen [93], suggested that the CV risks of coxibs are similar to those seen with t-NSAIDs. Indeed, five subsequent observational studies [83], [94–97], two systematic reviews of observational studies [98], [99], and a meta-analysis of randomized trials [12] concluded that both t-NSAIDs and COX-selective NSAIDs share the same CV risks, i.e. an increase in AMI, CHF, and risk of sudden death. The lack of cardioprotective effect of naproxen and the inability of low-dose ASA to counterbalance the CV effects of coxibs, pointed out in the Hippisley-Cox study [94], challenged the hypothesis that the increase in CV risk of selective COX-2 inhibitors could be due to their ‘prothrombotic’ activity. However, a recent large (1.9 million patients) inception cohort study [100] reported that concomitant use of aspirin reduced the risk of CV events when given with rofecoxib, celecoxib, sulindac, meloxicam, and indomethacin, but not when given with ibuprofen. If confirmed by additional studies, these findings suggest that COX-selective NSAIDs are preferable to t-NSAIDs in rheumatic patients with CV risk who require low-dose aspirin [101].

Although the combined clinical experience is suggestive of an increased CV risk with coxibs in situations of elevated thrombotic risk[102] and a gradual increase of risk over extended periods of exposure [103], the speculative thrombogenicity of selective COX-2 inhibitors has not been documented in clinical trials to date. Indeed, studies showing an increased AMI risk should be interpreted with caution, because affected patients had multiple risk factors for hypercoagulability, and the association between coxib treatment and thrombotic episodes could have been only temporal rather than causal in nature. The importance of underlying risk factors has been clearly emphasized by the Nurses’ Health Study, which examined the influence of t-NSAIDs on the risk of major CV events (non-fatal AMI, fatal coronary heart disease (CHD), non-fatal and fatal stroke) [96]. While confirming the elevated CV risk of frequent (>22 days per month) t-NSAID use, the study revealed that the risk was particularly evident among current smokers (in whom endothelial dysfunction triggers platelet adhesiveness and aggregation [104]) and was absent among those who had never smoked (Table III). Patients with pre-existing CV disease are obviously at the highest risk, but a randomized controlled trial has not been conducted in this population. However, a recent cohort study [105] confirmed that both rofecoxib and celecoxib are associated with an excess risk of AMI for current users with a history of myocardial infarction. Interestingly, in this study only rofecoxib was associated with increased AMI risk in patients without a previous event.

Table III.  CV risk in t-NSAID users: effect of smoking and frequency of drug intake [96].)

	Relative risk	95% Confidence intervals
t-NSAIDs	1.44	1.27–1.65
Current smoker	1.82	1.38–2.42
Past smoker	1.58	1.28–19.5
Never smoked	1.11	0.88–1.41
>15 packets/week	1.86	1.27–2.73

The clinician should therefore not forget that poor long-term control of CV risk factors, such as hypertension, dyslipidemia, diabetes, smoking, and obesity, is more deleterious in terms of CV mortality than the administration of a t-NSAID. Both COX-selective and t-NSAIDs should be used with caution in patients with underlying prothrombotic states (in whom low-dose aspirin or appropriate anticoagulant treatment should be used as clinically indicated) and/or concomitant risk factors.

Myth 7: Corticosteroids or anticoagulants increase GI risk in combination with NSAIDs, but it is safe to co-prescribe NSAIDs with other medications

The interactions of NSAIDs with angiotensin-converting enzyme (ACE) inhibitors, angiotensin type 1 (AT₁) receptor antagonists, diuretics, and selective serotonin reuptake inhibitors (SSRIs) are widely underappreciated. Apart from the attenuation of the effects of antihypertensive drugs, coxibs and t-NSAIDs increase the risk of renal impairment when given concomitantly with ACE inhibitors/AT₁ receptor antagonists and diuretics [106].

Selective serotonin reuptake inhibitors (SSRIs) are associated with an increased risk of bleeding [22], [107], leading to hospital admission [108]. Gastrointestinal bleeding may be generalized [109] or localized to the upper gastrointestinal tract [110–112]. Other studies, however, did not find particular association with bleeding [113], [114].

SSRIs may impair platelet aggregation, and depletion of serotonin is postulated as the most likely mechanism [115]. The risk of bleeding is further increased when SSRIs are taken concomitantly with t-NSAIDs or aspirin. Observational studies show a RR for UGIB ranging from 1.3 to 3.6 for SSRIs alone, RR 12.2 to 15.6 for users of SSRIs and t-NSAIDs together, and RR 5.2 to 7.2 for SSRIs and aspirin users [22]. Furthermore, the risk ratio for gastrointestinal adverse effects overall was 12.4 [116]. A recent meta-analysis of 4 observational studies [117] involving 153,000 patients showed an OR of 2.36 (95% CI 1.44–3.85) for SSRI-associated UGIB, which increased to 6.33 (95% CI 3.40–11.8) with concomitant t-NSAIDs use. In patients aged over 50 years with no UGIB risk factors, the NNH per year is 411 for SSRIs alone, and 106 with concomitant t-NSAIDs use. Analysis of 101 spontaneous reports shows that UGIB occurred after a median of 25 weeks with SSRIs; ~67% of these patients were on t-NSAIDs [117].

Although bleeding with SSRIs is rare, it is increasing and estimated to occur at a frequency of 1 in 100 to 1 in 1,000 patient-years of exposure to SSRIs, with the elderly in the higher range [115]. SSRIs are the most commonly prescribed antidepressants, and excess GI bleeding attributable to SSRI use was assumed to be 3.1 per 1000 treatment years, increasing to 4.1 per 1000 treatment years among octogenarians and to 11.7 per 1000 treatment years among those with prior UGIB [118]. The association has been an increasing concern since these agents are widely used—especially in the elderly [119], where NSAIDs already carry a considerable GI risk.

Physicians should be aware of SSRI-induced bleeding, and a preventive strategy should be considered in SSRI users at risk, especially the elderly or those with a history of GI bleeding, and those taking t-NSAIDs or aspirin and/or clopidogrel [22]. Although concomitant SSRIs and t-NSAIDs strongly increase the risk of gastrointestinal events and should be avoided [120], no study has investigated whether patients already on a SSRI or other anti-depressant and requiring an anti-inflammatory drug should be prescribed a t-NSAID or a coxib. The role of gastric mucosal protection with a PPI has similarly not been explored in patients taking SSRIs. However, a recent nation-wide, register-based, matched, case-control study in Finland found that the adjusted odds ratio (AOR) for GI bleeding decreased with the type of concomitant NSAID, with AOR 4.66 (3.48–6.24) for concurrent use of SSRIs with t-NSAIDs, 4.17 (2.44–7.12) for SSRI and semi-selective NSAIDs (nimesulide, nabumetone, meloxicam, and etodolac), and 2.56 (1.37–4.76) for COX-selective NSAIDs [121]. This suggests that a coxib rather than a t-NSAID may decrease the bleeding risk in SSRI users.

Risk assessment should be based on the individual patient, and physicians should consider any concomitant drug use in current SSRI or t-NSAID users. They must judge both risks and benefits in relation to CV, GI, and neuropsychiatric issues. SSRI-associated bleeding risk will probably be decreased, but not abolished, with acid suppression, but the evidence is lacking. Moreover, aspirin or clopidogrel are also commonly prescribed in the elderly, and concomitant SSRIs with aspirin are a concern; close monitoring is needed, and a PPI may be required. Prospective studies are needed to determine whether SSRIs increase bleeding risks when used long-term concomitantly with aspirin/clopidogrel, and whether co-prescribed PPIs decrease risk in high-risk patients.

Conclusions

The rapid growth of information concerning the effects of COX inhibitors in the different organ systems has increased the complexity of clinical management of patients who need NSAIDs. Current evidence often challenges historical myths that may perpetuate habits and ‘beliefs’ in clinical practice. We have reviewed some of these ‘myths’ and drawn the following evidence-based conclusions that should help guide physicians who manage patients with pain:

• The evidence from both RCTs and observational studies indicates that the risk of GI complications is present with both short-term and long-term use of t-NSAIDs from the very first dose. Therefore, current strategies to prevent GI complications should be implemented regardless of the duration of therapy.

• The most frequent GI adverse effect of t-NSAID therapy is abdominal pain or dyspepsia, which often leads to interruption of treatment and serious worsening of the quality of life. Current evidence indicates that coxibs are associated with a lower incidence of this adverse effect than t-NSAIDs. Albeit not overwhelming, this advantage could be clinically relevant in the individual patient.

• Traditional-NSAIDs induce damage in the upper GI tract. The growing evidence indicates that they also induce a wide range of lesions, damage, and complications in the small and large bowel. Whereas PPIs combined with t-NSAIDs and coxibs are associated with a decreased risk of upper GI toxicity, the strategies to reduce NSAID damage in the lower GI tract have not been defined. Current evidence suggests that coxib therapy may be the preferred option, but further studies are needed.

• A substantial number of patients who need NSAIDs are also taking low-dose aspirin for cardiovascular prophylaxis. Aspirin increases the risk of upper gastrointestinal complications when taken concomitantly with either t-NSAIDs or COX-2-selective inhibitors. The current challenging question is whether the combination of low-dose aspirin and COX-2-selective inhibitors is a better therapeutic option than the combination of a t-NSAID with low-dose aspirin. While there are no clinical trials that have specifically addressed this question, available large GI outcome studies have shown that when aspirin was allowed, a coxib plus aspirin was associated with a non-significantly lower ulcer rate and lower ulcer complications rate when compared with t-NSAIDs. Moreover, some, but not all, observational studies have shown that a coxib plus low-dose aspirin was associated with lower GI risk than a t-NSAID plus low-dose aspirin. Thus, from a GI perspective, the combination of aspirin plus a coxib seems the preferred option compared with aspirin plus a t-NSAID, for patients at high GI risk who require aspirin for CV prophylaxis. Future RCTs specifically designed to compare the GI and CV safety of a coxib plus aspirin and a t-NSAID plus aspirin are awaited to confirm these conclusions.

• The renovascular effects of NSAIDs are well known. Current evidence suggests that t-NSAIDs and coxibs have a similar incidence of these adverse effects, but with molecule-specific quantitative differences between the various drugs. Blood pressure should be monitored and managed appropriately in patients on NSAIDs. In patients with severe congestive heart failure NSAIDs should be avoided.

• The increased cardiovascular risk of COX-2-selective inhibitors has been well documented in RCTs and observational studies. Whereas this risk may be different according to dose and patient base-line cardiovascular risk, more recent evidence points out that at least some, if not all, t-NSAIDs may also increase that risk. COX-selective and non-selective t-NSAIDs should be used with caution in patients with underlying prothrombotic states and/or concomitant cardiovascular risk factors.

• In t-NSAID users great attention should be paid to drug-to-drug interactions, especially with ACE inhibitors, AT₁ receptor antagonists, and SSRIs. Since these last drugs are associated with an increased risk of UGIB, their combination with t-NSAIDs is of particular concern, and close follow-up is needed. Limited evidence suggests that co-administration of a PPI may be required and that a coxib could represent a safer alternative to t-NSAIDs.

Acknowledgements

This work was carried out thanks to an unrestricted educational grant from Pfizer Europe, who supported the meetings amongst the authors. The Company did not have any role in design, planning, or execution of the review, or in writing the manuscript. The terms of the financial support from Pfizer included freedom for the authors to reach their own conclusions and an absolute right to publish the results of their work, irrespective of any conclusions reached.

CS and AL were involved with the original concept and planning of the review. Each author performed searches and led on data extraction, analysis, and preparation of a given section of the manuscript. RH and CS co-ordinated the writing and prepared the first and subsequent versions of the whole paper. All authors read and approved the final manuscript. Declaration of interest: The authors have received consulting and/or lecture fees from pharmaceutical companies and other organizations. The authors have received research support from charities and government sources at various times. No author has any direct stockholding in any pharmaceutical company. The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.