Safe prescribing of non-steroidal anti-inflammatory drugs in patients with osteoarthritis – an expert consensus addressing benefits as well as gastrointestinal and cardiovascular risks

Each statement is followed by the ‘grade’ of supporting evidence and the results of the two votes: i) agreement with the statement and ii) agreement with the grade of evidence followed by a brief summary. A consensus was considered reached when more than 75% of participants voted a, b, or c (agree strongly, agree moderately, or agree mildly).

Statement 1:OA impacts quality and quantity of life; it should therefore be treated appropriately.Level of Agreement: Strong Agreement (vote statement: a, 48.5%; b, 36.4%; c, 12.1%; d, 3%). Level of Evidence: B (vote grade: A, 42.4%; B, 48.5%; C, 9.1%).

Pain strongly and negatively impacts QoL, with increasing evidence that it may also impact quantity of life. Chronic pain interferes with the domains of physical and social functioning, emotional and mental health, energy, vitality, and general health [76]. In addition, it has been shown to be indirectly associated with a reduced cumulative survival due to CV and respiratory causes [77].

QoL is adversely impacted by OA and this has been shown in large, comprehensive, cross-sectional, or longitudinal studies [78], and this is supported by many other studies confirming the same conclusions. OA patients suffer a spectrum of symptoms, which impair normal daily functions including sleeping, walking, climbing stairs, opening containers, preparing food, or self-caring to variable extents [78]. There is evidence that chronic musculoskeletal disorders, including OA, have an overall greater adverse impact on QoL than other chronic diseases such as CV, respiratory, cerebrovascular, and neurologic disorders, GI diseases, and cancer [79].

Three large, comprehensive, longitudinal studies confirm the effect of OA on longevity [77,80-82]. Reduced physical function, especially walking disability, is a major risk factor for mortality in OA patients [81]. Thus, physical activity is good for health [80]. The potential impact of disability on longevity has also been demonstrated in other musculoskeletal inflammatory conditions. In a cohort of patients with ankylosing spondylitis followed for up to 33 years, the highest mortality was associated with greater disability and with non-use of NSAIDs [83].

Large observational studies and analyses of RCTs or pooled data confirm the important benefits of treating OA [84-86]. Treatment of OA improves physical performance and is beneficial to several domains of arthritis function, including pain, stiffness, function, and global ratings of performance. In addition, it improves the quality of sleep [87-89].

Statement 2:Many OA patients requiring NSAID therapy are not being treated appropriately according to their GI risk profile.Level of Agreement: Strong Agreement (vote statement: a, 75.0%; b, 21.9%; c, 3.1%). Level of Evidence: A (vote grade: A, 68.8%; B, 28.1%; C, 3.1%).

There is convincing evidence supporting the use of gastroprotective strategies in at-risk patients treated with NSAIDs, which has been adopted by different guidelines worldwide [90,91]. Patients at increased GI risk can be identified by a history of complicated or uncomplicated ulcers, advancing age (patients older than 60 to 65 years, depending on guidelines), use of concomitant medication (especially low-dose aspirin or anticoagulants), and the presence of H. pylori infection [61]. NSAID-treated OA patients with risk factors can be exposed to inappropriate therapy as a result of not receiving gastroprotective therapy, not being adherent to the prescribed therapy, or getting non-indicated prevention strategies. Diverse studies with different methodological approaches in different cohorts of patients have reported important findings in this regard. Very low rates of prescription of gastroprotective therapies according to national or international guidelines have been reported, although these rates have increased progressively [92-96]. In the Netherlands, correct prescription rose from 6.9% in 1996 to 39.4% in 2006 in high-risk NSAID users, whereas over-prescription rose from 2.9% to 12.3% [97].

Similar rates were reported in a cross-sectional study of patients prescribed NSAIDs in the United States of America, where only 27.2% of high-risk patients were prescribed a gastroprotective compound according to guidelines. Among patients from VA hospitals with at least two risk factors, adherence to guidelines was 39.7%; among those with three risk factors, adherence was 41.8%. The likelihood of adherence was further decreased if they were prescribed NSAIDs for ≥90 days [98]. Review of medical charts in one large cross-sectional study (n = 17,105) of OA patients found that, in over half of the population examined, NSAID prescriptions did not follow guidelines. Specific areas, where the recommendations were not followed or were overlooked, were in patients with both high GI and CV history (74% inappropriate) and in those with a high GI risk alone (49% inappropriate). However, other recommendations were followed. The study showed high rates of PPI co-prescription with ns-NSAIDs in patients with increased GI risk. However, half of patients with low GI risk and no CV history were still treated with ns-NSAIDs plus a PPI or a COX-2 selective NSAID, contrary to current guidelines [54]. A recent study in Canada has reported that concordance with guideline recommendations increased for celecoxib and decreased for ns-NSAIDs after rofecoxib withdrawal, whereas co-prescription of gastroprotective agents with ns-NSAIDs remained suboptimal, with only 45.6% of at-risk patients receiving these drugs [99].

Adherence by patients to the prescribed drug is another problem. Early reports showed that over one third of patients did not take the gastroprotective agents as prescribed [94]. More recent studies reported similar or better rates for prescription lasting <3 months [100,101], but others reported much lower adherence rates [92,93]. Appropriate prescription and optimal adherence are important for NSAID users; evidence indicates that patients with risk factors who do not receive or follow appropriate prevention strategies have an increased risk of GI complications [100,102]. A recent study involving three European databases found that, among NSAID treated patients with low adherence (<20% of the time with gastroprotection), the odds ratio (OR) was 2.39 (95% CI, 1.66–3.44) for all upper GI events and 1.89 (95% CI, 1.09–3.28) for upper GI bleeding alone when compared to patients who had high levels of adherence (>80% of the time on NSAIDs with gastroprotection) [96]. This increased risk among patients with low adherence was also found in high-risk patients who received a COX-2 selective inhibitor alone whereas guidelines recommend combination of a COX-2 selective inhibitor with a PPI [103].

Statement 3a:The efficacy of ns-NSAIDs and COX-2 selective inhibitors in pain is comparable in patients with OA.Level of Agreement: Strong Agreement (vote statement: a, 87.5%; b, 12.5%). Level of Evidence: A (vote grade: A, 87.5%; B, 9.4%; C, 3.1%).

Statement 3b:The efficacy of ns-NSAIDs and COX-2 selective inhibitors in pain is comparable in patients with rheumatoid arthritis (RA).Level of Agreement: Strong Agreement (vote statement: a, 84.4%; b, 9.4%; c, 6.3%). Level of Evidence: A (vote grade: A, 87.5%; B, 9.4%; C, 3.1%).

The discussion about whether ns-NSAIDs or COX-2 selective inhibitors should be preferred in patients with RA or OA is usually dominated by the possible GI and CV events. Although these adverse effects may have dramatic manifestations, they occur only in a minority of patients. Other issues, like tolerability, adherence to therapy, and cost/price of the drug, may also play a role. However, and perhaps of primary importance, the analgesic and anti-inflammatory effect is crucial for patients suffering from pain due to RA or OA.

In at least five high quality RCTs, a comparable effectiveness has been shown:

It is worth emphasizing that, although these drugs seem to have – as a class and at a population-based level – a comparable effect, individual treatment responses may be variable and dose-dependent. Thus, in daily practice the choice between ns-NSAIDs and COX-2 selective inhibitors will depend on the CV and GI risk profile, drug-tolerability, clinical experience of the physician with the given drug, cost/price, and pharmaco-economic considerations, as well as individual patient response.

Statement 4:NSAID use is associated with increased risk of adverse events throughout the entire GI tract; this is associated with substantial mortality.Level of Agreement: Strong Agreement (vote statement: a, 75.0%; b, 21.9%; c, 3.1%). Level of Evidence: A (vote grade: A, 87.5%; B, 12.5%).

The GI adverse effects of NSAIDs have been well documented in several studies [111-115], meta-analyses [116-118], and Cochrane reviews [119]. The majority of these studies have reported adverse events in the upper GI tract. The meta-analysis by Ofman et al. [118], which reviewed severe upper GI complications in almost 800,000 patients taking oral NSAIDs for at least 4 days, showed an OR of 5.36 (95% CI, 1.79–16.1) from 16 RCTs (versus placebo), RR of 2.7 (95% CI, 2.1–3.5) from 9 cohort studies, and OR of 3.0 (95% CI, 2.5–3.7) from 23 case control studies for severe upper GI complications, which included perforations, clinically relevant ulcers, and bleeding [118].

Similarly, low-dose (≤325 mg daily) aspirin is associated with major upper GI bleeding: the RR was 2.1 (95% CI, 1.6–2.7) in the meta-analysis by McQuaid and Laine [120] and 2.6 (95% CI, 2.2–2.9) in the large Danish cohort study, carried out by Sørensen et al. [121]. Lanas et al. [122] recently reported an OR of 1.55 (95% CI, 1.27–1.90) and found that PPI use reduced the risk of major GI bleeding (OR, 0.34; 95% CI, 0.21–0.57).

A predictable and consistent GI blood loss has been shown in healthy volunteers taking ibuprofen (800 mg t.i.d.) [115]. Bleeding was observed in 27/31 subjects (87%) and averaged 4.5 to 5.0 mL/day (SD, 12; range, 0–65 mL/day) with an onset of 3 to 5 days after starting the drug [115].

In the recent CONDOR study, a large, prospective RCT (n = 4,400), the cumulative proportion of patients with adjudicated, clinically significant events throughout the GI tract was significantly greater in those patients taking diclofenac in combination with omeprazole, despite the partial protection provided by the PPI in the upper GI tract, than in patients taking celecoxib [108].

The effect of NSAID use on mortality has been less well studied and a figure of 16,500 deaths in the US has been widely quoted since the original estimate was published [123]. However, in a more recent report of results from two concurrent Spanish studies [114], the incidence of hospital admission due to major GI events of the entire (upper and lower) GI tract was 121.9 events/100,000 person/years and admissions for upper GI complications were six-fold higher than for lower GI tract events. Mortality rates attributed to NSAIDs/low-dose aspirin use were substantial, at 21.0 and 24.8 cases/million people, respectively, and up to one-third of all NSAID/aspirin deaths could be attributed to low-dose aspirin use. Mortality rates associated with either major upper or lower GI events were similar, 5.57% (95% CI, 4.9–6.7) and 5.62% (95% CI, 4.8–6.8), respectively. However, upper GI events were more frequent. Moreover, three further studies reported mortality figures, which range from 3.8 to 11% in patients with peptic ulcer bleeding associated with NSAID use [124-126]. Furthermore, while the adverse events associated with NSAID use can lead to mortality, this could be due to associated causes other than GI adverse events (e.g., CV events) [127].

Statement 5:NSAID-induced adverse events in the lower GI tract are not prevented by PPIs.Level of Agreement: Strong Agreement (vote statement: a, 75.0%; b, 18.8%; c, 3.1%; d, 3.1%). Level of Evidence: B (vote grade: A, 43.8%; B, 56.3%).

While it is clear that PPIs reduce the development of peptic ulcer and ulcer complications in patients taking NSAIDs and/or aspirin, their beneficial effect, which is related to their antisecretory activity [128], is not expected beyond the duodenum [30]. The appreciation that NSAID-associated GI damage does extend beyond the duodenum dates back to the early 90’s, when a few observational studies and the first large RCT prevention study (i.e., the Misoprostol Ulcer Complication Outcomes Safety Assessment (MUCOSA) trial) were published [129]. In the more recent Vioxx™ Gastrointestinal Outcomes Research (VIGOR) trial, more than 40% of the NSAID-related events occurred in the lower (i.e., small bowel and colon) GI tract [130]. A systematic review [131] of 47 studies (18 randomized, 14 case-control, 8 cohort, and 7 before/after studies) found that patients taking ns-NSAIDs had significantly more adverse effects versus no NSAID use in 20/22 lower GI integrity studies (dealing with permeability, inflammation, and microscopic lesions), 5/7 visualization studies, 7/11 bleeding studies (OR, 1.9–18.4 in case-control studies), 2/2 perforation studies (OR, 2.5–8.1), and 5/7 diverticular disease studies (OR, 1.5–11.2). As reported in the Spanish studies (described under Statement 4), over the past decade there has been a progressive change in the overall picture of GI events leading to hospitalization, with a clear decreasing trend in upper GI events and a slight but significant increase in lower GI events [41].

The availability of video capsule endoscopy has allowed a precise quantification of the incidence and characterization of small intestinal damage. Indeed, available studies [132,133] show that about 75% of NSAID users display intestinal mucosal injury, with most denuded areas identified in the proximal part of the small bowel and all ulcers identified in the distal part [134].

In healthy volunteers [133,135,136] and patients [132], omeprazole did not prevent NSAID-associated intestinal damage, evaluated by video capsule and/or fecal calprotectin measurement. The failure of PPIs to protect the small bowel is due to the fact that NSAID-enteropathy is not a pH-dependent phenomenon [30]. Indeed, although inhibition of mucosal prostaglandin synthesis with NSAID use occurs along the entire digestive tract, there are significant differences between the distal and proximal GI tract in the concurrence of other pathogenic factors that may add to mucosal damage. Among the most evident are the absence of acid (which plays a pivotal role in upper GI damage) and the presence of bacteria and bile in the intestine, which may trigger specific NSAID-related pathogenic mechanisms in the distal GI tract [137]. Some recent experimental evidence [138] suggests that PPIs might actually worsen the NSAID intestinal damage by inducing dysbiosis, an adverse event repeatedly described in humans [139,140].

Like non-selective compounds, COX-2 selective NSAIDs damage the small bowel, but the frequency and severity of events are generally lower. Indeed, a systematic review found that COX-2 selective inhibitors had significantly less effect versus ns-NSAIDs in 3/4 GI integrity studies, one endoscopic study (RR mucosal breaks, 0.3), and two randomized studies (RR lower GI clinical events, 0.5; hematochezia, 0.4) [131]. The better intestinal tolerability of the selective agent, celecoxib, persisted even when the ns-NSAID was combined with a PPI [135,136]. In addition, switching RA patients on long-term ns-NSAIDs to celecoxib resulted in a significant reduction of small bowel injury [141]. While two large outcome studies (the VIGOR and CONDOR trials) showed a reduced risk of more serious events in the lower or the whole GI tract, for rofecoxib and celecoxib, respectively [108,130], this benefit was not confirmed for etoricoxib in the Multinational Etoricoxib versus Diclofenac Arthritis Long-term (MEDAL) program [142].

The impact of upper GI tract adverse effects can be mitigated by the use of PPIs. The lack of efficacy of these agents in preventing lower GI tract damage is however an unmet clinical need that requires addressing. The use of non acidic COX-2 selective inhibitors, such as celecoxib, may help to reduce the risk of damage throughout the entire GI tract, but much work is needed to define the best preventive strategy for the lower GI tract in NSAID users.

Statement 6:Celecoxib is associated with fewer adverse events throughout the entire GI tract compared to ns-NSAIDs.Level of Agreement: Strong Agreement (vote statement: a, 62.5%; b, 18.8%; c, 18.8%). Level of Evidence: A (vote grade: A, 46.9%; B, 43.8%; C, 9.4%).

A meta-analysis of RCTs showed that COX-2 selective inhibitors were associated with significantly fewer gastro-duodenal ulcers (RR, 0.26; 95% CI, 0.23–0.30) and clinically important ulcer complications (RR, 0.39; 95% CI, 0.31–0.50) than ns-NSAIDs [143]. Other studies have also shown that celecoxib is as safe to the upper GI tract as ns-NSAIDs plus a PPI [144].

It is now well known that NSAIDs induce mucosal damage to the small and large bowel, including inflammation, erosions, ulcers, bleeding, perforation, and obstruction [61,145]. Epidemiological studies have reported an increased risk of lower GI bleeding and perforation with NSAID and aspirin use [146,147]. More recently, biochemical and endoscopy capsule studies have shown a high frequency of mucosal inflammation and mucosal breaks in both healthy subjects [133] and patients taking NSAIDs [132]. Although PPIs are increasingly used to prevent NSAID-related GI adverse events, they do not protect from lesions beyond the duodenum (see Statement 5). This may explain the increasing number of hospitalizations due to complications of the lower GI tract, seen in some studies, whereas the corresponding numbers for upper GI complications are decreasing (see Statement 4) [41].

Capsule endoscopy studies [135,136] have demonstrated that, compared to ns-NSAIDs plus a PPI, celecoxib alone was associated with less mucosal damage of the small bowel in healthy volunteers. In the 6-month, double-blind, randomized CONDOR trial, in patients with OA or RA at increased GI risk [108], fewer (0.9%) subjects receiving celecoxib (200 mg b.i.d) met the criteria for the primary endpoint (clinically significant upper and lower GI events) compared to those receiving diclofenac (75 mg b.i.d.) plus omeprazole (3.8%; hazard ratio [HR] = 4.3; 95% CI, 2.6–7.0). Also, fewer patients taking celecoxib withdrew early because of GI adverse events compared to patients taking diclofenac plus omeprazole. Most of the outcome events were a drop in hemoglobin ≥2 g/dL attributed to identified lesions from the upper GI tract or to presumed small bowel lesions. These data have been confirmed in a recent 6-month randomized, open-label, blinded endpoint study that measured clinical outcomes throughout the GI tract [148]. Celecoxib use was associated with a lower risk of clinically significant upper and lower GI events than ns-NSAIDs (OR, 1.82; 95% CI, 1.31–2.55) [148]. However, a recent epidemiological study carried out in Taiwan has shown celecoxib to be associated with an increased risk of complications in the lower GI tract, similar to other NSAIDs [149]. These data need to be confirmed, since epidemiological studies are inevitably associated with different biases and confounding factors that may be difficult to control.

A relevant question is whether data on the safety of celecoxib in the lower GI tract can be extrapolated to other COX-2 selective inhibitors, since NSAID-associated mucosal damage in the lower GI tract depends not only on COX-1/COX-2 inhibition, but also on the physicochemical properties and the entero-hepatic circulation of individual NSAIDs [30].

In a 28-day trial conducted in healthy volunteers, Hunt et al. [150] found that fecal blood loss with etoricoxib (120 mg o.d.) was similar to placebo and significantly lower than that found in patients treated with ibuprofen (800 mg t.d.s.). However, the MEDAL program [142] did not confirm a decrease in lower GI complications with etoricoxib versus diclofenac.

Statement 7:The combination of celecoxib plus low-dose aspirin is associated with a lower risk of adverse events in the upper GI tract, as compared with ns-NSAIDs plus low-dose aspirin.Level of Agreement: Strong Agreement (vote statement: a, 43.8%; b, 28.1%; c, 6.3%; d, 9.4%; e, 3.1%; f, 9.4%). Level of Evidence: B (vote grade: A, 25.0%; B, 40.6%; C, 25.0%; D, 3.1%; E, 6.3%).

It is well established that low-dose aspirin exerts a life-saving antithrombotic effect, particularly in secondary prevention [151]. However, a large cohort study [121] and a meta-analysis of 35 RCTs, including 87,581 patients [122], have clearly shown that its use is associated with an increased risk of GI bleeding. This risk is further increased when aspirin is combined with clopidogrel or anticoagulants [122] and appears to be independent of the formulation of aspirin used (e.g., buffered or enteric coated) [152].

The well-known risk of upper GI events associated with the use of ns-NSAIDs is significantly enhanced by concomitant use of low-dose aspirin, a quite frequent clinical situation in older patients having both OA and CV co-morbidities [153,154]. A large, hospital-based, case-control study, performed in Spain, found that adding low-dose aspirin to ns-NSAIDs, increased the risk of upper GI bleeding by more than two-fold [155]. Similarly, a Canadian retrospective cohort study found a 62% increase in hazard ratio of hospitalization for GI events in patients taking these combinations, although the combination of celecoxib with low-dose aspirin was associated with a lower risk (HR, 0.62; 95% CI, 0.48–0.80) than the association of ns-NSAIDs plus low-dose aspirin [156].

When COX-2 selective inhibitors were examined as a class, conflicting results were reported by different studies. In the case of three epidemiological studies, one (a hospital-based, case-control) found a similar RR of upper GI bleeding when low-dose aspirin was combined with COX-2 selective inhibitors or ns-NSAIDs [155], while the others (nested case-control studies, performed in databases from the UK) observed an increased RR of upper GI complications when ns-NSAIDs or COX-2 selective agents were used together with an antiplatelet agent [157,158]. However, in a meta-analysis of four RCTs including 17,276 patients, the RR for perforation, ulcer, and bleeding between the combination low-dose aspirin/COX-2 selective inhibitors and low-dose aspirin/ns-NSAIDs was 0.72 (95% CI, 0.62–0.95) [143]. However, these were post hoc analyses and not randomised comparisons, which suggest a possible bias by patient selection.

As expected, the analysis of the risk associated with the combination of low-dose aspirin with individual COX-2 selective inhibitors also yielded different results. Indeed, with the exception of the Celecoxib Long-Term Arthritis Safety Study (CLASS) trial (where high doses, i.e., 800 mg/day, of the drug were employed) [159], all the available studies (be they epidemiological or RCTs) [156,160-162], as well as a meta-analysis of 31 RCTs (n = 39,605) [163], point to a lower upper GI risk of low-dose aspirin in combination with celecoxib than with the same antiplatelet agent combined with an ns-NSAID. However, it must be clear that low-dose aspirin potentiates the GI risk of either a selective COX-2 inhibitor or ns-NSAID and that, in patients with high GI risk, these combinations may still be harmful and gastroprotection with a PPI seems appropriate and beneficial [31].

In contrast, in the MEDAL program, the pre-specified pooled intent-to-treat analysis of the three RCTs (i.e., Etoricoxib versus Diclofenac sodium Gastrointestinal tolerability and Effectiveness trial (EDGE)-I, EDGE-II, and MEDAL), comparing etoricoxib with diclofenac in an overall population of 11,418 OA/RA patients taking low-dose aspirin the HR for overall upper GI clinical events was 0.78 (95% CI, 0.60–1.1), indicating a lack of significant difference between the two anti-inflammatory drugs [164]. Consistent with this finding, the cumulative rate of patient discontinuation due to clinical and laboratory upper GI adverse events was also not statistically different in the two treatment arms [165]. No formal studies have evaluated the GI risk of dual antiplatelet therapy in patients taking either ns-NSAIDs or coxibs.

Recent video capsule studies have shown that low-dose aspirin is also harmful to the small intestine [166-168]. The combination of aspirin with ns-NSAIDs would likely be more damaging. However, studies providing such evidence are lacking and, more importantly, the clinical relevance of these mucosal lesions need to be defined in the context of serious and life-threatening outcomes such as bleeding and perforation.

Statement 8:The risk of CV events associated with celecoxib use is similar to that associated with the use of most ns-NSAIDs.Level of Evidence: Strong Agreement (vote statement: a, 68.8%; b, 15.6%; c, 9.4%; d, 6.3). Level of Evidence: A (vote grade: A, 53.1%; B, 40.6%; C, 6.3%).

Although OA is not apparently an independent risk factor for CV disease, many OA patients are elderly and concomitant CV disease is not uncommon. In a nested case-control study (n ≥1,400,000) by Graham et al. [169], an increased CV risk was associated with rofecoxib as well as ns-NSAIDs. In RCTs, rofecoxib demonstrated a higher incidence of CV adverse events than naproxen in non-aspirin users with RA [109] or than placebo in patients with colorectal adenomas (Adenomatous Polyp Prevention on Vioxx™ (APPROVe) trial) [170]. For celecoxib, an increase in CV events was noted in patients with colorectal cancer when compared with placebo [171]. However, in these trials, the absolute numbers were low. In a RCT in Alzheimer’s disease, CV events were higher with naproxen compared to celecoxib or placebo [172]. In the CLASS study performed in OA and RA patients, there was no difference in CV events between celecoxib (400 mg b.i.d.) and ns-NSAIDs [173]. In the MEDAL trials, etoricoxib showed a similar cumulative incidence of thrombotic CV events to diclofenac [106]. A large meta-analysis of cohort and nested case-control studies also found an increased risk of CV events for all ns-NSAIDs and COX-2 selective inhibitors [46]. Other meta-analyses concluded that COX-2 selective agents and ns-NSAIDs have similar CV risk [47,174,175]. A recent network meta-analysis, which aimed to compare the CV risk of ns-NSAIDs and COX-2 selective inhibitors, found a similar CV risk between these two classes of anti-inflammatory compounds [48]. Naproxen appeared to be the least harmful, but this advantage has to be weighed against GI toxicity. It also must be considered that the absence of an increased CV risk observed in RCTs and meta-analysis with naproxen, when compared to placebo, was based on a high naproxen dose (500 mg b.i.d) [47].

Putting the CV risk in OA patients into context, the incidence of the GI risk is higher than the CV risk, and COX-2 selective inhibitors have a lower GI risk than ns-NSAIDs. Taking all the available data from clinical trials, meta-analyses, and cohort studies into account, the overall CV risk is increased for both ns-NSAIDs and COX-2 selective inhibitors. For each COX-2 selective inhibitor, however, the reduction of complicated upper GI events was numerically greater than any increase in Antiplatelet Trialists’ Collaboration events (fatal or non-fatal myocardial infarction or stroke, or vascular death) [176]. There is no evidence of major differences between ns-NSAIDs and COX-2 selective inhibitors. Hitherto, there is no published RCT which has been specifically designed to compare CV risk between ns-NSAIDs and COX-2 selective inhibitors. Results of the Prospective Randomized Evaluation of Celecoxib Integrated Safety versus Ibuprofen or Naproxen (PRECISION) study will provide important data on the comparative CV safety of celecoxib, ibuprofen, and naproxen. It is worthwhile mentioning that this trial, performed in a high CV risk population, is the first study specifically designed to assess the CV safety of anti-inflammatory drugs [177]. Unfortunately, the trial has been delayed because of slow enrollment and full results are not expected until 2016 [178].

Statement 9:COX-2 selective inhibitors do not interfere with the antiplatelet effect of low-dose aspirin. Level of Agreement: Strong Agreement (vote statement: a, 65.6%; b, 21.9%; c, 6.3%; d, 6.3%). Level of Evidence: A (vote grade: A, 77.4%; B, 16.1; C, 3.2%; D, 3.2%).

The benefits of low-dose aspirin use in secondary prevention clearly outweigh the risk [151]. However, this is not the case for primary prevention [179], where recommendations for aspirin use should be individualized, taking into account the balance between benefits and risks, as well as individual patient values and preferences [180]. Given that aspirin is a life-saving drug, discontinuing it or not adhering to the correct administration schedule enhances the risk of CV and cerebral events by more than three-fold [181,182]. This risk was magnified by up to 90-fold in patients with intracoronary stents [182].

Ns-NSAIDs, being COX-1-inhibitors, all impair thromboxane A₂ synthesis and, as a consequence, platelet aggregation, although the magnitude and duration of this effect varies amongst the different compounds [183]. With the exception of diclofenac [184,185] and meloxicam [186], almost all ns-NSAIDs can interfere with the anti-aggregant effect of aspirin [185,187-189].

While the pharmacodynamic, negative interaction between ns-NSAIDs and low-dose aspirin has been clearly established by studies in healthy volunteers and patients, the clinical consequences of such interaction are still not definitely ascertained. Indeed, available epidemiological studies provided conflicting results, with only three out of six reports showing a reduction of the cardio-protective effect of aspirin [190-195]. However, the few RCTs available are consistent in their findings that ns-NSAID use does worsen the CV outcome in patients taking low-dose aspirin. Indeed, in the Physicians’ Health Study, the regular use of these agents was associated with an increased risk (RR, 2.86; 95% CI, 1.25–6.56) of acute recurrent myocardial infarction [196]. Similarly, in the Therapeutic Arthritis Research and Gastrointestinal Event Trial (TARGET), designed to assess the GI and CV safety of lumiracoxib versus naproxen and ibuprofen, a post hoc subgroup analysis of high CV risk patients (n = 3,042) taking low-dose aspirin (60%) found a higher CV event rate in ibuprofen users compared to lumiracoxib users (1.48 versus 0.85 events per 100 patient/years) [197]. Besides the CV harm, concomitant administration of ns-NSAIDs and low-dose aspirin can be followed by stroke recurrence in patients with prior cerebrovascular events [188].

COX-2 selective NSAIDs spare platelet COX-1 activity [198] and do not affect platelet aggregation [184,198-201] or bleeding time [200]. Similarly, COX-2 selective inhibitors do not interfere with the anti-aggregant activity of low-dose aspirin both in healthy subjects [184,186,187,198] and patients with coronary heart disease [202,203]. Consistent with these results, in vitro studies on human platelets have shown that a low affinity for COX-1 and a high degree of COX-2 selectivity confers a low potential to block aspirin inhibition of platelet COX-1 [204].

Taking all the above lines of evidence into account, COX-2 selective inhibitors should represent the anti-inflammatory drugs of choice for patients taking low-dose aspirin for CV or cerebrovascular prevention [205], provided the anti-inflammatory therapy is deemed necessary and cannot be avoided with alternative therapies. It should be emphasized, however, that the European Medicines Agency’s (EMA) Committee for Medicinal Products for Human Use stated - in 2005 - that “COX-2 inhibitors must not be used in patients with established ischemic heart disease and/or cerebrovascular disease (stroke)” [206]. Therefore, the benefits (i.e., lack of blunting of aspirin protection and the anti-inflammatory/analgesic activity) should be balanced against the possible CV risks, which ultimately depend on individual patients’ clinical characteristics and co-medications.