History of Traditional Non-steroidal Anti-inflammatory Drugs
For more than 3,500 years, bark from willow trees and herbs such as meadowsweet, and many other plants, have been used as both pain relievers and antipyretics.1 In the 19th century, the active ingredient salicin was isolated and used in the treatment of acute and chronic rheumatism.2 French chemists were able to produce salicylic acid from salicin and, in 1897, an acetyl group was attached and thus acetylsalicylic acid (aspirin) was invented.1 Due to its potent anti-inflammatory properties, a use for aspirin was soon found in the treatment of rheumatoid arthritis (RA), but although effective compared to the standards of the day, treatment regimens started at 6 g/day and side-effects were significant, primarily upper gastrointestinal complaints and tinnitus.3 The quest began for a drug that was not a steroid, with the same anti-inflammatory properties as aspirin but without the side-effects. Ibuprofen was discovered in 1961 and, over the next 28 years, 42 new types of traditional non-steroidal anti-inflammatory drugs (tNSAIDs) were introduced, including butylpyrazolidines, acetic acid derivatives, oxicams, propionic acid derivatives and fenamates.4 However, the big sellers, such as ibuprofen diclofenac, naproxen and piroxicam, were all patented in the 1960s and subsequent competitors often had too little effect or too many side-effects.
Cyclooxygenase Enzymes
The effects of non-steroidal anti-inflammatory drugs (NSAIDs) are mediated by inhibition of the cyclooxygenase (COX) enzymes (Figure 1).5 The main task of COX enzymes is to convert fatty acid substrates (mainly arachidonic acid) to endoperoxides, which can then be metabolised by different downstream synthetases to various prostanoids, namely prostaglandins, thromboxanes and prostacyclins.6 The dominance of these downstream synthetases in various tissues determines which prostanoid is produced.
COX-1 is a housekeeping enzyme that is constitutively present in most cells and is involved in several important tasks, including protection of the gastric mucosa. COX-1 is present in high levels in platelets, which predominantly produce thromboxane A2 (TXA2), a compound with considerable prothrombotic properties, vasoconstrictor effects and the capacity to induce endothelial dysfunction.6 COX-1 is also prevalent in endothelial cells, where it can be upregulated by factors such as shear stress. In the endothelium, COX-1 is involved in the production of prostacyclin (PGI2), a potent vasodilator and inhibitor of platelet aggregation and lipid accumulation. The balance between PGI2 and TXA2 is critical for maintaining haemostatic balance (Figure 2A).
The cardioprotective effects of aspirin result from the irreversible and selective inhibition of COX-1, as well as a short half-life in the body. Because the platelets have no nucleus, they cannot produce new enzymes to generate the prothrombotic TXA2, whereas the endothelium will keep making new enzymes to create the antithrombotic PGI2. Moreover, platelets are relatively more exposed to the drug before hepatic first-pass metabolism than the endothelium of extrahepatic vascular beds.5 Thus, aspirin will tip the haemostatic balance in favour of PGI2 (Figure 2B). Aspirin is unique among COX-inhibiting drugs in that it irreversibly inactivates COX-1 through covalent modification of the active site, whereas other NSAIDs bind reversibly and non-covalently.7
Apart from the kidneys, brain, gut, lungs and thymus, COX-2 is scarcely expressed in most tissues in the normal state, and it is effectively lacking in platelets and vessels.5,8 However, it can be induced by several factors, such as inflammatory stimuli and pathological conditions, including cancer. At sites of inflammation, it is involved in the production of prostaglandin E2 (PGE2), which helps mediate some of the cardinal features of inflammation, such as pain, oedema and fever.8
Despite sharing the common mechanism of inhibiting both COX-1 and COX-2, tNSAIDs differ in their selectivity towards these enzymes (Figure 3). Understanding these differences in enzyme affinity is important for understanding their variability in anti-inflammatory, analgesic and side-effect profiles.
Introduction of Selective COX-2 Inhibitors
By selectively inhibiting the COX-2 enzyme, it was hypothesised that one could achieve potent anti-inflammatory effects without increasing the risk of gastrointestinal adverse events typically associated with tNSAIDs.9 In addition, according to the haemostatic balance produced by PGI2 and TXA2 described above, the potential of cardiovascular (CV) side-effects with selective COX-2 inhibitors (COXIBs) is low. Between 1993 and 1998, at least nine COXIBs were patented.
The VIGOR randomised controlled trial (RCT) was published in 2000.10 It included more than 8,000 RA patients and showed that gastrointestinal complaints were significantly lower in patients randomised to rofecoxib (Vioxx, Merck) than in those who took naproxen. A reduced frequency of MIs was also observed in the naproxen group, and the authors discussed how this could have been due to the similarities between naproxen and aspirin.10
Almost 4 years later, in September 2004, rofecoxib was withdrawn from the market after results from the APPROVe trial were presented.11 That RCT, in which 2,586 patients with previous colorectal adenomas were randomised to rofecoxib or placebo, found not only a reduced risk of recurrent adenomatous polyps but also an RR of 1.92 (95% CI [1.19–3.11]) for thrombotic events in the rofecoxib group. Only a few months later, a cumulative meta-analysis of the risk of CV events and rofecoxib was published in The Lancet.12 The authors concluded that: ‘rofecoxib should have been withdrawn several years earlier. The reasons for why manufacturer and drug licensing authorities did not continuously monitor and summarize the accumulating evidence need to be clarified’.12 After 1998, only one new COXIB (polmacoxib, 2015) has entered the market.13
In 2013, the Coxib and traditional NSAID Trialists’ (CNT) Collaboration published a meta-analysis of individual participant data from 280 RCTs of NSAIDs versus placebo (124,513 participants) and 474 RCTs comparing one NSAID directly to another NSAID (229,296 participants).14 The authors reported that COXIB users had an RR of 1.37 for any major vascular event, 1.76 for major coronary events and 2.28 for heart failure hospitalisation risk. Notably, the risk varied for different COXIBs and different dosages.14
Today, only celecoxib is available in the US, and it carries a warning from the Food and Drug Administration (FDA) of an increased risk of serious CV thrombotic events, MI and stroke.15 Furthermore, celecoxib should be used with caution in patients with heart failure. In the EU celecoxib, parecoxib and etoricoxib are sold. The European Medicines Agency (EMA) has labelled them all as contraindicated in patients with congestive heart failure (New York Heart Association [NYHA] II–IV), established ischaemic heart disease, peripheral arterial disease and/or cerebrovascular disease.16
Risk of Cardiovascular Events and Traditional NSAIDs
In a meta-analysis of RCTs published in 2006, investigators aimed to assess the risk of CV events conferred by tNSAIDs.17 Compared with placebo, the risk of vascular events was significantly increased in users of diclofenac (RR 1.63), but not significantly different for naproxen (RR 0.92) or ibuprofen (RR 1.51). The results were confirmed in a second meta-analysis of observational studies published the same year.18 Subsequent research, including two large RCTs published in 2016, demonstrated that the CV risk associated with the use of tNSAIDs, including ibuprofen, was at least as large as that for the use of celecoxib.19,20
Diclofenac, ibuprofen and naproxen were also studied in a meta-analysis from the CNT Collaboration that used individual participant data and was, in fact, a continuation of the 2006 meta-analysis of RCTs described above.14 Only diclofenac showed a significantly increased risk for major vascular events (RR 1.37), whereas the use of both diclofenac and ibuprofen was significantly associated with major coronary events (RR 1.70 and 2.22, respectively).14 All three tNSAIDs were associated with a risk of hospitalisation for heart failure (RRs of 1.85, 2.22 and 1.87 for diclofenac, ibuprofen and naproxen, respectively).
Although the CV risks are not uniform across all NSAIDs, the FDA currently advises that the data are not sufficiently strong to determine that the risk of any particular NSAID is definitely higher or lower than any other.21 The FDA also advises that NSAIDs should be labelled with information that CV risks can occur as early as in the first weeks of use and that the risks increase with higher doses. Conversely, the EMA has published a report concluding that naproxen appears to have the lowest CV risk of all the NSAIDs.22 Moreover, the EMA found diclofenac to be associated with the highest risk of CV events, with risks similar to that of COXIBs. Finally, in 2016, the Working Group for Cardiovascular Pharmacotherapy of the European Society of Cardiology published a review and position paper on the CV safety of non-aspirin NSAIDs.23 They concluded with four key positions regarding the use of NSAIDs and associated CV risks (Table 1).
Mechanisms Underlying the Increased Cardiovascular Risk in Users of NSAIDs
Results from clinical trials, biomarker studies and experimental animal models all point to COX-2 as the key to the increased CV risk in users of NSAIDs.5 Importantly, the risk does not seem to be mitigated by the COX-1 inhibition by tNSAIDs. Moreover, a meta-analysis found no clear relationship between the specificity of COXIBs and the risk of CV events.24 The underlying mechanisms have not been fully elucidated, but it appears that the PGI2–TXA2 equilibrium presented above is not the complete explanation. For instance, the two COX isoforms may have different roles in the formation of the antithrombogenic PGI2 and may, in turn, have different effects on the vasculature in different tissues.5 COX-2 is constitutively expressed in the kidneys, where it is an important regulator of renal homeostasis, including water and sodium balance.25 The enzyme participates in the modulation of the renin–angiotensin–aldosterone system, a key system governing blood pressure regulation. Thus, inhibition of COX-2 activity can affect several homeostatic processes, potentially leading to an increase in blood pressure in some individuals. Evidence suggests that constitutive COX-2 also exerts cardioprotective effects in various other tissues, such as the brain, gut and thymus, which could play roles in modulating CV risk.5
Furthermore, by inhibiting COX enzymes and thereby reducing prostanoid production, more arachidonic acid may be available for the leukotriene pathway, a shift that has been associated with proinflammatory effects and may contribute to the process of atherogenesis.26 It is also likely that the specific dosage and pharmacokinetic properties (e.g. half-life) of the NSAIDs are critical contributors to the variations seen in CV risk. Higher doses, often used for managing acute conditions, have been associated with especially high rates of CV events, as has longer-term treatment, even at lower doses.23 Moreover, the half-life of an NSAID may affect CV safety, with drugs such as celecoxib, which has a shorter half-life, possibly allowing for more rapid restoration of vascular prostanoid levels and thereby reducing the duration of CV impact. This contrasts with COXIBs like rofecoxib, which, due to a longer half-life, may have a prolonged effect on the CV system and, as a result, present a higher risk of CV events.27 Finally, the indication for treatment appears to be important, such as the treatment of patients with inflammatory joint diseases.28
Risk of Venous Thromboembolism with NSAID Use
Although the pathophysiology of venous thromboembolism (VTE) differs significantly from atherosclerotic CV disease, NSAIDs may contribute to endothelial dysfunction, a condition that can enhance hypercoagulability, one of the components of Virchow’s triad.29 Indeed, a 2015 meta-analysis reported an RR of 1.8 for VTE in tNSAIDs users and an RR of 1.99 in COXIB users.30 However, the data regarding the risk of VTE with NSAID use primarily come from observational studies, which introduces a significant challenge with confounding by indication. The indications for NSAID use overlap substantially with the risk factors for VTE, and the reliability of statistical adjustments for these confounding variables can often be questioned.
NSAIDs in Patients with Rheumatoid Arthritis
Although NSAIDs have not been shown to reduce acute-phase reactants or radiographic progression in patients with RA, and they are not recommended as primary treatments for disease modification, their role in symptom management is well established.31,32 A comprehensive systematic review and meta-analysis conducted in 2021, which evaluated data from 21 RCTs encompassing 10,503 patients, indicated that NSAIDs can effectively reduce pain and improve physical function and tender joints, as well as patients’ and physicians’ global assessment of health.33
In an analysis of the entire Norwegian population spanning from 2008 to 2017, we observed marked differences in NSAID prescription patterns between RA patients and the general population.34 Specifically, RA patients received a median of 8 NSAID prescriptions over the course of observation, with a median treatment duration of 50 days. In contrast, individuals within the general population received a median of 5 NSAID prescriptions with a median treatment duration of 20 days.34 These findings underscore the increased reliance on NSAID therapy among RA patients, with nearly 39,000 RA patients being exposed to NSAIDs during 12.2% of the observed time frame, compared with 3.0% exposure among almost 4.6 million adults without inflammatory joint diseases.34 These data highlight the importance of monitoring and managing the potential risks associated with prolonged NSAID use in RA patients.
Evidence of Less Pronounced Increase in Cardiovascular Risk with NSAIDs in Rheumatoid Arthritis Patients
Patients with RA are at an increased risk of developing CV disease due to a complex interplay between chronic systemic inflammation and a high prevalence of CV risk factors, such as smoking, hypertension, dyslipidaemia and physical inactivity.35 The persistent inflammatory state not only accelerates atherosclerosis but also exacerbates other CV disease risk factors, creating a compounded effect on the CV system. Given that the management of RA often involves the sustained use of NSAIDs, there is a particular concern regarding the potential side-effects of these medications in this high-CV-risk population.
It is important to note that the VIGOR trial, which first reported an increased CV risk with rofecoxib versus naproxen, specifically involved RA patients.10 A study from the British General Practice Research Database published in 2002 showed that RA patients with prescriptions for naproxen had a reduced risk of acute major thromboembolic CV events (OR 0.61), although the risk was significantly increased in those taking diclofenac (OR 1.68) and not statistically different in those taking ibuprofen (OR 1.05).36 In a primary care-based inception study from a cohort of patients with inflammatory polyarthritis published in 2009, survival rates were higher among NSAID users than those who were not on NSAIDs.37
The individual participant data meta-analysis by the CNT Collaboration found the adjusted risk for major CV events was not significantly increased in arthritis patients who used tNSAIDs (RR 1.15, 95% CI [0.55–2.44] for non-naproxen tNSAIDs and 0.63, 95% CI [0.28–1.41] for naproxen).14 These results were substantiated in a 2015 meta-analysis of controlled studies and RCTs published where the RR of CV events with tNSAIDs was 1.08 (95% CI [0.94–1.24]) and 1.36 (95% CI [1.10–1.67]) for COXIBs.38
In 2014, a study analysing data from the entire Danish population between 1997 and 2009 was published, which included 17,320 RA patients matched 1:4 by age and sex with 69,280 controls.39 During a median follow-up of 4.9 years, there were 6,283 CV events. The study demonstrated that although the HR for CV disease was 1.51 (95% CI [1.36–1.66]) with NSAID use in the general population, the HR was 1.22 (95% CI [1.09–1.37]) among RA patients.39 The trend for a reduced CV risk with NSAID use in RA patients compared with the general population was present for all individual tNSAIDs and COXIBs.39
Regarding the risk of pulmonary embolism (PE), we found in a nationwide Norwegian study that the incidence rate ratio (IRR) for PE in RA patients was 0.74 during tNSAID treatment, compared with 1.68 in the general population.34 The rates of PE during COXIB use were increased in both RA patients and the general population, although the effect was less pronounced in RA patients (IRR 1.75 for RA patients versus 2.80 for the general population).34
Although large-scale randomised trials provide robust evidence that minimises bias and confounding, observational studies may offer broader, real-world insights. However, it is important to note that observational studies, and meta-analyses derived from them, are unreliable because confounding by indication is inherent and uncontrollable. Despite their differences, both evidence types indicate a consistent trend: COXIB and non-naproxen tNSAID use in RA patients may carry an increased CV risk, yet it seems less pronounced than in the general population.
Possible Explanations for Less Pronounced Cardiovascular Risk with NSAIDs in Rheumatoid Arthritis Patients versus the General Population
The underlying mechanisms for the less prominent increase in CV risk in RA patients using NSAIDs compared with the general population have not been fully elucidated. However, there are several candidate theories.
First, the indications for NSAID use likely differ in RA patients because they will often take them to alleviate symptoms associated with increased disease activity, whereas other populations are more likely to use NSAIDs for non-inflammatory conditions. The generally increased CV risk in RA patients results from a complex interplay between chronic inflammation and high rates of modifiable CV risk factors.40 There is a significant increase in the CV risk in RA patients per time spent in acute disease flares.41 One may speculate that NSAID use may reduce the time with acute disease flares, thereby lowering the risk of CV events.
Second, as described earlier, COX-2 expression is heightened in inflammatory environments, which is evident in the increased COX-2 concentrations seen in inflamed synovial tissues.42 This means that different COX-1 and COX-2 expressions in RA patients versus the general population could lead to different CV effects of NSAID use.
Third, RA patients often have disabilities that hinder them from leading social and physically active lives. Therefore, alleviation of pain and stiffness with NSAIDs may help these patients participate in cardioprotective lifestyles.43,44
Balancing Risks and Benefits with NSAID Use
The therapeutic goal in managing RA and other painful conditions is to alleviate pain and, in some cases, suppress inflammation. NSAIDs, including COXIBs and tNSAIDs, can play an important role in achieving this objective. However, their use is accompanied by a spectrum of potential side-effects, including CV risk, that must be considered for each individual patient. For those with an elevated CV risk, the use of NSAIDs entails an especially careful decision. However, the incremental elevation in the risk of CV events may be deemed acceptable by some patients if it means significant relief from debilitating symptoms. Each patient’s situation is unique, with individual thresholds for acceptable risk and different potential for benefit.
The risk profiles of various NSAIDs vary, with COX-1 selectivity tending to increase gastrointestinal risks, whereas COXIBs elevate CV risks. Therefore, the choice of NSAID often involves balancing CV safety against gastrointestinal tolerance. Clinicians should consider the necessity of adding proton pump inhibitors to the treatment regimen to mitigate the risk of gastrointestinal complications, particularly in patients at elevated risk of bleeding.23
Apart from the comprehensive risk assessment of the patient, the clinician should also consider the intended use of the NSAID. For instance, there is some scientific evidence to support the clinical experience that etoricoxib and diclofenac may be more effective in pain management for certain conditions.45 Yet, due to their side-effect profiles, their use should be limited in duration. Conversely, low-dose (≤500 mg/day) naproxen could be more suitable for prolonged treatment.
Finally, patient autonomy and preference are paramount in this decision-making process. Clinicians should engage in open dialogue with patients, ensuring they are fully informed about the predictions and probabilities of associated risks versus the anticipated symptomatic relief. Through shared decision making, patients can actively participate in their treatment plans, which can increase adherence and satisfaction with care.
Conclusion
Although NSAIDs are effective in managing many common ailments in a variety of clinical settings, their safety profile, particularly regarding CV risk, necessitates careful consideration. Evidence suggests a nuanced relationship between NSAID use and CV outcomes, with risks manifesting differently across patient populations and individual NSAID profiles. Although COXIBs have been associated with higher CV risks, tNSAIDs also confer varying levels of risk that require strategic management.
Significantly, patients with RA experience a less pronounced elevation in CV risk from NSAID use compared with the general population, potentially due to the complex interplay of systemic inflammation, disease activity and pain management in RA. In order to refine our approach to NSAID use in RA and other inflammatory rheumatic diseases, more data that delineate the unique effects these medications exert on CV risk within these high-risk CV populations are warranted. Basic science and prospective clinical studies designed to isolate the variables contributing to CV outcomes in RA patients using NSAIDs will enhance our capacity to predict and manage these risks.
In summary, careful monitoring and personalised treatment are crucial when prescribing NSAIDs for rheumatic diseases to optimise benefits while minimising CV risks. Further research is imperative to guide safer NSAID use in these vulnerable populations.46