Is Increased Cardiovascular and Bleeding Risk the Price for Pain Relief?

No Free Lunch

Juan J. Badimon, PHD; Carlos G. Santos-Gallego, MD


J Am Coll Cardiol. 2020;76(5):530-532. 

The safety of nonsteroidal anti-inflammatory drugs (NSAIDs) has been the topic of much debate. As NSAIDs cause upper gastrointestinal bleeding, selective cyclo-oxygenase (COX)-2 inhibitors gained popularity due to lower bleeding events. However, there is no free lunch in medicine, and COX-2 inhibitors were associated with higher cardiovascular risk than nonselective NSAIDs.[1] This forced the withdrawal of rofecoxib and valdecoxib from the market, and guidelines discouraged NSAID use in cardiovascular patients.[2] The pathophysiological mechanism of this NSAID-induced cardiovascular risk is COX-2 suppression of prostacyclin synthesis, which shifts the prothrombotic–antithrombotic balance on endothelial cell surfaces toward thrombosis.[2]

Subsequent observational studies from Denmark confirmed increased cardiovascular events with NSAIDs,[3,4] and reported this higher cardiovascular risk occurred regardless of the time elapsed after myocardial infarction (MI)[3] and even during short NSAID exposures[4] (treatment duration of only 1–2 weeks for ibuprofen). Two clinical trials investigated which NSAID is less harmful when their use is inevitable (e.g., osteoarthritis or rheumatoid arthritis). In the PRECESSION trial,[5] celecoxib at low doses (200 mg) was noninferior to ibuprofen or naproxen with regard to cardiovascular safety, whereas celecoxib also exhibited lower risk of bleeding and of renal events than ibuprofen or naproxen. In the SCOT (Standard care vs. Celecoxib Outcome Trial),[6] cardiovascular events were also similar in celecoxib versus nonselective NSAIDs.

In this issue of the Journal, Kang et al.[7] advance this field with a real-life observational study on the risks in post-MI patients taking NSAIDs in South Korea. The authors perform a nationwide cohort study enrolling all patients with first-diagnosis of MI in 2009 to 2013. The authors take advantage of the Korean Health Insurance Review and Assessment Service (HIRA), a national organization that reviews claims for 98% of Korean residents. Prescriptions claim for NSAIDs (celecoxib, meloxicam, aceclofenac, diclofenac, naproxen, ibuprofen, dexibuprofen, and zaltoprofen) were identified.

A total of 108,232 patients with first-time MI were enrolled, and followed for 2.3 ± 1.8 years. Concomitant NSAID treatment significantly increased the risk of both cardiovascular events (hazard ratio [HR]: 6.96; 95% confidence interval: 6.24 to 6.77) and also bleeding events (HR: 4.08; 95% CI: 3.51 to 4.73), and this happened even at short treatment duration (1 week, confirming previous Danish studies[4]). The most clinically relevant finding is that, among all NSAID subtypes, cardiovascular and bleeding risks were lowest with celecoxib (HR: 4.65; 95% CI: 3.17 to 6.82 and HR: 3.44; 95% CI: 2.20 to 5.39, respectively) and meloxicam (HR: 3.03; 95% CI: 1.68 to 5.47 and HR: 2.8; 95% CI: 1.4 to 5.6, respectively). Therefore, although NSAID treatment should be avoided in patients with MI, if the use of NSAIDs is inevitable due to comorbidities, the prescription of celecoxib and meloxicam could be considered as alternative options.

The results of this paper[7] need to be considered in the context of other studies. First, these data are in accordance with the results of celecoxib not showing greater cardiovascular risk than other NSAIDs (SCOT[6] and PRECESSION[5] trials), but they expand the previous findings by including higher-risk patients (SCOT[6] only included patients free of cardiovascular disease, whereas PRECESSION[5] excluded patients with recent MI) and all NSAID doses (whereas PRECESSION[5] exclusively studied low dose of celecoxib). Second, this study breaks new ground by demonstrating cardiovascular safety of meloxicam (and not only of celecoxib), probably because of its higher COX-2 selectivity. Third, this paper[7] rejects the cardiovascular safety of naproxen, which had been suggested classically[2] and in the previous Danish data,[3,4] but that was not evident in this study;[7] in fact, both bleeding and cardiovascular risk were markedly elevated with naproxen versus other NSAIDs, as in PRECESSION,[5] and could be attributed to naproxen nonselective COX inhibition. The strongest aspect of this article is the thorough and comprehensive evaluation of the Korean population by use of the nationwide prescription claims database that includes 98% of the population (indeed, the largest contemporary cohort on this issue). Furthermore, although reimbursement databases may have intrinsic limitations because of the uncertainty of diagnostic codes, the accuracy of the ICD-10 diagnostic code for MI is 90% in East Asian countries,[8] increasing the data reliability. Second, clinical trials have severe inclusion and exclusion criteria, which limits the generalizability of the conclusions (hence the recent interest in pragmatical clinical trials); the inclusion of all the population reflects real-world clinical practice. Third, the expansion of the safety issues of NSAID to a different ethnicity is of the utmost importance given that most data about NSAIDs were extracted from Western populations but the risk of thrombosis/bleeding post-MI varies according to ethnicity.[9] In summary, this study allows the generalizability of the adverse events of NSAIDs to all patients by constituting global evidence encompassing different population groups.

The statistical analysis is strong. The potential confusion factor of the dynamic timing of the events related to NSAID intake (as retrospective analysis do not always allow to ascertain whether patients were actually taking a NSAID at the time of the event) is solved by the elegant solution of considering NSAIDs as a time-varying covariate. Importantly, the authors exclude NSAID time-periods after experiencing cardiovascular/bleeding events of interest. Finally, follow-up duration was counted separately for cardiovascular and bleeding outcome; that avoids the bias of patients with bleeding episodes being censored at that point and subsequent cardiovascular events (or lack thereof) being ignored.

The choice of endpoints deserves some discussion. First, cardiovascular mortality and all-cause death, the most relevant endpoints, were not considered because the claims database does not include death certificate information and it is impossible to differentiate between fatal and nonfatal episodes. Second, cardiovascular events include MI, ischemic stroke, or systemic arterial embolism (SAE); whereas SAEs are hardly included as outcomes in any other study. The fact that an additional sensitivity analysis excluding SAE from the primary composite outcome shows consistent results with the main analysis is reassuring.

As no study is perfect, this article also presents some limitations. First, clinically relevant adverse events of NSAIDs, such as acute renal failure or hypertension, were not assessed. Second, enrollment period was 2009 to 2013, when "older" medications (clopidogrel and VKA) were widespread; this explains the lack of information about the interaction of NSAID and contemporary state-of-the-art therapies, such as direct oral anticoagulants, prasugrel, and ticagrelor. The third limitation is the effect of information bias. This study was based solely on prescription claims, so over-the-counter NSAIDs could not be computed. Furthermore, patients do not necessarily take their medications consecutively, hence intake may last longer and patients may be exposed later than the claim database might indicate. Fourth, the nonrandomized observational nature of the HIRA database causes significant differences in baseline between the groups, and unmeasured confusion factors may have not been detected. The control for confounding by indication (i.e., that the patients taking NSAIDs were more prone to be sick than those not on NSAIDs) may not have been adequate. The condition treated with NSAIDs might be directly associated with cardiovascular risk (e.g., rheumatoid arthritis is known to increase MI[10]), and it may be this condition and not the NSAIDs that raises cardiovascular risk. Finally, this claim database does not offer clinical information to adjust (e.g., angiographic features, left ventricular function).

Ultimately, this important study[7] confirms that NSAID use post-MI significantly increased cardiovascular and bleeding risks, expands this fact to the Asian population, and provides information guiding treatment decisions. Furthermore, celecoxib and meloxicam showed the least increase in adverse outcomes versus other NSAIDs, thus they may deserve a more central place in the anti-inflammatory armamentarium. There is no free lunch in medicine, so if NSAID therapy in patients post-MI is inevitable, we have to understand which price we are paying for pain relief.