Postgraduate Medicine

ISSN: 0032-5481 (Print) 1941-9260 (Online) Journal homepage: [Link]

Cardiovascular safety of non-steroidal anti-

inflammatory drugs revisited

Chris Walker & Luigi M. Biasucci

To cite this article: Chris Walker & Luigi M. Biasucci (2018) Cardiovascular safety of non-
steroidal anti-inflammatory drugs revisited, Postgraduate Medicine, 130:1, 55-71, DOI:
10.1080/00325481.2018.1412799

To link to this article: [Link]

Accepted author version posted online: 05

Dec 2017.
Published online: 15 Dec 2017.

Submit your article to this journal

Article views: 223

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

[Link]
POSTGRADUATE MEDICINE, 2018
VOL. 130, NO. 1, 55–71
[Link]

CLINICAL FEATURE
REVIEW

Cardiovascular safety of non-steroidal anti-inflammatory drugs revisited

Chris Walkera and Luigi M. Biasuccib
a
Global Product Director, Pfizer, Walton Oaks, UK; bInstitute of Cardiology, Universita’ Cattolica, Roma, Italy

ABSTRACT ARTICLE HISTORY

Non-steroidal anti-inflammatory drugs (NSAIDs) have been widely used to treat inflammatory pain Received 15 March 2017
for decades. More recently, newer NSAIDs were developed to target the inducible isoform of Accepted 30 November 2017
cyclooxygenase (COX), COX-2, with the aim of reducing gastrointestinal toxicity. While the COX-2 KEYWORDS
selective inhibitors were effective in reducing pain and gastrointestinal harm, they soon were Non-steroidal
associated with an increased risk of adverse cardiovascular events. Initially, the view emerged anti-inflammatory drugs;
that selective inhibition of COX-2, and sparing of COX-1, was responsible for the increased cardi- cardiovascular; renal;
ovascular harm observed. However, as more data from different human populations has become osteoarthritis; COX-2;
available this view has begun to be challenged. This review examines the current understanding of COX-2 selective inhibitors
the role of prostaglandins and COX-1 and COX-2, particularly in platelets, the vasculature, and the
kidney together with an overview of the cardiovascular and renal safety of both traditional NSAIDs
and COX-2 selective inhibitors. Available data from active comparator randomized controlled trials,
including the data from the PRECISION trial investigating the long term cardiovascular safety of
patients exclusively with elevated baseline cardiovascular risk, are presented. The data, when
considered holistically, support the idea that all NSAIDs carry some level of cardiovascular risk, be
they traditional NSAIDs or COX-2 selective agents. There is also some evidence of heterogeneity of
effect with NSAIDs particularly in relation to effects on blood pressure, with no clear demarcation
based on the degree of COX-2 selectivity.

Introduction Agency (EMA) (valdecoxib and lumiracoxib).The primary

concern centered on an apparent excess of cardiovascular
Nonsteroidal anti-inflammatory drugs (NSAIDs) have been a
harm. In the case of valdecoxib, there were the dual con-
significant part of a physician’s armory in treating pain and
cerns of cardiovascular harm and serious skin reactions [8],
inflammation for decades Their use is well recognized in
while for lumiracoxib, the concern was solely focused on
current international guidelines for the treatment of both
severe adverse liver reactions [9].
osteoarthritis and ankylosing spondylitis [1–3].The effects of
COX-2 selective inhibitors have generally been studied
NSAIDs were initially attributed to inhibition of a single
more intensively and rigorously than their traditional NSAID
form of the cyclooxygenase (COX) enzyme, cyclooxygen-
class members, and a common belief became established in
ase-1 (COX-1). Identification of a second isoform – cycloox-
the literature that unopposed inhibition of COX-2 was respon-
ygenase-2 (COX-2) [4], primarily induced in response to
sible for the increased cardiovascular harm observed [10].
inflammatory processes, stimulated interest from pharma-
In this review, the authors consider both the current under-
ceutical companies in producing medicines targeting COX-
standing of the roles of the various prostaglandins, and their
2, and, by inference, inflammatory processes. By sparing the
inhibition by COX targeting NSAIDs in potentially explaining
COX-1 isoform, which is largely constitutively expressed, it
the cardiovascular harm observed, as well as covering the key
was anticipated that medicines reducing gastrointestinal
data from different sources in human populations that have
harm and with minimal effect on platelet function would
become available to address this question.
be produced [5,6].
Rofecoxib and celecoxib were licensed in the late 1990s
and were followed by parecoxib/valdecoxib, etoricoxib, and Role of prostaglandins
lumiracoxib in the new millennium. All medicines demon-
Prostaglandins are lipids derived from arachidonic acid, with a
strated similar efficacy to traditional NSAIDs, with the antici-
crucial role in maintaining homeostatic function and in the
pated reduction in gastrointestinal toxicity [7]; however,
mediation of the inflammatory response (Figure 1) [11].
they were beset by challenges with unexpected harms,
Production is dependent on the activity of both COX isoforms
with subsequent withdrawal of some medicines either spon-
[12–14] with biosynthesis relying on the conversion of arachi-
taneously by the manufacturer (rofecoxib), or in response to
donic acid to prostaglandin G2 to prostaglandin H2 (PGH2) and
regulatory requests by the US FDA and European Medicines
then thromboxane together with the four bioactive forms of

CONTACT Chris Walker [Link]@[Link] Pfizer, Walton Oaks, Dorking Road, Tadworth, Surrey KT20 7NS, UK
© 2017 Informa UK Limited, trading as Taylor & Francis Group
56 C. WALKER AND L. M. BIASUCCI

Figure 1. Pathway of prostanoid formation and illustration of functions. Prostanoid production varies widely in different tissues (not all roles are illustrated on the
figure above – e.g important roles in mediating pain and inflammation).

prostaglandins as catalyzed by tissue specific synthases: pros- and inducible functions in the body, including bone formation
taglandin E2 (PGE2), prostacyclin (PGI2), prostaglandin D2 (PGD2), and resorption [21], female reproduction [22,23], and gastric
and prostaglandin F2α (PGF2α) [11]. The four prostaglandins and protection [19,24].
thromboxane are collectively termed prostanoids and are
known to fulfill a range of functions in a number of different
tissues. Prostanoids exert their effects by activating G-protein-
coupled receptors, with the majority of prostanoids linked to
Inhibition of prostaglandins
one or two receptor subtypes; PGE2 being the more complex It is not currently clear which properties of COX inhibition may
exception with four subtypes of G-protein receptor currently result in increased cardiovascular risk. In theory, the selective
recognized [11]. inhibition of COX-2, through a reduction in the vasodilatory
PGD2 is primarily expressed in the central nervous sys- properties of PGI2 with no counterbalancing inhibition on the
tem, PGF2α plays a significant role in the female reproduc- production of vasoconstrictory thromboxane (via COX-1) could
tive cycle, in airways, and in eyes [11], while PGE2 plays a be predicted to disturb a homeostatic balance in favor of
key role in a range of tissues including kidney, vascular increased thrombogenicity [10,25]. The exact mechanisms
smooth muscle cells, and brain. The two remaining prosta- put forward have evolved over time [26,27] and suggest that
noids, PGI2 and thromboxane, are believed to have antag- there is a complex interplay of factors including COX-2 selec-
onistic functions (Figure 2). Thromboxane, shown to be tivity and drug exposure, as determined by dose, duration of
almost exclusively derived from platelet COX-1, promotes action, and duration of treatment, which act synergistically to
platelet aggregation and adhesion, and smooth muscle con- determine the level of risk with a particular medicine. Long-
traction [11,15]. PGI2 is believed to be produced by a range acting, highly COX-2 selective medicines would intuitively
of sources and causes a vasodilatory response as well as being associated with more cardiovascular risk compared
inhibiting platelet aggregation [16–18]. For the most part, with short-acting, nonselective medicines There is, however,
this review will focus on the effects of the prostaglandins a significant body of data that indicates that both of these
most closely associated with cardiovascular harm – namely types of NSAID are associated with cardiovascular harm [28,29]
PGE2, PGI2, and thromboxane. perhaps suggesting the interactions between cardiovascular
The premise set out in the introduction, namely that COX-2 risk and NSAIDs are more complex.
and COX-1 were seen as having solely inducible and constitu- A detailed review of all of the biology of cyclooxygenases
tive function, respectively, is now understood to be an over- in each individual cell type potentially contributing to cardio-
simplification, with COX-2 having a clear role in physiological vascular harm is beyond the scope of this review and is
processes [19,20]. The COX isoforms have diverse constitutive identified in other detailed, excellent reviews [18,25].

Figure 2. Antagonistic effects of thromboxane and prostacyclin in the vasculature.

 POSTGRADUATE MEDICINE 57

The following sections describe a summary of the effects of

NSAIDs on platelets, vasculature, and also the kidneys, with a
view to understanding the potential importance of prostaglan-
dins at these three important sites. While the main focus is on
the role of prostaglandin mediators, a potential role for nitric
oxide pathways is also briefly discussed in the section on the
vasculature where it is recognized to have a clear role in
cardiovascular health that may be influenced by NSAIDs.

Effects of NSAIDs on platelets

Perhaps the clearest understanding of the effects of
NSAIDs comes from research on platelets. Aspirin at low
doses (≤325 mg) selectively and irreversibly inhibits COX-1, Figure 3. COX-2 selectivity.
leading to a permanent inhibition of thromboxane-depen- a
Relative COX-2 selectivity as shown by concentration required to inhibit 80% of COX-1 and
COX-2 activity (assessed by human modified whole blood assay). Higher values indicate
dent platelet function [30,31]. As platelets are anucleate, greater COX-1 selectivity, lower values indicate greater COX-2 selectivity [50,51].
no new protein expression is possible, and a unique antith-
rombotic effect is achieved by regular low doses of aspirin
that produce a cumulative and almost complete inhibition
Effect of NSAIDs on the vasculature
of platelet COX-1, together with a much lower effect on
endothelial COX. This complete inhibition of platelet COX-1 Cells of the vascular endothelium are recognized as a major site
is useful in secondary prevention of heart disease [32]. At of prostacyclin synthesis [56]; however, the relative contributions
higher doses, aspirin inhibits both isoforms of the COX of the two different isoforms of COX, and therefore the effect of
enzyme, producing both broader anti-inflammatory effects different NSAIDs inhibiting the production of vasodilatory pros-
[33,34] and the gastrointestinal toxicity observed with tacyclin, have been much debated. A view that has predomi-
other NSAIDs. nated in the literature relies on the notion that medicines that
Traditional NSAIDs also have an effect on platelet function reduce prostacyclin without significant effect on thromboxane
[35,36]; however, their effects are reversible, transient, and levels will promote a pro-thrombotic state. Concerns were raised
related to the half-life of the nsNSAID [37,38]. In addition, early in the development of COX-2 selective inhibitors where
traditional NSAIDs particularly ibuprofen, indomethacin, tia- studies in healthy volunteers demonstrated significant reduc-
profenic acid, and naproxen), when administered prior to or tions in metabolites of prostacyclin but not thromboxane
concurrently with aspirin, have been shown to consistently [57,58] suggesting COX-2 selective inhibitors would precipitate
reduce the magnitude and duration of aspirin’s inhibitory this unbalanced reduction in prostacyclin. Support for the poten-
effect on platelet aggregation through a process of competi- tial pro-thrombotic effect of COX-2 inhibition has been provided
tive inhibition at the COX-1 active site [31,39–45]. Clinical data by a significant number of studies in mice with global [59,60] or
in this area are mixed, with some putative evidence for a tissue targeted deletions [61,62] of either the prostacyclin recep-
detrimental effect of concomitant aspirin use in conjunction tor [59] or the COX-2 gene [60–62] demonstrating an athero-
with ibuprofen [40], which was also seen in a prospective thrombotic phenotype.
analysis of clinical trial data with regular use of NSAIDs gen- Research from other laboratories has not supported this view.
erally [46], but not in a large cohort study [47] or a case– The observation that levels of prostacyclin metabolites in the
control study [48]. urine do not correlate with those in the circulation [18,63,64],
In contrast, medicines demonstrating some selectivity for while hotly debated [56,65], casts some doubt on the link
the COX-2 enzyme, such as meloxicam [49] and diclofenac between COX-2 and prostacyclin synthesis in the vasculature.
[45], have no effect on aspirin-induced inhibition of platelet Studies in humans have also suggested that both forms of COX
function in populations of healthy volunteers (see Figure 3 may play a significant role in the production of prostacyclin in
for selectivity ratios of some of the most commonly studied the normal and diseased vasculature [66,67] with Zidar et al., in
NSAIDs). Indeed, selectivity of NSAIDs is best thought of as the more recent publication, strongly challenging the ‘COX-1
a continuum with medicines being either truly nonselective constitutive’/‘COX-2 inducible’ paradigm [67]. Other research,
(e.g. ibuprofen), demonstrating some selectivity for the which focused on cell cultures [68,69] and various animal models
COX-2 enzyme (e.g. meloxicam), or being truly COX-2 selec- (including rabbits [70], rats [69], and mice [71]) has supported the
tive (etoricoxib). importance of COX-1 in prostacyclin production in conditions
For the currently available medicines classed as COX-2 ranging from mild oxidative stress through to recognized ather-
selective by regulators (celecoxib/etoricoxib), there was no osclerotic disease. It would therefore appear that the role of COX-
effect on COX-1-mediated thromboxane A2-dependent plate- 1 in the pathogenesis of atherosclerosis may have been under-
let function in healthy volunteers at supratherapeutic doses recognized with an excessive focus on the role of COX-2. It is
[35,52], no effect on the antiplatelet effects of low-dose aspirin noteworthy that for many of these authors, COX-2 is still believed
[41,53,54], and no effect on the antiplatelet effect of concur- to play an important role in the development of cardiovascular
rent aspirin and clopidogrel (celecoxib only) [55]. disease at the atherosclerotic plaque [70–72]. However, there is
58 C. WALKER AND L. M. BIASUCCI

currently a lack of clarity on the relative contributions of COX-1 on the use of NSAIDs in these populations are of an observa-
and -2 to the production of prostacyclin in the vasculature. tional nature, and will be briefly described in a later section.
The contradictory findings highlighted are almost certainly
explained, at least in part, by factors that include the differences
Effects on the kidney: clinical trials
in the organism or tissue under investigation and the presence
or absence of disease in the test system. However, an intriguing Initial short-term studies (≤2 weeks’ duration) of first-generation
possibility remains, that a causal link between prostacyclin COX-2 selective inhibitors at licensed [85] and unlicensed doses
synthesis (or even other types of prostaglandins) and COX 2 [85–87] in both young and elderly patients demonstrated similar,
inhibition that produces an increased thrombotic risk is at a site small but tangible effects on GFR, blood pressure, and creatinine
other than the vasculature. Recent work on gene expression levels, as with the traditional NSAID comparators utilized
patterns suggests that there is more COX-2 expressed in the (naproxen, diclofenac, or indomethacin). Notably, the effect on
kidney and thymus than in the vasculature itself [72] with only GFR in young patients was largely transitory, returning to baseline
one gene in heart and aorta altered in COX-2 knockout mice, levels within 3–7 days [85], while the effects on a healthy elderly
compared with an alteration of over 1000 different genes in the population were consistently observed over the short course of
kidney [73] including genes that block the synthesis of the studies [86–88]. In trials considering patients on a diet contain-
endothelial nitric oxide which is widely recognized for its pro- ing normal levels of sodium, celecoxib, rofecoxib, and traditional
tection against thrombosis and atherosclerosis [74]. NSAIDs demonstrated a similar 2–4 mm rise in systolic blood
While this work [72,73] is not without controversy, as the test pressure (SBP) [89,90] while a high licensed dose (90 mg) of the
system has recently been criticized for its relevance [27], there COX-2 selective inhibitor etoricoxib produced a 7 mm increase in
are a number of other completely independent studies in ambulatory SBP at the same day 14 time point [89].
humans which suggest that NSAIDs can have different effects Randomized controlled trials, largely of 6–12 weeks duration,
on endothelial function [75–77] with the presence of a positive in patients largely with OA and rheumatoid arthritis [RA], initially
effect one COX-2 selective inhibitor on endothelial function in supported the picture that all medicines in the NSAID class
patients with coronary artery disease [75] being absent in a carried a broadly similar level of renal harm (Table 1). Post-hoc
similarly conducted study in patients exposed to another med- pooled analysis of trials, typically of 12 weeks duration, of more
icine in the same COX-2 selective subclass [76]. The significance than 13,000 celecoxib patients and over 4700 etoricoxib patients
of these, apparently, cyclooxygenase-independent effects com- (both COX-2 selective agents were studied, largely, at the
bined with the better recognized cyclooxygenase-dependent licensed doses), demonstrated that investigator reported
effects has yet to be fully elucidated; however, it remains edema and hypertension were observed with a similarly low
plausible that there is potential interplay between the two frequency with both COX-2 selective agents and traditional
pathways as elegantly suggested over a decade ago [74]. NSAIDs (Table 1) [91–93].

Effects on the kidney: phase 4 trials and meta-analyses

Effects of NSAIDs on the kidney
Larger, longer duration phase 4 trials suggested differences in
Both COX-1 and COX-2 are expressed in the kidney where they renal effects with different NSAIDs. Retrospective analysis of
are involved in the regulation of fluid balance and blood nearly 8000 patients (median exposure, 9 months) with normal
pressure. Their exact locations in the human kidney are still baseline renal function from the Celecoxib Long-term Arthritis
of some debate – while COX-1 is constitutively found in Safety Study (CLASS) identified a significantly lower incidence
glomerular vessels, collecting ducts, and interstitial cells [78], of hypertension or new-onset hypertension with a suprather-
COX-2 expression is believed to be localized to the macula apeutic dose of celecoxib (800 mg/day) compared with ibu-
densa, podocytes, and renal vasculature [78–81]. COX-1 func- profen (2400 mg/day) (2.7% and 4.2%, respectively; p < 0.05).
tions mainly in the control of renal hemodynamics and glo- No significant difference was seen with celecoxib compared
merular filtration rate (GFR), while COX-2 is believed to with diclofenac (150 mg/day) [94]. Significantly more ibupro-
primarily affect salt and water excretion [82]. PGE2 and PGI2 fen patients (12%) required initiation of antihypertensive or
are the prostanoids with the best-defined renal functions/roles diuretic medications during the study than patients in the
[81]; with PGE2 mediating sodium reabsorption in the distal celecoxib (9.8%) or diclofenac (9.2%) groups [94].
renal tubule. Both PGE2 and PGI2 also are believed to increase Similarly, post hoc analysis of a large trial of lumiracoxib
potassium secretion, primarily through the secretion of renin (400 mg/day) compared with the maximum licensed doses of
and activating the renin-angiotensin system [81]. In the ‘non- ibuprofen (2400 mg/day) and naproxen (1000 mg/day)
diseased’ kidney of a middle-aged, healthy adult, the role of revealed small population-based change in SBP at 4 weeks
renin would not usually be critical; however, the interplay (see Table 2) [95]. Additionally, over the entire study, signifi-
between renin and COX-2 would be expected to manifestly cantly greater numbers of ibuprofen patients experienced
increase in importance in the elderly where the upregulation either investigator-reported hypertension (de novo) or aggra-
of COX-2 compensates for the reduced production of renin vation of existing hypertension (compared with the lumira-
with age [83]. This compensatory state, where the kidney is coxib arm of the study; Table 2) [95].
described as ‘prostaglandin-dependent’, is a hallmark of many A post hoc analysis of the Multinational Etoricoxib and
conditions expected in the elderly including dehydration, Diclofenac Arthritis Long-term (MEDAL) trial, reporting the relative
renal failure, and congestive heart failure (CHF) [84]. The data safety of another COX-2 selective inhibitor, etoricoxib (60–90 mg),
 POSTGRADUATE MEDICINE 59

Table 1. Effects of NSAIDs on the kidney: Individual pooled (post hoc) renal analyses of Phase 3 data for COX-2 inhibitors.
Incidence of adverse events, %
Aggravated
Reference No. of trials No. of patients Duration Treatment/comparator Peripheral edema Hypertension hypertension
Whelton et al., 2000 [91] >50 >13,000 12 weeksa Celecoxib (n = 5704) 2.1 0.8 0.6
Placebo (n = 1864) 1.1 0.3 0.4
tNSAIDsc (n = 2098) 2.1 0.7 0.3
2 yearsb Celecoxibd (n = 5155) 5.0 3.1 1.1
Lower extremity edema Hypertension
Curtis et al., 2004 92 8 4770 12 weeks Etoricoxib 60 mg/d (n = 658) 3.2 4.0
Etoricoxib 90 mg/d (n = 889) 1.5 3.4
Etoricoxib 120 mg/d (n = 472) 1.3 4.7
Placebo (n = 1491) 1.9 2.0
Naproxen (n = 1034) 2.3 2.9
Ibuprofen (n = 226) 1.8 6.6
a
Limited to the five, 12-week, North American randomized controlled trials.
b
Long-term open-label follow-up.
c
nsNSAID comparators were ibuprofen, naproxen, or diclofenac (data were not shown by individual medicine or by individual celecoxib dose).
Daniels et al., 1999 [93] assessed OA clinical trial database of 3595 patients where the incidence of lower extremity edema was: placebo, 1.1%; 12.5 mg rofecoxib,
3.6%; 25 mg rofecoxib, 3.8%; ibuprofen, 3.8%; and diclofenac, 3.4%. Rofecoxib 50 mg and nsNSAIDs had similar effects on mean DBP and SBP.
d
In analysis of celecoxib clinical trials, 78.6% of patient exposure relates to a therapeutic dose (200–400 mg daily), 12.8% of patient exposure relates to a
subtherapeutic dose, and 8.5% of patient exposure to a supratherapeutic dose (800 mg day).
tNSAIDs: traditional nonsteroidal anti-inflammatory drugs.

compared with diclofenac, demonstrated an increase in SBP of [102,103] and one 12-week study including younger
approximately 3 mm Hg with the doses of etoricoxib utilized in OA patients (mean age 62–64) with controlled hypertension
and RA patients compared with more modest increases of and type 2 diabetes [104] suggested small but significant
approximately 1.1 mm Hg with diclofenac after 4 months [98]. increases in blood pressure at a population level with
Differences in blood pressure translated into a significantly greater rofecoxib (25 mg/day) compared with celecoxib (200 mg/
incidence of hypertension and edema-related discontinuations day). The difference in ‘cuff blood pressure measurements’
with etoricoxib when considered over the entire study period in in the two 6-week trials was supported by stronger ambu-
the broader MEDAL program (Table 2) [101]. latory blood pressure monitoring data in the 12-week trial
The results of meta-analyses further corroborate the idea of of diabetic patients with controlled hypertension, with
heterogeneity among NSAIDs in terms of renal outcomes significant rises in population SBP reported for rofecoxib
(Table 2). Analysis of 114 clinical trials of different NSAIDs including but not celecoxib or naproxen (p < 0.05 in pairwise com-
over 116,000 patients indicated that celecoxib was associated with parisons between rofecoxib and other medicines; Table 3)
a significantly lower risk of both hypertension (relative risk, RR [102–104].
[95% CI], 0.83 [0.71–0.97]) and renal dysfunction, defined as a The increased frequency of blood pressure disturbance
significant change in serum urea or creatinine levels, clinically across all treatment groups in these three trials (ranging
diagnosed kidney disease, or renal failure (0.61 [0.40–0.94]), com- from 7% to 30% in Table 3 compared with, typically, 2–4%
pared with controls (either placebo, other NSAIDs, or non-NSAID in Tables 1 and 2 including all patient types) highlights the
treatments) [99]. Rofecoxib was found to be associated with a difficulty in managing patients with cardiovascular comor-
significantly increased risk of the composite renal events, as well bidities. Given that small increases in SBP at a population
as each of the component parts (renal dysfunction, hypertension, level have been shown to contribute to significantly higher
and edema) when compared with the same control groups [99]. A frequencies of cardiovascular outcomes [105,106] any
later meta-analysis that considered trial data and focused specifi- potential differences in blood pressure effects between
cally on the risk of hypertension demonstrated an increased risk of NSAIDs should not be overlooked, particularly in the man-
hypertension with COX-2 selective inhibitors (pooled) versus pla- agement of chronic conditions.
cebo but not versus traditional NSAIDs [100]. A trend was noted
toward increased risk with COX-2 selective inhibitors in the latter
comparison that was principally due to a significantly increased Cardiovascular outcomes with NSAIDs
risk of blood pressure elevation in patients treated with rofecoxib
or etoricoxib (see Table 2 for individual comparisons) [100]. These COX-2 selective inhibitors, efficacy, and safety profiles were
findings should be treated with some caution as blood pressure initially investigated in trials that either sought to better
was not a predetermined endpoint and hypertension was not understand their gastrointestinal toxicity, or to establish
always clearly defined. efficacy in an investigational setting. A large number of
trials were conducted, typically with patient populations at
low-to-moderate cardiovascular risk, and sometimes at
doses of the COX-2 selective inhibitor that would be con-
Phase 4 studies in arthritis patients with elevated
sidered supratherapeutic for the treatment of musculoske-
baseline cardiovascular risk
letal conditions. None were specifically designed with
Two 6-week studies focusing on elderly patients (mean cardiovascular outcomes as a predefined endpoint and, as
age 73–74 years) all with baseline controlled hypertension such, the relatively small numbers of cardiovascular events
 60
C. WALKER AND L. M. BIASUCCI

Table 2. Effects of NSAIDs on the kidney: Phase 4 data from large long-term studies and independently conducted meta-analyses.
Reference No. of trials No. of patients Duration Treatment Comparator Selected safety findings
Whelton et al., 2006 [94] 1 (post hoc analysis) 7968 6–15 months Supratherapeutic Ibuprofen Any hypertension adverse event (new onset or aggravation of existing hypertension),
celecoxib 2400 mg/d celecoxib 2.7% vs. ibuprofen 4.2%, p < 0.05 (celecoxib vs. diclofenac 2.7% vs. 2.6%,
(800 mg/d) Diclofenac p = ns)
150 mg/d
TARGET [95,96] 1 9156 12 months Lumiracoxib Naproxen De novo hypertension (throughout study): 5.0% lumiracoxib vs. 7.8% ibuprofen
(and BP 400 mg/d 1000 mg/d (p < 0.0001); 5.2% lumiracoxib vs. 5.4% naproxen (p = ns)
measurements Ibuprofen Aggravated hypertension (throughout study): 7.6% lumiracoxib vs. 10.3% ibuprofen
at 4 weeks) 2400 mg/d (p < 0.0002); 7.9% lumiracoxib vs. 9% naproxen (p = ns)
4 week changes in SBP: 0.57 mm Hg for lumiracoxib, compared with 3.14 mm Hg with
ibuprofen (p < 0.0001); 0.48 mm Hg for lumiracoxib compared with 1.80 mm Hg for
naproxen (p < 0.0001).
MEDAL Program [97] 3 34,701 18 months Etoricoxib 60 or Diclofenac Discontinuations due to renal dysfunction ranged from 0.4% to 2.3% with etoricoxib and
(mean) 90 mg/d 150 mg/d 0.4–1.8% with diclofenac
Discontinuations due to hypertension ranged from 2.2% to 2.5% with etoricoxib and 0.7–
1.6% with diclofenac
MEDAL trial [98] 1 23,498 BP measurements Etoricoxib 60 or Diclofenac Investigator reported BP measurements (increase in SBP): 1.7–3.4 mm Hg (mean 3.0 mm
at 4 months 90 mg/d 150 mg/d Hg) for etoricoxib vs. −0.6–1.6 mm Hg (mean 1.1 mm Hg) for diclofenac
Zhang et al., 2006 [99] 127 116,094 <1–208 weeksa Celecoxib Vs. controls (no Renal dysfunctionb, RR (95% CI): celecoxib, 0.61 (0.40–0.94); rofecoxib, 2.31 (1.05–5.07)
Rofecoxib use) Hypertension, RR (95% CI): celecoxib, 0.83 (0.71–0.97); rofecoxib 1.55 (1.29–1.85)
Peripheral edema, RR (95% CI): celecoxib, 1.09 (0.91–1.31); rofecoxib, 1.43 (1.23–1.66)
Chan et al., 2009 [100] 51 130,541 ≥4 weeks COX-2 selective tNSAIDs Hypertension, RR (95% CI): rofecoxib, 1.53 (1.34–1.75; p = 0.04), n = 9088; celecoxib, 0.89
inhibitors (0.77–1.01; p = 0.22), n = 15,053; etoricoxib, 1.52 (1.39–1.66; p = 0.01), n = 12,837
a
The majority of trials were between 6 and 13 weeks.
b
Renal dysfunction included significant changes of serum urea, creatinine, or clinically diagnosed kidney disease or renal failure.
BP: blood pressure; CI: confidence interval; DBP: diastolic blood pressure; tNSAIDs: traditional nonsteroidal anti-inflammatory drugs; ns: not significant; RR: relative risk; SBP: systolic blood pressure.
 POSTGRADUATE MEDICINE 61

Table 3. Effects of NSAIDs on the kidney: Phase 4 trials in patient populations enriched for cardiovascular risk factors.
Reference No. of trials No. of patients Duration Treatment Comparator(s) Selected safety findings
Whelton et al., 2001 [102] 1 810 6 weeks Celecoxib 200 mg/d Rofecoxib 25 mg/d Edema incidence, celecoxib 4.9% vs.
rofecoxib 9.5%, p = 0.014
Increase in SBPa, celecoxib 11% vs.
rofecoxib 17%, p = 0.032
Whelton et al., 2002 [103] 1 1092 6 weeks Celecoxib 200 mg/d Rofecoxib 25 mg/d Edema incidence, celecoxib 4.7% vs.
rofecoxib 7.7%; p < 0.05
Increase in SBPa, celecoxib 6.9% vs.
rofecoxib 14.9%; p < 0.01
Sowers et al., 2005 [104] 1 404 12 weeks Celecoxib 200 mg/d Rofecoxib 25 mg/d Significant increase in mean
Naproxen 500 mg twice daily population SBP with rofecoxib
(130.2 to 133.9 mm Hg; p < 0.001)
but not celecoxib (131.8–
132.2 mm Hg) or naproxen
(133.7–133.0 mm Hg)
Progression to ambulatory
hypertension from controlled
hypertension was significantly
greater with rofecoxib than with
celecoxib (30% vs. 16%; p = 0.05)
a
An increase in SBP of >20 mm Hg and an absolute value >140 mm Hg.
SBP: systolic blood pressure.

observed should be seen as hypothesis generating rather treated with rofecoxib (25 mg/day) and celecoxib (400 and
than hypothesis testing. 800 mg/day) in the APPROVe and APC trials, respectively
[107,108]. The initial trial publications reported cardiovascular
endpoints using the well-recognized Anti-Platelet Trialists’
Placebo-controlled trials
Collaboration (APTC) endpoint, or a broader composite that
Both rofecoxib and celecoxib were trialed in patients with a considered other thromboembolic events and cardiovascular
history of colorectal adenoma (Adenomatous Polyp Prevention outcomes over 3 years. APPROVe and APC trials demonstrated
on Vioxx [APPROVe] [107], Adenoma Prevention with Celecoxib an excess of cardiovascular outcomes with rofecoxib (46 events
[APC] [108], and Prevention of Colorectal Sporadic with rofecoxib vs. 26 events with placebo [107]), celecoxib
Adenomatous Polyps [PreSAP] [109]) with a view to demon- 800 mg/day (400 mg twice daily; 20 events with celecoxib vs.
strating a benefit previously seen with high doses of celecoxib 6 events with placebo [108,113]), and celecoxib 400 mg/day
[110]. All three trials were ‘placebo only’ comparisons, and (200 mg twice daily; 17 events with celecoxib 400 mg/day vs. 6
enrolled patients with a mean age of 60 years and largely events with placebo [108,113]; Table 5). The results of both
similar baseline demographics and cardiovascular comorbidities studies were further corroborated by analysis of data over
(Table 4). A fourth investigational study, the Alzheimer’s Disease 5 years, including a significant period off treatment [112,114].
Anti-inflammatory Prevention Trial (ADAPT) [111], focused on In the 3-year APPROVe trial analysis, separation of rofecoxib
an older population with a median age of 74 years, with gen- from placebo occurred from 18 months, with an earlier separa-
erally higher baseline cardiovascular morbidity as evidenced by tion (approximately 4–5 months) in a non-adjudicated com-
the higher usage of low-dose aspirin (56% compared with posite endpoint of CHF, pulmonary edema, and cardiac failure
16–31% in the polyp prevention trials; Table 4). [107]. In APC, the separation of celecoxib 800 mg/day from
All four trials were terminated early following an excess of placebo began to occur at around 12 months [108]. Notably,
independently adjudicated cardiovascular events in patients both trials demonstrated a mean increase in population SBP,

Table 4. Patient characteristics in placebo-controlled trials of COX-2 selective inhibitors.

APPROVe [107] APC [108] PreSAP [109] ADAPT [111]
Patients, N 2586 with history of 2035 with history of 1561 with adenomas 2528 (aged >70 years) with family history of
colorectal adenomas colorectal neoplasia removed Alzheimer’s disease
Mean age, years 59 60 61 74 (median)
Treatment
COX-2 selective inhibitor Rofecoxib 25 mg/da Celecoxib 400 mg/d and Celecoxib 400 mg/d Celecoxib 400 mg/d
800 mg/d
Supratherapeutic dose No Yes No No
Low-dose aspirin users 16–17% 29–31% 17% 56%
Medical history
Tobacco use 22% 14–18% 23–24% 56%
Hypertension 34–36% 39–42% 35–39% 43%
Use of lipid lowering 26–29% 27–28% 17% NR
drug
Diabetes 9% 9–10% 10–11% NR
a
The study initially included an additional treatment arm of rofecoxib 50 mg/d. The protocol was amended shortly after the trial began to include patients taking
cardioprotective aspirin and this treatment arm was removed from the study at this time. Those patients already treated with rofecoxib 50 mg/d were not
included in any analysis [112].
NR: not reported.
62 C. WALKER AND L. M. BIASUCCI

Table 5. Cardiovascular events in placebo-controlled clinical trials of NSAIDs.

Trial Treatment No. of patients Safety outcome Incidence RR/HR (95% CI) vs. placebo
APPROVe [107] Rofecoxib 25 mg/d 1287 Thrombotic adverse events 46 events (3.6%) 1.92 (1.19–3.11)
Placebo 1299 26 events (2.0%)
APC [108,113] Celecoxib 800 mg/d 671 Composite APTC endpoint 20 events (3.0%) 3.4 (1.4–8.5)
(supratherapeutic dose)
Celecoxib 400 mg/d 685 17 events (2.5%) 2.8 (1.7–7.2)
Placebo 679 6 events (0.9%)
PreSAP [109] Celecoxib 400 mg/d 933 Serious cardiovascular events 23 events (2.5%) 1.30 (0.65–2.62)
Placebo 628 12 events (1.9%)
ADAPT [111] Celecoxib 400 mg/d 726 Composite APTC endpoint 28 events (5.54%) 1.10 (0.67–1.79)
Naproxen 440 mg/d 719 40 events (8.25%) 1.63 (1.04–2.55)
Placebo 1083 37 events (5.68%)
AE: adverse event; APTC: Anti-Platelet Trialists’ Collaboration; CI: confidence interval; CV: cardiovascular; HR: hazard ratio; RR: relative risk.

as measured anecdotally at 1- and 3-year time intervals, with myocardial infarction [MI], cardiac arrest, stroke, and pulmon-
the largest increase in blood pressure compared with placebo ary embolism) during short-term use (≤14 days) of the inject-
occurring in the celecoxib 800 mg/day group (>5 mm Hg able prodrug parecoxib and its oral active metabolite
increase in SBP at 3 years) [107,108,113]. (valdecoxib) compared with standard of care [115,116].
In contrast, no significant difference was observed at the Similarly the antiproliferative potential of celecoxib identi-
population level [113] between celecoxib 400 mg/day (adminis- fied in vascular smooth muscle cells [117] has encouraged the
tered once daily) and placebo in the PreSAP trial [109]; however, investigation of its potential in the prevention of restenosis
investigator-reported hypertension was significantly more com- after angioplasty [118–120]. While there were clear benefits
mon in the celecoxib arm (RR [95% CI], 1.35 [1.08–1.69]) [109]. with celecoxib (400 mg/day) in terms of a reduction in late
In the ADAPT trial, which considered a broader composite luminal loss of stents, over 3–6 months [118,119] and over
outcome, including APTC endpoints, CHF, and transient ischemic 2 years [120], the larger trial, which used a higher loading dose
attack, the incidence with celecoxib (28 events, 5.54%) was not of celecoxib initially, also demonstrated an excess of nonfatal
significantly different to placebo (37 events, 5.68%) [111]. MI and cardiac death compared with controls (6 events vs. 1
However, significantly greater risk was observed with the rela- event; p = 0.03) [119].
tively low dose of naproxen utilized (440 mg/day) compared with
placebo (8.25% vs. 5.68%, respectively; Table 5) [111].
Active comparator trials
Active comparator trials of COX-2 selective inhibitors also
Placebo-controlled trials in high-risk settings
report anecdotal safety data on cardiovascular outcomes.
COX-2 selective inhibitors have also been studied in patients Trials were conducted in patients with RA or OA (or sometimes
with greater baseline cardiovascular risk. Perioperative trials in a combined patient population; Table 6) and were generally
coronary artery bypass graft surgery (CABG) patients illu- around 9–12 months in duration. Criticism of these trials often
strated an elevated risk of cardiovascular events (mainly focus on the supratherapeutic doses of the COX-2 selective

Table 6. Patient characteristics in comparator trials of COX-2 selective inhibitors.

VIGOR [121,122] CLASS [123,124] TARGET [96] MEDAL [97]
Patients, N 8076 with RA 3987 (27% RA, 73% OA) 9156 with OA 34,701 (28% RA, 72% OA)
Mean age, years 58 60 63 63
Treatment
COX-2 selective inhibitor Rofecoxib 50 mg/d Celecoxib 800 mg/d Lumiracoxib 400 mg/d Etoricoxib 60 or 90 mg/d
Supratherapeutic dose Yes Yes Yes No
tNSAID comparator(s) Naproxen 1000 mg/d Ibuprofen 2400 mg/d Naproxen 1000 mg/d Diclofenac 150 mg/d
Diclofenac 150 mg/d Ibuprofen 2400 mg/d
Low-dose aspirin users Not permitted 21% 24% 35%
Medical history
Tobacco use NR 15% 10% 12%
Hypertension NR 39% 45% 47%
Hypercholesterolemia NR 16% NR NR
Dyslipidemia NR NR 20% 29%
Diabetes NR 8% 8% 11%
CV disease history NR 9% 10% 12%
Prior-MI NR 3% 2% NR
CHF NR 1% NR NR
CV outcomes, HR/RR (95% CI)a
APTC endpoint NR 1.1 (0.7–1.6)b 1.14 (0.78–1.66) 0.96 (0.79–1.16)
All thrombotic CV events 2.38 (1.39–4.00) NR NR 0.95 (0.81–1.11)
Confirmed or probable MI NR NR 1.31 (0.70–2.45) NR
a
HR or RR (95% CI) for COX-2 selective inhibitor compared with all nsNSAIDs in that study.
b
Defined in this study as serious cardiovascular thromboembolic events (myocardial infarction, stroke, CV deaths, and peripheral events).
APTC: Anti-Platelet Trialists’ Collaboration; CHF: congestive heart failure; CI: confidence interval; COX: cyclooxygenase; CV: cardiovascular; HR: hazard ratio; MI:
myocardial infarction; NR: not reported; tNSAID: traditional nonsteroidal anti-inflammatory drug; OA: osteoarthritis; RA: rheumatoid arthritis; RR: relative risk.
 POSTGRADUATE MEDICINE 63

agents utilized, and the long duration of study – both of which 50 events (0.55%) with the combined traditional NSAIDs [96].
are seen as inconsistent with clinical practice. Pairwise analysis of the different treatment arms also high-
The Vioxx Gastrointestinal Outcomes Research (VIGOR) trial, lighted no differences (19 events with lumiracoxib vs. 23 with
conducted in a population of more than 8000 RA patients, ibuprofen; 40 events with lumiracoxib vs. 27 events with
compared a supratherapeutic dose of rofecoxib (50 mg/day) naproxen) [96].
against the nsNSAID naproxen (1000 mg/day) for a median of The final trial analysis, MEDAL, was a pooled analysis of
9 months [121]. Gastrointestinal benefits were observed for three clinical trials of etoricoxib (60 or 90 mg/day) in patients
rofecoxib when compared with naproxen; however, the trial with OA and RA compared with the maximum licensed dose
illustrated a significantly increased risk of MI with rofecoxib of diclofenac (150 mg/day) [97]. A key endpoint was the
compared with naproxen (0.4% vs. 0.1%; RR [95% CI], 0.2 [0.1– incidence of all thromboembolic events (APTC was a second-
0.7]). Notably, patients in VIGOR were not permitted to take ary endpoint). Although the hazard ratio (95% CI) for etori-
low-dose aspirin. In a subgroup analysis excluding those coxib (320 events in all thrombotic events category) compared
patients in the trial who met the FDA criteria for use of aspirin with diclofenac (323 events in all thrombotic events category)
for cardiovascular prophylaxis there was no significant differ- was not significantly greater for either all thrombotic cardio-
ence in the rate of MI (0.2% with rofecoxib vs. 0.1% with vascular events (0.95 [0.81–1.11]) or APTC events (0.96 [0.79–
naproxen). Commentators at the time speculated that the dif- 1.16]), guidelines from the American Heart Association have
ferences could be due to an antithrombotic effect of naproxen suggested that this was not entirely unexpected given the
or a pro-thrombotic effect of rofecoxib [122]. The same pre- relatively high selectivity of diclofenac for COX-2 [126], which
liminary review of the VIGOR trial data (as taken from the FDA effectively meant that they believed the program was a con-
website) indicated that the RR (95% CI) of a thrombotic cardi- sideration of two COX-2 selective inhibitors [127].
ovascular event (including MI, unstable angina, cardiac throm-
bus, resuscitated cardiac arrest, sudden or unexplained death,
Cardiovascular outcomes: meta-analyses of
ischemic stroke, and transient ischemic attack) with rofecoxib
randomized controlled trials
treatment compared with naproxen was 2.38 (1.39–4.00) [122].
The analysis also highlighted the increase in mean population Most of the trial data considered, with the exception of the
SBP and diastolic blood pressure (DBP) with rofecoxib (SBP, data from CABG and stent settings, have been included in a
4.6 mm Hg; DBP, 1.7 mm Hg) compared with naproxen (SBP, number of meta-analyses that have used different methodo-
1 mm Hg; DBP, 0.1 mm Hg) [122]. logical techniques and endpoints to make comparisons with
Celecoxib, in contrast, was studied in a population of placebo or active comparator(s) [28,29,128]. Meta-analyses
approximately 8000 OA and RA patients (Table 6) at a typically show three consistent features: the lowest numerical
supratherapeutic dose (800 mg/day) compared with the max- risks are associated with naproxen, there is an elevated risk
imum licensed doses of two traditional NSAIDs (ibuprofen with any other nsNSAID or COX-2 selective inhibitor vs. pla-
2400 mg/day and diclofenac 150 mg/day). The trial failed to cebo, and similar risks are observed when nsNSAIDs and COX-
show a conclusive benefit in terms of the pre-specified primary 2 selective inhibitors are directly compared.
gastrointestinal outcomes [123,125]; however, a post hoc analy- Primary results from the three most comprehensive ana-
sis assessing the incidence of serious cardiovascular events (MI, lyses, two of which are from the same academic group, are
stroke, cardiovascular death, and peripheral events) did not highlighted in Table 7 [28,29,128]. The largest meta-analysis
detect any significant difference in cardiovascular events with conducted to date, considering over 200,000 person-years of
celecoxib (52 events, 1.3%) compared with the two traditional data, demonstrated a 35–40% increase in risk with all active
NSAIDs combined (49 events, 1.2%), with RR (95% CI) of 1.1 comparators compared with placebo, with the exception of
(0.7–1.6) or when individual comparisons were made between naproxen, which demonstrated no difference in risk [28].
medicines [124]. Although there was no readily available assess- Further analyses (presented in a supplementary appendix),
ment of blood pressure effects from the CLASS trial at this time, considered celecoxib as a distinct agent (rather than pooled
the commentary conducted in the prior year demonstrated a together with other COX-2 selective inhibitors) and demon-
significantly higher incidence of MI with both rofecoxib and strated that a significant risk (RR [95% CI]) existed with all
celecoxib, compared with a large placebo group in a primary doses pooled together (1.36 [1.00–1.84]) and with the
prevention setting [122]. The authors also recommended that supratherapeutic dosing regimen of 800 mg/day (2.96 [1.21–
all future clinical trials of COX-2 selective inhibitors include an 7.25]), but there was no significant risk associated with the
assessment of cardiovascular risk [122]. licensed dosing regimens of 400 mg/day (1.29 [0.81–2.04]) and
The manufacturers of lumiracoxib and etoricoxib broadly 200 mg/day (0.95 [0.03–3.00]) [28].
adopted this advice when designing later trials of their med- Another meta-analysis using a network technique to draw
icines. In the Therapeutic Arthritis Research and comparisons where no direct data exists demonstrated an
Gastrointestinal Event Trial (TARGET), comparisons between a increase in the incidence of the composite APTC outcome
supratherapeutic dose of lumiracoxib (400 mg/day) and the for all medicines; however, this was statistically significant for
maximum licensed doses of ibuprofen and naproxen yielded only lumiracoxib, rofecoxib, and ibuprofen compared with
no significant differences in the APTC composite endpoint placebo [29]. All three meta-analyses recognized that the
over the year-long study in more than 18,000 patients main component driving the increase in risk was the incidence
(Table 6) [96]. The APTC event rates were generally low, with of coronary events (fatal and nonfatal MI) with the authors of
59 events (0.65%) reported with lumiracoxib compared with the largest analysis predicting that, compared with placebo,
64 C. WALKER AND L. M. BIASUCCI

Table 7. Meta-analyses of cardiovascular events in clinical trials of NSAIDs.

Reference No. of trials Safety outcome Treatment Comparator RR (95% CI)
Kearney et al., 2006 [128] 138 Serious vascular events (myocardial Naproxen Placebo 0.92 (0.67–1.26)
infarction, stroke, or vascular death) Ibuprofen Placebo 1.51 (0.96–2.37)
Diclofenac Placebo 1.63 (1.12–2.37)
All COX-2 selective inhibitors Placebo 1.42 (1.13–1.78)
All COX-2 selective inhibitors All tNSAIDs 1.16 (0.97–1.38)
Trelle et al., 2011 [29] 31 APTC composite outcome Ibuprofen Placebo 2.26 (1.11–4.89)
Lumiracoxib Placebo 2.04 (1.13–4.24)
Rofecoxib Placebo 1.44 (1.00–1.99)
Diclofenac Placebo 1.60 (0.85–2.99)
Etoricoxib Placebo 1.53 (0.74–3.17)
Celecoxib Placebo 1.43 (0.94–2.16)
Naproxen Placebo 1.22 (0.78–1.93)
Bhala et al., 2013 [28] 474 Major vascular events (nonfatal MI, Diclofenac Placebo 1.41 (1.12–1.78)
nonfatal stroke, or vascular death) Ibuprofen Placebo 1.44 (0.89–2.33)
Naproxen Placebo 0.93 (0.69–1.27)
All COX-2 selective inhibitors Placebo 1.37 (1.14–1.66)
Celecoxib (all doses) Placebo 1.36 (1.00–1.84)
Celecoxib 800 mg/d (supra-therapeutic dose) Placebo 2.96 (1.21–7.25)
Celecoxib 400 mg/d Placebo 1.29 (0.81–2.04)
Celecoxib 200 mg/d Placebo 0.95 (0.03–3.00)
Celecoxib (all doses) Diclofenac 0.94 (0.54–1.63)
Celecoxib (all doses) Ibuprofen 1.01 (0.48–2.13)
Celecoxib (all doses) Naproxen 0.93 (0.46–1.88)
APTC: Anti-Platelet Trialists’ Collaboration; CI: confidence interval; COX: cyclooxygenase; MI: myocardial infarction; tNSAIDs; traditional nonsteroidal anti-inflamma-
tory drugs; RR: relative risk.

use of any COX-2 selective agent or diclofenac would produce events with celecoxib; 144 with naproxen; 155 with ibuprofen
an additional three major vascular events (per 1000 patients [130]. Importantly, there was no suggestion of a lower risk of
per year), of which one would be fatal [28]. Following this cardiovascular outcomes with higher doses of naproxen, in
meta-analysis, EMA effectively introduced contraindications contrast to earlier meta-analyses [28,128].
for the use of diclofenac in patients with existing cardiovas- Secondary endpoint analyses of efficacy variables from the
cular disease – harmonizing the product information with the trial did show a statistical benefit of using naproxen when
labeling for agents considered COX-2 selective. Similarly, gui- compared with either celecoxib or ibuprofen in one of three
dance was adopted by the European Society of Cardiology parameters assessed (visual analog scale); however these sta-
placing similar caution on the use of diclofenac to those tistical differences were numerically small, and far below the
medicines traditionally classified as being COX-2 selective [50]. thresholds necessary to illustrate a clinically meaningful differ-
ence [130]. Additionally, in the large sub group (~46% of
patients) stratified by low dose aspirin use, there was no
A new clinical trial data set
evidence for the ablation of the antiplatelet effect of aspirin
Recently, data from the Prospective Randomized Evaluation of in any pairwise comparison [130].
Celecoxib Integrated Safety versus Ibuprofen Or Naproxen The study has been criticized for low adherence rates to
(PRECISION) trial [129] of moderate doses of celecoxib (mean treatment and for potential imbalances in the dosing regi-
dose 209 mg/day), compared with ibuprofen (mean dose mens [131]. The adherence rates, while lower than in shorter
2045 mg/day) and naproxen (mean dose 852 mg/day) in an clinical trials where pain is the main symptom, are believed to
arthritis population with existing cardiovascular disease or at be comparable to the rates observed in clinical practice [132].
high risk of developing cardiovascular disease was published Similarly, the doses used were also representative of those
[130]. In this academically driven trial, more than 24,000 that are licensed for the indications studied in the countries
patients were studied for a minimum period of 18 months, where the trial was conducted [130]. Residual questions do
with the goal of observing at least 580 independently adjudi- remain, however, around 400 mg dose of celecoxib as this
cated cardiovascular endpoints meeting the APTC criteria in regimen was relatively untested in the PRECISION trial (the
an ITT population (analysis truncated at 30 months) and at 200 mg bd dose was only licensed for use in RA patients in the
least 420 APTC endpoints in the modified intention to treat countries where the trial was conducted) [132].
(mITT) population (analysis extends to 42 months) [130]. Intriguingly, analyses of the secondary and tertiary endpoints
Importantly, the analysis prespecified that non-inferiority do suggest that there may be subtle differences between the
would only be demonstrated if the hazard ratio did not medicines at the dosing regimens studied in terms of cardiovas-
exceed 1.12 in both populations and the one-sided 97.5% cular and renal safety [130]. At the time of writing, a prespecified
confidence interval was ≤1.33 in the ITT population and substudy of the PRECISION trial measuring ambulatory blood
≤1.40 in the mITT population [130]. pressure (PRECISION-ABPM) in a subset of around 440 patients
Non-inferiority was successfully demonstrated, with similar was published [133]. This substudy demonstrated a significant
numbers of APTC outcomes in the ITT (188 events with cel- difference in the primary endpoint of population SBP at 4 months
ecoxib [2.32% of patients]; 201 events with naproxen [2.52%]; [133]. While celecoxib decreased the average SBP measured over
218 events with ibuprofen [2.71%]) and mITT populations (134 24 h by 0.3 mmHg, ibuprofen and naproxen increased it by 3.7 and
 POSTGRADUATE MEDICINE 65

1.6 mmHg, respectively [133]. The resulting difference of diclofenac, naproxen and rofecoxib) within 7 days of initiating
−3.9 mmHg between celecoxib and ibuprofen was significant therapy, as well as for longer time frames [141]. However, the
(p = 0.009) [133], providing further support for heterogeneous data over the shorter time periods (≤7 days) may have been
effects on blood pressure of medicines in the broad NSAID class effected by protopathic bias.
[100,104].
Two recent trials using a prospective blinded endpoint
design also investigate the incidence of cardiovascular out-
Observational data from patients exclusively with
comes between celecoxib and a mixed traditional NSAID treat-
elevated cardiovascular baseline risk
ment arm in Japanese (~10,400 patients; mean follow-up
~1.9 years) [134] and European arthritis cohorts (~7300 Despite the inherent weaknesses of observational data, these
patients; median follow-up 3 years) with low baseline cardio- studies are useful to characterize risks in populations largely
vascular risk [135]. In the Japanese study, the incidences of all excluded from randomization in clinical trials before the con-
cardiovascular events are described as similar (79 in the cel- duct of PRECISION. Tables 8 (raised cardiovascular risk but no
ecoxib group; 84 in the ‘mixed NSAIDs group’), with no pre- prior event) and 7 (prior MI) demonstrate available data in this
determined statistical plan to assess non-inferiority [134]. In area. Risks observed in Table 8 are typically not statistically
the European cohort, non-inferiority was only demonstrated in significant [48,142–144], based on small samples sizes, while
the ITT population (125 APTC events with celecoxib vs. 124 the risks of the patient experiencing a second infarct (see
events in the ‘mixed NSAIDs group’), with the low numbers of Table 9) are generally higher than those shown in Table 8,
events captured in the ‘on treatment analysis’ dictating that with a 40–50% increase in risk compared with no use being
this analysis was viewed indeterminately [135]. more typical [145–149].

Observational data
Observational data in patients with CHF and acute
Observational data investigating the cardiovascular safety of
kidney injury
NSAIDs have also proved useful in characterizing cardiovascu-
lar risks as healthcare databases often contain much larger Observational data also exist focusing on hospitalization for
numbers of patients than would be evaluable in a clinical trial heart failure [150,151] and recurrent heart failure (Table 10)
setting. Caution must be employed when considering obser- [151–154]. In nearly all cases the mean age of the patient
vational findings as there is often a lack of correction for population was above 75 years. Increases in risk observed
various biases [136]. Notably, the indication for treatment are typically lower than the approximate doubling in risk
may be related to the risk of future health outcomes (indica- (compared to nonuse) reported in a meta-analysis of rando-
tion bias), or the medicines may be used to treat symptoms mized trials where the patients were typically over a decade
that are actually early manifestations of the outcome of inter- younger [28].
est (protopathic bias). While indication bias is potentially The largest study – a nested, case control considering over
avoidable by use of a control population with the disease of 92,000 hospital admissions for heart failure from data sets in
interest [137], the potentially significant challenges dictate four different European countries (Netherlands, Italy, Germany,
that we focus only on the largest of observational analyses and the United Kingdom) demonstrated current use of any
here. NSAID was associated with a 19% increase in heart failure;
Thirty case control studies considering more than 184,000 however, a more detailed analysis of the individual drug
cardiovascular events, and 21 cohort studies describing over effects demonstrated a significantly increased risk of first hos-
2.7 million individuals with different levels of baseline cardio- pitalization for heart failure with several nsNSAIDs, including
vascular risk were analyzed by McGettigan et al. [138]. ibuprofen and naproxen, and two COX-2 selective inhibitors
Significantly elevated risks of cardiovascular events were (etoricoxib and rofecoxib) but not for other individual NSAIDs
observed with rofecoxib (45% increase) and diclofenac (40% including celecoxib [151].
increase) compared with no use of NSAIDs [138]. Risks were A large Danish study considering recurrent CHF illustrated
also significantly higher for ibuprofen, celecoxib, and significant increases in risk, ranging from 16% to 40%, when all
naproxen (18%, 17%, and 9%, respectively, compared with doses of traditional and selective NSAIDs were considered. The
no use); however, low dose ibuprofen (generally <1200 mg risk of recurrent heart failure was highest for doses of diclofe-
day in the dozen studies analyzed) was not associated with an nac > 100 mg and doses of rofecoxib > 25mg (HR 1.42 and
increased cardiovascular risk [138]. 1.86, respectively, see Table 10) [154].
Subsequent analyses focusing on the endpoint of MI [139], Observational studies have also provided information on
or stroke [140], demonstrated an increased risk of either event the risks of hospitalization due to acute kidney injury [155–
with use of rofecoxib, and possibly diclofenac, but not other 158]. Heterogeneity is seen across different NSAIDs, with no
NSAIDs (including celecoxib) [139,140]. In the analysis consid- clear correlation with increased COX-2 selectivity [155–158],
ering MI, there was little or no increase in risk associated with and renal outcomes are more consistently observed with
exposures of less than 30 days with any medicine evaluated higher doses [157] or multiple use [156] of NSAIDs, with
[139]. In contrast, a 2017 analysis using individual patient data medicines that are nonselective (ibuprofen) [156,158] and
from large patient data sets suggested that risks of acute MI highly selective (rofecoxib) appearing to be associated with
occurred with all NSAIDs studied (celecoxib, ibuprofen, some of the highest risks of harm.
 66

Table 8. Cardiovascular risk of tNSAIDs and COX-2 selective inhibitors in a high cardiovascular risk community population.
Celecoxib Naproxen Rofecoxib Diclofenac
Observational study n Effect estimate 95% CI n Effect estimate 95% CI n Effect estimate 95% CI n Effect estimate 95% CI
Johnsen et al., 2005 [142]a 56 1.15b 0.85–1.54 18 1.38b 0.82–2.34 92 1.58b 1.24–2.00 – – –
Fischer et al., 2005 [48]c,d – – – 51 0.95e 0.72–1.26 – – – – – –
Varas-Lorenzo et al., 2009 [143]
Patients with hypertension 66 1.11e 0.82–1.50 13 1.22e 0.61–2.44 39 1.24e 0.83–1.85 39 0.94e 0.63–1.42
Patients with coronary heart disease 54 1.00e 0.68–1.46 9 1.35e 0.53–3.42 35 1.53e 0.92–2.53 34 0.95e 0.60–1.49
Shau et al., 2012 [144]f NR 1.81e 1.07–3.05 NR 1.30e 0.81–2.10 – – – NR 1.33e 1.11–1.60
C. WALKER AND L. M. BIASUCCI

a
High cardiovascular risk was defined as at least one previous hospitalization for cardiovascular disease, hypertension, diabetes mellitus, or coronary revascularization or at least one previous prescription for cardiovascular
drugs, insulin, or oral hypoglycemic drugs.
b
Risk ratio.
c
High cardiovascular risk was defined as patients with diagnosed diseases that predispose them to acute myocardial infarction.
d
The study also reported the odds ratio (95% CI) with ibuprofen, 1.00 (0.84–1.19), n = 140.
e
Odds ratio.
f
High cardiovascular risk was defined as patients with a diagnosis of hypertension.
CI: confidence interval; tNSAIDs; traditional nonsteroidal anti-inflammatory drugs; NR: not reported.

Table 9. Cardiovascular risk of tNSAIDs and COX-2 selective inhibitors in a post-myocardial infarction community population.
Celecoxib Ibuprofen Naproxen Rofecoxib Diclofenac
a a a a
Observational study n Effect estimate 95% CI n Effect estimate 95% CI n Effect estimate 95% CI n Effect estimate 95% CI n Effect estimatea 95% CI
Gislason et al., 2006 [145] 42 1.36b 0.73–2.53 136 1.32b 1.02–1.72 – – – 59 2.46b 1.42–4.24 61 1.67b 1.15–2.42
Brophy et al., 2007 [146] 82 1.40c 1.06–1.84 – – – 8 1.56c 0.68–3.58 58 1.59c 1.15–2.18 – – –
Sorensen et al., 2008 [147] 42 1.50 1.10–2.05 59 1.25 1.07–1.46 – – – 59 1.63 1.27–2.10 61 1.54 1.23–1.93
Olsen et al., 2012 [148]
1 y post-MI 93 1.55 1.27–1.91 391 1.23 1.11–1.36 44 1.44 1.07–1.94 88 1.45 1.18–1.79 177 1.57 1.36–1.83
2 y post-MI 43 1.51 1.12–2.04 213 1.63 1.42–1.87 18 1.16 0.98–2.48 56 2.30 1.76–2.99 95 1.95 1.42–1.87
3 y post-MI 31 1.59 1.12–2.27 132 1.43 1.20–1.71 8 0.99 0.49–1.98 26 1.60 1.09–2.36 53 1.56 1.19–2.04
4 y post-MI 19 1.33 0.85–2.09 99 1.30 1.07–1.59 7 1.12 0.53–2.25 22 1.72 1.13–2.62 61 2.03 1.57–2.62
5 y post-MI 15 1.32 0.79–2.19 77 1.24 1.00–1.56 7 1.59 0.76–3.34 18 1.82 1.14–2.90 38 1.60 1.16–2.20
>5 y post-MI 10 0.80 0.43–1.48 125 1.42 1.24–1.62 17 1.10 0.68–1.77 13 1.50 0.87–2.59 109 1.73 1.43–2.09
Olsen et al., 2015 [149]d 59 1.46 1.13–1.89 395 1.42 1.28–1.57 21 0.86 0.52–1.36 21 1.06 0.69–1.63 200 1.65 1.44–1.90
a
Hazard ratio unless otherwise noted.
b
Odds ratio.
c
Rate ratio.
d
Considers a broader CV composite than other studies reported here; considering CV death, nonfatal recurrent MI, and ischemic stroke, transient ischemic attack, and systemic arterial emboli.
CI: confidence interval; MI: myocardial infarction; tNSAIDs; traditional nonsteroidal anti-inflammatory drugs.
 POSTGRADUATE MEDICINE 67

Concluding remarks

1.15–1.24

0.51–1.33

1.24–1.48
1.21–1.48
1.17–1.73
95% CI

–
NSAID-dependent inhibition of different prostaglandins has
the potential to play a critical role in the function of plate-
lets, vasculature and the kidney as well as their prostaglan-
Diclofenac
Effect estimate

din targets in the diseased joints [30–32,35–38,42–45,58–

1.19

0.82

1.35
1.34
1.42
–

–
64,66,67,78–84]. Initial focus on the antagonistic effects of

Odds ratio for recurrent heart failure. Numbers of patients in each group are not reported but across the entire analysis a total of 8353 patients were hospitalized because of recurrent heart failure.
prostacyclin and thromboxane was used to suggest that
unopposed COX-2-dependent inhibition of prostacyclin

Hazard ratio for recurrent heart failure. Numbers of patients in each group are not reported but across the entire analysis a total of 39,984 patients were hospitalized because of heart failure.
3228

may produce a pro-thrombogenic vascular effect with

NR
NR
NR
26
–

–
n

Odds ratio for recurrent heart failure compared with celecoxib. The study reported the odds ratio (95% CI) with all tNSAIDs together compared with celecoxib, 1.21 (0.92–1.60), n = 73.
COX-2 selective agents [59–62,121]. However, inhibition of
1.28–1.44

0.96–1.42
1.19–2.11

1.26–1.55
1.20–1.49
1.46–2.35
0.35–2.42
prostaglandins in the kidney is also likely to be pivotal [78–
95% CI

1.5–2.2

84]. Aligned to this notion there are data from both large
outcome trials [94,101,133] and meta-analyses [99,100] sug-
gesting that NSAIDs are relatively heterogeneous in relation
Effect estimate
Rofecoxib

to their effect on blood pressure, with post hoc analysis of

1.36

1.17
1.58

1.40
1.33
1.86
0.91

phase 4 trials illustrating significantly worse blood pressure

1.8

effects for ibuprofen when compared with either lumira-

coxib or celecoxib [94,95], or for etoricoxib when compared
Rate ratio of hospital admission for congestive heart failure. The study reported the rate ratio (95% CI) with all tNSAIDs together, 1.4 (1.0–1.9), n = 47.

with diclofenac [98]. A correlation between blood pressure-

1213
143

242
139

NR
NR
NR
NR
n

raising effects and poorer cardiovascular outcomes is also

tentatively seen in the long-term placebo-controlled studies
1.07–1.27

0.59–1.86

1.00–1.40
0.97–1.44
0.88–1.57
0.32–2.38
95% CI

for COX-2 selective inhibitors [108,109].

–

While early long-term placebo-controlled studies in

Higher doses are celecoxib > 200 mg/d; ibuprofen > 1200 mg/d; naproxen > 500 mg/d; rofecoxib > 25 mg/d; diclofenac > 100 mg/d.
Lower doses are celecoxib ≤ 200 mg/d; ibuprofen ≤ 1200 mg/d; naproxen ≤ 500 mg/d; rofecoxib ≤ 25 mg/d; diclofenac ≤ 100 mg/d.

humans demonstrated an excess of cardiovascular end-

Naproxen
Effect estimate

points with COX-2 selective agents, these trials almost uni-

1.16

1.05

1.18
1.18
1.18
0.87

CI: confidence interval; MI: myocardial infarction; tNSAIDs; traditional nonsteroidal anti-inflammatory drugs; NR: not reported.

versally lacked a traditional NSAID comparator arm. Studies,

–

prospectively designed to assess the relative cardiovascular

safety of traditional NSAIDs and COX-2 selective agents,
have since confirmed that cardiovascular harm is also
590

NR
NR
NR
NR
19
n

seen to a similar degree with the studied traditional

1.12–1.23

0.66–3.21

1.10–1.23
1.10–1.23
1.04–1.33
1.05–1.70

NSAIDs (ibuprofen, naproxen, and diclofenac) both in popu-

95% CI
Table 10. Risk of heart failure with tNSAIDs and COX-2 selective inhibitors in a community population.

lations with mixed baseline cardiovascular risk and in

–

patients with elevated baseline risk [96,97,130]. The largest

data sets from observational studies and meta-analyses of
Effect estimate
Ibuprofen

observational studies also suggest that use of traditional

1.18

1.46

1.16
1.16
1.18
1.34

NSAIDs is not free of cardiovascular risk, with diclofenac

–

consistently appearing at the higher end of the spectrum of

harm [138,139,141].The overall picture emerging is one
where cardiovascular and renal risks observed with tradi-
2012

NR
NR
NR
NR
–

–
n

tional NSAIDs and COX-2 selective agents exist across the

spectrum of medicines in the broad class and are not
0.90–1.02

1.12–1.39
1.09–1.41
1.04–1.33
0.53–2.06
95% CI

0.8–1.3

driven solely by the COX-2 selectivity of the medicine

Odds ratio for recurrent heart failure compared with celecoxib.
–
–

under consideration.
Effect estimate
Celecoxib

Reference
Reference

Funding
Odds ratio of hospital admission for heart failure.
0.96

1.24
1.24
1.26
1.05
1.0

This manuscript was sponsored by Pfizer.

1253
116

177
132

NR
NR
NR
NR
n

Declaration of interest
Mamdani et al., 2004 [150]a

Gislason et al., 2009 [154]e

Hudson et al., 2007 [153]d
Hudson et al., 2005 [152]c

C Walker is an employee of Pfizer and holds stock options with Pfizer.

Arfe et al., 2016 [151]b

Arfe et al., 2016 [151]h

LM Biasucci is a recipient of a research grant from Astra-Zeneca and in

Recurrent heart failure
Observational study

the last 3 years has received fees for speeches from Pfizer, Bayer and
Siemens. Medical writing support was provided by Joshua Fink, PhD, of
Engage Scientific Solutions, and funded by Pfizer. The authors have no
Higher doseg
Lower dosef
Heat failure

other relevant affiliations or financial involvement with any organiza-

All doses

tion or entity with a financial interest in or financial conflict with the

subject matter or materials discussed in the manuscript apart from
those disclosed.
b

g
h
e
a

f
68 C. WALKER AND L. M. BIASUCCI

References 24. Brzozowski T, Konturek PC, Konturek SJ, et al. Role of prostaglan-
dins in gastroprotection and gastric adaptation. J Physiol
1. Braun J, van den Berg R, Baraliakos X, et al. 2010 update of the Pharmacol. 2005;56 Suppl 5:33–55.
ASAS/EULAR recommendations for the management of ankylosing 25. Grosser T, Yu Y, Fitzgerald GA. Emotion recollected in tranquility:
spondylitis. Ann Rheum Dis. 2011;70:896–904. lessons learned from the COX-2 saga. Annu Rev Med.
2. Zhang W, Doherty M, Arden N, et al. EULAR evidence based 2010;61:17–33.
recommendations for the management of hip osteoarthritis: report 26. Grosser T, Fries S, FitzGerald GA. Biological basis for the cardiovas-
of a task force of the EULAR Standing Committee for International cular consequences of COX-2 inhibition: therapeutic challenges
Clinical Studies Including Therapeutics (ESCISIT). Ann Rheum Dis. and opportunities. J Clin Invest. 2006;116:4–15.
2005;64:669–681. 27. Grosser T, Ricciotti E, FitzGerald GA. The cardiovascular pharmacol-
3. Jordan KM, Arden NK, Doherty M, et al. EULAR Recommendations ogy of nonsteroidal anti-inflammatory drugs. Trends Pharmacol Sci.
2003: an evidence based approach to the management of knee 2017;38:733–748.
osteoarthritis: report of a Task Force of the Standing Committee for 28. Bhala N, Emberson J, Merhi A, et al. Vascular and upper gastro-
International Clinical Studies Including Therapeutic Trials (ESCISIT). intestinal effects of non-steroidal anti-inflammatory drugs: meta-
Ann Rheum Dis. 2003;62:1145–1155. analyses of individual participant data from randomised trials.
4. Rosen GD, Birkenmeier TM, Raz A, et al. Identification of a cycloox- Lancet. 2013;382:769–779.
ygenase-related gene and its potential role in prostaglandin for- 29. Trelle S, Reichenbach S, Wandel S, et al. Cardiovascular safety of
mation. Biochem Biophys Res Commun. 1989;164:1358–1365. non-steroidal anti-inflammatory drugs: network meta-analysis. BMJ.
5. Hawkey CJ. COX-2 inhibitors. Lancet. 1999;353:307–314. 2011;342:c7086.
6. Everts B, Wahrborg P, Hedner T. COX-2-specific inhibitors–the 30. Patrono C, Coller B, FitzGerald GA, et al. Platelet-active drugs: the
emergence of a new class of analgesic and anti-inflammatory relationships among dose, effectiveness, and side effects: the
drugs. Clin Rheumatol. 2000;19:331–343. Seventh ACCP Conference on Antithrombotic and Thrombolytic
7. Moore RA, Derry S, Phillips CJ, et al. Nonsteroidal anti-inflammatory Therapy. Chest. 2004;126:234S–64S.
drugs (NSAIDs), cyxlooxygenase-2 selective inhibitors (coxibs) and 31. Catella-Lawson F, Reilly MP, Kapoor SC, et al. Cyclooxygenase
gastrointestinal harm: review of clinical trials and clinical practice. inhibitors and the antiplatelet effects of aspirin. N Engl J Med.
BMC Musculoskelet Disord. 2006;7:79. 2001;345:1809–1817.
8. European Medicines Agency. EMEA public statement on the sus- 32. Patrono C, Garcia Rodriguez LA, Landolfi R, et al. Low-dose aspirin
pension of the marketing authorisation for bextra (valdecoxib) in for the prevention of atherothrombosis. N Engl J Med.
the European Union. [cited 2017 Sep 27]. Available at: [Link] 2005;353:2373–2383.
[Link]/docs/en_GB/document_library/Public_statement/ 33. Roth GJ, Machuga ET, Ozols J. Isolation and covalent structure of
2009/12/[Link] the aspirin-modified, active-site region of prostaglandin synthe-
9. European Medicines Agency. Questions and answers on the recom- tase. Biochemistry. 1983;22:4672–4675.
mendation to withdraw the marketing authorisations for lumira- 34. Patrono C, Rocca B. Aspirin and other COX-1 inhibitors. In: Gresele
coxib-containing medicines. [cited 2017 Sep 27]. Available at: P, Born GVR, Patrono C, et al., editors. Antiplatelet agents.
[Link] Handbook of experimental pharmacology. 210: Springer Berlin
Referrals_document/lumiracoxib_107/[Link] Heidelberg; 2012. p. 137–164.
10. Funk CD, FitzGerald GA. COX-2 inhibitors and cardiovascular risk. J 35. Leese PT, Hubbard RC, Karim A, et al. Effects of celecoxib, a novel
Cardiovasc Pharmacol. 2007;50:470–479. cyclooxygenase-2 inhibitor, on platelet function in healthy adults: a
11. Ricciotti E, FitzGerald GA. Prostaglandins and inflammation. randomized, controlled trial. J Clin Pharmacol. 2000;40:124–132.
Arterioscler Thromb Vasc Biol. 2011;31:986–1000. 36. Van Hecken A, Schwartz JI, Depre M, et al. Comparative inhibitory
12. Vane JR. Inhibition of prostaglandin synthesis as a mechanism of activity of rofecoxib, meloxicam, diclofenac, ibuprofen, and
action for aspirin-like drugs. Nat New Biol. 1971;231:232–235. naproxen on COX-2 versus COX-1 in healthy volunteers. J Clin
13. Smith WL, DeWitt DL, Garavito RM. Cyclooxygenases: structural, cel- Pharmacol. 2000;40:1109–1120.
lular, and molecular biology. Annu Rev Biochem. 2000;69:145–182. 37. Shah A, Woodruff M, Agarwal V, et al. Pharmacokinetics, safety, and
14. Vane JR, Bakhle YS, Botting RM. Cyclooxygenases 1 and 2. Annu tolerability of BAY 12-9566 and nonsteroidal anti-inflammatory
Rev Pharmacol Toxicol. 1998;38:97–120. agents (naproxen, ibuprofen) during coadministration in patients
15. Hamberg M, Svensson J, Samuelsson B. Thromboxanes: a new with osteoarthritis. J Clin Pharmacol. 2001;41:330–339.
group of biologically active compounds derived from prostaglan- 38. Evans AM. Pharmacodynamics and pharmacokinetics of the pro-
din endoperoxides. Proc Natl Acad Sci U S A. 1975;72:2994–2998. fens: enantioselectivity, clinical implications, and special reference
16. Moncada S, Gryglewski R, Bunting S, et al. An enzyme isolated from to S(+)-ibuprofen. J Clin Pharmacol. 1996;36:7S–15S.
arteries transforms prostaglandin endoperoxides to an unstable sub- 39. Rao GH, Johnson GG, Reddy KR, et al. Ibuprofen protects platelet
stance that inhibits platelet aggregation. Nature. 1976;263:663–665. cyclooxygenase from irreversible inhibition by aspirin.
17. Moncada S, Higgs EA, Vane JR. Human arterial and venous tissues Arteriosclerosis. 1983;3:383–388.
generate prostacyclin (prostaglandin x), a potent inhibitor of plate- 40. MacDonald TM, Wei L. Effect of ibuprofen on cardioprotective
let aggregation. Lancet. 1977;1:18–20. effect of aspirin. Lancet. 2003;361:573–574.
18. Mitchell JA, Warner TD. COX isoforms in the cardiovascular system: 41. Renda G, Tacconelli S, Capone ML, et al. Celecoxib, ibuprofen, and
understanding the activities of non-steroidal anti-inflammatory the antiplatelet effect of aspirin in patients with osteoarthritis and
drugs. Nat Rev Drug Discov. 2006;5:75–86. ischemic heart disease. Clin Pharmacol Ther. 2006;80:264–274.
19. Wallace JL, Devchand PR. Emerging roles for cyclooxygenase-2 in 42. Gladding PA, Webster MW, Farrell HB, et al. The antiplatelet effect
gastrointestinal mucosal defense. Br J Pharmacol. 2005;145:275– of six non-steroidal anti-inflammatory drugs and their pharmaco-
282. dynamic interaction with aspirin in healthy volunteers. Am J
20. Khan KN, Paulson SK, Verburg KM, et al. Pharmacology of cycloox- Cardiol. 2008;101:1060–1063.
ygenase-2 inhibition in the kidney. Kidney Int. 2002;61:1210–1219. 43. Gengo FM, Rubin L, Robson M, et al. Effects of ibuprofen on the
21. Norrdin RW, Jee WS, High WB. The role of prostaglandins in bone in magnitude and duration of aspirin’s inhibition of platelet aggrega-
vivo. Prostaglandins Leukot Essent Fatty Acids. 1990;41:139–149. tion: clinical consequences in stroke prophylaxis. J Clin Pharmacol.
22. Olson DM. The role of prostaglandins in the initiation of parturition. 2008;48:117–122.
Best Pract Res Clin Obstet Gynaecol. 2003;17:717–730. 44. Anzellotti P, Capone ML, Jeyam A, et al. Low-dose naproxen inter-
23. Alfirevic Z, Keeney E, Dowswell T, et al. Labour induction with feres with the antiplatelet effects of aspirin in healthy subjects:
prostaglandins: a systematic review and network meta-analysis. recommendations to minimize the functional consequences.
BMJ. 2015;350:h217. Arthritis Rheum. 2011;63:850–859.
 POSTGRADUATE MEDICINE 69

45. Schuijt MP, Huntjens-Fleuren HW, de Metz M, et al. The interaction 68. Bolego C, Buccellati C, Prada A, et al. Critical role of COX-1 in
of ibuprofen and diclofenac with aspirin in healthy volunteers. Br J prostacyclin production by human endothelial cells under modifi-
Pharmacol. 2009;157:931–934. cation of hydroperoxide tone. Faseb J. 2009;23:605–612.
46. Kurth T, Glynn RJ, Walker AM, et al. Inhibition of clinical benefits of 69. Toniolo A, Buccellati C, Pinna C, et al. Cyclooxygenase-1 and pros-
aspirin on first myocardial infarction by nonsteroidal antiinflamma- tacyclin production by endothelial cells in the presence of mild
tory drugs. Circulation. 2003;108:1191–1195. oxidative stress. PLoS One. 2013;8:e56683.
47. Curtis JP, Wang Y, Portnay EL, et al. Aspirin, ibuprofen, and mortal- 70. Wong E, Huang JQ, Tagari P, et al. Effects of COX-2 inhibitors on
ity after myocardial infarction: retrospective cohort study. BMJ. aortic prostacyclin production in cholesterol-fed rabbits.
2003;327:1322–1323. Atherosclerosis. 2001;157:393–402.
48. Fischer LM, Schlienger RG, Matter CM, et al. Current use of non- 71. Li S, Liu B, Luo W, et al. Role of cyclooxygenase-1 and −2 in
steroidal antiinflammatory drugs and the risk of acute myocardial endothelium-dependent contraction of atherosclerotic mouse
infarction. Pharmacotherapy. 2005;25:503–510. abdominal aortas. Clin Exp Pharmacol Physiol. 2016;43:67–74.
49. Meek IL, Vonkeman HE, Kasemier J, et al. Interference of NSAIDs 72. Kirkby NS, Lundberg MH, Wright WR, et al. COX-2 protects against
with the thrombocyte inhibitory effect of aspirin: a placebo-con- atherosclerosis independently of local vascular prostacyclin: identi-
trolled, ex vivo, serial placebo-controlled serial crossover study. Eur fication of COX-2 associated pathways implicate Rgl1 and lympho-
J Clin Pharmacol. 2013;69:365–371. cyte networks. PLoS One. 2014;9:e98165.
50. Schmidt M, Lamberts M, Olsen AM, et al. Cardiovascular safety of 73. Ahmetaj-Shala B, Kirkby NS, Knowles R, et al. Evidence that links
non-aspirin non-steroidal anti-inflammatory drugs: review and loss of cyclooxygenase-2 with increased asymmetric dimethylargi-
position paper by the working group for Cardiovascular nine: novel explanation of cardiovascular side effects associated
Pharmacotherapy of the European Society of Cardiology. Eur with anti-inflammatory drugs. Circulation. 2015;131:633–642.
Heart J Cardiovasc Pharmacother. 2016;2:108–118. 74. Hermann M, Ruschitzka F. Coxibs, non-steroidal anti-inflammatory
51. Warner TD, Giuliano F, Vojnovic I, et al. Nonsteroid drug selectiv- drugs and cardiovascular risk. Intern Med J. 2006;36:308–319.
ities for cyclo-oxygenase-1 rather than cyclo-oxygenase-2 are asso- 75. Chenevard R, Hurlimann D, Bechir M, et al. Selective COX-2 inhibi-
ciated with human gastrointestinal toxicity: a full in vitro analysis. tion improves endothelial function in coronary artery disease.
Proc Natl Acad Sci U S A. 1999;96:7563–7568. Circulation. 2003;107:405–409.
52. Dallob A, Hawkey CJ, Greenberg H, et al. Characterization of etor- 76. Title LM, Giddens K, McInerney MM, et al. Effect of cyclooxygenase-
icoxib, a novel, selective COX-2 inhibitor. J Clin Pharmacol. 2 inhibition with rofecoxib on endothelial dysfunction and inflam-
2003;43:573–585. matory markers in patients with coronary artery disease. J Am Coll
53. Wilner KD, Rushing M, Walden C, et al. Celecoxib does not affect Cardiol. 2003;42:1747–1753.
the antiplatelet activity of aspirin in healthy volunteers. J Clin 77. Widlansky ME, Price DT, Gokce N, et al. Short- and long-term COX-2
Pharmacol. 2002;42:1027–1030. inhibition reverses endothelial dysfunction in patients with hyper-
54. Ouellet M, Riendeau D, Percival MD. A high level of cyclooxygen- tension. Hypertension. 2003;42:310–315.
ase-2 inhibitor selectivity is associated with a reduced interference 78. Komhoff M, Grone HJ, Klein T, et al. Localization of cyclooxygenase-
of platelet cyclooxygenase-1 inactivation by aspirin. Proc Natl Acad 1 and -2 in adult and fetal human kidney: implication for renal
Sci U S A. 2001;98:14583–14588. function. Am J Physiol. 1997;272:F460–8.
55. Lee W, Suh JW, Yang HM, et al. Celecoxib does not attenuate the 79. Adegboyega PA, Ololade O. Immunohistochemical expression of
antiplatelet effects of aspirin and clopidogrel in healthy volunteers. cyclooxygenase-2 in normal kidneys. Appl Immunohistochem Mol
Korean Circ J. 2010;40:321–327. Morphol. 2004;12:71–74.
56. Ricciotti E, Yu Y, Grosser T, et al. COX-2, the dominant source of 80. Nantel F, Meadows E, Denis D, et al. Immunolocalization of cycloox-
prostacyclin. Proc Natl Acad Sci U S A. 2013;110:E183. ygenase-2 in the macula densa of human elderly. FEBS Lett.
57. Catella-Lawson F, McAdam B, Morrison BW, et al. Effects of specific 1999;457:475–477.
inhibition of cyclooxygenase-2 on sodium balance, hemodynamics, 81. Harris RC. COX-2 and the kidney. J Cardiovasc Pharmacol. 2006;47
and vasoactive eicosanoids. J Pharmacol Exp Ther. 1999;289:735–741. Suppl 1:S37–42.
58. McAdam BF, Catella-Lawson F, Mardini IA, et al. Systemic biosynthesis of 82. Weir MR. Renal effects of nonselective NSAIDs and coxibs. Cleve
prostacyclin by cyclooxygenase (COX)-2: the human pharmacology of a Clin J Med. 2002;69(Suppl 1):SI53–8.
selective inhibitor of COX-2. Proc Natl Acad Sci U S A. 1999;96:272–277. 83. Stichtenoth DO, Frolich JC. COX-2 and the kidneys. Curr Pharm Des.
59. Egan KM, Lawson JA, Fries S, et al. COX-2-derived prostacyclin con- 2000;6:1737–1753.
fers atheroprotection on female mice. Science. 2004;306:1954–1957. 84. Sica DS, Schoolwerth AC, Gehr TW. Pharmacotherapy in congestive
60. Cheng Y, Wang M, Yu Y, et al. Cyclooxygenases, microsomal pros- heart failure: COX-2 inhibition: a cautionary note in congestive
taglandin E synthase-1, and cardiovascular function. J Clin Invest. heart failure. Congest Heart Fail. 2000;6:272–276.
2006;116:1391–1399. 85. Rossat J, Maillard M, Nussberger J, et al. Renal effects of selective
61. Yu Y, Ricciotti E, Scalia R, et al. Vascular COX-2 modulates blood cyclooxygenase-2 inhibition in normotensive salt-depleted sub-
pressure and thrombosis in mice. Sci Transl Med. 2012;4:132ra54. jects. Clin Pharmacol Ther. 1999;66:76–84.
62. Tang SY, Monslow J, Todd L, et al. Cyclooxygenase-2 in endothelial 86. Whelton A, Schulman G, Wallemark C, et al. Effects of celecoxib and
and vascular smooth muscle cells restrains atherogenesis in hyper- naproxen on renal function in the elderly. Arch Intern Med.
lipidemic mice. Circulation. 2014;129:1761–1769. 2000;160:1465–1470.
63. Flavahan NA. Balancing prostanoid activity in the human vascular 87. Dilger K, Herrlinger C, Peters J, et al. Effects of celecoxib and diclofenac
system. Trends Pharmacol Sci. 2007;28:106–110. on blood pressure, renal function, and vasoactive prostanoids in
64. Kirkby NS, Lundberg MH, Harrington LS, et al. Cyclooxygenase-1, young and elderly subjects. J Clin Pharmacol. 2002;42:985–994.
not cyclooxygenase-2, is responsible for physiological production 88. Swan SK, Rudy DW, Lasseter KC, et al. Effect of cyclooxygenase-2
of prostacyclin in the cardiovascular system. Proc Natl Acad Sci U S inhibition on renal function in elderly persons receiving a low-salt
A. 2012;109:17597–17602. diet: a randomized controlled trial. Ann Intern Med. 2000;133:1–9.
65. Mitchell JA, Warner TD. Reply to Ricciotti et al.: evidence for vas- 89. Schwartz JI, Thach C, Lasseter KC, et al. Effects of etoricoxib and
cular COX isoforms. Proc Natl Acad Sci U S A. 2013;110:E184. comparator nonsteroidal anti-inflammatory drugs on urinary
66. Belton O, Byrne D, Kearney D, et al. Cyclooxygenase-1 and −2- sodium excretion, blood pressure, and other renal function indica-
dependent prostacyclin formation in patients with atherosclerosis. tors in elderly subjects consuming a controlled sodium diet. J Clin
Circulation. 2000;102:840–845. Pharmacol. 2007;47:1521–1531.
67. Zidar N, Odar K, Glavac D, et al. Cyclooxygenase in normal human 90. Schwartz JI, Vandormael K, Malice MP, et al. Comparison of rofe-
tissues–is COX-1 really a constitutive isoform, and COX-2 an indu- coxib, celecoxib, and naproxen on renal function in elderly subjects
cible isoform? J Cell Mol Med. 2009;13:3753–3763. receiving a normal-salt diet. Clin Pharmacol Ther. 2002;72:50–61.
70 C. WALKER AND L. M. BIASUCCI

91. Whelton A, Maurath CJ, Verburg KM, et al. Renal safety and toler- 111. ADAPT Research Group. Cardiovascular and cerebrovascular events
ability of celecoxib, a novel cyclooxygenase-2 inhibitor. Am J Ther. in the randomized, controlled Alzheimer’s Disease Anti-Inflammatory
2000;7:159–175. Prevention Trial (ADAPT). PLoS Clin Trials. 2006;1:e33.
92. Curtis SP, Ng J, Yu Q, et al. Renal effects of etoricoxib and com- 112. Baron JA, Sandler RS, Bresalier RS, et al. Cardiovascular events
parator nonsteroidal anti-inflammatory drugs in controlled clinical associated with rofecoxib: final analysis of the APPROVe trial.
trials. Clin Ther. 2004;26:70–83. Lancet. 2008;372:1756–1764.
93. Daniels B, Gertz B, Morrison B, et al. Renal safety profile of rofe- 113. Solomon SD, Pfeffer MA, McMurray JJ, et al. Effect of celecoxib on
coxib. A specific inhibitor of Cox-2, in controlled clinical trials. cardiovascular events and blood pressure in two trials for the pre-
[Abstract]. Arthritis Rheum. 1999;42:S143. vention of colorectal adenomas. Circulation. 2006;114:1028–1035.
94. Whelton A, Lefkowith JL, West CR, et al. Cardiorenal effects of 114. Bertagnolli MM, Eagle CJ, Zauber AG, et al. Five-year efficacy and
celecoxib as compared with the nonsteroidal anti-inflammatory safety analysis of the Adenoma Prevention with Celecoxib Trial.
drugs diclofenac and ibuprofen. Kidney Int. 2006;70:1495–1502. Cancer Prev Res (Phila). 2009;2:310–321.
95. Farkouh ME, Verheugt FW, Ruland S, et al. A comparison of the 115. Ott E, Nussmeier NA, Duke PC, et al. Efficacy and safety of the
blood pressure changes of lumiracoxib with those of ibuprofen and cyclooxygenase 2 inhibitors parecoxib and valdecoxib in patients
naproxen. J Clin Hypertens (Greenwich). 2008;10:592–602. undergoing coronary artery bypass surgery. J Thorac Cardiovasc
96. Farkouh ME, Kirshner H, Harrington RA, et al. Comparison of Surg. 2003;125:1481–1492.
lumiracoxib with naproxen and ibuprofen in the Therapeutic 116. Nussmeier NA, Whelton AA, Brown MT, et al. Complications of the
Arthritis Research and Gastrointestinal Event Trial (TARGET), car- COX-2 inhibitors parecoxib and valdecoxib after cardiac surgery. N
diovascular outcomes: randomised controlled trial. Lancet. Engl J Med. 2005;352:1081–1091.
2004;364:675–684. 117. Yang HM, Kim HS, Park KW, et al. Celecoxib, a cyclooxygenase-2
97. Cannon CP, Curtis SP, FitzGerald GA, et al. Cardiovascular outcomes inhibitor, reduces neointimal hyperplasia through inhibition of Akt
with etoricoxib and diclofenac in patients with osteoarthritis and signaling. Circulation. 2004;110:301–308.
rheumatoid arthritis in the Multinational Etoricoxib and Diclofenac 118. Koo BK, Kim YS, Park KW, et al. Effect of celecoxib on restenosis after
Arthritis Long-term (MEDAL) programme: a randomised compari- coronary angioplasty with a Taxus stent (COREA-TAXUS trial): an
son. Lancet. 2006;368:1771–1781. open-label randomised controlled study. Lancet. 2007;370:567–574.
98. Krum H, Swergold G, Curtis SP, et al. Factors associated with blood 119. Kang HJ, Oh IY, Chung JW, et al. Effects of celecoxib on restenosis
pressure changes in patients receiving diclofenac or etoricoxib: after coronary intervention and evolution of atherosclerosis (Mini-
results from the MEDAL study. J Hypertens. 2009;27:886–893. COREA) trial: celecoxib, a double-edged sword for patients with
99. Zhang J, Ding EL, Song Y. Adverse effects of cyclooxygenase 2 angina. Eur Heart J. 2012;33:2653–2661.
inhibitors on renal and arrhythmia events: meta-analysis of rando- 120. Chung JW, Yang HM, Park KW, et al. Long-term outcome of adjunc-
mized trials. JAMA. 2006;296:1619–1632. tive celecoxib treatment after paclitaxel-eluting stent implantation
100. Chan CC, Reid CM, Aw TJ, et al. Do COX-2 inhibitors raise blood for the complex coronary lesions: two-year clinical follow-up of
pressure more than nonselective NSAIDs and placebo? An updated COREA-TAXUS trial. Circ Cardiovasc Interv. 2010;3:243–248.
meta-analysis. J Hypertens. 2009;27:2332–2341. 121. Bombardier C, Laine L, Reicin A, et al. Comparison of upper gastro-
101. Cannon CP, Curtis SP, Bolognese JA, et al. Clinical trial design and intestinal toxicity of rofecoxib and naproxen in patients with rheu-
patient demographics of the Multinational Etoricoxib and Diclofenac matoid arthritis. VIGOR Study Group. N Engl J Med. 2000;343: 1520–
Arthritis Long-term (MEDAL) study program: cardiovascular out- 1528. 2 p following 8.
comes with etoricoxib versus diclofenac in patients with osteoarthri- 122. Mukherjee D, Nissen SE, Topol EJ. Risk of cardiovascular events
tis and rheumatoid arthritis. Am Heart J. 2006;152:237–245. associated with selective COX-2 inhibitors. JAMA. 2001;286:954–959.
102. Whelton A, Fort JG, Puma JA, et al. Cyclooxygenase-2–specific 123. Silverstein FE, Faich G, Goldstein JL, et al. Gastrointestinal toxicity
inhibitors and cardiorenal function: a randomized, controlled trial with celecoxib vs nonsteroidal anti-inflammatory drugs for osteoar-
of celecoxib and rofecoxib in older hypertensive osteoarthritis thritis and rheumatoid arthritis: the CLASS study: a randomized
patients. Am J Ther. 2001;8:85–95. controlled trial. Celecoxib Long-term Arthritis Safety Study. JAMA.
103. Whelton A, White WB, Bello AE, et al. Effects of celecoxib and 2000;284:1247–1255.
rofecoxib on blood pressure and edema in patients > or =65 124. White WB, Faich G, Whelton A, et al. Comparison of thromboem-
years of age with systemic hypertension and osteoarthritis. Am J bolic events in patients treated with celecoxib, a cyclooxygenase-2
Cardiol. 2002;90:959–963. specific inhibitor, versus ibuprofen or diclofenac. Am J Cardiol.
104. Sowers JR, White WB, Pitt B, et al. The Effects of cyclooxygen- 2002;89:425–430.
ase-2 inhibitors and nonsteroidal anti-inflammatory therapy on 125. Juni P, Rutjes AW, Dieppe PA. Are selective COX 2 inhibitors super-
24-hour blood pressure in patients with hypertension, osteoar- ior to traditional non steroidal anti-inflammatory drugs? BMJ.
thritis, and type 2 diabetes mellitus. Arch Intern Med. 2002;324:1287–1288.
2005;165:161–168. 126. European Medicines Agency. PRAC recommends the same cardiovas-
105. Staessen JA, Thijs L, Fagard R, et al. Predicting cardiovascular risk cular precautions for diclofenac as for selective COX-2 inhibitors.
using conventional vs ambulatory blood pressure in older patients [cited 2017 Sep 27]. Available at: [Link]
with systolic hypertension. Systolic Hypertension in Europe Trial en_GB/document_library/Press_release/2013/06/[Link]
Investigators. JAMA. 1999;282:539–546. 127. Antman EM, Bennett JS, Daugherty A, et al. Use of nonsteroidal antiin-
106. Clement DL, De Buyzere ML, de Bacquer DA, et al. Prognostic value flammatory drugs: an update for clinicians: a scientific statement from
of ambulatory blood-pressure recordings in patients with treated the American Heart Association. Circulation. 2007;115:1634–1642.
hypertension. N Engl J Med. 2003;348:2407–2415. 128. Kearney PM, Baigent C, Godwin J, et al. Do selective cyclo-oxyge-
107. Bresalier RS, Sandler RS, Quan H, et al. Cardiovascular events asso- nase-2 inhibitors and traditional non-steroidal anti-inflammatory
ciated with rofecoxib in a colorectal adenoma chemoprevention drugs increase the risk of atherothrombosis? Meta-analysis of ran-
trial. N Engl J Med. 2005;352:1092–1102. domised trials. BMJ. 2006;332:1302–1308.
108. Solomon SD, McMurray JJ, Pfeffer MA, et al. Cardiovascular risk 129. Becker MC, Wang TH, Wisniewski L, et al. Rationale, design, and
associated with celecoxib in a clinical trial for colorectal adenoma governance of Prospective Randomized Evaluation of Celecoxib
prevention. N Engl J Med. 2005;352:1071–1080. Integrated Safety versus Ibuprofen Or Naproxen (PRECISION), a
109. Arber N, Eagle CJ, Spicak J, et al. Celecoxib for the prevention of cardiovascular end point trial of nonsteroidal antiinflammatory
colorectal adenomatous polyps. N Engl J Med. 2006;355:885–895. agents in patients with arthritis. Am Heart J. 2009;157:606–612.
110. Steinbach G, Lynch PM, Phillips RK, et al. The effect of celecoxib, a 130. Nissen SE, Yeomans ND, Solomon DH, et al. Cardiovascular safety of
cyclooxygenase-2 inhibitor, in familial adenomatous polyposis. N celecoxib, naproxen, or ibuprofen for arthritis. N Engl J Med.
Engl J Med. 2000;342:1946–1952. 2016;375:2519–2529.
 POSTGRADUATE MEDICINE 71

131. FitzGerald GA. Imprecision: limitations to interpretation of a large 145. Gislason GH, Jacobsen S, Rasmussen JN, et al. Risk of death or
randomized clinical trial. Circulation. 2017;135:113–115. reinfarction associated with the use of selective cyclooxygenase-2
132. Nurmohamed MT. Therapy: cardiovascular safety of celecoxib, inhibitors and nonselective nonsteroidal antiinflammatory drugs
naproxen and ibuprofen. Nat Rev Rheumatol. 2017;13:136–138. after acute myocardial infarction. Circulation. 2006;113:2906–2913.
133. Ruschitzka F, Borer JS, Krum H, et al. Differential blood pressure 146. Brophy JM, Levesque LE, Zhang B. The coronary risk of cyclo-
effects of ibuprofen, naproxen, and celecoxib in patients with oxygenase-2 inhibitors in patients with a previous myocardial
arthritis: the PRECISION-ABPM (Prospective Randomized infarction. Heart. 2007;93:189–194.
Evaluation of Celecoxib Integrated Safety Versus Ibuprofen or 147. Sorensen R, Abildstrom SZ, Torp-Pedersen C, et al. Use of selective
Naproxen Ambulatory Blood Pressure Measurement) Trial. Eur cyclooxygenase-2 inhibitors and nonselective nonsteroidal antiin-
Heart J. 2017;38(44):3282–3292. flammatory drugs in high doses increases mortality and risk of
134. Hirayama A, Tanahashi N, Daida H, et al. Assessing the cardiovas- reinfarction in patients with prior myocardial infarction. J
cular risk between celecoxib and nonselective nonsteroidal antiin- Cardiovasc Nurs. 2008;23:14–19.
flammatory drugs in patients with rheumatoid arthritis and 148. Olsen AM, Fosbol EL, Lindhardsen J, et al. Long-term cardiovascular
osteoarthritis. Circ J. 2014;78:194–205. risk of nonsteroidal anti-inflammatory drug use according to time
135. MacDonald TM, Hawkey CJ, Ford I, et al. Randomized trial of switch- passed after first-time myocardial infarction: a nationwide cohort
ing from prescribed non-selective non-steroidal anti-inflammatory study. Circulation. 2012;126:1955–1963.
drugs to prescribed celecoxib: the Standard care vs. Celecoxib 149. Schjerning Olsen AM, Gislason GH, McGettigan P, et al. Association
Outcome Trial (SCOT). Eur Heart J. 2017;38:1843–1850. of NSAID use with risk of bleeding and cardiovascular events in
136. Signorello LB, McLaughlin JK, Lipworth L, et al. Confounding by patients receiving antithrombotic therapy after myocardial infarc-
indication in epidemiologic studies of commonly used analgesics. tion. JAMA. 2015;313:805–814.
Am J Ther. 2002;9:199–205. 150. Mamdani M, Juurlink DN, Lee DS, et al. Cyclo-oxygenase-2 inhibi-
137. Lapi F, Piccinni C, Simonetti M, et al. Non-steroidal anti-inflamma- tors versus non-selective non-steroidal anti-inflammatory drugs
tory drugs and risk of cerebrovascular events in patients with and congestive heart failure outcomes in elderly patients: a popu-
osteoarthritis: a nested case-control study. Intern Emerg Med. lation-based cohort study. Lancet. 2004;363:1751–1756.
2016;11:49–59. 151. Arfe A, Scotti L, Varas-Lorenzo C, et al. Non-steroidal anti-inflam-
138. McGettigan P, Henry D. Cardiovascular risk with non-steroidal anti- matory drugs and risk of heart failure in four European countries:
inflammatory drugs: systematic review of population-based con- nested case-control study. BMJ. 2016;354:i4857.
trolled observational studies. PLoS Med. 2011;8:e1001098. 152. Hudson M, Richard H, Pilote L. Differences in outcomes of patients
139. Salvo F, Antoniazzi S, Duong M, et al. Cardiovascular events asso- with congestive heart failure prescribed celecoxib, rofecoxib, or
ciated with the long-term use of NSAIDs: a review of randomized non-steroidal anti-inflammatory drugs: population based study.
controlled trials and observational studies. Expert Opin Drug Saf. BMJ. 2005;330:1370.
2014;13:573–585. 153. Hudson M, Rahme E, Richard H, et al. Risk of congestive heart failure
140. Varas-Lorenzo C, Riera-Guardia N, Calingaert B, et al. Stroke risk and with nonsteroidal antiinflammatory drugs and selective cyclooxy-
NSAIDs: a systematic review of observational studies. genase 2 inhibitors: a class effect? Arthritis Rheum. 2007;57:516–523.
Pharmacoepidemiol Drug Saf. 2011;20:1225–1236. 154. Gislason GH, Rasmussen JN, Abildstrom SZ, et al. Increased mortal-
141. Bally M, Dendukuri N, Rich B, et al. Risk of acute myocardial infarc- ity and cardiovascular morbidity associated with use of nonsteroi-
tion with NSAIDs in real world use: bayesian meta-analysis of dal anti-inflammatory drugs in chronic heart failure. Arch Intern
individual patient data. BMJ. 2017;357:j1909. Med. 2009;169:141–149.
142. Johnsen SP, Larsson H, Tarone RE, et al. Risk of hospitalization for 155. Chou CI, Shih CJ, Chen YT, et al. Adverse effects of oral nonselective
myocardial infarction among users of rofecoxib, celecoxib, and and cyclooxygenase-2-selective NSAIDs on hospitalization for acute
other NSAIDs: a population-based case-control study. Arch Intern kidney injury. Medicine (Baltimore). 2016 Mar; 95(9):e2645.
Med. 2005;165:978–984. 156. Lafrance JP, Miller DR. Selective and non-selective non-steroidal
143. Varas-Lorenzo C, Castellsague J, Stang MR, et al. The use of selec- anti-inflammatory drugs and the risk of acute kidney injury.
tive cyclooxygenase-2 inhibitors and the risk of acute myocardial Pharmacoepidemiol Drug Saf. 2009;18:923–931.
infarction in Saskatchewan, Canada. Pharmacoepidemiol Drug Saf. 157. Schneider V, Levesque LE, Zhang B, et al. Association of selective
2009;18:1016–1025. and conventional nonsteroidal antiinflammatory drugs with acute
144. Shau WY, Chen HC, Chen ST, et al. Risk of new acute myocardial renal failure: a population-based, nested case-control analysis. Am J
infarction hospitalization associated with use of oral and parenteral Epidemiol. 2006;164:881–889.
non-steroidal anti-inflammation drugs (NSAIDs): a case-crossover 158. Winkelmayer WC, Waikar SS, Mogun H, et al. Nonselective and
study of Taiwan’s National Health Insurance claims database and cyclooxygenase-2-selective NSAIDs and acute kidney injury. Am J
review of current evidence. BMC Cardiovasc Disord. 2012;12:4. Med. 2008;121:1092–1098.