The risk of coronary thrombosis with cyclo-oxygenase-2 inhibitors does not vary with polymorphisms in two regions of the cyclo-oxygenase-2 gene
Dr David Henry, Institute for Clinical Evaluative Sciences, Sunnybrook Health Sciences Centre, 2075 Bayview Avenue, Toronto M4N3M5, Canada. Tel.: +1 41 6480 4297, Fax: +1 41 6480 6048, E-mail: firstname.lastname@example.org
WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT
• The risk of cardiovascular events with non-steroidal anti-inflammatory drugs (NSAIDs) varies with individual drugs and with dosage. Little is known about other factors that can modify risk.
• Polymorphisms of the COX-2 gene conceivably could alter the risk of myocardial infarction and also the risk from NSAIDs.
WHAT THIS STUDY ADDS
• Two common polymorphisms of the COX-2 gene, rs 20417 (G > C) and rs 5275 (T > C), were not associated with variation in the risk of acute coronary syndrome, and individually did not alter the risk associated with use of NSAIDs.
• Haplotype analysis identified a trend to protection against NSAID-related cardiovascular risk with the ‘low risk’ haplotype, but the effect was weak and if confirmed, the clinical utility would be limited.
AIMS To investigate whether polymorphisms of the cyclo-oxygenase-2 (COX-2) gene modify the adverse cardiovascular effects of COX-2 inhibitors.
METHODS A case control study was conducted in the Hunter Region of New South Wales, Australia. Cases (n= 460) were hospitalized with acute coronary syndrome (ACS). Controls (n= 640) were recruited from the electoral rolls. Structured interviews gathered information on variables including recent ingestion of non-steroidal anti-inflammatory drugs (NSAIDs). Targeted genotyping of rs 20417(G > C) and rs5275 (T > C) polymorphisms was performed by real-time polymerase chain reaction using allele-specific probes
RESULTS Ingestion of any NSAID in the week prior to interview was associated with an elevated risk for ACS: adjusted odds ratio 1.8 (1.2, 2.5). The rs 20417 and rs 5275 polymorphisms were not singly associated with risk for ACS: adjusted odds ratios 1.1 (0.80, 1.5) and 1.2 (0.88, 1.5), respectively. Individually, the polymorphisms did not modify the risk of ACS with the drugs. When analyses were conducted by haplotype, the adjusted odds ratio with celecoxib or rofecoxib in individuals who had one or two copies of the ‘low risk’ haplotype (no GT) was 1.2 (0.29, 5.0), compared with 2.1 (1.1, 4.0) with the ‘high risk’ haplotype (one or two copies of GT).
CONCLUSIONS We found little evidence of a gene/drug interaction. We found a statistically non-significant trend toward a lower risk of coronary events with NSAIDs in the presence of the ‘low risk’ haplotype. Even if confirmed, the clinical utility of the finding would be limited as this haplotype is carried by a minority of the population.
The adverse cardiovascular effects of non-steroidal anti-inflammatory drugs (NSAIDs) have been studied extensively since the increase in cardiovascular risk with rofecoxib was first identified in 2000 [1–5]. Meta-analyses of randomized trials and observational studies have confirmed the increased risk of vascular occlusion with rofecoxib at commonly used doses and indicated that there was a risk of thrombosis with celecoxib at higher doses [6–8]. Significantly, it appears that some of the older non-selective NSAIDs, such as diclofenac, may cause this problem . The only widely investigated drug that has been shown consistently not to increase risk of thrombosis in usual doses is naproxen [5, 7].
The biology of this adverse effect is complex. Most of the selective and non-selective NSAIDs achieve a high degree of COX-2 inhibition . At the level of the vascular endothelium this will depress production of prostacyclin and the resultant risk probably depends on the extent to which the drugs also suppress COX-1 mediated synthesis of thromboxanes . Sustained high level suppression of COX-1 is needed to counter the loss of prostacyclin . This is hard to achieve with most NSAIDs and will depend on the dose of drug ingested, and the pharmacokinetic/pharmacodynamic relationships that determine tissue responses to each agent. While it has been difficult to show directly that NSAIDs block prostacyclin production by constitutive COX-2 in the vascular endothelium in humans, the balance of evidence favours an important role for the drugs . The consequences include vasoconstriction, the loss of a general restraining effect on platelet activation, elevated blood pressure, an atherogenic effect, and possibly a failure of COX-2 dependent adaptive responses to ischaemic pre-conditioning [10–12].
The excess risk to individuals of having a vascular occlusion triggered by use of a NSAID will depend on their background risk and the relative risk of thrombosis with that particular drug. Factors that are known to modify the relative risk are the individual drug (e.g. elevated with rofecoxib, not with naproxen) and the ingested dose. A dose related increase in relative risk (RR) has been seen with both rofecoxib and celecoxib [6–8].
Conceivably, patient related factors may also alter the relative risk of thrombosis on treatment with COX-2 inhibitors. These could alter the pharmacokinetics or pharmacodynamics of the drugs [13–15]. Genetically determined alterations in the activities of enzymes that are pharmacological targets have been reported to produce variations in the effects of NSAIDs [16, 17].
There has been extensive study of the associations between single nucleotide polymorphisms (SNPs) of the COX-2 gene and risk of cardiovascular events. The rs20417 (G > C) polymorphism, commonly known as −765 G > C, in the promoter region has been studied most, with conflicting results. Early studies found apparently protective effects, with reduced risk of myocardial infarction and stroke [18, 19]. Later studies have contradicted these findings [20, 21]. The rs5275 (T > C) polymorphism, also known as +6365T/C and +837T > C, in the 3′untranslated region of the COX-2 gene has been less studied. Though one investigation found no association with calcified plaque , both polymorphisms have been shown to have significant effects on the expression of COX-2, with a consequent reduction in synthesis of various inflammatory mediators [23, 24], potentially altering both the therapeutic and adverse effects of COX-2 inhibiting drugs such as NSAIDs. It would be expected that the haplotype represented by the ‘low risk’ genotype at both locations might provide some protection against the adverse cardiovascular effects of NSAIDs.
We have investigated the latter possibility by measuring variations in the relative risk of thrombosis associated with NSAIDs among individuals who possess the rs20417 (G > C) and rs5275 (T > C) polymorphisms of the COX-2 gene.
We undertook a case control study to investigate the risk of acute coronary syndrome (ACS: acute myocardial infarction or unstable angina pectoris) with ingestion of NSAIDs (both selective COX-2 inhibitors, coxibs, and conventional non-selective NSAIDs). When the study commenced in August 2003, celecoxib and rofecoxib were the most widely used NSAIDs in Australia. Use of meloxicam was slight and valdecoxib, lumiracoxib and parecoxib were not marketed.
Cases were individuals admitted with ACS (August 2003 to October 2006) to three hospitals in the Hunter and Central Coast regions of New South Wales. These are the main hospitals serving a predominantly Caucasian population of just over 1 million persons. Cases had to meet the criteria for ACS developed by the PRISM study group . They were identified through daily scrutiny of computerized admission lists, attendance at morning report and enquiries of the clinical staff working in medical and cardiology wards. The index day for cases was the day of admission to hospital.
Controls were recruited from the Hunter Community Study (HCS). This is a prospective cohort of community-dwelling men and women aged 55–85 years selected at random (November 2004 to May 2005) from the State Electoral Rolls . Medical history and clinical variables were collected at a research clinic staffed by a research nurse/associate and by questionnaire, and blood was drawn and stored at −80°C for DNA isolation. The HCS employed the same questions about NSAID use that were employed with the cases. The date of interview was used as the index day for community controls.
Research nurses used a structured protocol to interview both cases and controls. Information collected included demographics, medical history, smoking history, alcohol intake, and medicines use, both prescribed and over the counter. Ingestion of analgesics was explored through a series of open questions asking about medicines used for painful conditions. These were followed by more direct questions and presentation of ‘flash’ cards, listing the generic and trade names of all relevant products on the Australian market. Each question could be repeated once for clarification. Details of all analgesics ingested within the week and month prior to admission were recorded. We have successfully used these methods in previous studies of cardiovascular complications of NSAIDs [27, 28].
Additional clinical and investigation data to corroborate the admission diagnosis and past medical history were collected from the medical records using a standardized data extraction form. Aliquots of blood (20 ml/person) were drawn from all cases and interviews were conducted within 1 week of admission, the majority within the first 3 days. The interviewers knew whether they were interviewing cases or controls, and were aware of the study aims. They understood the importance of adhering strictly to the wording of the protocol. Cases understood that we were interested in medicines consumption, particularly medicines for pain, but were unaware of the hypothesis in respect of anti-inflammatory drugs and thrombosis.
All study procedures were approved by the Hunter New England Health and North Sydney Central Coast Research Ethics Committees and the Human Research Ethics Committee of The University of Newcastle. All participants gave informed written consent.
DNA was extracted from the cellular fraction of peripheral blood and genotyping performed by real-time polymerase chain reaction (PCR) assays using Taqman MBG allele specific probes and assays developed by Assay-by-Design (Applied Biosystems, Foster City, California, USA). Data were acquired on the Rotor-Gene RG-3000 and analysed by end-point allelic discrimination using Rotor-Gene Real Time Analysis Software v6.1 (Corbett Research, Sydney, Australia). A subset of samples was verified by restriction fragment length polymorphism (RFLP) using previously published methods .
We studied the single and joint effects of consumption of NSAIDs and the presence of the two polymorphisms of the COX-2 gene. The risk of ACS in users compared with non-users was estimated from the exposure odds ratio (OR) for ingestion of the drugs of interest in the week prior to the index day. The same approach was used to estimate the OR for ACS in those who exhibited the polymorphisms of interest. OR estimates were adjusted for potential confounders (age, gender, body mass index, risk factors for, or previous history of cardiovascular events) by multivariable logistic regression analysis.
We studied the joint effects of NSAIDs and polymorphisms by estimating the OR with the drugs in the presence of the individual genotypes. We added an interaction term between NSAID and genotype in the logistic regression. The polymorphisms were modelled both individually as well as jointly by inferring the haplotype using the E-M algorithm as implemented in the SimHap program . In these analyses we looked at the joint effects of consumption of NSAIDs and the presence of either one or two copies of the risk haplotype. The GT haplotype, i.e G at rs20417 and T at rs5275, was used as the analysis category since it was the ‘high risk’ haplotype, even though these represent the common alleles at each SNP.
The characteristics of the cases and community controls are summarized in Table 1. Cases and controls were of similar age and BMI. As expected, cases were more likely to be male, had a higher rate of current smoking, and were more likely to have a history of vascular events or other relevant cardiovascular problems (Table 1). These factors were adjusted for in the subsequent analyses.
Table 1. Baseline characteristics of cases and controls
|Current smoker||No||606 (94.7%)||328 (71.3%)||<ref>||<ref>|
|Yes||34 (5.3%)||132 (28.7%)||7.2 (4.8, 10.7)||8.6 (5.5, 13)|
|Gender||Male||300 (46.9%)||340 (73.9%)||3.2 (2.5, 4.2)||2.4 (1.8, 3.2)|
|Female||339 (53.1%)||120 (26.1%)||<ref>||<ref>|
|Age (years)||Mean (SE)||66.0 (0.29)||65.1 (0.63)||0.99/year (0.98, 1.0)||1.0/year (0.99, 1.0)|
|Body mass index (kg m−2)||Mean (SE)||28.8 (0.19)||28.6 (0.25)||0.99/unit (0.97, 1.0)||1.0/unit (0.98, 1.0)|
|Vascular ischaemia history†||No||519 (81.1%)||239 (52%)||<ref>||<ref>|
|Yes||121 (18.9%)||221 (48%)||4.0 (3.0, 5.2)||4.3 (3.1, 5.8)|
The crude and adjusted analyses of variables of interest are presented in Table 2. Ingestion of a coxib or any NSAID in the week prior to the index day was associated with a 60 to 80% relative increase in risk of ACS. These analyses were relatively unaffected by adjustment for other risk factors (Table 2). Neither of the COX-2 polymorphisms was individually associated with any increase or reduction in risk of ACS. Individual estimates of heterozygote and homozygote variant groups for each of the two SNPs were very similar, hence they were pooled, i.e. a dominant model was used.
Table 2. Univariate and multiple regression analyses: COX-2 inhibitors and individual COX-2 polymorphisms and their relationship to acute coronary syndrome
|Coxibs||46 (7.2)||46 (10.0)||1.4 (0.86, 2.1)||1.6 (1.0, 2.6)|
|Any NSAID||122 (19.1)||123 (26.7)||1.6 (1.2, 2.2)||1.8 (1.2, 2.5)|
|rs20417 G > C||162 (25.3)||124 (27.0)||1.1 (0.83, 1.4)||1.1 (0.80, 1.5)|
|one or two copies of G|
|rs5275 T > C||333 (52.0)||247 (53.7)||1.1 (0.84, 1.4)||1.2 (0.88, 1.5)|
|one or two copies of T|
The joint effects of the COX-2 inhibitors and the individual polymorphisms of the COX-2 gene are presented in Table 3. In order to explore more clearly the joint effect of drug use and polymorphisms, data were displayed using the 2 × 4 table suggested by Botto & Khoury . There were no apparent interactions between the individual polymorphisms and the use of either coxibs or any NSAID in the previous week. As seen in Table 3, the effect of either class of drugs was similar, regardless of the SNP present.
Table 3. Joint effects of COX-2 inhibitors and the single COX-2 polymorphisms
|Coxibs||46 (7.2)||46 (10.0)||1.4 (0.86, 2.1)||1.6 (1.0, 2.6)|
|Coxib ingestion plus rs20417G > C*||12 (1.9)||13 (2.8)||1.5 (0.64, 3.3)||1.5 (0.61, 3.9)|
|one or two copies of G|
|Coxib ingestion plus rs5275T > C||25 (4.0)||28 (6.2)||1.6 (0.90, 2.8)||1.8 (0.98, 3.4)|
|one or two copies of T|
|Any NSAID||122 (19.1)||123 (26.7)||1.6 (1.2, 2.2)||1.8 (1.2, 2.5)|
|Any NSAID||31 (4.9)||40 (8.7)||2.0 (1.2, 3.5)||2.5 (1.4, 4.8)|
|plus rs20417G > C*|
|one or two copies of G|
|Any NSAID plus rs5275T > C||65 (10.3)||72 (15.9)||1.7 (1.2, 2.5)||2.0 (1.3, 3.2)|
|one or two copies of T|
However, when we performed analysis by haplotype there was an apparent effect of genotype on the risk of acute coronary syndrome. (Table 4) Again, estimates for heterozygotes or homozygotes for the high risk haplotype were very similar and these were pooled. The OR with either coxib (rofecoxib or celecoxib) in individuals who had no copies of the risk haplotype was 1.2 (95% CI 0.29, 5.0), compared with 2.1 (1.1, 4.0) for those who had one or two copies of the risk haplotype. A similar but weaker effect was seen with use of any NSAID: 1.8 (0.69, 4.5) vs. 2.2 (1.3, 3.8), respectively. However, it is important to note that 95% confidence intervals for these estimates overlap to a large degree and formal tests for interaction were negative.
Table 4. Analyses of joint effects of COX-2 inhibitors use and COX-2 haplotypes
|No drug||82 (13.1)||49 (10.8)||1.0 (Reference)||1.0 (Reference)|
|Coxib||8 (1.3)||3 (0.7)||0.6 (0.16, 2.5)||1.2 (0.29, 5.0)|
|GT (one or two copies)|
|No drug||497 (79.5)||357 (79.0)||1.2 (0.82, 1.8)||1.3 (0.81, 1.9)|
|Coxib||38 (6.1)||43 (9.5)||1.9 (1.1, 3.3)||2.1 (1.1, 4.0)|
|No drug||73 (11.7)||40 (8.8)||1.0 (Reference)||1.0 (Reference)|
|Any NSAID||17 (2.7)||12 (2.7)||1.3 (0.56, 3.0)||1.8 (0.69, 4.5)|
|GT (one or two copies)|
|No drug||431 (69.0)||290 (64.2)||1.2 (0.81, 1.9)||1.3 (0.83, 2.1)|
|Any NSAID||104 (16.6)||110 (24.3)||1.9 (1.2, 3.1)||2.2 (1.3, 3.8)|
Because of the possibility of residual confounding in the case control design we also performed a ‘case-only’ analysis to explore the possibility of interaction (Table 5). This analysis yielded an interaction OR for the presence of the risk haplotype and COX-2 use of 2.2 although the confidence interval overlapped 1 (0.66, 7.6). While not significant, the direction of this estimate supports the idea that there may be interaction between the COX-2 risk haplotype and COX-2 inhibitor use, as noted in the full case control analysis. The interaction OR for other NSAIDS was close to 1.
Table 5. Case only study examining the COX-2 haplotypes by NSAID interaction on risk of acute myocardial infarction
|Any NSAID||No GT||10|
|OR (any NSAID) = 1.4 (0.71, 2.6)|
|OR (Coxib) = 2.3 (0.69, 7.6)|
|OR (any NSAID) = 1.4 (0.72, 2.7)|
|OR (Coxib) = 2.2 (0.66, 7.6)|
This study confirmed an increase in risk of ACS with recent ingestion of NSAIDs. The relative risk estimates were in line with those from other epidemiological studies and randomized trials. (5–7) We found weak evidence of a gene/drug interaction. We did not see this with joint exposure of drugs and individual SNPs, but saw minor variation in the risk of ACS with NSAIDs when we analysed by haplotype. If these observations are confirmed, it would indicate that in the 10–15% of the Australian population without the high risk haplotype (GT variants at both loci), the risk of myocardial infarction with NSAIDs is not elevated. However this finding, if true, would have limited clinical utility as the drugs are very widely used and the majority of the population possesses the ‘high risk’ genotype.
We were not able to confirm the reported protective effect of the rs20417 (G > C) polymorphism in the promoter region of the COX-2 gene. This has been extensively investigated, with mixed findings. Our results were consistent with several studies that have found no association between this SNP and risk of coronary thrombosis [20, 21]. We have separately reported that the rs20417 (G > C) polymorphism did not modify the risk of stroke although, compared with the wild types, both this variant and the rs5275 (T > C) polymorphism in the un-translated region of the COX-2 gene were associated with better functional outcomes after stroke .
Our study found no independent effect of the rs5275 (T > C) polymorphism. As far as we know, this is one of the first attempts to study the link between this polymorphism and ACS. Like the rs20417 polymorphism, rs5275 (T > C) is associated with reduced expression of COX-2 , which might have been expected to provide some protection against vascular thrombosis. However, our data do not support this hypothesis.
This leaves open the question of why low expressers of COX-2 might have reduced risk of the adverse effects of NSAIDs on the coronary circulation? Perhaps the most obvious answer is that in atherosclerotic plaque, COX-2 expression is up-regulated, leading, among other things, to increased prostaglandin E2 (PGE2) synthesis. This stimulates proteinase expression by plaque macrophages, degrading components of the extracellular matrix and promoting plaque ulceration and rupture [33, 34]. In this unstable situation, inhibition of prostacyclin (particularly if thromboxane synthesis is unimpaired) could trigger a vascular occlusion. Platelet function may also be relevant. Platelet activation and increased turnover are features of inflammatory conditions including atherosclerosis. Activated platelets induce COX-2 expression in adherent monocytes, leading to production of pro-thrombotic eicosanoids including PGE2 and thromboxane B2[35, 36]. Compared with normal expressers, low expressers of COX-2 might have diminished risk of these COX-2-driven prothrombotic processes and a reduced susceptibility to the adverse effects of NSAIDs. Conceivably, such individuals may also be less dependent on protective ischaemic pre-conditioning (which appears to be COX-2 dependent), rendering them less vulnerable to the effects of drugs that inhibit COX-2 but spare COX-1 [12, 37].
These possibilities are compatible with a general view of the importance of a balance between COX-2 dependent production of prostacyclin in the endothelium and COX-1 dependent production of thromboxanes by platelets [10, 11]. In practice, most of the NSAIDs in common use achieve high levels of COX-2 inhibition during routine use, but probably differ in the extent to which they spare COX-1 [9, 11]. For instance naproxen, which has been found repeatedly not to increase cardiovascular risk, achieves a high degree of COX-1 inhibition and its kinetics are likely to ensure that this is sustained throughout the dosing cycle .
This is one of the first investigations of a putative genetically based variation in cardiovascular risk from NSAIDs. Major early studies of COX-2 genotypes and cardiovascular risk did not report a NSAID consumption history [12, 19] or reported non-use of NSAIDs . However, other genetically based associations for this drug class have been studied, including the risk of gastrointestinal bleeding and the response to treatment with NSAIDs to prevent colorectal adenomas [15–17].
In this study, we investigated the impact of polymorphisms in the promoter and an un translated region of the COX-2 gene. These could be viewed as ‘pharmacodynamic’ response variables, but genetically determined variations in the pharmacokinetics of NSAIDs, leading to reduced biotransformation of the drugs, could also increase the risks with these drugs . Putative variants are polymorphisms of drug metabolizing enzymes: CYP2C9 (celecoxib, piroxicam and possibly diclofenac); CYP2C8 (naproxen, ibuprofen) and UGT1A (indomethacin, ibuprofen diclofenac, naproxen) [15, 39, 40].
The strengths of this study are that the cases were carefully validated, and drug consumption history was determined by direct interview, rather than reliance on administrative records. Controls were recruited from the same community as the cases. An identical protocol was used to take drug consumption histories from cases and controls; the latter were recruited from the same population as the cases. The direct contact also enabled collection of DNA from cases and controls, which has not been possible in any previously published pharmaco-epidemiological study of the cardiovascular effects of NSAIDs. It should also be noted that this was a targeted study, which analysed two genotypes selected a priori on the basis of published evidence and their likely functional significance.
The study has several limitations. The major concern is the rather small population size and the modest differences detected upon subgroup analysis. With the many disease associations with genetic variants studies that have recently been published, and the poor reproducibility of many of the results, the requirement for further supportive studies is clear . This would require reproduction of the results in an independent population and/or functional demonstration (in patient blood samples) of COX-1/2 levels and activity, something that was not possible here. In addition, our study was too small to study effects of varying doses of the drugs.
This was a case control study, meaning it is dependent on self report of drug ingestion and the possibility of various recall biases. It was not a matched study. The control group was drawn from the community in which cases occurred and the differences in age, gender and other variables were expected in view of the aetiology of coronary events. These factors were fully adjusted for in the analyses. If we had tried to match at an individual level, we would have reduced the statistical power of the study. Despite the adjustments we made, there is a possibility of residual confounding. However, it is not immediately apparent what would cause this, and the case only analyses, which are not affected by case/control differences, tend to support the conclusions of the main analysis.
Finally, while this paper was under consideration we identified two recent studies that examined possible interactions between polymorphisms of the COX genes and use of NSAIDs. In a large nested case cohort study in Denmark, Vogel and colleagues found no interaction between two polymorphisms of the COX-2 gene: rs5275 (studied here) and rs689466, and recent consumption of NSAIDs . In a case only study of subjects with ACS, St Germaine and colleagues in Quebec, Canada studied a total of 14 polymorphisms of the COX-2 gene and found no interaction with use of NSAIDs . Significantly, St Germaine and colleagues also found no interaction between polymorphisms of CYP2C9 and NSAID use. However, they found interactions between two independent polymorphisms of the COX-1 gene and use of NSAIDs, concluding that this needed confirmation in larger cohort studies . Currently, this latter finding is the strongest evidence for genetically determined variations in the risk of cardiovascular events with use of NSAIDs, albeit with COX-1 rather than COX-2 polymorphisms.
In conclusion, We found little evidence of a gene/drug interaction between two COX-2 polymorphisms and use of NSAIDs. We found a trend towards a lower risk of coronary events with NSAIDs in the presence of the ‘low risk’ haplotype but this is unlikely to have clinical utility, at least in Caucasian populations among whom the polymorphisms are present in a minority.
MS, LL and LB have received funds for research from Pfizer.
The study was supported by grants from the National Health and Medical Research Council of Australia and the National Heart Foundation of Australia. Financial support for reagents used in the genotyping of samples was provided to Dr Michael Seldon's laboratory by Pfizer Australia. None of the funders had any involvement in the design analysis or reporting of the study. We are grateful to all who participated in the Hunter Community Study