Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Myocardial infarction: aspirin, NSAIDs, and COXIBs


Hands up anyone not presently confused about whether coxibs or NSAIDs do or do not cause heart attacks? We have had lots of papers, randomised trials, meta-analyses, and observational studies, none of which wholly give an answer that makes complete sense. Regulatory bodies and their experts are having to re-evaluate information and guidance, so there is little hope of ordinary folk understanding what is going. A large Danish observational study [1] does much to help.


Denmark provides an excellent source of observational data because it has a national health service and superb facilities to produce unambiguous linkage between what happens in hospitals and in the community.

Over four years all patients with a discharge diagnosis of myocardial infarction over 20 years old living in a community of 1.4 million people were identified. Ten controls were selected for each case matched by age and sex. Prescriptions for NSAIDs (except low dose 200 mg ibuprofen tablets), coxibs and high dose aspirin (200-500 mg) were obtained. Use was classified by current (filled prescription within 30 days), new users with a first prescription within 30 days, and recent users (30-90 days) or past users (prescription filled more than 90 days before an event).

Possible confounding factors were obtained from discharge and prescription history, and the confounding factors used to adjust crude relative risks. About 20 separate confounders were used for adjustment. Relative risks were calculated with reference to non-use of coxib or NSAID, which was lower in cases than controls, and which consequently resulted in an increase in relative risk of about 12%.

Risk of myocardial infarction was defined where possible. High risk was defined as at least one previous hospital admission for cardiovascular disease, diabetes, hypertension, and prescriptions for cardiovascular or diabetes medicines. Low risk was absence of these factors.


There were 10,280 cases and 103,000 controls. The average age was 70 years, with a range of 19 to 101 years. About 60% were men. Cases were much more likely than controls to have discharge diagnoses of cardiovascular disease, hypertension, diabetes, and respiratory problems. They were more likely to have prescriptions for platelet inhibitors, antihypertensive drugs, diabetes medicines, lipid lowering drugs and nitrates. Use of NSAIDs or coxibs was 55% in cases and 49% in controls.

Coxib and NSAID use is shown in Figures 1 and 2. Over 30% of cases and controls were former users of NSAIDs. Crude relative risks compared to non-use are shown in Figure 3 for all coxibs combined, and for all NSAIDs combined.

Figures 1 and 2: Percent of cases and controls using coxibs (top) and NSAIDs (bottom)

Figure 3: Crude relative risk for all coxibs and NSAIDs by type of use compared to nonuse

For current users the adjusted relative risk was significantly increased with use of aspirin, non-naproxen NSAIDs, rofecoxib, and other coxibs, but not naproxen or celecoxib (Figure 4). Point estimates ranged from about 1.3 for aspirin and celecoxib to 1.8 for rofecoxib. The smallest number of cases in this category was 26 for naproxen, and for other analyses there were always more than 50 cases.

Figures 4 and 5: Relative risk of MI admission for current users (top) and new users (bottom), compared to nonuse

For new users, the adjusted relative risk was significantly increased for aspirin, coxibs and non-naproxen NSAIDs (Figure 5). For naproxen there were only 4 cases who were new users. Point estimates ranged from 1.3 for aspirin to 3.4 for other coxibs. The number of cases was smaller in this analysis, mostly between 20 and 60.

The relative risk for recent and former use was always lower than for current or new use. Significantly increased relative risk occurred only for former users of coxibs other than celecoxib or rofecoxib (relative risk 1.2; 95% confidence interval 1.01 to 1.4), former users of high dose aspirin (1.3; 1.1 to 1.4), and recent users of non-naproxen NSAIDs (1.2; 1.1 to 1.3).

Analysis according to cardiovascular risk was restricted to a sub-group for whom there was information to calculate risk. Relative risk was significantly higher for rofecoxib and non-naproxen NSAIDs in low risk and high risk persons (Table 1), though numbers used to calculate this were small. There was no association between the number of filled prescriptions and risk of myocardial infarction for any drug category.

Table 1: Relative risk in persons with low and high cardiovascular risk. Statistically significant results bold, shaded

Adjusted relative risk (95% CI)
Low risk
High risk
2.0 (0.89 to 4.3)
1.4 (0.82 to 1.6)
Other NSAIDs
2.0 (1.6 to 2.5)
1.6 (1.4 to 1.8)
1.8 (0.92 to 3.1)
1.2 (0.85 to 1.5)
2.8 (1.7 to 4.5)
1.6 (1.2 to 2.0)
Other coxibs
1.8 (0.98 to 3.3)
1.1 (0.8 to 1.6)


Until now we have had only analyses with relatively small numbers of events, and with different comparators. By and large, in people with arthritis, meta-analyses of randomised trials and observational studies agree that there is no large difference between coxibs and NSAIDs, other than rofecoxib doses of 50 mg a day or more, where there was a consistent effect. Comparisons with placebo in randomised trials have been limited by small numbers of events in studies of limited duration.

In three-year studies comparing aspirin or coxibs with placebo to prevent colorectal adenomas, the published information to date shows consistently increased rates of cardiovascular events, with dose response, even for low dose aspirin. In three-year studies comparing NSAIDs or coxibs with placebo in Alzheimer disease prevention, there is no difference between coxibs and placebo.

The strengths of this Danish study include its large size, with 10,000 cases, its population-based design, and the fact that it looked at coxibs, NSAIDs, and high dose aspirin. It collected information on many confounding variables, and used them to adjust relative risk. This is important, because cases were different from controls, with much more cardiovascular disease, hypertension, and diabetes.

The major weakness is that it could not collect information on dose, likely to be a major factor. Nor could it collect information on lifestyle or diet, again important determinants of cardiovascular risk.

The study tells us that current and new users of coxibs, NSAIDs, and aspirin are associated with a higher risk of hospital admission for myocardial infarction. That risk may lower with naproxen and celecoxib, though the strength of that conclusion is limited by having only 26 cases for naproxen and 71 for celecoxib. Even a very large study like this has size limitations, because only about 2% of cases and controls were current users of coxibs. Analysis by individual drugs further reduces the number of events for analysis.

A similar, large case-control study from the UK [2] had almost identical conclusions. It found a 20-30% increased risk for first heart attack with use of most coxibs and NSAIDs within three months. The results with celecoxib were not statistically significant, as here. There was no significant increased risk for coxibs beyond three months, but there was a residual risk beyond three months with diclofenac and NSAIDs which were neither ibuprofen nor naproxen.

The best, simplest, conclusion is that coxibs, NSAIDs, and aspirin confer an excess risk of myocardial infarction. Consistently, though, no statistically significant increased risk is being found for celecoxib, though again without dose information. Increased risk may have been difficult to spot previously because a small risk can be lost where the overall risk is high, and we weren't looking. Until recently we have not been looking for cardiovascular ill effects of NSAIDs and aspirin.

A great deal of thinking will need to be done. There will be, and have been, suggestions that all these drugs, including over the counter analgesics, should be withdrawn. But in both these large studies half the patients were present or former users of NSAIDs or coxibs. Alternatives are few, with problems of their own.

The aim should be to maximise efficacy and effectiveness, while minimising both common and reversible adverse events and rare but serious ones. There are balances that need to be struck, new ways of thinking to be found, new research initiated about better and safer use of these medicines, but that may be baying for the moon.


  1. SP Johnsen et al. Risk of hospitalization for myocardial infarction among users of rofecoxib, celecoxib, and other NSAIDs. Archives of Internal Medicine 2005 165: 978-984.
  2. J Hippisley-Cox, C Coupland. Risk of myocardial infarction in patients taking cyclo-oxygenase-2 inhibitors or conventional non-steroidal anti-inflammatory drugs: population based nested case-control analysis. BMJ 2005 330:1366-1342.

previous or next story