Skip navigation

Coxibs, aspirin, polyps, and cardiovascular harm


Clinical bottom line

Aspirin and coxibs increase the risk of cardiovascular events in patients at risk of colorectal adenomas. The absolute increased risk is about 0.5% a year of treatment.


JA Baron et al. A randomized trial of aspirin to prevent colorectal adenomas. NEJM 2003 348: 891-899.

RS Bresalier et al. Cardiovascular events associated with rofecoxib in a colorectal adenoma chemoprevention trial. NEJM 2005 352: 1092-1102.

SS Solomon et al. Cardiovascular risk associated with celecoxib in a clinical trial for colorectal adenoma prevention. NEJM 2005 352: 1071-1080.


Use of NSAIDs and aspirin reduce the risk of colorectal cancer and adenomas. Most of that information comes from observational studies, many of which show an association between NSAID or aspirin use, and lower rates of colorectal cancer or adenoma.

Despite there being a lot of observational data, much of which has been well collected and presented, this does not mean that we can be sure that NSAIDs or aspirin prevent colorectal cancer or adenomas. A Bandolier review of studies published up to 2004 concluded that overwhelming evidence for the efficacy of aspirin, NSAID, or coxib on adenomatous polyp development was lacking, though aspirin in three trials had a number needed to treat with 150 to 325 mg aspirin for one to three years of 16 (95% confidence interval 9 to 56) for one patient to be free of adenomas.

One of the three aspirin trials reported increased heart attacks and stokes with aspirin compared with placebo over three years, while the others made no comment on these adverse events. Two long-term studies with coxibs have now published, and also show increase cardiovascular events in these patients. These are examined and pooled below. There may be other studies not yet published, so we do not yet have the full picture.


All three studies were designed to last about three years, and to investigate effects of coxib or aspirin versus placebo on colorectal adenomas in patients with recent history of histologically documented adenomas. Details of patients in the three trials are shown in Table 1. There was considerable consistency between them, though the earlier aspirin trial did not report on some details.

Table 1: Details of patients in the three trials

81 or 325 mg
25 mg
200 or 400 mg
Number (all patients)
Duration of follow up (months)
minimum 34
Mean age
Age range
Male (%)
Low dose aspirin (%)
Antihypertensive medicine (%)
High cardiovascular risk (%)
History of hypercholestrolaemia (%)
History of diabetes (%)
Cigarette use (%)

The three trials reported cardiovascular event outcomes differently. The aspirin study reported deaths, myocardial infarctions, and stroke. We do not know whether deaths and MI were the same persons, so in the analysis below only MI and stroke were pooled.

The rofecoxib and celecoxib studies reported various outcomes, most importantly the antiplatelet trialist collaborators (APTC) outcome of combined incidence of death from cardiovascular, haemorrhagic or unknown origin, nonfatal MI, nonfatal ischaemic or haemorrhagic stroke. Other outcomes were also detailed.


In each of these three trials, the incidence of APTC events or MI plus stroke was higher with aspirin or coxib than with placebo (Figure 1). Overall 1.4% of participants had an event over about three years with placebo, and 2.8% with aspirin or coxib. Treating patients with colorectal polyps with any dose of aspirin or coxib for about three years doubled the risk of an APTC event with a relative risk of 2.2 (95% confidence interval 1.5 to 3.4).

The number of people needed to be treated for about three years for one more to be harmed in this way by aspirin or coxib was 71 (47 to 150).

Figure 1: APTC events or MI + stroke with aspirin (darker yellow) or coxib (lighter yellow) compared with placebo.

For aspirin and celecoxib where two doses were used, there was a dose-response relationship, with higher event rates at higher doses (Figure 2).

Figure 2: Crude incidence rates for APTC events for placebo and particular doses of aspirin or coxib (number of events at end of bar)

Death from any cause was no different with aspirin or coxib (32/3392, 0.9%) than with placebo (19/2350, 0.8%). The relative risk was 1.1 (0.6 to 2.0).

The rofecoxib study investigated and reported on hypertension. Mean systolic blood pressure values increased by 3.4 mmHg with rofecoxib, compared with no change with placebo. The increased risk of thrombotic events was similar at different categories of blood pressure increase. Serious hypertension, serious oedema, and serious heart failure were more common with rofecoxib than with placebo. Combined they occurred in 26/1287 patients (2%) on rofecoxib and 3/1299 patients (0.2%) on placebo.


This is clearly a high risk of a cardiovascular death, or nonfatal heart attack or stroke in people who were at risk of developing colorectal adenomas. Table 1 shows that the two coxib studies enrolled some patients at higher risk of cardiovascular problems, though this was not clear from the aspirin study, in which aspirin use was an exclusion criterion.

The absolute difference was about 0.5% per year, equivalent to an additional risk of 1 in 200 persons. Given the relatively young age of the patients in these trials mostly below 60 years, a 10-year additional risk of about 5% is much too high, especially for outcomes like heart attach, stroke, or death. The two coxib studies do not tell us whether there was any benefit in terms of reductions in adenomas or cancer to offset any increased risk.

Another unknown is the realationship between increased cardiovascular risk with aspirin and coxibs in these patients, and that occurring in patients with arthritis for whom there is little evidence of increased risk, except at elevated doses of rofecoxib. The fact that the increased risk seems to be present with aspirin might make us think that perhaps patients with a history of colorectal polyps may be different from those we usually treat.

And finally, there may be other studies that have not yet reported, but in which the increased risk is not as high as in these three studies. As more are reported, Bandolier will attempt to update this analysis (April 2005).