Aspirin is an old drug. Some 2000 years ago the Greeks used the bark and leaves of the willow tree (which contains salicylic acid) to relieve pain and fever. At the end of the 19th century acetyl salicylic acid started being produced on an industrial scale and aspirin soon became a widely used painkiller. In the late 1960s it was found that a single dose of aspirin irreversibly inhibits the normal aggregation of platelets by suppressing the cycloxygenase mediated synthesis of thromboxane A2. The effect of aspirin persists until newly formed platelets have been released; their biological lifespan is about nine days.
Sometimes, adverse drug reactions can be turned to advantage, and the antithrombotic effects of aspirin have been widely exploited. Aspirin has been given successfully both to healthy people and to patients with coronary artery disease to prevent myocardial infarction, to patients with acute myocardial infarction to reduce vascular mortality, to patients with atrial fibrillation to prevent stroke and systemic embolism, and to patients with a history of stroke or transient ischaemic attack to prevent further ischaemic events.1–5
It has been suggested that aspirin should be given to all men aged 50 years or older and to all women after the menopause.6 There are, however, two related reasons why this enthusiasm needs to be dampened. Firstly, aspirin, through its ability to block the synthesis of prostaglandins, may damage the gastrointestinal mucosa.7 Erosions may be trivial, but they may progress to ulcers, which in turn may bleed or perforate, and may even kill. This happens more often than many doctors like to believe.8 Secondly, reduced thrombus formation results in a greater tendency to bleed. Thus, in a patient who is taking aspirin any ulcer that may arise may bleed even more extensively.
It seems, therefore, a logical aim to use the minimum dose of aspirin that inhibits thrombus formation and also minimises the risk of any gastrointestinal complications. Thus, physicians have been treating their patients with low dose aspirin on the understanding that they did more good than harm to them.
There is now strong evidence in this issue of the BMJ from yet another systematic review of randomised trials that this may not necessarily be so (p1183).9 These authors tested the risk of gastrointestinal haemorrhage from long term (average duration 28 months) treatment with subanalgesic doses of aspirin used for preventing ischaemic events. There were two main results. Firstly, gastrointestinal haemorrhage occurred on average in 2.5% of patients exposed to aspirin compared with 1.4% who were not; this difference was statistically significant. Secondly, there was no evidence of dose responsiveness over a wide range of doses (50 to 1500 mg/day). How do these numbers compare with others, and what are the implications?
A previous systematic review reported a considerably lower incidence of gastrointestinal bleeds with non-steroidal anti-inflammatory drugs including aspirin, and there was evidence of dose-responsiveness for bleeds that were related to aspirin.8 The difference between the two meta-analyses is probably due to different definitions of adverse events. For instance, in the physician's health study,1 a large randomised trial included in both meta-analyses,8,9 of 11037 patients given aspirin 325 mg every other day for 60 months, as many as 3.6% had symptoms of haematemesis or melaena. This was the level of harm that Derry and Loke were extracting for the purpose of their systematic review.9 Using this definition throughout, the number needed to harm for haemorrhage with aspirin compared with placebo was about 100, and there was a lack of dose responsiveness.9 However, in the same randomised trial 10 times fewer patients taking aspirin (0.34%) had a potentially fatal haemorrhage compared with those taking a placebo.1 Using this more serious level of harm, there was evidence of dose responsiveness8: the incidence of serious bleeds (and perforations) was 0.3% with 325 mg aspirin every other day for 60 months,1 0.6% with 1 g/day for 36 months,10 and 0.9% with 2.5-5.2 g/day for two months.11 An earlier meta-analysis has also shown a consistent tendency (although not statistically significant) for a smaller risk of gastrointestinal bleeds with smaller doses of aspirin (<300 mg/day).12
The results of these meta-analyses are not contradictory but complementary.8,9,12 Indeed, the most important message in Derry and Loke's paper is that there is no gain without pain. And as with many systematic reviews, their's leaves more questions open than it answers. Thus, the research agenda is set: Who should be given what dose of aspirin, and for how long? In patients with a history of stroke or transient ischaemic attack, the minimal effective dose of aspirin to prevent further vascular accidents remains unknown.13 Nor do we know how long patients have to take aspirin. In the prevention of recurrent stroke aspirin seems to be of benefit independent of the patient's age.5 However, in elderly patients with atrial fibrillation the benefit of prophylactic aspirin to prevent strokes is unproved.4 Also the risk of both gastrointestinal complications and perhaps congestive heart failure with non-steroidal anti-inflammatory drugs may increase with increasing age.14,15
Finally, there is a methodological message. Derry and Loke analysed data from almost 66000 patients chronically exposed to a wide range of different doses of aspirin. It is unlikely that the same body of data would ever be tested in a single randomised controlled trial. Innovative models are needed to estimate rare events with confidence, and systematic reviews currently provide the best solution. In the light of Derry and Loke's analyses, it may be more appropriate for some people to eat an apple rather than an aspirin a day.6