Paracetamol is widely used as an over the counter analgesic because it is perceived to be safer than aspirin. In carefully selected trial populations, aspirin therapy carries a relatively low (but statistically significant) absolute risk increase in gastrointestinal adverse events. The risk from aspirin may potentially be higher in individuals who have risk factors for peptic ulcer or gastrointestinal haemorrhage. As such, paracetamol may well be the preferred choice in many instances.Keywords: acetylsalicylic acid, adverse drug reaction, analgesia, gastrointestinal, pain reliefIn this issue of JCPT, Steiner and Voelker (1) look at the gastrointestinal tolerability of short-term aspirin use for acute pain and conclude that on safety grounds, there is no good reason to prefer paracetamol over aspirin. In other words, aspirin offers an acceptable adverse effects profile, and we should not be averse to use aspirin for acute pain, especially as there is ample evidence for its analgesic efficacy. Of course, preparations of paracetamol and aspirin are widely available on the shelves of convenience stores, and for the patient with a headache, it would be nice to have some evidence to guide the choice of treatment. Does Steiner and Voelker's meta-analysis help us reach a definitive conclusion?Let's take a closer look at the data in this metaanalysis. There were nine randomized doubleblind trials of single-dose aspirin (1000 mg) vs. placebo in patients with headache or pain following dental extraction. The authors combined the results in an individual patient data meta-analysis to calculate pooled relative risks (RR) and numberneeded-to-harm (NNH). We need to pay particular attention to the exclusion criteria which meant that patients with intolerance to NSAIDs, or a history of peptic ulceration ⁄ gastric bleeding were excluded. Patients with any disorder of kidney, liver, lung or heart were also excluded, and by any measure, one would consider the trial participants to have had a very low baseline risk of gastrointestinal (GI) adverse events.Nevertheless, meta-analysis of the trials found a significantly increased risk of GI adverse events with aspirin compared to placebo, RR of 1AE7 [95% Confidence Intervals (CI) 1AE7-2AE4], and an NNH of 42 (95% CI 25-111). Steiner and Voelker then conducted an analysis on a subset under the heading of 'adverse drug reactions' (ADRs), defined as adverse events that were judged by the investigator to have some likelihood of a causal relationship with the treatment. This sub-analysis does not show a statistically significant increase in risk of GI ADRs, and the authors 'argue that ADRs are more meaningful, and more relevant to patients' choices.' However, in an adequately blinded, placebocontrolled randomized trial, the subjective act of classifying an event as an ADR merely adds a layer of bias. It simply isn't necessary to introduce the preconceptions and expectations of trial investigators here. Indeed, one wonders how many myocardial infarctions in the rofecoxib trials would have es...