Does the varied use of NSAIDs explain the differences in the risk of myocardial infarction?