We should be cautious about embracing data before it is published in the academic press, and must always avoid treating correlation as causation.