Making sense of these data illustrates the basic problem of epidemiology: how to distinguish events that are causally related from events that are coincidentally related. In this case—12 reported deaths over four years, during which about a million patients took about 30 million doses of Adderall or Adderall XR—a meaningful analysis would compare the annual death rate that might be attributed to the medication with the annual death rate from similar causes for children in the general population. But for how long, on average, do children take Adderall and Adderall XR? If the answer is one year, then the annual death rate is 1.2 per 100,000 users per year. But if the medication is taken for four years on average, then the annual rate drops to 0.3 per 100,000 users. Similarly, what is our comparison group? The drug manufacturer tallies and reports "sudden, unexplained" deaths, but annual vital statistics for the population as a whole track "cardiac deaths" that are subdivided by age ranges that might well not exactly match the range for the children who died while taking these drugs. Assume, however, that the ranges are similar enough to make a comparison, and assume some worst-case assumptions. That gives you:

1 million medication users in a cohort of about 41 million similarly aged children;

4 annual deaths among users and 257 annual cardiac deaths in the cohort as a whole;

or an annual death rate of 4 per 1 million for Adderall-treated kids compared with 6.3 per 1 million cardiac deaths for the cohort.