.... but it sounds like you're saying that, for example, exposure to benzene and increased occurrence of myeloid leukemia among people with different ethnicities and hereditary risks is purely coincidental.
Not at all. I have stressed that there is a statistical correlation, which means (at p<0.05) that this is NOT purely coincidental. But a statistical correlation does not mean that we understand the cause. In the case of viruses we know how the virus inserts differing instructions into the cell DNA, altering cell division rates - this is obviously very close to promoting a tumour. We know that ionizing radiation can break up DNA, which the cell's error-correcting processes may then reassemble incorrectly, but this is a much more random activity. In the case of mutagens we know very little about what is happening, and we have quite a few competing theories, many of which will turn out to be wrong.
My comment was a throwaway line, but I was trying to say something quite precise. I get quite worried about the use of stats, particularly in things like 'evidence-based medicine', because they are so easily misused by activists. Stats are useful for pointing you in the direction of an answer, but I don't believe that you have the answer until you understand the mechanism. Sometimes it is well worth taking action on the basis of statistical correlation alone - John Snow knew nothing about the germ theory of disease when he had the Broad Street pump handle removed - but it is also possible to go off in a wrong direction. I am deeply sceptical by instinct and training, and like to stress that correlation is not causation wherever I can...