I think there are too many confounding factors to make a strong claim either way. For example, the older a mother is the more risk there is of a child having do

I think there are too many confounding factors to make a strong claim either way. For example, the older a mother is the more risk there is of a child having downs syndrome, and that in turn is associated with higher risk of ALL/AML. So without considering other factors, you may be led to think that the younger siblings might have a higher risk (remember that the paper suggested that a first hit mutation may be required before the infection). My point is just that to test an epidemiologic hypothesis you have to really think about all the different confounding factors that could bias your results.