I'm not familiar with the Androgel formulation -- I know the generality of what it is comprised of, but do not know what happens on the surface of the skin when it is applied and dries, other than that they claim 10% delivery -- and so cannot say much.
But can say that even if say 1/10th of the applied amount (seems unlikely) wound up going to another person, and somehow wound up on them applied just as effectively as directly applying the gel (seems unlikely) then they'd be getting a 1/10th dose.
Back in the 50s and 60s anabolic steroids actually were given clinically to children in a non-trivial number of cases. Oral synthetics, true. But the fact was they could tolerate quite a bit, every day, without outcomes that disturbed the doctors at all. Even recently there was a study on use of oxandrolone for use in young girls with Turner's syndrome. It is not the case that children are, as in the House episode, hypersenstive to androgen. I don't think they should be given it, certainly not if there is no really compelling medical reason and there rarely if ever is, but they are not hypersensitive.
So even if the Androgel delivered 10% of a dose, which seems a probably-implausibly-high estimate, occasionally getting that seems unlikely to me to yield these outcomes.
Besides, why only 8 cases if that was the mechanism?
And 1% might be a more rational guess on amount of delivery.
The mechanism of left-the-drug-in-reach child-imitates-parent, on the other hand, seems to me (not to repeat myself) quite likely to generate say 8 cases what with the large number of Androgel users and not every parent being ultracareful and not every young child having enough sense to not do such a thing.