Hmm, so I get that you can't compare between the age groups, because that would result in confounding the age variable. But, could you also just compare within an age group and not across them? Then you wouldn't be confounding for age, you would actually be controlling for age, which is how I interpreted the answer.