> they do not take into consideration that other things also cause autism
Fortunately, we have figured out how to tease out the influence of a single variable among many: test vs. control groups. If the sample size is large enough, most variables will average out to have the same effect in both groups.