This is a very reasonable objection to the Andreasson study, which is one of the reasons I linked the Zammit study instead: it controlled for that variable and the correlation persisted anyway. I really encourage you to read the study, they do a number of other clever things too.