Part of the problem here is that it's difficult to study alcohol's effect on fetuses, because it would be unethical to give it to pregnant subjects as part of a clinical trial. We know that excessive drinking is toxic to a fetus, but we don't know where the line is. So any experiment where pregnant people were asked to drink would be a massive breach of ethics. But that doesn't mean we can't study it at all. Some people do drink while pregnant, especially outside of the United States. If we could follow enough of those parents and compare their babies to those of people who didn't drink at all, we could start to get a sense of how dangerous a few drinks might be. You would certainly have trouble getting people to accurately report how much they drink, especially in countries where it's taboo, but an abundance of potentially flawed data would still be better than the relative absence we have now. Meta-analyses like the one described above can help pinpoint the truth from a mess of conflicting, imperfect studies.