[michaericalribo]

This is true, however, the sample size is quite large, they controlled for confounders, and they address the possibility that this is all a coincidence:

> Whilst we controlled analyses for all known confounders and were more thorough in this than any previously published study, we cannot exclude the possible of residual confounding. However, the sensitivity analyses estimated that unobserved confounding would need to be of a greater strength than any recognized observed confounder, including age and smoking, to obviate the association between alcohol and brain health, which seems implausible.

As an informal / heuristic argument, it seems practical to conclude there is some causal relationship.

[jeeeb]

Controlling for confounders and demonstrating causality are different things though. They explicitly acknowledge that they don’t demonstrate causality as a study weakness.

People who drink alcohol may simply tend to have a slightly different brain structure.

In this regard it would be more interesting to see a large scale longitudinal study demonstrating an effect in the same individual over time.

Combined with population level studies and a demonstrated mechanism, I think that would be fairly definitive.