I was answering to the "Correlation is not causation" statement in the comment. Epidemiological studies usually have to satisfy most of the "Bradford Hill" criteria in order to claim a certain association.

More information: http://www.drabruzzi.com/hills_criteria_of_causation.htm