You should interpret it the same way you interpret any other study that demonstrates only correlation and fails to control for obvious confounders.