Headline should read: “Retrospective data analysis interpreted to imply a possible relationship after statistical tinkering.” While this comes from a legit paper, the implication of causality in the headline is problematic.
From the actual paper: “We use matching techniques to control for confounding and Cox proportional hazards regression models...” This is neither an experimental study, nor an obvious, straightforward difference. Further, even if there is a 1.5 odds ratio, it could run the other way: “Mothers with loading for autism spectrum disorders more likely to be identified as cannabis users while pregnant.”