You can manipulate statistics to show a link between just about any pair of variables. Massage this. Try a dozen different formulas to calculate significance. And tada! You have the link you were digging for.
The article I read about this study this morning stated: "But after accounting for all lifestyle factors, the researchers found the link to dementia was statistically insignificant." Despite their best efforts, I'm sure. Yet the clickbait headline and lead still blare out "dementia!!"