From the article:
There are some important things to keep in mind here, like the fact that 15 of the 17 studies were case-control studies. This kind of research cannot prove cause and effect, and it often does not look at things that might have affected both the exposure and the outcome.
A number of the studies that were looked at were of low quality, which is something that the authors also highlight.
Findings were inconsistent across studies, but those of higher quality suggested that associations in unadjusted models might have been due to factors that could have influenced the results.