Comment by svara

It does not require more explanation: publication bias means null results aren't in the literature; do enough small low quality trials and you'll find a big effect sooner or later.

Then the supposed big effect attracts attention and ultimately properly designed studies which show no effect.