This is a serious mischaracterization. The paper found that one flawed trial inflated fluoxetine’s estimated benefit. Removing it reduced the effect substantially, but the pooled estimate still favored fluoxetine and excluded zero. “No better than nothing” is not what these data show.
