The studies that found no benefit missed a drug known to work, too
Sertraline missed placebo in the 340-person national trial, and fluoxetine finished behind placebo on the main depression score in a second.
What the study found
In both US trials that found nothing, the prescription drug failed too
That pattern points at the trials rather than at the herb. Sertraline missed placebo at p=0.18 in the 340-person national trial, and fluoxetine finished behind placebo on the main depression score in a second. A study that cannot detect a working drug cannot rule one out.
The caveat · One extract, one dose, in each trial
Both trials tested a single extract at a single dose, so neither says anything about preparations that differ in strength. The 340-person trial reports no trend toward benefit whatsoever, which an insensitive trial would not be guaranteed to produce. The smaller trial enrolled about a quarter fewer people than it planned to.
What we don't know
Two large US trials found no benefit where pooled trials find one, but in both the prescription antidepressant also failed to beat placebo. The open question is whether those trials could detect any effect at all.