★★★Well replicated — this finding’s sources average out to meta-analyses, systematic reviews or high-powered trials.
The trials disagree on how well it works, and size is what splits them
Sorting the placebo-controlled trials by size splits the answer: 1.28 times as many responders in the nine larger ones, 1.87 in the nine smaller.
What the study found
The smaller the trial, the bigger the benefit it reported
Sorting the placebo-controlled trials by size splits the answer: 1.28 times as many responders in the nine larger ones, 1.87 in the nine smaller. A 2026 network analysis of 163 trials flagged its own St John's Wort estimate, roughly twice the established one, as unrealistic.
The caveat · Size also tracks rigour and funding
Trial size stands in for several things at once, including how carefully a trial was run and who paid for it, so this does not isolate a single cause. The 2026 estimate rests on three trials in 207 people, all of one proprietary extract.