Screening process
Every number below is measured from the pipeline's own artifacts rather than entered by hand. Papers we could not obtain are counted openly — an unread paper is a limitation of the review, not an exclusion from it.
The full-text screen is this review's eligibility pass. An abstract-level relevance gate was deliberately skipped: weight is usually reported only in a safety table, so an abstract screen would reject exactly the papers the review exists to find — a false negative that no later stage can recover.
Study designs among eligible papers
Exposure strata
Never pooled together: drinking-water studies associate higher lithium with LOWER obesity, and anorexia trials give lithium in order to induce weight gain.
How weight was measured
Continuous kg and “proportion gaining ≥7%” are different questions — the disagreement between the two is the central finding of this literature, so they are never merged.