Supplement 2: Data, evidence account and Stata reanalysis

This archive contains aggregate extracted research summaries, not individual
participant data or copies of copyrighted source publications.

Selection/: all 8,799 database dispositions, current 43 primary efficacy reports,
77 context/illustrative records, unresolved reports and additional citation leads,
with public identifiers, actual source-read levels and reasons. These records and
reports are not independent trial counts. selection_counts.json reconciles them.
Evidence/: complete public study profiles, result-specific risk-of-bias and finite
certainty tables, numerical result crosswalk and references.
Protocol/: original dated protocol and amendments, plus retrospective final
reporting decisions. Read these limitations before interpreting pooled results.
Stata/: reproducible scripts, aggregate inputs, source provenance and numerical
outputs. Set the working directory to Stata/ and run do run_all.do in Stata 19+.
No user-written packages are required. Save unrelated data before running.

The 45 rows include repeated times, shared arms and alternate estimands. Consult
the evidence crosswalk rather than treating these as 45 independent trials.
baseline/pools.csv contains four overlapping exploratory subsets with three
methods each; these are not four independent confirmations or the whole body.
New figures are generated by make_comparison_figures.do and are unpooled.
The validation file records 857 passed reproduction checks; arithmetic agreement
does not verify completeness of screening, source access, or independent review.

Use was AI assisted as disclosed in the manuscript. No repository DOI is claimed.
Copyrighted full-text publications remain accessible through their cited sources.
