Dated final reporting and synthesis decisions, 1 October 2026

These decisions were made after the results were known and are not prospective.
The 22 September protocol and amendments are retained unchanged for transparency.

1. The final report uses systematic evidence mapping and comparator-specific
narrative synthesis. No complete pooled psychological-treatment or CBT effect is
claimed. Historical overlapping waiting-list pools are supplementary exploratory
sensitivities selected by source availability. Their membership and estimators
must not be chosen according to statistical significance.
2. Retrieval is closed at 1 October 2026 for this version. The 349 unretrieved
database reports and two retrieved but unclassifiable reports remain unresolved,
not excluded as ineligible. The 60 supplementary citation leads remain separately
identified. This closure is a reporting cutoff, not proof of exhaustive retrieval.
3. All 43 primary efficacy reports are accounted for in 42 conservative synthesis
units. The two Shabahang reports are grouped without asserting independence or
proving a single trial. Weck 2013 is linked supporting evidence. No unique total
participant count is asserted across uncertain overlap or eligible subgroups.
4. Numerical results are restricted to 45 source-supported rows. Reported effects
without a Stata row and eligible studies without usable outcomes remain in the
evidence account. Fourteen named units have no current Stata row. No whole-review
adult-only sensitivity, complete remission/harms/service-use synthesis or
comprehensive treatment ranking is claimed.
5. The contextual stream is supplied as an inventory with selected illustrations.
It is not a completed thematic synthesis of all contextual reports. Design-specific
appraisal coverage is limited and explicitly recorded. Historical case reports are
an illustrative appendix; the initial title rules did not exhaustively capture them.
6. There are 57 result appraisals, 40 exact matches to the 45 numerical rows, and
five limited-applicability rows. Ten finite certainty profiles are reported. They
must not be described as comprehensive treatment-class certainty assessments.
7. AI assistance and single-author conduct are disclosed. Separate AI passes and
independent implementations of the arithmetic are not independent human review.
