Searches through 22 September 2026 identified 8,152 PubMed and 7,673 Europe PMC occurrences. Removal of 7,026 duplicate occurrences left 8,799 unique records. All titles were recorded as read during AI-assisted first-pass screening; 803 available abstracts or source excerpts were also read in that pass. The screening process did not constitute independent duplicate human review. The inherited first-pass exclusions comprised 6,571 records (6,181 title-only and 390 with abstract/excerpt reading); they were not full-text exclusions.

At the 1 October 2026 retrieval cutoff, the database record ledger contained 1,698 additional exclusions, 36 primary efficacy reports, 75 contextual reports, 65 linked supporting reports, three protocols without completed results, and 351 records whose eligibility remained unresolved. Of the latter, 349 reports had not been retrieved through the documented available routes, and two retrieved reports remained unclassifiable (Jones 2002, clinical-entry definition; Bumstead 1879, historical diagnostic ascertainment). Unavailable reports were not classified as ineligible. The 1,698 later exclusions comprise different reading stages and should not be described collectively as full-text exclusions.

After integration of the documented supplementary sources, 43 primary efficacy reports were mapped to 42 conservative family units. The two Shabahang reports were retained separately but assigned to one conservative unit because participant independence was unresolved; the count is not a claim of 42 confirmed independent randomized trials. Weck 2013 was linked to the existing Weck 2015/NCT01119469 family and did not add a primary end-of-treatment trial. Seventy-seven contextual reports comprised 66 main clinical context reports and 11 illustrative case or case-series reports in a separate appendix. Context reports were not equated with independent cohorts.

Registry searches returned 516 query memberships representing 380 unique registry records; 341 received initial exclusions and 39 underwent retained-record follow-up. Registry entries were not counted as additional trials merely because they had different identifiers. Forward tracing from three selected seed reports returned 274 occurrences and 242 unique index records: 59 matched the database master and 183 required separate consideration. Eleven retained leads were adjudicated; the remaining 172-record pass yielded 112 explicit exclusions, 57 awaiting-classification leads and three protocol/population-link leads. These 60 unresolved supplemental leads had no exact DOI/PMID match to the database master and remain separately listed; none was promoted to included evidence. Backward tracing identified 38 report leads, of which 31 matched the database master and seven were additional report leads. These search-source counts overlap at report and family level and must not be added as numbers of studies.

The illustrative case appendix is supplementary and is not an exhaustive systematic census of historical or single-patient reports. Some original first-pass design exclusions preceded the dated clarification permitting illustrative cases. Relevant-section reading, full-report acquisition and design-specific appraisal are recorded separately. Closing retrieval at a stated date does not mean that every potentially eligible report was obtained, that all citation leads were resolved, or that all context evidence received full appraisal.
