Cytoplasmic nucleophosmin immunohistochemistry against NPM1 mutation status in acute myeloid leukaemia: diagnostic accuracy, boundary conditions and sources of discordance (PRISMA-DTA)

Systematic review and meta-analysis of diagnostic test accuracy (PRISMA-DTA) of NPM1 localisation immunohistochemistry (cytoplasmic versus nucleus-restricted nucleophosmin) on bone marrow trephine, clot or cell-pellet sections against molecular detection of NPM1 mutation in the same patients with acute myeloid leukaemia, framed around the boundary conditions under which the surrogate holds (specimen, fixation, decalcification, section thickness, positivity threshold, patient spectrum, verification completeness and the breadth of the molecular reference). Six databases (PubMed, Embase, Web of Science, Scopus, CENTRAL, Europe PMC), a historical-terminology arm mirrored in four databases, forward citation searching in Scopus, Web of Science and Google Scholar, backward citation searching of eight seed reports and of five reviews, and PubMed related articles, all completed 2–7 September 2026 without limits, with one supplementary full-text search (Scholar Gateway, Wiley) reported as such. Two reviewers screened independently, the first arm by arm during retrieval and the second on the consolidated record set afterwards, blinded (agreement 95.2%, kappa 0.40); discordant records were resolved between the reviewers on the recorded evidence. Every 2×2 cell was extracted verbatim; no cell was inferred. The prespecified primary model is a bivariate binomial-normal model with a univariate fallback and a DerSimonian–Laird cross-check; smear, skin and extramedullary series are strata, never pooled. This registration is made after the searches, screening, extraction and primary analysis were completed; the attached protocol states the written plan of 2 September 2026, the state of the evidence at registration (seven independent cohorts, 858 paired observations) and every deviation from the plan, each with its date and its place in the deposited working record.