Comparator agreement and surveillance evidence gaps of CRISPR-Cas and related programmable nuclease assays for African swine fever: a systematic review and sparse-data meta-analysis

Background High agreement between African swine fever virus (ASFV) programmable nuclease assays and molecular comparators does not establish surveillance accuracy when the validity and independence of reference methods are questionable. Methods PubMed, Web of Science Core Collection, as well as CNKI journals and dissertations were searched through 1 July 2026. We reviewed paired classification results, audited recruitment and index/reference targets, and evaluated using QUADAS-2 and adapted diagnostic GRADE. Marginal beta-binomial models served as the primary approach for sparse-event analysis; the Reitsma model was adopted as a conventional secondary reference method. Results Twenty-nine studies yielded 29 non-overlapping comparison arms comprising 1,262 diagnostic-material observations; these observations were not necessarily derived from independent animals. The median arm sample size was 30. All arms included at least one zero cell, and 25 of these arms reported no discordant test results. Reference target sequences could be ascertained for eight arms. Among these eight arms, six shared the reported target genes with the index assay, whereas the remaining two used different target genes. The remaining 21 target relationships remained unresolved, and all 29 reference-standard judgments were unclear. Studies relying on selected or archived specimen panels and experimentally infected materials, alongside those with incomplete recruitment documentation, did not provide verified prospective consecutive surveillance cohorts. Comparator-relative agreement could not identify the accuracy of independently determined infection status or the magnitude of reference bias. Due to risk of bias, indirectness, imprecision, and suspected publication bias, the certainty of their representative surveillance applications was very low. Conclusion Published validation panels show high comparator agreement, while true surveillance accuracy and incremental diagnostic benefit remain unestablished. Prospective multicenter validation with independent reference adjudication and representative recruitment is required before adoption as an alternative to existing methods. Systematic review registration https://www.crd.york.ac.uk/PROSPERO/view/CRD420261413928 , Identifier: CRD420261413928.