Replication package: Biodiesel in heavy-duty engines: why a meta-analysis could not be computed

Protocol, search strategy, screening register, retrieval census, extraction dataset, evidence cascade, risk-of-bias pilot, PRISMA 2020 assessment, synthesis code and figure code supporting a prespecified attempt to meta-analyse the effect of biodiesel and its blends, against a fossil-diesel reference tested in the same study, on the NOx and particulate emissions of heavy-duty engines, with blend level as the moderator. The synthesis was not performed, and why it could not be is the result. A Scopus search executed on 26 September 2026 returned 1,822 records, of which 138 were biodiesel candidates and 89 passed screening. The 64 reports whose abstracts declare both a diesel reference and a blend level were prioritised for retrieval and 61 were obtained by institutional subscription and read in full. Of those 61, 36 state a number of repetitions, 25 state some measure of dispersion, only 8 state it as usable numbers, and only 9 measure at engine-out or post-DOC. Applied cumulatively the requirements give 61, 36, 19, 7, 5 and 0: five studies reach the synthesis threshold of five declared in advance, and none of those five measures at engine level. The two sets are disjoint, so no outcome could be pooled - not because the experiments are poorly reported, but because statistical completeness and measurement position do not co-occur. Thirteen flat files, numbered in reading order because Zenodo does not preserve folders. Full texts of the primary studies are not redistributed, for copyright reasons; every retrieval attempt is recorded. See 00_README.md for the inventory, the reproduction steps and the conventions a re-user must respect - in particular that the deduplication key is the Scopus EID and never the DOI, that every proportion has a denominator of 61, and that measurement basis and measurement position decide what a ratio means