Which carcinogenicity study should I use? Automated identification of reliable studies

Deriving compound-specific acceptable intakes for Class 1 mutagenic impurities under ICH M7 depends on selecting the most reliable carcinogenicity study. In practice, this selection is frequently subjective, inconsistent and time-consuming. We describe an automated workflow, implemented within LCDB Plus, Lhasa Limited's carcinogenicity decision-support platform, to identify the most robust study for a given compound/species/sex combination. The Lhasa reliability grading scheme has been extended to include an A+ category, distinguishing studies that are fully adequate for dose-response modelling. A Pareto optimisation algorithm has been applied to identify the most robust study of those available. The Lhasa TD₅₀ model incorporates the Akaike Information Criterion and Pearson's χ² goodness-of-fit test, explicitly flagging borderline dose-response trends. Applied to 18,167 studies in LCDB version 2025.1, the algorithm identifies a single most robust study in 97.6% of compound/species/sex cases. Selections were validated against expert-derived study choices from the ICH M7 addendum and published regulatory acceptable intakes. This workflow reduces study selection burden - supporting the transition to reproducible, non-animal decision frameworks consistent with the 3Rs and recent regulatory strategies for replacing animals in science - while delivering recommendations aligned with experts, enabling reproducible acceptable intake derivation and supporting submissions for ICH M7 Class 1 impurities.