Require genetic and target-trial evidence before funding any repurposing phase 3

MA.32 (metformin in breast cancer) was negative despite extensive observational support that was later attributed to immortal-time and confounding biases. Mendelian randomisation using drug-target genetic proxies, target trial emulation with active-comparator new-user designs in large health records, and pre-registered replication across independent databases can test causal plausibility for a fraction of a trial's cost. The proposal is a funding rule: repurposing phase 3 trials receive public funding only after a standardised evidence gate (genetic support where a proxy exists, at least two concordant target trial emulations, and a plausible dose-response) has been passed and published.