Pharmacogenomics is often implemented as though certainty accumulates monotonically from variant detection to prescribing action. In practice, these levels can separate. A genotype may be analytically secure while its functional consequence remains substrate-dependent, altered by copy-number structure or comedication, or incompletely transferable across populations. A gene–drug association may be pharmacologically compelling while the magnitude of clinical benefit remains dependent on comparator therapy, endpoint, treatment phase, and the availability of alternatives. Guideline organizations may additionally translate the same underlying evidence into different actions. This Current Opinion argues that pharmacogenomic implementation should therefore not be triggered by molecular certainty alone. Instead, implementation decisions should distinguish analytical certainty, functional certainty, gene–drug association certainty, clinical effect-size uncertainty, and therapeutic context. High-certainty examples such as selected DPYD–fluoropyrimidine and TPMT/NUDT15–thiopurine relationships demonstrate that clinically actionable pharmacogenomics is achievable, but they should not be treated as templates for every candidate gene–drug pair. The article develops a proposed molecular-to-clinical certainty architecture in which downstream uncertainty remains visible rather than being hidden behind a binary “actionable” label. This approach reframes implementation as a decision problem requiring convergence between reliable molecular information and a sufficiently clear clinical consequence.
IJPRAS
