Pharmacovigilance often treats a statistically prominent drug–event association as if it were inherently closer to clinical or regulatory action. That inference is unsafe. Disproportionality, machine-learning prioritisation, longitudinal healthcare data, case-level assessment, and other signal methods answer questions about whether an association is unusual, reproducible, temporally coherent, or evidentially credible. They do not by themselves determine whether an intervention is warranted. Actionability additionally depends on the seriousness, magnitude, reversibility and preventability of harm; background risk; exposed-population vulnerability; therapeutic alternatives; time sensitivity; and the feasibility and consequences of intervention. This Current Opinion argues for an explicit separation between signal strength and clinical actionability. Statistical strength should describe the credibility and maturity of the safety association, whereas actionability should describe the consequence-sensitive case for doing something under current uncertainty. The distinction explains two recurrent pharmacovigilance problems: strong signals that remain insufficiently decision-ready and modest signals for which proportionate early action may nevertheless be justified. The proposed separation is a conceptual framework, not a validated scoring system, causal algorithm, or regulatory rule.
IJPRAS
