When does a statistical fact become an ethical imperative?