Though, Gödel's completeness theorem I suppose does relate provability to truth in that it shows that (for systems with the right kind of rules of inference) provability is equivalent to something being true in all models...
Still, are you sure Tarski's undefinability theorem isn't more relevant to your point?