We can say that epistemology essentially concerns itself with two things: truth and our “relationship to truth”. Many philosophers, however, would say that truth simpliciter (rather than our knowledge of truth) is a metaphysical (... or a semantic... or a philosophy of language or mind) issue, not primarily one of our knowledge and therefore epistemology.
In terms of truth being a metaphysical issue.
This is the case because various philosophers see truth as either some kind of property or even some kind of entity. Not only that: a property that is mind-independent (even if minds can come to know the truth). Strictly speaking, then, truth is a question of ontology, which is a branch of metaphysics.
As for truth being deemed an issue for the philosophy of language.
We can say that this is the case because truth is seen as a property of sentences or statements. Yet even here we smuggle in metaphysics by talking of the properties of statements. What kind of property is that? How does the property truth attach itself to sentences? What is the relation between the statement and its truth-property?
The philosopher of language can circumvent metaphysics by saying that truth is literally a linguistic property, not a metaphysical one. That is, saying that truth is linguistic (or semantic) is a reference to the linguistic fact that we can add the predicate ‘is true’ to the statement “Dolphins are mammals”. Truth is a semantic endorsement of a prior truth-claim or statement. Thus truth- deflationists, for example, argue that we can easily drop the “is true” from the statement without any loss of power or meaning. In that case, the semanticist can simply get rid of truth altogether (as is the case with semantic naturalisers).
Earlier I made the twofold split between truth and our knowledge of truth. We can say, however, that truth must come before our knowledge of truth. Or the metaphysics or semantics of truth must come before epistemology simply because of the fact that we can’t have knowledge of a truth if we don’t already know what truth is. However, we can epistemologise (as it were) truth by arguing that truth is somehow a function of our knowledge of truth (or of the tools and procedures required to discover truth).
A similar manoeuvre can be found in semantics.
Many logical positivists, for example, originally argued that the meaning of a sentence is the means we use to verify it. And verification is an epistemological issue in that we require epistemic tools and procedures in order to verify a statement or proposition. So just as meaning was seen as a function of epistemic verification, we can now argue that the truth of a statement is a question - or function - of the epistemic tools and procedures we require to either construct truth or discover it. And if all we have are such epistemic tools and procedures (i.e., not a property we can call ‘truth’), then we can become truth-deflationists (or truth-naturalisers) by simply getting rid of truth altogether (or, at the least, getting rid of the predicate ‘is true’). Thus we can conclude that naturalised epistemology basically gets rid of the metaphysics of truth.