This white paper by Pete Gough discusses out of specification and out of trend results and how they should be investigated.