The completeness of cancer registration in follow-up studies--a cautionary note.