Correspondence between symptoms and preference-based health status measures in the STOP study.