Reliability of personal interview data in a hospital-based case-control study.