This reabstracting study was initiated to estimate accuracy and reliability of quality control for the Daegu Cancer Registry data in 1997-1998. The study population was 7,837 incident cases registered as malignant primary tumor in the Daegu Cancer Registry and two samples were selected for agreement study of the tumor and demographic information and for usual residence survey, separately. The 10 items reabstracted from medical records of 696 cases were compared to original data and classified as major and minor disagreement by the Centralized Cancer Patient Data System. The usual residence of 143 cases of one district was traced back actually through the computerized personal identification system of the government. The highest rate of major disagreements, 12%, was for morphology, followed by 12.0% for date of diagnosis and 5.6% for site. Demographic items had good reproducibility (exact agreement rates > 95%). Usual residence of 13 cases (9.1%) was out of Daegu. These results suggest that periodic assessment of the validity is necessary for quality control of the Daegu cancer registration data by the standardization of the cancer registration method including data collection method.