Data quality assurance

Data quality was assured by adequate design, and the questionnaire was pre-tested on 10% of the entire sample (n = 50) that were not included in the survey. The baseline survey was administered after amendment of ambiguities identified in the questionnaires. The data collectors and supervisors were given 2 days of intensive training to familiarize themselves with the tools and methods of data and blood specimen collection. The collected data were checked carefully for completeness, accuracy, and clarity every day by a supervisor, and the principal investigator monitored the overall activities of data collection. One interviewer and one medical laboratory technologist together with the principal investigator were involved in data collection at each center. We examined the reliability and internal consistency of the dietary and physical activity questionnaire data based on Cronbach’s α coefficient; the Cronbach’s α were 0.82 and 0.85 for the dietary and physical activity instruments, suggesting that the level of internal consistency was high. With regard to reproducibility, the two sets of responses from the patients in the test–retest group were inspected using the intraclass correlation coefficient. For example, the intraclass correlation coefficient for the physical activity questionnaire was 0.90, indicating satisfactory test–retest reliability.