Outcome

Every individual's identification number was matched with the National Death File for CRC mortality and the National Cancer Registry file for CRC incidence, respectively.30^,31 Instead of immediate colonoscopy, a comprehensive cancer registry system would be more appropriate to develop a prediction model for expected CRC incidence in 5-10 years. Furthermore, no colonoscopy was indicated for negative FIT, and cancer registry would be the standard for CRC incidence. Finally, the National Death File-based mortality would be the ultimate criteria in evaluating CRC mortality. The time interval from FIT to the development of CRC incidence or mortality would be measured once CRC occurred.

A step-wise Cox proportional hazards regression analysis was carried out to identify risk predictors that were significantly associated with increased risk of CRC in multivariable models. We included physical inactivity and low intake of fruit and vegetables as risk factors of CRC since its risk approximated statistical significance. Hazard ratios (HRs) and 95% confidence interval (CI) were estimated for each variable. NNS (number needed to scope to find one cancer during study period) used in this study was approximately a reciprocal of age-adjusted incidence rate/1000 person-years.