Since non-comparability of screened and non-screened individuals is the main threat to the validity, randomized trial is the optimal assessment for screening programs. Randomization controls potential confounders when the sample size is sufficiently large. Self-selection bias is also controlled as individuals are allocated to either group at random after they have agreed to participate in the trial. Lead time bias can be taken in account by adjusting for the average lead time when comparing screening versus symptom- detected individuals or preferably by comparing age-specific mortality rates for both groups. Trials can also control the length bias by comparing the mortality experience of the groups after repeated screening. While randomized trials can provide the best and most valid evidence concerning the efficacy of a screening program, their use is limited because of the problems of costs, ethics and feasibility and most evidence on the effects of screening programs come from nonexperimental study designs.