Design issues in cohort studies