A New Method for the Analysis of Cohort Studies : Implications of the Multistage Theory of Carcinogenesis Applied to Occupational Arsenic Exposure