In the present paper, a general multivariate approach, estimating cancer latency using data from a case-control study with time-related factors, is proposed based on the theory for risk analysis of states. Making use of a forward-analysis and backward-synthesis strategy and excess exposure fraction, this approach can simultaneously accomplish multivariate logistic regression analysis and estimation of susceptible exposure ages and latency distribution. The latency is resolved into a accumulative period and preneoplastic period. It gives a theoretical interpretation for primary and secondary prevention of cancer.