Estimating the expectation of life in cancer survival studies with incomplete follow-up information.