From principles of survival analysis statistical approaches are identified for a major problem with occupational mortality studies when few deaths, as with cause specific mortality, and no exposure data on the individuals are available. Specifically, this limitation is the general inadequacy of a geographic population as a comparison group for employed populations. Qualitative aspects of the so-called healthy worker effect are briefly presented, while the essential features of statistical strategies for part of this issue are sketched in detail. Although the occupation epidemiology literature focuses on selection bias as the source of the problems with the standardized mortality ratio, the potential inadequacy of the proportional hazards model for occupational mortality evaluations is emphasized here.