Under certain circumstances a proportionate mortality ratio (PMR) equals the cause-specific standardized mortality ratio (SMR) divided by the overall SMR. In this paper, the relationship between PMR and SMR is derived stepwise with focus on the conditions that must be fulfilled for each step of the derivation to be valid. It appears essential that the following conditions be fulfilled: (1) the compared populations are closed throughout the period of study; (2) all deaths (or a representative sample) in the compared populations are included in the analysis; (3) every dead person is assigned to exactly the same exposure category that he or she would have been assigned to in a follow-up study; and (4) possible diagnostic misclassification is not exposure-dependent. Furthermore, for dead people within the comparison group, the relative frequency of the cause of death of interest must be constant over age. In view of this it is recommended that PMR studies of occupational mortality are designed to cover only a short period; to use data from national or regional death certificate registries only; to be restricted to lifetime occupations; to make use of a hand-picked comparison group that has had the same access to medical care and diagnosis as the index group; and to restrict the PMR analysis to age-at-death strata in which the relative frequency of the studied cause of death is constant in the comparison group of deceased persons.