The relative mortality of Type 1 (insulin-dependent) diabetes in Denmark during the period 1933-1981 was studied using a modification of Cox's regression model on the basis of two patient populations, ascertained in different ways and independently of each other. Initial analysis showed that the two groups could be combined completely into one common analysis. Relative mortality was the same for both sexes. The additional variables studied were age at diagnosis, current age, calendar year at diagnosis and calendar time during follow-up. All these interrelated variables were accounted for in the analysis. The analysis showed that relative mortality (a) decreased with increasing age at diagnosis; (b) increased from 1933 to a maximum in about 1965, after which it decreased; (c) increased with increased duration of diabetes to a maximum at 15-25 years, after which it declined.