The time course and dose dependency of the incidence of bone-sarcomas among 900 German patients treated with high doses of radium-224 is analysed in terms of a proportional hazards model with a log-normal dependency of time to tumor and a linear-quadratic dose relation. The deduced dose dependency agrees well with a previous analysis in terms of a non-parametric proportional hazards model, and confirms the temporal distribution which has been used in the Radioepidemiological Tables of NIH. However, the linear-quadratic dose-response model gives a risk estimate for low doses which is somewhat less than half that obtained under the assumption of linearity.

Dedicated to Prof. W. Jacobi on the occasion of his 60th birthday

Work performed under Euratom contracts BI6-D-083-D, BI6-F-111-D, U.S. Department of Energy contract DE-AC 02-76 EV-00119, the U.S. National Cancer Institute