A parametric model is used to investigate the latency time of leukaemia observed in patients treated for Hodgkin's disease. In specifying the treatment effect on leukaemia-free survival, account was taken of a fraction of long-term survivors and of time-changing risk associated with the relapse of the primary disease. The model is applied to data collected in the International Database on Hodgkin's Disease. It permits estimation of the contributions of primary and of relapse treatment to the overall risk of induced leukaemia. Baseline hazards appear to be identical after primary and relapse treatments supporting the concept that induced leukaemia have common origin. The probability to induce leukaemia by MOPP chemotherapy is the same, regardless whether used as primary or relapse treatment.