Competing risks determining event-free survival in early breast cancer.

To evaluate the natural history of a disease and the effects of therapeutic interventions, it is important to determine which are the causes of treatment failure and to assess the extent to which each cause contributes to the total failure rate. The paper presents a new biostatistical technique to decompose the total event-free survival of a diseased population into cause-specific failure rates. The technique was based on a competing risk approach thereby avoiding biases related to assumptions of independence between different types of event. Such assumptions are inherent in the conventional Kaplan-Meier or actuarial methods. The competing risk method was used to analyse the pattern of failure among 2,850 pre- and postmenopausal patients with early-stage breast cancer and the results were compared to those obtained using conventional methods. The following events were analysed: loco-regional recurrence, distant metastasis, contralateral breast cancer, other new primary malignancies, and intercurrent deaths. The rate of new primary malignancies was found to be significantly higher in post- than in premenopausal patients (6% vs 3% at 10 years). In low-risk, node-negative postmenopausal patients the incidence of recurrences from breast cancer were found to be no greater than other types of events. This observation highlights the significance of the effect of different adjuvant therapies not only on the disease itself but also on the risk of second primary malignancies and other intercurrent diseases. In general, it was found that the conventional statistical methods tended to overestimate the event-specific rates. In conclusion, the method based on competing risks permits an unbiased analysis of all types of events determining the total event-free survival. It is thus useful for the description of the natural history of breast cancer as well as other diseases.

RCT Entities:   Population Interventions Outcomes