Analyzing prognostic factors in breast cancer using a multistate model

In breast cancer clinical research, an important goal is to analyze how factors are seen to affect the disease process. Meanwhile, the disease progression is not fully modelled using standard analysis since transitions between intermediate events such as local-regional recurrences (LRR) or metachronous contralateral breast cancer (MCBC) are not considered. In the present study, the progression of disease was modelled using a multistate model. By this approach, we assessed transitions during the course of the disease and studied prognostic factors for each transition. The model was applied to 6,185 patients with unilateral ductal invasive breast cancer, clinical stage I through III, treated between 1981 and 1988 at the Curie Institute. At first diagnosis, high clinical stage, high histological grade, positive lymph nodes, and age less than 40 years were associated with increased risks of LRR, metastases, or death. Except age, the same factors remained predictive for metastases or death following LRR. Chemotherapy for the first cancer was associated with a decreased risk for developing MCBC. As the time interval from diagnosis of the primary tumor to that of a local or contralateral recurrence increased, the risk of metastases or death decreased. Nodal status for the first tumor and clinical stage for the contralateral tumor increased the risk of metastases or death following MCBC. Conversely, the risk decreased for patients who received adjuvant hormone therapy following MCBC. In conclusion, the multistate model offers us a much more appropriate way to study prognostic factors for each transition in breast cancer disease.