• © 2009 by American Society of Clinical Oncology

Reply to D.-Y. Chang et al

We thank Chang et al¹ for their comments regarding our study.² As mentioned in our article and their letter, the shortcomings of this study and others concern the retrospective nature of the study and, particularly, the potential selection bias of patients who were offered locoregional treatment. We do agree that despite adjusted analyses that attempt to account for selection biases driving the use of surgery or exclusive locoregional radiotherapy, the improvement in survival seen in the group of patients who were offered treatment of the primary tumor cannot be definitively attributed to locoregional therapy (LRT) because in most published studies, the women in the surgery or LRT groups were younger, had smaller tumors and fewer metastatic sites, and were more likely to have bone/soft tissue metastases rather than visceral disease. Because brain metastasis is a poor prognostic factor, we explored the association between LRT and the risk of death, after excluding the 18 patients with brain metastasis. As mentioned in the online appendix, a similar association between LRT and a lower risk of death was observed, with a median survival time of 32 months and a 3-year overall survival rate of 43.4% among the 318 women who received LRT, compared with 21 months and 27.4% among the 245 women who did not receive LRT (P ≤ .0001). The variable brain metastasis was not incorporated in the multivariate analysis as patients with brain metastasis were excluded from the Cox model. Regarding our multivariate analysis, we tested the variable visceral metastases (yes/no) that was a prognostic factor in univariate analysis in the whole population. We do not consider that it was a fundamental mistake as patients without visceral metastases were those with bone metastasis alone or associated with extraregional lymph nodes or skin or subcutaneous nodes. Results of the Cox model were similar with regard to the variable bone metastasis only. We do agree that performance status, comorbidity, and response to systemic agents could affect the outcome of patients and influence treatment strategy. It is possible that patients selected for treatment of the primary tumor were considered to have longer life expectancy, and that those who received locoregional treatment had better prognostic features or a better response to a first-line systemic therapy than those who received systemic therapy alone. These variables were not prospectively recorded in our database and therefore were not reported. However, the effect of LRT was similar after we excluded patients who survived less than 6 months. Moreover, this is to our knowledge the first study to explore an association between LRT and the risk of late death (≥ 1 year after diagnosis), in order to limit the treatment assignment bias.

Well-designed prospective studies are now needed to formally test the hypothesis generated by these retrospective reviews and to identify metastatic breast cancer patients who are most likely to benefit from treatment of the primary tumor.