Double-peaked time distribution of mortality for breast cancer patients undergoing mastectomy

Purpose. To gather information on the natural history of breast cancer from the time-distribution of deaths of patients undergoing mastectomy alone. Patients and methods. A total of 1173 patients, who entered controlled clinical trials carried out at the Milan Cancer Institute and underwent radical or modified radical mastectomy without any adjuvant therapy for operable breast cancer, were examined. The risk of death at a given time after surgery was studied utilizing the death-specific hazard rate. The risk distribution was assessed relative to tumor size, axillary lymph node involvement, and menopausal status. Results. The hazard rate for death presented an early peak at about the 3rd-4th year after surgery and a second late peak near the 8th year. The double-peaked pattern was almost completely generated by N+ patients, while N- patients did not show relevant structures. Pre-menopausal patients showed an initial mortality wave covering about 6 years, with maximum height at the 4th year, followed by a peak 8 years after surgery, while post-menopausal patients showed an early high mortality surge peaking at the 3rd year, followed by a modest increase at the 8th year. Detailed analysis revealed that post-menopausal patients with early mortality had significantly larger tumors and higher nodal involvement, while no special trait characterized the corresponding pre-menopausal patients. Moreover, patients of the late mortality peak were more likely to have suffered early local-regional or contra-lateral recurrence or to be pre-menopausal patients recurring anywhere at the second recurrence peak. Conclusion. The double-peaked hazard curve confirmed the occurrence of discontinuous features in the natural history of breast cancer for patients undergoing mastectomy. Indeed, the mortality pattern maintained definite signs of the previous double-peaked structure of recurrences. However, death events did not parallel the corresponding recurrence events and, moreover, pre and post-menopausal patients revealed dissimilar survival after recurrence, at least for early deaths. These findings, showing disconnection of mortality pattern from recurrence pattern for subsets of patients, suggest that parameters other than those influencing the recurrence risk may determine the survival of recurred patients.