Time distribution of the recurrence risk for breast cancer patients undergoing mastectomy: further support about the concept of tumor dormancy.

PURPOSE: To gather information on metastatic growth from the time-distribution of first treatment failure in breast cancer patients undergoing mastectomy alone.
METHODS: The risk of recurrence at a given time after surgery was studied utilizing the cause-specific hazard function. Recurrence was categorized as first treatment failure at any site, local-regional recurrence, distant metastases, and contralateral tumor. The risk distribution was assessed relative to tumor size, axillary lymph node involvement, and menopausal status.
RESULTS: A total of 1173 patients treated between 1964 and 1980 with mastectomy alone and no adjuvant therapy were studied. The hazard function for first failure presented an early peak at about 18 months after surgery, a second peak at about 60 months and then a tapered plateau-like tail extending up to 15 years. A similar risk pattern was detectable for both local recurrence and distant metastases, while the curve of contralateral breast tumors showed a near flat plateau. The risk of early local-regional and distant recurrences was much lower for tumors less than 2 cm in diameter than for larger tumors; the risk of late recurrence was similar for small and large primaries. Node-positive patients showed peaks four to five times higher than node-negative patients. Sub-dividing node-positive patients into 1-3 and > 3 node-positive subsets did not substantially change the general picture of tumor recurrence. The hazard functions for premenopausal and postmenopausal patients were virtually superimposable.
CONCLUSIONS: The multipeak hazard curve suggests that the process resulting in overt clinical metastases may have discrete features. Primary tumor size could affect in different ways early and late metastases, while axillary node status should be related to the risk level, not to the risk pattern, and menopausal status does not seem to significantly affect the hazard distribution. Moreover, contralateral breast tumors, occurring at constant risk throughout the time, should be considered as second primary cancers. These findings could be reasonably explained by a tumor dormancy hypothesis, which assumes that micrometastases may be in different biological steady states, most of which do not imply tumor growth. Tumor or microenvironment changes could induce metastatic growth after given mean transition times from surgery and originate a discrete pattern of the hazard function.