Differential influence of prognostic factors on the occurrence of metastases at various anatomical sites in human breast cancer

Publisher Website
Google Scholar
Abstract
The present study was initiated by data on the anatomical occurrence of metastases from breast cancer. Metastases may occur at various organs, here grouped into ten categories, and simultaneous occurrences at several sites are common. Our aim was to identify and compare risk factors for development of metastases at each site.The influence of the various risk factors for recurrence at one specific site can be modelled as a standard competing risk problem, using well‐established survival analysis techniques such as the Cox regression model. Recently a generalization to joint occurrences at more than one site has been proposed. The more general models allowcomparisonof the differential prognostic influence of various risk factors on recurrence at different sites.We applied these methods to data from the Danish Breast Cancer Cooperative Group on recurrence after breast cancer, and examined the effect of degree of anaplasia, number of positive lymph nodes, size of primary tumour, skin or deep fascial invasion, age of the patient and adjuvant treatment regimen, as possible risk factors.There were no differential effects of type of therapy on recurrence in different sites. However, chemotherapy had a marginally significant positive effect at all sites. Moreover the number of positive lymph nodes was associated with an increased risk of metastases, homogenous across sites, except for brain where an opposing trend was found. Similarly the degree of anaplasia was associated with a somewhat increased risk of metastases, homogenous except for brain. For brain metastases a significantly more pronounced effect of degree of anaplasia was found.Comparison of the results obtained from the two types of models reveals that the new more general models confirm fewer differences between the influence of the prognostic factors than the traditional regression models suggest.