Abstract
Background
Despite years of research, it is still unclear which women with node‐negative (N–) breast cancer will need adjuvant chemotherapy and which women are being treated unnecessarily. Our goal was to determine which factors best predicted disease free survival (DFS) or cancer‐specific overall survival (OS) and, therefore, select the correct patients for treatment. A total of 11 parameters were measured: estrogen receptor (ER), progesterone receptor (PR), age, race, ploidy status, %G0/G1 (% non‐DNA synthesis), %S (% S‐phase), cathepsin D status, size, stage, and histologic grade.
Results
In this prospective study, we followed 556 N– patients diagnosed between 1991 and 1996. The tumors were 56% ER+, 51% PR+, 30% diploid, with a mean %S of 8.9%. The level of cathepsin D ranged from 0.50 to 155 pmol/mg of protein with a mean of 42.9 pmol/mg of protein. There were 87 recurrences (16%) and 72 cancer deaths (13%), with a median follow‐up of 7.8 years. Ploidy status (p = 0.01), S‐phase activity (p = 0.003), G1 phase activity (p = 0.02) and age (p = 0.01) were able to significantly predict DFS in a univariate manner. All of the measurable factors were significant or borderline significant in predicting OS in a univariate manner except for age, race, and ER status. In multivariate analysis with S‐phase included, it was the only remaining factor in DFS and OS; with S‐phase excluded, age and ploidy status remained as factors for DFS in stepwise regression, while PR, size, and cathepsin D were the remaining factors that predicted cancer‐specific OS. The effect of adjuvant treatment on prognosis was also analyzed.
Conclusions
Both biochemical and clinical parameters have the potential to predict prognosis for N– breast cancer. In this large prospective clinical trial, with a median follow‐up of 7.8 years, no individual marker adequately predicted the prognosis for an individual patient. %S activity was the best independent marker, but only 77% of the tumors provided this value. Subset analysis provided improved prognostication, but there were limits to its utility. These data represents a definitive study starting in 1991 and ending in 2002. © 2004 Wiley‐Liss, Inc.