The TNM stage proved to be the most significant independent prognostic factor for determining survival. In our study series, around 75% of all patients were diagnosed in stages I and II. Regular mammography combined with regular clinical breast examination may offer the best opportunity to increase the percentage of early stage cases detected. The frequency distribution of stage and the 5-year relative survival rates by stage were similar to the data reported by the American Cancer Society (American Cancer Society, 2001
). The observed decline in overall 5-year survival by stage was also seen within each histology or grade group. In all recent multiple regression analyses, stage had a major prognostic role. This is probably due to the fact that stage reflects the interaction between host and tumour. From this point of view, some variables connected with the host such as age, or connected with the tumour, such as histology and grade, could stand out as prognostic factors.
Histology was found to be prognostic in breast cancer. Lobular carcinoma had a better prognosis than ductal carcinoma for separate or combined stages in our study, while Mclaughlin et al (1995)
had a similar result for combined stages. Although they could not account for the stage distribution differences within histology groups, these differences of survival can be explained as follows: grade 1 lesion was more common in women who had lobular carcinoma than in those who had ductal carcinoma (51 vs
20% among the graded cases); and conversely, grade 3 lesion was identified less frequently in lobular carcinoma (12 vs
Histologic grade was an important determinant of prognosis that also allowed risk stratification within a given tumour stage. The proportion of high-grade cases increased with advancing stage, whereas the percentage of low-grade cases decreased (). This distribution would suggest that tumour with advanced stage would be more likely to present at a higher grade. The survival rates decreased with advancing grade within each stage. This survival trend confirmed results reported by Carriaga and Henson (1995)
. For most breast cancers, histologic grade was a more meaningful prognostic feature than histologic classification, and helped to explain the favourable prognosis of most special histologic types. About 10% of low-grade tumours recurred within 5 years, compared with about 30% of high-grade cancers (International Union Against Cancer, 2001
We found that overall survival was poorer for those younger than 40 years, and this difference was independent of tumour stage and type of treatment. This finding was consistent with other studies; for example Albain et al (1994)
showed significantly worse prognosis for younger patients even after other prognostic factors were considered by multivariate analysis. The association of young age at diagnosis with a worse prognosis in our series can be explained by a higher proportion of poorly differentiated cancers (50%) as compared with other age groups (aged 40–49, 40; 50–69, 29; and 70+ years, 24%), suggesting an aggressive breast cancer phenotype. There are various possible explanations for the poor survival experience of young women with breast cancer: (1) lack of competing causes of death (Ederer et al, 1963
); (2) higher frequency of undifferentiated tumours, more poorly differentiated cancer, microscopic lymph node involvement and negative hormonal receptor status (Bonnier et al, 1995
) or (3) more cases diagnosed with stage II or III cancer (Gajdos et al, 2000
). (In our series, 55% of tumours for women aged 0–39 years were diagnosed in stage II). Each of these hypotheses requires thorough investigation.
Our finding that patients aged 50–69 years had the best outcome can probably be explained by the relatively good local control (through mammography screening) for this age group in Canada, even though there is no significant difference in survival among age groups 40–49, 50–59 and 60–69 years (whose 5-year RSRs were 84, 88 and 88%, respectively). Some studies (Shapiro et al, 1982
; Miller et al, 1992
) have indicated that mammography screening has reduced breast cancer mortality and improved survival by detecting cancers at earlier, more treatable stages. In our study, nearly 96% of women with stage I cancers survived at least 5 years; this stage accounted for 50% of the women aged 50–69 years. In keeping with our results, Golledge et al (2000)
described that patients aged 60–69 years experienced the best survival, although this favourable outcome was principally restricted to axillary lymph node negative patients.
Determining the effect of patient age on breast cancer prognosis is confounded by many factors, such as screening rates, menopausal status and differences in treatment. Consequently, significant differences in study designs have resulted in a lack of consensus regarding the prognostic effect of patient age. La-Rosa et al (1996)
observed that women less than 35 years had a better prognosis at 5 years from diagnosis. Holli and Isola (1997)
found that the 5-year RSR was highest in women aged 46–50 years, whereas they found no significant difference between younger and older age groups. A different finding for 5-year age-specific survival by stage was reported from the Rhode Island Tumor Registry, where women aged 40 years or less had a worse prognosis than other age groups, except for those with stage I disease (Chung et al, 1996
). A limited number of young patients included in some studies, differences in patient selection, age grouping and analysis of outcome may contribute to the conflicting results for the relationship between age and prognosis.
In our study series, proportionally more stage I patients (39%) experienced surgery with radiotherapy, whereas more patients with stages II, III and IV diseases underwent surgery with radiotherapy and chemotherapy (57, 65 and 32%, respectively). Selection of therapy depends not only on the stage of the disease, but also on age, menopausal status, grade, histology, estrogen-receptor (ER) and progesterone-receptor (PR) status, HER2/neu gene amplification and general health. These factors may have resulted in selection bias in this study. Preventive mastectomy is an option to prevent breast cancer for women who are at very high risk for breast cancer. Possible candidates for this procedure are women with a strong family history of breast cancer and those who have a mutation in genes p53, BRCA1, or have gene BRCA 2. Chemotherapy can be considered for patients with hormone receptor-negative disease or advanced cancer. Hormonal therapy with or without chemotherapy is usually assigned for receptor-positive cancers.
Surgery plus radiotherapy plus chemotherapy would in general be given to higher risk patients (with four or more positive axillary nodes, large primary tumours, ER or PR negative, grade 2–3, 35 years of age and younger, etc.) than would surgery plus radiotherapy alone; this was confirmed by our study in terms of grade and age (the other variables were not included in this study) within each stage (
). In addition, radiotherapy might have been given post mastectomy to patients with higher risk profiles within a given stage, or with breast-conserving surgery to lower risk patients. Treatment protocols were significantly linked to the age at diagnosis for stage I, II and III (P
0.01, ): the percentage of patients who received the three combined treatments mainly decreased as age increased. Treatment differences between younger and older patients may be partially attributed to the poor physical condition and the comorbid diseases of the older patients. Histology was not significantly related to selecting treatment for stage I, II and III ().
Distribution of treatment by age at diagnosis, grade, histologic type and stage, ORCC, 1994–1997
Hormonal positive status predicts response to hormonal therapy. Many studies have shown that women with ER- or PR-positive cancers have a better prognosis than patients whose cancers do not have these receptors. An interaction of treatment with receptor status may partially explain the inferior results for chemotherapy-treated patients. However, since hormonal receptor status was missing for 93% of the patients in this series, the impact of hormonal receptor status on outcome could not be accurately assessed.
There are some other limitations to the study. Firstly, in the assessments of multiplicative effects of patient characteristics (i.e. interactions), the study did not have adequate power to produce conclusive evidence for the study subgroups with small sample size or inadequate end points (numbers of deaths). However, the results were suggestive and can provide direction for future research. For example, in the calculation of RSRs by age group for patients with stage I cancer, the RSR estimates were not reliable because of either the low frequency of death in this group or the small number of patients in the youngest age subgroup, so they need to be re-evaluated in a large series. To compensate for such a deficit in expected sample size and end points, two major strategies were used in this study: first, collapsing two adjacent groups into one group and, second, reducing the number of intervals during the analysis. Secondly, since information on cause of death was missing for 86% of the individuals in this series, we used a relative survival model, but the results could not be confirmed by Cox proportional regression. However, the estimate of relative survival is closer in theory to net survival, and empirical evidence shows that relative survival approximates net survival more closely than other methods (Esteve et al, 1990
An important strength of this study is the ability to investigate more comprehensively the impact of demographic, histologic and therapeutic factors on survival with breast cancer. The size of the study is sufficiently large to examine effect modifications and perform survival analysis across different subgroups of breast cancer. Perhaps more importantly, the results will provide further evidence for the debates on age influences on survival and the importance of grade on prognosis. In addition, one-quarter of patients diagnosed with overt metastases were alive at 5 years. The data thus provide fairly clear circumstantial evidence that the society, which puts resources and energy into breast cancer prevention and control through supporting integrated policy development, surveillance, research, education, diagnosis and treatment, has seen improved results. The development of new chemotherapeutic agents, the use of new radiation techniques and the implementation of multimodality therapy in advanced disease (81% of stage IV patients received chemotherapy combined with radiation and/or surgery in this study) have been observed to improve survival in patients with more advanced stages of disease.
In conclusion, our study has found that age and TNM stage at diagnosis, histologic grade and treatment were independent significant prognostic factors for breast cancer, whereas histologic subtype was statistically significant in the univariate analysis but not after adjusting simultaneously for other prognostic factors. Even more information was obtained when prognostic factors were examined in combination. Within stage, significantly wide variation in survival was seen due to age or treatment. The fact that lobular carcinoma had a better prognosis than ductal carcinoma can be explained by more grade 1 and less grade 3 cases in lobular carcinoma. The worse prognosis for young patients than other ages can be explained by their higher proportion of poorly differentiated cancers. Stage I patients aged 50–69 years having the best survival may be due to the earlier diagnosis achieved through screening. Although the analyses involved small sample of some categories, our findings support the existing literature concerning the prognostic effects of those factors. The determination of prognosis depends on the accurate assessment of prognostic factors and the appropriate choice of therapeutic and supportive intervention. Further studies are needed to confirm the results in a large sample and possibly to ensure the inclusion of other factors such as menopausal status, oestrogen receptor levels, progesterone receptor levels, number of nodes and waiting time for treatment.