Time-trends in survival in young women with breast cancer in a SEER population-based study
- First Online:
- Cite this article as:
- Ademuyiwa, F.O., Groman, A., Hong, C. et al. Breast Cancer Res Treat (2013) 138: 241. doi:10.1007/s10549-013-2425-1
- 459 Views
Mortality improvements in young women with breast cancer (BC) may be attributable to treatment advances; screening likely plays a less significant role as mammography is not recommended <40. We examined time-trends in outcome in a cohort of young women. Our goal was to determine the contributions of treatment and screening to mortality improvements and evaluate whether differential outcomes by ER status exist. Using SEER, patients (73,447) were divided into three categories by diagnosis year (1990–1994, 1995–1999, 2000–2004) and also categorized as <40 or 40–50 years. Multivariate analysis was done to investigate the association of survival with time period for both age groups by ER status. Hazard ratios (HR) for mortality in women 40–50 with ER positive BC declined over time. With 1990–1994 as referent, the HR in 1995–1999 was 0.77 (0.69–0.86) and 0.65 (0.59–0.71) in 2000–2004 (p < 0.001). Women <40 with ER positive BC also had improvements over time. In ER negative patients, the degree of improvements over time was less than that seen in ER positive women. We report a survival disparity over time in young women by ER status. Patients with ER negative disease have not had the degree of improvements over time as seen in ER positive disease. Therefore, mortality improvements in young women with ER positive BC may be attributed to treatment advances with endocrine agents.
KeywordsBreast cancerYoung ageSurvivalTrends
In the United States, approximately 27 % of either non-invasive or invasive breast cancers occur in women under 50 years of age . Among adolescent and young adult women, breast cancer is the most frequently diagnosed malignancy . Breast cancer in young women remains poorly understood and is believed to represent a more biologically aggressive disease with a higher frequency of adverse histopathologic characteristics and worse outcomes [3, 4]. These cancers are more likely to be high-grade, triple negative, HER2 positive, and have the presence of lymphovascular space invasion [5–9]. They are also more likely to be associated with a positive family history  and black race . Young age also appears to be independently prognostic of a worse outcome compared to an older age at presentation [12–14].
Improved breast cancer outcomes have been seen in recent years [15–19], presumably due to widespread breast cancer screening and improved treatment strategies. Breast cancer mortality rates have decreased by 2.2 % each year from 2000 to 2009 for all ages combined . Specifically, for women under the age of 50 years, death rates have decreased by 3 % per year as opposed to a yearly decrease of 1.9 % in women ≥50 years. The slower mortality gains in older women may be explained by a higher frequency of comorbidities, poorer utilization of screening mammograms, and a lower probability of receiving standard of care treatments [21–23]. It is unclear if the more rapid gains in survival in younger women are due to screening or treatments. Since routine mammography is not generally recommended for women under the age of 40, it is reasonable to conclude that mortality improvements in this group are attributable to treatment advances with screening playing a less important role. Consequently, we sought to examine the time trends in breast cancer outcome, as well as presentation, in a population-based cohort of young women to determine the relative contributions of treatment and screening to improvements documented in the literature. Since patients with estrogen receptor (ER) negative breast cancer generally have a worse outcome than those with ER positive disease, we examined changes in outcome over time by ER status. In addition, we evaluated whether younger age remained prognostic for poorer outcome in this cohort.
Patients and methods
Patient and tumor characteristics at the time of diagnosis were compared across three time periods (1990–1994, 1995–1999, and 2000–2004) using the Wilcoxon rank sum and Pearson χ2 tests for continuous and categorical variables, respectively. A cox proportional hazard model was used to evaluate time trends in overall survival (OS) by ER status among women <40 and those 40–50 years of age separately. The primary endpoint was OS. We censored patients at 5 years to keep follow up consistent. Patients alive at the date of last follow up were censored. We adjusted the analyses for race, tumor grade, histology, tumor size, progesterone receptor (PR) status, radiation, surgery, American Joint Committee on Cancer (AJCC) stage, and numbers of positive lymph nodes as these were considered confounding factors. Age, period of diagnosis, and ER status, and the two-way and three-way interaction terms were included in the model as explanatory variables. To evaluate a potential interaction between ER status and age, Kaplan–Meier (KM) estimates of survival were initially generated and a log-rank test was used to test for differences in OS by ER status and age-group. Cox proportional hazard models were subsequently used to calculate multivariate hazard ratios and 95 % CIs with adjustments and age, period of diagnosis, and ER status, and the two-way and three-way interaction terms were included in the model as explanatory variables. The Akaike information criterion (AIC) was used to determine the best model fit (smallest is best). The model with the smallest AIC was the model with two- and three-way interactions between age, period of diagnosis, and ER status. Four contrasts were implemented to determine the effect of period of diagnosis within the age-group and ER status categories. To evaluate time trends in breast cancer presentation, logistic regression analysis was used to generate odds ratios (OR) and 95 % CIs for developing AJCC stages I, II, III/IV breast cancer and breast cancers of varying tumor size across the time periods being examined for women <40, and for women age 40–50 years. As the SEER registries expanded during this study period, we performed a sensitivity analysis limiting the dataset to registries in existence during 1990–1994. The results did not differ from the entire dataset; thus, we present results of the larger dataset. All associations were considered statistically significant at an alpha <0.01 (p value 0.01). All statistical analyses were performed using the SAS statistical software package (version 9.3; SAS Institute, Inc., Cary, NC).
Patients’ characteristics by year of diagnosis
Time trends in mortality
Mortality risk over time by age group and ER status
5 year OS (95 % CI)
10 year OS (95 % CI)
95 % CI
95 % CI
95 % CI
Time trends in breast cancer characteristics
Trends in breast cancer characteristics over time by age
ORb (95 % CI)
OR (95 % CI)
OR (95 % CI)
>1 to <3 cm
>3 to ≤5 cm
>1 to <3 cm
>3 to ≤5 cm
Impact of age on outcomes
Multivariate model results for 5 year survival
95 % CI
In this study, we aimed to determine the relationship between age and time trends in survival and breast cancer presentation in a population-based cohort of young women stratifying by ER status. To achieve this aim, we analyzed 73,447 patients aged 50 years and under from the SEER database. Several findings emerged. First, our data clearly demonstrate that improvements in breast cancer outcome over time have occurred in women ≤ 50 years with ER positive disease, with mortality reductions being similar in women aged 40-50 as well as younger than 40 years. Although improvements in women with ER negative breast cancer were also seen, the degrees of improvements were much less than those observed in women with ER positive breast cancer. To our knowledge, no other study has reported similar findings. Results of recent epidemiologic data showing improvements over time for all ages  are likely driven by the much larger proportion of ER positive breast cancer that occurs. Several possible explanations for our findings exist. The greater degree of improvements over time in ER positive women may be attributable to treatment with endocrine therapies with screening playing a less important role. This is further buttressed by the observation that mortality reductions were also seen in ER positive women younger than 40 years, a group that is not likely to undergo routine screening. In fact, although it did not focus specifically on patients with ER positive disease, a study evaluating trends in breast cancer mortality in several European countries demonstrated reductions in breast cancer deaths of up to 16–29 % between 1989 and 2006 . The authors concluded that since the European countries studied had wide differences in screening practices, the observed similarities in mortality reduction were likely attributable to improvements in treatment and the healthcare systems and less likely attributable to screening. The degree of improvements in patients with ER positive disease seen over time could also be related to the release of the Early Breast Cancer Trialists’ Collaborative Group overview analysis showing a substantial benefit from the use of 5 years of tamoxifen in premenopausal ER positive patients, which rapidly became the new standard of care. Another plausible and contributory explanation may be that women with ER negative disease may derive less survival benefits from screening mammography than those with ER positive disease. It is noteworthy that studies have shown that women with ER negative disease tend to have breast cancers diagnosed as interval cancers [25, 26], likely due to the rapid proliferation characteristics such tumors possess. With respect to the presentation of breast cancer over time, our data showed that women <40 years, who are statistically more likely to have ER negative disease, have actually presented with more advanced disease over time. One should not conclude, however, that mammograms have no role in detecting breast cancer in ER negative patients as determination of the ER status is done after diagnosis. Women who may be at an elevated risk for ER negative breast cancer, such as those with a strong family history, BRCA mutation carriers, or those of African descent, may require more intense screening with the use of magnetic resonance imaging in combination with mammography. In addition, as slower survival gains have been made in this cohort of women, improved adjuvant therapies need to be developed.
We also observed an increasing proportion of breast cancer over time in women aged 40–50 years compared with women <40 years. Among all women in this study, the proportion of those aged 40–50 years and diagnosed during 1990 and 1994 was 75.3%; 77.5 % between 1995 and 1999; and 78.9 % between 2000 and 2004. This is likely due to the increasing usage of screening mammograms in this age group over time. This is also supported by the increased probability of being diagnosed with earlier disease observed in this age group over time.
Lastly, unlike ER positive breast cancer patients, younger age did not appear to be associated with OS after controlling for multiple factors in patients with ER negative disease. There was much less 5 year OS variation by age in ER negative patients than that seen in ER positive patients, suggesting a poorer prognosis irrespective of age. The association between poorer breast cancer outcomes and younger age may be stronger for ER positive breast cancers where a plausible explanation for this relationship exists. It is known that younger premenopausal women are more likely to regain ovarian function after cytotoxic chemotherapy than older women, and those who do regain this function have an inferior outcome compared to those who do not [27, 28]. This may be less important in ER negative breast cancers that are believed to be hormone-independent. The other consideration is that ER negative breast cancers represent a more biologically heterogeneous disease than ER positive breast cancer and thus the contribution of younger age or other factors as negative prognostic indicators may not be as apparent. Indeed, we had previously reported on the absence of a significant relationship between obesity and clinical outcomes in patients with triple negative breast cancers .
Although these data provide a population-based representation of trends in breast cancer outcome and presentation by age, our findings must be interpreted in light of several limitations. As SEER does not collect detailed information on treatment, we were not able to control for receipt of chemotherapy. It is possible that patients in the younger age group and/or with ER negative disease in this study were more likely to be treated with chemotherapy as opposed to older patients or those with ER positive breast cancer. If that was the circumstance, one may have expected improved outcomes to be seen in those more likely to receive chemotherapy. Another key variable for which we could not account due to similar data collection reasons was HER2 status. In addition, we were not able to report on breast cancer-specific survival, as the reliability of cause of death coding by SEER has been questioned. During a preliminary analysis of this dataset, only 1.9 % of the patients was coded with a breast cancer death. This is an underrepresentation of the true numbers of breast cancer deaths particularly in a younger age group that usually does not have multiple competing causes of deaths. Therefore, as this age group is not one that is usually fraught with comorbidities, OS represents an accurate estimation of the burden of cancers in this group. Analyzing survival data over time can be subject to problems associated with stage migration, due to more sophisticated diagnostic techniques. Patients in a more recent time period may be more likely to present with more advanced disease due to improved detection of smaller metastasis. Indeed, our data did show that the proportion of those with stage III/IV disease did increase slightly from 9.2 % in 1990–1994 to 10.4 % in 2000–2004. Similarly, patients in the more distant past may be misclassified as having earlier stage breast cancer with a worse outcome due to the fact that they actually had later stage disease which was unable to be detected. However, there is no biological reason as to why this would differ by age group. Approximately 25 % of our study population did not reach 5 years of follow up; the proportion not reaching 5 years follow up was higher in the most recent time period. This potentially introduces a source of bias to the results. Lastly, we acknowledge that missing data can bias the analysis of registry data. Our analysis showed that the proportion of missing ER data was slightly higher in the earlier time periods. While several statistical methods to account for missing data do exist, we are hesitant to employ such techniques to reassign unknown ER status which is a key classifier of breast cancers. Therefore, we included only those patients with known ER status.
In summary, this study extends the literature on trends in breast cancer survival in young women by reporting a disparity by ER status. As we have demonstrated that greater mortality improvements over time have occurred in women ≤50 years with ER positive disease, the extent to which improvements have occurred in women with ER negative disease remains an open question. While it is reassuring that improvements in breast cancer outcome have occurred over time for some young women, further research is needed to fully understand breast cancer in younger women particularly for those who are afflicted with ER negative disease. We suggest increased attempts to involve this group of women in clinical trials.
Conflict of interest