Introduction

With the decrease in all-cause mortality in populations aged 85 years or older, the oldest-old population has rapidly increased in the United States1. It is estimated that the total number of adults aged 85 years and above double in 2060 compared with that in 20162. As life expectancy has increased, cancer morbidity has become a major concern among this population. In the US, the proportion of the oldest-old population only accounted for approximately 2% of the population; however, the proportion increased to 8% among newly diagnosed cancer cases1. This disparity exhibited a high cancer burden among this population. It is estimated that in 2019, 140,690 new cancer cases were diagnosed in individuals aged 85 years and older, including 61,830 males and 78,860 females1. Additionally, the number of cancer survivors in the oldest-old population is estimated to increase by 4.7 million in 20403, which is twofold higher than that in 2019.

The Global Cancer Statistics (2020) reported that breast cancer has become the most diagnosed cancer in females in most countries4. In the US, the incidence rate of breast cancer among females exhibited an increasing trend from the year 2010 to 2019, with an annual percent change (APC) of 0.5%5. Contrarily, another study showed that the incidence rate of breast cancer in females aged 85 years or above decreased since 2009, with an APC of − 2.1%1. Besides the incidence, the distribution of stages, surgery treatment, and 5-year relative survival differed between patients aged 85 years and higher or equal to 85 years1. Nevertheless, knowledge regarding the details of epidemiology and prognostic factors of breast cancer in females aged 85 years and older in the US is limited, especially at the molecular-level heterogeneity. In this study, we aim to provide information on females with breast cancer aged 85 years and older in the US, including data on the incidence, prevalence, stage distribution, initial treatment, and prognostic factors using the most recent data years (2010–2019) from the surveillance, epidemiology, and end-result (SEER) database.

Methods

Study population

The population-based cancer data is collected by the National Cancer Institute’s (NCI) SEER program in the US. In this study, we obtained data from SEER 17 registries, November 2021 submission (2000–2019), representing approximately 28% of the US population. The inclusion criteria were as follows: female patients aged 85 years and above, microscopically confirmed, not reported by autopsy or death certificate. The primary tumor site codes include C50.0 to C50.9. Because the details of receptor subtypes of breast cancer have been available since 2010, all data were extracted since then. Furthermore, only patients with only one primary breast cancer through lifetime were included. Carcinoma in situ was not included.

The receptor subtype was categorized into 5 groups based on the expression of hormone receptor (HR) and human epithelial growth factor receptor 2 (HER2), namely HR+/HER2−, HR+/HER2+, HR−/HER2+, HR−/HER2−, and unknown. The race/ethnicity was categorized into the following 4 mutually exclusive groups: White, Black, American Indians/Alaska Native (AI/AN), and Asian/Pacific Islander (API).

Incidence and prevalence rates

The SEER*Stat software (version 8.4.0; https://seer.cancer.gov/seerstat/) was used to calculate the incidence rate (2010–2019) and the annual prevalence rate (2010–2018, because the prevalence date is limited to January 1, 2019). Both rates were age-standardized based on the US standard population in 2000. Additionally, both rates were reported for the overall population and by receptor subtype, and/or race. The data are expressed as per 100,000 women.

The APC of incidence was calculated to determine trends or changes in rates over time, which was quantified by the NCI’s Joinpoint Regression Program (version 4.9.0.0; https://surveillance.cancer.gov/joinpoint/). In this method, a least square regression line was fitted to the natural logarithm of the rates using the calendar year as a regressor variable. This value was compared to 0 to indicate statistical significance.

Stage distribution

The stage distribution (2010–2019) was summarized based on the reported SEER combined stage in the SEER*Stat, which consisted of the following 4 categories: the local, regional, distant, and unknown stages. To summarize, the local stage refers to aggressive tumors that are completely confined to the organ of origin. The regional stage refers to neoplasms that extend to organs with or without regional lymph node infiltration. The distant stage refers to a lesion far from the primary tumor. Additionally, the shifts in stages over time were also summarized.

Initial treatment

First-course treatment data for breast cancer cases between 2010 and 2019 were summarized. Chemotherapy included conventional chemotherapy, targeted therapy, and immunotherapy because the most common targeted therapies were classified as chemotherapy in the SEER database6.

Nomogram

We construct and validate a visual nomogram to predict the 3-year overall survival (OS) among female patients with invasive breast cancer aged 85 years or older. Only cases between 2010 and 2014 were included. The following variables were used for analysis: age, race, tumor size, receptor subtype, SEER stage, and treatment pattern.

All selected patients were randomly divided into the training and validation cohorts at a ratio of 7:3. The baseline clinical characteristics between the two cohorts were compared by performing Pearson’s chi-square test. Univariable and multivariate Cox regression analyses were performed to determine the prognostic factors affecting OS in the training cohort. Subsequently, a nomogram was constructed to predict the 3-year OS probability of each patient using the significant factors, including age, tumor size, receptor subtype, SEER stage, and treatment pattern. The validation of the nomogram was evaluated by the consistency index (C-index), the area under the receiver operating characteristic curves (AUC), and the calibration curve both in the training and validation cohorts.

We divided all patients into three groups according to the 3-year OS predicted probability of the nomogram (high risk, predicted probability equals to or lower than 0.3; median risk, predicted probability between 0.3 and 0.6; and low risk, predicted probability higher than 0.6). Survival analyses were performed using the Kaplan–Meier method and the log-rank test among the three groups. A two-sided P-value of < 0.05 was considered statistically significant. All the analyses were performed using R software (version 4.2.1; https://www.r-project.org/).

Results

Study population overview

In total, 18,137 female patients with invasive breast cancer aged 85 years and older were identified from SEER 17 registries from 2010 to 2019. Among them, HR+/HER2− accounted for 68.7%, followed by HR−/HER2− (9.3%), HR+/HER2+ (7.4%), and HR−/HER2+ (3.1%). Patients with unknown receptor subtypes accounted for approximately 11.5%. The details of the clinical characteristics are summarized in Table S1.

Incidence

The incidence of female breast cancer among the oldest-old population was 181.82 (95% CI 179.18–184.49) per 100,000 women (Table 1). The incidence decreased from 184.73 per 100,000 women in the year 2010 to 177.71 per 100,000 women in the year 2019 (a change of − 3.8%; Fig. 1A), with an APC of − 1.0 (95% CI − 1.8 to − 0.1, P = 0.036; Table S2).

Table 1 The incidence rate of invasive breast cancer aged 85 years and older, United States, 2010 to 2019.
Full size table
Figure 1
Figure 1
Full size image

Incidence rate and annual prevalence of invasive breast cancer aged 85 years and older, United States, 2010 to 2019. (A) Incidence of overall and by receptor subtype, (B) incidence by race, (C) prevalence of overall and by receptor subtype. AI/AN American Indians/Alaska Native, API Asian/Pacific Islander.

Among different receptor subtype subgroups, the incidence of HR+/HER2− was the highest, which was much higher than HR+/HER2+, HR−/HER2+, HR−/HER2−, and unknown subtype (124.95 vs. 13.40 vs. 5.67 vs. 16.97 vs. 20.82 per 100,000 women; Table 1). Among different race subgroups, the incidence of the black population was 206.01 per 100,000 women, followed by the white population (187.36), API population (120.58), and AI/AN population (99.02) (Table 1). No significant change in incidence was found among each receptor subtype or race subgroup (Fig. 1A,B; the APC for each subgroup is presented in Table S2), except a decreasing trend of incidence in unknown receptor subtype subgroup (32.55 to 12.61 per 100,000 women from 2010 to 2019; APC: − 8.6, 95% CI − 11.1 to − 6.1, P < 0.001). The details of APC for each subgroup are summarized in Table S2.

Further, the analyses stratified by receptor subtype and race subgroups were performed. In terms of HR+/HER2−, the white and black populations showed similar incidences, which were much higher than those of the AI/AN and API populations (Fig. S1A). For HR−/HER2−, the black population showed the highest incidence, followed by the white and API population (Fig. S1D). For HR+/HER2+, HR−/HER2+, and unknown receptor subtypes, the incidences were similar among white, black, and API populations (Fig. S1B,C,E). The details of APC stratified by receptor subtype and race subgroups are summarized in Table S2.

We then determined the incidence of site-specific metastasis at diagnosis, including bone, brain, liver, and lung metastases. The incidence of bone metastasis was the highest among patients with HR+/HER2−, followed by lung and liver metastasis (Fig. S2A). For patients with HR+/HER2+, HR−/HER2+, and HR−/HER2−, the incidences of bone, liver, and lung metastases were similar (Fig. S2B–D). Furthermore, the incidences of brain metastasis in all receptor subtypes were extremely low. No distinct change in the incidence of site-specific metastasis was found. The details of APC for site-specific metastasis are summarized in Table S2.

Prevalence

The annual prevalence rate of female breast cancer in the oldest-old population decreased from 185.18 to 171.50 per 100,000 women from the year 2010 to 2018 (a change of − 7.4%), which was mainly attributed to the decrease of unknown receptor subtype (Fig. 1C). The annual prevalence of HR+/HER2−, HR+/HER2+, HR−/HER2+, and HR−/HER2− showed no distinct change. The annual prevalence of HR+/HER2− was consistently 8–11 times higher than that of Lumina B and HR−/HER2−, which was 20–29 times higher than that of HR−/HER2+ .

Stage distribution

The frequency stage distribution varied among different receptor subtypes. Compared with other receptor subtypes, HR+/HER2− exhibited the highest frequency of the local stage but the lowest frequency of the distant stage (Fig. 2). HR−/HER2+ exhibited a relatively higher frequency of the regional stage. Stage distributions between HR+/HER2+ and HR−/HER2− were similar.

Figure 2
Figure 2
Full size image

Stage distribution of invasive breast cancer aged 85 years and older, United States, 2010 to 2019.

The frequency distribution of all stages for HR+/HER2− was stable from 2010 to 2019 (Fig. S3A). For HR+/HER2+, the frequency of the local stage decreased over the years (from 60.9 to 46.9%), whereas the frequency of the regional stage increased over the years (from 27.8% to 38.1%) (Fig. S3B). Conversely, the frequencies of the local and regional stages for HR−/HER2− showed opposite trends compared with those for HR+/HER2+ (Fig. S3D). The frequency distribution of all stages for HR-/HER2+ varied over the years, but no particular trend was observed (Fig. S3C).

Initial treatment

Breast-conserving surgery (BCS) has been the most common treatment regime for female breast cancer, accounting for approximately 29.8%, followed by mastectomy (20.9%) and BCS+ radiotherapy (12.4%) (Fig. 3). BCS was more commonly performed in patients with local-stage cancer than in those with regional- or distant-stage cancer (42.1% vs. 14.1% vs. 3.6%). Mastectomy was more commonly performed in patients with regional-stage cancer than in those with local- or distant-stage cancer (32.6% vs. 18.5% vs. 8.0%). Chemotherapy and radiation were seldom used for these patients. Approximately 29.2% of the patients did not receive any treatment, and the number of patients with distant-stage (62.6%) and unknown-stage (87.1%) cancer was considerably higher than those with other stages.

Figure 3
Figure 3
Full size image

Patterns of initial treatment. BCS breast-conserving surgery, chemo chemotherapy (includes targeted therapy and immunotherapy), RT radiation. aA small number of these patients received chemotherapy.

The shifts in initial treatment trends are presented in Table S3. From 2010 to 2019, the number of patients undergoing mastectomy decreased markedly both in the local (from 25.9 to 13.4%) and regional (from 45.7 to 24.6%) stages, whereas the number of patients receiving no treatment in the local (from 15.6 to 24.9%) and regional (from 14.5 to 33.7%) stages increased. Additionally, the number of patients undergoing BCS in the local stage (from 39.6 to 44.9%) increased.

Nomogram

Between 2010 and 2014, 8904 patients were enrolled for constructing a nomogram, of which 6258 and 2682 were assigned to the training and validation cohorts, respectively. The clinical features and comparison of the two cohorts are presented in Table S4. Balanced data were obtained from the two cohorts. Furthermore, the results of the univariable and multivariate Cox regression analyses of the training cohort data showed that age, tumor size, receptor subtype, SEER stage, and treatment pattern were associated with 3-year OS (Table 2). Next, a nomogram was constructed based on these major factors (Fig. 4A). A certain point was assigned to each variable according to the point scale, and the total points were used to calculate the probability of the 3-year OS for each patient.

Table 2 Univariable and multivariable Cox analyses for overall survival.
Full size table
Figure 4
Figure 4
Full size image

A nomogram for predicting the 3-year overall survival. (A) The nomogram, (B) training calibration, (C) validation calibration. SEER Surveillance, Epidemiology, and End Results, BCS breast-conserving surgery, chemo chemotherapy (includes targeted therapy and immunotherapy), RT radiation, OS overall survival.

The nomogram was validated both in the training and validation cohorts. The C-index was 0.71 (95% CI 0.70–0.72) for the two cohorts. The calibration curves showed consistency between the predicted and observed probabilities in the two cohorts (Fig. 4B,C). Additionally, the predictive ability (AUC) of the nomogram was 0.74 for the training cohort and 0.75 for the validation cohort (Fig. S4). Moreover, all patients were divided into three groups on the basis of the probability of the predicted 3-year OS using the nomogram (Fig. 4A). The survival analyses based on the risk stratification showed that patients with high-risk and median-risk cancer exhibited much poorer survival than that exhibited by patients with low-risk cancer (Fig. 5).

Figure 5
Figure 5
Full size image

Survival analyses according to the risk stratification. (A) The training cohort, (B) The validation cohort. mCSS median cancer-specific survival.

Discussion

We are the first to report the epidemiological data on females with invasive breast cancer aging higher or equal to 85 years old in the US. Breast cancer incidence in the oldest-old population was 181.82 per 100,000 women, which was higher than incidence in women aged 20 years or older5. The Breast Cancer Statistics, 2022 in the US showed that invasive breast cancer incidence was higher in women aged 50 years and older than in women aged 20–49 years7. Similarly, ductal carcinoma incidence in situ was higher in women aged 50 years and older than in women aged 20–49 years old. These findings indicated the presence of a high breast cancer burden in older women in the US.

Breast cancer incidence in women aged 85 years and older decreased from 2009 to 20151, and the present results showed that the incidence continued to decrease from 2015 to 2019. The main subtype in women aged 85 years and older was HR+/HER2−, which accounted for 68.7% of all subtypes. Two factors, namely excess body weight and a reduced fertility rate, may contribute to this7. Patients with obesity showed higher estrogen levels, which could increase breast cancer incidence8. As an important target organ, the growth of the mammary gland can be affected, which may lead to the abnormal proliferation and transformation of mammary cells and even cancerization. Overweight and excessive mammographic screening could be associated with the increased incidence of estrogen receptor-positive breast cancer9,10. Reproductive factors, including age at first birth, parity, and breastfeeding, can affect breast cancer risk by altering the levels of circulating hormones such as estrogen11. Herein, a decreased fertility rate increased the risk of estrogen accumulation. Additionally, significantly decreased levels of an unknown receptor subtype (APC: − 8.6, 95% CI − 11.0 to − 6.1, P < 0.001) implied the improved ability of precise diagnosis. Nevertheless, the high amount of data on receptor subtypes in women aged 85 years and older (approximately 11.5%) was missing, and this trend was more common in this population than in women aged 25–84 years. Hence, more attention should be paid to addressing this issue. Conversely, the distribution of the identified receptor subtypes was similar between women aged 25–84 years and women aged 85 years and older12.

In addition to the receptor subtype, the incidence varied across different races. In women aged 85 years and older, the black population exhibited a slightly higher incidence than did the white population, followed by the incidence in the API and AI/AN population. Overall, the white population exhibited the highest incidence, followed by the black, AI/AN, and API populations7. The Breast Cancer Statistics, 2022 showed that the incidence was higher in the white population from 2000 to 2019 than in the black population. Nonetheless, the disparity may vary among different receptor subtypes and age groups when the present results are combined with those of the Acheampong report12. For patients with HR+/HER2−, the white population exhibited the highest incidence in the 40–54, 55–69, and 70–84-year-old-age groups. In the 25–39 and 85+-year-old-age groups, the incidence in the white and black populations was similar. Conversely, the highest incidence was observed in the black population in 40–54, 55–69, and 70–84-year-old-age groups in patients with HR−/HER2+. For patients with HR+/HER2+, the incidence in the white and black populations was similar in all age groups. For HR−/HER2−, the black population exhibited the highest incidence in all age groups. In addition to genetic factors, individual-level socioeconomic status and neighborhood social factors might contribute to the racial disparity in the incidence13,14,15.

Females with breast cancer aged 85 years and older exhibited relatively higher proportions of distant-stage and unknown-stage tumors compared with the proportions exhibited by females aged less than 85 years; however, the proportion of local-stage cancer was lower1. The present results showed a staged shift from regional-stage cancer to local-stage cancer in HR−/HER2− cases, which was consistent with the trend of stage shift in overall female breast cancer cases7. Notably, we found a stage shift from local-stage cancer to regional-stage cancer in HR+/HER2+ cases. The reason behind this is unknown and should be clarified in future studies.

For the treatment of breast cancer, surgery is the most common and important treatment approach. The selection of the type of surgical approach for females with breast cancer aged 85 years and older with different stages was similar to that for the overall female breast cancer cases, except that chemotherapy and radiation approaches were seldom selected for the oldest-old population7. The reason was the poor physical conditions and the higher proportion of comorbidity. Herein, we found numerous patients with local-stage cancer opting for BCS instead of mastectomy during the initial treatment course. Inconsistently, previous studies showed that mastectomy would be more likely selected by patients younger than 40 years or patients with larger and/or more aggressive tumor characteristics16,17,18. The possible reasons for opting for mastectomy as the first treatment included reluctance to undergo radiation and concerns over recurrence19,20. Notably, the proportions of not opting for any treatment increased in the oldest-old population, which was more pronounced in patients with local- and regional-stage cancers. Targeted therapy and immunotherapy could be recommended to these patients because these treatment regimens have exhibited good outcomes and fewer side effects21,22. Notably, a few patients who underwent surgery probably did not undergo axillary lymph node dissection, especially if they had the ER+/HER2− disease. Therefore, the nodal status may be underestimated in those who did not undergo surgery or axillary surgery and may be overestimated in women who underwent axillary surgery.

Herein, we first developed a visual nomogram to predict the 3-year OS rate in females with breast cancer aged 85 years and older based on five common clinical characteristics, namely age, receptor subtype, SEER stage, tumor size, and treatment pattern, which showed good discrimination and calibration. Further risk stratification effectively helped identify patients at a high risk of poor survival. Hence, the developed nomogram can serve as an auxiliary prediction tool in future clinical practice. Patients with a high risk of poor survival should receive more intense treatment and/or be closely followed up. Except for malignancies, the OS of the oldest-old population was markedly affected by comorbidities, such as hypertension, diabetes, coronary heart disease, and chronic obstructive pulmonary disease, which were not specified in the SEER database. Hence, future studies should integrate these factors into the nomogram to improve its efficacy. Additionally, part of patients categorized as receiving no treatment may die of other cause early rather than cancer. And these patients could not get anti-cancer treatment. Hence, these patients may not benefit from the nomogram.

Nevertheless, this study has some limitations that should be acknowledged. First, missing data on receptor subtypes was as high as 11.5%, which might affect the accurate evaluation of the four receptor subtypes to some extent. However, this was a reflection of real-world data, which suggested that more attention should be paid to accurately diagnosing this population. Second, the data from the SEER 17 registries accounted for 28% of the overall population, suggesting that this study only partly represented the epidemiological data, and it might not fully and accurately reflect the epidemiological data of the whole country. Finally, with limited data resources, we could not validate the nomogram using an external cohort, which needs to be addressed in the future.

Conclusion

This is the first epidemiological study on females with breast cancer aged 85 years and older in the US using the most recent data from SEER 17 registries (2010–2019). Breast cancer incidence varied across receptor subtypes and races. The developed predictive nomogram helped effectively identify patients with poor survival.