Nomograms predict survival in elderly women with triple-negative breast cancer: A SEER population-based study

Abstract

BACKGROUND:

The effective treatment of breast cancer in elderly patients remains a major challenge.

OBJECTIVE:

To construct a nomogram affecting the overall survival of triple-negative breast cancer (TNBC) and establish a survival risk prediction model.

METHODS:

A total of 5317 TPBC patients with negative expression of estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2) who were diagnosed and received systematic treatment from 2010 to 2015 were collected from the American Cancer Surveillance, Epidemiology and End Results (SEER) database. They were randomly divided into training set ( $n=$ 3721) and validation set ( $n=$ 1596). Univariate and multivariate Cox regression analysis were used to identify prognostic features, and a nomogram was established to predict the probability of 1-year, 3-year and 5-year OS and BCSS. We used consistency index (C-index), calibration curve, area under the curve (AUC) and decision curve analysis (DCA) to evaluate the predictive performance and clinical utility of the nomogram.

RESULTS:

The C-indices of the nomograms for OS and BCSS in the training cohort were 0.797 and 0.825, respectively, whereas those in the validation cohort were 0.795 and 0.818, respectively. The receiver operating characteristic (ROC) curves had higher sensitivity at all specificity values as compared with the Tumor Node Metastasis (TNM) system. The calibration plot revealed a satisfactory relationship between survival rates and predicted outcomes in both the training and validation cohorts. DCA demonstrated that the nomogram had clinical utility when compared with the TNM staging system.

CONCLUSION:

This study provides information on population-based clinical characteristics and prognostic factors for patients with triple-negative breast cancer, and constructs a reliable and accurate prognostic nomogram.

Keywords

Triple negative breast cancer nomogram overall survival breast cancer-specific survival SEER database

1. Introduction

The prevalence of breast cancer is age-dependent, such that older age is associated with increased morbidity [1]. The effective treatment of breast cancer in elderly patients remains a major challenge. Older breast cancer survivors contend with health issues and concerns around wellbeing that are related to cancer treatment, the anagen process, and other considerations around illness, treatment, and survivorship. Health concerns among breast cancer survivors include concomitant diseases, osteoporosis, a range of treatment-associated symptoms, effects on physical functioning, concerns around cognitive function, nutritional deficits, and reduced physical activity [2].

Older women are more frequently diagnosed with estrogen-driven tumors, which have a more favorable prognosis and can be effectively treated with targeted hormone therapy [3]. Although the proportion of more favorable breast cancer subtypes increases with age, a high proportion of older patients are still diagnosed with high-risk tumors [4]. According to the International Organization for Geriatric Oncology (SIOG) recommendations, the old Oncology patients were Classified as two age groups for age at diagnosis ( $<$ 70 years and $\geqslant$ 70 years) [5]. Older patients, aged 70 years or older, with breast cancer are less likely to receive chemotherapy, although the use of chemotherapy has been increasing in this population since developments in cancer treatment that have occurred since 2000 [6, 7, 8]. As the higher average life expectancy reaches 80 years in many countries, especially for women, considering increased patients aged $\geqslant$ 70 years, herein, it is a clinical challenge to strike a balance between decision-making, undertreatment and overtreatment in patients aged $\geqslant$ 70 years with breast cancer.

Triple-negative breast cancer (TNBC) is characterized by the absence of estrogen receptor (ER) and progesterone receptor (PR) expression as well as a lack of human epidermal growth factor receptor 2 (HER2) overexpression; this cancer subtype comprises approximately 10–15% of all diagnosed breast cancer subtypes as well as more than one-third of all breast cancer-related deaths [9, 10, 11]. TNBC is difficult to treat and is associated with a poor prognosis resulting from a lack of validated targeted treatments concomitant with the highly aggressive biological behavior of this cancer subtype [12, 13]. TNBC is one of the most invasive subtypes of breast cancer and, as suggested above, one of the most challenging to treat [14]. Based on the clinical complexity of older breast cancer patients and the highly aggressive nature of triple-negative breast cancer, there is a need to create a predictive survival instrument in order to facilitate effective clinical decision-making within this population.

The Surveillance, Epidemiology, and End Results (SEER) program is a cancer registry that covers approximately 30% of the US population. Through the National Data Center associated with SEER, this database contains comprehensive patient information, including demographic and clinical data as well as comprehensive follow-up data. We note that TNBC occurring in older patients is a specific category of triple negative breast cancer.

In this study, we used the SEER database to evaluate clinical characteristics and prognostic factors with regard to TNBC occurring in elderly women. Nomograms have been identified as a novel and reliable tool that incorporate information on demographic and clinicopathological factors in order to make accurate prognostic predictions for many cancer types. Studies have reported on these synthetic and visualized predictive models (i.e., nomogram prediction models) with regard to a wide range of cancers [15, 16, 17, 18]. The objective of the current study was to establish a nomogram accurately and reliably predicting overall survival (OS) and breast cancer-specific survival (BCSS) in elderly patients with triple-negative breast cancer.

2. Materials and methods

2.1 Data collection

The data in this study were obtained from the SEER-linked database, which collects cancer incidence data from 18 population-based cancer registries. SEER-Stat software (version 8.3.9, National Cancer Institute, Bethesda, Maryland, USA) was used to filter and collect information on representative patients (see http://seer.cancer.gov/). The random non-release back extraction for random grouping was used in a ratio of 1:2. And the package is dplyr package of R language. The even number entered the training set ( $n=$ 3721) and the odd number entered the validation set ( $n=$ 1596).

2.2 Cohort selection

The cases selected in our study were elderly women ( $\geqslant$ 70 years) with triple-negative breast cancer. Since the SEER database has only collected information on molecular subtypes and breast cancer metastases since 2010, we selected patients diagnosed with breast cancer between 2010 and 2015 for the current evaluation. The major inclusion criteria for this study were as follows: (1) cases that were pathologically diagnosed with primary breast cancer, (2) age $\geqslant$ 70 years, (3) cases with information on both OS and BCSS, and (4) the breast cancer molecular subtype of the enrolled patients was classified as TNBC. A total of variables were selected for evaluation in this study, including follow-up data (OS and BCSS) and the following covariates: race, age, laterality, grade, AJCC (American Joint Committee on Cancer) 7th edition (2010+) stage I-IV, Tumor Node Metastasis (TNM) stage, metastasis status (bone, brain, lung, liver), surgery, radiotherapy, chemotherapy, and several other covariates as shown in Fig. 1.

Figure 1.

Flowchart describes inclusion criteria and assignment of training and validation cohorts.

2.3 Statistical analysis

The patients were randomly divided into training and validation cohorts at a ratio of 7:3 (training cohort $n=$ 3,721, validation cohort $n=$ 1,596). Candidate predictors for survival in older TNBC patients were selected based on clinical importance, scientific knowledge, and information on predictors as identified within previously published articles. Univariate and multivariate survival analyses were performed using a Cox proportional hazards regression model [19]. Firstly, a univariate analysis was conducted with each potential prognostic factor as the only dependent variable; prognostic factors with a $P$ -value of $<$ 0.05 were then included in a multivariate Cox proportional hazards regression model. Following this, the independent prognostic factors screened by multivariate Cox regression analysis were applied to construct a prognostic nomogram; OS and BCSS probabilities were evaluated at 1, 3, and 5 years for older patients with TNBC.

For validation of the final nomogram, the Harrell concordance index (C-index) was used to verify predictive precision [20]. The value of the C-index can range from 0.5 to 1.0, with 0.5 indicating random chance and 1.0 indicating perfect ability to accurately distinguish outcomes using the model. Prognostic predictive ability was validated through a receiver operating characteristic (ROC) curve used for assessing the area under the curve (AUC), constructing calibration curves, and conducting decision curve analysis (DCA). We plotted Kaplan-Meier curves based on the tertile score stratifications predicted from the model and conducted comparisons using the log-rank test.

The cutoff values for the predicted tertile score stratifications were determined using X-tile software (version 3.6.1; Yale University School of Medicine, New Haven, CT, USA) [21]. The elementary cohort was randomly separated into training and validation cohorts using R statistical software (version 4.1.2, http://www.rproject.org/). Nomogram model construction was achieved using the “rms” and “survival” packages in R.

3. Results

3.1 Clinicopathological characteristics in the training and validation cohorts

A total of 5,317 women aged 70 years or older who were diagnosed with triple-negative breast cancer between 2010 and 2015 were selected from the SEER database and were randomly separated into two cohorts, a training cohort ( $N=$ 3,721) and a validation cohort ( $N=$ 1,596). The training cohort was used to develop a nomogram, while the validation cohort was used for verification. More than half of the patients were in the 70–74 (37.9%) and 75–79 (26.4%) age ranges. The majority of participants were Caucasian (78%). The validation cohort had similar characteristics to the training cohort, demonstrating that the allocation of patients was performed according to the principle of randomization. The clinicopathologic characteristics of the training and validation cohorts are shown in Table 1.

Table 1
The demographic and baseline characteristics of the Cohort with 70 years or older with triple-negative breast cancer

Characteristics	All patients ( $n=$ 5317)		Training cohort ( $n=$ 3721)		Validation cohort ( $n=$ 1596)
	Number of patients (%)		Number of patients (%)		Number of patients (%)
Age
70–74 years	2015	(37.9)	1376	(37.0)	639	(40.0)
75–79 years	1405	(26.4)	1014	(27.3)	391	(24.5)
80–84 years	1042	(19.6)	727	(19.5)	315	(19.7)
85+ years	855	(16.1)	604	(16.2)	251	(15.7)
Race
American Indian/Alaska Native	21	(0.4)	14	(0.4)	7	(0.4)
Asian or Pacific Islander	323	(6.1)	217	(5.8)	106	(6.6)
Black	825	(15.5)	581	(15.6)	244	(15.3)
White	4148	(78.0)	2909	(78.2)	1239	(77.6)
Origin of primary
Left	2757	(51.9)	1922	(51.7)	835	(52.3)
Right	2560	(48.1)	1799	(48.3)	761	(47.7)
Grade
Grade I	170	(3.2)	109	(2.9)	61	(3.8)
Grade II	1267	(23.8)	887	(23.8)	380	(23.8)
Grade III	3835	(72.1)	2694	(72.4)	1141	(71.5)
Grade IV	45	(0.8)	31	(0.8)	14	(0.9)
Stage Group
I	2198	(41.3)	1511	(40.6)	687	(43.0)
II	2039	(38.3)	1456	(39.1)	583	(36.5)
III	768	(14.4)	531	(14.3)	237	(14.8)
IV	312	(5.9)	223	(6.0)	89	(5.6)
T stage
T1	2511	(47.2)	1715	(46.1)	796	(49.9)
T2	1989	(37.4)	1439	(38.7)	550	(34.5)
T3	406	(7.6)	288	(7.7)	118	(7.4)
T4	411	(7.7)	279	(7.5)	132	(8.3)
N stage
N0	3677	(69.2)	2568	(69.0)	1109	(69.5)
N1	1051	(19.8)	741	(19.9)	310	(19.4)
N2	323	(6.1)	235	(6.3)	88	(5.5)
N3	266	(5.0)	177	(4.8)	89	(5.6)
M stage
M0	5005	(94.1)	3498	(94.0)	1507	(94.4)
M1	312	(5.9)	223	(6.0)	89	(5.6)
Surgry
No	429	(8.1)	314	(8.4)	115	(7.2)
Yes	4888	(91.9)	3407	(91.6)	1481	(92.8)
Radiation
None/Unknown	2832	(53.3)	1975	(53.1)	857	(53.7)
Yes	2485	(46.7)	1746	(46.9)	739	(46.3)
Chemotherapy
No/Unknown	3010	(56.6)	2086	(56.1)	924	(57.9)
Yes	2307	(43.4)	1635	(43.9)	672	(42.1)
Liver metastases
No	5242	(98.6)	3665	(98.5)	1577	(98.8)
Yes	75	(1.4)	56	(1.5)	19	(1.2)
Lung metastases
No	5180	(97.4)	3624	(97.4)	1556	(97.5)
Yes	137	(2.6)	97	(2.6)	40	(2.5)

Table 1, continued
Characteristics	All patients ( $n=$ 5317)		Training cohort ( $n=$ 3721)		Validation cohort ( $n=$ 1596)
	Number of patients (%)		Number of patients (%)		Number of patients (%)
Bone metastases
No	5190	(97.6)	3629	(97.5)	1561	(97.8)
Yes	127	(2.4)	92	(2.5)	35	(2.2)
Brain metastases
No	5292	(99.5)	3704	(99.5)	1588	(99.5)
Yes	25	(0.5)	17	(0.5)	8	(0.5)

Table 2

Univariate and multivariate analysis with Cox regression for OS

Variables	Univariate analysis			Multivariate analysis
	HR (95% CI)		$p$ value	HR (95% CI)		$p$ value
Age
70–74 years	Reference			Reference
75–79 years	1.44	(1.22–1.69)	$<$ 0.001	1.4	(1.19–1.65)	$<$ 0.001
80–84 years	2.23	(1.9–2.61)	$<$ 0.001	1.81	(1.53–2.15)	$<$ 0.001
85+ years	3.4	(2.9–3.97)	$<$ 0.001	2.27	(1.91–2.7)	$<$ 0.001
Origin of primary
Lift	Reference			Reference
Right	0.92	(0.83–1.03)	0.163	NA	NA
Race
Other	Reference			Reference
Asian or Pacific Islander	0.49	(0.22–1.06)	0.070	NA	NA
Black	1.49	(0.31–1.38)	0.259	NA	NA
White	2.49	(0.25–1.09)	0.083	NA	NA
Grade
Grade I	Reference			Reference
Grade II	1.58	(1.0–2.5)	0.048	1.28	(0.81–2.02)	0.298
Grade III	2.18	(1.4–3.39)	0.001	1.69	(1.08–2.64)	0.022
Grade IV	3.55	(1.84–6.85)	$<$ 0.001	2.19	(1.12–4.27)	0.022
T stage
T1	Reference			Reference
T2	2.57	(2.24–2.96)	$<$ 0.001	1.96	(1.69–2.27)	$<$ 0.001
T3	5.81	(4.82–7.02)	$<$ 0.001	2.86	(2.32–3.53)	$<$ 0.001
T4	8.52	(7.12–10.19)	$<$ 0.001	3.33	(2.68–4.14)	$<$ 0.001
N stage
N0	Reference			Reference
N1	2.32	(2.03–2.64)	$<$ 0.001	1.55	(1.34–1.79)	$<$ 0.001
N2	3.12	(2.58–3.76)	$<$ 0.001	2.05	(1.67–2.5)	$<$ 0.001
N3	5.2	(4.29–6.31)	$<$ 0.001	2.86	(2.3–3.55)	$<$ 0.001
Surgery
NO	Reference			Reference
Yes	0.13	(0.12–0.15)	$<$ 0.001	0.33	(0.28–0.39)	$<$ 0.001
Chemotherapy
NO	Reference			Reference
Yes	0.73	(0.65–0.82)	$<$ 0.001	0.6	(0.52–0.68)	$<$ 0.001
Radiation
NO	Reference			Reference
Yes	0.44	(0.39–0.49)	$<$ 0.001	0.59	(0.52–0.67)	$<$ 0.001
Liver metastases
NO	Reference			Reference
Yes	12.66	(9.55–16.78)	$<$ 0.001	3.2	(2.31–4.44)	$<$ 0.001

Table 2, continued
Variables	Univariate analysis			Multivariate analysis
	HR (95% CI)		$p$ value	HR (95% CI)		$p$ value
Lung metastases
NO	Reference			Reference
Yes	8.95	(7.17–11.18)	$<$ 0.001	1.51	(1.14–1.99)	0.003
Bone metastases
NO	Reference			Reference
Yes	7.79	(6.17–9.84)	$<$ 0.001	1.49	(1.11–2.00)	0.008
Brain metastases
NO	Reference			Reference
Yes	15.22	(9.25–25.05)	$<$ 0.001	2.93	(1.72–4.99)	$<$ 0.001

Figure 2.

Nomograms for predicting 1-, 3-, and 5-year OS (A) and BCSS (B) for patients with the prognosis factors. The total points are calculated by summing up the points for each factor. The predicted probabilities of OS and BCSS can be obtained by projecting the location of the total points to the bottom scales. OS, overall survival; BCSS, breast cancer-specific survival.

Table 3

Univariate and multivariate analysis with cox regression for BCSS

Variables	Univariate analysis			Multivariate analysis
	HR (95% CI)		$p$ value	HR (95% CI)		$p$ value
Age
70–74 years	Reference			Reference
75–79 years	1.38	(1.14–1.67)	0.001	1.46	(1.2–1.78)	$<$ 0.001
80–84 years	1.67	(1.36–2.05)	$<$ 0.001	1.65	(1.34–2.02)	$<$ 0.001
85+ years	2.65	(2.18–3.23)	$<$ 0.001	2.28	(1.86–2.8)	$<$ 0.001
Origin of primary
Lift	Reference			Reference
Right	0.95	(0.83–1.1)	0.506	NA		NA
Race
Other	Reference			Reference
Asian or Pacific Islander	1.06	(0.26–4.4)	0.935	NA		NA
Black	1.58	(0.39–6.38)	0.521	NA		NA
White	1.18	(0.29–4.74)	0.814	NA		NA
Grade
Grade I	Reference			Reference
Grade II	1.44	(0.8–2.59)	0.230	1.07	(0.59–1.95)	0.8209
Grade III	2.31	(1.31–4.1)	0.004	1.43	(0.8–2.56)	0.2242
Grade IV	2.86	(1.17–6.99)	0.021	1.27	(0.51–3.17)	0.6095
T stage
T1	Reference			Reference
T2	3.55	(2.91–4.33)	$<$ 0.001	2.42	(1.96–2.97)	$<$ 0.001
T3	9.46	(7.43–12.03)	$<$ 0.001	3.7	(2.83–4.83)	$<$ 0.001
T4	14.03	(11.13–17.67)	$<$ 0.001	3.91	(2.97–5.15)	$<$ 0.001
N stage
N0	Reference			Reference
N1	3.26	(2.76–3.86)	$<$ 0.001	1.79	(1.49–2.14)	$<$ 0.001
N2	4.59	(3.67–5.74)	$<$ 0.001	2.56	(2.01–3.25)	$<$ 0.001
N3	8.22	(6.59–10.26)	$<$ 0.001	3.4	(2.65–4.35)	$<$ 0.001
Surgery
NO	Reference			Reference
Yes	0.11	(0.09–0.13)	$<$ 0.001	0.32	(0.26–0.4)	$<$ 0.001
Chemotherapy
NO	Reference			Reference
Yes	1.06	(0.92–1.23)	0.393	NA		NA
Radiation
NO	Reference			Reference
Yes	0.49	(0.43–0.57)	$<$ 0.001	0.65	(0.55–0.75)	$<$ 0.001
Liver metastases
NO	Reference			Reference
Yes	15.67	(11.54–21.28)	$<$ 0.001	2.89	(2.03–4.12)	$<$ 0.001
Lung metastases
NO	Reference			Reference
Yes	12.83	(10.14–16.22)	$<$ 0.001	1.76	(1.3–2.37)	$<$ 0.001
Bone metastases
NO	Reference			Reference
Yes	10.99	(8.57–14.09)	$<$ 0.001	1.44	(1.05–1.99)	0.024
Brain metastases
NO	Reference			Reference
Yes	21.55	(13.05–35.58)	$<$ 0.001	2.89	(1.65–5.07)	$<$ 0.001

3.2 Univariate and multivariate Cox proportional hazards analyses

Univariate and multivariate Cox proportional hazard regression analyses were used to explore prognostic factors for OS and BCSS. Following this, all factors showing statistical significance in the univariate analysis were subjected to multivariate analysis within a Cox regression model The factors independently associated with OS in the multivariate analysis were age, grade, T stage, N stage, surgery, radiation, chemotherapy, bone metastases, lung metastases, liver metastases, and brain metastases (all $P<$ 0.05), which are shown in Table 2, whereas age, T stage, N stage, surgery, radiation, bone metastases, lung metastases, liver metastases, and brain metastases were statistically significantly associated with BCSS (Table 3). With regard to age subgroups, risk estimates were found to be statistically significantly higher in the 80–85 and $\geqslant$ 85 years age groups as compared with the younger groups.

R statistical software was applied to develop a nomogram for OS and BCSS (i.e., in order to predict patient prognoses), and all of the major variables specified above were used to establish the nomograms for OS and BCSS. Nomograms constructed based on the final Cox models were used to predict survival probabilities (i.e., 1 -, 3 -, and 5-year OS and BCSS) for older patients with TNBC (Fig. 2). By aggregating the total scores from all the aforementioned variables and casting these scores to the total point scale, we were able to predict outcome probabilities for OS and BCSS in older patients with TNBC.

3.3 Calibration and validation of the prognostic nomograms

C-index values, calibration curves, ROC curves, and DCA curves were used to identify the efficacy of the predictive nomograms. In the training cohort, the C-index for the OS nomogram was 0.797 (95% confidence interval [CI], 0.784–0.810) and the C-index for the BCSS nomogram was 0.825 (95% CI, 0.811–0.840). In the validation cohort, the OS C-index for the nomogram was 0.795 (95% CI, 0.775–0.814) and the BCSS nomogram index was 0.818 (95% CI, 0.795–0.841). These results indicate that the nomograms were repeatable and accurately predicted both OS and BCSS.

Figure 3.

Calibration curves for the 1-, 3-, and 5-year. (A, B, C) calibration curves for OS in the training cohort; (D, E, F) calibration curves for OS in the validation cohort; (G, H, I) calibration curves for BCSS in the training cohort; (J, K, L) calibration curves for BCSS in the validation cohort. OS, overall survival; BCSS, breast cancer-specific survival.

Figure 4.

ROC curves for the 1-, 3-, and 5-year. (A, B, C) ROC curves for BCSS in the training cohort; (D, E, F) ROC curves for OS in the validation cohort; (G, H, I) ROC curves for BCSS in the training cohort; (J, K, L) ROC curves for OS in the validation cohort.OS, overall survival; BCSS, breast cancer-specific survival,ROC, receiver operating characteristic curve; AUC, areas under the ROC curve.

Moreover, in both the training and validation cohorts, the calibration curves of nomograms predicting 1-, 3-, and 5-year OS and BCSS as well as the probability of recurrence indicated that the model had good predictive value (Fig. 3). The 1-, 3-, and 5-year ROC results showed that the sensitivities and specificities of the OS and BCSS nomograms were higher than those of the AJCC TNM staging system in both the training and validation cohorts (Fig. 4). The 1-year, 3-year and 5-year AUC values for the validation cohort OS nomograms were improved as compared with prognostic evaluations conducted using the TNM staging system (86.8%, 81.7%, 79.4% vs. 82.3%, 76.0%, and 72.6%, respectively). In addition, in the validation cohort, the predictive performance of nomograms in terms of BCSS was superior to that of TNM staging (1-year BCSS: 88.0% vs. 85.9; 3-year BCSS: 83.9% vs. 81.1%; 5-year BCSS: 81.5% vs. 78.6%). Moreover, the DCA results suggested that the prognostic nomograms constructed herein had high net benefits with regard to clinical decision-making (Fig. 5).

Figure 5.

DCA curves for the 1-, 3-, and 5-year. (A, B, C) DCA curves for BCSS in the training cohort; (D, E, F) DCA curves for OS in the validation cohort; (G, H, I) DCA curves for BCSS in the training cohort; (J, K, L) DCA curves for OS in the validation cohort. DCA, decision curve.

Figure 6.

Survival of older TNBC patients according to different risk groups. In the validation cohort, OS (A) in nomogram-based low-, moderate- and high-risk subgroups; BCSS (B) in nomogram-based low-, moderate- and high-risk subgroups. In the training cohort, OS (C) and BCSS (D) in nomogram-based low-, moderate- and high-risk subgroups, respective. DCA, decision curve analysis; TNBC, triple-negative breast cancer; OS, overall survival; BCSS, breast cancer-specific survival.

3.4 Performance of nomograms stratified based on risk scores

To evaluate the discrimination capability of the model, we stratified by tertiles of the predicted probability scores within the nomograms and then plotted these findings as Kaplan-Meier curves for OS and BCSS (Fig. 6). According to the tertile cut-off values for the OS nomogram, the dataset was divided into low-risk (score $\leqslant$ 181), moderate-risk (181 $<$ score $\leqslant$ 272), and high-risk (score $>$ 272) groups. Using a similar approach, the BCSS group was divided into three subgroups of elderly patients with triple-negative breast cancer based on the derived scores ( $\leqslant$ 100, 100 to 186, and $>$ 186). In the validation cohort, the low-risk group had the highest OS rates of 97.1% at 1 year, 87.5% at 3 years, and 78.8% at 5 years, followed by the moderate-risk group (85.3%, 56.3%, and 39.2%) and the high-risk group (47.8%, 19.4%, and 12.1%). There was a statistically significant difference in survival outcomes among the three groups (Fig. 6).

In terms of BCSS, the low-risk group had the highest BCSS rates of 97.9% at 1 year, 90.3% at 3 years, and 85.9% at 5 years, followed by the moderate-risk group (89.1%, 68.3%, and 56.7%) and the high-risk group (49.3%, 20.4%, and 8.8%). Similar results were observed in the validation cohort. Survival curves showed that risk stratification differentiated well between OS and BCSS in all subgroups ( $P<$ 0.05).

4. Discussion

TNM staging is the most commonly implemented breast cancer staging system and is used as a basic prognostic factor within prognostic evaluations. Nomograms have been shown to provide a clearer prediction of prognoses as compared with TNM staging for some cancers [22, 23, 24]. It has been documented that nomograms play a key role in the prediction of breast cancer patients and can effectively distinguish the tumor area of breast cancer patients and can help in the diagnosis of breast cancer [25, 26]. In the present study, the derived nomograms were validated The C-indices of the OS and BCSS nomograms were statistically significantly higher in both the training and validation cohorts, showing better discrimination ability and improved results as compared with those reported in previous studies [27]. ROC analysis indicated that the novel nomogram presented better sensitivity and specificity with regard to predicting OS and BCSS in older patients with TNBC. In addition, calibration curve and DCA curve evaluations demonstrated good consistency as well as a favorable net clinical benefit with regard to the evaluated model. Several studies have shown that the accuracy of prognostic models can be statistically significantly improved by integrating AJCC staging with other clinical prognostic indicators [28], and T-staging and N-staging each played an important role in our nomogram drawings.

As a pivotal prognostic factor for breast cancer, distant metastasis sites are closely associated with poor survival outcomes. Our results suggest that bone metastasis, lung metastasis, and brain metastasis are independent risk factors that can increase the scope and accuracy of predictive nomograms. Bone metastasis is the most common type of metastasis occurring in breast cancer.

Kono et al. demonstrated that breast cancer patients with single bone metastases exhibited a longer OS than those with other single-organ metastases [29]. Moreover, the median survival time of metachronous patients following bone metastasis was 7.54 years in metachronous patients [30]. Lung metastases occurring in BC have been recorded as the second most commonly occurring metastasis in breast cancer, and are associated with a median survival time of 22 months after treatment [31]. The median survival time of BC patients exhibiting untreated metastasis to the liver is only approximately 4–8 months [31]. As compared with other common cancer types, brain metastasis occurring within breast cancer is rare, and the median associated survival time is only approximately 4 to 6 months [32].

In this study, the hazard ratios for OS and BCSS exhibited an increasing trend from the 70–74 year subgroup to the 75–79 year, 80–84 year, and 85+ year subgroups. Other research groups have reported that BCSS is not as severe as OS in the subgroup of elderly breast cancer patients aged $\geqslant$ 80 years, and that the poor survival prognoses in this subgroup may be due to other causes that are unrelated to BC itself [33]. Namely, competing causes of death are known to increase with age [34].

Regarding treatment factors, we found that surgery, chemotherapy, and radiotherapy were independent prognostic factors for the progression and survival of older women with triple-negative breast cancer. However, chemotherapy did not lead to improved BCSS within our study. As shown in our model, among older TNBC patients, surgical resection of primary tumors was found to statistically significantly prolong both OS and BCSS. Some researchers have shown that the addition of chemotherapy to treatment regimens for older patients with TNBC offers substantial benefits in terms of overall survival [35]. However, chemotherapy was not an independent prognostic factor for BCSS in our study, consistent with the findings of a previous investigation [36]. Studies have indicated that even healthy older women (who are recommended chemotherapy) do not tolerate this treatment as well as younger women, demonstrating an increased likelihood of potential side effects, hospitalizations, and short-term mortality [37, 38]. Studies have shown that older patients with T1-2N0 triple-negative breast cancer treated with postoperative radiotherapy might show better overall survival in studies controlling for different baseline clinical characteristics and employing propensity score matching [39].

Our data indicate that the novel nomograms developed herein have better prognostic value and predictive accuracy as compared with the TNM staging system in terms of evaluations in older patients with TNBC. Therefore, we conclude that the nomograms delineated in this study provide a more accurate individualized survival assessment for elderly patients with triple-negative BC.

4.1 Limitations

This study has several limitations. Firstly, detailed information on systemic treatment was not collected, especially with respect to specific chemotherapy regimens, radiotherapy regimens, targeted therapies, and hormone therapy. Secondly, the SEER database does not provide information on Ki-67, BRCA1/2 gene mutations, or polygenic status, all of which are important genetic factors as well as strong prognostic parameters with regard to BC trajectories [40]. Thirdly, this was a retrospective epidemiologic investigation, and selection bias may therefore be unavoidable. Fourth, we lacked external data validation with another independent large-scale dataset, but performed effective internal verification.

5. Conclusion

Based on a large population recruited from within the SEER database, we established nomograms predicting survival outcomes in elderly patients with triple-negative BC. These validated nomograms provide a visual prognostic assessment for each prognostic factor and will help physicians predict 1-, 3-, and 5-year OS and BCSS in elderly patients with triple-negative BC. Our findings inform future research direction as well as directly informing medical guidelines.

Ethics statement

Since the data were obtained from the SEER database, informed consent and ethical approval were not required.

Funding

This work was supported by the National Natural Science Foundation of China (No. 8217103088).

Author contributions

RF and BL contributed to the study design, data collection, and manuscript writing. XW and WH conceptualized the study and was involved in result interpretation and manuscript writing. RF and DL statistical analyzed the data. JZ, YY and XC contributed with a critical revision of the manuscript. All authors contributed to the article and approved the submitted version.

Data availability statement

Publicly available datasets were analyzed in this study. This data can be found in the Surveillance, Epidemiology, and End Results (SEER) database (https://seer.cancer.gov/).

Footnotes

Acknowledgments

Not applicable.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

References

DeSantis

Fedewa

Goding Sauer

Kramer

Smith

Jemal

. Breast cancer statistics, 2015: Convergence of incidence rates between black and white women. CA Cancer J Clin. 2016 Jan-Feb; 66(1): 31-42. doi: 10.3322/caac.21320. Epub 2015 Oct 29. PMID: 26513636.

Coughlin

Paxton

Moore

Stewart

Anglin

. Survivorship issues in older breast cancer survivors. Breast Cancer Res Treat. 2019 Feb; 174(1): 47-53. doi: 10.1007/s10549-018-05078-8. Epub 2018 Dec 1. PMID: 30506112.

Lodi

Scheer

Reix

Heitz

Carin

Thiébaut

Neuberger

Tomasetto

Mathelin

. Breast cancer in elderly women and altered clinico-pathological characteristics: a systematic review. Breast Cancer Res Treat. 2017 Dec; 166(3): 657-668. doi: 10.1007/s10549-017-4448-5. Epub 2017 Aug 12. PMID: 28803352.

Cabrera-Galeana

Soto-Perez-de-Celis

Reynoso-Noverón

Villarreal-Garza

Arce-Salinas

Matus-Santos

Ramírez-Ugalde

Alvarado-Miranda

Meneses-García

Lara-Medina

Torres-Dominguez

Bargalló-Rocha

Mohar

. Clinical characteristics and outcomes of older women with breast cancer in Mexico. J Geriatr Oncol. 2018 Nov; 9(6): 620-625. doi: 10.1016/j.jgo.2018.04.003. PMID: 29691196.

Biganzoli

Battisti

NML

Wildiers

McCartney

Colloca

Kunkler

Cardoso

Cheung

de Glas

Trimboli

Korc-Grodzicki

Soto-Perez-de-Celis

Ponti

Tsang

Marotti

Benn

Aapro

Brain

EGC

. Updated recommendations regarding the management of older patients with breast cancer: a joint paper from the European Society of Breast Cancer Specialists (EUSOMA) and the International Society of Geriatric Oncology (SIOG). Lancet Oncol. 2021 Jul; 22(7): e327-e340. doi: 10.1016/S1470-2045(20)30741-5. Epub 2021 May 14. PMID: 34000244.

Derks

MGM

van de Velde

CJH

Giardiello

Seynaeve

Putter

Nortier

JWR

Dirix

Bastiaannet

Portielje

JEA

Liefers

. Impact of Comorbidities and Age on Cause-Specific Mortality in Postmenopausal Patients with Breast Cancer. Oncologist. 2019 Jul; 24(7): e467-e474. doi: 10.1634/theoncologist.2018-0010. Epub 2019 Jan 3. PMID: 30606886; PMCID: PMC6656441.

Leone

Freedman

Lin

Tolaney

Vallejo

Leone

Winer

Leone

. Tumor subtypes and survival in male breast cancer. Breast Cancer Res Treat. 2021 Aug; 188(3): 695-702. doi: 10.1007/s10549-021-06182-y. Epub 2021 Mar 26. PMID: 33770314.

Lao

Lawrenson

Edwards

Campbell

. Treatment and survival of Asian women diagnosed with breast cancer in New Zealand. Breast Cancer Res Treat. 2019 Sep; 177(2): 497-505. doi: 10.1007/s10549-019-05310-z. Epub 2019 Jun 5. PMID: 31168758.

Bianchini

Balko

Mayer

Sanders

Gianni

. Triple-negative breast cancer: challenges and opportunities of a heterogeneous disease. Nat Rev Clin Oncol. 2016 Nov; 13(11): 674-690. doi: 10.1038/nrclinonc.2016.66. Epub 2016 May 17. PMID: 27184417; PMCID: PMC5461122.

10.

Zou

Xie

Zheng

Liu

Tang

Tian

Deng

Zhang

Wong

Tan

Liu

Xie

. Leveraging diverse cell-death patterns to predict the prognosis and drug sensitivity of triple-negative breast cancer patients after surgery. Int J Surg. 2022 Nov; 107: 106936. doi: 10.1016/j.ijsu.2022.106936. Epub 2022 Sep 20. PMID: 36341760.

11.

Geyer

Pareja

Weigelt

Rakha

Ellis

Schnitt

Reis-Filho

. The Spectrum of Triple-Negative Breast Disease: High- and Low-Grade Lesions. Am J Pathol. 2017 Oct; 187(10): 2139-2151. doi: 10.1016/j.ajpath.2017.03.016. Epub 2017 Jul 20. PMID: 28736315; PMCID: PMC5809519.

12.

Gan

Liu

Jiang

Xin

Liu

Qin

Cheng

Liu

Zhang

Wang

Zhang

Wang

. Radiomics features based on automatic segmented MRI images: Prognostic biomarkers for triple-negative breast cancer treated with neoadjuvant chemotherapy. Eur J Radiol. 2022 Jan; 146: 110095. doi: 10.1016/j.ejrad.2021.110095. Epub 2021 Dec 4. PMID: 34890936.

13.

Siegel

Miller

Jemal

. Cancer statistics, 2018. CA Cancer J Clin. 2018 Jan; 68(1): 7-30. doi: 10.3322/caac.21442. Epub 2018 Jan 4. PMID: 29313949.

14.

Garrido-Castro

Lin

Polyak

. Insights into Molecular Classifications of Triple-Negative Breast Cancer: Improving Patient Selection for Treatment. Cancer Discov. 2019 Feb; 9(2): 176-198. doi: 10.1158/2159-8290.CD-18-1177. Epub 2019 Jan 24. PMID: 30679171; PMCID: PMC6387871.

15.

Balachandran

Gonen

Smith

DeMatteo

. Nomograms in oncology: more than meets the eye. Lancet Oncol. 2015 Apr; 16(4): e173-80. doi: 10.1016/S1470-2045(14)71116-7. PMID: 25846097; PMCID: PMC4465353.

16.

Zhao

Ding

Qiu

Peng

Pan

Zhang

Shi

Zhou

Luo

. Development and validation of a nomogram for preoperative prediction of cervical lymph node involvement in thyroid microcarcinoma. Aging (Albany NY). 2020 Mar 13; 12(6): 4896-4906. doi: 10.18632/aging.102915. Epub 2020 Mar 13. PMID: 32170046; PMCID: PMC7138557.

17.

Iasonos

Schrag

Raj

Panageas

. How to build and interpret a nomogram for cancer prognosis. J Clin Oncol. 2008 Mar 10; 26(8): 1364-70. doi: 10.1200/JCO.2007.12.9791. PMID: 18323559.

18.

Chen

Liu

Yang

Liu

You

Dong

Lyu

. Development and Validation of a Nomogram for Predicting Survival in Male Patients With Breast Cancer. Front Oncol. 2019 May 14; 9: 361. doi: 10.3389/fonc.2019.00361. PMID: 31139562; PMCID: PMC6527749.

19.

Ripatti

Palmgren

. Estimation of multivariate frailty models using penalized partial likelihood. Biometrics. 2000 Dec; 56(4): 1016-22. doi: 10.1111/j.0006-341x.2000.01016.x. PMID: 11129456.

20.

Harrell

Jr. Lee

Mark

. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996 Feb 28; 15(4): 361-87. doi: 10.1002/(SICI)1097-0258(19960229)15:4<361::AID-SIM168>3.0.CO;2-4. PMID: 8668867.

21.

Camp

Dolled-Filhart

Rimm

. X-tile: a new bio-informatics tool for biomarker assessment and outcome-based cut-point optimization. Clin Cancer Res. 2004 Nov 1; 10(21): 7252-9. doi: 10.1158/1078-0432.CCR-04-0713. PMID: 15534099.

22.

Wang

Xia

Gong

Wang

Yan

Wan

Liu

Shi

Lau

Shen

. Prognostic nomogram for intrahepatic cholangiocarcinoma after partial hepatectomy. J Clin Oncol. 2013 Mar 20; 31(9): 1188-95. doi: 10.1200/JCO.2012.41.5984. Epub 2013 Jan 28. PMID: 23358969.

23.

Pan

Yang

Chu

Zhang

Tao

Chen

Lyu

. Nomograms for predicting long-term overall survival and cancer-specific survival in lip squamous cell carcinoma: A population-based study. Cancer Med. 2019 Jul; 8(8): 4032-4042. doi: 10.1002/cam4.2260. Epub 2019 May 21. PMID: 31112373; PMCID: PMC6639254.

24.

Pan

You

Feng

Zhao

Lyu

. Development and validation of a nomogram for predicting cancer-specific survival in patients with Wilms’ tumor. J Cancer. 2019 Aug 28; 10(21): 5299-5305. doi: 10.7150/jca.32741. PMID: 31602280; PMCID: PMC6775601.

25.

Zhao

Hua

Geng

Guo

Suo

Zhou

Wang

. A Shortcut Weighted Fusion Pyramid Network for Microcalcification Detection in Breast Mammograms. Technol Health Care. 2023; 31(3): 841-53. doi: 10.3233/THC-220235. PMID: 36442221.

26.

Wang

Zou

Zeng

Song

Zhou

Ding

Ming

. Combination of Variational Mode Decomposition and Coherent Factor for Ultrasound Computer Tomography. Technol Health Care. 2022; 30(S1): 163-71. doi: 10.3233/THC-228016. PMID: 35124594; PMCID: PMC9028653.

27.

Liu

Yue

Huang

Duan

Zhao

Liu

Xiang

. Risk Stratification Model for Predicting the Overall Survival of Elderly Triple-Negative Breast Cancer Patients: A Population-Based Study. Front Med (Lausanne). 2021 Sep 21; 8: 705515. doi: 10.3389/fmed.2021.705515. PMID: 34621757; PMCID: PMC8490672.

28.

Huang

Liang

Tian

Liang

Chen

Liu

. Development and Validation of a Radiomics Nomogram for Preoperative Prediction of Lymph Node Metastasis in Colorectal Cancer. J Clin Oncol. 2016 Jun 20; 34(18): 2157-64. doi: 10.1200/JCO.2015.65.9128. Epub 2016 May 2. Erratum in: J Clin Oncol. 2016 Jul 10;34(20):2436. PMID: 27138577.

29.

Kono

Fujii

Matsuda

Harano

Chen

Wathoo

Joon

Tripathy

Meric-Bernstam

Ueno

. Somatic mutations, clinicopathologic characteristics, and survival in patients with untreated breast cancer with bone-only and non-bone sites of first metastasis. J Cancer. 2018 Sep 8; 9(19): 3640-3646. doi: 10.7150/jca.26825. PMID: 30310523; PMCID: PMC6171013.

30.

Parkes

Warneke

Clifton

Al-Awadhi

Oke

Pestana

Alhalabi

Litton

Hortobagyi

. Prognostic Factors in Patients with Metastatic Breast Cancer with Bone-Only Metastases. Oncologist. 2018 Nov; 23(11): 1282-1288. doi: 10.1634/theoncologist.2018-0085. Epub 2018 Aug 17. PMID: 30120166; PMCID: PMC6291319.

31.

Smid

Wang

Zhang

Sieuwerts

Klijn

Foekens

Martens

. Subtypes of breast cancer show preferential site of relapse. Cancer Res. 2008 May 1; 68(9): 3108-14. doi: 10.1158/0008-5472.CAN-07-5644. PMID: 18451135.

32.

Brufsky

Mayer

Rugo

Kaufman

Tan-Chiu

Tripathy

Tudor

Wang

Brammer

Shing

Yood

Yardley

. Central nervous system metastases in patients with HER2-positive metastatic breast cancer: incidence, treatment, and survival in patients from registHER. Clin Cancer Res. 2011 Jul 15; 17(14): 4834-43. doi: 10.1158/1078-0432.CCR-10-2962. PMID: 21768129.

33.

Hwang

Jung

Hyeon

Park

Ahn

Nam

Kim

Lee

Choi

Cho

. A nomogram to predict pathologic complete response (pCR) and the value of tumor-infiltrating lymphocytes (TILs) for prediction of response to neoadjuvant chemotherapy (NAC) in breast cancer patients. Breast Cancer Res Treat. 2019 Jan; 173(2): 255-266. doi: 10.1007/s10549-018-4981-x. Epub 2018 Oct 15. PMID: 30324273.

34.

Sparano

Gray

Makower

Pritchard

Albain

Hayes

Geyer

Jr. Dees

Goetz

Olson

Jr. Lively

Badve

Saphner

Wagner

Whelan

Ellis

Paik

Wood

Ravdin

Keane

Gomez Moreno

Reddy

Goggins

Mayer

Brufsky

Toppmeyer

Kaklamani

Berenberg

Abrams

Sledge

Jr., Adjuvant Chemotherapy Guided by a 21-Gene Expression Assay in Breast Cancer. N Engl J Med. 2018 Jul 12; 379(2): 111-121. doi: 10.1056/NEJMoa1804710. Epub 2018 Jun 3. PMID: 29860917; PMCID: PMC6172658.

35.

Crozier

Pezzi

Hodge

Janeva

Lesnikoski

Samiian

Devereaux

Hammond

Audisio

Pezzi

. Addition of chemotherapy to local therapy in women aged 70 years or older with triple-negative breast cancer: a propensity-matched analysis. Lancet Oncol. 2020 Dec; 21(12): 1611-1619. doi: 10.1016/S1470-2045(20)30538-6. PMID: 33271091.

36.

Lan

Zheng

Shao

Luo

Yang

Wang

. The Role of Adjuvant Chemotherapy in Metaplastic Breast Carcinoma: A Competing Risk Analysis of the SEER Database. Front Oncol. 2021 Apr 26; 11: 572230. doi: 10.3389/fonc.2021.572230. PMID: 33981594; PMCID: PMC8107469.

37.

Rosenstock

Lei

Tripathy

Hortobagyi

Giordano

Chavez-MacGregor

. Short-term mortality in older patients treated with adjuvant chemotherapy for early-stage breast cancer. Breast Cancer Res Treat. 2016 Jun; 157(2): 339-350. doi: 10.1007/s10549-016-3815-y. Epub 2016 May 4. PMID: 27146586.

38.

Barcenas

Niu

Zhang

Buchholz

Elting

Hortobagyi

Smith

Giordano

. Risk of hospitalization according to chemotherapy regimen in early-stage breast cancer. J Clin Oncol. 2014 Jul 1; 32(19): 2010-7. doi: 10.1200/JCO.2013.49.3676. Epub 2014 May 27. PMID: 24868022; PMCID: PMC4164758.

39.

Haque

Verma

Hsiao

Hatch

Arentz

Szeja

Schwartz

Niravath

Bonefas

Miltenburg

Brian Butler

Teh

. Omission of radiation therapy following breast conservation in older (=70 years) women with T1-2N0 triple-negative breast cancer. Breast J. 2019 Nov; 25(6): 1126-1133. doi: 10.1111/tbj.13443. Epub 2019 Jul 4. PMID: 31273872.

40.

Giuliano

Connolly

Edge

Mittendorf

Rugo

Solin

Weaver

Winchester

Hortobagyi

. Breast Cancer-Major changes in the American Joint Committee on Cancer eighth edition cancer staging manual. CA Cancer J Clin. 2017 Jul 8; 67(4): 290-303. doi: 10.3322/caac.21393. Epub 2017 Mar 14.

Nomograms predict survival in elderly women with triple-negative breast cancer: A SEER population-based study

Abstract

BACKGROUND:

OBJECTIVE:

METHODS:

RESULTS:

CONCLUSION:

Keywords

1. Introduction

2. Materials and methods

2.1 Data collection

2.2 Cohort selection

3. Results

3.1 Clinicopathological characteristics in the training and validation cohorts

Table 1 The demographic and baseline characteristics of the Cohort with 70 years or older with triple-negative breast cancer

3.3 Calibration and validation of the prognostic nomograms

4. Discussion

4.1 Limitations

5. Conclusion

Ethics statement

Funding

Author contributions

Data availability statement

Footnotes

Acknowledgments

Conflict of interest

References

Table 1
The demographic and baseline characteristics of the Cohort with 70 years or older with triple-negative breast cancer