Factors associated with the onset and survival of subsequent primary breast cancer in female non‐metastatic breast cancer survivors

This study aimed to investigate the risk factors for the onset of subsequent primary breast cancer (SPBC) in women with a previous diagnosis of early‐stage breast cancer (BC) and to construct a prognostic prediction model for patients with SPBC. Using the Surveillance, Epidemiology, and End Results‐17 (SEER‐17) database, we conducted a retrospective cohort analysis on women with initial primary early‐stage BC from 2004 to 2015. Standardized incidence ratio (SIR) was calculated to determine the risk of subsequent primary cancer (SPC). A competing risk model was built to identify the risk factors for the onset of SPBC. And risk factors associated with breast cancer‐specific mortality in SPBC patients were evaluated and presented in the form of nomogram. Compared with the general population, the overall risk of SPC for all sites was significantly elevated in women with early‐stage BC (SIR = 1.21, 95% CI: 1.20–1.23), and breast is the most frequent site. Age, race and ethnicity, year of diagnosis, history of other tumors, histological type, surgery, radiation, chemotherapy, tumor size, positive lymph nodes numbers and ER status were independent risk factors (p < .05) for the onset of SPBC. A new prognosis nomogram demonstrated good discrimination after internal validation with a C‐index of 0.869 (95% CI: 0.859–0.880), and showed favorable consistency and clinical usefulness. The incidence of SPBC and prognosis of patients with SPBC were well estimated based on a large cohort. Our nomogram model had excellent prediction performance and could be a useful tool to predict prognosis.

disease-specific survival rate has been reported especially for early-stage BC patients. 4,5provement in the curability for BC raises the importance of care for subsequent disease, especially subsequent primary cancer (SPC), [6][7][8] and female breast is the most frequent site, which accounts for approximately 30% of all subsequent cancers. 9,10The increased risk might be attributed to behavioral factors, long-term side effects of treatment regimens, and the presence of an inherited genetic predisposition, such as BRCA1/2 mutation, PTEN mutation, and Cowden disease. 11However, the risk factors for subsequent primary breast cancer (SPBC) in BC survivors and the associated survival are poorly known.[14][15] In addition, previous studies have confirmed that developing SPC was a poor prognostic factor for patients with BC 10,16,17 and SPC undoubtedly affects the quality of life of BC survivors.However, the current guidelines for the management of BC patients focus primarily on metastasis and recurrence of the initial primary cancer, leading to inevitable neglect for other sites of developing subsequent primary cancers.This highlights a crucial need for addressing these research gaps, and BC survivors should be advised of their increased risk for developing a certain cancer in their lifetime.
Therefore, the aims of our study were to systemically identify risk factors for SPBC in a large cohort of women with early-stage BC.And for clinical convenience, we develop and validate a competing-risk nomogram 18 that integrates the clinicopathological risk factors of the two primary cancers for the individual survival prediction of earlystage BC survivors who were diagnosed with SPBC.

| Data sources
The Surveillance, Epidemiology, and End Results (SEER) database (http://seer.cancer.gov)collects cancer incidence data from population-based cancer registries covering approximately 47.9 percent of the U.S. population. 19We gained access to the SEER database via a standard request to SEER, and used SEER*Stat software (http:// seer.cancer.gov/seerstat) to access the SEER 17 registries (2021 submission).

| Study population
Patients included in this study were women aged 18 years and older with stages I-III BC documented in SEER as their first primary cancer diagnosis between January 1, 2004, and December 31, 2015.This calendar year interval was chosen due to unified staging system and adequate follow-up time.And the follow-up time began with baseline and ended with diagnosis of SPBC, last contact or death without developing SPBC or end of the study period (December 31, 2019), whichever occurred first.
The exclusion criteria were as follows: (1) patients diagnosed by autopsy only or death certificate only; (2) age at diagnosis was older than 70 years old; (3) unknown causes of deaths or patients died from accidents; (4) follow-up less than 6 months, in order to ensure the therapeutic effect; (5) patients with unknown variables, or variables without explicit values.Data were de-identified and publicly available under a data use agreement with the US National Cancer Institute and informed patient consent was not required to access and use SEER data. 20

| Outcome measures
Patients with SPBC required a minimum two-month period after the primary diagnosis because multiple primary tumors are difficult to distinguish from metastases or recurrences during this period, and the current study uses the SEER definition to define multiple primary tumors. 11And the patients in our study were divided into three groups according to their outcomes at the end of follow-up: (1) censored observations (also called "alive group"): patients who were alive till the last follow-up or lost to follow-up without developing SPBC; (2) SPBC group: patients who experienced SPBC followed the SEER definition rules; (3) mortality group: patients who were dead without evidence of SPBC during the follow-up time.

| Covariates
Analyses were conducted for pre-specified subgroups based on clinically meaningful characteristics.

| Clinical characteristics
SEER data were also used to obtain information on clinical characteristics, including AJCC stage (American Joint Committee on Cancer (AJCC) 6th edition), estrogen receptor (ER), progesterone receptor (PR) and tumor human epidermal growth factor receptor 2 (HER2) status (categorized as negative and positive), tumor size, primary site, laterality, pathological grade (categorized as I, II and III/IV), histological type (categorized as infiltrating duct carcinoma, lobular carcinoma, infiltrating duct and lobular carcinoma, infiltrating duct mix with other types of carcinoma and other types), T A B L E 1 Characteristics of all included female patients with early-stage BC. treatment forms and positive lymph node number (categorized as 0, 1-3, 4-9, ≥10 and non-detected).
The forms of treatment were surgery type, chemotherapy, and radiotherapy.Patients included in this study were classified as nonsurgical group and surgical group, who required to receive cancerdirected surgery documented in SEER, classified as breast-conserving surgery or mastectomy.
In addition, in identifying risk factors associated with the survival of patients with SPBC, two additional covariates have been considered.They are respectively time interval from diagnosis and ipsilateral breast.

| Statistical analysis
Firstly, the distributions of all baseline characteristics were summarized by calculating the median and range for continuous variables and frequencies for categorical variables, respectively.
Secondly, the relative risk of SPC among early-stage BC was compared to the general population and presented as the standardized incidence ratio (SIR).The SIR and corresponding 95% confidence interval (CI) were calculated using the MP-SIR module in SEER*stat software.And cumulative mortality functions were calculated and plotted to show the cumulative incidence 21 of different outcomes in all included BC patients, and a hazard curve was also plotted to reveal the changing risk of developing SPBC.
And then, Fine-Gray's competing-risk regression analysis [22][23][24] was used to identify the risk factors for SPBC in BC patients.
Developing SPBC was treated as the event of interest, death from any causes before developing SPBC was considered as a competing event, and the sub-distribution hazard ratio (SHR) with the corresponding 95% CI was acquired.Variables with p < .05 in the univariate analyses were selected for the multivariate analysis.
The BC patients were then divided into the following three    Non-cancer death was considered as competing risk event for cancer-specific death.And for clinical use, a nomogram for predicting BCSM was conducted. 18,25del performance was assessed by internal validation, which was performed by discrimination and calibration by the bootstrap resampling method. 26The discrimination of the nomogram was measured by the concordance index (C-index) and the time-dependent receiver operating characteristic (ROC) curve with 1000 bootstrap resamples.The calibration was graphically assessed with calibration curves (1000 bootstrap resamples) to determine the degree of coincidence between the actual probability and the predicted probability.
A decision curve analysis was also performed to determine the clinical usefulness of the nomogram by calculating the net benefits at different threshold probabilities. 27,28l statistical analyses mentioned above were performed using R software (version 4.1.2).A two-sided p < .05 was considered statistically significant.

| Cause-specific cumulative incidence function over time
The cause-specific cumulative incidence among BC patients included in this study is illustrated in Figure 2A.By far the highest cumulative incidence was caused by mortality, followed by SPBC.The SPBC-specific hazard curve is shown in Figure 2B, which indicated that the incidence of SPBC is not constant.The steepest decline of the hazard rate was seen within the first 14 months, and then rose progressively and changed dynamically over time.

| Risk factors for the onset of SPBC
As shown in

| Competing regression analysis of prognostic factors for BCSM in patients with SPBC
Survival curves (Figure 3) were analyzed by log-rank test and a statistically significant difference between survival curves was found (p < .001).Patients in the BC-only group had better overall survival (OS) than those in the other two groups, which indicated that patients with SPBC had a poor prognosis.
To further investigate the relationship between BCSM and the prognostic factors, a new Fine-Gray's competing risk model was constructed.Apart from the initial primary cancer-related information, the details about SPBC were also taken into consideration.The baseline characteristics of patients with SPBC are shown in Table S1 and the results of univariate and multivariate analyses are shown in Table 4.
In the univariate analysis, exception for radiation of SPBC, all other factors had a risk for BCSM with statistically significant.
Multivariate analysis was then performed including the variables which revealed statistically significant difference in the univariate analysis.
In detail, the multivariate analysis result indicated that longer time F I G U R E 3 Overall survival curves for patients included in the study.The differences in overall survival rates were statistically significant between the three groups ( p < .001).
T A B L E 4 Univariate and multivariate Fine-Gray's competing risk model analysis for BCSM.

| Development of the competing-risk nomogram
As shown in Table 4,

| DISCUSSION
With the improvement in diagnostic techniques, advances in hormone therapy and development of systemic treatment, survival after BC diagnosis has improved throughout the world. 29With prolonged survival, there is an increased likelihood that patients will receive a diagnosis of SPC as a result of underlying genetic or other risk factors related to breast cancer and treatment.
In our analysis, female patients with early-stage BC were at higher risk for a new cancer than in the general population, and the SPBC accounted for a large proportion, which were consistent with previous studies. 7,8,30A possible explanation for the increase is that these patients are under more intensive surveillance after treatment.
5][36][37] In addition, most of the existing studies have focused on the SPC effect of the survival, 17,38 but failed to evaluate associated survival in BC survivors with SPBC and few studies considered subsequent BC diagnoses as primary outcome indicator.
To our knowledge, it is the first study to treat subsequent BC diagnoses as the event of interest, and explore the risk factors associated with the onset of SPBC.And the established competing-risk nomogram is also the first known to incorporate multiple clinically meaningful factors simultaneously, and for specific prediction on the prognosis of early-stage BC patients with SPBC.As the actual incidence of SPBC in the elderly is higher than reported, which may make the result of our study an underestimate, 39 we excluded women whose first BC was diagnosed after the age of 70 years. 40,41Besides, we mainly focus on patients without metastasis at the initial diagnosis, with consideration of low long-term survival and complicated prognostic factors of metastatic (stage IV) BC.
Overall, several notable findings deserve mention in the current study.Firstly, our study found that young age of the patient at the initial diagnosis is a prominent risk factor for developing SPBC, which was in line with previous studies. 42,43Young age at diagnosis is a common feature of cancer in individuals with a genetic susceptibility, 44 and it is also known that for individuals with deleterious mutations in BRCA1 and /or BRCA2, there is an elevated risk of developing additional BC. 45[48] Previous studies had confirmed that radiotherapy was positively associated with the incidence of SPC, which might be attributed to the  antitumor immune responses. 49The possible reasons may explain this unexpected result include the following: in general, the SEER regis- However, there are still some limitations to this study.(1) Some metastases or recurrences may be mistaken as SPC, even though SEER registry established strict rules for documenting multiple primary tumors.(2) Our study suffered from lack of specific information on chemotherapy and hormone treatment, and thus, we could not perform the analysis on different regimens.(3) Some potential risk factors were not available, such as body mass index, family history, and genetic mutation. 49Consequently, further investigation is warranted to assess the risks based on detailed information on treatment, gene expression and other confounders.

| CONCLUSIONS
In conclusion, our study found that female patients with early-stage BC had an increased risk of developing SPBC.And a convenient nomogram was developed to offer prognostic assessment for those patients with SPBC.Additional studies are warranted to validate our findings and unravel the underlying mechanisms of multiple primary cancers.
subgroups: BC only (BC only group), subsequent BC after initial BC (SPBC group) and all residual patients were aggregated in the other group.Kaplan-Meier survival curves were used to compare survival difference among subgroups and analyzed using the log-rank test.To further investigate the relationship between T A B L E 2 Risk of SPC compared with the US general population.

F I G U R E 1
The observed events numbers and SIR for top 10 SPC sites.F I G U R E 2 Cause-specific cumulative incidence function over time.(A) The cumulative function of SPBC, and Death in early-stage BC patients.(B) The SPBC-specific hazard curve in early-stage BC patients.
interval from the initial BC diagnosis may benefit survival.And patients with poor histological grade had a higher risk of BCSM (in terms of histological grade of IPBC, grade III/IV versus grade I: SHR = 1.468 (1.125-1.915), in terms of histological grade of SPBC, SHR (95% CI) as high as 1.347 (1.058-1.715)and 2.119 (1.594-2.817)relative to grade I group).Moreover, patients who were diagnosed with lobular carcinoma ) = 1.415 (1.070-1.871)]relative to those diagnosed with infiltrating duct carcinoma.And regarding the second tumor, patients in the surgery group and chemotherapy group had lower mortality risk than patients in the no surgery and no chemotherapy group ( p < .05).And the larger the tumor size or the larger the number of positive lymph nodes, the greater the likelihood of BCSM ( p < .05).And we also further evaluated the association between HER2 status and BCSM using data from January 1, 2010, onward.It was observed that BCSM was associated with HER2 receptor status of the second tumor [HER2-positive vs HER2-negative: SHR = 0.482 (0.274-0.848)], but was not associated with the initial tumor ( p = .21).
13 factors were statistically significant in multivariate analysis, and a competing-risk nomogram was constructed based on the 13 factors (Figure4A), and the calculation formula for the nomogram is presented in Note S1.The time-dependent ROC curve of the nomogram to predict BCSM at 3-year, 5-year and 10-year in the cohort is presented in Figure4B, with an area under the receiver operating characteristic curve (AUROC) of 0.924 (95% CI: 0.898-0.950),0.929 (95% CI: 0.915-0.944),and 0.895 (95% CI: 0.882-0.907),respectively.The nomogram was validated with 1000 bootstrap resamples to calculate a robust C-index, which yielded an averaged C-index of 0.869 (95% CI: 0.859-0.880)(FigureS1).The calibration curves (1000 bootstrap resamples) also showed good agreement among the estimations with the nomogram and actual observations (Figure4C).DCA was performed to evaluate the clinical net benefit using the competing-risk nomogram.Figure5compares the net benefit of the predictive model to those in two hypothetic scenarios.Within a considerable range, the clinical net benefit of the competing-risk nomogram was larger than that under the assumption of not using the model or screening all patients, which indicated that the competing-risk nomogram is clinically useful.

F
I G U R E 4 Competing-risk nomogram and the corresponding discrimination and calibration curves.(A) Nomogram to predict probability of BCSM at 3, 5, and 10 years.(B) The 3-year, 5-year, and 10-year time-dependent ROC curves of the competing-risk nomogram.(C) Calibration curves of the competing-risk nomogram.
tries collect information on radiotherapy and chemotherapy given as part of the first course of treatment only, and certain types of treatment data are incomplete.Besides, unlike clinical trials, many factors involved in determining the course of treatment were not be captured in the registry data, such factors include: patient preferences, physician recommendations, comorbidities, and proximity to treatment providers.And addressing the comparative the efficacy or effectiveness of treatment without controlling for the factors is also not recommended by the SEER database.50Furthermore, other factors such as histological type, tumor size, positive lymph nodes numbers and ER status were associated with developing SPBC and affected the survival of patients with SPBC ( p < .05),which was consistent with conventional wisdom that poor clinical features of BC were associated with poor clinical outcomes.Consequently, early-stage BC survivors should be advised of their higher-than-average risk for a subsequent cancer and there is also a growing need to promote effective cancer screening along with healthy life-styles accordingly.Moreover, the current guidelines pay insufficient attention to BC survivors with SPBC, no clear standardized follow-up plan is developed for SPBC, and these patients may experience greater psychological distress.With the help of our nomogram, a more accurate survival prediction was provided for informed decision-making based on patients' individual characteristics.
Univariate and multivariate Fine-Gray's competing risk model analysis for SPBC.
T A B L E 3 -conserving surgery, 146 393 (51.30%) patients were treated with chemotherapy and 165 498 (58.00%) were treated with radiation, showing that nearly half of the patients received the cancer-directed treatment.3.2 | Risk of SPC at top 10 sites compared with the general populationAs shown in Figure1and Table2, the overall risk of SPC for all sites A total of 285 349 early-stage BC patients from the SEER database were included in the present study.Median duration of follow-up was 98 months (range 66-137 months).The detailed clinicopathological characteristics are summarized in Table 1.Regarding age, most of the patients were over 35 years old, which was relatively balanced.Within the histological type, infiltrating duct carcinoma (n = 221 869, 77.75%) was dominant.During initial treatment, 166 453 (58.33%) patients underwent T A B L E 3 (Continued) Abbreviations: BC, breast cancer; BCT, breast-conserving therapy; ER, estrogen receptor; IDALC, infiltrating duct and lobular carcinoma; IDC, infiltrating duct carcinoma; IDMO, infiltrating duct mix with other types of carcinomas; LC, lobular carcinoma; ME, mastectomy; PR, progesterone receptor; SPBC, subsequent primary breast cancer.aUnmarriedpatients include single (never married); Married patients include married (including common law) and unmarried or domestic partner; Others include divorced, widowed and separated.b Other primary sites include C50.0-Nipple, C50.6-Axillary tail of breast, C50.8-Overlapping lesion of breast, C50.9-Breast, NOS.breast

Table 3
, in the univariate analysis, age, marital status, race and ethnicity, year of diagnosis, history of other tumors, histological