Prognostic value and mode selection of locoregional treatment in Stage-IV breast cancer patients

This study aimed to assess the actual prognostic significance of different locoregional treatment (LRT) (surgery and radiotherapy) modalities for stage-IV breast cancer (BC) patients and construct a competing risk nomogram to make precise predictions of the breast cancer-specific death (BCSD) risk among LRT recipients. A total of 9279 eligible stage-IV BC patients from the Surveillance Epidemiology and End Results (SEER) database were included in this study. Initially, we evaluated the impact of LRT on survival both before and after the propensity score matching (PSM). Then, we used the Cox hazard proportional model and competing risk model to identify the independent prognostic factors for LRT recipients. Based on the screened variables, a comprehensive nomogram was established. Kaplan–Meier curves demonstrated that LRT significantly prolonged overall survival (OS) and breast cancer-specific survival (BCSS) (P < 0.001). In addition, patients treated with surgery combined with postoperative radiotherapy (PORT) possessed the optimal survival (P < 0.001). Regardless of the surgical modalities, primary tumor resection combined with radiotherapy could ameliorate the prognosis (P < 0.05). Subgroup analysis showed that in patients with T2-T4 stage, PORT had a survival benefit compared with those undergoing surgery combined with preoperative radiotherapy (PRRT) and surgery only. Based on the screened independent prognostic factors, we established a comprehensive nomogram to forecast BCSD in 1 year, 2 years and 3 years, which showed robust predictive ability. PORT was associated with a lower BCSD in stage-IV BC patients. The practical nomogram could provide a precise prediction of BCSD for LRT recipients, which was meaningful for patients’ individualized management.


Introduction
In 2020, the morbidity of breast cancer (BC) has surpassed lung cancer as the leading cause of malignancies worldwide (Li et al. 2019;Sung et al. 2021).Besides, the mortality of BC has ranked second among cancer-related deaths in females and brought a heavy burden to the society.Although the public awareness of BC screening has gradually increased, there are still 5-10% patients with distant metastases at initial diagnoses (DeSantis et al. 2016).Among stage-IV BC patients, the 5-year survival probability is about 20%, and only less than 10% patients can achieve long-term survival (Gobbini et al. 2018).However, recent research demonstrated that the prognosis of stage-IV BC patients has significantly improved in the past decade (DeSantis et al. 2016;Gradishar et al. 2020;Lee et al. 2020).Presently, systemic treatment remains the core therapeutics Hanxiao Cui and Xueting Ren have contributed equally to this work and share first authorship.
for stage-IV BC patients, including chemotherapy, endocrine and targeted therapy.The main purpose of treatment is to relieve locoregional symptoms, improve life quality and prolong survival (Lane et al. 2019;Loeven and Brown 2021).
Locoregional treatment (LRT), containing primary tumor surgery and (or) radiotherapy, was once supposed to mitigate local symptoms such as bleeding and pain, without survival benefits (Teshome 2018).However, several retrospective studies demonstrated that primary tumor resection for appropriate patients could significantly improve survival (Arciero et al. 2019;Asaad et al. 2021;Lin et al. 2020;Thomas et al. 2016).Additionally, advanced breast cancer patients who received postoperative radiotherapy possessed better survival outcomes (Kim et al. 2020(Kim et al. , 2021;;Zhang et al. 2021).Nevertheless, controversial results were observed in four prospective trials (Badwe et al. 2015;Fitzal et al. 2019;Soran et al. 2018; Ulaş Kahya et al. 2022).In the Turkey's MF07-01 trial, a notable 5-year overall survival improvement was found in patients with LRT (Soran et al. 2018).However, LRT failed to ameliorate prognosis in the other three trials (Badwe et al. 2015;Fitzal et al. 2019;Ulaş Kahya et al. 2022).Due to the advances in systemic therapeutics, the clinical utility of LRT needs to be reassessed.
In this study, we comprehensively analyzed the impact of different LRT modalities on survival in stage-IV BC patients using the data from Surveillance, Epidemiology and End Results (SEER) database.Furthermore, we constructed a competing risk nomogram to make precise prediction of survival in LRT recipients.

Data collection
This study included primary female stage-IV BC patients diagnosed from 2010 to 2017.The exclusion criteria were listed as follows: (1) patients without histological confirmation, (2) patients with unknown status (race, grade, tumor size, lymph nodes, marital status, subtype, metastatic status and radiotherapy), (3) patients diagnosed by death certificate or autopsy, (4) patients lacking complete survival data.Ultimately, a total of 9279 patients were enrolled in this study.Age, race, marital status, grade, tumor size (T), lymph nodes (N), BC subtype, metastatic pattern, surgery data, chemotherapy and radiotherapy were selected as variables.

Ethics statement
The data of this study were downloaded from the SEER database (SEER*Stat version 8.3.9), which was one of the largest publicly available cancer databases in the America.As there was no private information involved, this study did not require informed consent.The Ethical Board of The Second Affiliated Hospital of Xi'an Jiaotong University approved this study.

Statistical analysis
Age was a continuous variable.According to the American incidence and mortality data of BC, we adopted appropriate age cutoffs and divided the patients into four categories (< = 40, 41-55, 56-70 and > 70).Descriptive statistics and Chi-square test were used to analyze distribution of clinicopathological characteristics and test statistical significance.In addition, we conducted a 1:1 propensity score matching (PSM) to reduce differences of baseline characteristics between groups.The primary endpoints of this study were overall survival (OS), breast cancer-specific survival (BCSS) and breast cancer-specific death (BCSD).Kaplan-Meier analysis and log-rank test were performed to compare the survival variations in different locoregional treatment modalities.Considering the impact of non-breast cancerspecific events (NBCSD), we simultaneously constructed Cox hazard proportional model and competing risk model to identify the optimal independent prognostic factors.Based on the screened variables, we established a competing risk nomogram to predict 1-, 2-and 3-year BCSD of LRT recipients.Calibration curves and time-dependent ROC curves were plotted to evaluate the consistency and accuracy of the model.All statistical analyses were completed by R software 4.0.3, and a two-sided P < 0.05 was deemed as statistically significant.

Baseline characteristics
A total of 9279 stage-IV BC patients were eligible for this study, of which 5151 (55.51%) patients received LRT, while the remaining 4128 (44.49%) patients did not.Through the analysis of the demographic data, we observed that although the number of patients undergoing LRT has decreased slightly, it still accounts for a considerable proportion (Fig. 1).In this study, significant distribution differences were found in age, race, marital status, grade, T, N, subtype, metastatic status and chemotherapy between LRT and non-LRT groups (P < 0.05) (Table 1).Patients treated with LRT tended to be younger, have higher histological grade, smaller tumor size, more advanced lymph nodes status, more aggressive subtypes and higher ratio of bone-only metastasis and receiving chemotherapy.Considering the potential bias caused by unbalanced baseline distribution, we included all the relevant variables for 1:1 PSM analysis.After PSM, the LRT and non-LRT groups comprised 3632 patients, respectively.No statistical differences were observed in distribution characteristics between the two groups (Table 1).
Among the 5151 patients who received LRT, 3036 (58.94%) died in this retrospective study.The cumulative incidence of BCSD was 53.80% (2771/5151), while the incidence of death from other causes was 5.14% (265/5151).To eliminate the impact of competing events, we constructed a competing risk model.In the univariate analysis, patients who underwent PORT possessed relatively lower BCSD, which was consistent with the results of Cox regression (P < 0.001) (Table 3).Additionally, we observed that ten variables (age, race, marital status, grade, T, N, subtype, metastatic status, chemotherapy and LRT modalities) remained independent prognostic factors in the competing risk multivariate analysis (Fig. 6).

Establishment of a nomogram using competing risk model
In a 7:3 ratio, we randomly divided the patients into the training and internal validation sets.Based on the above screened indices, we established a competing risk nomogram to predict the BCSD at 1-, 2-and 3-year intervals (Fig. 7).For a given patient, each variable corresponded to a specific score, and the sum of the scores was compared with the probability scales to forecast 1-, 2-, and 3-year BCSD (1-year = 8.39%, 2-year = 17.50%, 3-year = 26.60%).High consistency was demonstrated between the predicted and practical BCSD using the calibration curves (Fig. 8A-B).Furthermore, the areas under the ROC curves were in the range of 0.75-0.80whether in the training or in the internal testing set (training: 1-year = 0.779, 2-year = 0.780, 3-year = 0.766; testing: 1-year = 0.790, 2-year = 0.792, 3-year = 0.757), showing the accuracy of the model (Fig. 9A-B).

Discussion
Using data from the SEER database, we focused on stage-IV BC patients to evaluate the impact of different locoregional treatment modalities on prognosis.Competing risk nomogram was established for accurate predictions of the BCSD in LRT recipients.Previous studies have explored whether primary tumor resection or postoperative radiotherapy could bring survival benefits to stage-IV BC patients, but there still lacked studies that comprehensively analyzed different LRT modalities and sequence (Arciero et al. 2019;Asaad et al. 2021;Kim et al. 2020Kim et al. , 2021;;Lin et al. 2020;Thomas et al. 2016;Zhang et al. 2021).Besides, this is the first large-scale retrospective study to predict the impact pf distinct LRT modality on survival based on a competing risk analysis.Through the analysis of the demographic data, we observed that between 2010 and 2017, 55.51% of the stage-IV BC patients received LRT.Although the number of patients undergoing LRT has decreased slightly, it still accounts for a considerable proportion.The possible reason for this decline may be the remarkable advances in systemic therapy.
In this study, we found that LRT could significantly prolong the survival of stage-IV BC patients.Similar results have been demonstrated in some retrospective studies (Choi et al. 2018;Pons-Tostivint et al. 2019).A multi-center research showed that LRT was related to improved OS in de novo metastatic BC patients, including those with visceral invasion at diagnosis (HR: 0.65, 95%CI 0.55-0.76),but no survival benefit was observed in triple negative BC patients (Pons-Tostivint et al. 2019).One SEER-based study found that postoperative radiotherapy was linked to survival advantages in de novo stage-IV BC patients (Zhang et al. 2021).A meta-analysis of 30 observational research showed that surgery could notably ameliorate the overall survival of metastatic BC patients (Xiao et al. 2018).The improved survival could be attribute to the following factors.Firstly,  illi et al. 2004;Danna et al. 2004).Secondly, radiotherapy mainly kills cancer cells through two ways, that is highenergy radioactive rays directly induces DNA breakage or induces the production of reactive oxygen in cancer cells (Durante and Formenti 2018;Vanpouille-Box et al. 2017;Yamazaki et al. 2020;Zhang et al. 2022).It reduces tumor burden by killing tumor cells in situ and regulating immune microenvironment though 'Radscopal effect' (Klug et al. 2013;Zhang et al. 2022).In spite of the support of retrospective studies, controversial results were observed in prospective studies.MF07-01 was the only prospective trial to prove the survival benefit of LRT, in which the median survival time of the surgical group was nine months longer than that of the non-surgical group (P = 0.005) (Soran et al. 2018).However, in the Austrian ABCSG-28/POSYTIVE cohort, the 3-year survival probability of 90 patients in the LRT and non-LRT groups was similar (Fitzal et al. 2019).The American EA2108 study demonstrated that early LRT significantly improved locoregional control (P < 0.001), whereas no prolonged OS (P = 0.570) and elevated life quality was observed in the LRT recipients (Ulaş Kahya et al. 2022).
Through further classification of the LRT modalities, we found that PORT possessed the optimal survival benefits, whereas there were no differences in survival between patients receiving PRRT and surgery only.Results of subgroup analysis demonstrated that radiotherapy on the basis of surgery contributed to improve survival, regardless of breast-conserving surgery, mastectomy and reconstruction.Therefore, it is of great value to select appropriate patients for vigorous LRT.Stratified analysis of the tumor size was performed and the results showed that PORT was beneficial to T2-T4 stage patients, while for T1 stage patients, there were no similar results.This discrepancy may be due to the relatively larger tumor size, higher histological grade, more extensive involvement of lymph nodes, and higher proportion of viscera metastasis, aggressive subtype and chemotherapy recipients in T2-T4 stage patients.On the basis of systemic therapy, postoperative radiotherapy for those patients with fairly good general conditions can minimize tumor burden, regulate tumor suppressive microenvironment through in situ and 'Radscopal effect' and thus improve patients' survival.Preoperative radiotherapy can lessen tumor volume, reduce the clinicopathological stage and effectively control the localized symptoms.However, in this study, PRRT recipients had no survival benefits compared with the patients treated with surgical intervention only.This may be owing to the limited patients receiving PRRT in this cohort, and its impact may not be thoroughly evaluated.Since stage-IV BC patients with LRT possessed a longer life expectancy, they inevitably face the risk of other cause-specific death.Therefore, it is indispensable to construct a competing risk model in LRT recipients to compare survival and predict prognosis.The cumulative incidence curves indicated that patients treated with PORT had a lower BCSD, which was consistent with the Cox hazard proportional model.Through the univariate and multivariate analyses of the two models, we included 10 independent predictors to establish the comprehensive competing risk nomogram.Although many nomograms have been developed in recent years to predict the survival of stage-IV BC patients, our model has some distinctive features (Li et al. 2017;Liu et al. 2021Liu et al. , 2022)).Firstly, we integrated the surgery and radiotherapy data and structured a novel parameter 'LRT'; thus, we could make personal predictions of the survival probability more accurately in the LRT recipients.Secondly, considering the impact of competitive events, this model could make predictions more objectively without overestimating the role of LRT.Thirdly, because the research objects were metastatic tumors, unlike the traditional survival analysis, the index we used to evaluate the prognosis was BCSD.Assessment of the model was performed, and the calibration curves and time-dependent ROC curves revealed fabulous uniformity and accuracy of the model, respectively.
There were also some limitations in this study.Firstly, it is hard to completely remove the bias in the retrospective studies through the available statistical analysis (Llorca and Delgado-Rodríguez 1999).Secondly, we only compared the survival differences between the surgery only, radiotherapy only, PORT and PRRT groups, but did not distinguish the specific radiotherapy methods on prognosis.Thirdly, the database lacked information of the patients' general health status, as well as the quantity and size of the metastatic lesions, which might influence the clinical management.Lastly, the number of stage-IV LRT recipients with complete medical records in our cancer center was small, so we were unable to carry out effective external verification.

Conclusion
Our study revealed that LRT was associated with better prognosis in stage-IV BC patients, especially in the PORT modality.The comprehensive nomogram integrating

Fig. 1
Fig. 1 Proportion of patients undergoing locoregional treatment from 2010 to 2017

Fig. 2 A
Fig. 2 A Kaplan-Meier curves of OS before PSM.B Kaplan-Meier curves of BCSS before PSM.C Kaplan-Meier curves of OS after PSM.D Kaplan-Meier curves of BCSS after PSM

Fig. 3
Fig. 3 Kaplan-Meier curves of A OS, B BCSS in different LRT modalities

Fig. 6 Fig. 7
Fig. 6 Forest plot visualizing multivariate Cox regression of BCSD in stage-IV BC patients

Fig. 8
Fig. 8 Calibration curves for predicting BCSD at 1-, 2-and 3-year intervals (A) in the training set, (B) in the validation set

Table 2
Univariate Cox hazard proportional model analysis of LRT recipients PORT surgery combined with postoperative radiotherapy, PRRT surgery combined with preoperative radiotherapy, HR hazard ratio, 95%CI 95% confidence interval, OS overall survival, BCSS breast cancer-specific survival Forest plot visualizing multivariate Cox regression of A OS, B BCSS in Stage-IV BC patients

Table 3
Univariate competing risk analysis