This heterogeneity is manifested in different molecular types, gene signatures, metastatic patterns and demographic variables, and these factors interrelate, leading to different outcomes between individuals (27). In a previous study on 7,986 patients with dnMBC from the SEER database from the years 2010–2016, hormone receptor (HR)⁺/human epidermal growth factor (HER)2⁻ was the most common subtype (57.6%), followed by HR⁺/HER2⁺ (19.0%), triple-negative breast cancer (TNBC; 13.6%) and HR⁻/HER2⁺ (9.8%) (27). In addition, 50.3% of dnMBC tumors were poorly differentiated or undifferentiated (27). The prognosis of patients with de novo invasive lobular carcinoma may be worse than that of those with de novo invasive ductal carcinoma (28). The majority of patients with dnMBC suffer from regional lymph node metastasis, suggesting that regional involvement may precede distant metastasis (29). The most common distant metastatic site is bone, accounting for 73.4% of sites, while the brain is the least common disseminated organ (6.9%) (27). Different subtypes of dnMBC demonstrate specific predilection in distant metastasis. A previous study demonstrated that luminal tumors were more likely to metastasize to bone [P=0.02; bone vs. other organs, including the liver, lungs and central nervous system (CNS)] (30). The upregulated HER2 subtype harbored a liver-homing feature (P=0.0004; liver vs. other sites), while TNBC subtypes were inclined to metastasize to the lungs (P=0.01; lung vs. other sites) (30).

During adjuvant systemic treatment, more treatment-resistant tumor clones develop, resulting in disease progression. Therefore, refractory MBC generally harbors a more dismal prognosis than dnMBC (27,32–36). Following active treatment, some patients with dnMBC can obtain a long survival time (37).

The Epidemio-Strategy-Medical-Economical (ESME)-MBC data platform (NCT03275311) included 16,702 patients with dnMBC from January 1, 2008 to December 31, 2014 (38). Of these patients, the median OS rates of patients with the HR⁺/HER2⁻ (n=9,907), HER2⁺ (n=2,861) and TNBC (n=2,317) subtypes were 42.12, 44.91 and 14.52 months, respectively. Elbaiomy et al (39) found that the TNBC subtype of dnMBC highly expressed breast cancer stem cell (BCSC) markers (CD44⁺ and CD24⁻), which predicted a poor treatment response, and shorter OS and progression-free survival (PFS) times. In the future, BCSC marker expression may be used to tailor the treatment strategy for patients with dnMBC, and BCSC target therapy may be a promising strategy (39). Dawood et al (40) reported that the number of patients with dnMBC surviving >2 years increased over time. A younger age, lower tumor grade, HR positivity, non-inflammatory disease and definitive primary resection contributed to this trend. As previously demonstrated, compared with patients with dnMBC aged 40–59 years, those aged <40 years had a longer survival time, particularly those with the TNBC subtype (41). An uninsured status and older age have also been shown to be associated with early mortality in patients with dnMBC (42). Immediate breast reconstruction has also been shown to not affect survival outcomes when compared with mastectomy in patients with dnMBC (43). In 2002, Khan et al (44) performed a comprehensive retrospective study of 16,023 patients with dnMBC in the duration of 1990–1993 using the National Cancer Database. Their findings revealed that four parameters determined the prognosis of patients with dnMBC, including surgical removal of the PT, the administration of systemic treatment, metastatic burden and tumor type (visceral or soft tissue). The retrospective study by Soares et al (45) demonstrated that breast surgery (hazard ratio, 0.45) and initial CNS metastasis (hazard ratio, 3.09) were independent prognostic factors for patients with dnMBC. In 2006, Babiera et al (11) analyzed 224 patients with dnMBC and revealed that only one metastatic site and no HER2 gene amplification indicated a more favorable OS, while estrogen receptor (ER) positivity and PT resection were closely linked to an improved metastatic PFS (MPFS). Similarly, a previous multivariate analysis of prognostic factors and OS in 395 dnMBC cohorts performed by Blanchard et al (46) found that definitive primary surgery, HR positivity and a low number of metastases were associated with an improved prognosis. King et al (47) re-evaluated the data of dnMBC cohorts enrolled in the Translational Breast Cancer Research Consortium (TBCRC) 013 program and screened out 101 patients with 21-gene recurrence score (RS) results. In the ER⁺/HER2⁻ (n=69) subgroup, a 21-gene low risk (RS <18) score was a strong indicator of an improved 2-year OS and time to first progression (47). The metastatic pattern has a profound impact on the prognosis of patients with dnMBC. The OS times of visceral, nodal and bone-only metastases are approximately <6 months, 18 months and 3–4 years, respectively (48). Lin et al (49) determined that the allocation and number of metastatic sites were independent prognostic factors for dnMBC. Therefore, they proposed the subdivision of the M1 stage criteria in a heterogeneous setting: i) M1a, a single metastatic site excluding the liver and brain; ii) M1b, multiple metastatic sites excluding the liver and brain, or liver only involvement; and iii) M1c, liver plus other metastatic sites, or brain only involvement. The M1a subtype benefited the most from PT resection, while the upstaging of M1 was related to a higher benefit from chemotherapy alone (49). In 2020, Plichta et al (50) proposed a novel dnMBC staging system for prognosis. The proposed recursive partitioning analysis including anatomical and biological factors to refine prognostic estimates. The distant metastatic site number and ER status were identified as the first and second stratification points, respectively, and the HER2 status, clinical T stage and tumor grade were incorporated for further divisions. Subsequently, dnMBC was classified into three stages: IVA, IVB and IVC, with a 3-year OS rate of >50, 30–50 and <30%, respectively (50).

The genomic analysis of dnMBC was previously conducted to reveal the drivers contributing to the early onset of metastasis (51). In HR⁺/HER2⁻ dnMBC, PIK3CA mutations were enriched (41.9%), indicating that PI3K inhibitors may play a critical role in the treatment of this subset. Additionally, KMTD2 and SETD2 mutations were more common, while TP53 and BRCA1 were less common in HR⁺/HER2⁻ dnMBC, compared with recurrent MBC. In TNBC dnMBC, MYB amplification was prevalent and high tumor mutation burden predicted a better prognosis (51). PTEN mutations were more frequent in dnMBC, which are closely associated with immune evasion (52). As previously demosntrated, 11.8% of patients with dnMBC harbored a BRCA2 mutation, indicating that BRCA germline testing was necessary (53).

In dnMBC, with the aim of optimizing disease monitoring, active image evaluation, such as positron emission tomography-computed tomography scanning and magnetic resonance imaging, for CNS screening and tumor marker surveillance is encouraged (54).

Thomas et al (2) conducted the largest retrospective study of dnMBC (n=21,372) using the SEER database spanning 24 years (1988–2011). Patients undergoing surgery as initial therapy (n=8,330) harbored a significantly longer median OS time than those who did not (n=13,042) (28 vs. 19 months) (2). Additionally, MF 07–01, the first randomized trial, also verified the OS benefit gained by upfront LRT in dnMBC (17). A total of 274 candidates were randomized into two groups: The upfront LRT followed by systemic treatment group (n=138), and the systemic therapy only group (n=136). The LRT arm demonstrated a notable OS advantage (median OS time, 46 vs. 37 months, P=0.005; 5-year OS rate, 41.6 vs. 24.4%). Unplanned subgroup analyses found that patients with the ER/PR⁺ (P=0.008) and HER2⁻ (P=0.01) subtypes, an age <55 years (P=0.007) and solitary bone-only metastases (P=0.04) benefited the most from LRT (17). Despite more favorable features in the LRT arm, such as a higher proportion of ER/PR⁺ (85.5 vs. 71.8%: P<0.05) and lower rates of TNBC (7.3 vs. 17.4%; P<0.05), the MF 07–01 trial highlighted that in a well-selected population of dnMBC, the upfront LRT may lead to a survival benefit (17). In the study by Shien et al (55), 344 patients with dnMBC were divided into two groups: The local surgery (n=160) and no local surgery (n=184) groups. A total of 150 (94%) patients in the surgery group had primarily local surgery. Following a median follow-up time of 33 months, the local surgery group attained a superior OS time compared with the no surgery group (median OS time, 27 vs. 22 months; P=0.049), notably in younger individuals (<50 years old) (median OS time, 35 vs. 24 months: P=0.021) (55).

In 2017, Barinoff et al (56) re-analyzed two prospective trials conducted in Germany with patients with MBC administered targeted therapy (ML16684: Started in 2000, trastuzumab for HER2⁺ patients; ML21165: Started in 2007, bevacizumab for HER2⁻ patients). Among the 570 patients with dnMBC (405 from ML16684 and 165 from ML21165), 2 patients had no data on local resection, 426 patients underwent upfront breast surgery and 142 patients did not undergo breast surgery (56). The surgery cohort had less patients with T4 staging and patients with a lower metastatic tumor burden, but more patients with N3 staging. Other characteristics, such as HR and HER2 status, Eastern Cooperative Oncology Group (ECOG) score and prior palliative management, were equally distributed in the surgery and no surgery cohorts. However, upfront primary surgery failed to bring a benefit for either PFS (surgery vs. no surgery, 13.55 vs. 11.76 months; P=0.18) or OS (surgery vs. no surgery, 34.1 vs. 31.7 months; P=0.23). Nonetheless, in the subgroup without visceral metastasis, the removal of the PT manifested an overt OS time advantage (45.7 vs. 27.2 months; P=0.026) (56).

On the contrary, the ABCSG-28 POSYTIVE trial, a prospective randomized study aiming to evaluate the effect of upfront surgery in untreated patients with dnMBC, drew the opposite conclusion (18). In 2011–2015, 90 untreated candidates with dnMBC were randomly assigned into two groups: The upfront PT surgery followed by systemic therapy (group A, n=45) and systemic therapy only (group B, n=45) groups. Upfront surgery failed to bring any survival benefit (group A vs. B; median survival time, 34.6 vs. 54.8 months, P=0.267; time to distant progression, 13.9 vs. 29.0 months, P=0.0668) (18). This trial was strengthened by a well-balanced design regarding critical prognostic factors, but was criticized for insufficient recruitment and early closing of the study, which underpowered the conclusion (18).

Co et al (57) and Kwong and Co (58) conducted a retrospective study on 172 patients with dnMBC in Hong Kong between 2007–2016. Following systemic intervention for 8–10 months, resection of the PT was conducted in 91 good responders, who gained a notably improved 5-year OS rate (43.9 vs. 33.9%; P=0.026) (57,58) Additionally, Rao et al (59) analyzed the optimal time for surgery for patients with stage IV breast cancer initially with a complete PT excision. In their study, 75 patients were divided into three groups according to the interval between the diagnosis of dnMBC and surgery of the PT: Group 1, 0–2.9 months (n=47); group 2, 3–8.9 months (n=14); and group 3, ≥9.0 months (n=14). Notably, group 2 had the longest MPFS (P=0.0008), despite not having an OS superiority (P=0.12), particularly in patients with a single metastatic site and negative margin (59). A rational explanation for this was that systemic chemotherapy was completed within 3–9 months after diagnosis and those good responders benefited the most from curative surgical management. However, this conclusion was limited by the small sample size and retrospective nature of the study (59).

Lane et al (8) conducted a large-scale retrospective study using the American College of Surgeons National Cancer Database for the years 2003–2012. In total, 24,015 patients with dnMBC who were alive 1 year after diagnosis were enrolled in their study. The candidates were divided into three groups according to the treatment sequence: The systemic treatment only (n=13,505), the upfront surgery before systemic treatment (n=4,552) and the surgery following systemic treatment (n=5,958) groups. The median OS times of these groups were 37.5, 49.4 and 52.8 months, respectively (8). Surgery, whenever it was performed, was closely associated with an improved prognosis, compared with systemic therapy alone (P<0.001). Furthermore, surgery following systemic intervention appeared to prolong the median OS time, compared with the surgery prior to systemic treatment group (8).

For aggressive subtypes of dnMBC, upfront surgery may lead to deleterious consequences. Patients with multiple visceral metastases benefit from first-line systemic treatment other than initial surgery (2). In 2015, the results of the TATA trial, an open-labelled randomized study, were published (19). A total of 350 patients with dnMBC in India responsive to first-line systemic treatment (336 chemotherapy and 14 endocrine) were randomly divided into two groups: The LRT (n=173) and no LRT (n=177) groups. Following a median follow-up time of 23 months, LRT to the PT failed to confer any survival benefit (19). The median OS times and the 2-year OS rates in the LRT and no LRT groups were 19.2 vs. 20.5 months (P=0.79) and 41.9 vs. 43.0%, respectively. The LRT modality also significantly improved the locoregional PFS time (median not attained vs. 18.2 months; P<0.0001), whereas it was detrimental to the distant PFS time (median, 11.3 vs. 19.8 months; P=0.012) (19). That trial was criticized due to insufficient systemic treatment according to the current guidelines. A total of 107 patients (31%) were HER2⁺, but only 9 patients (8%) underwent anti-HER2 therapy. In addition, only a few patients were administered taxane-based chemotherapy. Moreover, the median OS time in the TATA trial was much shorter than that reported by others such as Lane et al (8) and Pons-Tostivint et al (16), partly due to the undertreated systemic therapy and later diagnosis of symptoms. Collectively, this underpowered the conclusions made, particularly in the era of a constantly updated systemic strategy.

Pons-Tostivint et al (16) published their latest results on 4,276 patients with dnMBC in the NCT03275311 trial in 2018. A total of 1,706 patients underwent LRT, while 2,570 patients did not. The median duration between MBC diagnosis and LRT was 5.7 months. Compared with the TATA trial, all the patients in the NCT03275311 trial received modern systemic therapy (16). In total, 94% of the HER2⁺ cohort was administered trastuzumab and 90% of the HR⁺ candidates were administered endocrine therapy. LRT attained a superior OS time in patients with the HR⁺/HER2⁻ (61.6 vs. 45.9 months; P<0.001) and HER2⁺ (77.2 vs. 52.6 months; P=0.008) subtypes, but had no effect on OS in the TNBC population. Additionally, in the visceral metastatic subset, LRT also reduced the risk of mortality (16). In a previous retrospective study on 622 patients with dnMBC performed by Cady et al (60), primary surgery (n=234) indicated a survival benefit compared with without surgery (n=388; P<0.0001). For the bone metastasis subset (n=255), the survival advantage was overt in the primary surgery group (P<0.0001), but this benefit was reduced after case-matching (n=168; P=0.0003), which reflected the selection bias. Similarly, in the visceral metastatic subgroup (n=159), the survival superiority conferred by primary resection was significant (P=0.0008), but this is not the case after case-matching (n=100; P=0.09) (60). With regards to the order of systemic treatment and primary surgery, it was found that 90% of patients who survived for 2-years were administered chemotherapy initially. However, this survival benefit was not present in the cohort of patients who received surgery immediately after systemic therapy, suggesting that patients with dnMBC who respond robustly to upfront systemic intervention are most likely to benefit from primary surgery (60).

NCT03870919 has been launched to evaluate the OS of combined modality with the CDK4/6 inhibitor, palbociclib, plus letrozole and the most adapted LRT in the treatment of patients with naïve HR⁺/HER2⁻ dnMBC. The ECOG 2108 (NCT01242800) study was designed to clarify the effects of local treatment in patients with stage IV breast cancer. In the 2020 ASCO meeting, the results of the long-awaited ECOG 2108 study were presented in which 390 patients with dnMBC were enrolled, of which 256 patients who responded to initial systemic therapy were randomized to either undergo LCT (surgery and radiation, n=125) followed by systemic therapy or continue with systemic treatment (n=131) (61). However, early LCT failed to bring any survival benefit (neither OS nor PFS). Without LRT, a 2.5-fold higher risk of local failure was posed. Therefore, the ECOG-ACRIN E2108 trial indicated that LRT to the PT should not be recommended to patients with dnMBC for the purpose of survival extension and only reserved for those with well-controlled systemic disease and a progressive PT (61). However, these results were limited for the following reasons: i) Multivariate analysis was not conducted; ii) the inclusion criterion was PT surgery with free margin. However, in the LRT group (n=125), only 109 patients underwent surgery and only 87 (80%) patients had R0 resection; and iii) the T4 and/or N2/3 patients accounted for 48% of the cohort. The effect of the high proportion of patients with a locally advanced status needs to be clarified (61).

Retrospective analyses indicated that for aggressive subtypes, such as multiple visceral metastases or those with heavy tumor burden, upfront LRT was detrimental. Timely and effective systemic treatment to make metastatic lesions controllable and asymptomatic followed by definitive LRT may infer the greatest survival benefit (62,63).

The role of adjuvant RT has been explored by a series of studies. In the study by Altundag (64), 118 patients with dnMBC administered LRT were further divided into three groups: RT only, 2 (1%) patients; surgery only, 54 (29%) patients; surgery followed by RT, 132 (70%) patients. Post-operative RT provided a significant survival advantage (64–66). The 5-year OS and PFS rates in the three groups were 0, 22 and 56% (P<0.001) and 0, 3 and 27% (P=0.001), respectively (66). Additionally, Kim et al (67) used the SEER database to address this issue. Following propensity-score matching, the addition of RT further improved the cancer-specific survival (CSS) of candidates who lived >6 months. The study by Arciero et al (25) also favored the addition of RT to surgery of the PT in dnMBC. In the multivariate analysis of the OS of patients with dnMBC who underwent surgery, the administration of RT was an independent factor of an improved outcome (P<0.001) (25).

By contrast, in the analyses performed by Lane et al (8), the addition of RT had no association with OS (P=0.83), which was revealed by an adjusted multivariate OS model. However, the majority of patients did not receive RT (59.75%) in this cohort, and the RT administered was not only confined to the PT area (breast, chest or regional lymph nodes; 17.89%), but also included the spine (9.19%) and extremity bones, including the pelvis (4.36%). Therefore, the conclusion of that study could not truly represent the value of RT to the PT region (8).

Tan et al (68) conducted a population-based study using the SEER database for the years 2004–2008. A total of 10,441 patients with MBC were divided into four subsets: i) R0 resection (both resection of primary and metastatic sites), 272 (2.61%) patients; ii) PT removal only, 4,025 (38.55%) patients; iii) metastatic lesion resection only, 409 (3.92%) patients; and iv) no surgery, 5,735 (54.93%) patients. The median OS times of the patients in the four groups were 51, 43, 31 and 21 months, respectively (P<0.001) (68). Furthermore, the HR status influenced the survival benefit from surgery. For the HR⁺ cohort, the median OS times and 5-year OS rates for R0 resection, primary resection, metastases resection and no resection were 66, 52, 38 and 28 months and 54.1, 44.9, 31.7 and 22.0%, respectively (P<0.001). However, for the HR⁻ cohort, the median OS times and 5-year OS rates for R0 resection, primary resection, metastases resection and no resection were 18, 24, 12 and 12 months and 26.7, 25.0, 6.8 and 11.8%, respectively (68). In the HR⁺ cohort, the R0 resection group attained a further survival benefit compared with the primary resection group (P<0.001), and the same was observed for the metastases resection group compared with the no resection group (68). However, this additional survival advantage disappeared in the HR⁻ cohort. These findings indicated that HR⁺ patients with dnMBC may attain more survival benefits from a more aggressive LRT approach (68).

The addition of post-operative RT to the PT region following surgery may lead to an improved outcome of patients with dnMBC. For HR⁺ patients with dnMBC, distant metastatic site resection could also lead to an improved survival outcome.

In 2006, Rapiti et al (10) conducted a study including 300 patients with dnMBC recorded at the Geneva Cancer Registry between 1977–1996. After adjusting for other cancer prognostic confounding factors, complete resection of the PT reduced the risk of breast cancer-specific mortality by 40% (P=0.049) (10). This benefit was particularly evident for patients with dnMBC and bone metastasis only. The metastatic pattern was a vital covariable that determined the prognosis of dnMBC (10). Wang et al (69) retrospectively analyzed 8,142 patients with dnMBC from 2010–2015 using the SEER database. By comparing the surgical (n=1,891) and non-surgical groups (n=6,251), they noted that surgery provided a survival advantage in those with a bone-only metastasis or with multiple metastases including the bone (P<0.05), but a survival inferiority in those with multiple metastases only in visceral organs (P<0.05) (69). The aforementioned MF 07-01 study, revealed that the solitary bone-only metastatic subset benefited from LRT (17). However, the other three well-known randomized trials, TATA (19), TBCRC 013 (47) and POSYTIVE (18), failed to identify the superiority brought by surgery. Notably, the proportion of bone-only metastasis in the TATA (19), TBCRC 013 (47), POSYTIVE (18) and MF 07-01 (17) studies was 28, 37, 38 and 46%, respectively. The highest ratio of patients with bone-only metastasis was found in the MF 07-01 study, which largely determined the positive results of the study (17).

Barinoff et al (56) found that the number of involved sites was closely associated with PFS and OS in the dnMBC cohort. Multivariate analysis determined that <3 metastatic sites led to an improved PFS and OS in patients with dnMBC (P=0.0005 and P=0.00051, respectively). Xiong et al (3) retrospectively reviewed 313 patients with dnMBC from January, 2006 to April, 2013 at the Sun Yet-sen University Cancer Center. Patients who underwent local surgery (n=188) had a notable median OS time advantage compared with those who did not undergo surgery (n=125) (78 vs. 37 months; P=0.002). Stratified survival analysis revealed that bone metastasis plus a PT <5 cm, ≤3 metastatic sites, or soft tissue metastasis predicted a benefit from local removal (3). Bafford et al (9) analyzed the impact of staging for breast surgical removal in patients with dnMBC. The 147 patients enrolled were divided into two groups: The surgery (n=61) and non-surgery (n=86) groups. After adjusting for age, number of metastatic sites, ER/Her2neu status and the use of chemotherapy, trastuzumab and endocrine therapy, the surgery group gained a significantly improved OS time compared with the non-surgery cohort (4.13 vs. 2.36 years; P=0.003). However, the number of metastases in the non-surgery group was markedly higher than in the surgery group (≥3, 84 vs. 41%) (9). Notably, in the surgery group, 36 patients underwent surgery before the diagnosis of metastatic disease. The 25 patients diagnosed with stage IV pre-operatively had a very similar OS time to those without surgery (2.4 vs. 2.36 years), while the patients undergoing surgery before metastatic diagnosis had a notably improved OS time (4.05 years) (9). This may be explained by the fact that those who underwent resection of the intact PT before a stage IV diagnosis were typically asymptomatic and with a limited metastatic burden, which contributed to a relatively optimal outcome (9). Previously, Lin et al (49) identified 8,582 patients with dnMBC from 2010–2014 using the SEER database, and further subdivided M1 into three groups. Patients with an M1a status (meaning a single metastatic site, except the liver and brain) gained the most benefit from PT surgery.

In a retrospective study conducted by Tan et al (68), resection of both the primary and metastatic lesions for stage IV BC led to the optimal survival of patients with HR⁺ breast cancer. In a single institutional analysis of 111 patients with dnMBC conducted by Samiee et al (70), HR⁺ and surgical removal both indicated a better prognosis (P<0.001 and P=0.041, respectively), using multivariate analysis. The ESME-MBC study demonstrated a median OS time of 42.12 months for newly diagnosed HR⁺/HER2⁻ patients with dnMBC (38). In another study, HR positivity predicted a favorable outcome in the surgical group independently, following multivariate analyses. ER positivity was found to be the only good prognostic predictor in surgical patients with dnMBC (57).

Rashaan et al (71) demonstrated that a younger age (P=0.03) and an absence of complications (P=0.03) were independent prognostic factors for improved survival in the dnMBC cohort receiving surgery. In the retrospective study by Shien et al (55), PT surgery conferred an OS advantage in young candidates (P=0.021), but not in older candidates (>51 years old, P=0.665). Blanchard et al (46) also demonstrated that the subpopulation of relatively long survivors of dnMBC gained more benefit from definitive surgery. For those patients who underwent surgery and survived for >2 years (n=119), survival was significantly improved compared with no primary surgery (n=51; P=0.0037) (46).

Yoo et al (7) developed and validated a survival prediction-scoring model using the Korean Breast Cancer Registry database, aiming to identify the subset of long-term survivors of dnMBC who underwent surgery of the PT. Their study excluded those who received resection of metastatic foci. The 2,232 enrolled patients were divided into three groups, according to the surgery type: Surgery (n=1,541), non-surgery (n=588) and partial surgery (n=103; only breast or axillary removal) groups. The surgery cohort had a significantly improved median survival time compared with the non-surgery group (53 vs. 31 months; P<0.001) (7). There was no significant difference with respect to the median OS time between the non-surgery and partial surgery cohorts (37 vs. 31 months; P=0.113). The OS prediction-scoring model for patients with dnMBC undergoing surgery of the PT was established based on the following seven factors: T stage, tumor grade, lymphovascular invasion (LVI), ER expression, and the levels of Ki-67, CA 153 and alkaline phosphatase (7). The patients in the surgery group were further grouped into four subsets: Subsets with a score of 0–3, 4–5, 6–7 or 8–10, with clearly defined 3-year survival rate outcomes (87.3%, score 0–3; 68.4%, score 4–5; 48.2%, score 6–7; 35.3%, score 8–10). The group with a score 0–3 had a significant advantage as regards OS, demonstrated both in the discovery cohort (87.3%; P<0.001) and in the validated cohort (85.9%; P<0.001). This tool provided a method to identify the candidates with dnMBC who would benefit the most from primary surgery (7).

Since dnMBC is highly heterogeneous, multiple factors, including the status of the patient, tumor pathology, biology and metastatic pattern, may help to determine the cohorts most likely to benefit from LRT. A younger age and a good general condition without complications, a smaller tumor, a lower tumor grade, no LVI, HR positivity, bone-only metastasis and low metastatic burden have been found to be associated with an improved outcome. In the future, a gene signature of dnMBC may help to identify the patients who will benefit the most from LRT. Additionally, an improvement in systemic intervention and new drugs may achieve the complete remission of metastatic lesions, which will result in more candidates who will benefit from definitive LRT. In clinical practice, the flow diagram with therapeutic options presented in Fig. 1 may aid in decision-making when treating patients with dnMBC.

Previously, some studies proposed that resection of the PT would promote the progression of distant metastasis, partly mediated by the angiogenesis rebounds and the growth factor surge responding to the surgical wounding (72,73). It was suggested that the PT secreted angiogenesis inhibitors, such as angiostatin, controlling the growth of metastatic lesions (74). PT removal would therefore alter the growth kinetics of the residual tumor (75). However, these theories resulted from an animal model with an unclear clinical translational implication (76).

The metastasis of a tumor is comprehensive, involving multiple factors and associations. Under certain incentives, such as hypoxia, tumor cells develop epithelial-interstitial transformation and increase their invasive ability (77). Accordingly, the interactions with stromal cells in the surrounding microenvironment degrade, restructuring the extracellular matrix and allowing tumor cells to invade the basement membrane and enter blood vessels. In this process, there are approximately three mechanisms involved. Firstly, the PT will secrete relevant cytokines, such as TGF-β, VEGF, matrix metalloproteinases and the remodeling enzyme, lysyl oxidase, and promote the migration of tumor cells. Secondly, endocrine signaling from the primary foci recruits mesenchymal stem cells (MSCs) from the bone marrow into the microenvironment of the primary foci to promote metastasis (78–80). Thirdly, pro-inflammatory and pro-angiogenic factors secreted by the primary foci, such as TNF-α, IL-8 and VEGF (78,81–83), which increase angiogenesis and vascular permeability of the primary foci, facilitate the invasion of blood vessels by tumor cells. Circulating tumor cells (CTCs) form clusters and migrate in the bloodstream (84–86). Various cytokines, chemokines and growth factors secreted by the primary foci and MSCs recruited from the bone marrow also enter the target organ in advance, which shapes the targeted microenvironment and forms a pre-metastatic niche to welcome the arrival of the CTCs (83,87–92). Organ specificity during metastasis is regulated by chemokines, exosomes, integrins and other factors, such as chemokine receptors (87–89,93–97). If CXCR4 is highly expressed by the PT, the bone, lungs and brain are prone to metastasis due to the extensive expression of CXCR4 ligand-CXCL12. The new metastasis may originate from the PT or from an existing metastasis (98–100). Crosstalk exists between the primary and secondary tumor. The PT secretes highly effective immunosuppressive molecules, such as TGF-β, VEGF, IL-10, inflammatory factors (such as IL-6 and IL-8) and angiogenesis inhibitors (such as angiostatin and endostatin), to regulate growth of the secondary tumor (Fig. 2) (78,81–83).

Total tumor burden has been revealed to be closely linked with survival (44,101). The level of CTCs is also an independent prognostic factor for the PFS and OS of patients with MBC (102,103). Similarly, the PT of dnMBC serves as another metastatic site or source of dissemination (10,13,44). Resection of the PT can block the continued production of CTCs, reduce subsequent metastasis and improve prognosis (46–49). In the study conducted by Rapiti et al (10), surgery of the PT with a positive margin failed to improve survival of patients with dnMBC compared with patients who did not undergo surgery. The 5-year CSS rates of patients who underwent primary surgery with negative margins, positive margins, an unknown margin status or who did not undergo surgery were 27, 16, 12 and 12%, respectively (P=0.0002) (10). Yoo et al (7) conducted a retrospective study on dnMBC with a larger sample size. Patients were further categorized into three groups according to the following: The surgery (n=1,541), partial surgery (n=103, no definite resection of the PT) and non-surgery (n=588) groups. The definite surgery cohort demonstrated a significant OS time advantage compared with the non-surgery group (53 vs. 31 months; P<0.001), which was not observed in the partial surgery group (37 vs. 31 months; P=0.113) (7). Additionally, Hazard et al (104) verified that the chest wall status had a direct effect on the OS of patients with dnMBC. Compared with a symptomatic chest wall, the presence of a free or asymptomatic status could notably decrease the hazard ratio of mortality (hazard ratio, 0.415). Multivariate analysis further supported that primary surgery markedly reduced the probability of uncontrolled chest wall status (104). Furthermore, MSCs recruited by the PT from the bone marrow release paracrine signals, such as C-C motif ligand 5, which interact with C-C motif receptor 5 on cancer cells, induce the metastasis of primary cancer cells and maintain the metastatic phenotype of cancer cells (79). Cancer stem cells (CSCs) in the PT, are able to self-renew, differentiate and activate, and are the main source of continuous dissemination and metastasis (52). These findings highlight that the complete clearance of the PT, eradicating the source of metastasis, may translate into a survival benefit (101).

Maximal tumor debulking markedly increases drug accessibility (101). Notably, necrotic, hypoxic and avascular tumors exhibit resistance to chemotherapy or RT. Hazard et al (104) demonstrated that surgery improved the local control (S vs. NS: 36/44 82% vs. 20/59 34%, P=0.001) and prolonged time to first progression (P=0.015) in dnMBC. In total, 18.2% (8/44) patients in the surgery group also experienced cancer relapse following LRT (104). The removal of these tumors could limit the emergence of treatment-resistant cells, resulting in an increased sensitivity to treatment (12,105). Additionally, CSCs are resistant to common anticancer strategies. Eradiation of the PT will therefore improve treatment efficacy (54–60,62).

The tumor evolution of some dnMBC patients is related to immune factors and benefit most from removal of the PT (2). In a 4T1-bearing mammary carcinoma mouse model, both antigen specific T-cell proliferation and antibody responses were markedly suppressed compared with tumor-free mice. In patients with BC, the levels of cytokines produced by CD4⁺ and CD8⁺ T cells, including IL-2, IFN-γ, TNF-α and IL-4, were markedly decreased compared with the healthy population (63,64). Tumor cells can synthesize and secrete various immunosuppressive factors, such as TGF-β and VEGF, which inhibit effector cells (65). Myeloid-derived suppressor cells play a vital role in immune suppression in patients with BC (66,67). Removal of the PT helps to restore both antibody and cell-mediated immune responses, which manifests as an activated macrophage and dendritic cell function, even in the presence of metastatic disease (100). Moreover, the PT harbors a certain humoral resistance that does not belong to or at least develops slowly in the metastatic lesions. The eradication of this dependent relationship existing between the primary and metastatic foci by PT removal causes the disseminated lesions to be more vulnerable to therapy (106).