Introduction

Today, breast cancer (BC) is not regarded as a single entity but as a heterogeneous disease with different biological subtypes. Classic histopathologic variables have been combined with rankings based on multi-gene expression signatures, an important step towards personalizing BC treatment. Amongst the most popular platforms are the genomic grade index (GGI; MapQuant Dx®), MammaPrint® [1], Oncotype DX® [2], the wound-response model [3], Veridex 76-gene signature [4], 92-gene signature THEROS [5] and the intrinsic subtype model [6]. The latter provides the most extensive biological information and allows the classification of BC into five intrinsic subtypes: Luminal A, luminal B HER2 negative, luminal B HER2 positive, HER2-overexpressing and basal-like subtypes. Of course, clinico-pathological surrogate definitions exist and the basal-like subtype is usually treated equivalent to the triple-negative subtype (TNT) [7–11]. The BC subtypes show different biological behavior in terms of survival and recurrence and have typical patterns of metastatic spread [12–14]. For example, patients with hormone receptor-positive BC are more likely to develop bone metastases, whereas visceral recurrence is associated with a lack of the hormone receptors [15, 16]. TNT seems to metastasize more likely viscerally [13, 17, 18], predominantly into the lungs [19]. The HER2-overexpressing subtype is more likely to metastasize to lung, liver and brain than HER2-negative subtypes. Moreover, HER2-positive BC tends to metastasize less often to the skeleton [20, 21]. In contrast, Kenneke and coworkers evaluated that all subtypes, except basal-like tumors, metastasize most commonly to the bone [22].

Regarding the pattern of metastasis, the different molecular subtypes of BC demonstrate variable prognostic behavior. Metastases were found less frequently in patients with the luminal A subtype and more in the HER2-overexpressing and TNT subtypes [13]. The advanced TNT BC has a poor prognosis and fewer targeted therapies [23–26]. In view of the poor prognosis of advanced BC, the aim of this study was to analyze the typical metastatic pattern of the subtypes of BC in Germany, the sites of metastasis, and the outcome to get a better understanding of the biology and prognosis of the different BC subtypes.

Methods

The BRENDA (breast cancer care under evidence-based guidelines) database used for this retrospective study included 9625 patients treated in the adjuvant setting without metastases at the Department of Gynecology and Obstetrics at the University of Ulm and 16 partner clinics in Germany from 1992 to 2008. It has already been described in various publications with exact inclusion and exclusion criteria [27–30]. Metastatic disease was defined by X-ray imaging, computed tomography, magnetic resonance imaging, ultrasound or bone scan and/or by histological detection. Data were received from the physicians responsible for follow-up care. Moreover, the statistical registers were contacted.

Surrogate definition

We used a surrogate definition that comprises the hormone receptors (HR) of estrogen (ER) and progesterone (PR), the HER2 receptor, and tumor grade (G). Low tumor grade is a grading of 1 and 2; 3 is high. Luminal A is HR positive, HER2 negative and low tumor grade. Luminal B HER2 negative (luminal B/HER−) is HR positive, HER2 negative and high tumor grade. Luminal B HER2 positive (luminal B/HER2+) represents HR positive/HER2 positive. The triple-negative subtype (TNT) is negative for HR and HER2. The HER2-overexpressing subtype is negative for HR and positive for HER2 [8, 9].

Statistical analysis

All categorical data were described using numbers and percentages. Comparisons of categorical variables between groups were made using χ 2 tests. Quantitative data were presented using median and range or mean and standard deviations. Overall survival (OAS) from the time of metastasis was defined as the interval between detection of the first distant metastasis and death. If the patient was lost to follow-up, data were censored at the date of the last known contact. When no information was available, the status was coded as missing data. Survival distributions and median survival times were estimated using the Kaplan–Meier product-limit method. The log-rank test was used to compare survival rates. The Cox proportional hazards model was used to estimate the hazard ratio and confidence intervals. The proportional hazards assumption was assessed by including the individual terms with time in the models. Multivariate Cox proportional hazards regression models were used to accommodate the differing risk factor distributions between groups. For classification, we used Exhausted CHAID (Chi-squared Automatic Interaction Detector), a type of decision tree technique, based upon adjusted significance testing (Bonferroni testing), to construct (non-binary) trees. p values less than 0.05 were considered statistically significant. Statistical analyses were two sided and carried out using R-3.20, NCSS 10 and SPSS 22.

Results

886 patients (9.2%) developed distant metastases during the observation period. Thereof, 61.3% had bone and 56.9% visceral metastases with the predominant locations: liver (37.0%), lung (30.6%), pleura (11.7%) and peritoneum (3.8%). The CNS was affected in 15.3% of the patients and 15.6% had other metastases (e.g., skin and soft tissues) (Table 1). The median observation time was 53 months (95% CI: 48–57 months) from date of primary diagnosis, and 27 months (95% CI: 24.5–29.5 months) from date of metastatic disease.

Table 1 Pattern of metastases according to the subtype of breast cancer in percent
Full size table

The pattern of metastasis and correlation to subtypes

According to the subtypes of BC determined in St. Gallen in 2011 [7], visceral metastases were found most often in TNT BC (68.7%) and the least in luminal A BC (48.3%). The same applies to lung and CNS metastases (50.4 versus 21.0%, respectively, 25.2 versus 9.4%). Liver metastases were formed mostly by the luminal B/HER2+ subtype (46.6%) and the least by the luminal A subtype (33.0%). In each subtype, except TNT, liver metastases appeared more often than lung metastases. We investigated patients with the TNT subtype; 35.1% with liver metastases and 50.4% with lung metastases. Bone metastases occurred most often in patients with luminal A (66.8%) and the least in patients with TNT subtype (38.9%). Pleural metastases were discovered most often in TNT (13.0%), and the least in the luminal B/HER2+ subtype (4.9%). Peritoneal metastases occurred most often in the luminal B/HER2− (5%) and least of all in the luminal B/HER2+ subtype (1%) (Table 1).

Subtypes of breast cancer and numbers of sites of metastases

Three or more different sites of metastases were found most frequently in patients with TNT (26.7%) and the least in women with luminal A subtypes (14.5%). The luminal B/HER2− metastasized most often to two different sites, the luminal B/HER2+ subtype the least. Luminal A (58.8%) were most likely and luminal B/HER2− subtypes (42.5%; Pearson χ 2: p = 0.003) least likely to develop one site of metastasis (Fig. 1).

Fig. 1
figure 1

The different subtypes of breast cancer and the number of metastatic sites in percent as cross-tabulation (Pearson χ 2: p = 0.003)

Full size image

Differences in the location of metastasis and breast cancer subtypes

A highly significant difference in the visceral manifestations was found between the luminal B subtypes (59%), luminal A subtype (48.3%) and the HER2-overexpressing/TNT subtypes (68.1%; p < 0.001) (Fig. 2a). Furthermore, the frequency of bone metastases differed significantly between the luminal (67.5%) and the HER2-overexpressing/TNT subtypes (41.4%; p < 0.001) (Fig. 2b). In contrast, there was little difference in the frequency of liver metastases in the luminal B/HER2−/TNT/luminal A subtypes compared to the HER2 positive subtypes (p = 0.236); 34.9 and 45.1%, respectively (Fig. 2c). Examinations of the lung metastases showed again highly significant differences (p < 0.001): 32.5% of the luminal B and HER2-positive subtypes, 50.4% of the TNT subtype and 21.0% of the luminal A subtype (Fig. 2d) developed lung metastases. Metastases of the CNS occurred in 23.0% of HER2-overexpressing/TNT/luminal B/HER2+ subtypes, whereas only 11.2% of the luminal A and luminal B/HER2− subtypes developed CNS metastases (p < 0.001; Fig. 2e). In contrast, there was only a slight difference in pleural metastasis behavior between the luminal B/HER2+ (4.9%) and the other subtypes (p = 0.419; Fig. 2f).

Fig. 2
figure 2figure 2

Classification of sites of cancer metastases by subcategories of breast cancer: visceral (a), bone (b), liver (c), lung (d), CNS (e) and pleural (f) metastases using the univariate Exhaustive CHAID (Chi-squared Automatic Interaction Detector) decision tree algorithm

Full size image

Overall survival stratified by subtypes of breast cancer

OAS of BC subtypes was analyzed in comparison to the luminal A subtype. Altogether, patients with the luminal A subtype have the best OAS from date of metastatic disease of all patients with metastases. Patients with luminal B/HER2+ subtype had a slightly lower OAS with a hazard ratio (HR) of 1.26 (95% CI: 0.96–1.65; p = 0.103). We observed significantly poorer OAS for patients with the luminal B/HER2− (HR 1.39; 95% CI: 1.13–1.72; p = 0.002), the HER2-overexpressing (HR 1.75; 95%: 1.30–2.36; p < 0.001) and the TNT subtype (HR 2.87; 95% CI: 2.27–3.64; p = 0.002) (Fig. 3). The number of months of OAS from date of metastatic disease is summarized in means and medians in Table 2.

Fig. 3
figure 3

Overall survival (OAS) from date of metastatic disease of all patients with metastases stratified by subcategories of breast cancer

Full size image
Table 2 Means and medians for the overall survival (OAS) in months from date of metastatic disease
Full size table

Patients with the TNT subtype (HR 3.25; 95%: 2.29–4.61; p < 0.001) also had the worst OAS from date of metastatic disease of patients with only one site of metastasis. The luminal A subtype had the best OAS followed by not significantly reduced OAS of the HER2-overexpressing (HR 1.05; 95% CI: 0.51–2.15; p = 0.891), the luminal B/HER2− (HR 1.30; 95% CI: 0.94–1.79; p = 0.113) and the luminal B/HER2+ subtypes (HR 1.34; 95% CI: 0.89–2.01; p = 0.166) (Fig. 4a).

Fig. 4
figure 4

a Overall survival (OAS) from date of metastatic disease of patients with only one site of metastasis. b Overall survival (OAS) from date of metastatic disease of patients with three or more different sites of metastases stratified by subcategories of breast cancer

Full size image

Interestingly, the OAS from date of metastatic disease of patients with three or more different sites of metastases is the best for patients with the luminal B/HER2+ subtype (HR 0.82; 95% CI: 0.51–1.33; p = 0.430). We observed no significant differences in the luminal A and the luminal B/HER2− subtype (HR 1.23; 95% CI: 0.83–1.82; p = 0.294). The HER2-overexpressing (HR 1.82; 95% CI: 1.00–3.32; p = 0.050) and the TNT subtypes (HR 1.90; 95% CI: 1.22–2.94; p = 0.004) had a significantly poorer OAS in this case (Fig. 4b).

Consecutively, we examined OAS from date of metastatic disease according to the different sites of metastases. Patients with the TNT subtype and liver metastases (HR 3.08; 95% CI: 2.11–4.49; p < 0.001) had the worst OAS compared to the luminal A subtype. An almost equal OAS existed for patients with the luminal B/HER2+ subtype (HR 0.98; 95% CI: 0.66–1.45; p = 0.919) with a slightly worse OAS for the luminal B/HER2− (HR 1.30; 95% CI: 0.93–1.77; p = 0.129) and the HER2-overexpressing subtype patients (HR 1.55; 95% CI: 0.98–2.45; p = 0.059) (Fig. 5a).

Fig. 5
figure 5

Overall survival (OAS) from date of metastatic disease of patients with liver metastases (a), lung metastases (b), bone metastases (c) and CNS metastases (d) stratified by subcategories of breast cancer

Full size image

OAS of patients with pulmonary metastasis was the best for the luminal A subtype, followed by the luminal B/HER2− (HR 1.23; 95% CI: 0.81–1.86; p = 0.338), the luminal B/HER2+ (HR 1.34; 95% CI: 0.80–1.86; p = 0.226) and the HER2-overexpressing subtypes (HR 1.92; 95% CI: 1.09–3.38; p = 0.025); the TNT subtype had the worst OAS (HR 3.07; 95% CI: 0.94–0.79; p = 0.113) (Fig. 5b). We also demonstrated that the TNT subtype with bone metastases always had the worst OAS (HR 3.77; 95% CI: 2.66–5.350; p < 0.001). OAS is the best for patients with the luminal A subtype followed by the luminal B/HER2+ (HR 1.40; 95% CI: 1.00–1.96; p = 0.048), the luminal B/HER2− (HR 1.57; 95% CI: 1.22–2.02; p = 0.001) and the HER2-overexpressing subtypes (HR 2.03; 95% CI: 1.33–3.11; p = 0.001) (Fig. 5c).

Our analysis of CNS metastases demonstrated a slightly better OAS for patients with the luminal B/HER2+ (HR 0.89; 95% CI: 0.49–1.60; p = 0.350) subtype than for patients with the luminal A subtype. Similar is the OAS of patients with the HER2-overexpressing subtype (HR 1.05; 95% CI: 0.51–2.15; p = 0.891) followed by a slightly worse OAS for the luminal B/HER2− subtype (HR 1.32; 95% CI: 0.74–2.35; p = 0.350) (Fig. 5d).

Figure 6 shows the OAS of patients with metastatic BC treated in the adjuvant setting. OAS from the date of primary diagnosis of breast cancer is the best for luminal A, followed by luminal B/HER2− (HR 1.43; 95% CI: 1.16–1.77; p = 0.001), luminal B/HER2+ (HR 1.50; 95% CI: 1.14–1.97; p = 0.001) and HER2-overexpressing subtypes (HR 1.97; 95% CI: 1.46–2.67; p < 0.001). The TNT subtype (HR 2.86; 95% CI: 2.26–3.62; p < 0.001) has a significantly lower OAS (Fig. 6).

Fig. 6
figure 6

Overall survival (OAS) of patients with advanced breast cancer from date of primary diagnosis of metastatic disease

Full size image

Table 3 summarizes the OAS and the metastasis-free survival (MFS) of the patients with advanced BC from date of primary diagnosis. Here, the MFS was 23 months (95% CI: 19.94–26.06; SEM = 1.56) for patients with luminal A, followed by luminal B/HER2− subtype with 19 months (95% CI: 15.83–22.17; SEM = 1.62). Interestingly, the MFS (median) of the luminal B/HER2+ subtype was only 15 months (95% CI: 11.38–18.62; SEM = 1.85), only 14 months for the HER2-overexpressing (95% CI: 9.25–18.75; SEM = 2.42) and only 13 months for the TNT subtype (95% CI: 10.76–15.42; SEM = 1.14) (Table 3).

Table 3 The overall survival (OAS) and the metastasis-free survival (MFS) of patients with advanced breast cancer from date of primary diagnosis
Full size table

Discussion

Metastasis of BC occurs in 20–30% of BC patients [31, 32]. It turns the local tumor growth into a systemic disease for lifetime [33]. In recent years, at least since the 12th International Breast Cancer Conference in St. Gallen, BC is no longer seen as a uniform disease but is divided into different subgroups according to the gene expression patterns [7]. Therefore, many biological characteristics and tools of BC are a focus of research. Molecular aspects such as gene expression profiles [34], microRNAs [35], circulating tumor DNA [36], circulating tumor cells [37, 38], tumor stem cells [39, 40], different signaling pathways in tumor cells [41], and different aspects of the environment with immune cells like myeloid-derived suppressor cells and regulatory T cells [42, 43] are playing increasingly important roles.

Therefore, it was a priority for us to examine closely the distinct pattern of metastasis of different BC subtypes to create a basis for understanding the biological characteristics. The surrogate definition [8, 9] was used because the KI-67 labeling, which differs between the luminal A and B subtype, was unavailable in the BRENDA specimen [7, 10] as in the previous literature [13]. Altogether, 9.2% of the patients developed distant metastases during the observation period. This is slightly less than that described in the literature [32]. This difference could be due to the variable time periods (1992–2008 vs. 1973–2003) and to the associated different adjuvant therapies.

Next, we analyzed the metastatic pattern depending on the subtypes. Pogoda et al. described metastasizing TNT BC: 15% in the CNS, 14% in the lungs, 11% in the skeleton, 8% in the liver and 14% loco-regional [44]. This corresponds to our data, which describe more visceral, lung and CNS metastases of the TNT than of the luminal A subtype. In a recent study, the metastatic pattern of the different subtypes in Germany was described with a somewhat different classification of distant metastases in “bone”, “visceral”, “bone and visceral” and “brain (with/without bone or visceral metastases)” metastases. This is comparable to our data [13]. The same applies to other literature [17, 19, 22, 45, 46]. Smid and coworkers observed slightly different data which may be due to some modifications in the subtype classification [47].

Already known are the following results of several actual trials: with respect to the first site of metastasis, the HER2-overexpressing subtype had a highly significant rate of liver metastases and the TNT subtype a greater risk of developing lung metastases than the luminal A subtype. CNS metastases seem to be associated with the HER2-overexpressing subtype; bone metastases are more frequent in luminal A subtypes [12]. In contrast to our data, Kast et al. showed bone-only metastases to be more likely in luminal A and luminal B/HER2− subtypes. Visceral metastases were associated with HER2-overexpressing, TNT, luminal B/HER2+ subtypes, while CNS metastases were associated with luminal B/HER2+ and TNT subtypes [13]. In this case, the results of Dent et al. match well with our data because they observed that visceral metastases were four times more likely to develop in the TNT subtype [18]. On the other side, Kennecke et al. described that HER2-positive subtypes were more likely to metastasize in the brain, liver and lungs [22]. These literature data are congruent to our results.

Kast and coworkers described the metastatic pattern of the different subtypes of BC for the first time in 304 metastasized patients in Germany between 2006 and 2011 [13]. In our study, the number of patients was higher with 886 patients from 1992 to 2008. Kast and coworkers analyzed data of BC patients treated in four different clinics in Dresden and Radebeul. Our study involved a greater overall area, the Department of Gynecology and Obstetrics at the University of Ulm and 16 partner clinics (all certified breast cancer centers) in Baden-Wuerttemberg in Germany. Therefore, our data could well complement the results from Dresden [13].

In the RegistHER2 trial, they found one site of metastasis in 51.3% of patients with the HER2-positive subtypes, two sites in 25.8%, three sites in 14.5% and four or more sites of metastasis in 8.3% [48]. The number of metastasis sites has already been analyzed for patients with brain metastases: 60.9% of the luminal subtype, 73.7% of the luminal B subtype, 64.9% of the HER2-positive subtype, but only 6.81% of the TNT subtype with brain metastasis had three or more sites of metastasis [49]. These data were supplemented by our investigations.

Lobbezoo and coworkers described the best survival with metastases in the HR-positive/HER2-positive subtype and the worst in the TNT subtype, independent of the site of metastasis, from 2007 to 2009 [46]. In contrast, Lee et al. analyzed a not significantly better survival of HR-positive subtypes compared to TNT or HER2-positive subtypes [50]. Significant differences in the median survival from time of first metastasis were described by Kenneke et al.: the longest survival was seen for the luminal A subtype, followed by the luminal B, the luminal/HER2, the HER2-overexpressing, the basal-like and the TNT non-basal subtype [22]. This coincides well with previous results [12] and our analyzed data. An equal median best survival was shown by Kast and coworkers for HER2-positive and luminal A subtypes [13]. TNT subtypes had the worst survival. The described differences in the HER2-positive subtypes could be due to the availability of specific HER2-positive-directed therapies [51–53].

Altogether, the best OAS in our study is for patients with the luminal A subtype, the worst is for patients with the TNT subtype. The survival of patients with liver metastases has been described to be the poorest for the TNT subtype [54]. A better prognosis for patients with HR-positive subtypes was described previously [55, 56] and matches our data. Although, we could only show the significant difference in OAS between the TNT and the luminal A subtype, Ge and coworkers failed to show any significant differences between the different subtypes [57]. The data of patients with lung metastases analyzed by Yhim et al. showed the best survival parameters for HR-positive subtypes and the worst for HER2-positive and TNT subtypes, corresponding to our data [58]. The survival of patients with bone metastases was observed by Lee et al. to be longer in HR-positive than in HER2-positive and TNT subtypes [50]. These results complement again our analysis. Regarding brain metastasis, a poorer OAS of the HER2-positive and the TNT subtypes was described previously [59, 60]. In contrast, Arslan et al. described a slightly better survival for the luminal B, followed by the HER2-positive, the luminal A and the TNT subtypes [61]. There were some differences to our data, maybe resulting from the fact that we observed the OAS from date of CNS, rather than brain metastases.

The analysis of the MFS of the different subtypes of BC showed very interesting results. The median MFS of the luminal B/HER2+ subtype, the HER2-overexpressing and the TNT subtypes is 13–15 months. This is a very short interval regarding the tumor doubling time. Recently, Coumans and coworkers analyzed a tumor doubling time of 1.7  ±  0.9 months in more than 38,000 BC patients. Furthermore, they summarized the previous literature without shorter tumor doubling times [62]. The tumor volume doubling time of primary BC tumors analyzed by mammograms or ultrasound was at least 100 days [63, 64]. Ryu and coworkers detected a 2.4-fold shorter tumor volume doubling time for TNT subtypes compared to ER-positive tumors [63]. The relatively short MFS of luminal B/HER2+, HER2-overexpressing and TNT subtypes leads to the question of whether there were really no metastases at the time of primary diagnosis. We recommend thorough staging examinations for patients with these subtypes always, and repeating the diagnostics immediately in case of doubt.

Conclusion

The different subtypes of BC had a typical pattern of metastasis: the subtypes of BC, the sites and the numbers of metastases were pivotal for prognosis. The better knowledge of the biology of BC subtypes will help to personalize therapeutic options and follow-up examinations.