|Home | About | Journals | Submit | Contact Us | Français|
Metastatic breast cancer is the most frequent cause of cancer death for women worldwide. In the last 15 years, a large number of new agents have entered clinical use, a result of the dramatic increase in our understanding of the molecular underpinnings of metastatic breast cancer. However, while these agents have led to better outcomes, they are also at the root cause of increasing financial pressure on healthcare systems. Moreover, decision making in an era where every year new agents are added to the therapeutic armamentarium has also become a significant challenge for medical oncologists. In the present article, we will provide an ample review on the most recent developments in the field of treatment of the different subtypes of metastatic breast cancer with a critical discussion on the slow progress made in identifying response biomarkers. New hopes in the form of ctDNA monitoring and functional imaging will be presented.
Breast cancer (BC) is the most common cancer type, as well as the first cause of cancer death among women worldwide, with an estimated 1.7 million new cases and 521,900 deaths in 2012 (Torre et al., 2015). Though most women present with localized potentially curable tumours, incurable and lethal relapses and de novo metastatic breast cancer (MBC) remain frequent in clinical practice (Welch et al., 2015).
The classic paradigm of MBC treatment, i.e. decision making that is based on pathological (hormonal receptor status and HER2 status) and clinical (patterns of dissemination, disease burden and presence/absence of symptoms) parameters, that we may call “stratified oncology” has not significantly changed in recent years (Cardoso et al., 2012, 2014). Meanwhile, advances in translational research have generated exponential growth in our understanding of the molecular underpinnings of MBC, including the characterization of molecular subtypes (Perou et al., 2000), discovery of numerous potential therapeutic targets and mechanisms of resistance to treatment (Wang et al., 2011; Stendahl et al., 2004; Pohlmann et al., 2009). Both in parallel and connected to these advances at the bench, in the clinic the therapeutic arsenal available for metastatic patients has increased dramatically.
At the end of the 20th century, available chemotherapy regimens provided a maximum progression free survival (PFS) of 10 months and an overall survival (OS) that rarely exceeded 20 months (Fossati et al., 1998). Later, newer chemotherapeutic agents pushed survival to above 20 months and distinctive indications for sequential versus combined use were developed, with the former becoming the standard, unless response rate was of the essence (Dear et al., 2013). Though recent years have seen a number of new classic cytotoxic agents such as eribulin come into use with positive results, it is the coming of age of targeted therapy, spearheaded by trastuzumab in the late 90s that has transformed and will continue to transform management of MBC in the coming years (Thomas et al., 2007; Mendes et al., 2015; Cortes et al., 2011).
Though “more and better drugs” is one of the paths towards better outcomes, simply adding more drugs does not solve the entire equation. Better matching of drug to patient, through the development of efficient biomarkers – a concept dubbed “personalized” or “precision” medicine – has the potential to both improve results and to reduce unnecessary treatment (and thus toxicity). However, at this juncture, frustratingly and despite intensive research and some hopeful candidates (Lee et al., 2016), the only predicative biomarkers in current clinical use backed by solid evidence remain HER2 and oestrogen receptor status (Table 1). For healthcare systems strained by rising treatment costs (Mariotto et al., 2011), biomarkers may improve the cost‐effectiveness of treatment, allowing the adoption of new treatments that would otherwise be considered too costly (Elsada et al., 2016). At the same time, improvements in functional imaging studies may better evaluate both the effect of and response to treatment, allowing for more precise decision making on the part of oncologists than is possible at the present time (van Kruchten et al., 2015).
Current treatment options, biomarkers and future developments.
In the present article, we will provide a review of the essential data that has led to the recent addition of new agents for MBC, discuss new treatment strategies for oligometastatic and overtly metastatic disease, as well as summarize to what extent oncologists can rely on “biomarkers” for the prescription of these expensive drugs.
Hormonotherapy (HT) is arguably the oldest form of target therapy and since the discovery of oestrogen receptor (ER), more than 50 years ago, major advances have occurred in the field (Jensen, 2004) (Table 1). Tamoxifen, a selective oestrogen receptor modulator, was the first compound that showed dramatic responses and a relatively good safety profile in patients with metastatic breast cancer (Ward, 1973). Agents that directly or indirectly target ER by different mechanisms, such as aromatase inhibitors (AI), luteinizing hormone releasing hormones (LHRH) agonists and the ER receptor degrader fulvestrant were also found to be effective in ER‐positive breast cancer (Klijn et al., 2001; Mauri et al., 2006). These new agents, added to the drug arsenal in the last decade may be slightly more effective than classic treatments options (Cardoso et al., 2013; Miller et al., 2007). Contrary to chemotherapy, in HT available results do not suggest that association of multiple agents is useful. Three trials evaluating non‐steroidal AI (anastrozole) and fulvestrant versus anastrozole alone as first line treatment showed conflicting results. While the SWOG 226 trial showed modest PFS gain of 1.5 months (Hazard ratio [HR] = 0.8, p = 0.007) and OS gain of 6.4 months (HR = 0.81, p = 0.049) in spite of 41% crossover, both the SOFEA and FACT trials did not show any advantage to the dual combination (Mehta et al., 2012; Bergh et al., 2012; Johnston et al., 2013). Perhaps the efficacy differences are related to the discrepancies in the treated populations between the studies. In the SWOG trial, prior adjuvant tamoxifen was provided to 40% of patients and de novo metastasis developed in 38%, while in the FACT trial, prior adjuvant tamoxifen was provided to 66% of patients and de novo metastasis developed in only 13%. This suggests that tamoxifen naïve patients without any development of acquired resistance may benefit from dual therapy at first line. Notably, these studies used the suboptimal dosage of fulvestrant 250 mg and not the current optimal dose of 500 mg which showed 20% reduction in risk of progression and was not associated with increased toxicity (Di Leo et al., 2010). Therefore, based on these trials dual anastrozole and fulvestrant is currently not indicated and sequential treatment remains the standard. Regarding the first line anti‐hormonal therapy of choice, in hormonotherapy naïve patients, the FIRST trial showed an advantage in terms of OS for the fulvestrant group over the anastrozole group (median OS, 54.1 versus 48.4 months, p = 0.04) (Ellis et al., 2015). Results of a large phase 3 confirmatory trial are eagerly awaited. In parallel, a new generation of oestrogen receptor degraders is being developed, such as RAD1901 (Garner et al., 2015). In pre‐clinical in vitro and in vivo murine models, these drugs appear to be at the same time more efficient and less toxic than fulvestrant. They are at the moment in phase 1 development (Harb et al., 2015).
Defining endocrine resistance is of great importance not only for treatment strategy but also for pragmatic stratification of patients entering in trials for advanced breast cancer. It is now agreed, based on expert opinion consensus, that patients progressing during the first two years of adjuvant hormonal therapy or during the first three months of hormonal therapy in the metastatic setting, should be classified as “primary endocrine resistant” (Cardoso et al., 2014). Different paths lead to endocrine resistance, including: (1) loss of the ER receptor; (2) mutations in the ER receptor; (3) up regulation of alternative signal transduction pathways (such as HER2). Extensive crosstalk between pathways form a true “web” of alternative options for continued proliferative signalling, granting flexibility and resilience to cancer cells (de Anda‐Jáuregui et al., 2015).
ER negativity is the de facto reason of primary endocrine resistance and acquired endocrine resistance develops when around 25% of patients loose ER expression during disease progression, highlighting the importance of biopsy and ER evaluation of metastatic lesions at first appearance (Gutierrez et al., 2005; Johnston et al., 1995). Some mutations in ER are associated with hormonal resistance and are enriched in metastatic samples compared with primary cancers. Most of these mutations will ultimately lead to increased transcriptional activity of the pathway, albeit by different mechanisms (Jeselsohn et al., 2014; Giguère, 2014; Segal and Dowsett, 2014). Oestrogen receptor (ESR1) mutations, as a central mechanism of partial or complete resistance to endocrine therapy, have great potential as biomarkers to predict response and guide treatment choices (Witzel et al., 2016). Mutations can indeed lead to independent activation that may resist inhibition (Weis, 1996; Li et al., 2013) and frequency is correlated with the extent of exposure to hormonotherapy and more specifically aromatase inhibition (Segal and Dowsett, 2014). Since the resistance they convey is often partial, their detection may point, after prospective trials, to the necessity of changing regimens or to increase dosage (Jeselsohn et al., 2015). Up regulation of alternative signal transduction pathways such as the EGFR/HER2, PIK3CA/mTOR was also identified as predictors of resistance to hormonotherapy. Blocking both ER and HER2 was associated with modest significant improvement in PFS in comparison to hormonal therapy alone in two randomized trials (Johnston et al., 2009; Kaufman et al., 2009). This association may be useful in the treatment of fragile patients and/or those with a low disease burden, for whom the physician would prefer to avoid chemotherapy.
Another interesting attempt to overcome ER resistance by blocking the EGFR/HER receptor was performed in a trial that combined gefitinib (EGFR tyrosine kinase inhibitor) and tamoxifen (Osborne et al., 2011). Unfortunately the combination was not found to be superior to tamoxifen although some positive signal was seen in endocrine therapy naïve and HER2+ patients. The usefulness of targeting the phosphatidylinositol 3‐kinase (PI3K)/AKT/mammalian target of rapamycin (mTOR) pathway was tested in the BOLERO‐2 study: the mTOR inhibitor everolimus and exemestane were compared to exemestane alone in patients with MBC who had failed previous AI. While PFS was superior for the combination arm (HR = 0.36, p < 0.001), no differences were seen in OS (Piccart et al., 2014; Baselga et al., 2012). Significantly, 11% of patients in the combination arm discontinued treatment due to adverse events in contrast to only 1% in the control arm, showing that, though clever, the concept of bypassing resistance by multi‐targeting may deprive hormonotherapy of its traditionally mild toxicity profile. The BOLERO‐2 results highlight the need for useful biomarkers that will identify the subset of patients that may derive substantial benefit from the addition of mTOR inhibitor. There are no validated biomarkers that predict response to everolimus (Lee et al., 2015a), at the present time, and research by Hortobagyi and colleagues using next generation sequencing, looking into 182 cancer associated genes in the BOLERO‐2 population did not pick up any useful positive predictive signals (Hortobagyi et al., 2013). However, women with two or more alterations in CCND1, PIK3CA, FGFR1/2 or PTEN, did not benefit from everolimus treatment. More recently, in the BELLE‐2 phase 3 trial of buparlisib, a pan‐PI3K inhibitor in association with fulvestrant versus fulvestrant, the association was marginally superior in terms of PFS (6.9 versus 5). Interestingly, the presence of PI3K mutant circulating tumour DNA (ctDNA) predicted an enhanced benefit (Baselga et al., 2015).
The relative success of the BOLERO strategy led to expanded interest in alternative proliferative pathways. Though most ER positive tumours are relatively “indolent”, a small subset has a generally lower Allred score, higher KI67 index and a more aggressive natural evolution (Luminal B), often proving to be primarily resistant to HT or only transiently sensitive (Ades et al., 2014). In these tumours, the dysregulation of the cyclin dependent kinases (CDK), a family of molecules that play a vital role in controlling the frequency of cellular division is an often exploited proliferation pathway (Malumbres and Barbacid, 2009). Indeed several clinical trials are currently evaluating selective CDK 4–6 inhibitors (ribociclib, abemaciclib, palbociclib) with HT in BC (Finn et al., 2016). The results of the PALOMA‐1 randomized phase 2 study, which tested palbociclib in combination with letrozole versus letrozole alone as first‐line treatment of ER‐positive, HER2‐negative, advanced breast cancer were recently reported (Finn et al., 2015). The addition of palbociclib to letrozole in this study, significantly improved PFS (20.2 versus 10.2 months, HR = 0.48, one‐sided p = 0.0004). Biomarker analysis, based on pre‐clinical studies showing that reduced expression of p16 and increased expression of cyclin D1 predicted greater sensitivity to palbociclib, did not result in positive correlations. The results of PALOMA 1 were recently confirmed in the phase 3 PALOMA 2 trial, with a median PFS was 24.8 versus 14.5 months (HR = 0.58 95 IC 0.46–0.72, p < 0.000001). The PALOMA‐3 study that compared palbociclib and fulvestrant to fulvestrant alone in patients who had progressed during hormonotherapy or relapsed after adjuvant treatment showed a PFS advantage of 5.4 months (Turner et al., 2015). There was no correlation between efficacy and oestrogen receptor expression levels nor presence of PI3K mutations (Cristofanilli et al., 2016). Toxicity was again enhanced with neutropenia in above 70% of patients in both trials, though mechanisms seem to be distinct from those of chemotherapy (Hu et al., 2015). Considering the above results, association of hormonotherapy and targeted agents are clearly a new way forward, and, though no head to head comparisons are available between everolimus and palbociclib, the toxicity profile of palbociclib and the lack of OS benefit with either compound at the moment favour a sequence with palbociclib first, followed by everolimus.
The resounding success of trastuzumab has changed the prognosis of HER2 positive MBC, but both primary and secondary resistance to trastuzumab remain very significant issues, leading to the development of strategies to bypass it.
Cancer cells that exploit the HER2 pathway proliferate and invade very effectively, a fact that explains the traditionally aggressive behaviour of these tumours. In the face of anti‐HER2 therapy, different pathways to resistance are available: (1) reduced binding between trastuzumab and HER2; (2) alternative activating mechanisms at receptor level or downstream; (3) crosstalk; and (4) failure to trigger immune‐mediated mechanisms to destroy tumour cells (Pohlmann et al., 2009; Advani et al., 2015).
To face this issue, a new generation of anti‐HER2 drugs have been created (Table 1). Lapatinib, a tyrosine kinase inhibitor, bypasses resistance by targeting the intracellular domain of the HER2 receptor and has been in clinical use for many years following positive trials in combination with capecitabine and with trastuzumab (Blackwell et al., 2010; Geyer et al., 2006). Its relevance in clinical practice has been recently diminished by newer agents, such as the anti‐heterodimerization (HER2/HER3) monoclonal anti‐body pertuzumab, that has come into clinical use after the landmark CLEOPATRA trial (Advani et al., 2015; Giordano et al., 2014). CLEOPATRA is a phase 3 study that evaluated adding pertuzumab to trastuzumab and docetaxel in first‐line MBC. Both PFS and OS were in favour of the pertuzumab group (OS 56.5 versus 40.8 months) (Swain et al., 2013). The combination of paclitaxel, trastuzumab and pertuzumab is another valid option (Dang et al., 2014).
In an attempt to overcome HER2 resistance by blocking the mTOR pathway, two trials (BOLERO‐1 and BOLERO‐3) were performed. BOLERO‐1 evaluated the addition of everolimus to trastuzumab + paclitaxel regimen in the first‐line setting (Hurvitz et al., 2014). BOLERO‐3 evaluated the addition of everolimus to trastuzumab + vinorelbine in the second‐line setting. While BOLERO‐3 demonstrated advantage in terms of PFS from 7 versus 5.8 months (HR = 0.78, p = 0.006) to the addition of everolimus, BOLERO‐1 did not confirm the benefit (Segal and Dowsett, 2014; Witzel et al., 2016). Interestingly, the BOLERO‐3 exploratory translational analysis suggested increased benefit from mTOR inhibition in patients with ER negativity (HR = 0.65) or PTEN loss (PFS 9.6 versus 5.3 months, p = 0.01) (André et al., 2014a). This led to a protocol amendment, before the data were unblinded, for analysis in the BOLERO‐1 in ER negative subgroup of patients which showed PFS of 20.2 versus 13.1 months (p = 0.0049) in favour of the everolimus addition, but it did not pass the predefined statistical significance threshold. In an exploratory analysis, candidate biomarkers namely high phosphorylated S6 (pS6), low PTEN, and PI3KCA mutations were evaluated. Results suggest that use of everolimus may be useful for patients with low PTEN or high pS6 levels (Jerusalem et al., 2013). Since there are better treatment alternatives, for both first line (pertuzumab containing regimen) and second line (TDM1), the results of the BOLERO 1 and 3 trials have not changed clinical practice. In another attempt to overcome resistance to anti‐HER2 therapy, the LUX‐breast 1 trial added the pan‐HER blocker afatinib in association with vinorelbine or trastuzumab and vinorelbine in patients who had previously received trastuzumab as adjuvant or first line metastatic treatment. The study was prematurely closed as toxicity in the afatinib arm was significant, and because the chance of establishing the superiority of the afatinib arm was judged to be minimal (Harbeck et al., 2016).
TDM1 or the addition of the highly toxic agent emtansine to trastuzumab, exploits HER2 overexpression as a path towards identification of cancer cells and selective delivery of chemotherapy (LoRusso et al., 2011).
The EMILIA trial showed that TDM1 significantly prolonged PFS and OS with less toxicity than lapatinib plus capecitabine in HER2+ advanced breast cancer previously treated with trastuzumab and a taxane. Evaluation of biomarkers showed that all subgroups analyzed had longer PFS and OS with TDM1 (Baselga et al., 2013). While in CLEOPATRA prior trastuzumab exposure was minimal (10%) and interval of ≥12 months from last trastuzumab (as adjuvant therapy) was required, in the EMILIA study, 100% of the patients were previously exposed to trastuzumab and if previously treated in the adjuvant setting a trastuzumab free interval of ≤6 months was mandatory (16% of patients). Results of two other phase 3 trials evaluating TDM1 were recently reported. The THERESA trial compared TDM1 to physician choice in previously treated patients. Again TDM1 showed superiority for both PFS and OS (PFS 6.2 versus 3.3 months, p < 0.0001), strengthening its role in trastuzumab treated patients (Krop et al., 2014). The final OS results showed significant advantage in the TDM‐1 arm, with 22.7 versus 15.8 months in the physician choice arm (Wildiers et al., 2015). In another trial (MARIANNE), 1095 patients with HER2+ breast cancer were randomized to first line treatment with TDM1 versus TDM1 + pertuzumab versus trastuzumab and taxane. The study met its non‐inferiority endpoint, showing similar PFS among the three arms but did not meet PFS superiority endpoint for TDM1 containing regimens. These results do not impact the current indications of pertuzumab and TDM1 in advanced HER2‐positive breast cancer.
The designs of the CLEOPATRA and EMILIA trials pose problems for determining the exact benefits of pertuzumab and TDM1 at the present time, since: (1) unlike in CLEOPATRA most patients with HER2 positive MBC have received adjuvant trastuzumab; (2) patients going into the second line are now receiving pertuzumab, and thus EMILIA results may not apply to them. These questions are unlikely to be resolved through new prospective trials, and thus several groups have started to build HER2 positive cancer registries to seek theses answers retrospectively (Tripathy et al., 2014). Therefore, for first line therapy or trastuzumab naive patients the current recommendation is pertuzumab + trastuzumab and docetaxel, and for second line and trastuzumab exposed patients, TDM1 is preferred.
Despite the advances in terms of number of agents, at the present time no biomarker other than HER2 status itself is available to guide treatment choice in these patients. This has led to increased interest and thus major trials such as the ones previously reported have included the search for predictive biomarkers within their protocols. In a recent report on the EMILIA trial sample, several biomarkers were evaluated in tumour tissue, including quantitative PCR analysis of EGRF, HER2, and HER3, as well as PTEN and PI3K mutations and PTEN expression. TDM1 was equally superior to lapatinib/capecitabine in all subgroups. Interestingly, PI3K mutations were associated with shorter PFS and OS for the lapatinib/capecitabine arm (Baselga et al., 2016). Similar results were obtained in a biomarker evaluation of CLEOPATRA, that demonstrated that pertuzumab consistently showed a PFS benefit, independent of biomarker subgroups (ER, HER2, HER3, and PIK3CA). PI3K wild‐type patients, however, had twice the PFS of PI3K mutated patients (Baselga et al., 2014). Studies focused on trastuzumab use in MBC showed that greater HER2 amplification (HER2 copy number and ERBB2/centromere) benefit more from trastuzumab (Fuchs et al., 2014; Gullo et al., 2009). Other markers have been studied in small retrospective studies, such as EGFR (a negative prognostic factor but not predictive of trastuzumab failure) (Lee et al., 2015b), MET and EGF copy number (both negative predictor of response to trastuzumab) (Minuti et al., 2012). While all of these studies are tissue based, with all the limitations that this entails, one interesting recently published study looked into circulating serum HER2 (actually the cleaved extracellular domain part of the HER2 receptor (sHER2). With a total of 1902 patients from 3 different randomized trials that tested lapatinib versus non‐lapatinib regimens in MBC, PFS improved by 3.4% for every 10 ng/ml increase of sHER2. Though the results of this study must be prospectively validated, it is of particular interest because of its evaluation on sHER2, a quantifiable and dynamic biomarker, through a theoretically simple to perform ELISA test (Lee et al., 2016).
The success of available anti‐HER2 therapies in prolonging survival has also increased the proportion of patients that suffer from central nervous system dissemination, either in the form of solid brain metastasis (BM) or meningeal carcinomatosis (McKee et al., 2016; Shen et al., 2015) but have visceral disease under control. Besides traditional local forms of therapy, two additional forms of therapy have generated interest – systemic treatment and intrathecal trastuzumab administration.
Emphasis on the potential benefit of systemic treatment is based on results suggesting that patients who received trastuzumab with BM have improved survival (Park et al., 2008). Examples of agents tested include the pan‐HER2 inhibitor neratinib which reduced incidence of BM by 50% in a recent phase 2 trial (Awada et al., 2016). Lapatinib has been the focus of particular interest in this setting as it crosses the blood–brain barrier, with a response rate of 30% when given with capecitabine in a small phase 2 trial (Bachelot et al., 2013). In the larger CEREBEL phase 3 study of lapatinib/capecitabine versus trastuzumab/capecitabine in the prevention of BM no benefit was detected. At the present time, systemic therapy for BM remains experimental and there are no special recommendations in the HER2 positive subgroup. For meningeal carcinomatosis, intrathecal trastuzumab has been evaluated with some evidence of positive effect, though prospective trials are needed (Zagouri et al., 2013; Park et al., 2015).
Triple negative breast cancer (TNBC) is an heterogeneous group of different tumours that suffer from lack of targetable receptors, and while prognosis of ER+ and HER2+ cancers improved during the last decade, that of TNBC remains dismal (Metzger‐Filho et al., 2012), with chemotherapy as largely its only treatment option. So far the only targets for treatment that have tangible, although limited results, have been angiogenesis and DNA repair, though other pathways are being studied, such as the androgen receptor (Jamdade et al., 2015; Kalimutho et al., 2015; Palma et al., 2015). Immunotherapy may in the coming years yield positive results in TNBC, as 20% of these tumours express PDL1 (Mittendorf et al., 2014) but it is at this stage strictly experimental, with the ongoing Keynote 012 phase 1 trial showing promising results (Buisseret et al., 2015; Pusztai et al., 2016).
A few new chemotherapeutic drugs have shown activity in MBC in the recent past years including ixabepilone, eribulin, nabpaclitaxel, etirinotecan and vinflunine (Ribeiro et al., 2012). Eribulin in the EMBRACE phase 3 trial against physician's choice of treatment in heavily pre‐treated patients showed an OS gain of 2.5 months, leading to its FDA approval. Nab‐paclitaxel has proved itself slightly superior to regular paclitaxel in one phase 3 trial, with a special advantage in terms of less adverse events. One phase 2/3 trial is ongoing, on a TNBC‐only population, comparing different combinations of active chemotherapy agents in the first line (Gradishar et al., 2005; Yardley et al., 2015), with the goal of establishing a standard regimen for first line treatment. Despite there being no specific interaction between TNBC and these drugs, as no alternatives are available these drugs are especially useful in TNBC treatment (André and Zielinski, 2012). Though active, etirinotecan failed to prove its superiority to physicians' choice of treatment in the recent BEACON trial. However, in a pre‐planned subgroup analysis, etirinotecan had interesting activity in the subgroup of patients with CNS disease, with a median OS of 10 months in the experimental arm versus 4.8 months in the control arm (Perez et al., 2015).
Despite results that are sometimes excellent in TNBC, chemotherapy, especially when taxane and platinum based, must be often stopped due to toxicity concerns. Therefore, in patients with proven responsiveness to chemotherapy, the strategy of maintenance or “early second line” may be valuable, using a regimen that is less toxic. The IMELDA trial focuses on this strategy, comparing sole antiangiogenic therapy (bevacizumab) with bevacizumab and capecitabine, in a population of patients who responded to 3–6 cycles of docetaxel and bevacizumab and showed both PFS (11.9 versus 4.3 months HR = 0.38, p < 0.0001) and OS (39.0 versus 23.7 months HR = 0.43, p = 0.0003) benefit in the capecitabine and bevacizumab arm (Gligorov et al., 2014).
The capacity to induce angiogenesis is one of the core characteristics of cancer cells, with a specially pronounced role in TNBC (Trédan et al., 2015). Bevacizumab, an anti‐VEGF anti‐body, may for this reason be useful in TNBC (Table 2). As demonstrated in a meta‐analysis of three randomized phase 3 trials, bevacizumab improved PFS but not OS when combined with first‐line chemotherapy for HER2‐negative metastatic breast cancer. In the subgroup of patients with TNBC, the PFS HR was 0.63 (95% CI 0.52–0.76, p < 0.0001) and one‐year survival rates were 71% with bevacizumab‐containing therapy versus 65% with chemotherapy alone (Miles et al., 2013a). This possible benefit from bevacizumab in TNBC was also observed in another analysis (Rossari et al., 2012). The RIBBON‐2 trial which evaluated bevacizumab with investigator choice of chemotherapy in previously treated MBC showed again improved PFS for the bevacizumab with suggested stronger effect for TNBC patients and for patients who received bevacizumab with taxanes (Brufsky et al., 2011). TANIA evaluated bevacizumab plus chemotherapy versus chemotherapy alone in second‐line treatment cancer after first‐line treatment with bevacizumab plus chemotherapy in HER2 negative population. The PFS was 6.3 versus 4.2 months (HR = 0.75, p = 0.0068) in favour of the bevacizumab arm (von Minckwitz et al., 2014).
Pivotal clinical trials supportive of the role of bevacizumab in triple negative breast cancer.
Altogether, these trials suggest that bevacizumab plays a modest role in the first and second line of metastatic ER negative breast cancer (Table 1). Ramucirumab, another antiangiogenic failed to show benefit in a large phase 3 trial with 22% of triple negative subpopulation (Mackey et al., 2015). As no good targeted therapies alternatives are available in TNBC, bevacizumab is reimbursed in few countries.
Can we better refine our selection of which patients to expose to expensive antiangiogenic agents, an issue that has become even more pressing after the withdrawal of approval for bevacizumab in MBC in the US and the UK? Several attempts have been made to look retrospectively and find biomarkers of response. Results suggest that VEGF‐A plasma levels correlate with a trend towards improved OS (Van Cutsem et al., 2012), an interesting candidate since there is an available ELISA test to dose it (Maru et al., 2013). In MBC, biomarker analysis of the AVADO study suggests that high baseline plasma levels of VEGF‐A and VEGFR‐2 are associated with greater treatment efficacy. Single nucleotide polymorphisms (SNP) analysis did not show any predictive value, contrary to previously published results in an analysis of the E2100 trial (Miles et al., 2013b; Schneider et al., 2008). The prospective phase 3 Meridian study is ongoing to evaluate the role of VEGF‐A as a predictive biomarker (Miles et al., 2013c).
5–10% of breast cancers are associated with a germinal mutation in the BRCA1 and/or BRCA2 genes. Approximately 70% of BRCA1 and 20% of BRCA2 mutated BC present as TNBC (Metzger‐Filho et al., 2012). Patients with certain BRCA1 and 2 mutations are more sensitive to PARP inhibitors in accordance with the concept of synthetic lethality and showed promising results in clinical trials of different PARP inhibitors (Sonnenblick et al., 2015). Currently the OlymipiAD and BRAVO trials recruit BRCA positive, TNBC (or HER2 negative) patients to randomized phase 3 trials that compare olaparib or niraparib, respectively, to physician's choice. Two interesting smaller phase 2 trials are also in recruitment. In the Ruby trial (NCT02505048), rucaparib will be given to MBC cancer patients that either have BRCA1 or 2 somatic mutations or that are positive for BRCAness as determined by a genomic signature that has already been tested in ovarian cancer with positive results (McNeish et al., 2015). BRCA‐like tumours share the core molecular characteristic of the tumours that arise in germline BRCA mutation carriers – a deficiency in the process of double‐strand break repair by homologous recombination and thus may respond to similar therapeutic approaches (Lord and Ashworth, 2016). Another approach is taken in the NCT02401347 (talazoparib beyond BRCA) trial of talazoparib in patients with MBC and somatic or germline mutations in a predefined set of homologous recombination (or associated) genes, excluding BRCA1 and 2. Thus in the coming years, PARP inhibitors may become a treatment option for BRCA germline mutated patients as well as for well selected non‐mutated patients.
Platinum chemotherapy which acts by generating double strand breaks in DNA may be especially useful in DNA repair impaired patients such as BRCA positive or BRCA like tumours. The TNT trial randomized TNBC (BRCA mutated or not) to docetaxel or carboplatin. 376 patients were accrued, with 8% BRCA mutated. Overall the trial was negative, but for the BRCA subgroup carboplatin was significantly more effective than docetaxel in terms of response rate (RR) (68.0% vs 33.3%, p = 0.03) (Tutt et al., 2014).
Tumour cells have varying capacities to form true metastases and this manifests clinically as the different possible presentations of MBC (Weichselbaum and Hellman, 2011). Though, in general, a significant number of patients present initially with a limited number of distant disease sites (Dorn et al., 2011), clinical evolution differs for many of these patients beyond this point, with some patients showing disease that rapidly spreads, while in other metastases progress slowly (Salama and Chmura, 2015). The existence of a subset of patients with a small number of metastatic lesion in one or two organs that are operable has led to attempts to cure metastatic disease, with some success, leading to such strategies becoming standard in colon cancer (Gallinger et al., 2013). Radiosurgery has also been use with some success (Tree et al., 2013) in the past, with ongoing prospective trials in BC being eagerly awaited to validate this strategy (Salama and Chmura, 2015). A new strategy that is being tested in trials is to combine radiotherapy with the new generation of immunotherapy to take advantage of the so called “abscopal effect” – local radiotherapy that induces long‐distance responses, a phenomenon likely connected to the capacity that radiotherapy has to present antigens and induce immune response (Reynders et al., 2015). However, not all patients that are submitted to local therapies are cured or even derive prolonged PFS or OS. The question is how can we predict in which patients oligometastatic disease is just the “tip of the iceberg” or “true oligometastatic” with slow development so that we can avoid using local therapies just because they are available (Treasure, 2012).
One traditional clinical strategy is “give it time” – systemic therapy and careful follow up of response patterns, with local treatment being used only for patients who have good and prolonged responses – a strategy that is, in MBC, corroborated by data. In one retrospective study with 86 MBC patients who were submitted to resection of hepatic lesions, predictive factors of prolonged survival after surgery were being ER positive, best response to pre‐surgery chemotherapy (with good responders deriving a larger benefit from surgery) and, for patients with post curative treatment relapse, a DFS of more than 2 years (Abbott et al., 2012). These findings also suggest that oligometastatic disease in BC, as genomic studies in different tumours have concluded, is biologically different from amply metastatic disease (Reyes and Pienta, 2015). In BC, there is evidence that 4 microRNAs, namely miR‐655‐3p, miR‐544a, miR‐127‐5p and miR‐369‐3p, all encoded in the 14q32 locus are consistently overexpressed in samples from oligometastatic patient who benefited from local therapies thus opening the door to possible development of biomarkers that can identify these patients through a more elegant method (Uppal et al., 2015).
Cancer evolution emerges as a consequence of accumulating driver mutations that can be different from disease site to disease site (spatial heterogeneity) as well as rise and disappear with time in Darwinian fashion (temporal heterogeneity) (Valastyan and Weinberg, 2011). A recent analysis of whole‐genome sequences of 560 breast cancers provided one of the largest overviews of the somatic genetic basis of BC. Ninety three mutated cancer genes implicated in genesis of the disease were identified, 12 base substitution mutational signatures and 6 rearrangement signatures were found (Nik‐Zainal et al., 2016). Refining our understanding of the genomic dependencies of tumours as well as the interactions between different driver molecular pathways will require an increased focus in the interaction between pre‐clinical and early clinical research.
The coexistence of different tumour clones with different sets of genetic aberrations suggests that therapeutic strategies targeted at predominant aberrations may not prove sufficient, especially in the metastatic setting were spatial and temporal evolution are more likely to be clinically significant from the start (Zardavas et al., 2015). Academic enterprises are therefore encouraged to build a longitudinal map of the clonal evolution of breast cancer that interrogates both primary and secondary disease sites. To achieve this, traditionally, multiple biopsies would be necessary (Gerlinger et al., 2012). Instead of re‐biopsy of different metastatic sites, a more convenient strategy could be blood based tests that analyze circulating tumour cells (CTCs) or circulating tumour DNA (ctDNA) (Crowley et al., 2013).
CTCs related research has contributed to the constant evolution of our understanding of the mechanisms of cancer progression, with a significant role in the construction of hypotheses that may explain clinically relevant phenomena such as late relapses and tumour heterogeneity (Pantel and Speicher, 2015; Schlange and Pantel, 2016). Moreover, CTCs are already well established prognostic factors in BC and can also be used, via protein and RNA expression, to define cellular subpopulation that diverge from the primary (Ignatiadis et al., 2015). Multiple groups have reported that HER 2+ CTCs can be detected in patients that have HER2 negative tumours in both the early and late settings (Ignatiadis et al., 2011; Fehm et al., 2007). This clear example of tumour heterogeneity exemplifies the potential of liquid biopsy, and has led to clinical trials of trastuzumab in such patients, including the ongoing Treat CTC study (in the early setting, though this instigating proof of concept trial could prove a powerful rationale for studies in MBC of a similar strategy). It is important to note, however, that trials evaluating the clinical utility of CTCs were not able to demonstrate yet an improvement in outcomes for this strategy (Ignatiadis and Piccart, 2012; Ignatiadis and Dawson, 2014). In parallel, a trial by Dawson and colleagues demonstrated that circulating tumour DNA could be an informative, specific, and highly sensitive biomarker of metastatic breast cancer (Dawson et al., 2013). The concept of the use of CTCs and or ctDNA to evaluate tumour genome in a dynamic manner during treatment is often called “liquid biopsy”. However, to leverage such new technologies in the setting of new drug development, new trial designs are paramount.
To further advance precision medicine, a new generation of clinical trials with innovative designs will most probably to a large extent replace the traditional paradigm of comparing a new cancer treatment chosen empirically versus the standard of care. Among these designs are included so called “basket” trials where on mutation/drug pair is tested on a range of different histologies and “umbrella” trials where multiple actionable mutation within one disease are matched to potential effective agents (Catenacci, 2015). In BC, examples of significant studies in this model include in the neoadjuvant setting in the I‐SPY‐2 study (Barker et al., 2009) and in the advanced setting in the SAFIR study (André et al., 2014b). Another example that combines both an innovative design with liquid biopsy is the “AURORA” trial recently launched by the Breast International Group (BIG) a large collaborative molecular screening programme, in which specimens, blood and plasma from 1300 patients with metastatic breast cancer will be analyzed by next generation sequencing for a panel of cancer‐related genes. Depending on the molecular profiles found, patients can be directed to clinical trials assessing molecularly targeted agents (Zardavas et al., 2014). Major obstacles remain, including regulatory issues, the complexities of trial design, the need for cooperation between different pharmaceutical companies as well as the limitations of currently available methods of genomic and transcriptomic analysis and of available targeted agents (Catenacci, 2015; Rodon et al., 2015). As the recent negative SHIVA trial has shown (Le Tourneau et al., 2015), limitations in our diagnostic tools, targeted agents and more importantly our understanding of the proliferative pathways may compromise innovative therapeutic strategies based on traditional or liquid biopsies.
Fundamentally, however, as precision medicine transitions from theory to practice, physicians should refrain from embracing non‐evidence based use of targeted therapies paired with micromanagement strategies of unproven clinical validity and utility. Interestingly, a recent study that investigated the current attitudes towards the integration of tumour genome sequencing in breast cancer management, showed that lack of evidence is still a major concern for wider use of a genotype‐driven approach in breast cancer (Gingras et al., 2016), suggesting an understanding among medical oncologists that, though promising, precision medicine has not yet fully come of age.
FDG PET/CT has become a widely used radiological test for diagnosis, staging and evaluation of response to therapy. In one study conducted in HER2 positive patients receiving neoadjuvant treatment (Gebhart et al., 2013), early (2 weeks) PET/CT metabolic response predicted pathologic complete response at surgery. A new generation of tracers with different affinities offers new options for functional imaging, with studies on ER positive tracer (18 F) fluoroestradiol (FES) and 89Zr‐trastuzumab (van Kruchten et al., 2015). These markers, besides potentially improving our understanding of tumour heterogeneity, may prove to be useful predictors of early clinical response and permit faster therapy changes in non‐responders.
In the ongoing ZEPHIR study, patients receiving TDM1 for HER2 positive MBC are submitted to baseline FDG PET and 89ZR‐trastuzuab PET (TRAS‐PET) as well as an early FDG PET after 2 weeks of treatment. The results obtained in the first 56 evaluable patients show that 1/3 of patients with supposedly HER2 positive disease have a “negative” TRAS‐PET and a very low probability of benefitting from the drug. Similarly no metabolic response on the early FDG PET correlates with a short time to treatment failure (Gebhart et al., 2016).
Another PET tracer, [18F] fluoroestradiol‐17β (18F‐FES), a radio‐labelled oestradiol analogue, has been developed to evaluate the expression and the binding capacity of ER in breast cancer tissue, and thus is potentially useful in a number of clinical settings. Mortimer and colleagues compared tumour biopsy and 18F‐FES CT/PET findings in 41 breast cancer patients. 18F‐FES identified 16 of 21 biopsy confirmed ER+ patients in this sample. For ER negative patient, concordance between imaging and biopsy was 100% (Mortimer et al., 1996). A further study suggests that it may reduce the number of biopsies during evaluation of possible BC recurrence (Koleva‐Kolarova et al., 2015). 18F‐FES PET was also used in a study to evaluate ER availability in MBC patients during fulvestrant use. 18F‐FES‐PET showed significant residual ER availability in tumours during fulvestrant therapy in 38% of patients, a finding associated with early tumour progression (van Kruchten et al., 2015).
The last 20 years of MBC treatment have been characterized by an ever increasing arsenal of drugs and biomedical tests that have led to prolongation of survival. While some new agents are well established as new standards of care, their optimal sequencing is debatable and will require further investigation. We still live in the era of stratified oncology, and classical criteria still are the main decision‐making tools, though, under the surface more and more research is focused on the development of biomarkers that will likely change this in the coming years, even if, for now, frustratingly, they have not despite considerable efforts. We have, therefore, a long way to complete the transformation to personalized oncology. For this to come true, we need new academic models of collaboration, involving pharmaceutical and insurance companies, and support of our patients, as well as improved prospectively evaluated biomarkers and tools that solve the conundrum of space and time tumour heterogeneity.
The authors are not aware of any affiliations, memberships, funding, or financial holdings that might be perceived as affecting the objectivity of this review.
Amir Sonnenblick was an ESMO translational research fellow.
Sonnenblick A., Pondé N., Piccart M., (2016), Metastatic breast cancer: The Odyssey of personalization, Molecular Oncology 10, doi: 10.1016/j.molonc.2016.07.002.