Patterns in target-directed breast cancer research

We undertake an analysis of ongoing BC targeted therapy trials registered to CT.gov to describe patterns of ongoing clinical research, highlight gaps in current research programs and identify ways of optimizing ongoing initiatives. A search of clinicaltrials.gov was conducted on September 4, 2013 to identify ongoing randomized phase II and III trials of targeted therapies in BC. A total of 280 trials were analyzed, the majority conducted in either human epidermal growth factor receptor 2 (HER2)-positive (n = 79, 28.2 %) or hormone receptor (HR)-positive (n = 104, 37.1 %) populations. Less than half of all trials were conducted in populations selected to match the agent under investigation (n = 126, 45 %). HER2-directed therapy is the single most investigated class of targeted agents (n = 73, 26.1 %), but trials investigating anti-angiogenic agents are also common (n = 49, 17.5 %). The most common new classes of agents under investigation in HR-positive and triple negative (TN)/BRCA-positive disease, are non-receptor protein kinase-inhibitors (n = 12; 11.5 %) and poly (ADP-ribose) polymerase inhibitors (n = 6; 30 %), respectively. The majority of regimens combine new targeted agents with either chemotherapy (n = 164, 58.6 %) or endocrine therapy (n = 113, 40.4 %); a total of 8 trials (2.8 %) investigated peptide-drug conjugates. The most frequently utilized end-points were pathological complete response in the neo-adjuvant setting (n = 36, 52.9 %) and time-to-event end-points in the adjuvant and advanced settings (77.3 and 72.6 %, respectively). Our findings suggest a need for more target-matched agent development, maintenance of a value-based focus in research and a need for the clinical development of agents to treat TN/BRCA-positive and HR-positive BC.


Background
Breast cancer (BC) is a significant health concern, with approximately 256,140 new diagnoses of BC in North America annually and 44,720 deaths in 2013 (DeSantis et al. 2013; Canadian Cancer Society's Steering Committee on Cancer Statistics 2013). Over 600 million dollars are invested in BC research in the United States (US) each year by, the National Cancer Institute alone (National Cancer Institute 2013), and female BC has received the highest allotment of US national expenditure for cancer treatment (National Cancer Institute 2012). For over a decade, a main objective of BC research has been the development of targeted agents designed to improve outcomes while decreasing toxicity (Jain 2014). Efforts to move from a "one size fits all" to a more personalized approach to therapy have resulted in a substantial, multi-faceted body of research. Examples of some of the more significant research gleanings related to trial populations, interventions and trial design are summarized in Table 1. Prominent among these is the discovery of target-matched treatment strategies, the development of targeted treatments in populations enriched for the biological target of interest [e.g., hormone-receptor (HR) or human epidermal growth factor receptor 2 (HER2)]. Recent data show that wide-spread use of target-matched strategies over the last 15 years have resulted in dramatic improvements in the prognosis of patients with estrogenreceptor (ER)-positive ( Table 1 Lessons learned over the past decade of target-directed research in breast cancer CT chemotherapy, ET endocrine therapy, FEC fluorouracil, epirubicin and cyclophosphamide, HER2 human epidermal growth factor receptor 2, HR hormone receptor, NAT neoadjuvant therapy, OS overall survival, pCR pathological complete response, T-D target-directed therapy, T-DM1 trastuzumab emtansine, TN triple negative a Patient selection is based on over-expression, mutation or other modification of one or more biomarkers or on a multi-biomarker profile/signature with prognostic or predictive value b Biomarker used to positively-select patients is targeted by the investigational T-D agent c Depends on use of pCR as surrogate for survival (pCR translates to disease-free survival and overall survival according to results of the NOAH trial)  d Overall survival (or surrogate) as primary end-point

Negative trial outcomes
Bevacizumab combinations in first-line (Miles et al. 2010;Miller et al. 2007;Robert et al. 2011) Positive trial outcomes EGF104535 (Guan et al. 2013), CLEOPATRA (Swain et al. 2012;Verma et al. 2012), EMILIA (Baselga et al. 2012c; Collaborative Group 1998; Davies et al. 2011) and HER2positive disease (Dawood et al. 2010;Yin et al. 2011;Harris et al. 2011), in both the early and advanced settings. Additionally, the discovery of 6 intrinsic biological BC subtypes (luminal A; luminal B; HER2-enriched; basallike; normal breast-like; and claudin-low) (Perou et al. 2000;Sorlie et al. 2001;Carey et al. 2006;Prat et al. 2010) has reshaped our understanding of disease biology and shifted our current approach to treatment. Treatment decisions are now guided by prognostic and predictive biomarkers [ER, progesterone receptor (PR) and HER2] which define 3 major therapeutic groups: HER2-positive disease (~20 % of all patients) (Arteaga et al. 2012;Ross et al. 2009), HR-positive disease (~75 %) (Anderson et al. 2011;Lim et al. 2012;Nadji et al. 2005), and triplenegative disease (TN, neither HER2, ER or PR-positive; ~15 %) (Foulkes et al. 2010). The National Institute of Health's clinicaltrials.gov (CT. gov) database is the most robust of international trial registries, serving as both a mandatory repository for information on clinical trials conducted under US regulation and a prerequisite for publishing study results in peer-reviewed journals (Hirsch et al. 2013). Although select data are populated by individual investigators and not always consistently reported, the database represents a unique resource through which to evaluate research. The database currently contains detailed information on more than 5000 clinical trials in BC from more than 90 countries (ClinicalTrials.gov 2014b), and ranks BC among the most investigated tumor types per incidence (Hirsch et al. 2013). However, given that clinical research in oncology is both costly and associated with the highest rates of drug attrition and trial failure (Begley and Ellis 2012;Hutchinson and Kirk 2011) we have undertaken an analysis of ongoing BC targeted therapy trials registered to CT.gov to describe patterns of ongoing clinical research, highlight gaps in current research programs and identify ways of optimizing ongoing initiatives.

Study selection
A total of 1545 matching records were downloaded for analysis, and 1265 studies were excluded (Fig. 1). The remaining data set of 280 trials was locked and parsed to facilitate analysis.

Populations and classes under development
The majority of trials were conducted in either HER2positive (n = 79, 28.2 %) or HR-positive (n = 104, 37.1 %) populations (Fig. 2a). Trials conducted in TN/ BRCA-positive disease accounted for 7.1 % (n = 20) of all research. Trials in all other populations accounted for 27.5 % (n = 77) of research. Less than half of all ongoing trials were conducted in target-matched (i.e., enriched for the biological target of the therapy under investigation) populations (n = 126, 45.0 %; Table 2). The most investigated classes of agents were HER2-inhibitors (n = 73, 26.1 %), endocrine agents (n = 52, 18.6 %) and anti-angiogenic agents (n = 49, 17.5 %; Fig. 2b). The proportion of research dedicated to the development of emergent agents was only slightly greater than the proportion addressing established agents (52.5 and 47.5 %, respectively), and consisted mostly of phase II studies (74.8 %).
There was a slightly greater proportion of phase II trials compared with phase III trials overall (n = 164, 58.6 %; Table 2). Phase II trials were most common in the neoadjuvant (n = 53, 77.9 %) and advanced (n = 96, 65.8 %) settings, while phase III trials were more common in the adjuvant setting (n = 51, 77.3 %).

Populations
Given the prevalence of expression and demonstrated ability to target HER2 and HRs, the proportion of research dedicated to populations defined by these biomarkers is appropriate. However, relative to the overall incidence of HR-positive and TN disease (~75 % (Lim et al. 2012;Anderson et al. 2011;Nadji et al. 2005) and ~15 % (Foulkes et al. 2010)), the amount clinical development in these settings is low and highlights a need for further research in these settings.

Elimination of chemotherapy
One of the great promises of targeted therapy was the potential to reduce or eliminate the need for chemotherapy and its indiscriminate effect on normal tissue. However, after more than a decade of research, the majority of trials conducted in non-HR-positive populations (n = 151, 85.8 %) combine targeted agents with chemotherapy. It is only recently that trials have begun to explore the removal of chemotherapy from targeted regimens for select populations; e.g., eliminating chemotherapy from HER2-directed regimens in elderly adjuvant patients (n = 1) or from dual HER2-targeted combinations in the advanced and neo-adjuvant settings (n = 2).
Recurrence scores based on clinicopathological features [e.g., Nottingham Prognostic Index (Blamey et al. 2007), Adjuvant! (Ravdin et al. 2001), and PREDICT (Wishart et al. 2011)] have also been useful in identifying patients who may forego adjuvant chemotherapy. Gene signature-based scores are now being validated in randomized phase III trials in intermediate-risk, HRpositive patients (TAILORx (ClinicalTrials.gov 2014a), RxPONDER (ClinicalTrials.gov 2014d), and MINDACT (ClinicalTrials.gov 2013)), and are expected to further define subsets of patients who may be spared the toxicity of chemotherapy Bogaerts et al. 2006). Antibody-cytotoxic conjugates are yet another important means by which chemotherapy-associated adverse effects can be reduced. In HER2-positive disease, a HER2-directed cytotoxic is replacing existing single-agent targeted therapy or cytotoxic-targeted combinations, and a GpNMB-directed cytotoxic is being developed for GpNMB-expressing TN disease (METRIC trial, NCT01997333) (Celldex Therapeutics 2012; Clini-calTrials.gov 2012, 2014c).

Combinatorial strategies
Effectively targeting oncogenic mutations or copy number alterations has proven challenging, with no new agents identified in the last 15 years. In this context, combinatorial approaches have become one of the most commonly explored strategies. In HER2-positive disease, mono-class regimens combining multiple targeted agents to more effectively block a given receptor have become the focus of ongoing combinatorial research. Dual-HER2-inhibition has held much promise in both the advanced (Baselga et al. 2012c;Verma et al. 2012) and neo-adjuvant  settings, yet findings from the ALTTO trial, showing a lack of improvement with the addition of lapatinib to standard adjuvant targeted therapy, calls into question the benefits of this approach in earlier settings . Results from the APHINITY trial (NCT01358877), assessing the addition of pertuzumab (rather than lapatinib) to adjuvant targeted therapy, will help clarify the role for combinatorial strategies in early disease.
In HR-positive disease, a main direction of research has been the development of multi-class regimens to inhibit secondary processes, such as treatment resistance (mTOR/PI3K-inhibitors; insulin growth factor receptor [IGF(R)]-inhibitors; epidermal growth factor receptor [EGFR]-inhibitors; fibroblast growth factor receptor [FGFR]-inhibitors), cell cycle regulation (CDK4/6-inhibitors) or effects of the tumor micro-environment (bisphosphonates). Breakthroughs such as the addition of the mTOR-inhibitor everolimus to exemestane in advanced BC resistant to prior non-steroidal aromatase-inhibitor therapy (Baselga et al. 2012b) and the addition of palbociclib to fulvestrant in patients with advanced BC progressing on prior ET therapy (Turner et al. 2015) illustrate the promise of combinatorial approaches in enhancing established targeted strategies. However, questions of tolerability and cost remain as combinatorial strategies are undertaken to more completely inhibit pro-oncogenic pathways.

Neo-adjuvant setting: platform for accelerated drug development
The neo-adjuvant setting provides a unique platform for targeted agent research, with opportunities for correlative studies and the potential for translating discovery into benefit in the adjuvant setting. Relative to drug development, improvements in pCR have been correlated with survival outcomes in HER2-positive and TN subtypes (Cortazar et al. 2014) and can be used as the basis for accelerated FDA approval (Prowell and Pazdur 2012). Although the FDA approved pertuzumab in the neoadjuvant setting, the results of the Neosphere trial did not show a statistically significant association between pCR and 3-year disease-free survival and progression-free survival (PFS) (Gianni et al. 2015). The increase in both number and proportion of clinical trials conducted in this setting over the last several years suggests an increased commitment to neo-adjuvant research, although data also suggests that it remains an underutilized strategy.
The negative results of the ALTTO study , evaluating an adjuvant dual-HER2blockade, bring into question the assumption that benefits in the neo-adjuvant setting (Piccart-Gebhart et al. 2013) automatically translate into adjuvant benefits. These findings underscore the complexity and challenges of accelerated drug development. Innovative approaches to neo-adjuvant research, using adaptive Bayesian designs and pCR as the primary end-point, to rapidly select active novel agents (e.g., ISPY2 trial (ClinicalTrials.gov 2015)), may lead to more efficient use of research resources by requiring fewer patients, although absolute magnitudes of benefit are difficult to assess using this type of approach and results require phase III confirmation.

Optimization of research resources
Although BC is the most investigated disease site (Hirsch et al. 2013), it is also an area of research associated with one of the highest rates of drug attrition and trial failure (Begley and Ellis 2012;Hutchinson and Kirk 2011). Presently, antiangiogenic agent trials (n = 49, 17.5 %) comprise almost a fifth of all ongoing research, and the total number of these trials is comparable the sum of all ET directed research (n = 52, 18.6 %). Clinical testing of anti-angiogenics in BC has been marked by failure to demonstrate clinically significant PFS and survival benefits and an increased risk of serious side effects (Miles et al. 2010;Robert et al. 2011;Hamburg 2011;Barrios et al. 2010;Baselga et al. 2012d;Mackey et al. 2013). Despite this, as of September 2013, a total of 25,784 patients were accrued to current anti-angiogenic trials, with planned accrual of an additional 3833 patients across 12 trials. A 2006 survey of leading developers estimates that the cost of enrolling a patient into a phase III trial is $26,000 (lifesciences world 2006; Stewart et al. 2010). Given these figures, the investment directed toward anti-angiogenic research has amounted to a staggering $770,042,000. As the hope of success continues to entice patients and clinicians alike to fully explore the benefits of a given class of therapy, prudence would call for a redirection of resources towards classes of agents that have demonstrated therapeutic benefit or for which a biomarker is available to guide therapy. This is best exemplified in the recent discovery of the 14-gene signature to identify immune-enriched patients who preferentially respond to trastuzumab therapy ).

Conclusions
Target-directed research is essential to ongoing research efforts in BC and our understanding of how to optimize these strategies continues to evolve. Our findings suggest that there is a continued need for target-matched agent development, maintenance of a value-based focus in research and a need for the clinical development of agents to treat TN/BRCA-positive and HR-positive BC.

Target-directed trial dataset
A search of the CT.gov website was conducted on September 4, 2013 to identify randomized phase II and III trials of targeted therapies in BC. We considered targeted therapies to be anti-cancer drugs with a clear cellular or molecularly-directed mechanisms of action that interfere with cell growth signaling or tumor blood vessel development, promote death of specific cell types, or stimulate the immune system to destroy specific cell types and/or deliver toxic drugs to cancer cells (National Cancer Institute 2014). All non-randomized, non-systemic, non-therapeutic, or withdrawn trials, as well as those conducted in a non-invasive setting, without a target-directed agent in the experimental arm, or with a primary completion date (date of primary outcome data collection, or date expected) before January 2012, were excluded.

Trial review and classification
Each trial was classified and analyzed based on the following 9 criteria, which were established based on the record title: (1) degree to which the investigational targeted agent is established (defined below), (2) number and (3) class of targeted agents in the investigational arm, (4) use of continued targeted therapy, (5) setting, (6) biological subtype of population, (7) status of trial, (8) study type and (9) end-points used. If the category was unclear, conditions and key words were assessed or the full CT.gov record was reviewed.
Established targeted-drugs, defined as those with at least one US Food and Drug Administration (FDA)approved BC indication as of September 4, 2013, are summarized in Table 3. Bone-modifying/remodeling agents and progesterone were considered established due to their historical and widespread use in BC treatment; all other agents were defined as emergent. Trials were categorized into 4 mutually exclusive groups based on the biomarker status of the trial population, in order of therapeutic relevance, as follows: HER2positive; HR-positive; TN or BRCA-mutated; (TN/ BRCA-positive); and other populations (HER2-negative trials with HR status unspecified; other subtypes and unselected; not defined by biomarker status).
To assess the degree to which neo-adjuvant trials have changed over the last 5 years, trials with primary completion dates between 2012 and 2016 were compared to those with primary completion dates between 2007 and 2011.

Authors' contributions
ST, MT, CS, JFB, and DM participated in the study conception and design. MT, DM and IM carried out collection and assembly of data. All authors participated in data analysis and interpretation and writing of the manuscript. All authors read and approved the final manuscript.