Abstract
Common resistance mechanisms to endocrine therapy (ET) in estrogen receptor (ER)–positive metastatic breast cancers include, among others, ER loss and acquired activating mutations in the ligand-binding domain of the ER gene (ESR1LBDm). ESR1 mutational mediated resistance may be overcome by selective ER degraders (SERD). During the first-in-human study of oral SERD AZD9496, early changes in circulating tumor cells (CTCs) and circulating tumor DNA (ctDNA) were explored as potential noninvasive tools, alongside paired tumor biopsies, to assess pharmacodynamics and early efficacy.
CTC were enumerated/phenotyped for ER and Ki67 using CellSearch in serial blood draws. ctDNA was assessed for the most common ESR1LBDm by droplet digital PCR (BioRad).
Before starting AZD9496, 11 of 43 (25%) patients had ≥5 CTC/7.5 mL whole blood (WB), none of whom underwent reduction to <5 CTC/7.5 mL WB on C1D15. Five of 11 patients had baseline CTC-ER+, two of whom had CTC-ER+ reduction. CTC-Ki67 status did not change appreciably. Patients with ≥5 CTC/7.5 mL WB before treatment had worse progression-free survival (PFS) than patients with <5 CTC (P = 0.0003). Fourteen of 45 (31%) patients had ESR1LBDm+ ctDNA at baseline, five of whom had ≥2 unique mutations. Baseline ESR1LBDm status was not prognostic. Patients with persistently elevated CTC and/or ESR1LBDm+ ctDNA at C1D15 had worse PFS than patients who did not (P = 0.0007).
Elevated CTC at baseline was a strong prognostic factor in this cohort. Early on-treatment changes were observed in CTC-ER+ and ESR1LBDm+ ctDNA, but not in overall CTC number. Integrating multiple biomarkers in prospective trials may improve outcome prediction and ET resistance mechanisms' identification over a single biomarker.
Determination of the molecular status of a cancer by evaluating circulating biomarkers has been designated as a “liquid biopsy.” Liquid biopsies offer the possibility of serial noninvasive monitoring of the overall tumor burden as opposed to traditional research tissue biopsies, which assess a single site of disease. In patients with metastatic breast cancer (MBC), elevated circulating tumor cells (CTCs) or ESR1 mutations assessed in circulating tumor DNA (ctDNA) have been previously associated with worse prognosis. In this work, we used these circulating markers to assess pharmacodynamic changes in patients with estrogen receptor (ER)–positive/HER2-negative MBC participating in a first-in-human study of the oral selective ER degrader (SERD), AZD9496. By integrating CTC and ctDNA, we were also able to explore the potential complementary utility of these circulating markers to detect more general mechanisms of resistance to ET noninvasively within a phase I trial of an oral SERD.
Introduction
Endocrine (or antiestrogen) therapies (ETs) are the favored initial choice of treatment for most patients with estrogen receptor (ER)–positive metastatic breast cancer (MBC). Commonly used ETs include the selective ER modulator (SERM) tamoxifen, third-generation aromatase inhibitors (AIs), and the selective ER degrader (SERD) fulvestrant (1).
Resistance to ET can occur de novo or be acquired during the course of therapy (2). Absence of ER expression is the single most potent mechanism of ET resistance (3). Other mechanisms of ET resistance include deregulation of the ER pathway, for example, by phosphorylation of factors downstream of ER, and/or activation of alternative prosurvival or proliferative pathways (4).
Several investigators have reported activating mutations in the ligand-binding domain (LBD) of ESR1, the gene encoding ER. These mutations induce ligand-independent ER activity, leading to apparent resistance to AIs and other estrogen-depletion strategies (5–8). The potential impact of ESR1LBD mutations (ESR1LBDm) on clinical outcomes together with implications of subclonality of these mutations has not been fully elucidated. However, preclinical and independent exploratory clinical studies have suggested that patients with ESR1LBDm have shorter progression-free survival (PFS) on subsequent AI therapy versus patients with ESR1LBD wild-type (WT) cancers and yet may retain relative (possibly dose-dependent) sensitivity to fulvestrant (5, 9–12). Increased dosing of fulvestrant might further improve its efficacy, but the bioavailability and pharmacokinetic (PK) limitations of this drug restrict the maximum feasible dose to 500 mg once-monthly intramuscular injection, and steady-state plasma concentrations are not reached until 3 to 6 months after first administration.
The development of an oral SERD with satisfactory bioavailability that could be given at higher relative doses than fulvestrant is highly desirable. AZD9496 (13) is one of several oral SERDs that are being tested in clinical trials such as RAD1901 (NCT02338349; ref. 14), GDC-9545 (NCT03332797), LSZ102 (NCT02734615), and SAR439859 (NCT03284957).
In this first-in-human study, the PK of AZD9496 was characterized by a rapid absorption (median tmax: 1.55–3.00 hours) and fast biphasic decline after reaching the maximum concentration with a mean terminal half-life (t½) of 1.4 to 5.7 hours (13). Following multiple doses of AZD9496, the exposures of AZD9496 were consistently and dose-dependently lower than for a single AZD9496 dose, presumably due to autoinduction of cytochrome P450 (CYP) isoenzymes (13).
Ideally, pharmacodynamic (PD) markers should be incorporated into early therapeutic development trials to determine if the drug is reaching and affecting its drug target. However, serial metastatic tissue biopsies are invasive (15) and expensive. Conversely, “liquid biopsies,” such as assays for circulating tumor cells (CTC) or cell-free tumor DNA (ctDNA), are relatively noninvasive and allow longitudinal assessment of circulating biomarkers (16). These markers may allow monitoring of PD changes, as well as identification of genetic and nongenetic determinants of drug response and development of drug resistance. We and others have reported the ability to accurately and reproducibly measure ER and Ki67 levels in CTC as well as ESR1LBDm from ctDNA (ESR1LBDm ctDNA; refs. 11, 17–19).
Here, we report the results of correlative studies of CTC-phenotype, ESR1LBDm status in ctDNA, and PD analyses of paired biopsies in blood and tissue samples collected from patients with ER-positive/HER2-negative MBC who participated in the first-in-human phase I dose-escalation trial of the oral SERD AZD9496 (13).
Materials and Methods
Study design and objectives
This correlative study was performed using specimens prospectively collected from patients who participated in a phase I, open-label, multicenter trial (NCT02248090 in clinicaltrials.gov), designed to investigate the safety and tolerability and PKs of increasing doses of the oral SERD AZD9496 (13). The main study was carried out in accordance with the principles of the International Conference on Harmonization guidelines for Good Clinical Practice, the Declaration of Helsinki, and all applicable laws. All subjects provided written-informed consent approved by their local Institutional Review Board for collection of serial tissue biopsies and serial blood draws as part of their participation in the clinical trial.
The primary objectives of this correlative trial were to investigate (1) the prognostic role of CTC enumeration and characterization prior to treatment and whether CTC enumeration, CTC-ER status, and CTC-Ki67 status could be used as a PD biomarker of AZD9496 activity; (2) the prognostic role of baseline ESR1LBD mutational status in ctDNA and early changes in mutational levels during treatment as potential PD biomarker and/or for early prediction of response; and (3) changes in ERα, progesterone receptor (PR), and Ki67 expression in optional paired tumor biopsies to assess the PD activity of AZD9496. Moreover, baseline variability of CTC and ctDNA was investigated.
Patient staging and follow-up
Details of eligibility, accrual, and conduct of the clinical trial have been reported in a study by Hamilton and colleagues (13). Briefly, eligibility was limited to patients with ER-positive/HER2-negative metastatic or locoregionally recurrent disease, not amenable to treatment with curative intent, who had progressed after ≥6 months of ET for ER-positive breast cancer (before protocol amendment August 21, 2015, must have spent ≥6 months on a line of ET in the advanced setting). Pre- or perimenopausal women must have started luteinizing hormone-releasing hormone agonist treatment at least 4 weeks before study treatment and must have continued this treatment throughout the study. Sex hormone–containing drugs such as hormone-replacement therapy (HRT), dehydroepiandrosterone, other androgens (e.g., oxandrolone), SERMs (e.g., raloxifene), and megestrol acetate were not permitted during the study. Of note, a washout from any cytotoxic chemotherapy, investigational agents, or other anticancer drugs for the treatment of advanced breast cancer from a previous treatment regimen or clinical study within 14 days of the first dose of study treatment was required. Although a fulvestrant-specific “wash-out” period was not defined, for the purpose of CTC analysis in this study, we defined fulvestrant “wash-out” as 120 days (∼3.3 half-lives).
Assessment of tumor response by RECIST 1.1 was performed every 8 weeks after the start of treatment for 24 weeks, and thereafter every 12 weeks (±1 week) until objective disease progression, as defined by RECIST 1.1, even if a patient discontinued treatment prior to progression. A “rolling 6” design was employed, in which each cohort of at least three and up to six patients received AZD9496 at escalating dose. Dosing began at 20 mg once daily up to 600 mg twice daily, which was regarded as the maximum feasible dose on the basis of the number of tablets required for each dose.
Collection of CTC, plasma for ctDNA analysis, and paired biopsies
Overall timeline of sample collection is shown in Supplementary Table S1. Blood samples were drawn twice prior to starting therapy: a screen time point (−28 to −1 days prior to treatment initiation) and a second sample on the day of treatment initiation (C1D1). The duration of a cycle was 4 weeks for the first six cycles and 6 weeks thereafter. In one patient, cycle 1, day 2 (C1D2) sample was used as a surrogate sample for ESR1LBDm assessment due to missed screening/C1D1 samples. Collectively, these specimens are designated as “baseline,” but results from individual time points are reported separately as stated throughout this report.
CTC.
For CTC evaluation, whole blood (WB) was collected into CellSave preservative tubes (Menarini Silicon Biosystems, Inc.) during a screening window (−28 to −1 days prior to C1D1), and subsequently on C1D1, cycle 1, day 15 (C1D15), and at treatment discontinuation. Three CellSave tubes were pooled and divided into three aliquots for CTC assessment [one to test CTC-ER status, one to test CTC-Ki67 status, and one to perform RNA extraction from CTC (data not reported), respectively]. Blood samples were maintained and shipped at room temperature to one of two central laboratories (University of Michigan or Cancer Research UK Manchester Institute) where they were processed within a maximum of 96 hours after blood draw.
ctDNA.
For ctDNA analysis, blood was drawn into 10-mL Streck Cell-Free DNA BCT (STRECK, INC) tubes at screening, C1D1, C1D2, C1D15, C3D1, day 1 of every alternate cycle after C3D1, and at treatment discontinuation. Screening or C1D1 samples were used for baseline determination of ESR1LBDm status (mutation detected at any of the two time points), but the C1D2 sample was used as a surrogate in case of missing screening/C1D1 samples. For determination of changes of ESR1LBDm levels at C1D15 versus baseline, C1D1 (when available) was used as baseline levels, otherwise screening was used. One tube of WB per time point was collected for ctDNA and shipped to a central laboratory (Covance) at ambient temperature range (6°C–37°C) for further processing.
Tissue.
Consent for paired research tumor biopsies was optional for patients enrolled in this study. The pretreatment samples were collected during screening or at disease progression on the therapy administered prior to study. Sites were requested to obtain the on-treatment sample on C2D1 ± 7 days, but the sample could be taken outside this time window if agreed with the sponsor (before protocol amendment August 21, 2015, the on-treatment sample was scheduled at C1D15; however, anytime between C1D7 and C1D28 was acceptable). Sites were instructed to collect the on-treatment biopsy between 2 and 12 hours after the latest dose of AZD9496. Patients could also consent to a tumor biopsy at disease progression, if clinically feasible.
Sample processing and assay methods
Blinding.
Laboratory personnel were not blinded to trial subject number, dose group, and time point of CTC and ctDNA samples. For the paired tumor tissue samples, the laboratory personnel and the pathologist were blinded to the time point, but not to trial subject number and dose group. Laboratory personnel for all samples types were blinded to the clinical outcomes.
CTC. Enrichment, enumeration, and characterization.
Blood specimens were analyzed for CTC enumeration and semiquantitative analysis of expression of ER and Ki67 using the CXC CellSearch Kit and CellSearch system (Menarini Silicon Biosystems, Inc.), as previously reported (17, 19). CTC were considered elevated if the aliquot contained ≥5 CTC/7.5 mL of WB according to previous literature (20). If an aliquot had elevated CTC, CTC staining was expressed on an arbitrary scale of 0 to 3+, as described previously (19). For each marker, 0 and 1+ was considered negative, and 2–3+ was considered positive (19). CTC enumeration, CTC-ER status, and CTC-Ki67 status were determined by two independent reviewers (E.M. Dolce, E.P. Darga, K. Aung, C. Paoletti, F. Butt, and N. Iqbal). Discordant results were reconciled by joint readings. Quality control between the two sites was performed every 3 to 4 weeks to ensure reading methods were concordant. CTC response was defined as a reduction to <5 CTC/7.5 mL WB at C1D15, as previously described (21, 22).
CtDNA and tissue: molecular analyses
ddPCR analyses: tumor tissue DNA and circulating free DNA extraction.
DNA was extracted from formalin-fixed paraffin-embedded (FFPE) sections using the QIAamp DNA FFPE Tissue Kit (QIAGEN) according to the manufacturer's protocol, and buffer ATE was applied to each column and incubated for 5 minutes at room temperature prior to elution into Eppendorf LoBind Microcentrifuge tubes. Blood samples were processed on day of arrival in two steps. In the first step, WB was centrifuged at approximately 2,000 G for 10 minutes using a prechilled centrifuge set to 4°C, and plasma was separated by pipette and transferred to a 15-mL Falcon tube. In the second step, plasma was centrifuged again at approximately 12,000 G or higher for 10 minutes using a prechilled centrifuge set to 4°C, separated by pipette, aliquoted, and frozen at −80°C until extraction. ctDNA was extracted from 2-mL plasma using the QIAamp Circulating Nucleic Acid Kit with the QIAvac 24 Plus vacuum manifold (QIAGEN) as previously described (23). Samples were stored at −20°C prior to droplet digital PCR (ddPCR) analysis.
ESR1 mutation analyses by ddPCR.
ddPCR was performed using the QX200 AutoDG Droplet Digital PCR System (BioRad) according to the manufacturer's protocol. Custom assays were designed by IDT, incorporating locked nucleic acid (LNA) bases into each probe to increase discrimination. Probes were from IDT, and primers (SePOP desalted) from Eurogentec (Supplementary Table S2). The D538G, E380Q, Y537C, Y537N, and Y537S ESR1 assays were run as singleplex assays. Multiplexing three assays (V534E, L536Q, and L536R) was achieved by modifying both probe concentrations and the concentration of FAM and HEX fluorescent labels. In any cases where the ESR1 multiplex result was not conclusive, ddPCR was repeated with relevant singleplex assays. If the baseline sample was negative for E380Q and the multiplex assay, no samples of subsequent time points were tested due to limited sample material. For the remainder of the assays (D538G, Y537C, Y537N, and Y537S), every time point was tested regardless of status at baseline. Each 20-μL ddPCR reaction contained 5 μL of circulating free DNA or FFPE-extracted DNA. Positive and negative controls were run in triplicate or quadruplicate on each plate. Note that 150 bp gBlocks (IDT) or sheared plasmids (AZ Bioscience) containing the mutation of interest were used as positive controls. Human genomic DNA (male) was used as WT control (Promega). Appropriate elution buffer and water served as negative controls. Thresholds were manually set for each sample using acceptance criteria defined during the optimization of each assay. QuantaSoft software (version 1.7.4; BioRad) was used to assign positive/negative droplets and convert counts into mutant copies/mL (23). Mutation levels were expressed as mutant DNA copies/mL plasma. “ESR1LBDm+” ctDNA indicates the presence of at least one ESR1LBDm (i.e., definite positive by the predefined acceptance criteria) in the ctDNA (at baseline or at other time point, as specified throughout the report). “Borderline” cases (i.e., <3 positive droplets, which is the predefined threshold of positivity) were considered “not detected” (or “negative”) unless otherwise specified. A “ctDNA response” was defined as ≥50% decline in mutational levels of the dominant mutation at C1D15 versus baseline.
Tissue. Immunohistochemistry analyses.
Tumor biopsies were processed into FFPE blocks prior to immunohistochemistry (IHC) analysis at the participants' institutions. IHC analysis was performed at AstraZeneca for semiquantitative ER, PR, and Ki67 (24). See Supplementary Table S3 for antibody details. IHC staining was performed using a LabVision Autostainer (Thermo Fisher Scientific, Inc.), following optimal pretreatment using a HistoS-3 microwave pressure retrieval system. Antibody-specific staining for ER and PR was scored manually by a board-certified pathologist at AstraZeneca (J. Geradts) and/or at a contracted external good clinical practice–accredited laboratory (Source BioScience). For both hormone receptors, the percentage of tumor cells was assessed with an intensity of negative, weak, moderate, and strong staining (0, negative; 1, weak; 2, moderate; 3, strong). Results for ER and PgR were recorded as an H score [sum of (1 x percentage weak) + (2 x percentage moderate) + (3 x percentage strong)] to a maximum of 300. Ki67 index was expressed as average % positive cells (count 300 cells).
PK.
The exploratory PK/PD analysis involved visual evaluation of changes in CTC count and CTC-ER status, CTC-Ki67 status, and ESR1LBDm ctDNA levels over treatment time in relation to AZD9496 dose and PK parameters (AUC, Cmax), assessed by standard noncompartmental methods. Absolute changes in biomarker at C1D15 (compared with baseline) were used to perform the assessment. The CA15-3 time-series biomarker data (assessed by local laboratories) were assessed through visual evaluation and model-based assessment using mixed effects modeling in R software v3.3.1. In the initial modeling step, linear and nonlinear models were investigated within the R software package “nlme” to describe the time course of the biomarker data. Subsequently, model-based evaluations were performed to test the statistical significance of AZD9496 dose or PK endpoints as explanatory covariates for the rate of change of CA15-3 over the time course of the study. Model selection was based on the likelihood ratio test (P = 0.01). The PK/PD relationship analysis with tissue markers was not performed.
Statistical analysis
All statistical analyses were performed with GraphPad Prism version 7.04 (GraphPad Software) or Microsoft Excel. Association of CTC levels and ctDNA with PFS was assessed using the log-rank test. PFS was defined as the time from start of treatment until objective disease progression as defined by RECIST 1.1 or death (by any cause in the absence of progression) regardless of whether the patient withdrew from study treatment or received another anticancer therapy prior to progression. Patients who had not progressed or died at the end of the study were censored at the time of their last evaluable RECIST assessment. Unless stated otherwise, P values were two tailed and considered significant if P < 0.05. Treatment was ongoing in six patients (13.3%) at the data cut-off of January 31, 2017. The study is reported according to the REMARK guidelines (25).
Results
Patients
All 45 patients enrolled in the phase I trial were also enrolled in this correlative study. Each subject had at least one blood sample collected at any time point for CTC and ctDNA assessment (Fig. 1). In particular, 43 patients had CTC collected at screening and/or C1D1, whereas two patients only had CTC collected at time of study discontinuation. In addition, 44 patients had ctDNA collected at screening and/or C1D1. For the patient, who did not have ctDNA collected at screening and C1D1, the sample from C1D2 was considered as baseline. Patient and disease characteristics were previously described (13). Data cutoff was on January 31, 2017.
CTC enumeration and ESR1LBDm+ ctDNA at baseline
Of the 43 patients for whom CTC were collected at screening and/or C1D1, 23 (53.5%) and 11 (25%) had ≥1 and ≥5 CTC/7.5 mL WB at one or the other, or both, baseline blood draws (screening or C1D1), respectively (Table 1; Fig. 2A). Fourteen of 45 (31%) patients had ESR1LBDm+ ctDNA at baseline (screening or C1D1 sample for 44 patients and C1D2 for one patient). Five of these 14 patients (36%) had more than one ESR1LBDm identified in the ctDNA in the same blood draw (Fig. 2B), all of which featured a D538G clone (5/5). For this report, the “dominant mutation” was considered the mutation with the highest mutational levels at baseline. When considering all ESR1LBDm detected at baseline, the most common mutations were D538G (6/14), followed by Y537S (5/14) and Y537N (5/14). Overall, of the 43 patients who had both CTC and ctDNA assessed, 23 (53.5%) patients had either elevated CTC numbers or ESR1LBDm+ ctDNA. In particular, 11 patients (26%) had elevated CTC (≥5 CTC/7.5 mL WB), and 12 (28%) had ESR1LBDm+ ctDNA at baseline (Table 1). Only four (9%) patients had both elevated CTC and ESR1LBDm+ ctDNA. Eight of the 32 (25%) patients with <5 CTC/7.5 mL WB had elevated ESR1LBDm+ ctDNA, whereas seven of 31 (23%) patients with “ESR1LBDm not detected” (ESR1LBDm− ctDNA) had elevated CTC levels (Table 1).
CTC enumeration (<5 vs. ≥5 CTC/7.5 mL WB) and ESR1LBDm status [ESR1LBDm− vs. ESR1LBDm+ (≥1)] in 43 patients who had both CTC and ctDNA assessed at baseline
. | CTC at baseline . | ||
---|---|---|---|
ESR1LBDm status at baseline . | <5 CTC/7.5 mL WB . | ≥5 CTC/7.5 mL WB . | Total . |
ESR1LBDm+ ctDNA | 8 | 4 | 12 |
ESR1LBDm- ctDNA | 24 | 7 | 31 |
Total | 32 | 11 | 43a |
. | CTC at baseline . | ||
---|---|---|---|
ESR1LBDm status at baseline . | <5 CTC/7.5 mL WB . | ≥5 CTC/7.5 mL WB . | Total . |
ESR1LBDm+ ctDNA | 8 | 4 | 12 |
ESR1LBDm- ctDNA | 24 | 7 | 31 |
Total | 32 | 11 | 43a |
Abbreviations: ESR1LBDm+, ESR1 mutation detected; ESR1LBDm−: ESR1 mutation “not detected.”
a43/45 patients had both CTC and ctDNA at baseline (two patients only had ctDNA, but not CTC assessed).
CTC enumeration and ESR1LBDm in ctDNA at baseline. A, CTC enumeration for ER aliquot in log scale only for patients with ≥1 CTC/7.5 mL WB. The red horizontal line represents 5 CTC/7.5 mL WB. No bar represents patient with one CTC only since the graph is in log scale. * Enumeration assessed in Ki67 aliquot; two patients were not included because CTC were only drawn at the time of discontinuation and had ≥5 CTC/7.5 mL WB. B, Prevalence and distribution of ESR1LBDm in ctDNA by ddPCR at baseline; * all the cases with ≥2 ESR1LBDm featured a D538G mutation. C, CTC enumeration of ER aliquot at screening and C1D1 (paired T test) in patients with ≥5 CTC/7.5 mL WB; * Enumeration assessed in Ki67 aliquot. D, ESR1 mutational levels (copies/mL plasma) at screening and C1D1 (paired T test). For patients with ≥2 ESR1LBDm, all mutations are reported. “Borderline” cases (i.e., <3 positive droplets, which is the predefined threshold of positivity) were included if at least one definite positive mutant result was detected at one of the two time points. Abbreviations: ESR1, estrogen receptor alpha gene; ND, not detected.
CTC enumeration and ESR1LBDm in ctDNA at baseline. A, CTC enumeration for ER aliquot in log scale only for patients with ≥1 CTC/7.5 mL WB. The red horizontal line represents 5 CTC/7.5 mL WB. No bar represents patient with one CTC only since the graph is in log scale. * Enumeration assessed in Ki67 aliquot; two patients were not included because CTC were only drawn at the time of discontinuation and had ≥5 CTC/7.5 mL WB. B, Prevalence and distribution of ESR1LBDm in ctDNA by ddPCR at baseline; * all the cases with ≥2 ESR1LBDm featured a D538G mutation. C, CTC enumeration of ER aliquot at screening and C1D1 (paired T test) in patients with ≥5 CTC/7.5 mL WB; * Enumeration assessed in Ki67 aliquot. D, ESR1 mutational levels (copies/mL plasma) at screening and C1D1 (paired T test). For patients with ≥2 ESR1LBDm, all mutations are reported. “Borderline” cases (i.e., <3 positive droplets, which is the predefined threshold of positivity) were included if at least one definite positive mutant result was detected at one of the two time points. Abbreviations: ESR1, estrogen receptor alpha gene; ND, not detected.
CTC count, ER positivity on CTC [(CTC-ER positive), % ER-positive CTC], and ESR1 mutational levels tended to remain stable in the absence of treatment (screening vs. C1D1, with a median interval period of 9 days; range, 3–21; paired T test; Fig. 2C and D). One patient had a dramatic reduction, and one patient had a substantial increase in ESR1LBDm ctDNA levels (Fig. 2D). Variation in CTC number and/or ER positivity was observed in three cases. As illustrated in Fig. 3A, patient No. 36 had 12.5% of CTC that were ER positive on 16 CTC at screening, which decreased to 0% on 13 CTC at C1D1. Similarly, patient No. 6 had 7.7% ER positivity on 13 CTC at screening which decreased to 0% on 12 CTC at C1D1. Likewise, patient No. 27 had 21.2% CTC-ER positive on 74 CTC at screening which reduced to 4.3% on 208 CTC at C1D1.
Serial changes in CTC enumeration and characterization during treatment with AZD9496 for patients with ≥5 CTC/7.5 mL WB (N = 11). CTC-biomarker enumeration and staining intensity for ER positivity (A) and Ki67 positivity (B). In both figures, the above bar graphs represent the number of CTC, and each group of bars represents a different patient at different time points; in the below bar graph, the individual colors within each bar provide the percentage of CTC that stained 0 ( blue), 1+ (
red), 2+ (
green), or 3+ (
purple) for ER and Ki67 expression within each patient. ⁁ patient within fulvestrant wash-out period at the time of trial enrollment; * patient was ESR1 (D538G) mutant at discontinuation; ** patient was ESR1LBDm− in ctDNA, but ESR1 (Y537C) mutant in tissue. For the remaining cases, ESR1LBDm status in tissue was not available. Abbreviations: C1D1-pre, cycle 1 day 1 pre-dose; C1D15-post, cycle 1 day 15 post-dose; Disc., discontinuation; ESR1: estrogen receptor alpha gene; Screen, screening.
Serial changes in CTC enumeration and characterization during treatment with AZD9496 for patients with ≥5 CTC/7.5 mL WB (N = 11). CTC-biomarker enumeration and staining intensity for ER positivity (A) and Ki67 positivity (B). In both figures, the above bar graphs represent the number of CTC, and each group of bars represents a different patient at different time points; in the below bar graph, the individual colors within each bar provide the percentage of CTC that stained 0 ( blue), 1+ (
red), 2+ (
green), or 3+ (
purple) for ER and Ki67 expression within each patient. ⁁ patient within fulvestrant wash-out period at the time of trial enrollment; * patient was ESR1 (D538G) mutant at discontinuation; ** patient was ESR1LBDm− in ctDNA, but ESR1 (Y537C) mutant in tissue. For the remaining cases, ESR1LBDm status in tissue was not available. Abbreviations: C1D1-pre, cycle 1 day 1 pre-dose; C1D15-post, cycle 1 day 15 post-dose; Disc., discontinuation; ESR1: estrogen receptor alpha gene; Screen, screening.
Early changes in CTC levels, CTC-ER status, and ESR1 mutational levels in ctDNA during treatment with AZD9496
CTC.
None of the 11 patients with ≥5 CTC/7.5 mL WB at baseline experienced a CTC response defined as a reduction to <5 CTC/7.5 mL WB (Fig. 3A). Of the 10 patients for whom CTC-ER analysis was available at baseline, five had CTC with some degree of ER positivity (2+ and 3+) prior to treatment (Fig. 3A). In two of these five patients (patients No. 30 and No. 27) treated at the dose of 150 mg twice daily and 250 mg BID, respectively, CTC-ER positivity was reduced in the subsequent sample (C1D15) compared with baseline (Fig. 3A).
CTC-ER status was negative in the remaining 5 patients with ≥5 CTC/7.5 mL WB prior to treatment (Fig. 3A). Samples from two patients (patients No. 44 and No. 26) were drawn within the “wash-out” period of fulvestrant, and therefore, the CTC may have exhibited downregulated-negative ER. The remaining three cases (patients No. 35, No. 13, and No. 17), who were outside the fulvestrant “wash-out” period had CTC-ER negative at all time points with an increase in CTC number on treatment. All but two patients with ≥1 CTC/7.5 mL WB, from whom blood specimens were drawn during the wash-out period of fulvestrant, had CTC-ER negative, except patients No. 15 and No. 26, who had one and two CTC-ER positive, respectively (Supplementary Table S4). The mean Ki67 positivity on CTC (CTC-Ki67 positivity) was 11.7% (range, 0%–21%) in all the patients at baseline and did not change appreciably over time [mean at C1D15 was 15.2% (range, 0%–45%); Fig. 3B]. There was no apparent correlation between changes in CTC-Ki67 status and CTC-ER status.
ESR1 mutational status in ctDNA.
ESR1LBD mutations exhibit variable dynamics during AZD9496 therapy (Fig. 4A). A decline of ESR1LBDm mutational levels (dominant mutation) of ≥50% at C1D15 versus baseline was observed in eight of the 14 (57%) patients with detectable ESR1LBDm at baseline (Fig. 4B), without apparent dose–response relationship (Fig. 4B). In three of these eight patients, the ESR1LBDm+ ctDNA became ESR1LBDm− ctDNA at C1D15. In patient No. 1, ESR1LBDm+ ctDNA was detected again at C3D1 when she discontinued treatment, whereas, in patients No. 29 and No. 21, clearance was durable until the last on-treatment sample tested (C7D1 and C9D1), respectively. Within the blood specimens from the five patients with multiple ESR1 mutations, serial levels demonstrated both convergent and divergent evolution of the different subclones during treatment with AZD9496 (Fig. 4C).
Dynamics of ESR1LBDm in ctDNA during treatment with AZD9496. A, ESR1 mutational levels tracking shown for 17 patients with longitudinal plasma collected throughout treatment and who were ESR1LBDm+ by ddPCR analysis at any time point (14 patients with ESR1LBDm+ before treatment and three patients with ESR1LBDm absent prior to treatment, but detected in on-treatment ctDNA samples). For patients with ≥2 ESR1LBDm, the mutation with highest levels (copies/mL plasma) pretreatment (“dominant mutation”) is shown. B, A decline of circulating ESR1LBDm of ≥50% at C1D15 compared to pretreatment was evident in eight of 14 (57%) of patients with no apparent dose–response relationship. *ESR1 LBDm status was “borderline” (i.e., below predefined threshold of positivity) at C1D1, but clearly detectable at another time point. #Still receiving treatment at data cutoff of January 31, 2017. C, ESR1LBDm tracking in five cases with ≥2 ESR1LBDm+ during treatment with AZD9496. In one case (patient No. 21), ESR1 (D538G and Y537N) mutations were detected at screen but not at C1D1. #Still receiving treatment at data cutoff of January 31, 2017. Abbreviations: C9D1, cycle 9 day 1; disc., discontinuation; ESR1: estrogen receptor alpha gene; ND: not detected.
Dynamics of ESR1LBDm in ctDNA during treatment with AZD9496. A, ESR1 mutational levels tracking shown for 17 patients with longitudinal plasma collected throughout treatment and who were ESR1LBDm+ by ddPCR analysis at any time point (14 patients with ESR1LBDm+ before treatment and three patients with ESR1LBDm absent prior to treatment, but detected in on-treatment ctDNA samples). For patients with ≥2 ESR1LBDm, the mutation with highest levels (copies/mL plasma) pretreatment (“dominant mutation”) is shown. B, A decline of circulating ESR1LBDm of ≥50% at C1D15 compared to pretreatment was evident in eight of 14 (57%) of patients with no apparent dose–response relationship. *ESR1 LBDm status was “borderline” (i.e., below predefined threshold of positivity) at C1D1, but clearly detectable at another time point. #Still receiving treatment at data cutoff of January 31, 2017. C, ESR1LBDm tracking in five cases with ≥2 ESR1LBDm+ during treatment with AZD9496. In one case (patient No. 21), ESR1 (D538G and Y537N) mutations were detected at screen but not at C1D1. #Still receiving treatment at data cutoff of January 31, 2017. Abbreviations: C9D1, cycle 9 day 1; disc., discontinuation; ESR1: estrogen receptor alpha gene; ND: not detected.
Three patients (patients No. 31, No. 37, and No. 12) who had ESR1LBDm− ctDNA at baseline had ESR1LBDm+ ctDNA at a later time point. The immediate prior therapy for these three patients was everolimus plus exemestane (patient No. 31; wash-out 35 days), fulvestrant ± BYL719 (patient No. 37; wash-out 35 days), and letrozole plus palbociclib (patient No. 12; wash-out 20 days). Overall, when considering any given time point, 17 of 45 (38%) patients had ESR1LBDm+ ctDNA (Fig. 4A).
Clinical outcomes associated with CTC and ctDNA analyses
The median duration on treatment with AZD9496 was 64 days (range, 22–643 days, across the wide range of doses examined; ref. 13; Supplementary Fig. S1), and six patients (13.3%) were still on treatment up to the data cutoff of January 31, 2017. Of the 39 patients who discontinued treatment, 38 discontinued due to progressive disease or death, and one due to an adverse event.
The prognostic effect of the baseline circulating biomarkers was investigated. Consistent with prior reports (21, 22, 26), patients with ≥5 CTC/7.5 mL WB at baseline had significantly worse PFS than those with <5 CTC/7.5 mL WB (median PFS of 54 vs. 164 days; log-rank P = 0.0003; Fig. 5A). In contrast, ESR1LBD mutational status at baseline was not predictive of outcome on AZD9496 in all patients (Fig. 5B) and when considering only the subgroup of patients with <5 CTC/7.5 mL WB (Fig. 5C).
Circulating biomarkers data and prediction of treatment outcome. Kaplan–Meier plots for (A) PFS of all patients according to CTC levels (<5 vs. ≥5 CTC/7.5 mL WB) at baseline. Red line, ≥5 CTC/7.5 mL WB; black line, <5 CTC/7.5 mL WB. B, PFS of all patients according to ESR1LBDm status [ESR1LBDm negative (−) vs. ESR1LBDm positive (+)] at baseline. Red line, ESR1LBDm+; Black line, ESR1LBDm−. C, PFS of patients with <5 CTC/7.5 mL WB according to ESR1 status (ESR1LBDm− vs. ESR1LBDm+) at baseline. D, PFS of all patients according to ESR1LBDm status (*ND = never detectable or “ctDNA response” defined as drop ≥50% at C1D15 in the dominant ESR1LBDm vs. residual ESR1LBDm) at C1D15. Red line, residual ESR1LBDm+; black line, ND or ctDNA response. E, PFS of patients with <5 CTC/7.5 mL WB according to ESR1LBDm status (*ND = never detectable or “ctDNA response” in the dominant ESR1LBDm vs. residual ESR1LBDm) at C1D15. Red line, residual ESR1LBDm+; black line, ND or ctDNA response. F, PFS of all the patients according to persistently elevated biomarker (** ≥5 CTC/7.5 mL WB and/or lack of ctDNA response defined as a decline of <50% of ESR1LBDm level) at C1D15. Red line, persistently elevated biomarkers at C1D15; black line, other. Abbreviations: ESR1, estrogen receptor alpha gene; ND, never detectable.
Circulating biomarkers data and prediction of treatment outcome. Kaplan–Meier plots for (A) PFS of all patients according to CTC levels (<5 vs. ≥5 CTC/7.5 mL WB) at baseline. Red line, ≥5 CTC/7.5 mL WB; black line, <5 CTC/7.5 mL WB. B, PFS of all patients according to ESR1LBDm status [ESR1LBDm negative (−) vs. ESR1LBDm positive (+)] at baseline. Red line, ESR1LBDm+; Black line, ESR1LBDm−. C, PFS of patients with <5 CTC/7.5 mL WB according to ESR1 status (ESR1LBDm− vs. ESR1LBDm+) at baseline. D, PFS of all patients according to ESR1LBDm status (*ND = never detectable or “ctDNA response” defined as drop ≥50% at C1D15 in the dominant ESR1LBDm vs. residual ESR1LBDm) at C1D15. Red line, residual ESR1LBDm+; black line, ND or ctDNA response. E, PFS of patients with <5 CTC/7.5 mL WB according to ESR1LBDm status (*ND = never detectable or “ctDNA response” in the dominant ESR1LBDm vs. residual ESR1LBDm) at C1D15. Red line, residual ESR1LBDm+; black line, ND or ctDNA response. F, PFS of all the patients according to persistently elevated biomarker (** ≥5 CTC/7.5 mL WB and/or lack of ctDNA response defined as a decline of <50% of ESR1LBDm level) at C1D15. Red line, persistently elevated biomarkers at C1D15; black line, other. Abbreviations: ESR1, estrogen receptor alpha gene; ND, never detectable.
As noted, none of the 11 patients with elevated CTC at baseline experienced a decline to <5 CTC/7.5 mL WB at C1D15. Using relative changes in CTC enumeration from baseline to C1D15 (±50% from baseline), numerically similar median PFS (mPFS) was seen between patients who had an increase (changes ≥50%) in CTC number (N = 5; mPFS of 45 days), stable (changes <50%) CTC number (N = 3; mPFS of 56 days), or a decrease (changes ≥50%) in CTC number (N = 3; mPFS of 54 days). Six of the seven patients with available CTC enumeration at treatment discontinuation had a ≥50% increase in CTC at C1D15 (range from 195% to 7,538%). By protocol, if a patient had <5 CTC/7.5 mL WB at screening/C1D1, subsequent blood draws were not collected for CTC analysis, and therefore, we could not address whether these patients had developed ≥5 CTC/7.5 mL WB on C1D15.
Patients with ESR1LBDm− at baseline or decline of ≥50% at C1D15 had a trend toward longer PFS compared with those with residual ESR1LBDm+ ctDNA at C1D15, when the analysis included all patients (N = 38; mPFS = 111 days vs. N = 6; mPFS = 54 days; log-rank P = 0.055; Fig. 5D). The subgroup of patients with <5 CTC/7.5 mL WB showed a similar trend (N = 28; mPFS = 166 days vs. N = 3; mPFS = 109 days), but numbers were too small to draw meaningful statistical conclusion (Fig. 5E). At C1D15, patients with either persistently elevated CTC or ESR1LBDm+ ctDNA had a worse PFS (N = 15; mPFS = 55 days) compared with patients with neither biomarker elevated (N = 30; mPFS = 166 days; log-rank P = 0.0007; Fig. 5F).
Of note, one patient (No. 43) achieved partial and durable response on AZD9496 accompanied by sustained reduction of the tumor marker CA15-3 (13). She had ESR1LBDm− ctDNA at baseline and during treatment, and 0 CTC/7.5 mL WB at baseline (Supplementary Fig. S1). The patient was treated with 250 mg BID AZD9496 and stopped treatment due to progressive disease after approximately 26.4 months on therapy.
Tissue markers
Paired biopsies from five patients were considered evaluable for PD analysis (Fig. 1; Supplementary Fig. S2). Timing of the on-treatment samples and number of hours after last dose of AZD9496 are shown in Supplementary Fig. S2A. These five patients were treated at various AZD9496 doses ranging from 40 to 400 mg BID.
We identified significant limitations (i.e., sample characteristics and/or time point of collection) associated with the paired tumor samples collected from metastatic lesions in this trial. A reduction in ER H score was detected in paired biopsies from two subjects (31% and 47%), of whom the first one was accompanied by reduction in Ki67 (52%). Details are described in Supplementary Fig. S2B.
Overall, variable and inconsistent changes were seen in biomarker expression (Supplementary Fig. S2B) without any clear dose–response. Taken together, this limited sample set did not allow a robust assessment of the PD activity of AZD9496, speaking to the importance of circulating tumor biomarker evaluation.
PK/PD relationship analysis
We examined whether there was a correlation between doses of AZD9496 and the PK parameters (AUC, Cmax) for AZD9496 and the changes over time for CTC, ESR1LBDm ctDNA levels, and CA15-3. No graphical correlation was found between the dose nor the PK parameters of AZD9496 and CTC number or status, ESR1LBDm ctDNA levels, or CA15-3 at any time point studied. With respect to model-based assessments of CA15-3, no significant relationships between dose or PK parameters (AUC, Cmax) and change in CA15-3 (linear slope parameter) could be detected over the course of the study at any time point.
Discussion
In this correlative study of a prospective first-in-patient phase I trial, we have investigated the PD activity of an oral SERD, AZD9496, through blood-borne and tumor tissue biomarkers. We have also explored potential predictive markers of outcome on AZD9496 in ER-positive MBC patients.
As expected, pretreatment CTC levels (≥5 CTC/7.5 mL WB) were a strong prognostic marker in this cohort of MBC, as seen in prior studies (21, 22). Twenty-five percent of patients had ≥5 CTC/7.5 mL WB at baseline, but none had a reduction of CTC numbers to <5 CTC/7.5 mL WB on treatment (C1D15). The lack of reduction suggests that, at least in this patient group, the agent failed to induce an early “CTC response,” defined as a reduction to <5 CTC/7.5 mL WB. It remains unclear whether this is due to the insufficient target inhibition of the agent or due to the C1D15 time point's being too early to assess changes in CTC during ET monotherapy. Additional data regarding CTC numbers and ET at different time points are under investigation within a prospective phase 2 clinical trial (COMETI P2 NCT01701050; ref. 27).
Of the 10 patients for whom CTC-ER status was available at baseline, only five had pretreatment CTC-ER positivity, whereas the others with CTC-ER negative were not informative for PD. In two of the five patients, there was a reduction of CTC-ER positivity at C1D15 versus baseline, although it is difficult to discern if this observation was due to a true PD effect or due to analytic variability, because we did not have an untreated patient group as control.
The other five patients with ≥5 CTC/7.5 mL WB at baseline had CTC-ER negative (Fig. 3A). We speculate that two of these cases may have represented “downregulated” CTC-ER, since they had received recent fulvestrant treatment with possible impact on ER expression in CTC due to its mechanism of action. Three cases, which were not recently treated with or were beyond the fulvestrant “wash-out” period, had CTC-ER negative at all time points, and all three experienced an increase in CTC number during treatment with AZD9496. We speculate that in these patients, ER-negative clones may have emerged, and therefore endocrine independent metastases, even though their primary cancers were ER positive. We have previously reported data regarding CTC enumeration and CTC characterization of ER and BCL2 in patients progressing on fulvestrant and have shown potential ability for CTC to provide insights into the potential mechanisms of resistance to this drug (17).
Because serial tissue Ki67 reduction is associated with benefit from neoadjuvant ET (28), we explored whether serial Ki67 values on CTC would reflect response to AZD9496. Proliferative status of CTC using Ki67 expression has been investigated in several diseases including breast cancer (19, 29–32). However, in this study, the percentage of CTC-Ki67 positivity within each patient at baseline was relatively low (<21%) and did not change substantially in any patient. Although we only collected five paired biopsies, of interest, a paired liver biopsy was obtained from patient No. 30 at roughly the same time as CTC collection, and tissue Ki67 decreased by approximately 50% in a set of paired baseline follow-up biopsies. This observation suggests that CTC-Ki67 status may not reflect tissue Ki67. We speculate, but would require further study to demonstrate, that CTC may stop proliferating when they are in circulation.
ESR1LBDm+ ctDNA was detected in 31% of patients at baseline, and about a third of these patients had at least two different ESR1LBD mutations. Ninety percent of all patients had received prior AI therapy, consistent with the evidence that profound estrogen depletion may select for ESR1m (9). We observed that neither CTC enumeration, CTC-ER status, nor ESR1LBD mutational levels changed significantly between screening and treatment initiation (C1D1), confirming the analytic validity of serial assays for these markers in the absence of intervening therapy in this disease and setting. To our knowledge, this is the first study reporting “double baseline” results on ctDNA mutational levels and CTC.
The presence or absence of ESR1LBDm+ ctDNA at baseline was not associated with outcome in this study. Serial tracking of ESR1LBDm showed variable patterns of dynamics during AZD9496 therapy. A “ctDNA response” (decline in ESR1LBDm levels of ≥50% at C1D15 compared with baseline) was observed in 57% of patients (eight of 14) who harbored ESR1LBDm at baseline, of whom three had a complete and durable clearance of the dominant mutation. Patients with lack of ctDNA response had shorter time of study versus patients with ctDNA response or ESR1LBDm− ctDNA. There was no apparent relationship between high mutational levels at baseline and later ctDNA response. In patients with “ctDNA response” at C1D15, ESR1LBDm levels increased at subsequent time point in some cases, whereas some others maintained suppressed ESR1LBDm levels throughout treatment, even at the time of progression. Although early changes of ESR1 mutational levels in ctDNA might be a valid tool to assess PD of an SERD, tracking ESR1LBDm+ ctDNA levels to anticipate relapse could be of limited value in some cases, possibly due to the fact that ESR1LBDm are frequently subclonal (i.e., they represent a subclone in a substantial fraction of patients, poorly representing the cancer overall). This observation is supported by recent findings by O'Leary and colleagues who investigated the dynamics of PIK3CAm and ESR1LBDm during treatment with fulvestrant plus palbociclib/placebo and described that ESR1LBDm can be lost even when patients are experiencing disease progression and may not determine clinical outcome (33). Taken together, these results indicate that genetic heterogeneity might be itself another mechanism of resistance (34). Among the ESR1LBDm+ cases, five had multiple ESR1LBDm in ctDNA, and mutation tracking revealed that both convergent and divergent evolution of different subclones is possible. It is uncertain whether this is dependent on varying affinity of AZD9496 to the different mutant receptors or other factors associated with clonal abundance of the concurrent ESR1LBD mutations. Of note, we do not report here on the more comprehensive assessment of tumor genetics. However, investigating the dynamics of tumor mutational burden or truncal mutations during treatment with AZD9496 and their correlation with clinical outcomes warrants further exploration. Additional genomic analyses of DNA derived from CTC would also help to clarify if these mutations are polyclonal or if they are present in the same DNA sequence. Prior studies suggest that both circumstances are possible (11, 35, 36).
When we correlated the blood-borne biomarkers with clinical outcome, we found that the presence of elevated CTC (≥5 CTC/7.5 mL WB) at baseline was a strong prognostic factor in this cohort (log-rank P = 0.0003). In contrast, ESR1LBDm status at baseline was not prognostic in all the patients or in those with <5 CTC/7.5 mL WB. Interestingly, patients with either persistently elevated CTC number or ESR1LBDm+ ctDNA at C1D15 compared with baseline had a worse PFS than patients with neither biomarker elevated (log-rank P = 0.0007). Although the small sample size limits the opportunity for a meaningful multivariate analysis, there was little overlap in the patient populations with respect to persistent elevations of CTC number and ESR1LBDm+ at C1D15. Only two of 15 patients in this analysis had both persistently elevated CTC number and ESR1LBDm status, and persistent elevations of both factors appeared to be independently prognostic. Of note, none of the patients with elevated CTC at baseline converted to nonelevated and residual ESR1m at C1D15 was prognostic (Fig. 5D).
Limitations of our exploratory study include the small sample size, the fact that patients were treated at different doses of AZD9496 and without a comparator arm, and the lack of conclusive data on the pharmacodynamically and clinically effective dose of AZD9496. Additional limitations are that we only assessed the most common ESR1LBD mutations, and we tested relatively limited volumes of plasma. None of the PD markers' early changes correlated with the AZD9496 dose or PK parameters. It remains uncertain whether C1D15 is a too early time point to assess PD changes during ET monotherapy. Only 11% of the patients (5/45) provided evaluable paired biopsies, preventing any conclusive data on the PD effect of AZD9496. These results illustrate the challenge of collecting serial tissue specimens from advanced cancer patients to enable a thorough assessment of proof of mechanism in early phase trials.
Despite these limitations, >95% of patients enrolled in this phase I trial provided blood samples for circulating biomarkers, whereas collection of tissue biopsies was limited to only few participants, emphasizing one of the potential advantages of liquid biopsies. In addition, we observed several potential general mechanisms of resistance to ET (37). The first potential resistance mechanism observed was the loss of ER expression on CTC in ER-positive MBC. In particular, of the 10 patients who had ≥5 CTC/7.5 mL WB at baseline, three patients had CTC-ER negativity after excluding the two patients who were within the fulvestrant “wash-out” period. A second potential mechanism was the identification of a mutation of the ESR1 gene either by ctDNA at baseline or in tissue, with five patients harboring >1 ESR1LBDm+ ctDNA. We could also hypothesize that additional mechanisms may be present when CTC-ER positivity is still present along with ESR1LBDm− ctDNA, but these mechanisms were not assessed in this current trial. It should be noted that, although one patient treated with 250 mg BID AZD9496 achieved a confirmed and durable partial response (13), there is still uncertainty on the dose–response relationship for AZD9496. Moreover, the impact of the PK characteristics of AZD9496 [a relatively short half-life (alpha half-life of 1–2 hours) and a dose-dependent decrease in exposure upon multiple doses, presumably due to CYP autoinduction] on the PD activity is unclear, limiting our ability to be conclusive on the correlation between these biomarker findings and benefit from AZD9496. A presurgical window of opportunity study (NCT03236974) is currently ongoing and is aimed to compare the PD effects of AZD9496 with those of fulvestrant in women with ER-positive early breast cancer awaiting surgery with curative intent, and to characterize the PK/PD relationship. In summary, possible mechanisms of ET resistance may differ for patients with CTC-ER positive versus CTC-ER negative. For the ER-positive component, failure to downregulate ER may be due to ESR1LBD mutations or because they received a nonefficacious dose of AZD9496. In the CTC-ER negative component, it may be due to the emergence of ER-negative/endocrine-independent clones, or perhaps due to a downregulation of CTC-ER positive accompanied by a concomitant upregulation of other secondary pathways.
Taken together, these results suggest that complementary and comprehensive characterization of CTC and ctDNA may be required for noninvasive early prediction of outcome during ET. Once validated, these tools have the potential to be used in prospective trials to select optimal therapy upfront, or after short exposure to treatment. The latter could involve switching to different ET agent, continuing the same ET and adding other targeted treatment such as mTOR or CDK4/6 inhibitors, or switching to non-endocrine treatment.
Disclosure of Potential Conflicts of Interest
G. Schiavon, T.H. Carr, M. Hoch, J. Lindemann, and S. Morgan have ownership interests (including patents) in AstraZeneca. E. Hamilton is a consultant/advisory board member for Genentech/Roche, Lilly, and Pfizer. A. Armstrong has ownership interests (including patents) in AstraZeneca. S.-A. Im is a consultant/advisory board member for AstraZeneca, Hanmi, Novartis, Pfizer and Roche. C. Dive reports receiving speakers bureau honoraria from and reports receiving commercial research grants from AstraZeneca. E.A. Harrington has ownership interests (including patents) at AstraZeneca. R. Baird reports receiving commercial research grants from AstraZeneca. D.F. Hayes has ownership interests (including patents) at Inbiomotion and OncImmune, reports receiving commercial research grants from Menarini Silicon Biosystems, and AstraZeneca, reports receiving commercial research support from Eli Lilly, Merrimack, Puma, and Pfizer. C. Paoletti received research funding from AstraZeneca and Pfizer and received travel paid by Menarini Silicon Biosystems, the maker of CellSearch, the system used to capture and enumerate CTC. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: C. Paoletti, G. Schiavon, M. Hoch, G. Marshall, S. Morgan, S.-A. Im, C. Dive, E.A. Harrington, J.C. Barrett, R. Baird, D.F. Hayes
Development of methodology: C. Paoletti, G. Schiavon, J. Geradts, G. Marshall, E. Hamilton, A. Armstrong, D.F. Hayes
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Paoletti, T.H. Carr, J. Geradts, J. Lindemann, S. Morgan, V. Rowlands, E. Hamilton, M. Patel, A. Armstrong, K. Jhaveri, S.-A. Im, C. Dive, R. Baird, D.F. Hayes
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Paoletti, G. Schiavon, T.H. Carr, J. Geradts, M. Hoch, J. Lindemann, S. Morgan, P. Patel, V. Rowlands, M. Patel, N. Iqbal, E.A. Harrington, J.C. Barrett, R. Baird, D.F. Hayes
Writing, review, and/or revision of the manuscript: C. Paoletti, G. Schiavon, E.M. Dolce, E.P. Darga, T.H. Carr, J. Geradts, M. Hoch, T. Klinowska, J. Lindemann, S. Morgan, V. Rowlands, K. Aung, E. Hamilton, M. Patel, A. Armstrong, K. Jhaveri, S.-A. Im, C. Dive, E.A. Harrington, J.C. Barrett, R. Baird, D.F. Hayes
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Paoletti, G. Schiavon, E.M. Dolce, E.P. Darga, T.H. Carr, S. Morgan, N. Sathiyayogan, K. Aung, E. Hamilton, M. Patel, S.-A. Im, N. Iqbal, F. Butt
Study supervision: C. Paoletti, G. Schiavon, T. Klinowska, J. Lindemann, S. Morgan, N. Sathiyayogan, M. Patel, A. Armstrong, C. Dive, E.A. Harrington, J.C. Barrett, D.F. Hayes
Acknowledgments
We would like to thank all the investigators and site staff, with special thanks to the patients and families. We also wish to thank the following contributors to this study: Stuart Findlay, Graham Fisher, Catherine Geh, Amandine Gojon, William Howat, Audrey Lawrence, Grace Lu, Tony Nash, Sabina Patel, Shyamprasad Perisetla, Alison Pritchard, Andrzej Rutkowski, Claire Smith, and Jonathan Tugwood. We also acknowledge Cancer Research UK, ECMC, and NIHR/BRC infrastructure funding for the Cambridge Cancer Centre.
This study was sponsored by AstraZeneca.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.