Clinical Significance of Circulating Tumor Cells in Hormone Receptor–positive Metastatic Breast Cancer Patients who Received Letrozole with or Without Bevacizumab

Hormone receptor–positive (HR⁺) breast cancer represents approximately 70% of all breast cancers (1, 2). The standard of care for metastatic breast cancer (MBC) includes sequential endocrine therapy (ET) alone or in combination with targeted agents (1, 2). In the Cancer and Leukemia Group B (CALGB, now part of Alliance for Clinical Trials in Oncology) 40503 trial, the addition of bevacizumab (an antibody to VEGF-A) to letrozole (an aromatase inhibitor) prolonged progression-free survival (PFS) but not overall survival (OS) in postmenopausal women with HR⁺ MBC (3). The mechanisms involved in resistance to ET are not fully understood and remain an active area of research (4). Biomarkers that can identify patients whose tumors are more likely to respond or develop resistance to ET and ET combinations are an unmet need.

Blood-based biomarkers, for example, circulating tumor cells (CTC), offer a minimally invasive approach for assessing prognosis and monitoring of disease burden in MBC (5, 6). Increased levels of CTCs prior to treatment is highly prognostic for disease progression and death (6–8). Moreover, failure to clear CTCs early in treatment is associated with poor response to therapy (6, 9–13). In principle, CTCs can facilitate monitoring of disease status and tumor response, and thus enable the potential use of more effective therapy earlier in the disease course.

Efforts to demonstrate the clinical utility of CTCs have been actively pursued (8). For example, the STIC CTC trial recently examined the potential role of CTCs in early treatment modification in HR⁺ HER2-negative MBC (14). Investigators found that switching to chemotherapy in patients with high levels of CTCs prior to treatment (≥5 CTC/7.5 mL of blood) resulted in significant improvements in PFS compared with patients who received standard hormone therapy (14). Smerage and colleagues conducted a single, prospective, randomized study (SWOG0500 trial) to examine whether serial monitoring of CTCs could guide treatment decisions in MBC (15). While the study failed to demonstrate that treatment modification based on CTC response at first follow-up could improve outcomes, it did confirm previous observations showing that patients who have high CTC counts at baseline (≥5 CTC/7.5 mL of blood) had worse outcomes, regardless of treatment, compared with those with <5 CTCs (15).

In this study, we hypothesized that CTCs could serve as a prognostic and predictive marker in HR⁺ MBC treated with letrozole or letrozole +bevacizumab in the first-line setting. To address this hypothesis, we performed an ancillary study in the CALGB 40503 trial to evaluate whether baseline and changes in serial CTC levels were associated with PFS and OS, and whether baseline CTCs could predict benefit from the addition of bevacizumab to letrozole (16).

This is a preplanned study to examine the clinical significance of CTCs in the CALGB 40503 trial (NCT00601900; ref. 3). This trial was a multicenter randomized phase III study that compared the efficacy of letrozole alone with letrozole given in combination with bevacizumab (antibody against VEGF-A) as first-line endocrine-based therapy in postmenopausal women with HR⁺ advanced breast cancers. Patients who received more than one prior chemotherapy for MBC were not eligible. Prior adjuvant or neoadjuvant chemotherapy was allowed. The study design and efficacy results have been previously reported (3). Patients were enrolled between December 2008 and December 2011. The institutional review boards at the NCI and at each site approved the study. All participants provided a written informed consent that included the use of collected specimens. The study was performed in accordance with the Declaration of Helsinki.

Blood was collected at 4 time points: baseline and before every third bevacizumab cycle (3-week cycles): 2 (T1), 3 (T2), and 4 (T3) or approximately 21-day intervals in the letrozole only arm. Samples were drawn into CellSave preservative tubes (Menarini Silicon Biosystems, LLC) at each participating site and shipped to the University of California San Francisco (John W. Park Laboratory) for analysis. CTC enumeration was performed by investigator (J.H. Scott) who was blinded to the clinical data.

CTCs were enumerated within 96 hours using the CellSearch system (Menarini Silicon Biosystems, LLC) following the manufacturer's instructions without modification (17). Briefly, 7.5 mL of blood was subjected to immunomagnetic enrichment to capture EPCAM-positive cells using the CellSearch Circulating Tumor Cell Kit. This was followed by immunofluorescence microscopy to enumerate CTCs, which were defined as nucleated (DAPI-positive) cells of epithelial origin (cytokeratin-positive and CD45-negative). Samples with ≥5 CTCs per 7.5 mL of blood were considered CTC positive.

The primary endpoint of the study was PFS, defined as the interval between study entry and first documented disease progression or death without progression. A secondary endpoint was OS, defined as time from study entry to death from any cause. Event-free patients were censored at their last clinical evaluation. Stratification factors (disease measurability and disease-free interval), age, and HER2 status were included as covariables in the multivariable models. Survival analysis was performed on follow-up data available as of July 31, 2019.

Patient and tumor characteristics were summarized according to CTC status and the proportions between groups (CTC-positive vs. CTC-negative) were compared using Pearson χ² test.

Survival curves were estimated by the Kaplan–Meier method and compared using the log-rank test (18). Multivariable Cox regression models adjusted for known prognostic factors were used to estimate HRs and 95% confidence intervals (CI). Restricted mean survival times and differences were calculated for different time points (19). The prognostic effect of changes in CTC status between baseline and other time points was tested using a time-dependent Cox model. Data collection and statistical analyses were conducted by the Alliance Statistics and Data Center. Statistical analysis was performed using R (version 3.6.0) and SAS software (version 9.4).

Of the 391 patients randomized in the CALGB 40503 (3), 48 were excluded due to missing adverse events, treatment, or disease evaluation data (Fig. 1A). Of the remaining 343, 3 had no stratification data and 46 did not have pretreatment CTC data and were excluded from the current analysis. The baseline analysis cohort consisted of 294 patients, of whom, 154 received letrozole + bevacizumab and 140 received letrozole only. No significant differences in characteristics were observed between patients in the original study cohort and those in this study (Supplementary Table S1). The proportion of patients with impaired functioning (Eastern Cooperative Oncology Group performance status 1) was higher in the group that was excluded from this study.

A summary of the patient and tumor characteristics of the study cohort by baseline CTC status is shown in Table 1. The median age was 58 years old; 52% received both letrozole and bevacizumab; 99% were ER positive and 19% were HER2 positive; 63% had measurable disease; and 49% received prior hormone therapy. The treatment arms were balanced within the CTC-positive and CTC-negative patient groups.

Evaluation for CTCs was performed in 7.5 mL of blood at 4 time points using CellSearch (Fig. 1B). Of the 294 patients, 92 (31.3%) were CTC positive at baseline. CTC positivity was significantly associated with bone-only metastasis (P < 0.01; Table 1). The overall CTC-positive rates decreased over time (T1: 23% of 233; T2: 20% of 172, T3: 15% of 196; Fig. 1C).

We examined the association of CTC levels at baseline and serial CTC measurements with clinical outcomes. The median follow-up time for the patients in this study was 39 months. The first three analyses below were performed irrespective of arm assignment.

CTC-positive patients had a significantly shorter median PFS (13.6 months, 95% CI, 8.4–16.9; Fig. 2A) and OS (32.5 months, 95% CI, 26.6–36.1; Fig. 2B) compared with CTC-negative patients (PFS: 16.9 months, 95% CI, 14.1–19.3; OS: 47.5 months, 95% CI, 42.9–50.2). In a multivariable Cox regression analysis that included other prognostic variables, CTCs remained a significant negative prognostic factor for PFS (HR, 1.79; 95% CI, 1.35–2.36; Fig. 2C; Supplementary Table S2) and OS (HR, 2.72; 95% CI, 1.98–3.73; Fig. 2D; Supplementary Table S2).

We assessed whether change in CTC status from baseline to the first time point (T1) during therapy was associated with patient outcome. We identified 4 groups according to serial CTC status: patients who were positive at baseline and remained (i) positive (42 of 219, 19%) or (ii) became negative (37 of 219, 17%) at the T1 measurement, and patients who were negative at baseline and (iii) became positive at T1 (11 of 219, 5%); or (iv) remained negative at T1 (129 of 219, 59%). There were significant differences in the PFS (P = 0.02; Fig. 2D) and OS among the four groups (P < 0.01; Fig. 2E). Multivariable Cox regression analysis revealed that patients who remained CTC-positive at T1 had a significant increased risk of progression (HR, 2.15; 95% CI, 1.43–3.23) and death (HR, 2.7; 95% CI, 1.66–4.38) compared with those who remained CTC negative at T1 (Supplementary Table S3). Similarly, patients who became CTC positive at T1 had a significant increased risk of death compared with those who remained CTC negative (HR, 3.2; 95% CI, 1.57–6.51).

We assessed whether change in CTC status throughout therapy was associated with patient outcome. In this analysis, the CTC status was treated as a time-dependent variable. At baseline, patients were classified as CTC positive or CTC negative. At each time point, patients were reassigned to CTC positive or CTC negative if their status changed. Multivariable Cox regression analysis revealed that patients with CTC-positive status at any time (baseline or at a follow-up time point) had significant increased risk of progression (HR, 2.2; 95% CI, 1.58–3.07) and death (HR, 3.4; 95% CI, 2.36–4.88) compared with patients who were CTC negative (Supplementary Table S4).

Next, we stratified patients into 4 groups according to CTC status (at baseline) and treatment arm (Table 2). Patients who were CTC positive and received letrozole only had the worse PFS (adjusted likelihood ratio P < 0.01) and OS (adjusted likelihood ratio P < 0.01; Fig. 3A).

For CTC-negative patients, there was no significant difference in median PFS (letrozole: 14.7 months vs. letrozole + bevacizumab: 18.4 months, adjusted likelihood ratio P = 0.18) and OS (letrozole: 45 months vs. letrozole + bevacizumab: 49.1 months, adjusted likelihood ratio P = 0.2) between treatment arms (Fig. 3B).

For CTC-positive patients, there was no significant difference in the median OS between arms (letrozole: 27.1 months vs. letrozole + bevacizumab: 33.6 months, adjusted likelihood ratio P = 0.5). Interestingly, median PFS was significantly longer in the letrozole + bevacizumab (18.0 months) versus letrozole (7.0 months; adjusted likelihood ratio P = 0.009).

We evaluated whether CTCs at baseline were predictive of treatment efficacy. The tests for interaction between baseline CTCs (positive vs. negative) and bevacizumab (yes vs. no) were not statistically significant for PFS (P = 0.87) or OS (P = 0.99).

We assessed whether change in CTC status from baseline to the first time point (T1) during therapy was associated with patient outcome in each of the treatment arms. We observed that patients in the letrozole arm who failed to clear CTCs (CTC⁺CTC⁺) had the worse PFS (adjusted likelihood ratio P = 0.05; Fig. 3C) and OS (adjusted likelihood-ratio P < 0.01; Fig. 3D). This observation was less apparent in the letrozole + bevacizumab arm. Patients who remained CTC-negative (CTC⁻CTC⁻) had the most favorable PFS (adjusted likelihood-ratio P < 0.01, Fig. 3E). Patients who were CTC negative at baseline and became CTC positive at T1 (CTC⁻CTC⁺) had the worse OS (adjusted likelihood ratio P < 0.01; Fig. 3F). Interestingly, OS of patients who were initially CTC positive and either became negative (CTC⁺CTC⁻) or remained positive (CTC⁺CTC⁺) was better compared with CTC⁻CTC⁺ group.

We calculated the restricted mean survival time differences at 6, 12, 18, and 24 months between patients who received letrozole + bevacizumab versus letrozole(Table 3). Results revealed significant PFS benefit with the addition of bevacizumab for both CTC-positive and CTC-negative patients. For example, at 24 months, disease progression was, on average, delayed by 5.9 months (95% CI, 2.4–9.4) and 2.5 months (95% CI, 0.1–5) in CTC-positive and CTC-negative patients, respectively.

Significant differences in mean OS of 2.1 months (95% CI, 0.3–3.8) and 3.0 months (95% CI, 0.4–5.6) were observed at 18 and 24 months, respectively, in CTC-positive patients who received letrozole + bevacizumab versus those who received letrozole alone. (Table 3). No significant differences were observed in earlier time points. Among the CTC-negative patients, there was no significant difference in mean OS between arms at all time points examined.

The CALGB 40503 trial was conducted to examine the efficacy of bevacizumab in extending PFS and OS when added to first-line letrozole in HR⁺ MBC (3). The study was activated in 2008 soon after the FDA granted accelerated approval of bevacizumab (in combination with first-line chemotherapy) for treatment of HER2-negative MBC. This approval was revoked in 2011 because of lack of evidence in prolonged OS, and while improvement in PFS was observed, the toxicities associated with bevacizumab remained significant (20). Subsequently, CALBG 40503 reported results that were consistent with this assessment (3). In contrast, a neoadjuvant study in early-stage breast cancer (NSABP B40: chemotherapy with or without bevacizumab in treating women with stage I, stage II, or stage IIIA breast cancer that can be removed by surgery) showed improvement in OS (21). These findings suggest that there may be subgroup(s) of patients who may benefit from bevacizumab treatment, and a biomarker which can identify this subset is clearly warranted.

We performed a prospective CTC study in patients enrolled in CALGB 40503. We enumerated CTCs in serially collected blood samples and evaluated the prognostic impact of these cells. Results of our study showed that baseline levels of CTCs were highly prognostic for both PFS and OS. Furthermore, we found that changes in CTC status between baseline and other time points were prognostic, that is, failure to clear CTCs (being consistently CTC positive) or a switch from CTC negative to CTC positive were associated with poor outcomes. These findings are consistent with results from previous studies (6–13).

In CALGB 40503, serial CTC information was not used to guide therapy. A clinical study in the first-line setting for treatment of HR⁺ MBC did show that changing to chemotherapy in patients with persistent increase in CTCs led to improvements in patient outcomes (15). In contrast, a previous study in unselected patients with MBC (all comers) failed to demonstrate that early change of chemotherapy regimen can improve survival in patients who failed to clear CTCs during first follow-up.

We examined the predictive value of CTCs by evaluating differences in survival of patients according to CTC status (i.e., CTC-positive or -negative at baseline) and treatment received. Comparison of the median PFS and OS revealed longer PFS and OS among CTC-positive patients who received bevacizumab compared with CTC-positive patients who only received letrozole. Furthermore, our exploratory analysis revealed significantly longer mean PFS (at all time points) and OS (at 18 and 20 months) in CTC-positive patients in the letrozole + bevacizumab versus the letrozole-only arm. In contrast, there was no significant difference in mean OS in CTC-negative patients between the two arms.

A large study by Cristofanilli and colleagues showed that patients with MBC with ≥5 CTCs per 7.5 mL blood (stage IV_aggressive) have significantly worse OS compared with those with <5 (stage IV_indolent; ref. 8). In our study, baseline CTCs in postmenopausal women with HR-positive MBC were highly prognostic not only for OS but also PFS. Taken together, these studies confirm CTCs as a strong negative prognostic factor in MBC regardless of breast cancer subtype.

Interesting but counterintuitive observations have been made regarding the clinical impact of CTC levels during treatment with bevacizumab (22–24). For example, Bidard and colleagues found that only baseline CTCs, but not on-treatment levels, were associated with progression in patients with MBC treated with bevacizumab and chemotherapy (24). More interestingly, Gazzaniga and colleagues noted that more than half of metastatic colorectal cancer patients who progressed on bevacizumab had undetectable CTCs (23). The investigators speculated that treatment with bevacizumab facilitated epithelial-to-mesenchymal transition in CTCs, and thus became undetectable by CellSearch, an epithelial-based assay (23), and because bevacizumab affects vessel endothelium (25), impairment of CTC intravasation was proposed to explain why CTC counts considerably decreased during treatment (24) even in patients who do not respond to bevacizumab (23).

In contrast to the observations made in previous studies (22–24), our study showed significant differences in PFS and OS of patients based on baseline CTCs and at first follow-up. Interestingly, patients who were CTC positive at baseline regardless of CTC status at T1 had comparable OS with those who were consistently CTC negative especially in the first 1.5 years of follow-up (Fig. 3F). In addition, these patients (CTC⁺CTC⁺ and CTC⁺CTC⁻) had significantly longer OS compared with those who were initially CTC negative and became positive (CTC⁻CTC⁺). The latter group was, however, very small (n = 5). Collectively, these observations suggest that serial analysis of CTCs may help identify groups of patients who could potentially benefit most from bevacizumab treatment.

Clinical studies have also examined circulating endothelial cells (CECs) as potential biomarkers of bevacizumab efficacy. Recent report by Vasseur and colleagues showed that high levels of CECs at baseline, but not during treatment, were associated with reduced PFS in patients with HER2-negative MBC treated with chemotherapy and bevacizumab (26). Contradictory results from previous studies from the same group (24, 27) as well as others (28), particularly on the direction of prognostic significance of CECs, have highlighted the need to further examine the clinical impact of these cells.

Our findings demonstrate the potential application of CTCs for patient stratification in clinical studies that investigate benefit from bevacizumab treatment, and potentially the benefit of adding other targeted agents to endocrine therapy in the metastatic setting. Understanding CTCs and treatment interactions could eventually help guide patient randomization and risk stratification to facilitate accurate testing of efficacy of novel agents for treatment of MBC. For example, inclusion of CTCs as a stratification factor can facilitate the enrichment of patients with specific risks and could in turn better identify patients most likely to benefit or not benefit from added therapy.

Our results suggest that elevated levels of CTCs at baseline may be predictive of benefit from bevacizumab treatment. Highly vascularized tumors may have the potential to better respond to bevacizumab (29). Moreover, it is hypothesized that extensive vascularization may promote the shedding of CTCs into the blood (30). We therefore speculate that increased CTC levels in the blood due to high vascularity may explain the associated benefit of bevacizumab in CTC-positive patients.

A limitation of the study was the modest sample size particularly in the CTC-positive subset (92/294 = 31%), and thus, validation in a larger cohort is warranted.

In summary, our findings demonstrate that CTCs are robust prognostic markers in postmenopausal women with HR⁺ MBC treated with letrozole or letrozole + bevacizumab in the first-line setting. Our results also suggest a potential OS benefit from adding bevacizumab to letrozole in patients with poor prognosis MBC as defined by CTC positivity at baseline. If confirmed, CTCs may be useful as predictive markers for treatment benefit from bevacizumab and may aid in patient selection for future clinical trials investigating the efficacy of bevacizumab and, importantly, other targeted agents in HR⁺ MBC.

K.V. Ballman reports grants from NCI during the conduct of the study and personal fees from Takeda (DSMB member), Johnson and Johnson (expert witness), Janssen Oncology (expert witness), and Eli Lilly (expert witness) outside the submitted work. J.W. Park reports personal fees from Genentech (honoraria) outside the submitted work. A. Partridge reports other from Novartis (travel support x 1 2019) outside the submitted work. L. Carey reports other from Roche (research funding for clinical trials and time spent as advisor) outside the submitted work. H.S. Rugo reports grants from Pfizer (funding for sponsored studies to UCSF), Novartis (funding for sponsored studies to UCSF), Lilly (funding for sponsored studies to UCSF), Genentech/Roche (funding for sponsored studies to UCSF), Macrogenics (funding for sponsored studies to UCSF), OBI (funding for sponsored studies to UCSF), Merck (funding for sponsored studies to UCSF), Eisai (funding for sponsored studies to UCSF), Immunomedics (funding for sponsored studies to UCSF), Daiichi (funding for sponsored studies to UCSF), Seattle Genetics (funding for sponsored studies to UCSF), Odonate (funding for sponsored studies to UCSF); nonfinancial support from Daiichi (travel support for educational meeting), Mylan (travel support for educational meeting), Pfizer (travel support for educational meeting), Merck (travel support for educational meeting), AstraZeneca (travel support for educational meeting), Novartis (travel support for educational meeting), Macrogenics (travel support for educational meeting), Novartis (travel support for educational meeting), and other from Samsung (limited consulting) and Puma (limited consulting) outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

M.J.M. Magbanua: Conceptualization, data curation, software, formal analysis, supervision, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing. O. Savenkov: Data curation, software, formal analysis, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. E.J. Asmus: Data curation, software, formal analysis, investigation, visualization, methodology. K.V. Ballman: Formal analysis, visualization, writing-original draft, writing-review and editing. J.H. Scott: Formal analysis, investigation, methodology. J.W. Park: Conceptualization, resources, supervision, funding acquisition, writing-review and editing. M. Dickler: Conceptualization, resources, funding acquisition, investigation. A. Partridge: Resources, funding acquisition, investigation, writing-review and editing. L.A. Carey: Resources, funding acquisition, investigation, writing-review and editing. E.P. Winer: Resources, funding acquisition, investigation, writing-review and editing. H.S. Rugo: Conceptualization, resources, supervision, funding acquisition, investigation, writing-review and editing.

Research reported in this publication was supported by the NCI of the NIH under award numbers U10CA180821 and U10CA180882 (to the Alliance for Clinical Trials in Oncology), U24CA196171, UG1CA233180, UG1CA233373, U10CA180820 (ECOG-ACRIN), and UG1CA233329, U10CA180888 (SWOG). Also supported in part by funds from Genentech. The authors thank the patients for participating in this trial. We wish to thank Bryan P. Schneider, M.D. (U10CA180820) and Debasish Tripathy, M.D. (U10CA180888) for championing the study in ECOG-ACRIN and SWOG National Clinical Trials Network, respectively. We also thank the following groups who provided support to Alliance for Clinical Trials in Oncology and Alliance Foundation Trials programs as listed in https://acknowledgments.alliancefound.org.

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.