Study design
Eligible patients were defined as women with metastatic breast cancer who agreed to participate in the multi-institutional CAMELLIA study (registered on ClinicalTrials.gov, Identifier NCT01917279), which was a prospective, randomized, open-labeled phase III study to explore the efficacy and safety of metronomic chemotherapy with capecitabine versus intermittent capecitabine as the maintenance therapy following first-line chemotherapy with capecitabine plus docetaxel in women with HER2-negative metastatic breast cancer at 32 clinical centers in China.
Eligible patients received capecitabine (1000 mg/m2 twice daily on days 1–14, every 3 weeks) plus docetaxel (75 mg/m2 on day 1, every 3 weeks) for a maximum of 6 cycles or until disease progression, intolerable adverse events, or patient withdrawal occurred. Patients with stable disease or a partial or complete response after initial chemotherapy were randomized to receive the maintenance chemotherapy with capecitabine of either conventional or metronomic dosage [20, 21]. A prospective translational study with longitudinal CTC analyses every 6 weeks during the first-line chemotherapy between November 2013 and July 2017 was designed to determine the prognostic value of dynamic CTC phenotype detection based on EMT marker composition. The study was approved by the local ethical committee, and informed consent was signed by every patient before entering the clinical trial.
The clinicopathological characteristics of the patients with detectable CTCs in the peripheral blood samples, including age, position and number of metastatic sites, HR status, and previous endocrinotherapy after confirmed tumor relapse were collected. Tumor response was assessed in accordance with the Response Evaluation Criteria in Solid Tumors (RECIST) guidelines version 1.1. Patients were required to undergo computed tomography, or magnetic resonance imaging if indicated, before first-line chemotherapy and after every two cycles of chemotherapy to evaluate disease and survival statuses. Disease-free survival (DFS) was defined as the duration between the date of breast cancer diagnosis and the date of clinical relapse confirmed by imaging. Progression-free survival (PFS) was defined as the duration between the date of enrolment and the date of clinically observed disease progression (according to RECIST criteria v1.1). Meanwhile, blood samples were collected synchronously until disease progression for the detection of serum tumor markers carcinoembryonic antigen (CEA) and carbohydrate antigen 153 (CA153) to determine CTC phenotype.
Isolation, classification, and enumeration of CTCs
As has recently been described in details [18], CTCs were isolated, classified, and counted using the CanPatrol CTC filtration system, which includes a filtration tube (SurExam, Guangzhou, Guangdong, China) containing a calibrated membrane with 8-μm diameter pores (SurExam), a manifold vacuum plate with valve settings (Millipore, Billerica, MA, USA), an E-Z 96 vacuum manifold (Omega, Norcross, GA, USA), and a vacuum pump (Auto Science, Tianjin, China). Prior to filtration, blood samples were treated with red blood cell lysis buffer (154 mmol/L NH4Cl, 10 mmol/L KHCO3, and 1 mmol/L ethylenediaminetetraacetic acid [EDTA] in deionized water; Sigma, St. Louis, MO, USA) to remove erythrocytes. Phosphate buffer saline (PBS) with 4% formaldehyde (Sigma) was subsequently applied to resuspend the remaining cells. The cell suspension was transferred to a filtration tube and pumped at 0.08 or more MPa to collect isolated CTCs on the membrane.
A multiplex RNA-in situ hybridization (RNA-ISH) assay based on branched DNA (bDNA) signal amplification was applied to classify and count CTCs. Four epithelial biomarkers (EpCAM and CK8/18/19), two mesenchymal biomarkers (Vimentin and Twist), and a leukocyte biomarker (CD45) were used to capture and characterize CTC subpopulations. A detailed RNA-ISH assay was performed as previously described [18]. Three types of fluorescently labeled probes were added and incubated with cells. The sequences of the capture probes and bDNA signal amplification probes had been previously published and were synthesized by Invitrogen (Shanghai, China) [18]. The cell nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI, Sigma), and the cells were analyzed with an automatic Axio Imager Z2 fluorescence microscope (Zeiss, Carl Zeiss Meditec AG, Germany). The red and green fluorescence signals represent the expression of epithelial and mesenchymal biomarkers, respectively. The white fluorescent signals represent the expression of CD45.
Statistical analyses
The clinicopathological characteristics of the recruited patients at baseline were described in percentages of categorical variables. The associations between the distribution of CTC subpopulations at baseline and clinicopathological characteristics (HR status, DFS, and previous endocrinotherapy after confirmed tumor relapse) of patients were analyzed using Mann–Whitney U test. In addition, χ2 test or Fisher’s exact test were used to compare the variations in CTC phenotype and disease status.
Receiver operating characteristic (ROC) curves were constructed to evaluate the performance of CTC phenotypes for predicting 1-year PFS rate. The area under each ROC curve (AUC) was calculated to assess the discriminating power, and Youden index (sensitivity + specificity − 1) was calculated to select the optimal cut-off values for CTC distribution. The Kaplan–Meier PFS curves were plotted to verify the cut-off criteria for prognosis evaluation based on EMT marker composition. Multivariate hazard ratios for PFS were estimated with Cox proportional hazards regression analysis.
The statistical analyses were conducted using SPSS software, version 22.0 (SPSS Inc., Chicago, IL, USA), and P values less than 0.05 were considered statistically significant. A nomogram was constructed to graphically visualize our predictive model according to the results of multivariable Cox regression analysis using the rms package in R version 3.4.1 (http://www.r-project.org/) [22]. The discriminative ability and accuracy of the nomogram were measured using concordance index (C-index) [23].