Journal Logo


Circulating Tumor Cells Dynamics in Pancreatic Adenocarcinoma Correlate With Disease Status

Results of the Prospective CLUSTER Study

Gemenetzis, Georgios MD∗,†; Groot, Vincent P. MD∗,†,‡; Yu, Jun MD, PhD∗,†; Ding, Ding MD, MS∗,†; Teinor, Jonathan A. BS∗,†; Javed, Ammar A. MD∗,†; Wood, Laura D. MD, PhD†,§; Burkhart, Richard A. MD∗,†; Cameron, John L. MD; Makary, Martin A. MD, MPH; Weiss, Matthew J. MD; He, Jin MD, PhD∗,†; Wolfgang, Christopher L. MD, PhD∗,†

Author Information
doi: 10.1097/SLA.0000000000002925


Pancreatic ductal adenocarcinoma (PDAC) is historically characterized by increased mortality rates.1 It is currently predicted to become the second leading cause of cancer-related deaths in the United States by 2030.2 This unfavorable prognosis is mainly associated with the advanced stage of disease at the time of diagnosis, due to a combination of late presentation and early metastasis.3 The absence of biomarkers with high diagnostic sensitivity and specificity in pancreatic cancer makes early identification and monitoring of disease progression challenging.4 However, the in-depth genetic characterization of pancreatic tumors5 and the introduction of possible molecular biomarkers for systemic disease assessment6 introduce a new potential in identification and treatment of PDAC.

The implication of circulating tumor cells (CTCs) in the metastatic mechanism has been proposed since the introduction of the “seed and soil” hypothesis by Stephen Paget in breast cancer.7 The hypothesis suggests that primary tumor cells invade adjacent vasculature, disseminate through circulation to colonize distant sites, and give rise to secondary metastases.8,9 A basic mechanism that allows CTCs to endure the dissemination process is epithelial to mesenchymal transition (EMT).10 EMT plays a major role in embryonic tissue differentiation and pathologic response to tissue injury.11 In cancer, epithelial cells respond to EMT-inducing signals from the tumor microenvironment and acquire mesenchymal characteristics and, subsequently, tumor-initiating potential.12 Involvement of CTCs in the metastatic process has been identified in the majority of solid tumors.13–16 In addition, recent studies in PDAC have reported on the potential of CTCs as a molecular biomarker of disease status by identifying an association of their number and phenotype with early disease recurrence, overall survival (OS), and presence of occult metastases.17–19

CTCs dynamics, or fluctuations of CTCs burden in circulation, is a promising longitudinal indicator of disease status. In metastatic breast cancer, CTCs dynamics have been identified as a potential biomarker of treatment response and disease progression.20–22 Similar data regarding utilization of CTCs dynamics in evaluating treatment effect in PDAC are limited. We have designed and conducted a prospective longitudinal cohort study on CTCs dynamics (NCT02974764), in which 200 enrolled patients with PDAC were followed with sequential phlebotomy along the course of their disease. The primary study endpoint was to identify and characterize the association between CTCs dynamics and disease status. Secondary study goals included the assessment of the predictive capacity of CTCs regarding early recurrence and disease-specific mortality, and the identification of CTCs dynamics potential as a biomarker of disease recurrence.


Patient Cohort

All participating PDAC patients in the CLUSTER study (CircuLating tUmor cellS in pancreaTic cancER) were enrolled consecutively between March 2016 and March 2018. As most patients with resectable PDAC undergo surgery based on clinical presentation, a biopsy-proven malignancy was not mandatory before enrollment. Patients with a personal history of concomitant nonpancreatic malignancy, prior pancreatectomy for PDAC, or age < 18 years were excluded. All enrolled patients signed a detailed informed consent and were followed from the time of diagnosis or presentation to the clinic until patient death. The Institutional Review Board has approved this prospective study.

Data Collection

The study design is available in Supplementary Figure 1, Peripheral blood sample collection was performed in prespecified time points, during each phase of patient care. Patients scheduled for surgery had their first blood sample drawn before incision and the next one between postoperative days 4 and 6. Blood draws would repeat in 2 or 3-month intervals, often in conjunction with scheduled treatment plans, until the patient's death, loss of follow-up, or consent revocation.

Patient demographic data were collected from the prospectively maintained institutional database in combination with clinical notes. A detailed documentation of clinical presentation at diagnosis and patient performance status according to the American Society of Anesthesiologists classification was performed, together with assessment of preoperative invasive procedures and imaging modalities. When patients received neoadjuvant therapy, information about the type of regimen and duration of therapy were collected. All surgical resections were performed by the same team of pancreatic surgeons. Perioperative data, including type and duration of surgery, length of stay, 30-day morbidity (per Clavien-Dindo classification23), and 90-day mortality were noted. Pathology reports were utilized to assess tumor stage and grade, resection margin status, perineural and lymphovascular invasion, and regional nodal status. Presence of malignant cells within 1 mm (≤1 mm) from the surgical margin was characterized as R1 resection. In the postoperative setting, all patients who underwent resection were followed at the outpatient surgical clinic every 3 months for the first 6 months after surgery, and every 6 months after that. Adjuvant chemotherapy treatment was decided on the basis of medical oncologist recommendations and patient preference. Postoperative imaging for recurrence monitoring included abdominal/pelvic and thoracic multidetector computed tomography (MDCT) every 3 to 6 months for the first 2 years and yearly thereafter, or earlier when indicated. Imaging evidence of metastatic or local disease progression were accounted as disease recurrence. Patient date of death was retrieved from institutional medical records, online obituaries, or the Social Security Death Index. All data were collected, de-identified, and analyzed in prospective fashion by an independent reviewer. In this observational prospective study, all clinical decisions were independent of CTCs analysis results.

CTCs Isolation and Characterization

For CTCs isolation and enrichment, we collected 10 mL of peripheral blood in ethylenediaminetetraacetic acid (EDTA) vacutainers and processed it with the Isolation by Size of Epithelial Tumor Cells assay (ISET; Rarecells)24 within 4 hours. The ISET assay allows entrapment of cells of interest with size >8 μm. Control cell-spiking experiments in healthy individual blood samples using commercially available pancreatic cancer cell-lines were also performed to assess the ISET cell isolation capacity. In addition, blood samples from 5 healthy volunteers were filtered and stained as controls, to assess the presence (or lack thereof) of circulating cells with epithelial phenotype. Finally, all isolated cells were fixed on the membrane and preserved in -20°C for further downstream applications (Supplementary Methods,

CTCs enumeration and phenotype characterization were performed with an optimized immunofluorescent staining protocol (Supplementary Methods, a combination of pan-cytokeratin and vimentin antibodies were utilized to assess epithelial and mesenchymal cell traits, respectively. In addition, in order to ensure a high degree of specificity in CTCs assessment, cells were also stained with an antibody “cocktail” (anti-CD45, CD11b, CD14, CD34) to separate different white blood cells (WBCs) populations from CTCs. CTCs were stratified as epithelial-type (pan-cytokeratin+, vimentin-, CD-), mesenchymal type (vimentin+, pan-cytokeratin-, CD-), and epithelial/mesenchymal-type (pan-cytokeratin+, vimentin+, CD-), as previously described.17 Detected cells were additionally assessed for morphology characteristics (nuclei and cytoplasmic structures). Restrictions in cell size were not applied. CTCs enumeration and characterization were performed by a single reviewer, who was blinded to patient clinical data and sample collection time points. Reported results are in CTCs per mL of blood.

Definitions and Statistical Analysis

Continuous variables were summarized with means or medians and standard deviations or ranges, respectively, and compared with unpaired t test or Wilcoxon signed-rank tests as appropriate. Categorical variables were summarized using proportions and counts and compared with Chi-square or Fisher exact test. Recurrence-free survival (RFS) was defined as the time interval between date of operation and either date of recurrence or death, which came first, or censored at last follow-up. Early recurrence was defined as recurrence within 1 year in patients who had at least 1-year follow-up after operation.25 OS and post-resection survival were defined as the time from date of diagnosis (OS) or resection (post-resection survival) to either death or censored at last follow-up. Optimal cut-off of CTCs for early recurrence prediction were calculated with Youden Index, which was also used to choose optimal cut-off values for alterations in CTCs dynamics. Multivariable logistic regression was used to estimate the odds ratio (OR) of CTCs for early recurrence, while adjusting for other covariates. Missing data less than 20% were imputed using multiple imputation (5 permutations). Kaplan-Meier curves were used to estimate median survival with corresponding 95% confidence intervals (95% CIs). The log-rank test was utilized for subgroup comparison. A P value < 0.05 was considered statistically significant. Statistical analysis was performed with R software (V.3.4.1).


Demographic and Perioperative Data

In total, 200 consecutive patients were enrolled prospectively in the CLUSTER study. The patient selection flowchart and exclusion criteria are available in Fig. 1. The final cohort of 136 resected patients was stratified into 2 groups: chemo-naive patients eligible for upfront surgical resection (n = 79, 58%) and patients who had undergone neoadjuvant therapy preoperatively (n = 57, 42%). The included patients (Table 1) had a median age of 68 years [interquartile range (IQR) 59 to 74] and were predominantly males (n = 71, 52%). Most tumors were located in the right pancreas (head and/or uncinate process, n = 102, 75%), and the median tumor size was 30 mm (IQR 22 to 42). Margin-negative resection (R0) was achieved in 85% of patients.

Patient selection flowchart.
Demographics, Perioperative Outcomes, and Clinicopathological Data of Resected Patients

Circulating Tumor Cells Characteristics

As previously described,17 we predominantly identified 2 distinct CTCs populations in the peripheral blood of patients with PDAC: cells that express an epithelial phenotype (eCTCs, pan-cytokeratin+, vimentin-, CD-) and cells that express a combined epithelial/mesenchymal phenotype (mCTCs, pan-cytokeratin+, vimentin+, CD-; Supplementary Figure 2, A third population of cells with purely mesenchymal phenotype (pan-cytokeratin-, vimentin+, CD-) was not recognized in any of the patient samples. Morphologically, detected CTCs had a median diameter of 13 μm (IQR 10 to 15), without significant size differences between the 2 phenotypes. This size was comparable to neighboring WBCs (pan-cytokeratin-, vimentin-, CD+), which measured between 10 and 18 μm. However, distinct differences were noted in the nuclear structure: CTCs had a characteristically large homogeneous nucleus covering >80% of the cell surface in 2-dimensional imaging, and WBCs were characterized by a large segmented and/or lobulated nucleus. CTCs were most often recognized as single cells, and in very few samples (n = 4, 1%), small clusters of 3 to 5 cells were identified (Supplementary Figure 3, Genetic sequencing in a small subset of patients confirmed that identified CTCs harbored mutations in 3 driver PDAC genes (KRAS, TP53, and SMAD4). Circulating cells with epithelial and/or mesenchymal phenotype were not identified in any of the healthy volunteer samples (Supplementary Figure 4,

Treatment Modalities and Alteration of CTCs Dynamics

A total number of 516 peripheral blood samples were collected from the 136 eligible patients (median n = 4, IQR 2 to 5). CTCs were detected in the pre-operative samples of 131 patients (96%), regardless of their pre-operative treatment status. Chemo-naive patients who were eligible for upfront resection had a median number of 11 total CTCs/mL of blood (tCTCs, IQR 6 to 15), breaking down to 9 eCTCs/mL of blood (IQR 5 to 12) and 2 mCTCs/mL of blood (IQR 1 to 3) in the preoperative samples (Table 2). Patients who had previously undergone neoadjuvant therapy were found to have significantly lower median CTCs numbers across all identified populations: 7 tCTCs/mL of blood (IQR 3 to 10, P = 0.007, Fig. 2A), 5 eCTCs/mL of blood (IQR 5 to 9, P = 0.007, Supplementary Figure 5A,, and 1 mCTC/mL of blood (IQR 0 to 2, P = 0.034, Supplementary Figure 5B, in the preoperative samples. Statistical correlations between the number of CTCs (tCTCs, eCTCs, and mCTCs) and tumor stage, size, location, and differentiation grade were not identified (P value varied between 0.19 and 0.68).

CTCs Numbers of Included Patients
Boxplots of comparison between tCTCs of (A) preoperative blood samples in chemo-naive patients and patients who underwent neoadjuvant therapy, (B) preoperative and postoperative samples in chemo-naive patients, and (C) preoperative and postoperative samples in patients who underwent neoadjuvant therapy. tCTCs indicates total circulating tumor cells/mL of blood.

In chemo-naive patients, surgical resection of the primary tumor resulted in significant decrease of CTCs (Table 2), with postoperative median numbers of 2 tCTCs/mL of blood (IQR 1 to 4, P < 0.001, Fig. 2B), 2 eCTCs/mL of blood (IQR 1 to 3, P < 0.001, Supplementary Figure 6A,, and 0 mCTCs/mL of blood (IQR 0 to 1, P < 0.001, Supplementary Figure 6B, A similar significant CTCs population reduction was observed in neoadjuvant patients measuring postoperatively at 2 tCTCs/mL of blood (IQR 1 to 4, P < 0.001, Fig. 2C), 2 eCTCs/mL of blood (IQR 1 to 3, P < 0.001, Supplementary Figure 7A,, and 0 mCTC/mL of blood (IQR 0 to 1, P = 0.023, Supplementary Figure 7B,

In addition, patients who were explored and found to have occult abdominal metastatic disease (OMD) intraoperatively had significantly higher CTCs numbers than patients who underwent surgical resection (Table 2): 20 tCTCs/mL of blood (IQR 12 to 24, P < 0.001, Supplementary Figure 8A,, 15 eCTCs/mL of blood (IQR 9 to 18, P < 0.001, Supplementary Figure 8B,, and 3 mCTC/mL of blood (IQR 2 to 6, P < 0.001, Supplementary Figure 8C, In patients with OMD and patients whose tumor was not resected due to local extension and major vessel involvement, no significant difference between pre- and post-exploration CTCs was identified (tCTCs P = 0.16, eCTCs P = 0.09, mCTCs P = 0.10).

CTCs as a Biomarker for 1-year Disease Recurrence and Mortality

Fifty-nine patients (44%) presented with disease recurrence within 1 year from surgery. Multiple-site metastases was the most common pattern (n = 23, 39%), followed by local recurrence (n = 15, 25%) and liver metastases (n = 11, 19%). Preoperative CTCs were significantly higher in the recurrence cohort (tCTCs P < 0.001, eCTCs P = 0.002, mCTCs P = 0.005; Table 3). The difference was also significant in the postoperative CTCs populations. In order to clearly identify the proportion of epithelial/mesenchymal phenotype cells in the samples, we calculated an mCTCs-to-eCTCs ratio : it was significantly higher in the preoperative samples of patients who presented with early recurrence (median 0.23, IQR 0.14 to 0.33) than in patients who did not recur within a year (median 0.10, IQR 0 to 0.22, P = 0.008).

Demographics and Clinicopathological Data Comparison Between Patients Who Presented With Recurrence Within 12 months From Surgery and Patients Who Did Not

We further sought to identify preoperative factors, including demographic and clinical parameters combined with CTCs dynamics, which will allow identification of aggressive disease biology and increased probability of early recurrence. To increase the accuracy of the predictive model, we performed a distinct univariable and multivariable logistic regression analysis for chemo-naive and neoadjuvant patients (Table 4). In addition, the multivariable analysis was performed separately for each of 2 mutually exclusive CTCs populations (eCTCs and mCTCs) and the total number of CTCs (tCTCs). On univariate analysis for the chemo-naive cohort, CA19-9, which is the primary clinical biomarker for PDAC, was not predictive of 1-year recurrence at an optimal cut-off value of 178 U/mL [OR 1.18 (95% CI 0.86–1.62), P = 0.30]. On the contrary, all 3 different CTCs populations were independently associated with an increased risk of early-recurrence in their separate multivariable models: mCTCs had the highest predictive ability with an area under the curve (AUC) of 0.69, 52% sensitivity, and 82% specificity (Supplementary Figure 9, Similarly, in patients who underwent neoadjuvant therapy, all 3 CTCs populations were the only independent parameters associated with early recurrence: in this cohort, the tCTCs predictive model had the highest accuracy with an AUC of 0.75, 72% sensitivity, and 75% specificity.

Univariable and Multivariable Logistic Regression Analysis for Associations Between Preoperative Risk Factors and Early Recurrence of Pancreatic Ductal Adenocarcinoma Within 12 Months from Resection

In the studied cohort, mortality within 1 year from surgery was 31% (n = 51). In chemo-naive patients, preoperative or post-operative eCTCs and mCTCs numbers were not associated with 1-year disease-specific mortality (P = 0.65, P = 0.89 and P = 0.33, P = 0.43, respectively). However, in the neoadjuvant cohort, the predictive role of CTCs was present: postoperative numbers (POD 4 to 7) of all 3 populations (tCTCs ≥4/mL, eCTCs ≥3/mL, and mCTCs ≥1/mL of blood) were associated with early death from PDAC (P = 0.006, P = 0.009, and P = 0.018, respectively - Supplementary Figure 10,

CTCs Longitudinal Dynamics and Prediction of Disease Recurrence

CTCs dynamics were assessed on the basis of changes in CTCs numbers and phenotype distribution in combination with clinical reports and imaging data, along the course of patient disease. All patients with at least 3 sampling time points were included in the developed prediction model: a baseline preoperative sample, a postoperative inpatient sample or before initiation of adjuvant therapy, and a sample acquired 3 to 6 months after surgery. Therefore, we retrospectively reviewed sequential CTCs dynamics in 114 patients, including 44 who had undergone neoadjuvant chemotherapy. In each time point, CTCs populations were compared with the baseline sample for identification of alterations.

Alterations in CTCs dynamics, that preceded imaging evidence, were documented in patients with disease recurrence (Fig. 3A). This pattern was consistent across the majority of those patients who were longitudinally followed (n = 45, 76%). In the developed prediction model, chemo-naive and neoadjuvant patients were split into separate groups and an increase cut-off value of 4 and 5 tCTCs/mL of blood was identified, respectively, for prediction of recurrence within 2 months. The generated AUC was 0.871 for chemo-naive patients (82% sensitivity, 85% specificity), and 0.684 in neoadjuvant patients (67% sensitivity, 83% specificity). On the contrary, an increase in CTCs populations was not observed in patients who did not present with disease recurrence (Fig. 3B). In addition, regardless of preoperative treatment status, patients who recurred had a disproportionate increase in mCTCs. More specifically, the median preoperative ratio was 0.23 (IQR 0.14 to 0.33), whereas at the time of recurrence, the median ratio value was 0.42 (IQR 0.29 to 0.57, P < 0.001). However, the ratio was not specific enough to distinguish patients who presented with local versus distant metastases (P = 0.303).

A, Alterations in CTCs dynamics in a 74-year old chemo-naive patient with PDAC. The patient underwent a pancreaticoduodenectomy on Day 0 (OR) and a significant reduction in all CTCs populations was documented. A further decrease during adjuvant chemotherapy (Gemcitabine/capecitabine, green frame) is also present. An initial increase in CTCs is observed on Day 254. However, our predictive model recognizes Day 388 as the time point of significant CTCs increase (red dotted vertical line). A CT-scan on Day 388 showed no signs of disease recurrence. The patient was diagnosed with local recurrence and pulmonary metastases in a follow-up CT on Day 472 (red solid vertical line). A significant increase in the mCTCs population fraction is also noted: in the preoperative sample, the mCTCs/eCTCs ratio was 0.17 and at recurrence 0.72. B, CTCs dynamics in a 71-year old male with PDAC, who received neoadjuvant therapy (modified FOLIRINOX). An initial decrease was noted with surgical resection (distal pancreatectomy on Day 0) and CTCs numbers remained low along the patient's follow-up. No imaging evidence of disease recurrence exists by Day 406. CHT indicates chemotherapy; CT, computed tomography; eCTCs, epithelial phenotype circulating tumor cells/mL of blood; mCTCs, epithelial/mesenchymal phenotype circulating tumor cells/mL of blood; mFFX, modified FOLFIRINOX; OR, operating room; tCTCs, total circulating tumor cells/mL of blood.


The utilization of CTCs as a “liquid biopsy” to assess the tumor burden and guide clinical decisions has developed into an emerging field of study.26 In pancreatic cancer patients, CTCs have been previously introduced as possible biomarkers for recurrence and survival.18,27 However, biological hurdles, such as extensive cell heterogeneity and physical challenges, associated with cell kinetics within circulation limit that potential.28,29 In this prospective study, we attempted to characterize the physical attributes and phenotypes of CTCs in PDAC patients, longitudinally assess their dynamics along the course of the disease, and develop a comprehensive model for their potential clinical significance. The two main pillars of our study were a 2-step isolation technique (size-based positive selection and immunofluorescent biomarker staining) and the stratification of identified cells into an epithelial and epithelial/mesenchymal phenotype.

In the studied cohort, more than 95% of resectable PDAC patients had at least 1 identified CTC/mL of peripheral blood, whereas the median number of total CTCs across all preoperative samples was 11/mL of blood. Experimental models have calculated the cell dissemination rate from a tumor to vary between 3 and 4 × 106 per g/day,30 yet most are passively shed and less than 0.01% have metastatic potential.27 On the basis of our observations, the median number of CTCs in the peripheral circulation of a patient with resectable PDAC can be estimated at approximately 50,000 at any given time. In addition, most of the CTCs were isolated as single cells with a median size comparable to that of WBCs. Previous studies have also identified the presence of CTCs clusters or microemboli in patients with PDAC.18 We observed small clusters of <5 cells in 1% of the peripheral blood samples. In PDAC, disseminated cells travel through the portal vein into the liver, and subsequently enter the right side and later the peripheral circulation. Even though CTCs clusters are able to pass through capillary-sized vessels,31 it is unlikely that they make it to the periphery, due to extreme hydrodynamic flow forces and shear stress.32 The occurrence of EMT and the loss of epithelial marker expression, such as E-cadherin, which sustains cell adhesions may have an additional negative effect. However, it is possible that large cohesive cell cohorts disseminating from PDAC become trapped in the hepatic lobule network and may eventually give rise to liver metastases, according to the mechanism of collective migration.33 Identification of CTCs clusters in the portal vein can support this hypothesis and provide an explanation for the increased incidence and aggressiveness of liver metastatic disease in PDAC.34,35

Most identified CTCs expressed exclusively an epithelial phenotype and most likely were the result of tumor apoptosis in the circulation. A variable number of CTCs expressing a mixed epithelial/mesenchymal phenotype (mCTCs) was also present in preoperative samples, yet in a smaller percentage (n = 94, 70%). On the contrary, CTCs with purely mesenchymal traits were not identified in any of the patients. These findings concede with the recently introduced “partial EMT” concept, where CTCs acquire mesenchymal traits, while retaining specific epithelial characteristics.13,36 This mixed phenotype allows a cellular plasticity that appears to be critical in the establishment of metastatic colonies.37,38 An ongoing study for genetic characterization of CTCs from our group will provide further insight on the genetic link between the primary tumor and metastatic lesions. KRAS mutations have been previously identified in PDAC CTCs verifying their malignant origin39,40; however, comprehensive DNA sequencing on a single cell level continues to be a challenging task.41

Alterations in CTCs dynamics were observed along the course of the patients’ disease. Initially, a positive effect of neoadjuvant chemotherapy and surgical resection of the primary tumor on CTCs population was recognized. More specifically, patients who had received neoadjuvant treatment had significantly lower numbers of CTCs across all phenotypes compared with the chemo-naive cohort eligible for upfront surgery (P < 0.001). The beneficial effect of systemic treatment on CTCs dynamics has been previously identified in breast and prostate cancer and is an emerging field in PDAC.16,42 In our study, a correlation between CTCs and tumor pathological characteristics, such as size, site, and differentiation grade, was not identified. We account this observed non-correlation to two main reasons: (1) the relatively small cohort of patients and (2) to the fact that we stratified into chemo-naive and neoadjuvant patients. However, these results appear to strengthen the hypothesis that CTCs may disseminate very early in tumor development.43

Surgical resection of the primary tumor had the largest effect on CTCs dynamics. When the main source of CTCs was removed, their numbers reduced significantly (P < 0.001), but in most patients, they did not decrease to 0 (79%). Currently, all data on CTCs kinetics within the circulation come from experimental models and show a limited half-life of approximately few hours.44 We hypothesized that 4 to 7 days after surgery would be a reasonable time interval to assess the effect of surgical resection on CTCs dynamics. The fact that CTCs did not disappear postoperatively in most patients may suggest that collateral sources function as dormant tumor cells (DTCs) reservoirs very early in the disease.45

The biophysical attributes of CTCs and their dynamics translated into significant biomarker potential. Regardless of the preoperative treatment status, patients who presented with early disease recurrence (within 12 months from surgery) had significantly higher tCTCs and mCTCs in both the preoperative and the postoperative samples. To further assess the significance of mCTCs, we calculated an ratio: patients who recurred early had significantly higher ratio than those who did not (P = 0.008). Moreover, patients who were found to have occult metastases also had significantly higher CTCs and ratio (both P < 0.001), concurring with recently reported results by Court et al.19 Our group has previously identified CA19-9 as a strong predictor for early recurrence in PDAC25; however, in this prospective study, preoperative CA19-9 was not suggestive of early recurrence, most likely because patients who underwent neoadjuvant therapy were included in the comparison. Nonetheless, CTCs proved a reliable biomarker for early recurrence, regardless of the preoperative patient status. These data indicate that aggressive tumor biology may be reflected in CTCs dynamics.

This hypothesis is further supported by the longitudinal changes in CTCs dynamics. We identified a consistent increase in CTCs numbers in patients who presented with recurrence, before imaging evidence of disease. Moreover, the ratio at the time of recurrence was significantly higher than the preoperative sample, suggesting a more aggressive biology, likely associated with treatment resistance and clonal drift.46 In current clinical practice, disease surveillance is performed with repeated imaging and CA19-9 assessment. A limited number of studies have evaluated CA19-9 as a longitudinal follow-up biomarker. CA19-9 appears to have a mediocre sensitivity and specificity and is not useful for patients who do not express the Lewis antigen.4 Our developed CTCs model can identify recurrence with an accuracy of 80%, more than 2 months earlier from imaging evidence of disease. However, specificity was not high enough to predict the site of recurrence (local vs distant).

Scientific interest on the importance of circulating biomarkers has erupted in recent years, primarily regarding the role of circulating tumor DNA (ctDNA). Recent studies have shown the significant potential of ctDNA on early cancer detection and follow-up of disease.6,47 However, ctDNA represents collectively all indolent or aggressive clones shedding in the circulation from CTCs, the primary tumor, or metastases. On the contrary, each CTC represents a unique clone and may provide direct insight on the clonal prevalence in the metastatic mechanism. Data from this study suggest that significant potential for CTCs as a PDAC biomarker is present. However, as cell kinetics and survival are significantly affected within the circulation, the only way to overcome these obstacles is to develop optimal isolation and enrichment assays.48 Longitudinal assessment of ctDNA in PDAC is currently underway in our institution and we will soon be able to make a direct comparison between the 2 biomarkers or combine them for increased sensitivity and specificity.

This prospective observational study has several limitations. First, use of an isolation by size assay for CTCs identification may result in possible loss of smaller cells (<8 μm). In addition, even though we reviewed cellular morphology and utilized multiple antibodies for phenotypic characterization of CTCs, it is possible that false positives could be present, due to heterogeneous expression of epithelial and mesenchymal markers in blood cells or nonspecific staining of dead cells. Moreover, vimentin is a widely used mesenchymal marker for cell characterization, yet, other mesenchymal markers are available and were not utilized in this study. Subsequent genetic analysis was only performed in a very small subset of patients. In addition, as follow-up of patients who were enrolled later in the study is not yet completed, there is a possibility for small changes in the reported results. Finally, most patients were diagnosed in outside institutions and were treated and followed locally after resection, which resulted in limited data for: (1) pre-neoadjuvant treatment CTCs values, (2) longitudinal CA19-9 values, and (3) CTCs dynamics during salvation chemotherapy after disease recurrence. However, the CLUSTER study is the first to provide a detailed longitudinal insight on the systemic component of pancreatic cancer through enumeration and phenotypical characterization of CTCs. The next step is description of these cells with DNA and RNA sequencing, and protein expression analysis. Moreover, the major goal is to achieve ex vivo culture of CTCs for further genetic analysis and pharmacotyping for tumor response. Culture of CTCs has proven very challenging with traditional culture methods, due to the rarity and the complex nature of these cells.49,50 Novel techniques, such as 3-dimensional culture with organoids, may prove useful for profiling patient CTCs in the near future.51


In the CLUSTER study, we came across a detailed macroscopic view of the systemic nature of PDAC and its aggressive metastatic potential by longitudinally observing CTCs dynamics. Systemic therapy was associated with lower CTCs numbers and surgical resection of the primary tumor resulted in significant decrease in CTCs populations. Increased preoperative CTCs numbers and especially cells with mesenchymal traits were predictive of early recurrence after surgical resection. In addition, an increase in CTCs was documented in PDAC recurrence, approximately 2 months before new imaging evidence of disease. The potential of CTCs dynamics as a biomarker is evident and further in-depth analysis of their cellular profile is justified.


Dr Nipun B. Merchant (Miami, FL):

I want to congratulate Dr Wolfgang for presenting another very provocative study advancing the field of pancreas cancer as the group from Johns Hopkins continues to do.

In reviewing this manuscript, perhaps one of the most important findings in their study cohort, unrelated to this study, is that 44% of patients had recurrence of disease within 1 year of surgery, and that, too, with the majority of patients having recurrence at multiple sites, and 31% of patients died within 1 year of surgery. This clearly emphasizes the fact that we still have not gotten a handle on this disease and further underscores the need for biomarkers of early detection as well as therapeutic response.

A rate-limiting step for biomarker development in pancreas cancer is a limited and difficult access to tissue, particularly for longitudinal analysis, highlighting the need for liquid biopsy based biomarker development.

Dr Wolfgang and colleagues have presented a prospective, longitudinal study showing that 2 major CTC subtypes, epithelial and mesenchymal, can be identified in all pancreas cancer patients and that surgical resection of the primary tumor results in significant reduction of CTC burden. Patients who presented with early disease recurrence in less than 12 months had significantly higher preoperative as well as postoperative CTC numbers, and patients who received neoadjuvant chemotherapy had significantly lower CTCs than untreated surgical resectable patients suggesting that these cells have a potential to be used as a biomarker of therapeutic response.

In their multivariable analysis, pre-op numbers of all CTC subpopulations were the only predictors of early recurrence within 12 months of surgery, and a risk assessment score accurately identified the disease recurrence 2 months earlier than detected on imaging.

This is an excellent paper, and I think the way we need to move forward to make progress in pancreas outcomes. However, it also generates several comments and questions. And based on this, I have some logistical, methodological, and clinical relevance questions.

Surgical resection of the primary tumor resulted in significant reduction but not disappearance of CTC burden. Do you think this could be related to the timing of your first specimen collection 4 to 6 days postoperatively?

Do CTCs ever disappear? Is persistence of CTCs over time alone a risk factor of recurrence as versus increasing number?

You did not see a statistical correlation between the number of CTCs and tumor stage, size, location, and differentiation, yet pre- and post-op CTCs and mesenchymal to epithelial CTC ratio were also significantly higher in the early recurrence cohort. Can you comment on this?

In treatment-naive patients, pre- or post-op CTC numbers were not associated with 1-year disease-specific survival even though they decreased between pre- and post-op levels. Yet, in those who received neoadjuvant therapy, the predictive role of CTCs was present. What do you attribute this difference to?

You emphasize CTCs mesenchymal phenotype, those that undergo EMT, thereby having an increased risk of metastatic potential. Both post-op samples and treatment-naive and neoadjuvant-treated patients decreased to no mesenchymal CTCs in many patients, yet before recurrence, the mesenchymal-to-epithelial CTC ratio increased. How does this occur, and where do these cells come from?

I am also intrigued by the fact that there was no difference seen in the mesenchymal-to-epithelial ratio from local versus metastatic recurrence. As mesenchymal CTCs would have a higher likelihood of hematogenous spread, one would think that you will have a higher likelihood of detection in the blood in metastatic patients.

You show that CTCs that you isolate appear to be tumor derived as you have sequenced KRAS, TP53, and SMAD4 from them. What are the advantages of isolating CTCs compared with other emerging biomarkers such as circulating tumor DNA? Also, how can you determine that there are no false-positive cells or artifacts?

Lastly, what are some of the biological hurdles such as extensive cell heterogeneity and physical challenges associated with the kinetics within the circulation that have to be overcome before we can see CTCs as a robust biomarker in pancreas cancer?

I want to congratulate you and your colleagues for continuing to push the field forward in pancreas cancer, and I would like to thank the Association for the privilege of the floor.

Response From Christopher L. Wolfgang:

Thank you, Dr Merchant, for the nice discussion and insightful questions. It is an honor to have you discuss our paper. I will summarize and put some of these questions together.

The first question was regarding the fact that in some cases the CTCs do not fall to zero following surgical resection.

The findings of residual CTCs on post-op day 4 to 6 was not surprising. Initially, we thought that maybe this was the result of residual cells that would clear over time. On the basis of animal models, CTCs have a half life of approximately, at most, 1 to 2 hours, and in some instances measured even lower. So, based on this work, if you had a surgical resection, you would you expect that within post-op day 4 or 7 that you would see CTCs go to 0, and we did not see that. I was very surprised to see that over time, as shown in the manuscript, that a given patient settles out to a new baseline and remains there, even in the absence of recurrence. This baseline may be 0, but in many cases, it is not. This is exciting because I believe we are looking directly at the dormant subclinical disease that can relapse in the future. In the patients who made it to 1 year without recurrence, some have as many as 5 CTCs at baseline. In a few years, we will be able to answer your question of whether or not such patients are at risk of late recurrences as our cohort matures. It brings up another interesting question of where is the reservoir of this subclinical disease? One potential site is the bone marrow and liver, which we are planning on evaluating in a subsequent study.

What I think is interesting about this study is that we are looking at something we have never seen before. I think this is subclinical or dormant disease that we never had an opportunity to evaluate before this work. What will be interesting is what happens in the population that did not recur within 1 year, and the ones that we see the residual CTCs. Potentially, we know that these patients are still at risk of recurring. Will they have a resurgence in those cells?... and will those cells be the ones that cause the metastatic disease? These questions will be something that we will be able to determine as this study progresses.

The next question related to the lack of a statistical correlation with the pathological features and the number of CTCs. In the beginning, I was a bit troubled by this. We know that there are pathological features that predict relatively aggressive behavior, but if you think about it, it is imprecise, and there are factors that we do not know that predict. For example, we have all had node-positive, margin-positive patients who lived a long time and margin-negative patients who died within a year. And what I think that we are seeing here is that the CTCs are a very sensitive window into the tumor biology, and maybe are completely independent of those pathological features, suggesting that this may be a more direct assessment of tumor biology. We will have to see how that plays out over time.

There was a question then about the lack of correlation with survival in the chemo-naive patients, whereas there was a correlation in the neoadjuvant patients, and that there was correlation, however, with both recurrences for both neoadjuvant and adjuvant.

This is a very interesting observation. CTCs correlated with recurrence at 1 year but not survival for chemo-naive patients. The mathematic explanation is that they are still alive, but with disease.

The biological explanation is less clear but much more interesting. One possible explanation, which is highly unsettling, is that upfront chemo subjected the systemic disease to a natural selection process and simply caused clonal selection for more aggressive clones. That does not mean neoadjuvant is worse treatment … we just selected for aggressive clones before surgery as opposed to after surgery. In other words, choose your poison … select for the bad clones before surgery or after surgery. It is a little disheartening. Now this is just a theory, but as time goes on, we will evaluate this question in this cohort. The bright spot in this potential explanation is now that we can directly assess these clones (with CTCs analysis), we may be able to mitigate the clonal selection with new cocktails of existing drugs. Such as one would do for drug-resistant bacteria with treatment based on culture data.

The next question was regarding the fact that there was no difference seen in the mesenchymal to epithelial ratio from local versus metastatic recurrence.

Good question and I do not know the answer. However, based on published work and work we are doing, my bias is that even so-called local recurrence is systemic. That is … it is systemic disease that comes back through circulation or perhaps perineural invasion and grows aggressively in the already primed stroma of the tumor bed.

The next question was basically the differences among the various forms of liquid biopsy and why we chose CTCs.

ctDNA represents an average of everything that is going on throughout the entire body with respect to the cancer, that is, it measures collectively the indolent clones, aggressive clones, and even perhaps is weighted toward what is dead. In contrast, each CTC represents a unique clone, it is not an average. Moreover, it is most likely a direct assessment of what causes metastases, not indirect like ctDNA.

That said, we have a ctDNA arm of this study not part of this manuscript. We will be able to make a direct comparison or perhaps be able to make a composite marker as future studies from this trial.

Finally, the biological hurdles and challenges…in particular tumor heterogeneity.

Yes…the heterogeneity poses logistical (and cost in assessment) problems, but with careful assessment and data management these can be overcome. The heterogeneity of what we see is exciting because it mirrors what we see clinically with the biological behavior. The heterogeneity in CTCs tell me we have the necessary data points to measure the diverse biologies of patients’ tumors. I think we will see that at the end of the day we do not need to eradicate all clones, just the bad one. Perhaps not even eradicate it but develop a vaccine toward that clone that converts the disease into a chronic condition.

The other “physical challenge” is what method of isolation to use. They all have their weakness and strengths. We used ISET because we did not want to bias ourselves by a priori antibody-based selection. However, we compromised by biasing ourselves to CTCs greater than 8 microns in size. I believe this calculated decision may have paid off because I am not sure we would have identified the mCTC otherwise. We have ongoing work using a new flowcell device developed in the applied physics laboratory at Johns Hopkins and we also use flow cytometry at times to test certain aspects of CTCs.

Dr Keith D. Lillemoe (Boston, MA):

Chris, great paper, really advancing this field and probably all of oncology.

Based on your data, it would seem that using this type of data for clinical decision making, must be getting close. In your patient example, could you have started chemotherapy earlier and maybe altered the outcome in terms of survival? Have you started using these data to alter your management of the patients?

Second of all, the one finding that probably affects surgeons most, is the dramatic data, that early recurrence could be predicted by CTCs. Will it ever get to the point where people will be turned down for surgery because of a high level of preoperative CTCs either before or after neoadjuvant therapy?

Response From Christopher L. Wolfgang:

Thanks, Dr Lillemoe. Very insightful. Are we altering therapy in response to the CTCs? Hopefully, one day we will be doing that, but not at this point.

What we envision for the next trial is something pretty simple but in the vein of precision medicine. That is to simply switch from FOLFIRINOX to Gem-based chemo or vice versa upon an increase in CTCs but no clinical evidence or relapse or progression.

Regarding will we use this to exclude patients from surgery, I do not think we should use these results to make that decision. I think it is being able to alter neoadjuvant approaches, targeted approaches, with surgical resection to get the best outcome based on looking at the circulating disease, because right now we do not. We focus on the local disease and we fail systemically. Now, we have a tool to look at the systemic disease and use that in patients who go to surgical resection.

Dr Selwyn Vickers (Birmingham, AL):

Chris, really a nice paper and great presentation. Two quick questions. As you know, there are data being very clear now that organoids may be a really powerful predictor of response. Can you grow organoids from these CTCs in order to predict disease response to therapy?

Second, has there been an inherent value, as you can now isolate these cells versus their surface markers to doing single-cell exome sequencing to determine the further correlation with biology?

Response From Christopher L. Wolfgang:

Great question. The organoid question first. We actually have an organoid effort and group in our laboratory. One of our surgeons went to train with Dave Tuveson for exactly this reason. Right now, we are able to establish organoids from primary tumors, but after 3 years of trying, it has become exceedingly difficult to create tissue culture from CTCs, and it probably means that only a small percentage of circulating cells are capable of originating metastases. If you do a calculation you can estimate that at any one time in an individual, there are probably about 50,000 circulating cells. The probability of drawing 10 circulating cells of blood and getting that cell that can create metastases and then having the right conditions is pretty low. But, ultimately, that would be the perfect goal, so we are attempting to do that.

Probably the thing to focus on are CTC clusters, which I had not talked about here, which are probably the most aggressive clones.

Yes, we are planning on doing molecular analysis on these cells such as RNA seq.

Dr Andrew Lowy (La Jolla, CA):

It strikes me that this mesenchymal population is what others would probably call pancreatic cancer stem cells, which have been characterized by a number of groups. So, I wonder if you have looked at co-expression of markers specifically for stem cell populations that have been shown to be enriched in proteins such as CXCR4, Musashi, and second, whether you would apply single-cell sequencing to your CTCs to try to look for potential novel targets, particularly for these presumably highly drug-resistant cells.

Response From Christopher L. Wolfgang:

Excellent question. In our previous work on these mesenchymal cells, they do have tumor-initiating markers such as aldehyde dehydrogenase, CD133, CD44, and ALDH - so they do overlap with what would be defined as tumor-initiating cells.

Just real quickly, because it was asked before by Dr Vickers on the single cell sequencing, we are doing that, but, as you know, right now, maybe only 10% or 15% of the mutations are targetable. So, it is important to do. We have the ability to do it. But I think there is only a minimal impact in how we change therapy. RNA, we are working on those data now.

Dr Pierre-Alain Clavien (Zurich, Switzerland):

Chris, congratulations for this great work. I have only one short question, which relates to an ethical issue. Through the search for CTC dynamics, you are extracting information about the disease and putative poor natural history in patients who are doing well. Did you inform the patient about the CTC findings and possible poor prognosis? And if so, what was the recommendation given to them?

Response From Christopher L. Wolfgang:

That is a great question. We have actually talked about this at the onset of the study, if we were to find something clinically important, what would we tell patients? We inform patients that all of the data that we generate are available to them, and we can talk to them about it.

As one of the clinicians caring for these patients, I am blinded to the data, and we have a system in the laboratory that the people collecting the clinical data are separated from the CTC analysis, whereas the ones doing the analysis are blinded from the clinical data, so that we actually have an unbiased study.

When people request their CTC results and the clinician will be unblinded and review it with them. We inform them that we do not know what it means right now. It does not alter the clinical management.


The authors would like to acknowledge Lindsay Manos, Lara Espin, Christi Walsh, Caitlin Brown, and Tiffany Zavadsky for their help in patient follow-up, Hao Zhang for his valuable assistance in flow cytometry, and Maritina Iliadi for figures layout and editing.


1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2017. CA Cancer J Clin 2017; 67:7–30.
2. Rahib L, Smith BD, Aizenberg R, et al. Projecting cancer incidence and deaths to 2030: the unexpected burden of thyroid, liver, and pancreas cancers in the United States. Cancer Res 2014; 74:2913–2921.
3. Wolfgang CL, Herman JM, Laheru DA, et al. Recent progress in pancreatic cancer. CA Cancer J Clin 2013; 63:318–348.
4. Daamen LA, Groot VP, Heerkens HD, et al. Systematic review on the role of serum tumor markers in the detection of recurrent pancreatic cancer. HPB (Oxford) 2018; 20:297–304.
5. Bailey P, Chang DK, Nones K, et al. Genomic analyses identify molecular subtypes of pancreatic cancer. Nature 2016; 531:47–52.
6. Cohen JD, Li L, Wang Y, et al. Detection and localization of surgically resectable cancers with a multi-analyte blood test. Science 2018; 359:926–930.
7. Paget S. The distribution of secondary growths in cancer of the breast. Lancet 1889; 133:571–573.
8. Hanahan D, Weinberg RA. The hallmarks of cancer. Cell 2000; 100:57–70.
9. Chaffer CL, Weinberg RA. A perspective on cancer cell metastasis. Science 2011; 331:1559–1564.
10. Chaffer CL, San Juan BP, Lim E, et al. EMT, cell plasticity and metastasis. Cancer Metastasis Rev 2016; 35:645–654.
11. Thiery JP, Acloque H, Huang RYJ, et al. Epithelial-mesenchymal transitions in development and disease. Cell 2009; 139:871–890.
12. Lamouille S, Xu J, Derynck R. Molecular mechanisms of epithelial-mesenchymal transition. Nat Rev Mol Cell Biol 2014; 15:178–196.
13. Yu M, Bardia A, Wittner BS, et al. Circulating breast tumor cells exhibit dynamic changes in epithelial and mesenchymal composition. Science 2013; 339:580–584.
14. Connor AA, McNamara K, Al-Sukhni E, et al. Central, but not peripheral, circulating tumor cells are prognostic in patients undergoing resection of colorectal cancer liver metastases. Ann Surg Oncol 2016; 23:2168–2175.
15. Carter L, Rothwell DG, Mesquita B, et al. Molecular analysis of circulating tumor cells identifies distinct copy-number profiles in patients with chemosensitive and chemorefractory small-cell lung cancer. Nat Med 2017; 23:114–119.
16. Antonarakis ES, Lu C, Luber B, et al. Clinical significance of androgen receptor splice variant-7 mRNA detection in circulating tumor cells of men with metastatic castration-resistant prostate cancer treated with first- and second-line abiraterone and enzalutamide. J Clin Oncol 2017; 72:e170–e171.
17. Poruk KE, Valero V 3rd, Saunders T, et al. Circulating tumor cell phenotype predicts recurrence and survival in pancreatic adenocarcinoma. Ann Surg 2016; 264:1073–1081.
18. Chang M-C, Chang Y-T, Chen J-Y, et al. Clinical significance of circulating tumor microemboli as a prognostic marker in patients with pancreatic ductal adenocarcinoma. Clin Chem 2016; 62:505–513.
19. Court CM, Ankeny JS, Sho S, et al. Circulating tumor cells predict occult metastatic disease and prognosis in pancreatic cancer. Ann Surg Oncol 2018; 25:1000–1008.
20. Cristofanilli M, Budd GT, Ellis MJ, et al. Circulating tumor cells, disease progression, and survival in metastatic breast cancer. N Engl J Med 2004; 351:781–791.
21. Wang C, Mu Z, Chervoneva I, et al. Longitudinally collected CTCs and CTC-clusters and clinical outcomes of metastatic breast cancer. Breast Cancer Res Treat 2017; 161:83–94.
22. Adachi Y, Yoshimura M, Nishida K, et al. Acute phase dynamics of circulating tumor cells after paclitaxel and doxorubicin chemotherapy in breast cancer mouse models. Breast Cancer Res Treat 2018; 167:439–450.
23. Clavien PA, Barkun J, de Oliveira ML, et al. The Clavien-Dindo classification of surgical complications: five-year experience. Ann Surg 2009; 250:187–196.
24. Laget S, Broncy L, Hormigos K, et al. Technical insights into highly sensitive isolation and molecular characterization of fixed and live circulating tumor cells for early detection of tumor invasion. PLoS One 2017; 12:e0169427.
25. Groot VP, Gemenetzis G, Blair AB, et al. Defining and predicting early recurrence in 957 patients with resected pancreatic ductal adenocarcinoma. Ann Surg 2018; [Epub ahead of print].
26. Wan L, Pantel K, Kang Y. Tumor metastasis: moving new biological insights into the clinic. Nat Med 2013; 19:1450–1464.
27. Nordgård O, Tjensvoll K, Gilje B, et al. Circulating tumour cells and DNA as liquid biopsies in gastrointestinal cancer. Br J Surg 2018; 105:e110–e120.
28. Cristofanilli M, Braun S. Circulating tumor cells revisited. JAMA 2010; 303:1092–1093.
29. Plaks V, Koopman CD, Werb Z. Cancer. Circulating tumor cells. Science 2013; 341:1186–1188.
30. Bockhorn M, Jain RK, Munn LL. Active versus passive mechanisms in metastasis: do cancer cells crawl into vessels, or are they pushed? Lancet Oncol 2007; 8:444–448.
31. Au SH, Storey BD, Moore JC, et al. Clusters of circulating tumor cells traverse capillary-sized vessels. Proc Natl Acad Sci U S A 2016; 113:4947–4952.
32. Headley MB, Bins A, Nip A, et al. Visualization of immediate immune responses to pioneer metastatic cells in the lung. Nature 2016; 531:513–517.
33. Friedl P, Locker J, Sahai E, et al. Classifying collective cancer cell invasion. Nat Cell Biol 2012; 14:777–783.
34. Groot VP, Rezaee N, Wu W, et al. Patterns, timing, and predictors of recurrence following pancreatectomy for pancreatic ductal adenocarcinoma. Ann Surg 2018; 267:937–945.
35. Groot VP, Gemenetzis G, Blair AB, et al. Implications of the Pattern of Disease Recurrence on Survival Following Pancreatectomy for Pancreatic Ductal Adenocarcinoma. Ann Surg Oncol 2018; 25:2475–2483.
36. Schliekelman MJ, Taguchi A, Zhu J, et al. Molecular portraits of epithelial, mesenchymal, and hybrid states in lung adenocarcinoma and their relevance to survival. Cancer Res 2015; 75:1789–1800.
37. Lambert AW, Pattabiraman DR, Weinberg RA. Emerging biological principles of meta-stasis. Cell 2017; 168:670–691.
38. Rasheed ZA, Yang J, Wang Q, et al. Prognostic significance of tumorigenic cells with mesenchymal features in pancreatic adenocarcinoma. J Natl Cancer Inst 2010; 102:340–351.
39. Brychta N, Krahn T, von Ahsen O. Detection of KRAS mutations in circulating tumor DNA by digital PCR in early stages of pancreatic cancer. Clin Chem 2016; 62:1482–1491.
40. Kulemann B, Rösch S, Seifert S, et al. Pancreatic cancer: circulating tumor cells and primary tumors show heterogeneous KRAS mutations. Sci Rep 2017; 7:4510.
41. Court CM, Ankeny JS, Sho S, et al. Reality of single circulating tumor cell sequencing for molecular diagnostics in pancreatic cancer. J Mol Diagn 2016; 18:688–696.
42. Smerage JB, Barlow WE, Hortobagyi GN, et al. Circulating tumor cells and response to chemotherapy in metastatic breast cancer: SWOG S0500. J Clin Oncol 2014; 32:3483–3489.
43. Rhim AD, Mirek ET, Aiello NM, et al. EMT and dissemination precede pancreatic tumor formation. Cell 2012; 148:349–361.
44. Labelle M, Hynes RO. The initial hours of metastasis: the importance of cooperative host–tumor cell interactions during hematogenous dissemination. Cancer Discov 2012; 2:1091–1099.
45. Stefanovic S, Diel I, Sinn P, et al. Disseminated tumor cells in the bone marrow of patients with operable primary breast cancer: prognostic impact in immunophenotypic subgroups and clinical implication for bisphosphonate treatment. Ann Surg Oncol 2016; 23:757–766.
46. Greaves M, Maley CC. Clonal evolution in cancer. Nature 2012; 481:306–313.
47. Jamal-Hanjani M, Wilson GA, McGranahan N, et al. Tracking the evolution of non-small-cell lung cancer. N Engl J Med 2017; 376:2109–2121.
48. Agarwal A, Balic M, El-Ashry D, et al. Circulating tumor cells: strategies for capture, analyses, and propagation. Cancer J 2018; 24:70–77.
49. Yu M, Bardia A, Aceto N, et al. Cancer therapy. Ex vivo culture of circulating breast tumor cells for individualized testing of drug susceptibility. Science 2014; 345:216–220.
50. Cayrefourcq L, Mazard T, Joosse S, et al. Establishment and characterization of a cell line from human circulating colon cancer cells. Cancer Res 2015; 75:892–901.
51. Boj SF, Hwang C-I, Baker LA, et al. Organoid models of human and mouse ductal pancreatic cancer. Cell 2015; 160:324–338.

circulating tumor cells; CLUSTER study; disease progression; dynamics; early recurrence; longitudinal; mesenchymal; pancreatic cancer; prediction; prospective; treatment response

Supplemental Digital Content

Copyright © 2018 Wolters Kluwer Health, Inc. All rights reserved.