Circulating tumor cells enumeration from the portal vein for risk stratification in early pancreatic cancer patients

cancers Article Circulating Tumor Cells Enumeration from the Portal Vein for Risk Stratification in Early Pancreatic Cancer Patients Javier Padillo-Ruiz 1,*, Gonzalo Suarez 1, Sheila Pereira 1, Francisco JoséCalero-Castro 1, Jose Tinoco 1, Luis Marin 1, Carmen Bernal 1, Carmen Cepeda-Franco 1, Jose Maria Alamo 1, Francisco Almoguera 1, Hada C. Macher 2, Paula Villanueva 1, Francisco JoséGarcía-Fernandez 3, Inmaculada Gallego 4, Manuel Romero 3, Miguel Angel Gomez-Bravo 1, Valeria Denninghoff 5,* and María JoséSerrano 6,7,8   Citation: Padillo-Ruiz, J.; Suarez, G.; Pereira, S.; Calero-Castro, F.J.; Tinoco, J.; Marin, L.; Bernal, C.; Cepeda-Franco, C.; Alamo, J.M.; Almoguera, F.; et al. Circulating Tumor Cells Enumeration from the Portal Vein for Risk Stratification in Early Pancreatic Cancer Patients. Cancers 2021,13, 6153. https:// doi.org/10.3390/cancers13246153 Academic Editor: James Hicks Received: 5 October 2021 Accepted: 3 December 2021 Published: 7 December 2021 Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. Copyright: © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/). 1Instituto de Biomedicina de Sevilla, University Hospital Virgen del Rocío, 41013 Seville, Spain; [email protected] (G.S.); [email protected] (S.P.); fjcaler[email protected] (F.J.C.-C.); [email protected] (J.T.); [email protected] (L.M.); [email protected] (C.B.); [email protected] (C.C.-F.); [email protected] (J.M.A.); [email protected] (F.A.); [email protected] (P.V.); [email protected] (M.A.G.-B.) 2Department of Molecular Biochemistry, University Hospital Virgen del Rocío, 41013 Seville, Spain; [email protected] 3Department of Gastroenterology, University Hospital Virgen del Rocío, 41013 Seville, Spain; fjgarciaf.hvr[email protected] (F.J.G.-F.); mr[email protected] (M.R.) 4Department of Oncology, University Hospital Virgen del Rocío, 41013 Seville, Spain; [email protected] 5Molecular-Clinical Lab, University of Buenos Aires (UBA)—National Council for Scientific and Technical Research (CONICET), Buenos Aires C1122AAH, Argentina 6Oncology Unit, Centre for Genomics and Oncological Research-GENYO, Pfizer, University of Granada, Andalusian Regional Government, 18016 Granada, Spain; [email protected] 7Integral Oncology Division, Instituto Biosantario Granada (iBS-Granada), Virgen de las Nieves University Hospital, 18012 Granada, Spain 8Department of Pathological Anatomy, Faculty of Medicine, University of Granada, 18071 Granada, Spain *Correspondence: [email protected] (J.P.-R.); [email protected].ar (V.D.) Simple Summary: Effective biomarkers are needed to enable personalized medicine for pancreatic cancer patients. This study analyzes the prognostic value, in early pancreatic cancer, of circulating tumor cells and clusters from the central venous catheter and portal blood. Circulating tumor cells were isolated using an immunomagnetic selection and were detected by microscopy using immunocytochemistry staining. In conclusion, the circulating tumor cell number in portal blood identifies a death risk in patients with early pancreatic cancer. Abstract: Background. Effective biomarkers are needed to enable personalized medicine for pancreatic cancer patients. This study analyzes the prognostic value, in early pancreatic cancer, of single circulating tumor cell (CTC) and CTC clusters from the central venous catheter (CVC) and portal blood (PV). Methods. In total, 7 mL of PV and CVC blood from 35 patients with early pancreatic cancer were analyzed. CTC were isolated using a positive immunomagnetic selection. The detection and identification of CTC were performed by immunocytochemistry (ICC) and were analyzed by Epi-fluorescence and confocal microscopy. Results. CTC and the clusters were detected both in PV and CVC. In both samples, the CTC number per cluster was higher in patients with grade three or poorly differentiated tumors (G3) than in patients with well (G1) or moderately (G2) differentiated. Patients with fewer than 185 CTC in PV exhibited a longer OS than patients with more than 185 CTC (24.5 vs. 10.0 months; p= 0.018). Similarly, patients with fewer than 15 clusters in PV showed a longer OS than patients with more than 15 clusters (19 vs. 10 months; p= 0.004). These significant correlations were not observed in CVC analyses. Conclusions. CTC presence in PV could be an important prognostic factor to predict poor prognosis in early pancreatic cancer. In addition, the number of clustered-CTC correlate to a tumor negative differentiation degree and, therefore, could be used as a diagnostic biomarker for pancreatic cancer. Cancers 2021,13, 6153. https://doi.org/10.3390/cancers13246153 https://www.mdpi.com/journal/cancers Cancers 2021,13, 6153 2 of 15 Keywords: circulating tumor cell; cluster; portal vein; central venous catheter; pancreatic cancer; early stage; death risk stratification Highlights •CTCs can be detected in the early stages of pancreatic cancer. •CTC inside a cluster is much higher in G3 than in G1–2 in both samples. • The larger the size of the tumor, the greater the number of total CTC for PV and CVC samples. • The number of CTC < 185 in PV (HR = 4.464; p= 0.016) and no vascular invasion (HR: 3.663; p= 0.013) were independent predictors of better long-term survival. 1. Introduction Pancreatic ductal adenocarcinoma (PDAC) is one of the most aggressive cancers associated with poor prognosis and high mortality. It has a natural tendency toward very early spread, even in resectable cases. Low-grade pancreatic cancers (G1) tend to grow and spread more slowly than high-grade (G3) cancers. Most of the time, G3 tends to have a poor prognosis compared to G1 or G2 cancers [ 1 – 3 ]. Therefore, it is necessary to find prognostic markers that identify the minimal residual disease and predict the risk of relapse. Seriate analyses of CA19-9 during chemotherapy represent practical and specific markers of response to the treatment itself. Other relevant prognostic parameters are the presence of perineural, vascular, and lymphatic invasion, which are essential characteristics for tumor growth and dissemination [1–3]. Unfortunately, there are currently no prognostic and predictive markers to indicate the presence of vascular, neural, or lymphatic invasion, hindering the clinical management of patients and, therefore, their possibility of cure. Among several biomarkers that could be evaluated preoperatively, single circulating tumor cells (CTC), as part of the liquid biopsy family, could be one of the proposed clinical markers. Several studies have reported CTC’s prognostic and predictive value in cancer patients, including pancreatic cancer [ 4 – 6 ]. In solid tumors, detection of CTC is usually performed using peripheral blood. Still, this search has not been possible in pancreatic cancer because blood flow first drains into the portal vein (PV), continues to the liver, and finally reaches the peripheral blood. Consistently, previous work has shown that CTC in patients with pancreatic cancer is detected with greater certainty in PV than in the peripheral venous blood of such patients [ 7 ]. Little is known about the characteristics and clinical implications of finding CTC clusters in PV. Although there are studies in advanced pancreatic cancer [ 8 , 9 ], only a few evaluate the role of CTC and clusters in the portal and central venous blood in patients with early pancreatic cancer as preoperative markers of risk stratification [ 7 , 10 ]. In this work, CTC and clusters were evaluated in patients with early pancreatic adenocarcinoma from the portal and central venous blood samples collected simultaneously during pancreaticoduodenectomy (PD). This study aimed to correlate CTC and clusters count from PV and central venous catheter (CVC) with tumor grade, preoperatively tumor size/CA19-9, and vascular/lymphatic/neural invasion. Additionally, to correlate this finding with overall survival (OS) and disease-free interval (DFS), including liver metastasis or local recurrence. 2. Material and Methods 2.1. Study Design and Sample Collection This work was a prospective cross-sectional and longitudinal study involving 35 patients with carcinoma in the head of the pancreas. Patients with metastasis were excluded, and all patients received gemcitabine as treatment. Twenty-two were men and 13 women with a mean age of 67.4 ± 9.8 years old. The Ethics Committee of the University Hospital Virgen del Rocio approved the study protocol. Informed consent was obtained from all patients. According to the American Joint Committee on Cancer (AJCC) TNM Classification of Malignant Tumors, the study included stages I (n= 15) and II (n= 20) Cancers 2021,13, 6153 3 of 15 of disease. Only patients in whom no metastasis was shown by 18F-Fluro-deoxyglucose positron emission tomography and computed tomography (FDG-PET-CT) underwent surgery. Biopsy was also performed for any suspicious lesions, and if an intraoperative analysis was positive for metastatic adenocarcinoma, PD was canceled. The exclusion criteria were metastatic cancer diagnosis, locally advanced tumor, pancreatitis episodes within the three months before surgery, and previous history of cancer within the past five years. The tumor size and CA 19-9 (268.3 ± 586.9 U/mL) were preoperatively evaluated. PD was performed in all 35 patients. Complete resection (R0) was achieved in 24 (69%) patients according to established protocol [ 11 , 12 ]. Based on their differentiation, tumors were divided into three grades: well-differentiated (G1) in 8 (23%) patients, moderately differentiated (G2) in 20 (57%), and poorly differentiated (G3) in 7 (20.0%). The microscopic vascular invasion was observed in 13 (37%) patients, lymphatic invasion in 9 (26%), and neural invasion in 18 (51%). Adjuvant chemotherapy based on gemcitabine-based formula was administered postoperatively after pathological diagnosis of pancreatic carcinoma for 35 patients. All cases were followed prospectively during 24 months. A CT scan was performed to diagnose liver metastases and local recurrence as a low-density mass in the liver or surgical area. 2.2. Validation of CTC Detection Cell culture Pancreatic cancer cell PANC-1 (ATCC ® CRL-1469, Manassas, VA, USA) were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM, Lonza BE12-604F, Basel, Switzerland) supplemented with 10% fetal bovine serum (F7524, Sigma-Aldrich, St. Louis, MO, USA) and 1% Penicillin/Streptomycin (15140122, GIBCO). Cell cultures were grown in a humidified incubator with 5% CO 2 at 37 ◦ C. Once the cells were needed, the cells were washed with Trypsin-EDTA (25300062, GIBCO, Thermo Fisher Scientific, Inc., Waltham, MA, USA). Pancreatic cancer cells (PCC) were used as a positive control. Two control groups were used for negative and positive control. For technical negative control, 7 mL aliquots of peripheral blood were collected from healthy donors (n= 3) and were mixed with approximately 300 PCC for positive control. All samples were isolated with the Isoflux ™ system and detected following the manufacturer’s protocol by Fluxion. The specificity of the anti-cytokeratin (CK) was evaluated using the pancreatic tumor PANC-1 cell line (ATCC ® CRL-1469, Manassas, VA, USA) as a positive control because it has a high CK protein expression; and the peripheral blood mononuclear cell (PMBC) as a negative control for its non-expression of EpCAM and CK (source: The human protein atlas). Clinical negative controls of PV samples would not be ethically approved because there are surgical risks in that sampling. Regardless of this group, we included CVC samples from a non-neoplastic control group based on their clinical history, such as those undergoing cardiac surgeries. Therefore, the sample from 8 patients was extracted from the pre-implanted CVC before cardiac surgery. We performed oncological biochemical markers of these patients in the sample extracted simultaneously: prostate-specific antigen (PSA), alpha-fetoprotein (AFP), carcinoembryonic antigen (CEA), carbohydrate antigen- 15.3 (CA15-3), carbohydrate antigen-19.9 (CA-19.9), carbohydrate antigen-125 (CA-125). As an inclusion criterion to consider as proper negative controls for the recruited patients, we determined that the oncological markers tested in the sample extracted simultaneously and in the same place as the one extracted for CTC evaluation had average values. 2.3. CTC Isolation, Detection, and Enumeration During surgery, before manipulating the tumor, 7 mL blood samples were simultaneously obtained from the CVC tip in the superior vena cava and from the PV by direct punction. The first blood-draw was discarded to exclude epithelial cells dislodged by the venipuncture (these factors play a role in the frequency of epithelial cells in the blood, as these steps may lead to unspecific shedding of epithelial cells). Both patient samples were collected in K2-EDTA Vacutainer tubes, maintained at room temperature, and processed 24 h after collection. Blood samples were enriched in peripheral mononuclear blood cells Cancers 2021,13, 6153 4 of 15 using gradient centrifugation with Histopaque ® -1119, and CTC were isolated using the IsoFlux platform. Isoflux platform has been designed to isolate the CTC based on a microfluidic process for immunomagnetic positive selection. The IsoFlux ™ system utilizes micrometer-scale beads, which have been shown to result in a magnetic moment that is sufficient for capturing cells even with low target expression [ 13 ]. The process is automatic, increasing the ability to capture the CTC. The Isoflux ™ Epithelial to Mesenchymal Transitions Circulating Tumor Cell Enrichment Kit (EMT Enrichment Kit, Izasa, Catalog N.910-0106, Werfen, San Diego, CA, USA) was used for performing CTC enrichment. In this kit, beads were conjugated with four different antibodies, targeting both epithelial and mesenchymal markers. The kit utilizes both anti-EpCAM and anti-EGFR-antibodies for the detection of epithelial cells, as well as anti-N-Cadherin and anti-Vimentin as mesenchymal markers. The EpCAM is a cell surface molecule known to be highly expressed in solid cancers. The immunomagnetic beads conjugated with antibodies were added to cells suspended in IsoFlux Binding Buffer and incubated for 90 min at 4 ◦ C, as indicated in the Fluxion protocol. Subsequently, they were subjected to immunomagnetic isolation with IsoFlux (Fluxion Biosciences Inc). After the sample was processed, the enriched cells were fixed and stained with the fluorescent reagents (Isoflux ™ Circulating Tumor Cell Enumeration Kit Izasa Catalog N.910-0093, Werfen, San Diego, CA, USA). The fluorescent reagents included were anti-CK-fluorescein isothiocyanate (FITC) specific for the intracellular protein CK, characteristic of epithelial cells; anti-CD45-Indocarbocyanine (Cy3) specific for leukocytes; and Hoechst 33342, a nucleic acid stain cell-permeant nuclear counterstain that emits blue fluorescence when bound to AT-rich regions of the minor groove in DNA. The CTC detection and enumeration were performed by fluorescence microscopy, and the images obtained were processed using image software based on the Hough transform (VR-CTC). We had described previously a method based on image processing to count sets of pixels showing a cell nucleus, cytokeratin expression, and no CD45 expression [ 14 ]. To count them, the Hough transform was used. The approach allowed a classification of the events counted in CTC, cluster, and clustered-CTC. Then, CTC were enumerated as morphologically intact CK+/CD45-/nucleated cells. The size of cells and clusters and the number of clustered CTC were also characterized. The mean coefficient of variation in CTC determination was less than 2% [0–1.41]. The sensitivity presented by this approach was 85.58%, and the specificity was 88.16% [ 14 ]. The results obtained were compared with counting undertaken by a technician. High correlations were demonstrated in total pancreatic tumor cells in healthy donors (R 2 = 0.995) and patients with pancreatic cancer (R2= 0.955) as well as in free cells (R2= 0.993 and R2= 0.975, respectively) [14]. 2.4. Statistical Analysis Single regression analysis was performed to assess the linearity of the two blood samples for detecting CTC. The correlation of CTC counts and tumor invasion (vascular, lymphatic, and neural) was analyzed using the Mann–Whitney U test and the Kruskal– Wallis H test. The distributions of patients above and below the cut-off level in CTC were compared using Fisher’s exact test. Spearman’s rank correlation coefficient examined the relationship between CTC counts and the tumor marker (Ca19.9). Survival analysis was done using the product-limit Kaplan–Meier and the log-rank test. Multivariate analysis was performed using a Cox proportional hazard model to test for independent prognostic variables. All statistical calculations were carried out using IBM Corp. released 2020, IBM SPSS Statistics for Windows, Version 27.0. Armonk, NY, IBM Corp. The required significance was p< 0.05. 3. Results 3.1. Validation of CTC Detection These experimental analyses included tumor cell lines and healthy donors as positive and negative technical controls, respectively. Negative control presented 0.67 ±0.31 CTC/mL in blood from healthy donors whereas positive control showed 37.15 ± 9.09 PCC/mL (PCC Cancers 2021,13, 6153 5 of 15 total = 185.75 ± 45.43). Positive controls showed that the detection rate of CTCs with this methodology is approximately 55%. The non-neoplastic control group included eight CVC samples from patients who underwent cardiac surgeries, had no history of cancer in their medical history, and had average results in their tumor biochemical markers taken from the sample simultaneously. These negative controls showed 0.87 ± 0.39 CTC/mL in CVC blood, which found no clusters. The median obtained was 0.50 CTC/mL with a range between 0.00 and 1.75. There are differences in CVC between the number of CTC/mL of patients with pancreatic cancer and non-neoplastic patients who underwent cardiac surgery (U-Mann Whitney p-value < 0.001). No clusters were found in the non-neoplastic group, unlike in samples from pancreatic cancer patients with a median (range) of 14.5 (3.8–35.5) cluster/mL. 3.2. Longitudinal Enumeration and Cluster Evaluation Single CTC and clusters were detected in all patients, both in PV and CVC (Figure 1). The analysis showing the correlation of CTC values on CVC vs. PV appeared in Table 1. In the three measures, Pearson’ s correlation had a coefficient r = 0.6 between PV and CVC. We found more CTC, clusters, and total CTC in CVC than PV, with no statistically significant differences (Table 1). The ratio and sizes are merely informative data used as quality control between trials, which should not differ between the samples 291.8 (120.0–500.0). Table 1. CTC characterization by Hough transform method. Research Variable PV (Median, Range) CVC (Median, Range) U-Mann Whitney p-Value Correlation Correlation p-Value Free CTC (cell/mL) 235.4 (101.3–375.3) 291.8 (120.0–500.0) 0.151 0.6 0.004 Cluster/mL 12.9 (4.8–33.0) 14.5 (3.8–35.5) 0.622 0.6 0.001 CTC inside a cluster 30.4 (12.5–89.4) 37.4 (11.3–89.5) 0.205 0.5 0.008 Total CTC (cell/mL) 310.0 (132.1–446.0) 405.7 (130.7–553.8) 0.239 0.6 0.001 Ratio Free CTC/CTC inside a cluster 2.6 (2.4–3.0) 2.5 (2.3–2.9) 0.051 0.2 0.306 Free CTC median size 7.5 (6.8–8.7) 7.8 (6.7–9.0) 0.981 0.3 0.229 CTC inside a cluster median size 10.2 (8.5–11.6) 10.3 (8.6–11.7) 0.990 0.3 0.074 When the data obtained were compared by discriminating according to the tumor grade, the values reached in the same tumor grade showed that in CVC, they are higher than in PV (Tables 2and 3). When comparing tumor grades for the same sampling site, we found that for both samples, the data that most attracts attention is CTC inside a cluster, which is much higher in G3 than in G1/2, despite not having significant differences. The ratio and sizes have no variation as we expected. Table 2. Correlation CTC and clusters according to the degree of tumor differentiation (CVC measurements). Research Variable G1–G2 G3 p-Value Free CTC (cell/mL) 279.2 (120.3–482.3) 357.0 (182.5–1020.8) 0.483 Cluster/mL 14.0 (3.3–31.7) 41.0 (17.9–46.8) 0.107 CTC inside a cluster 36.7 (8.7–73.7) 97.3 (54.5–116.3) 0.071 Total CTC (cell/mL) 399.0 (129.0–538.5) 481.0 (243.0–1117.7) 0.318 Ratio Free CTC/CTC inside a cluster 2.6 (2.4–2.9) 2.4 (2.2–3.1) 0.483 Free CTC median size 7.8 (6.6–9.1) 8.6 (6.7–10.2) 0.521 CTC inside a cluster median size 10.3 (8.7–11.6) 10.6 (7.8–13.2) 0.908 Table 3. Correlation CTC and clusters according to the degree of tumor differentiation (PV measurements). Research Variable G1–G2 G3 p-Value Free CTC (cell/mL) 240.3 (102.2–373.3) 227.0 (140.8–559.0) 0.841 Cluster/mL 10.5 (4.3–32.7) 7.5 (30.0–80.3) 0.310 CTC inside a cluster 29.7 (9.2–74.2) 78.5 (21.3–260.6) 0.201 Total CTC (cell/mL) 314.5 (133.0–406.7) 305.5 (162.1–819.6) 0.725 Ratio Free CTC/CTC inside a cluster 2.6 (2.3–2.9) 2.8 (2.6–3.9) 0.150 Free CTC median size 7.3 (6.5–8.5) 8.5 (7.0–9.6) 0.310 CTC inside a cluster median size 9.7 (8.1–11.5) 10.2 (9.5–12.4) 0.335 Cancers 2021,13, 6153 6 of 15 Cancers 2021, 13, x 7 of 16 Figure 1. Epifluorescence microscopy pictures. The CTC detection and enumeration were performed by fluorescence microscopy, and the images obtained were processed using image software based on the Hough transform (VR-CTC). The CTC recovery was stained with: (a) anti-CK-fluorescein isothiocyanate (FITC) specific for the intracellular protein cytokeratin (characteristic of epithelial cells) (Green); (b) anti-CD45-Indocarbocyanine (Cy3) specific for leukocytes (Red); (c) Hoechst 33342, used for nuclear staining (Blue); and (d) merged image subject to logical operations where we performed a logical multiplication between the thresholding of channels Blue (nuclear) and Green (CK). Finally, we removed the pixels of the thresholding of channel Red (CD45). We eliminated artifacts from the resulting mask using Matlab tools. The white circle in d3 shows the merged image of CTC, which is CK+/CD45−/nucleated cells; in this way, image pixels are Hoechst+, CK+, and CD45—corresponding to a CTC. Data from (1) negative control (PMBC), (2) positive control (PANC- 1 cell line), (3) CTC detected in PVC, and (4) CTC detected in CVC. The preoperative CA 19-9 levels and tumor size measured by CT scan were analyzed and correlated with CTC and CTC clusters. In both samples, CVC and PV, we did not find a significant association (Tables 4 and 5). We used the Spearman correlation coefficient (Rho) for these analyses, which takes values from +1 to −1. A Rho of +1 indicates a perfect association of ranks, a Rho of zero indicates no association between levels, and an r of −1 indicates a perfect negative association of grades. The closer Rho is to zero, the weaker the Figure 1. Epifluorescence microscopy pictures. The CTC detection and enumeration were performed by fluorescence microscopy, and the images obtained were processed using image software based on the Hough transform (VR-CTC). The CTC recovery was stained with: ( a ) anti-CK-fluorescein isothiocyanate (FITC) specific for the intracellular protein cytokeratin (characteristic of epithelial cells) (Green); ( b ) anti-CD45-Indocarbocyanine (Cy3) specific for leukocytes (Red); ( c ) Hoechst 33342, used for nuclear staining (Blue); and ( d ) merged image subject to logical operations where we performed a logical multiplication between the thresholding of channels Blue (nuclear) and Green (CK). Finally, we removed the pixels of the thresholding of channel Red (CD45). We eliminated artifacts from the resulting mask using Matlab tools. The white circle in d3 shows the merged image of CTC, which is CK+/CD45 − /nucleated cells; in this way, image pixels are Hoechst+, CK+, and CD45—corresponding to a CTC. Data from ( 1 ) negative control (PMBC), ( 2 ) positive control (PANC-1 cell line), (3) CTC detected in PVC, and (4) CTC detected in CVC. The preoperative CA 19-9 levels and tumor size measured by CT scan were analyzed and correlated with CTC and CTC clusters. In both samples, CVC and PV, we did not find a significant association (Tables 4and 5). We used the Spearman correlation coefficient (Rho) for these analyses, which takes values from +1 to − 1. A Rho of +1 indicates a perfect association of ranks, a Rho of zero indicates no association between levels, and an r of − 1 indicates a perfect negative association of grades. The closer Rho is to zero, the weaker Cancers 2021,13, 6153 7 of 15 the association between the ranks. Therefore, relationships identified using correlation coefficients should be interpreted as associations and not as causal relationships [ 15 ]. In both samples, the value of CA 19-9 has a positive relationship with CTC/cluster. Table 4. Tumor and CA 19-9 preoperatively evaluated by CTC characteristics with Spearman correlation coefficient (Rho) (CVC measurements). Research Variable CA 19-9 (Rho) Tumor Size * (cm) (Rho) Free CTC (cell/mL) 0.5 0.1 Cluster/mL 0.2 −0.1 CTC inside a cluster 0.2 −0.1 Total CTC (cell/mL) 0.4 0.4 Ratio Free CTC/CTC inside a cluster −0.3 −0.3 Free CTC median size −0.5 −0.1 CTC inside a cluster median size −0.6 −0.6 * Measured by CT scan. Table 5. Tumor and CA 19-9 preoperatively evaluated by CTC characteristics with Spearman correlation coefficient (Rho) (PV measurements). Research Variable CA 19-9 (Rho) Tumor Size * (cm) (Rho) Free CTC (cell/mL) 0.4 0.3 Cluster/mL 0.3 0.3 CTC inside a cluster 0.3 −0.1 Total CTC (cell/mL) 0.4 0.4 Ratio Free CTC/CTC inside a cluster 0 0 Free CTC median size −0.1 −0.1 CTC inside a cluster median size −0.2 0 * Measured by CT scan. Other factors related to survival are neural, vascular, and lymphatic invasion (Tables 6and 7). There was no significant correlation between CTC measurements with neural, vascular, or lymphatic invasion. Still, we are struck by the high number of CTCs (free and total) in PV samples in patients with vascular invasion concerning those who do not have it. The biological mechanism of tumor hematogenous dispersal is associated with vascular invasion. Table 6. Correlation of CTC and CTC clusters with microscopic invasion (CVC measurements). Research Variable Free CTC (Cell/mL) p-Value Cluster/mL p-Value Total CTC (Cell/mL) p-Value Vascular invasion 304.9 (151.6–656.4) 0.662 12.1 (3.1–29.5) 0.368 421.8 (162.7–697.9) 0.692 No vascular invasion 291.8 (116.1–480.9) 26.6 (4.3–41.0) 384.4 (123.3–522.5) Lymphatic invasion 304.9 (79.1–455.7) 0.504 31.1 (2.7–42.6) 0.565 421.8 (87.8–501.6) 0.629 No lymphatic invasion 291.8 (133.1–582.6) 13.3 (3.8–31.0) 384.4 (161.0–627.4) Neural invasion 350.5 (84.9–498.5) 0.728 13.4 (2.7–42.1) 0.667 443.1 (86.8–537.7) 0.728 No neural invasion 275.8 (164.5–625.3) 14.5 (10.5–32.8) 384.4 (198.2–668.0) Table 7. Correlation of CTC and CTC clusters with microscopic invasion (PV measurements). Research Variable Free CTC (Cell/mL) p-Value Cluster/mL p-Value Total CTC (Cell/mL) p-Value Vascular invasion 321.7 (217.4–451.3) 0.104 16.3 (5.5–41.5) 0.362 374.2 (271.0–508.7) 0.089 No vascular invasion 184.5 (97.2–322.3) 12.3 (3.2–29.9) 247.2 (125.4–401.0) Lymphatic invasion 204.7 (74.0–514.3) 0.489 23.8 (5.8–52-5) 0.397 350.5 (88.9–635.6) 0.939 No lymphatic invasion 248.2 (118.2–377.3) 12.3 (4.3–31.3) 305.5 (138.2–419.2) Neural invasion 247.3 (94.3–375.3) 0.905 13.8 (5.5–34.5) 0.528 327.3 (120.6–492.4) 0.798 No neural invasion 222.5 (113.0–370.8) 10.3 (3.1–32.2) 285.2 (141.1–404.7) Cancers 2021,13, 6153 8 of 15 3.3. Prognostic Factors for Long-Term Survival Regarding OS, the median survival time was 18 months (range 12.5–23.5 months) without postoperative mortality. Thirty of the 35 patients (86%) survived for more than six months, with 20 (57%) of them still alive one year after the primary diagnosis. Once the media (m) and standard derivation were calculated, based on Altmann optimal cut-off approximation, we estimated the p67 (or p33) value consulting in a probability table of the normal distribution the value of Pr (z < z) = 0.67, that corresponds to 0.724 [ 16 ]. Therefore, we estimate that quartile as m + 0.724s. Usually, the literature uses Q3 (p75) to consider the high value and Q1 (p25) to low values. We decided to use p33 instead of p25 to increase the sensibility at lower specificity. In our opinion, to evaluate possible cut-off points, we did not recommend using extreme values on each side. Therefore, it is excluded between 5 and 10% of each endpoint. According to these results, to predict a G3 pancreatic adenocarcinoma presence, we established a cut-off for cluster number of 20 clusters/mL and CTC inside a cluster of 64 CTC/cluster. The prognostic value of CTC was set in the 33rd percentile (p33), resulting in a cut-off point of 185 (SD 45.3) CTC and 15 (SD 6.8) clusters [15]. Regarding PV measurements, the median patient’s survival was significantly longer in patients with less than 185 CTC (24.5 months (IC 19.6-29.4) vs. 10 months (CI 95% 7.4–12.5); p= 0.018) and less than 15 clusters (19 months (CI 95% 15.8–22.2) vs. 10 months (CI 95% 7.2–12.8); p= 0.004) (Table 8and Figure 2). These findings were not found in CVC samples. Patients with vascular (p= 0.005) or lymphatic invasion (p= 0.044) were associated with less survival. However, the neural invasion did not correlate with OS (Figure 3). To identify prognostic factors for long-term survival, we used a stratified univariate and multivariable Cox regression. The multivariable analysis showed that both the number of CTC < 185 (HR = 4.464; p = 0.016) and no vascular invasion (HR: 3.663; p = 0.013) were independent predictors of better long-term survival (Table 9). Table 8. Median survival by groups (months). Research Variable Deaths N (%) Median Survival 95% CI Log-Rank-Value <185 portal CTC 5 (38.5) 24.5 19.6–29.4 0.018 ≥185 portal CTC 14 (63.6) 10.0 7.4–12.5 <15 portal clusters 8 (42.1) 19.0 15.8–22.2 0.040 ≥15 portal clusters 11 (68.8) 10.0 7.2–12.8 No vascular invasion 9 (40.1) 22.5 17.7–27.2 0.005 Vascular invasion 10 (76.9) 10.0 5.3–14.7 No lymphatic invasion 11 (42.3) 18.0 15.6–23.8 0.044 Lymphatic invasion 8 (88.8) 10.0 5.6–14.4 No neural invasion 7 (41.2) 21.5 15.5–27.4 0.249 Neural invasion 12 (66.7) 13.0 4.9–21.0 Degree of differentiation 0.003 G1–G2 15 (53.5) 19.0 4.6–23.1 G3 2 (40.0) 16.0 7.7–24.7 Global survival 18.0 12.5–23.5 Cancers 2021,13, 6153 9 of 15 Cancers 2021, 13, x 10 of 16 G1–G2 G3 Global survival 15 (53.5) 2 (40.0) 19.0 16.0 18.0 4.6–23.1 7.7–24.7 12.5–23.5 Table 9. Prognostic factors for long-term survival. Covariates were patient age, disease stage (I–II), portal CTC ≥ 185, portal clusters ≥ 15, microscopic invasion (vascular, lymphatic, neural), and tumor degree of differentiation (G1–3). Research Variable Univariate Multiple HR 95% CI p-Value HR 95% CI p-Value Sex (male) 2.050 0.789–5.325 0.140 Age 1.000 0.949–1.054 0.988 - Stage I vs. II 1.566 0.555–4.415 0.396 - ≥185 portal CTC 3.236 1.135–5.221 0.028 4.464 1.316–15.152 0.016 ≥15 portal clusters 2.486 0.989–6.247 0.053 1.330 0.443–4.528 0.624 Vascular invasion 3.568 1.346–9.457 0.011 3.663 1.321–10.204 0.013 Lymphatic invasion 2.418 0.969–6.031 0.058 2.512 0.940–6.711 0.066 Neural invasion 1.698 0.666–4.328 0.268 - G1–2 vs. G3 0.841 0.196–3.771 0.841 - CI: confidence interval; HR: adjusted hazard ratio. Figure 2. Overall survival according to left) PV-CTC’s cut-off; and right) PV-cluster number´s cut-off. Blue line: low cellular expression; Red line: high cellular expression. ≥185 CTC HR = 3.263 (1.135–5.221); p = 0.028. ≥15 clusters HR = 2.486 (0.989–6.247); p = 0.053. (a) (b) Figure 2. Overall survival according to left) PV-CTC’s cut-off; and right) PV-cluster number ´ s cut-off. Blue line: low cellular expression; Red line: high cellular expression. ≥ 185 CTC HR = 3.263 (1.135–5.221); p= 0.028. ≥ 15 clusters HR = 2.486 (0.989–6.247); p= 0.053. Cancers 2021, 13, x 10 of 16 G1–G2 G3 Global survival 15 (53.5) 2 (40.0) 19.0 16.0 18.0 4.6–23.1 7.7–24.7 12.5–23.5 Table 9. Prognostic factors for long-term survival. Covariates were patient age, disease stage (I–II), portal CTC ≥ 185, portal clusters ≥ 15, microscopic invasion (vascular, lymphatic, neural), and tumor degree of differentiation (G1–3). Research Variable Univariate Multiple HR 95% CI p-Value HR 95% CI p-Value Sex (male) 2.050 0.789–5.325 0.140 Age 1.000 0.949–1.054 0.988 - Stage I vs. II 1.566 0.555–4.415 0.396 - ≥185 portal CTC 3.236 1.135–5.221 0.028 4.464 1.316–15.152 0.016 ≥15 portal clusters 2.486 0.989–6.247 0.053 1.330 0.443–4.528 0.624 Vascular invasion 3.568 1.346–9.457 0.011 3.663 1.321–10.204 0.013 Lymphatic invasion 2.418 0.969–6.031 0.058 2.512 0.940–6.711 0.066 Neural invasion 1.698 0.666–4.328 0.268 - G1–2 vs. G3 0.841 0.196–3.771 0.841 - CI: confidence interval; HR: adjusted hazard ratio. Figure 2. Overall survival according to left) PV-CTC’s cut-off; and right) PV-cluster number´s cut-off. Blue line: low cellular expression; Red line: high cellular expression. ≥185 CTC HR = 3.263 (1.135–5.221); p = 0.028. ≥15 clusters HR = 2.486 (0.989–6.247); p = 0.053. (a) (b) Cancers 2021, 13, x 11 of 16 (c) (d) Figure 3. Overall survival according to (a) vascular invasion, HR = 3.568 (1.346–9.457); p = 0.011; (b) lymphatic invasion, HR = 2.418 (0.969–6.031); p = 0.058; (c) neural invasion, HR = 1.698 (0.666–4.328); p = 0.268; and (d) tumor grade differentiation, HR = 0.841 (0.196–3.771); p = 0.841. Blue line: no invasion or low grade (G1–2); red line: invasion or high grade (G3). 3.4. Prognostic Factors for Long Term Local and Systemic Progression Excluding incomplete resections, 23/35 (96%) patients had local recurrence diseasefree survival for more than six months, with 22 (92%) of them being still local progressionfree one year after the primary diagnosis. However, the median disease-free interval related to local recurrence was longer in patients with <185 CTC (28.5 vs. 25.0 months; p = 0.647) and less than 15 clusters (29.0 vs. 22.5 months; p = 0.787) in PV, differences were not significant. The vascular or neural invasion also failed to achieve correlation. Only the lymphatic invasion was associated with less local recurrence disease-free survival (12.0 vs. 23.0 months; p = 0.001). In fact, in the multivariable Cox regression analysis, the only lymphatic invasion was an independent predictor of disease-free interval for local recurrence (p = 0.031). Regarding systemic progression (liver metastases), 18 of the 35 patients (75%) had systemic progression-free survival for more than one year after the primary diagnosis. Regarding local recurrence, although the median disease-free interval related to metastases was longer in patients with <185 CTC (20.5 vs. 18.0 months; p = 0.636) and less than 15 clusters (24.0 vs. 18.0 months; p = 0.383) measured in PV, differences were not significant (Figure 4). Figure 3. Overall survival according to ( a ) vascular invasion, HR = 3.568 (1.346–9.457); p= 0.011; ( b ) lymphatic invasion, HR = 2.418 (0.969–6.031); p= 0.058; ( c ) neural invasion, HR = 1.698 (0.666–4.328); p= 0.268; and ( d ) tumor grade differentiation, HR = 0.841 (0.196–3.771); p= 0.841. Blue line: no invasion or low grade (G1–2); red line: invasion or high grade (G3).