In our work we include only adenocarcinomas because they are far more comum than the pancreatic tumors of endocrine origins. However, it is worthwhile say that pancreatic tumors of endocrine origin are less aggressive than carcinomas, with distinct prognostic parameters . The epidemiology for pancreatic cancer (adenocarcinoma) is heterogeneous, differing from one geographic region to another and in comparison to other types of tumors better studied, such as breast, prostate, lung and colorectal, little is known about the cellular and molecular prognostic factors of this disease. This knowledge gap is mainly due to the high aggressiveness of pancreatic tumors, which results in fewer than 5% of patients with 5-year survival after diagnosis. Additionally, pancreatic cancer presents with very subtle symptoms due to the retroperitoneal location of the organ . Our study showed that in the studied population, only 3.17% of the patients achieved a survival rate of more than 4 years and that the average survival was 9.5 months, which is similar to the pancreatic cancer world statistics [17–20].
According to previous studies, men are more likely to develop pancreatic cancer, and the difference between the genders is higher in developed countries than in developing countries. The same trend was found in our study, in which 52% of the patients were men and 48% were women. Although pancreatic cancer is not among the ten most common cancers, its incidence tends to increase over time. In developing countries, this is mainly due to two primary reasons, increased life expectancy and the adoption of risk behavior, including high calorie and fatty food intake and consumption of tobacco and alcoholic beverages, at early ages [17, 20].
Klöppel et al.  reported that 80% of pancreatic cancer cases are diagnosed in patients between 60 to 80 years of age. In our study, the average age of diagnosis was 65 years and ranged from 37 years to 95 years, which is consistent with previous reports in the literature.
With respect to self-declared ethnicity, 80% of the patients declared themselves white, 15.3% intermediate and 4.6% black. This is in contrast to other studies that have highlighted the black “race” as a risk factor for the pancreatic cancer development [17, 20]. However, Takikita et al.  observed that in the cohort they studied, 75% of the tumors were well- or moderately differentiated and were significantly more common in whites than in other races. This is despite a lack of detailed information in the medical records of patients, as was the case in our study, and which is barrier for epidemiological population studies. Additionally, the high degree of mixing of the Brazilian population between three main parental populations, Native American, African and European, must be considered. As such, in scientific studies, the determination of ethnicity based on self-declaration or through the amount of pigment in the skin of the individual is debatable and is a very subjective classification, especially in populations with a high degree of miscegenation .
By multivariate analysis of the clinical-pathological data, we identified some independent prognostic factors for pancreatic cancer. An age higher than 74 years was an independent factor (p=0.03) because the higher the age of the individual, the greater the chance of developing neoplasms. The risk with respect to old age is already well known due to the duration allowing for the accumulation of mutations in the cells, mainly in the pool of stem cells, which multiply slowly. Over time, these cells acquire the necessary mutations to maintain a balance between them and the organ microenvironment . Another significant prognostic factor was a history of DM 2 (p=0.014) and chronic pancreatitis (p=0.02), which demonstrated that they may be risk factors for pancreatic cancer. These data are in agreement with other several studies that have also identified DM 2 and chronic pancreatitis as risk factors [23–26].
Several studies have been conducted to clarify the relationship between DM 2, chronic pancreatitis and the risk of developing pancreatic cancer. Prizment et al.  showed that a particular single nucleotide polymorphism (SNP) associated with diabetes may also be associated with pancreatic cancer risk. Additionally, they observed that there were more cases of DM 2 in patients diagnosed with pancreatic cancer than among control patients who did not have a diagnosis of cancer. Another hypothesis, proposed by Braun, Bitton-Worms and LeRoith , was that insulin resistance, chronic inflammation and oxidative stress influence the development of pancreatic cancer because these processes are associated with both DM 2 and chronic pancreatitis. When the pancreatic beta cells are aberrantly hyperactive, the pancreatic tissue is exposed to high levels of insulin, which in turn has growth-promoting and mitogenic action, inducing cell proliferation. The tumor cells have a mechanism for capturing glucose that is independent from insulin, which confers a metabolic advantage. It is known that the high concentration of glucose in tissues leads to the formation of reactive oxygen species (ROS) that activate pro-inflammatory cytokines, which in turn, stimulate angiogenic factors such as VEGF . By immunohistochemistry with CD34 and D2-40 antibodies, the blood and lymphatic vessels were identified, counted and tabulated. For statistical purposes, the values were dichotomized as low and high BVD or LVD (Figure 1). By comparing the LVD and the clinical-pathological parameters (Table 2), we found an inverse relationship between LVD and death events (p=0.002). No relationship was found between BVD and the clinical-pathological parameters. A similar result was found by Kaplan-Meyer survival curve analysis (Figure 3), in which low LVD was also related to poor patient survival and a bad prognosis (p=0.021).
The vascularization is considered an important step for tumor progression , so it is surprising the lack of association between BVD and prognostic parameters. However, the absence of a relationship between BVD and prognostic parameters found in our study was also previously reported by Kawauchi et al.  in synovial sarcomas. In squamous cell carcinoma of the larynx, Sullu et al. (2010) found that blood vessel density was significantly higher in high-grade tumors but did not correlate with other prognostic parmeters . This lack of correlation between BVD and prognostic parameters could be due the tendency for pancreatic cancer to metastasize via the lymphangiogenic pathway. However, it also must be emphatizided that vascular diffusion density may also be involved in this process since only tumour cells within a distance less than 20 micro from the nearest neighbouring vessel are considered important for prognostic purpose .
The relationship between LVD and tumor behavior is controversial and has only been recently reported in the literature. Sleeman and Thiele  established a relationship between LVD and lymph node metastases in different types of tumors. Additionally, studies have demonstrated a positive relationship between peritumoral and/or intratumoral LVD, lymph node metastases and poor prognosis in squamous cell head and neck, endometrial and gastric tumors. On the other hand, no relationship was observed in pancreatic, ovarian and transitional cell bladder adenocarcinomas. Other reports have yielded conflicting results, such as in breast, lung, colorectal and oral esophageal cell tumors .
In pancreatic endocrine tumors, Rubbia-Brandt et al.  found a relationship between LVD and tumor clinical behavior. However, there was no relationship between LVD and the presence of lymph node metastases. The expression of VEGF-C, an inducer of lymphatic vessels, was lower in benign tumors than in well-differentiated, malignant tumors. These data show that high LVD is not required for the proliferation of lymphatic vessels and that the malignant transformation of the tumor leads to the induction of pro-lymphangiogenic factors. Additionally, despite the tendency for an increased number of lymphatic vessels in the tumor over time, the invasion of tumor cells into the lymph capillaries occurs early in the tumorigenic process . Similarly, Sipos et al.  reported that the LVD in malignant tumors was increased compared to that found in normal pancreas or in chronic inflammation. However, the intratumoral LVD was low, and the lymph vessels showed decreased proliferation, rather than being stimulated due to the presence of inflammatory factors. As such, functional intratumoral lymphatic vessels are not required during lymphatic invasion, and their loss of function can be caused by several different mechanisms .
Because we observed an inverse relationship between LVD and survival, we hypothesized that low LVD is due to the accelerated growth of the tumor and the eventual loss of function of the lymphatic vessels that were present. Accordingly, tumors with low LVD are more likely to be fast growing and of a more advanced stage, which would explain the low survival rate of these patients. The function of these vessels could have been lost because of the internal pressure caused by rapid tumor growth, which would lead to the collapse of the vessels, the invasion of tumor cells in the lymph vessels or other mechanisms that have yet to be elucidated . Some studies have suggested that despite the low intratumoral LVD, one can still find a high peritumoral LVD. These may be the vessels through which lymphatic metastasis occurs, resulting in the spread of tumor cells to the regional lymph nodes and, in the case of pancreatic tumors, the liver [33, 36].
In agreement with the research by Yin et al. , which demonstrated the importance of the lymphangiogenesis in the evolution of diabetes and pancreatic inflammation, our work found these diseases to be independent prognostic factors for pancreatic cancer. Furthermore, the formation of new lymph vessels is an early process that develops before tumorigenesis, and as the tumor grows, the lymph vessels are inhibited or disabled. Our study also reports the same prognostic importance of LVD that was found by Wang et al. .
Of the 100 pancreatic tumor cases that were used to construct the TMA and subsequently analyzed by immunohistochemistry for VEGF expression, 71% showed cytoplasmic VEGF expression. By Kaplan-Meyer survival curve analysis, the expression of VEGF in tumors proved to be significantly associated (p=0.023) with poor prognosis and a reduction in patient survival time.
It is known that the VEGF family of proteins contributes to the development of new blood vessels through the process of angiogenesis, which was widely studied and discussed by Folkman and reviewed by other researchers. Angiogenesis leads to the spread of tumor cells by hematogenous routes and, consequently, to metastasis [39–42]. The involvement of other organs by metastases significantly decreases the patient's survival. Additionally, the overexpression of VEGFs by the tumor or cells in the extracellular matrix leads to the rapid growth of the tumor mass because these factors have mitogenic properties that stimulate DNA synthesis in endothelial cells and provide better nutrition and tumor oxygenation through the formation of new blood vessels .
The family of VEGFs is composed of VEGF-A, VEGF-B, VEGF-C, VEGF-D, VEGF-E, VEGF-F, and placental growth factor (PlGF). VEGF-A was the first to be discovered and, therefore, was originally named VEGF. This factor was directly linked to angiogenesis that occurs in physiological processes, such as embryonic development, and in pathological processes, which include inflammation, wound healing and tumor growth. The function of VEGF-B, which is expressed in skeletal and cardiac muscle and brown adipose tissue, is not understood. However, unlike VEGF-A, its expression is not stimulated by low temperatures or by hypoxia. VEGF-C and-D are linked to lymphangiogenesis and, unlike VEGF-A, are only expressed in lymphatic endothelial cells in physiological situations. The most recently discovered factors are VEGF-E, which is encoded by a virus, and VEGF-F, which is expressed in the venom of a snake .
In our research, we used a VEGF antibody that, according to the manufacturer, has affinity for the protein isoforms with splicing in amino acids 121, 165 and 189, which correspond to VEGF-A [44, 45]. The relationship that was observed between the presence of VEGF-A and low survival can be accounted for by the mitogenic activities of VEGF-A, which contribute to the rapid growth of the tumor and, consequently, poor prognosis.
The independent prognostic factors identified in our study, advanced age and a history of DM 2 or chronic pancreatitis, are supported by the relationship between inflammation and the increased expression of VEGF. During inflammatory diseases, the chronic high concentrations of insulin, ROS, cytokines and inflammatory mediators, such as TNF-α and COX-2, lead to an increase the expression of pro-angiogenic factors, the most important being VEGF-A [23, 24].
Even though VEGF-A is directly linked to angiogenesis, recent studies have shown that this factor is also involved in lymphangiogenesis during tumor development. The biological effects of the VEGF family members occur through the interaction of the factors, which are secreted by tumor cells or matrix inflammatory cells, with transmembrane receptors located in the vascular and lymphatic endothelial cells. The interaction of the ligand with receptor leads to dimerization and the autophosphorylation of the receptor intracellular domains that, in turn, generates a cascade of reactions involving proteins related to cellular survival and division. VEGF-A can interact with the receptors VEGFR-1, VEGFR-2 and the neuropilins 1 and 2. However, it is through interaction with the VEGFR-2, found specifically in lymphatic endothelial cells, that lymphangiogenesis is induced .
VEGF-A can promote lymphangiogenesis either indirectly or directly. Indirectly, it can recruit inflammatory cells that, in turn, will produce VEGF-C and VEGF-D, which are known to promote lymphangiogenesis. Directly, VEGF-A will interact with its receptor, VEGFR-2, which is present in pre-existing lymphatic vessels, activating them . According to Wirzenius et al. , while the binding of VEGF-A to VEGFR-3 affects the spreading of the vessel network, its binding to VEGFR-2 produces an increase in lymphatic vessels size, making them able to drain the interstitial fluid of the tumors. In this way, VEGF-A promotes the development of an ideal pre-metastatic niche for the initiation of metastases .
VEGF expression is increased both at the transcription and translation levels in pancreatic tumor tissues, compared to surrounding non-tumoral tissue. Its expression, according to Liang et al. , is related to tumor size, stage and lymph node metastases, demonstrating it as an important prognostic marker of tumor behavior.
In our study, we demonstrated that the LVD is inversely related to survival, while the expression of VEGF-A is directly related to survival. In accordance with the current literature, we postulate that the expression of VEGF-A is an early event in the development of cancer. Inflammatory processes, which are known to increase the likelihood of cancer development, also promote the overexpression of VEGF-A [24, 25, 48]. The higher the expression of VEGF in the tumor during its development, the faster the growth of the tumor. In turn, this causes the previously developed lymph vessels to be inactivated by the increased internal pressure of the tumor and, consequently, they will collapse and be destroyed .