Prognostic significance of cytokeratin 19 expression in pancreatic neuroendocrine tumor: A meta-analysis

Background Pancreatic neuroendocrine tumor (PNET) comprises 1–2% of all pancreatic tumors, but its incidence is increasing. Although many studies have investigated the correlation between cytokeratin 19 (CK-19) and PNET, the prognostic significance of CK-19 expression in PNET is inconclusive. Methods Eligible studies were retrieved from Pubmed, Elsevier, Embase, Cochrane Library and Web of Science databases. All relevant data were extracted to analyze the relationship between CK-19 and PNET. We utilized a fixed or random effects model to calculate the pooled odds ratio (OR) with 95% confidence intervals (CI). Results Pooled data indicated CK-19 expression was significantly associated with poor 3- and 5-year overall survival (OS) for PNET, but not for 1-year overall survival. Additionally, positive CK-19 expression was correlated with large tumor size, advanced differentiation grade in World Health Organization-2010 (WHO-2010) and WHO-2004, vascular invasion, lymph node metastasis and liver metastasis. Conclusions Positive CK-19 expression can be used as a predictor of poor prognosis of PNET.


Introduction
Pancreatic neuroendocrine tumor (PNET) accounts for 1-2% of all pancreatic neoplasms [1]. In the United States, an age-adjusted annual incidence of PNET is 0. 3

Criteria for inclusion and exclusion
All studies included in this meta-analysis fulfilled the following criteria: (1) PNET was confirmed by pathology and not restricted by age or ethnicity; (2) CK-19 expression was measured by immunohistochemistry (IHC); (3) clinical trials or reports were published in English; (4) relevant data was provided directly or could be calculated indirectly; (5) the study with the highest quality assessment was enrolled when trials were performed in the same patient samples. Abstracts, editorials, letters, expert opinions, conference records, book sections, reviews without original data, case reports or studies without control groups were excluded. Studies were also excluded if: (1) the articles were about animals or cell lines; (2) the outcomes or parameters of patients were not clearly reported; (3) articles were overlapping. First, the title and abstract were screened to see whether they fulfilled the inclusion criteria. Second, the full text was further assessed after the initial screening. Finally, the eligibility of studies was verified by two reviewers (DC and JC).

Data extraction and literature quality assessment
Two reviewers (DC and JC) extracted valid data independently from eligible studies, and any discrepancy was resolved by consensus. Relevant characteristics were: (1) first author's name; (2) publication date; (3) number of patients included in this meta-analysis; (4) characteristics of the study population, such as age, gender and clinicopathological features; (5) PNET stage according to WHO-2010 and WHO-2004 classification; (6) methods for evaluating CK-19 expression; (7) manufacturers of antibody; (8) percentage of CK-19 expression; (9) whether overall survival data was provided (Table 1).
Our two reviewers assessed the quality of each selected study using the Newcastle-Ottawa scale (NOS) [12]. The evaluation of the methodology included three aspects: selection, comparability, and outcome or exposure. Final scores ranged from 0 (the least eligible) to 9 (the most eligible). The study would be ruled out if the score was less than 3.

Statistical analysis
The statistical analysis was performed by Review Manager (RevMan) software (version 5.3; Cochrane collaboration, http:ims.cochrane.org/revman/download) and STATA (version 12.0, Stata Corp. College Station, Texas). We pooled statistical variables contained in the original studies directly and obtained variables from available data indirectly or by reading the Kaplan-Meier survival curve according to the method by Parmar MK [13]. The Odd ratio (OR), together with 95% confidence interval (CI), was analyzed to estimate the relationship between CK-19 expression and the prognosis of PNET. A combined OR<1 suggested a worse survival rate, and for clinicopathological features, a combined OR>1 indicated a poor survival outcome. Heterogeneity among enrolled studies was checked with a Chi-square-based Q statistical test. And the I 2 statistic, ranging from 0% to 100%, was also calculated to measure the inter-study heterogeneity. If a P<0.10 and/or I 2 >50%, indicating the presence of heterogeneity, a random-effects model was used. Otherwise, a fixed-effects model was chosen. The publication bias was evaluated by the funnel plots made by Egger's test and Begg's test. If the plots were asymmetrical, the stability of our meta-analysis results needed to be assessed using trim and fill analyses. P<0.05 in the Q statistical test was considered statistically significant.

Selection of trials
A total of 141 studies were retrieved based on the initial search criteria. 34 duplicate articles were excluded. Another 89 studies were excluded because they were case reports, book sections, reviews, animal studies, conference records or abstracts, or had no relationship with the topic or no full text. After reading the full text, we excluded 8 more studies because the information about survival or clinicopathological features was insufficient. At the end of the screening, 10 retrospective studies met inclusion criteria and were used in this meta-analysis [6,[14][15][16][17][18][19][20][21][22] (Fig 1).
Metastasis is another aspect for the evaluation of PNET. In this meta-analysis, the correlation between CK-19 expression and lymph node and liver metastasis was described [6,20,22]. The combined ORs were 5.96 (95% CI: 2.18-16.34, Z = 3.47, P = 0.0005) for lymph node metastasis and 2.96 (95% CI: 1.08-8.12, Z = 2.11, P = 0.04) for liver metastasis (Fig 6). No statistical heterogeneity was found in either analysis (I 2 = 0% and 0%). The result showed that PNET with positive CK-19 expression was more likely to have lymph node and liver metastasis than those without CK-19 expression.

Publication bias
In this meta-analysis, Begg's test and Egger's test were used to evaluate potential publication bias. Based on the Begg's test, no publication bias was found in any study. However, publication bias was found in the studies of 3-year overall survival (P = 0.028) and 5-year overall survival (0.014) in Egger's test.
30-60 years. Moreover, men and women are equally affected, but poorly differentiated PNET occurs more frequently in men [5]. PNET originates from the multipotent ductular or acinar stem cells and nonislet cells [24,25]. Four hereditary syndromes (multiple endocrine neoplasia type 1 (MEN1), von Hippel-Lindau syndrome (VHL), neurofibromatosis type 1 (NF1), and tuberous sclerosis) are associated with PNET [26]. The clinical features depend on whether tumors are functional or non-functional. The functional tumors produce relevant hormone and exhibit a certain endocrine syndrome, while non-functional ones present symptoms, such as abdominal pain, anorexia, nausea, and jaundice, due to the tumor mass, invasion of adjacent tissue or distant metastasis. Currently, radiological, metabolic, serum, and endoscopic strategies are combined to make more precise diagnosis. Surgical resection is curative for localized or early stage PNET. For advanced PNET, systemic treatment or targeted treatment has been implemented, such as somatostatin analogs, agents targeting the vascular endothelial growth factor (VEGF) signaling pathway and the mechanistic target of rapamycin (mTOR). Cytotoxic chemotherapy is another option for patients [27].
Cytokeratins (CK) are a kind of intermediate filament protein. CK-19 is an acidic protein type I, a subgroup of CK. The abnormal expression of CK-19 is associated with different kinds of cancers, including breast cancer, hepatocellular carcinoma, gallbladder cancer andpapillary thyroid carcinoma, by interaction with other biomarkers. CK-19 maintains the function of keratin filament assembly by the phosphorylation of Ser-35 [28]. CK-19 also plays a key role in HCC with the expression of invasion-related/metastasis-related markers (VASP, LAMB1, PDGFRA), biliary marker (CD133, GSTP1, JAG1) and members of microRNA family 200, especially in PDGFRA-LAMININ B1-CK-19 cascade [29,30]. In lung cancer, CK-19 intracellularly binds to HER2 to promote HER2 activation [31]. Additionally, CK-19 is related to V600E in papillary thyroid carcinoma [32]. CK-19 is mainly expressed in the pancreatic ductal epithelium. During pancreatic morphogenesis, CK-19 is expressed in duct-like pancreatic precursor cells, including pancreatic islet cells between 12 and 16 weeks of fetal development, which will develop into exocrine acinar and islet beta-cell with CK-19 negative expression [10,11,33]. Furthermore, PNET with biomarkers of pancreatic precursor cells presents poorly differentiated grade and malignant behavior, while those that do not express biomarkers show better prognoses [6]. Hence, CK-19 can be a key indicator for PNET prognosis.
With the intensive awareness of patients for PNET and improved technology of computer tomography, the detection of PNET is increasing. Although PNET has a relatively good survival rate due to its indolent behavior [34], it is difficult to predict localized behavior, metastasis and recurrence. The prediction of malignant behavior and prognosis has not been successful or efficient using the WHO-2010 classification or the European Neuroendocrine Tumor Society (ENETS) staging system [35,36]. Therefore, it is urgent to find a new biomarker to evaluate the prognosis of PNET. A large number of studies have focused on CK-19 to evaluate its relationship to the prognosis of PNET. But some debates remain.
Our meta-analysis of the 10 selected studies revealed a total of 856 tumor samples from 854 PNET patients were included, of which, 428 showed positive CK-19 expression. The results indicated that CK-19 expression was significantly associated with 3-year and 5-year overall survival but not 1-year. Compared with ductal adenocarcinoma in the pancreases, PNET patients' survival rate is apparently longer. Therefore, during the 1 st or even 3 rd years, it is not very important to evaluate the PNET patients' survival. This may partially explain why there is no significant correlation between the CK-19 expression and 1-year overall survival. We also performed a meta-regression analysis to detect the effect of confounding factors on the impact of CK-19 on survival outcome. Since most data regarding demographic and clinicopathological features were not available from all original articles, only functional status was examined in the meta-regression analysis. It had no significant effect on the relationship between CK-19 expression and PNET survival outcome. The above results suggest that positive CK-19 expression is an indicator of reduced survival rate of PNET.
For the differentiation grade, WHO-2004 classification system was applied previously [37]. In WHO-2004, PNETs were categorized as "benign behavior", "uncertain behavior" and "malignant" (including well-differentiated endocrine carcinomas and poorly differentiated endocrine carcinomas) based on the combination of grade, stage and adjunct prognosticators (vascular and perineural invasion). But here, grade and stage of PNET were separated out. Grade was evaluated by WHO-2010 classification by mitotic activity and Ki-67 index [4]. ENETS and American Joint Committee on Cancer (AJCC) were used to assess the stage [38,39]. In this meta-analysis, results presented that positive CK-19 expression was correlated with malignant tumors in WHO-2004 and grade 3 in WHO-2010, suggesting positive CK-19 expression could predict advanced differentiation grade.
Whether 2 cm belonged to large or small size groups, there was a significant relationship between positive CK-19 expression and large tumor size, indicating a poor prognosis. Moreover, it was found that CK-19 expression was significantly correlated with vascular invasion, lymph node and liver metastasis. That's to say, PNET with positive CK-19 is more metastatic and adjacently invasive. Similarly, CK-19 plays an important role in the invasion of HCC [29,30].
However, some limitations should be elaborated. First, heterogeneity does exist in this meta-analysis because of different basic characteristic among the enrolled studies. A randomeffects model was used to weaken the unfavorable effect of variation among studies. Second, some relevant data were extracted from the studies indirectly, which could lead to unavoidable bias. For example, some overall survival data is from Kaplan-Meier survival. Third, relevant data from related studies was limited because of insufficient or incompatible statistical methods in these papers. Two studies described distant metastasis or tumor metastasis or metastasis at diagnosis without distinguishing between lymph node and liver metastasis [14,18]. The data of lymph node metastasis was inconsistent and was excluded in one study [14]. Fourth, different antibodies and definition standards were used to detect CK-19 expression. The unconformity could also result in inevitable heterogeneity. Fifth, subgroup analysis was not applicable because of the relatively small sample size. Finally, only studies published in English were enrolled. Therefore, a potential ethnic demographic bias may exist.
In this meta-analysis, the relation between the CK-19 expression and overall survival (1-year, 3-year and 5-year) and clinicopathological features, such as tumor size, differentiation grade, vascular and perineural invasion, lymph node and liver metastasis, was studied to assess the impact of CK-19 expression on PNET prognosis. We conclude that CK-19 expression is significantly correlated with poor overall survival and is useful for diagnosing clinicopathology. CK-19 can predict the prognosis of PNET patients.