Identification of research trends concerning application of stent implantation in the treatment of pancreatic diseases by quantitative and biclustering analysis: a bibliometric analysis

Data obtaining

All publications came from PubMed without the restrictions of languages. The PubMed database has been used to retrieve data in some of the biomedical research (Le et al., 2019; Le, 2019). PubMed is chosen not only because of the authority and breadth of the literature, but also the normative nature of the Medical Subject Headings (MeSH) keywords, more importantly. MeSH has been applied to index and catalog articles in PubMed. In our study, we collected literature on the application of stents in pancreatic diseases on May 15th, 2018, in order to ensure more current research results. Our research strategy was as follows: (“stents”[MeSH Terms] OR “stents”[All Fields] OR “stent”[All Fields]) AND ((“pancreas”[MeSH Terms] OR “pancreas”[All Fields] OR “pancreatic”[All Fields]) OR (“pancreatitis”[MeSH Terms] OR “pancreatitis”[All Fields])) and “2018/05/15” [PDAT]. Publication trends were retrieved from GoPubMed (http://www.gopubmed.org) (Doms & Schroeder, 2005).

Literature screening criteria

If a paper concerning application of stents in pancreatic diseases was an original article, we would accept the literature. Meanwhile, media coverage and science briefings were excluded. Furthermore, two researchers separately examined the papers by title, abstract and full text. One researcher excluded 20 articles, and the other researcher excluded 19 articles. And the agreement was 95%, which suggested a strong correlation (Mandrekar, 2011). Finally, title, author, institution, country, publication year and MeSH terms of available articles were saved into a new file in XML.

Data extraction and analysis

XML file was imported into BICOMB for data extraction (Dehdarirad, Villarroya & Barrios, 2014; Hu & Zhang, 2015; Lei et al., 2008). And authors, journals and the frequency ranking of MeSH terms were determined (Le & Ou, 2016; Le, Ho & Ou, 2019). According to the H index, the terms were first sorted in descending order of terms. Then the high-frequency major MeSH terms were identified if a term with frequency greater than or equal to its sequence number (h) from the list of high frequency terms, and h was the threshold for intercepting high frequency terms. Then, the relationships between the high-frequency major MeSH terms and the source literature were determined utilizing biclustering analysis. Also, a binary matrix was produced using the source literature set generated by BICOMB and the high-frequency MeSH terms as columns and rows.

Cluster analysis

Then, double clusters and visual analysis were performed by “gCLUTO” version 1.0 software. “gCLUTO” is a graphical cluster toolkit and graphical front-end of the “CLUTO” data clustering library (Karypis Lab, 2014; Li et al., 2015). The clustering analysis was employed to assess the high-frequency MeSH terms. The clustering method was used to repeat the bisection, cosine as the similarity function, and I2 as the clustering criterion function. By use of different numbers of clusters, two clusters were performed to differentiate the first-rank number of clusters. And the visualizations of high frequency and high-frequency bifocal results with MeSH article were constructed by use of Alpine and Matrix. By means of the semantic corrections between the MeSH terms and the content of typical articles in every group, the relevant topics on the application of stents in pancreatic diseases were obtained. And we made a visualized matrix biclustering of high-frequent major MeSH terms and PubMed Unique Identifiers (PMIDs) of articles on the application of stents in pancreatic diseases.

Strategic diagram analysis

A two-dimensional table is depicted by plotting themes based on centricity and density. The X-axis stands for centrality, namely the closeness between keywords within this category and those within other categories. It indicates the degree of interaction between a subject area and other subject areas. The Y-axis represents density, namely the closeness of the keywords within each category. And it indicates that this category maintains and develops its own capabilities (Callon, Courtial & Laville, 1991). The above eight categories were assigned to the four quadrants based on the results of the cluster analysis. In addition, excel was utilized to generate strategic diagram.

Social network analysis

The high frequency MeSH terms co-occurrence matrix was imported into the Ucinet 6.0 (Analytic Technologies Co., Lexington, KY, USA) software. And the social network analysis method was utilized to analyze the subject and knowledge structure of the application of stents in pancreatic diseases. Then the high-frequency MeSH term network was visualized by NetDraw 2.084 software. The nodes represent MeSH terms, and the links stand for the co-occurrence frequency of these terms. And we measured the degree, betweenness and closeness centralities of every node. At the same time, author relationship network was constructed by above methods.

Overall evaluation

Based on GoPubMed, we obtained the literature information according to the search strategy: stents [MeSH] and pancreas [MeSH] or “pancreatic diseases” [MeSH]. Figure 1A depicts the distribution of the publication year of corresponding papers. The first article was published in 1977. As time went by, the volume of publications increased year by year. By 2015, it had a downward trend. Figure 1B shows the volume of paper outputs concerning the application of stents in pancreatic diseases in the first 20 countries. And the map was generated by an online website (pixelmap.amcharts.com). The number in the map is the quantity of associated publications for every country or region. The United States stands first with 1,167 publications. Furthermore, we summarized the annual distribution of MeSH terms associated with the application of stents in pancreatic diseases (Fig. 1C). Different colors represent different highly frequent major MeSH terms. We found that these MeSH terms had roughly the same development trend every year from 1985-2018, indicating that they had close associations. As shown in Table 1, the top 29 authors with a cumulative percentage of 27.9483 are listed. “Baron TH” (84, 2.0468%), “Kahaleh M” (81, 1.9737%) and “Isayama H” (65, 1.5838%) are the top three authors. From 1977 to 2018, the 25 most active journals published publications on the application of stents in pancreatic diseases account for 49.92% of all publications. Table 2 demonstrates the 25 most productive journals, as the core journals in the research fields on the application of stents in pancreatic diseases under Bradford’s Law. “Gastrointestinal endoscopy”, “Endoscopy”, “World journal of gastroenterology” are the most active three journals.

High-frequent major MeSH terms

A total of 4,087 articles were selected until May 15th, 2018. Eighty-three high-frequency MeSH terms were extracted from the listed publications, with a cumulative percentage of 57.5291 (Table 3). “Stents” (2238, 3.8488%), “Treatment Outcome” (1038, 1.7851%) and “Retrospective Studies” (758, 1.3036%) are the top three MeSH terms.

Cluster analysis

The double cluster analysis results were visualized into mountain visualization and hierarchical cluster tree. In the mountain visualization, the peak and matrix visualizations express the high-frequency MeSH terms. Each cluster represents a peak marked by cluster number 0–7 in Fig. 2, and the related clusters are described according to the volume, color and height of the peaks. The volume of the peak is directly proportional to the number of MeSH terms in the cluster. Meanwhile, the internal standard deviation of a cluster object is represented by the color of the peak. Blue stands for the high deviation and red represents the low deviation. The peak is the position relative to the other clusters. The closer the distance between the two peaks, the higher the similarity between the two clusters. The height and similarity of each cluster are proportional to each other.

In Fig. 3, the row labels represent high-frequency MeSH terms, and the PMIDs locate the column labels at the right and bottom of the matrix. The color of each grid suggests the frequency of appearance in a paper. The darker the red, the greater the frequency. Eighty-three high-frequency major MeSH terms are distinguished into eight clusters in matrix visualization. The top and left of the hierarchical tree respectively indicate the relationships among the major MeSH terms and the associations among the papers. Meanwhile, the corresponding article is obviously shown for each high frequency MeSH terms in each cluster.

Strategic diagram

The centrality and density of the 8 clusters are listed in Table 4. The details of MeSH terms and clusters are shown in Table 5. In Fig. 4, x-axis represents the centrality, and y-axis stands for the density on the strategy diagram. The four quadrants clockwise from the upper right corner express the first quadrant, the second quadrant, the third quadrant and the fourth quadrant. As shown in Fig. 4A, the clusters in the first quadrant are suggested to be central topics in the network (due to their strong connection with other clusters) and have intense internal relationships (due to high degree of development). The clusters in the second quadrant are peripheral, however, already well-developed topic. The clusters in the third quadrant are both peripheral and undeveloped. The clusters in the fourth quadrant are central and undeveloped, but they are becoming mature to some extent (Indolfi et al., 2016).

Figure 4B depicts that cluster 1 and cluster 3 are located in the first quadrant, suggesting that the cluster densities and centrality degrees are all high, that is to say, the MeSH terms in cluster 1 and cluster 3 are closely linked, and research tends to be well-developed. And the orientation is high, indicating that it is at the center of the research network. Cluster 4 and 7 are located in the second quadrant with high density and low centrality, indicating that internal links are close together with a clear topic. The research on this topic is shown to be relatively well-developed, with little correlation with other research. Cluster 0, 2 and 6 are located in the third quadrant, with low density and centrality. MeSH terms of Cluster 0, 2 and 6 are the margins of the entire field. The internal structure is relatively loose and research is yet developed. Cluster 5 is located in the fourth quadrant with low density and high centrality, indicating that it has close relations with other research. However, the research is not found to be well-developed. The research on this topic has potential value, and is now in the exploratory stage; however, more research is required.

Social network analysis

As shown in Fig. 5A, we constructed the author relationship network. There are 29 nodes which represent 29 authors. The size and location of nodes suggests the decisive role of an author. Links indicate the connection between two authors. In Fig. 5A, the node “Itoi T” was the largest one, which was located in the center of the social network, followed by “Isayama H” and “Sasaki T”. Therefore, these authors could play a critical role in the field of the application of stents in pancreatic diseases. Their articles could represent the maturity of the research area and hot spots. Figure 5B depicts that the network relationships among 83 high-frequent major MeSH terms. The size of nodes suggests the centrality of high-frequent major MeSH terms. In the meanwhile, the thickness of the lines demonstrates the co-occurrence frequency of keywords pairs.

Discussion

We took advantage of GoPubMed to analyze the publication trends in the field of pancreatic stents. Before 2015, the volume of relevant publications was continuously rising and relative research interest was fluctuating rising. However, beginning with 2015, the volume of publications and relative research interest both showed a downward trend, which suggests that the researchers’ interest have shifted and more innovation needs to be explored in the pancreatic stents. In addition, we also focused on the countries and author of research outputs. The United States, Japan and Germany remain to be the countries with the largest number of publications on pancreatic stents. The results indicated the developed countries occupied main position in the field. After measuring the top 29 authors on pancreatic stents, we made the author relationship network. The authors in the field have close cooperation, emphasizing the importance of cooperation. By paying attention to these authors, we would have a general understanding of the research direction and hotspots in this field. In order to further track research trends, journals are also the focus of attention. Therefore, we measured the most active journals, considering as the central journals in the relevant fields such as Gastrointestinal endoscopy, Endoscopy, World journal of gastroenterology. The high-frequency MeSH terms may reflect the research hot spots. The 83 high-frequency major MeSH terms were achieved by the co-occurrence in the same paper, which represented the research content in the field. Yearly distribution trends on different MeSH terms had the same fluctuating trend.

Eighty-three hot major MeSH terms were clustered into eight clusters. The network revealed that these MeSH terms existed complex relationship network. Endoscopic retrograde ERCP in acute and chronic pancreatitis and imaging methods as an auxiliary method of stent placement are located in the second quadrant. Cluster 1 and 3 are located in the first quadrant, including the complications of stent placement in bile duct neoplasms and pancreatic neoplasms and stents for the prevention of pancreatic fistula following pancreaticoduodenectomy. The two topics are current research center and hot topics for pancreatic stents. And cluster 0, 2, 6 are located in the third quadrant, which suggesting that the three topics are at the margin and not yet mature, including stents placement in pancreatic neoplasms, the postoperative complications after stent placement therapy such as pancreaticoduodenectomy and pancreatic ducts changes for patients in chronic pancreatitis. In the meanwhile, complications such as pancreatitis associated with stent implantation could have potential research value in the fourth quadrant, which are the research center, however, not yet mature. Therefore, the topic could become potential hotspots in the future science research. Then the 8 topics would be introduced respectively.

Stents placement in pancreatic neoplasms

Increasing numbers of patients with resectable pancreatic neoplasms are receiving neoadjuvant therapy such as stents placement. Tumor growth in pancreatic neoplasms often leads to invasion of other organs and biliary obstruction, resulting in repeated stent placement (Shi et al., 2019). The self-expandable metal stents possess effectiveness and safety in achieving durable biliary drainage for patients with pancreatic neoplasms (Aadam et al., 2012; Van der Horst et al., 2014). For example, the covered self-expanding metal stents is used for the therapy of biliary tract hemorrhage induced by advanced pancreatic cancer-induced portal biliary disease (Kim et al., 2016).

The complications of stent placement in bile duct neoplasms and pancreatic neoplasms

As for pancreatic neoplasms, preoperative biliary drainage (PBD) promotes complications compared with surgery without PBD. The result could be associated with the plastic stents utilized. However, metal stents might decrease the PBD-associated complications (Tol et al., 2016). It has been confirmed that biliary stents could remarkably increase liver volume in both hilar and distal bile duct neoplasms (Lee et al., 2014). Endoscopic retrograde biliary drainage of metal bile duct stents are widely used for biliary obstruction. The application of bile duct stents has also led to an increasing number of complications. The main complications of pancreatic stents include migration, stent occlusion, and pancreatic ductal changes (ASGE Technology Assessment Committee et al., 2013).

Postoperative complications after stent placement such as pancreaticoduodenectomy

Pancreatic fistula is a leading complication following pancreaticoduodenectomy. Pessaux et al. (2011) have reported that external pancreatic duct stent reduces pancreatic fistula rate following pancreaticoduodenectomy. Obstructive jaundice is one of the known risk factors for treatment failure following hepatectomy for patients with hilar cholangiocarcinoma. In palliative care, self-expanding metal stents have a rapid reduction in bile duct pressure and reduce complication rates, while providing patients with adequate and rapid biliary drainage (Grünhagen et al., 2013).

Stents for the prevention of pancreatic fistula following pancreaticoduodenectomy

It is necessary to prevent pancreatic fistula after pancreaticoduodenectomy in stent placement. The incidence of pancreatic fistula in patients undergoing pancreaticoduodenectomy is as high as 56% and is considered to be a main factor on morbidity and mortality in patients following pancreaticoduodenectomy (Dong et al., 2016; Brown et al., 2014). And external duct stents placement could reduce the occurrence for clinically relevant postoperative pancreatic fistula (Motoi et al., 2012).

Prophylactic pancreatic duct stent can reduce the incidence of post-ERCP pancreatitis (PEP) and complications such as pancreatitis associated with stent implantation

Endoscopic retrograde ERCP was first introduced in 1968. As a diagnostic tool, it was used to assess the disorders of pancreas (Riff & Chandrasekhara, 2016). As a most common complication of ERCP, the incidence of PEP is still as high as 15% in high-risk cases (Elmunzer, 2017). A small number of patients could develop severe pancreatitis. Pancreatitis is a common and serious complication for endoscopic retrograde ERCP. Prevention of pancreatitis after ERCP remains the focus of clinical and research. Relevant strategies could decrease the occurrence of post-ERCP pancreatitis including patient selection, risk stratification, surgical techniques, pancreatic stenting, and drug prophylaxis. Placement of the pancreatic stent is a relatively new and increasingly popular method of reducing the risk of pancreatitis after ERCP (Shi et al., 2014). Prophylactic pancreatic stent placement decreases the incidence of pancreatitis after ERCP in high risk patients and reduces the severity of this condition (Freeman, 2007). In summary, placement for pancreatic duct stent decreases the incidence of pancreatitis (Sofuni et al., 2011).

Pancreatic duct changes in patients with chronic pancreatitis

It is essential to prevent pancreatic duct changes such as pancreatic leakage or pancreatic duct patency after pancreaticoduodenectomy. In duct-to-mucosa anastomosis, placement of the stent could be an effective mean of dilating the pancreatic duct  (Téllez-Aviña et al., 2018). Pancreatic stent is used to improve painful, obstructive chronic pancreatitis (Samuelson et al., 2016).

Stent placement in endoscopic pancreatic pseudocyst drainage

Pancreatic pseudocyst is one of the common local complications of acute and chronic pancreatitis. And endoscopic pancreatic pseudocyst drainage has been widely applied in the treatment of pancreatic pseudocysts  (Madder et al., 2016). Endoscopic drainage has the advantages of small invasiveness, short recovery time, low cost and low complication rate (Shah et al., 2015), like interventional endoscopic ultrasonography has been increasingly used to manage pseudocyst formation (Vilmann et al., 2015). As an example, Varadarajulu et al. (2013) have found that, compared with surgical bladder anastomosis, patients with endoscopy pancreatic pseudocyst drainage experience rarely recurrence of pseudocyst during follow-up.