Trends analysis of cancer topic of Cochrane systematic reviews: a bibliometric analysis

Purpose: To comprehensively analyze the scientic outputs of cancer topic of Cochrane systematic reviews (Cancer-CSR). Patients and methods: Cochrane Database of Systematic Review and Web of Science Core Collection were retrieved limited from Jan. 1, 2009 to Dec. 12, 2018. CiteSpace IV and Excel 2018 were applied to analyze and visualize the literature information. Results: Ultimately, 607 Cancer-CSR were retrieved, 32 countries, 179 institutions and 260 authors involved. The number of publications in Cancer-CSR has been increasing over the past decades (25(2009)-77(2018)). UK, USA, Canada, Australia, and Germany worked closely with other countries, especially UK (n=361) has taken the lead in this eld. The top 10 contributive institutions, which were almost came from developed countries, collaborated closely with other institutions. Cochrane Database Syst Rev, C Hdb Sys and J Clin Oncol were the top three journal/book with the highest co-citation. The top three co-cited references were the two different version of Cochrane handbook for systematic reviews and the guidelines of Review Manager. The biggest cluster of keywords “cytoreductive surgery (CRS)” and the latest clusters “visual inspection” and “non-steroidal anti-inammatory drug” were the most promising hotspots. Conclusions: Cancer-CSR has been increasing. Most of the reviews were came from the developed countries as well as the institutes in these countries. The knowledge base of were the methodology studies of systematic review, epidemiological data of cancer, and the reporting guideline of systematic reviews. The adjuvant therapy combined CRS, the screening of skin cancer and the management of cancer-related pain were the hotspots.


Introduction
With the development of evidence based medicine, the systematic review has been gradually considered as the high quality evidence (1) . In 1993, British physician and epidemiologist Archie Cochrane and his colleagues developed the Cochrane Collaboration, which represents the watershed movement responsible for the biggest advances in systematic review methodology (2,3) . Two years later, the Cochrane Database of Systematic Reviews (Cochrane Database Syst Rev) launched, a database of systematic reviews and meta-analyses which summarize and interpret the results of medical research, as well as contains different types of high-quality, independent evidence to inform healthcare decision-making (4)(5)(6) , and the 2017 Impact Factor of which is 6.754 (7) . A Cochrane systematic review is a systematic review of research in health care and health policy that is published in the Cochrane Database Syst Rev, and has been an important resource of guideline making (6,8) The scienti c research of cancer in Cochrane Database Syst Rev has never been stopped. Cancer affects all of humankind, from a global perspective, which is the second leading cause of death worldwide. By 2018, nearly one in six deaths was caused by cancer (9.5 million) (9,10) . It is worth noting that the burden of chronic diseases is disproportionately increasing in low-income countries and populations, and it is reported that nearly 80% of non-communicable disease-related deaths in low-and middle-income countries are caused by chronic diseases (11) . The total annual economic cost of cancer care in the United States in 2017 is 147370 million, up 15.4 percent compared to 2010, according to Cancer Prevalence and Cost of Care Projections in NIH (12) . Giving to the grave situation, cancer threats the whole humankind and puts a huge economic burden on every country.
In order to assess trends in research activities of the cancer topic of Cochrane systematic reviews (Cancer-CSR), bibliometric analysis has been used, which is a quantitative analysis combining mathematics and statistics (13) . Bibliometric analysis focuses on the bibliometric characteristics of research in a particular eld, helping investigators grasp the development priorities and trends in the eld and guiding their follow-up work (14) .
This study performed a bibliometric analysis of the Cancer-CSR to establish the current knowledge visualization map, analyze the characteristics and trends of research topics in this eld, compare the composition of topics in different dimensions, and nd the research hotspots of keywords cluster analysis.

Data source and collection
Records found in Cancer-CSR were searched through the Web of Science Core Collection (WoSCC) database from Jan. 1, 2009 to Dec. 12, 2018. All retrievals were done within one day in order to avoid the bias caused by the updating of the daily database. The data from WoSCC were checked with the records in Cochrane Database Syst Rev by two reviewers (Yang KL and Gao Y) independently.

Tools and methods
EndNote X8 (Thomson Reuters (Scienti c) LLC Philadelphia, PA, US) was applied to manage the retrieved data, and CiteSpace IV and Excel (Microsoft, 2016) were applied to analyze and visualize the literature information.
CiteSpace IV , developed by Prof. Chaomei Chen, is a Java application that combines information visualization methods, bibliometrics, and data mining algorithms in an interactive visualization tool for extracting patterns from citation data, which is to facilitate the analysis of emerging trends in a knowledge domain (15,16) . In this study, the individual network was derived from the 50 most highly cited papers in a one-year slice, and the cluster labels were showed by the log-likelihood ratio (LLR). Clusters of the same color represent co-citations or keywords made within the same time slice. The size of the node can re ect the number of the node is cited or appeared; the tree ring history of the node represents the number of papers published in different years. The wider the ring is in a certain year, the more frequently it is cited or appeared in the corresponding year; The larger the size of the cluster (that is, the more members contained in the cluster), the smaller the number of the labels (17) .

Data analysis and Visualization
The retrieved data were analyzed for the annual growth, co-citation analysis, and co-occurrence of countries, institutes, and keywords. The following index was focused on analysis: (1) Centrality: The centrality of a node is a graphical theoretical property quantifying the importance of a node's position in the network. A commonly used centrality metric is the "betweenness centrality" (18) , which measures the percentage of the shortest path in the network to which a given node belongs. The high centrality node is around by the purple circle in the visualization.
(2) Visually salient nodes (16) : a. Landmark node: the landmark node can be rendered by different visual space attributes such as size, height, or volume, which are commonly used in network visualization; b. Pivot node: the pivot node, which joints between different networks, is either the common nodes shared by two networks or the gateway nodes that are connected by internetwork links. According to the content of which, the included reviews were classi ed into different topic categories.

The basic characteristics
"Gastroenterology & hepatology"(n=111(18.3%)), "Gynaecology" (n=99(16.3%)), and "Blood disorders" (n=62(10.2%)) were the top three category among them. In addition, on the basis of the "review type" of Cochrane Library, most of the included reviews focused on Intervention, and only two of the included records were overview (Table 1).  The selected reviews encompassed 32 countries. In the light of the list of top 10 countries launching Cochrane review in cancer eld (Table 2), the publication of UK took up the most number (361), followed by the USA (97), Australia (75), Germany (71) and Netherlands (67). Among these 10 countries, the centrality of China (publications=52, centrality=0.05) was only a quarter of the centrality of Canada (publications=51, centrality=0.21) which had nearly the same number of publications as China. As was shown in Fig. 2A, the tree ring history of UK was the biggest one and was the richest in color. The nodes around by purple circle were UK, Canada, USA, Australia, and Germany, and the latest countries rstly published in 2018 were Austria, and Egypt.

Distribution of countries and institutes
The total of 178 institutes has involved in the co-occurrence network of institutes, as indicated in the Table 3. Among the top 10 institutes publishing reviews, six were located in UK, one in USA, one in Germany, one in Holland, and one in China. Newcastle University contributed most (52(6.1%)), with the biggest node in Fig. 2B, followed by Oxford University (39(4.5%)), University Hospital of Cologne (28(3.3%)), Royal United Hospital (25(2.9%)), Churchill Hospital (22(2.6%)) and Manchester University (22(2.6%)). The nodes of all these six institutes were around by purple circle, and the centrality of Oxford University (centrality=0.38) was the highest.

Analysis of co-cited journals and references
There were 138 co-cited journals included. Among the top 10 co-cited journals/books, seven journals/books were from UK and three were from USA (Table 4)  As for the co-cited references, 672 references were divided into seven clusters (Fig. 4A). Due to the limitation of the software, the system automatically displayed the clustering with more than 10 members, while the clustering that did not meet this condition would not be displayed. The clusters were listed from 2007 to 2012: "psychosocial intervention" (#23, n=5), "active treatment" (#8, n=16), "cancer survivor" (#0, n=69), "hiv-1 infection" (#12, n=13), "liver metastases" (#2, n=39), "following computed tomography" (#15, n=9), and "cancer pain" (#3, n=35). The landmark node was Higgins JPT, 2011 and the pivot nodes were labeled in the gure. The top 10 co-cited references were listed in Table 5. The references rstly cocited in 2018 were listed in Fig. 4B.

General information
In this study, the bibliometric analysis of Cancer-CSR was performed. A total of 607 publications in Cancer-CSR were retrieved using the online WoSCC database, covering many elds in cancer, such as gastrointestinal tumors (19) , gynecological oncology (20) , haematological malignancies (21) , palliative and supportive care (22) , and so on. Compared with other disciplines, the number of publications of cancer was higher than that in other topics in Cochrane Database Syst Rev (7) . This might be due to the Cochrane Review Group in Cochrane Library, which contained a variety of cancers, such as colorectal cancer group, breast cancer group, lung cancer group and so on (23) . The Cochrane Review Group provides authors with methods and editorial support to prepare for the Cochrane review and manages the editing process. Since 2018, the Cochrane Review Group has been organized into eight Cochrane Networks to generate highquality systematic reviews e ciently and timely, addressing the most important research questions to decision makers (8) .
UK, USA, Canada, Australia, and Germany were the countries with high centrality, which worked closely with other countries. UK ranked rst in productivity, with the highest centrality, containing more than half of the top 10 contributive institutes. Although the publications of USA were only one-third of UK, the centrality was almost the same. China and Brazil were the only two developing countries among the top 10 contributive countries. However, as for the centrality, China was the last but one among the top 10 contributive countries, which indicated that the importance of the studies we Chinese researchers involved in was not high. On the other hand, the overall methodological and reporting quality of systematic reviews by authors from China were similar to those from the USA (24) . So, the possible reason may be that the topics of the Chinese Cancer-CSR studies are not attractive and innovative enough. Egypt and Austria were the rst time to publish the reviews in Cancer-CSR during latest ten years, which focused more on the chemotherapy and multimodality therapy in cancer eld (25,26) .
With respect to the institutes involved, Newcastle University, one of the Ivy League schools and has the top medical school in UK, which was founded in 1834 as the College of medicine and surgery, was the most productive institution. The reviews of Newcastle University focused on the surgical treatment (27) , chemotherapy (28) and prognosis (29) of gynecological cancer. The most cited review was the evaluation of the e cacy, safety, and cost-effectiveness of primary cytoreductive surgery in patients with advanced epithelial ovarian cancer surgery (30) , published in 2011. The top 10 contributive institutions collaborated closely with other institutions, except for the Sichuan University and the University of Nottingham.
Checking the incidence rate of the districts of top 10 institutes and 10 countries, we also found that the incidence of all kinds of cancer in these districts was above the world average level (197.9) (31) .

Citation information
Among the top 10 co-citied journals/books, different journals usually favored publication on different scienti c subjects. For example, J Clin Oncol, Cancer, Ann Oncol and Eur J Cancer preferred every respect of cancer, from epidemiology, carcinogenesis and biology through to innovations in cancer treatment and patient care; Lancet preferred surgical oncology; Cochrane Database Syst Rev, BMJ, and N Engl J Med were comprehensive journals; Stat Med focused on statistical the methods to a particular medical problem, and C Hdb Sys was the most signi cant book to all the researchers who intended to conduct systematic reviews, which is the o cial guide that describes in detail the process of preparing and maintaining Cochrane systematic reviews on the effects of healthcare interventions The co-cited references were usually considered as the basics of the subject. Being-in Top 10 co-cited references, three versions of C Hdb Sys (Higgins JPT., 2011; 2009; 2008) were included, owing to the Cochrane collaboration has driven the greatest advancement in system review methods, which has brought together more than 13,000 members and over 50,000 supporters come from more than 130 countries dedicated to systematic review (8) . The second most co-cited reference was about Review Manager (RevMan) (32) , the most frequently used tool in meta-analysis, which facilitated preparation of protocols and full reviews, including characteristics of studies, comparison tables, study data and so on, and was also appropriate to the reviews of diagnostic test accuracy studies, reviews of studies of methodology and overviews of reviews. The rest top 10 co-cited references including methodology studies (33,34) , epidemiological data of cancer (35) , and the reporting guideline of systematic reviews (36) .
The latest co-cited references were focused on the diagnosis and treatment of skin cancer in adults (37)(38)(39) .
All these co-cited references constituted the knowledge base of this eld.

Research hotspots
The cluster of keywords could be used for detecting research hotspots, and monitoring the study frontiers transitions in a certain eld (40) Here, we listed three hotspots of Cancer-CSR as follows: #0 Cytoreductive surgery: Cytoreductive surgery (CRS) is a surgery which removes all visible lesions as much as possible and used to reduce tumor burden of patients and prolong survival, which has been applied in many kinds of cancer (41)(42)(43) . Because of the unsatisfactory effects of traditional treatment, better treatment regimens were needed (44) . The included studies about CRS covered the ovarian cancer (45)(46)(47) , colorectal cancer (48) , neuroblastoma (49) , breast cancer (50) and lung cancer (51) , which mostly focused on the different adjuvant therapy, different chemotherapeutic drugs and different dose of chemotherapy. However, the evidence found was insu cient in these aspects, so new clinical trials are needed to solve these problems. #7 visual inspection: Nonmelanoma of skin accounted for 5.8% of all new cancer cases in 2018 as well as melanoma of skin accounted for 1.6%. There were an estimated 1,000,000 new cases and 65,000 deaths worldwide, and the male incidence was about twice that of women. NMSC was the most commonly diagnosed cancer in North America, with Australia and New Zealand being the highest incidence rates countries, which partly because of the higher risk, but also because the detection rate of skin cancer (especially NMSC) was rising (52) . One burgeoning approach that has impacted skin cancer detection was re ectance confocal microscopy (RCM), which noninvasively shows nuclear and cellularlevel morphology in human skin in vivo (53,54) . Compared with the dermoscopy, RCM had similar sensitivity but the speci city was two times superior, and it could reduce the number of biopsies of benign lesions as a second-level examination following dermoscopy (55)(56)(57) . Nevertheless, there were some pitfalls. For example, the differentiation of melanocytes from langerhans cells in melanocytic lesions lacks speci city, the misdiagnosis due to speci c lesion or missed detection because of sampling error, as well as the clinical experience of the doctors (58) . Considering the pitfalls and the large size of current microscopes, RCM has been constantly improving. On the one hand, the continuous improvement of technique focused on a smaller, lower-cost and easier-to-use microscopes, which was indispensable for dissemination (59) . On the other hand, training the ability of dermatologists and pathologists to read and interpret RCM results at the bedside or remotely was the key to widespread acceptance and adoption of RCM imaging (58) . What's more, there were four records in Cancer-CSR concentrated on the applying of RCM for diagnosing skin cancer in adults with any suspicious lesion (60-63) , which were all published in 2018.
#10 Non-steroidal anti-in ammatory drug: Increased survival rates for cancer patients, coupled with increasing complexity of cancer and the introduction of new therapies, made pain treatments more challenging (64,65) . Non-steroidal anti-in ammatory drugs (NSAIDs) were widely used to treat postoperative fever and surgical pain. What's more, there was increasing evidence that non-steroidal antiin ammatory drugs have anticancer activity which could reduce the risk of cancer (66, 67) . However, there was no su cient evidence of the included records about reducing cancer-related pain in children and adolescents (68) , and no high-quality evidence to support the use of NSAIDs alone or in combination with opioids for the three step WHO cancer pain ladder (69) .

Strengths and limitations
To the best of our knowledge, this study is the rst bibliometric analysis the Cancer-CSR. Although all the reviews were written in English causing incomplete understandings and not in detail, the process of data analysis was still relatively objective. The research trends and hot spots of Cancer-CSR re ected that in systematic reviews in the cancer eld to some extent, but it was not comprehensive enough, for the Cochrane collaboration had greater scrutiny for systematic reviews, so there were not a lot of publications.

Conclusions
Cancer research is increasing in Cochrane Database Syst Rev. Most of the reviews were came from the developed countries as well as the institutes in these countries, like UK and USA. Co-cited references including C Hdb Sys and the guidelines of RevMan constituted the knowledge base of Cancer-CSR in the aspect of the methodology studies of systematic review, epidemiological data of cancer, and the reporting guideline of systematic reviews. What's more, the adjuvant therapy combined CRS, the screening of skin cancer and the management of cancer-related pain were the hotspots which needed more su cient evidences to support the clinical decision making.

Abbreviations
Cancer-CSR=cancer topic of Cochrane systematic reviews; CRS=cytoreductive surgery; LLR=log-likelihood ratio; WoSCC=Web of Science Core Collection.

Declarations
Ethics approval and consent to participate: Not applicable Consent for publication: Not applicable Availability of data and material: The datasets used and analyzed during the current study are available from the corresponding author upon reasonable requests.
Competing interests: The authors declare that they have no competing interests.  The analysis of references. (A) The co-citation map of references from reviews in Cancer-CSR, (B) The timezone view of co-cited references from reviews in Cancer-CSR