Two Decades of Research in Artificial Pancreas: Insights from a Bibliometric Analysis

Abstract

Background:

Artificial pancreas is a well-known research topic devoted to achieving better glycemic outcomes that has been attracting increasing attention over the years. However, there is a lack of systematic, chronological, and synthesizing studies that show the background of the knowledge generation in this field. This study implements a bibliometric analysis to recognize the main documents, type of publications, research categories, countries, keywords, organizations, and authors related to this topic.

Methods:

Web of Science core collection database was accessed from 2000 to 2020 in order to select high-quality scientific documents based on a specific search query. Bibexcel, MS Excel, Power BI, R-Studio, VOSviewer, and CorText software were used for a descriptive and network analysis based on the local database obtained. Bibliometric parameters as the h-index, frequencies, co-authorship and co-ocurrences were computed.

Results:

A total of 756 documents were included that show a growing scientific production on this topic with an increasing contribution from engineering. Outstanding authors, organizations, and countries were identified. An analysis of trends in research was conducted according to the scientific categories of the Web of Science database to identify the main research interests of the last 2 decades and the emerging areas with greater prominence in the coming years. A keyword network analysis allowed to identify the main stages in the development of the AP research over time.

Conclusions:

Results reveal a comprehensive background of the knowledge generation for the AP topic during the last 2 decades, which has been strengthened with international collaborations and a remarkable interdisciplinarity between endocrinology and engineering, giving rise to a growing number of research areas over time, where computer science and medical informatics stand out as the main emerging research areas.

Introduction

Benefits of intensive insulin therapy (IIT) to reduce diabetic complications has been evidenced from several clinical studies. In 1993, the Diabetes Control and Complications Trial Research Group, laid the groundwork for this affirmation, based on results obtained from 1441 type 1 diabetes (T1D) patients, about the IIT performance on delaying the onset and reducing the severity of diabetes complications compared to conventional therapy. ¹

Nowadays, “Artificial pancreas” is the most common designation used to refer a closed-loop system to control the glucose concentration of subjects with T1D implementing an automatic IIT. This system has also been named as “artificial endocrine pancreas,” “extracorporal artificial pancreas,” “closed-loop insulin delivery system,” “artificial β-cell system,” among others. Some of these names rely on either the route used for the insulin delivery and for the glucose concentration measurement. An artificial pancreas is considered an automatic system that delivers insulin continuously, trying to mimic the pancreatic β-cells response of a healthy subject, requires a continuous glucose measurement system, and implements a decision algorithm that computes the amount of insulin to be infused.

The artificial pancreas (AP) concept, development, assessment, and commercialization carried out during the last decades has attracted significant attention, where innovations and improvements of the insulin pump and glucose monitoring technology, modelling of glycemic dynamics, control algorithms, testing platforms, and clinical trials, have been some of the main research foci.

A renewed interest in the development of AP systems at the beginning of this millennium was promoted by several diabetologists gathered to formally restart the closed-loop initiative, developments in minimally invasive modern technology, and studies demonstrating the feasibility of promising AP designs for the control of glucose concentration. ²

Despite AP is an emerging research area of diabetes management, there is a lack of systematic, chronological, and synthesizing studies that show how this field has evolved over time. Although there are several consolidated and useful computational tools for bibliometric analysis, ³ there is only one study related to diabetes modelling and artificial pancreas using the Scopus database. ⁴ Other studies have been focused to Diabetes in a particular geographical area. ⁵ However, there are not studies summarizing the related research progress in the last 2-decade using the Web of Science database and multiple bibliometric tools of analysis and visualization so that key players and research trends can be recognized. Therefore, this study aims to implement a bibliometric analysis to determine the evolution of AP in terms of papers, citations, research categories, and keywords, to analyze the resulting international collaboration, and to recognize the prominent researchers and organizations on this research topic.

Materials and Methods

Data Collection

Data were collected on February 5, 2021, from Web of Science (WOS) Core Collection, including the Science Citation Index Expanded, the Social Sciences Citation Index, the Arts & Humanities Citation Index. and the Emerging Sources Citation Index. WOS has been accepted as a source of high-quality data and it is commonly used for scientific publications analysis.^6,7

The local database was built from a combination of keywords, Boolean logical operators and wildcards to find plural and other forms of words, resulting in the following search query: ((“artificial pancreas” OR “automated insulin delivery” OR “closed?loop insulin delivery” OR “artificial β?cell*” OR “artificial beta?cell*”) AND (“glucose control” OR “glyc$emic control”)). This equation was used in the topic field including title, abstract and keywords. Some types of documents as meeting abstract, early access, news, and editorial material, were excluded for quality reasons.

Information Analysis Tools

Bibexcel® software was used to extract and organize the records from the local database, ⁸ including the country and main organization fields. ⁸ An h-index for the author field of the local database was estimated using BibExcel® through the number of publications and citations. ⁹ A structured database and interactive visualization were made in MS-Excel® and power BI® software, respectively. ¹⁰ Ggplot2 in R-Studio was used for the analysis of trends.

Finally, an analysis of co-authorship networks for authors, countries, and co-occurrence for all keywords, was performed in the VOSviewer software. ¹¹ The CorText web application (www.cortext.net) was also used to visualize the network map of the relationship between the top 20 of organizations/authors.

Descriptive Analysis

AP Documents and Citations

The search was carried out from 2000 to 2020 obtaining 756 documents. The type of document “article” presented the highest number of documents and citations, 63.8% and 71.7%, respectively. “Review” was the second type of document, both in number of documents and citations, 19.4% and 23.8%, respectively; and finally, “proceedings paper,” “book chapter,” and “letter” documents with the smallest proportion.

Figure 1 shows the temporal distribution of the number and type of documents related to the AP topic from 2000 to 2020. The maximum number of documents and citations occurred in 2019 (with 111) and 2014 (with 2383), respectively, showing an exponential growth in this timeframe. The highest number of review documents occurred in 2018 (21%), which can be considered as an excellent sign for this research topic. A high number of review documents shows the relevance of a topic, since this type of document collects a large part of research information. ¹²

Figure 1.

Annual distribution of the Artificial Pancreas topic by type of documents and citations.

Trends in Research

The trend in research can be analyzed through the subject areas used in WOS collection. Figure 2 shows the number of WOS categories (WC) related to the AP topic for the last 2 decades. In 2000 there were only 4 categories, all of them clinical, while in 2020 there were 38 categories, most of them with a clear engineering focus. It is observed a growth trend traced through a linear regression showing a positive slope of 1.51 (R² = 0.88). The reader can find the complete list of categories per year and the corresponding number of documents in the Supplemental material.

Figure 2.

Number of WOS categories related to the AP topic over time.

Figure 3A shows the main WC found in the local database with at least 15 documents in total over the 2 decades and a prominent linear positive slope (greater than 0.5). Figure 3A also included the only WC that presented a negative slope. Figure 3B shows the featured emerging WC whose total number of documents was between 10 and 50, and had at least 70% of their total documents since 2015.

Figure 3.

Trends of the (A) main and (B) emerging WOS categories related to the AP topic over time. Each emerging category graph includes the relative percentage of documents since 2015. m, slope of the linear regression; TD, total documents.

Table 1 presents the documents with the highest number of times cited (TC) found in the local database for each of the main and emerging WOS categories shown in Figure 3.

Table 1.

Documents with the highest number of Times Cited (TC) from each of the Main and Emerging WOS Categories.

Category	Title	TC	Organization	Year
Main WOS categories
Endocrinology & Metabolism	Fully automated closed-loop insulin delivery versus semiautomated hybrid control in pediatric patients with type 1 diabetes using an artificial pancreas 13	337	Medtron MiniMed; Yale Univ	2008
Automation & Control Systems	Challenges and recent progress in the development of a closed-loop artificial pancreas 14	95	Rensselaer Polytech Inst	2012
Engineering, Electrical & Electronic	Periodic zone-MPC with asymmetric costs for outpatient-ready safety of an artificial pancreas to treat type 1 diabetes 15	77	Harvard Univ	2016
Engineering, Biomedical	An improved PID switching control strategy for type 1 diabetes 16	107	Sansum Diabet Res Fdn; Univ Calif Santa Barbara; Univ Padua	2008
Emergent WOS categories
Computer Science, Artificial Intelligence	Linear parameter varying (LPV) based robust control of type-I diabetes driven for real patient data 17	43	Obuda Univ	2017
Computer Science, Information Systems	Meal detection in patients with type 1 diabetes: a new module for the multivariable adaptive artificial pancreas control system 18	50	IIT; Univ Chicago; Univ Illinois	2016
Medical Informatics
Mathematical & Computational Biology
Pharmacology & Pharmacy	Insulin delivery methods: Past, present and future 19	73	GMERS Medial Coll; Univ Colorado; Univ Illinois	2016
Medicine, General & Internal	Artificial pancreas treatment for outpatients with type 1 diabetes: systematic review and meta-analysis 20	106	Aristotle Univ Thessaloniki; Univ Cambridge; Univ Oxford	2018
Pediatrics	Automated hybrid closed-loop control with a proportional-integral-derivative based system in adolescents and adults with type 1 diabetes: individualizing settings for optimal performance 21	28	Medtron MiniMed; Stanford Univ; Univ Colorado Denver; Univ Western Australia; Yale Univ	2017

Prominent Authors

According to the local database, 2014 authors have contributed to the AP research, where 1361 authors (67.6%) have a single publication. Table 2 shows the top 10 authors ordered by TD, detailing the contribution percentage, the number of citations, the h-index, MPY and the corresponding ranking position of each author. Hovorka R from the University of Cambridge (United Kingdom) is the most productive author, followed by Cobelli C from the University of Padua (Italy). The top countries by author affiliations in Table 2 are the United States (Harvard University, Sansum Diabetes Research Institute, Stanford University, University of Colorado), United Kingdom (University of Cambridge, Manchester University), Italy (University of Padua, University of Padova, University of Pavia), France (University of Montpellier), and Japan (Kochi University).

Table 2.

Top 10 most productive authors and corresponding organizations in the Artificial Pancreas topic. Several authors share the same rank due to corresponding contribution.

Author (organization)	TD (% contribution)	TC (ranking)	h-index (ranking)	MPY (TD)
Hovorka R (University of Cambridge)	57 (1,24)	2255 (1)	25 (2)	2014, 2019 (8)
Cobelli C (University of Padua)	50 (1,09)	2097 (2)	26 (1)	2015 (9)
Dassau E (Harvard University)	45 (0,98)	1658 (5)	22 (4)	2019 (7)
Doyle FJ (Harvard University, Sansum Diabetes Research Institute)	39 (0,85)	1712 (3)	24 (3)	2014, 2016 (7)
Renard E (University of Montpellier)	30 (0,65)	1674 (4)	19 (5)	2014, 2019 (4)
Hanazaki K (Kochi University)	29 (0,63)	427 (67)	13 (13)	2016 (5)
Wilinska ME (University of Cambridge)	29 (0,63)	1241 (6)	19 (5)	2019 (6)
Buckingham BA (Stanford University)	27 (0,58)	935 (18)	17 (6)	2017 (9)
Maahs DM (Stanford University)	27 (0,58)	713 (30)	15 (8)	2017 (9)
Del Favero S (University of Padova)	26 (0,56)	1192 (7)	15 (8)	2015 (6)
Thabit H (Manchester University)	26 (0,56)	1047 (11)	16 (7)	2014, 2018 (7)
Magni L (University of Pavia)	25 (0,54)	1063 (10)	15 (8)	2015 (5)
Forlenza GP (University of Colorado)	22 (0,48)	366 (79)	11 (11)	2017, 2019 (6)

MPY, most productive year; TC, times cited; TD, total documents.

Highly Cited Papers

Highly cited papers are considered of scientific excellence, top performance, and useful to benchmark the research performance according to its field, where the number of citations per article is outstanding. These papers usually define the research path in the literature. According to local database, the Medtronic-Yale’s group ¹³ ranks first with 337 times cited (TC), introducing the Medtronic ePID system evaluated in 17 adolescents with T1D during 34 h of closed-loop control in the hospital setting. Second, the Boston-Harvard-Massachusetts’s group ²² with 322 TC, introducing the safety and effectiveness of a bionic pancreas (delivering insulin and glucagon hormones) during 5 days in 20 adults and 32 adolescents with T1D in the outpatient setting. Third, the Montpellier-Padua-Virginia’s group ²³ with 295 TC, introducing a review of the state of the art related to the beginnings of the AP, its progress over time and challenges to face. Fourth, the Hannover-Tel Aviv-Ljubljana’s group ²⁴ with 247 TC, introducing the safety and efficacy of an AP system assessed in 56 adolescents with T1D at a diabetes camp. Finally, the Cambridge’s group ²⁵ with 219 TC, introducing a review of electromechanical closed-loop approaches, AP prototypes, and associated glucose sensors. Additional information about highly cited papers can be consulted in the Appendix.

Three of the 5 papers above, are related to clinical trials covering several testing conditions (hospital and ambulatory settings), while the other 2 correspond to review papers about the AP state of the art and closed-loop approaches. It is noteworthy that the 2 most cited papers on this list did not come from the prominent authors included in Table 2.

Network Analysis

AP Characterization through Keywords Evolution

Figure 4 shows the co-occurrence network of authors keywords based on the average year of publication of the documents in which a keyword appears. ²⁶ A minimum of 10 co-occurrences was considered to create this figure, resulting in 35 meets from 1084 keywords in total.

Figure 4.

Keyword network of the AP topic over time. The circle size represents the number of co-occurrences of a keyword, while the color represents the avg.pub.year parameter considered in VOSviewer. Keywords in blue and yellow are linked to early and recent years, respectively.

International Collaboration

Figure 5 shows the collaboration network between countries using VOSviewer. Here, a minimum of 10 documents per country was considered, obtaining 19 countries in total. The largest number of country-collaboration pairs (connecting lines in Figure 5) is led by the United States (17), followed by Germany (16), the United Kingdom (England) (15), Italy (14), and France (12). Regarding the number of TD in country-pairs (thickness of the line in Figure 5), the United States is in the first 3 places with Italy (35 TD), England (20 TD), and France (20 TD). Countries with the highest number of total collaborations on the AP subject were United States with 164, Italy with 90, England with 78, Germany with 75, and France with 63. Other countries had less than 50 collaborations.

Figure 5.

Country collaboration network. The circle size represents the total number of collaborations per country, the number of lines represents the number of countries collaborating, and the thickness of the line represents the number of collaborations between countries.

Figure 5 was created from a clustering and mapping technique through VOSviewer,^11,27 where countries were classified by color. The yellow cluster corresponds to the highest number of collaborations (222: Italy 90, France 63, Netherlands 41, China 17, and India 11), followed by the red cluster (185: Germany 75, Austria 36, Israel 36, Slovenia 27, and Denmark 11), and the Purple cluster (165: United States 16 and Japan 1). Others clusters have less than 130 collaborations.

Author and Organization Influence

Figure 6A shows the author collaboration network obtained through VOSviewer. ¹¹ A maximum of 50 authors per document and a minimum of 10 documents per author was considered, without TC restriction, resulting in 82 meets authors in total. The first name of authors was reduced to their initials. Authors were grouped by VOSviewer into 8 clusters: the red included 22 authors, 3258 collaborations and is led by University of Padua, the green included 15 authors, 1706 collaborations and is led by University of Cambridge, the blue included 15 authors, 1367 collaborations, and is led by Harvard University, and the yellow included 9 authors, 618 collaborations, and is led by Tel Aviv University. Others clusters had less than 400 collaborations. Most clusters were interconnected forming a main network whose most representative organizations, in addition to those mentioned above, were University of Girona (purple) and McGill University (orange). Figure 6A also shows 2 clusters (brown and gray) that did not collaborate with the main network neither among them. In particular, the bigger of these isolated clusters corresponds to a network that included 7 authors from Japan with 350 collaborations, and is led by Kochi University (which also appears in Table 2).

Figure 6.

Collaboration networks of (A) authors and (B) top 20 organizations/authors on the AP topic. (A) The closer the authors are to each other, the greater their relationship in terms of authorship. (B) Solid triangles and balls represent organizations and authors, respectively, while the transparent circles group the closest relationships.

Figure 6B shows the relationships between the top 20 authors (balls) and the top 20 organizations (triangles) found in the local database through the CorText Manager web application. Both top lists and the size of each node were defined by the number of documents. Each of the 40 nodes extracted is linked to some other by a line based on a distributional proximity measure. ²⁸ Moreover, nodes of authors and organizations were grouped in clusters (translucent circles) by their high proximity or relationship.

Discussion

The results show a growing interest in the AP topic since 2000 with a highly collaborative research. The exponential growth in the number of documents of Figure 1 could be influenced by different causes, for example, the development of new metabolic models of the glucose-insulin system, such as those proposed by Hovorka et al., ²⁹ or Dalla Man et al., ³⁰ through which the development of new AP systems evaluated and approved at a preclinical level through in-silico trials were greatly encouraged.^31–36 The commercial development of modern minimally invasive continuous glucose monitoring technology^37–40 and the spread of continuous subcutaneous insulin infusion therapy through insulin pumps^41–43 also promoted the implementation and clinical validation of artificial pancreas systems designed for the outpatient setting. ⁴⁴ On the other hand, organizations such as the Juvenile Diabetes Research Foundation (JDRF) and the National Institute of Health (NIH), financed the development of AP systems based on items of specific roadmaps for a massive and safe use (https://www.jdrf.org/impact/research/abstracts/ and https://www.niddk.nih.gov/about-niddk/research-areas/diabetes/type-1-diabetes-special-statutory-funding-program/about-special-diabetes-program).

Figure 1 also shows the annual distribution of citations, growing steadily over the years through 2014 and decreasing since 2015. This is a common behavior since citations require time to collect, particularly for more recent publications. Annual citations showed the highest number in 2014 (15.2% in total), being ²² by Russell et al., and ⁴⁵ by Frier, the most prominent documents with 13.5% and 8.6% of total citations for this year, respectively. Those papers reflect 2 main interest of the scientific community in the AP topic. The first relates to multiday studies under unrestricted outpatient conditions of advanced AP systems, such as the bihormonal one (insulin and glucagon), and the second relates to review papers, particularly aiming to hypoglycemia.

Regarding the WC, its number increased near to 950% in 2020 respect to 2000 (Figure 2). This increase over time shows the interdisciplinarity addressed by the AP topic, where categories like “Automation and Control Systems,” “Engineering, Electrical and Electronic,” and “Engineering, Biomedical,” were progressively integrated and received a greater role in this topic (Figure 3A). However, the “Endocrinology and Metabolism” category continues to primarily support the AP research with a prominent trend throughout the years. “Surgery” was the only WC with a negative linear slope from the entire list, which could suggest a decreasing interest on this alternative to reach an appropriate glucose control management. The emerging WC presented in Figure 3B, include a strong computer science focus, being “Computer Science, Artificial Intelligence,” “Computer Science, Information Systems,” and “Medical Informatics” the leading categories of this list according to their outstanding percentage of documents since 2015. This result suggests a growing interest in the AP research with a greater focus on computer science over the next years.

The list of documents related to the main WOS categories in Table 1 presents a common focus on applied control engineering, where “Endocrinology & Metabolism” excels as the category of greatest scientific interest in the AP topic. Clinical trials are usually published in this category. On the other hand, the documents from emerging WOS categories related to computer science and pediatrics of Table 1, present refined methods seeking to improve the performance of precedent AP realizations, while reviews and meta-analyzes are highlighted in “Pharmacology & Pharmacy” and “Medicine, General & Internal” categories. It is especially striking that in the “Surgery” category of Figure 3A, all relevant documents belong to Japanese researchers from Kochi University or Kochi Medical School.

The network analysis helps to explain the evolution of the AP topic over time, in particular Figure 4, where we identified 3 main stages. The first refers to the beginnings of AP systems and early hospital setting tests including critically-ill patients: “control algorithm,” “intensive insulin therapy,” “blood glucose control,” and “glucose monitoring.” Some of the highly cited papers are related to this stage.^13,46 In stage two, the development and use of AP systems is consolidated through systems for specific assessment conditions of interest, as the nocturnal hypoglycemia: “continuous glucose monitoring,” “insulin pump,” “model predictive control,” and “glucagon.” Again, some highly cited papers can be linked to this stage.^22,24 However, the third stage shows the maturity of this topic on a broader spectrum of evidence that includes other patient cohorts as well as comparisons against real-life performance settings: “children,” “type 2 diabetes,” “automated insulin delivery,” “technology,” “clinical trial,” and “exercise.” Some of the most representative papers related to this third stage were recognized in the local database since 2015 according to the number of citations, as follows: randomized crossover clinical trials,^47–50 reviews and meta-analysis,^20,51 and long-term feasibility studies. ⁵²

The keywords over time allow to reflect the development of the AP from very limited approaches in the first stage, followed by a consolidation of the field in the second stage, and the path envisaged for advanced systems in the third stage. This is also explained from the main and emerging WOS categories analyzed previously from Figure 3 and Table 1, gaining further experience in regulatory processes (Endocrinology & Metabolism category) along with an applied control engineering (Automation & Control Systems, and Engineering categories) that has evolved to allow that AP systems, originally developed for adult T1D subjects in research-facilities, can be used, without going through all stages of development, in studies with cohorts or test environments for which they were not initially considered.

The author’s influence in the AP topic is an extension of the network analysis that helps to understand how this topic has evolved over time. Figure 6A showed the most representative authors by number of collaborations in each cluster. These authors and their corresponding organizations are also part of Table 2, showing that the main knowledge generation comes from this great network, which lead the bibliographic production of the AP topic through a strong international collaboration, as can be confirmed in Figure 5. The red cluster of Figure 6A also presented the largest number of internal collaborations (ie, only organizations from the same cluster), with 1357 in total, followed by the green, blue and yellow clusters with 747, 443, and 274 internal collaborations, respectively. Blue cluster was the only with a greater number of external collaborations respect to the internal ones (about 52.1% of total). The remaining clusters showed a greater number of internal than external collaborations: red 71.3%, green 77.9%, yellow 79.7%, orange 88.9%, purple 86.7%, and gray and brown 100%. Overall, about 70.5% of all collaborations were internal, suggesting that research groups work primarily with the same collaborators in most studies. On the other hand, it was found a strong relationship of Figure 6B with Figure 6A regarding the distribution of clusters of most representative organizations and authors. Figure 6B shows 7 clusters, of which 3 stand out for their strong external collaboration: blue, yellow and soft green, being the University of Virginia the main pivot to integrate the Italian and French organizations with those of the United States. Moreover, organizations from England, Canada, Israel, and Germany have been collaborating mainly with American organizations such as the JAEB Center for Health Research, Stanford University, University of Colorado, the Sansum Diabetes Research Institute, Harvard University, and University of California, Santa Barbara. Japan organizations show no international collaborations. Figure 6B also show highly influential organizations, other than the affiliations of the most productive authors, such as the University of Bern, University of Tel Aviv, Schneider Children’s center of Israel, Kochi Medical School, Oregon Health and Science University, McGill University, and University of Montreal.

Regarding isolated clusters, the Japan experience (gray cluster in Figure 6A), presents several AP studies for a specific development designed for glycemic control of surgical and critical patients in intensive care units (ICU). Their main studies were ⁵³ with 47 citations and ⁵⁴ with 42 citations. Despite working in the same research area, all studies have been conducted locally by the organizations of this cluster, which explains the cluster isolation. In this regard, other AP designs for the ICU setting, as the one proposed by the GLYCOSTAT project, are currently under development considering several clinical hurdles delineated by leading experts in the field. ⁵⁵

Challenges and Future Research

Although this study covers a large part of the research progress in the AP topic during the last decades, obtaining consistent results according to our background, one remaining challenge is the metadata analysis by joining and standardizing information from different databases (ie, WOS and Scopus). Furthermore, since there is a lack of standardization of keywords, or exclusive terms related to this topic, the results may vary substantially to the search terms, the search query, or the database used for the bibliometric analysis. These challenges could be addressed in future research covering the information in a comprehensive way that reduces the variability inherent in this kind of studies.

Conclusion

A bibliometric analysis on the AP topic was carried out from scientific publications made during the period 2000-2020. Prominent documents, authors, organizations and countries were identified from a descriptive analysis revealing an overall understanding about this technological treatment for glycemic control. A considerable level of collaboration between countries was evidenced in the research progress, with a growing influence of engineering that fostered an interdisciplinary development of the topic. Results revealed a comprehensive background of the knowledge generation for the AP topic during the last 2 decades, showing a growing number of research areas, where computer science and medical informatics stand out as the main emerging research areas.

Appendix

Table A.

Most cited documents in the Artificial Pancreas topic.

Title	Year	TC	Source	Country	Organization
Fully automated closed-loop insulin delivery versus semiautomated hybrid control in pediatric patients with type 1 diabetes using an artificial pancreas 13	2008	337	Diabetes Care	USA	Medtron MiniMed; Yale Univ
Outpatient glycemic control with a bionic pancreas in type 1 diabetes 22	2014	322	New England Journal of Medicine	USA	Boston Univ; Harvard Univ; Massachusetts Gen Hosp
Artificial pancreas past, present, future 23	2011	295	Diabetes	France; Italy; USA	Montpellier Univ Hosp; Univ Montpellier I; Univ Padua; Univ Virginia
Nocturnal glucose control with an artificial pancreas at a diabetes camp 24	2013	247	New England Journal of Medicine	Germany; Israel; Slovenia	Kinder & Jugendkrankenhaus; Schneider Childrens Med Ctr Israel; Tel Aviv Univ; Univ Childrens Hosp; Univ Ljubljana
Continuous glucose monitoring and closed-loop systems 25	2006	219	Diabetic Medicine	UK	Univ Cambridge
Continuous glucose monitoring: a review of successes, challenges, and opportunities 56	2016	212	Diabetes Technology & Therapeutics	USA	Biomed Informat Consultants LLC
Hypoglycaemia in diabetes mellitus: epidemiology and clinical implications 45	2014	206	Nature Reviews Endocrinology	UK	Univ Edinburgh
Closed- loop artificial pancreas systems: engineering the algorithms 57	2014	195	Diabetes Care	USA	Sansum Diabet Res Inst; Univ Calif Santa Barbara
Closed-loop insulin delivery - the path to physiological glucose control 58	2004	191	Advanced Drug Delivery Reviews	USA	Sensor R&D
Fully integrated artificial pancreas in type 1 diabetes: modular closed-loop glucose control maintains near normoglycemia 59	2012	182	Diabetes	France; Italy; USA	Sansum Diabet Res Inst; Univ Calif Santa Barbara; Univ Montpellier; Univ Padua; Univ Pavia; Univ Virginia

TC, times cited.