On the 1st of August 2023, a comprehensive search was conducted on the Web of Science Core Collection (WOSCC) to retrieve all citations published from the 1st of January 2012 to the 31st of December 2022. The search was executed using the following formula: TS=(“Continuous blood glucose monitoring” OR “Continuous glucose monitoring” OR “Implantable CGM system” OR CGM OR FGM OR “Flash glucose monitoring” OR “Ambulatory glucose monitoring” OR “Continuous glucose sensors” OR “Real-time glucose monitoring” OR rtCGM OR “Subcutaneous glucose monitoring” OR “Continuous glucose measurement” OR “Continuous glycemic monitoring” OR “Continuous glucose sensing” OR “Continuous glucose meters”) AND TS=(Diabetes* OR “Diabetes mellitus*”), while limiting document types to “Article” or “Review Article”. Articles and reviews written in English were considered, while meeting abstracts, early access articles, editorial material, letters, collections, proceeding papers, news items, book chapters, hardware reviews, and withdrawn publications were excluded. The initial screening process yielded a total of 3680 original English articles, comprising 3207 articles and 473 reviews, which were deemed to be potential candidates for inclusion in the study.

To ensure the precision and caliber of the acquired data, a dual review process was undertaken by two researchers, Laixi Kong and Maoting Guo, who independently scrutinized the abstracts and keywords of literature to obtain the most pertinent articles. The objective of this study was to investigate the clinical practice of CGM in diabetes; thus, these clinical practice encompass various aspects: improvements in blood glucose control following CGM use, the utilization and interpretation of CGM-related metrics, remote monitoring and telemedicine, artificial pancreas (closed-loop systems), and the integration of CGM with multiple insulin pump systems. Exclusion criteria encompass topics such as technical design in CGM sensors, sensor material research, and unrelated reviews. Following manual screening, a total of 3024 papers were deemed suitable for inclusion in this study. From each publication, the title, publication year, country or region, institution, author, journal, references, author and keywords were methodically extracted. Further details pertaining to the literature extraction process are presented in Figure 1 .

Initially, a basic statistical analysis of the dataset was conducted utilizing Rstudio (v.4.2.2). The “bibliometrix” format was employed to store the data, and the “biblishiny” package was utilized to extract a range of features related to the research literature between 2012 to 2022 (13). These features, including Main Information, Most Relevant Authors, Author’s Production Over Time, Most Global Cited Documents, served for quantitative analysis. Subsequently, CiteSpace 6.1.R6 was utilized to cluster the keywords of institutions present in the literature, perform dual map overlay analysis of journals, unveil keyword clustering analyses within the text, identify the strongest cited bursts, and construct co-cited references timeline maps of publications. In addition, VOS viewer was employed to identify the collaborative networks of countries and institutions, evaluate the keywords pertaining to the subject, and visualize the post-analysis of the results.

In this review, a comprehensive analysis of 3024 literature sources was conducted, and the resulting search data was used to plot the trends in studies related to the application of continuous glucose monitoring (CGM) to clinical practice in diabetes using R studio. As illustrated in Figure 2 , the analysis revealed a consistent annual increase in the volume of research articles on this subject from 2012 to 2019, followed by a sharp rise in the number of publications from 2019 to 2021, suggesting a heightened interest in research pertaining to clinical practice of CGM in diabetes during this period. Notably, 2021 recorded the highest output of 555 articles. Furthermore, a linear trend line of annual publications was developed to gain further insights into the output trend, resulting in the equation Y=44.745X+6.4364, where Y represents the annual publications and X denotes the year. This model exhibits a coefficient of determination (R²) of 0.8819. Figure 3 presents an overview of the analyzed articles, encompassing a total of 46243 references and an average publication year of 4.37. Moreover, each article garnered an average of 22.88 citations, while the annual publication growth rate was 15.83%.

A total of 91 countries have conducted studies on the topic at hand. The Figure 4A indicated that the United States had the highest number of articles published (1029), followed by the United Kingdom (331) and China (264). The top 10 countries in terms of output were summarized in the Table 1 , with the United States exhibiting the highest centrality (0.17), H-index (86), and Citations Per Papers (32.00), surpassing other countries by a significant margin. Although China and Japan ranked high, their H-index and centrality were comparatively lower than those of other countries. The international cooperation relationship of each country was visualized using the CiteSpace, as shown in Figure 4B , where nodes represented countries and node size reflected the amount of national issuance. The purple portion of the circle represented centrality, with the United States positioned at the center, indicating frequent cooperation with other countries. Furthermore, the circle of the United States was the largest, indicative of the most influential issuance in the region.

Table 2 illustrates the top 10 institutions with the highest literature output, where in HARVARD UNIVERSITY, UNIVERSITY OF COLORADO SYSTEM, and UNIVERSITY OF COLORADO ANSCHUTZ MEDICALCAMPUS emerged as the top three institutions with the highest number of published articles (151, 144, and 132, respectively). Notably, HARVARD UNIVERSITY exhibited a significantly higher H-Index compared to other two institutions, indicating its dominant influence in publishing articles. Seven out of the top 10 institutions were affiliated with the United States.

The collaborative relationships between institutions were disclosed through the use of CiteSpace, as depicted in Figure 5A . The connecting line between each of the two labels in Figure 5B shows that the institutions in the same country cooperate closely.

Figure 6A presents the roster of the ten most pertinent authors within this research domain, with particular emphasis on the 61 articles affiliated with Roy W. Beck. A nuanced comprehension of the potency of influence and the yearly evolution of publications among these ten authors over the past decade is facilitated by Figure 6B . Evidently, Roy W. Beck sustains a conspicuously high echelon of scientific impact within this research sphere (14). Remarkably, it is salient that seven studies associated with him have ascended to constitute the upper echelon of the ten most frequently cited articles within this field (15–21).

In light of these dynamics, collaborative networks of research materialize as instrumental conduits for researchers to augment the breadth of their investigative pursuits or to conjoin forces with cohorts engaged in cognate inquiries. Accordingly, a judicious author threshold of 107 was established. Employing VOSviewer, we proceeded to visualize the extant panorama and gradation of author interplay within this domain, with the ensuing outcomes being expounded in Figures 6C, D .

Upon analyzing the literature’s cited and citing journals, it was possible to determine the influential journals in the field. Table 3 illustrated the top ten cited and citing journals, with DIABETES TECHNOLOGY & THERAPEUTICS holding the highest rank as the first citing journal, followed by DIABETES CARE and DIABETES RESEARCH AND CLINICAL PRACTICE. Among the cited journals, DIABETES CARE held the top spot, followed by DIABETES TECHNOLOGY & THERAPEUTICS and J DIABETES SCI TECHNOL. In 2022, DIABETES CARE held the highest impact factor among the citing journals, with a score of 17.152, followed by DIABETES TECHNOLOGY & THERAPEUTICS with a score of 7.337.

Furthermore, in plotting the Dual map overlay using CiteSpace, the journals that contributed to publications on the clinical practice of CGM in Diabetes Mellitus were analyzed. The resulting map in Figure 7 was divided into two halves, with the left side representing the research area of the cited journals and the right side depicting the research area of the citing journals. The colored curves between the nodes on the left and right halves illustrated the relationship between the highly active research areas of the two journals. The examination of the graph indicated the presence of two discernible green curves, which implied that publications pertaining to medicine, medical and clinical domains have a higher likelihood of being referenced by journals that focus on molecular, biological, and genetic areas, as well as health, nursing, and medical fields.

In this study, we utilized VOSviewer to visualize the 100 high-frequency keywords in the literature pertaining to the topic of interest. A threshold of 41 occurrences was set for the selection of these keywords. The resulting visualization in Figure 8A demonstrated that darker color blocks corresponded to higher frequency of occurrence of the respective keyword in the literature. Furthermore, proximity of a color block to the center of the yellow block indicated higher citation frequency and cited frequency. The top ten hot keywords, as ranked by frequency, were presented in Table 4 . Notably, “Type 1 diabetes”, “Glycemic control”, “Hypoglycemia” were among the top keywords, suggesting their significance as hot topics in this research field over the past decade. And Figure 8B illustrates the chronological depiction of keyword clustering analysis, offering a timeline perspective. The diagram portrays various clusters denoted by distinctively colored horizontal lines on the right side, each corresponding to a collection of keywords. The nodes positioned along these horizontal lines symbolize individual keywords. Notably, the spatial arrangement of these nodes along the horizontal axis signifies the inaugural appearance year within the scholarly literature for the associated keyword, thereby constituting a comprehensive temporal representation of the keyword cluster’s evolutionary progression. The cluster “0# glycemic variability” is the largest, Next is “#1 artificial pancreas”, “#2 physical activity”, “#3 type 2 diabetes”, “#4 flash glucose monitoring” and “#diabetes mellitus”

The CiteSpace is capable of identifying keywords that experience significant changes in frequency during a specific time period, commonly known as emergent words. Keywords that exhibit a delayed emergence and extended duration are indicative of the most recent research trends in a given field, enabling a temporal review of research hotspots and the projection of future trends. The default configuration of CiteSpace was substituted with the ensuing modes: “Year Per Slice” set to 1, “Top N%” set to 30.0%, and “Minimum Duration” set to 1. After conducting an analysis on the keywords with citation bursts, we determined that the 8 strongest burst keywords should be displayed as illustrated in Figure 9 . During the period spanning from January 2012 to December 2022, the ensuing keywords surfaced as outcomes: fluctuation (2012-2018), plasma glucose (2012-2016), hyperglycemia (2013-2016), cardiovascular disease (2013-2016), reduction (2014-2017), intensive treatment (2018-2020), time in range (2020-2022), and young adults (2020-2022). In the preceding two years, the keywords “Time in Range” and “Young Adult” have surfaced and persisted throughout 2020 and beyond. Of the two, “Time in Range” has exhibited the most intense outbreak with a value of 20.47, indicating that it currently represents the primary research focus and potentially marks a pivotal juncture with notable implications for future inquiry.

A timeline map of co-cited references was constructed using CiteSpace with the aim of comprehending the principal research topics and their progression within the field. The outcomes of the keyword clustering analysis of the references were exhibited on the right-hand side of Figure 10 , with “#flash glucose monitoring” comprising the most significant cluster. On the left-hand side, the citation relationship among each reference was presented over time, wherein larger nodes signified more frequent citations, and node color indicated the time when the reference was cited. The top 10 most frequently cited references were enumerated in Table 5 .

In brief, the annual production of scholarly articles in this field exhibited an upward trend overall, with a significant and substantial surge observed in 2020, likely due to the outbreak of the COVID-19 pandemic. The pandemic-induced public health measures have altered people’s lifestyles, potentially impacting the glycemic control of individuals with diabetes by limiting physical activity to some extent (22). Furthermore, the use of glucocorticoid therapy may exacerbate hyperglycemia once severe infections such as COVID-19 pneumonia have developed. Consequently, a study has suggested that a system that integrates telemedicine and continuous glucose monitoring (CGM) can effectively manage blood glucose levels and prevent adverse outcomes (23, 24). As a result, CGM has gained increasing adoption in clinical settings, with a peak in related research output in 2021. In the analysis of countries within a particular research area, the United States emerged as the leading contributor in terms of number of publications, centrality, H-index, and citations. This suggested that the United States possessed a greater level of influence within this research area and engages in frequent collaborations with other countries. Based on the aforementioned analysis, it is recommended that research teams hailing from Asian countries seek to augment their international influence by engaging in heightened cooperation with their counterparts in European and American nations. The dual map overlay depicted in Figure 7 reveals a wide range of subject areas covered by cited and citing journals, indicating untapped potential for further exploration within this research area. By employing a clustering analysis approach and examining the emergence of keywords and references, we have been able to identify research hotspots between 2012 and 2022 and forecast future trends in this research area. Notably, the most cited reference is a review authored by Tadej Battelino, Thomas Danne et al. This international consensus validated the feasibility of using the TIR index as a clinical endpoint and outcome measure to supplement HbA1c in various relevant populations, and the target thresholds outlined in the article serve as a valuable framework and point of reference for the clinical application of CGM (15).

The fundamental essence of an academic field can be encapsulated by its keywords, and through visual analysis of these keywords, one can discern the prevailing research trends and trajectories (25). Based on the high-frequency keywords extracted and the keyword clustering timeline mapping generated by CiteSpace, the primary research areas in this field during the past decade can be identified. These areas include continuous glucose monitoring (CGM), glycemic variability, type 1 diabetes, and hypoglycemia. Significantly, the other three hot keywords were all generated based on CGM.

In the realm of diabetes management, the advent of continuous glucose monitoring (CGM) technology has bestowed unprecedented prospects for the monitoring and regulation of patient glycemic levels. Traditional intermittent approaches to glucose monitoring have progressively exhibited their inherent limitations, rendering the comprehensive capture of blood glucose fluctuations throughout a patient’s diurnal existence a challenging endeavor. In contrast, the real-time and uninterrupted monitoring attribute intrinsic to CGM technology introduces a novel instrument for therapeutic guidance, both for medical practitioners and their patients. The study of glycemic variability has emerged as a pivotal domain of investigation in contemporary times, and concomitant with the evolution of CGM technology, its integration within clinical practice has experienced a notable escalation in recent years (26, 27). Recent findings have indicated a correlation between elevated glycemic variability (GV) and the advancement and escalation of vascular complications in diabetic patients, heightened susceptibility to hypoglycemic episodes, as well as a decline in the quality of life (QOL) for affected individuals (28–30). The metrics encompassing Glycemic Variability are presently acknowledged as a significant gauge of glycemic management (31). This underscores the imperative of delving into the systematic examination of continuous glucose monitoring (CGM) data, and underscores the criticality of adeptly harnessing and interpreting CGM data to effectively serve its role within the realm of clinical practice.

The assessment of transient glycemic fluctuations is frequently derived from continuous glucose monitoring standard deviation (CGM.SD), a readily computable metric commonly employed to quantify short-term glycemic variability. However, it is worth noting that CGM.SD is influenced by the prevailing mean glucose levels, thereby rendering it susceptible to this parameter. Conversely, the coefficient of variation (CV), derived from both the standard deviation and mean glucose, serves to ameliorate this inherent limitation by compensating for the aforementioned sensitivity to mean glucose levels (15). Furthermore, within clinical settings, diverse indices such as the mean amplitude of glycemic excursion (MAGE), J-index, low blood glucose index/high blood glucose index, average daily risk range, and mean of daily differences (MODD) are employed to evaluate distinct facets of glycemic fluctuations in patients (32–36). Additionally, the Time in Range (TIR) parameter, denoting the proportion of time during which blood glucose levels remain within specified glycemic thresholds, while not strictly categorized as a glycemic variability metric, assumes significance as a supplementary clinical target and an outcome measure for HbA1c assessment across a spectrum of diabetes mellitus presentations, as established by international consensus (15). Consensus opinions have also established a link between TIR and the risk of diabetic complications, such as the close association between TIR and the risk of microvascular complications (19), as well as the good correlation between TIR and HbA1c. Additional research has further substantiated the notion that HbA1c inadequately captures data pertaining to the fluctuation of blood glucose levels or the duration of time spent within the hypoglycemic or hyperglycemic spectrum. Consequently, the Time in Range (TIR) metric is presently being embraced as a favored metric for prognosticating the susceptibility to diabetic complications, delineating outcomes in clinical investigations, and evaluating glycemic management in patient cohorts (37, 38).

As the keyword emergence shown in Figure 9 , the emerging keywords for 2020 to 2022 are “Time in Range” and “Young Adult”. Research conducted in the realm of clinical practice concerning Continuous Glucose Monitoring (CGM) throughout this timeframe has been predominantly centered around acquiring more precise and up-to-the-minute data pertaining to glycemic regulation. Additionally, investigations have been dedicated to the viability of employing Time in Range (TIR) as a quantifiable parameter, alongside examinations concentrated on the demographic of young adults. Envisioning the future, the burgeoning prominence of TIR as a measurable criterion is anticipated to persistently mirror the evolving methodology within the domain of diabetes management. Progressions in technological innovations coupled with an increasingly profound comprehension of the intricacies inherent in diabetes are poised to propel these transformations forward, culminating in a sustained emphasis on ameliorating long-term prognosis and enhancing the quality of life for individuals afflicted with diabetes.

Although Continuous Glucose Monitoring (CGM) can furnish real-time insights into blood glucose levels and trends, alongside retrospective analyses of glycemic regulation patterns and glycemic metrics over specific temporal intervals (4, 39), their assimilation into clinical practice falls short of reaching optimal levels (40). The suboptimal adoption can be attributed, in significant part, to the dearth of software possessing the capacity for relatively straightforward and standardized statistical and graphical depiction, as well as interpretation, of glycemic data, thereby engendering uncertainty and reluctance among clinicians towards integrating CGM into their professional milieu (41, 42). Consequently, to surmount these obstacles and harness the full potential of continuous glucose monitoring (CGM) data within clinical settings, a method christened as the “ambulatory glucose profile” (AGP) was devised. The AGP is a tool utilized for assessing short-term glycemic variability indices in diabetic patients. By analyzing CGM data, it generates charts depicting median, interquartile range (IQR), and other statistical values, thereby providing a comprehensive evaluation of intra-day and inter-day glycemic fluctuations for patients (43, 44). A methodical examination of AGP reports proves to be a valuable and pragmatic approach, enabling real-time and comprehensive assessment of glycemic control and the effectiveness of any therapeutic modifications (45, 46). Through meticulous scrutiny of AGP charts, clinical practitioners can attain enhanced comprehension of patients’ glycemic patterns, identify potential issues, and discern opportunities for refining treatment regimens. Thorough AGP analysis aids in pinpointing pivotal factors for achieving optimal glycemic control, thereby furnishing robust support for formulating appropriate therapeutic adjustments and further integrating CGM data into routine clinical practice (47, 48). With the introduction of the AGP approach, clinicians are better equipped to expound upon and communicate glycemic data, collaboratively establish personalized treatment objectives with patients, and monitor their progress throughout the course of treatment. This endeavor fosters closer doctor-patient relationships, heightens patients’ awareness of glycemic control, and ultimately augments the efficacy of diabetes management. In addition, there are a number of Software Packages and Tools that support comprehensive analysis of CGM data (49).

Ultimately, through a bibliometric analysis of this research domain, we can discern with clarity that the focal point of clinical practice research transcends the mere analysis of various metrics and has surpassed conventional data analysis. As depicted by the keyword clustering analysis in Figure 8 , investigations are progressively expanding into the application of cutting-edge technologies such as artificial pancreas, machine learning, and artificial intelligence. These studies not only furnish diabetic patients with more advanced therapeutic modalities but also usher in novel possibilities for technological innovation and advancement within the realm of medicine. Looking ahead, we can anticipate witnessing further breakthroughs in these domains, heralding a positive impact on the well-being and quality of life for individuals afflicted by diabetes.