Roadmap to the Effective Use of Continuous Glucose Monitoring: Innovation, Investigation, and Implementation

Abstract

For 25 years, continuous glucose monitoring (CGM) has been evolving into what it is now: a key tool to both measure individuals’ glycemic status and to help guide their day-to-day management of diabetes. Through a series of engineering innovations, clinical investigations, and efforts to optimize workflow implementation, the use of CGM is helping to transform diabetes care. This article presents a roadmap to the effective use of CGM that outlines past, present, and possible future advances in harnessing the potential of CGM to improve the lives of many people with diabetes, with an emphasis on ensuring that CGM technology is available to all who could benefit from its use.

The incidence and prevalence of diabetes is increasing globally; yet, today, the overall quality of diabetes care is far from optimal (1), pushing us to explore new tools for and approaches to diabetes management. Glucose monitoring is one of the core components of diabetes management, along with personalized medication selection, insulin delivery systems, lifestyle management, and a focus on reducing diabetes distress and addressing the social determinants of health (SDOH) that significantly affect diabetes care and quality of life.

The Evolution of Glucose Monitoring

Remarkable strides have been made in the how we monitor glycemia, progressing from urine testing to fingerstick blood glucose monitoring (BGM) to, over the 30 years since the Diabetes Control and Complications Trial, reliance in large part on A1C assessment to evaluate glycemic status and guide diabetes management. These strides represent 40 years of progress in glucose monitoring, but it was not until continuous glucose monitoring (CGM) was incorporated into the care process for type 1 diabetes between 2008 and 2015 that we had a glucose monitoring modality that was engaging, enlightening, and began to connect people with diabetes and health care professionals (HCPs).

Since the introduction of CGM, there have been so many “Ah-Ha!” moments for people with diabetes (e.g., “It changed my life!” “It opened my eyes.” “It gave me peace of mind.”) and for HCPs (e.g., “I never knew an A1C of 8.3% could have a CGM profile that looked so abnormal and needed immediate attention!”). There is even a CGM education–based initiative titled, The AH-HA! Project (2).

A1C, with the most robust and longitudinal data, is currently the most commonly used marker of glycemic risk for long-term, diabetes-related vascular complications. However, CGM is the best tool to assess and manage the occurrence of acute glycemic complications of diabetes (i.e., dangerous hypoglycemia or extreme hyperglycemia), and data are also steadily building that correlate CGM metrics to long-term vascular complications (3,4). To reiterate, whereas A1C is a good long-term risk measurement tool, CGM has evolved to be an important measurement tool for both acute and long-term risks and is also an effective personalized diabetes management tool (5). Today, many clinicians are setting glycemic management goals based on the glucose management indicator (GMI), a CGM-derived approximation of A1C previously referred to an “estimated A1C.” Clinicians are increasingly also setting management goals based on the CGM-derived time in range (TIR) and time below range (TBR) metrics.

The U.S. Food and Drug Administration first approved CGM in 2001 (6). Since then, there have been major technical advances in interstitial glucose-sensing technology, including decreased sensor size, increased convenience and accuracy, elimination with some CGM systems of the need for calibration using BGM, approval of nonadjunctive status for some systems, and improved interoperability.

Five years ago, it seemed to me that the stage was set for CGM to transform type 1 diabetes care and also to have a significant impact on type 2 diabetes management, but that these paradigm changes were going to come about through a step-by-step process (7). Looking back at the past 25 years and also contemplating the steps still ahead that will be needed to realize CGM’s true potential, I created a roadmap for the effective use of CGM, which is presented in Figure 1 and explained in detail in the remainder of this article. This roadmap is organized in a general chronological progression starting with approval of the first CGM system in the United States in 2001 and laying out the steps or areas of CGM-focused research and development that have built on each other, allowing for the incredible progress we have seen to date and can anticipate continuing to see in the future.

Figure 1.

Roadmap to the effective use of CGM: innovation, investigation, and implementation. For most people with diabetes, TIR is defined as time spent with glucose 70–180 mg/dL and depicted on the AGP report in green; TBR is time spent with glucose <70 mg/dL and is depicted on the AGP report in red. *Adapted from ref. 9. CDCES, certified diabetes care and education specialist; RPM, remote patient monitoring. ©2023 HealthPartners Institute, International Diabetes Center. All rights reserved. Used with permission.

A CGM Roadmap: Cycles of Innovation, Investigation, and Implementation

After a decade of CGM data and nomenclature standardization, the roadmap illustrates a path forward to develop effective widespread clinical implementation and workflow strategies. Seeing the potential value of CGM in clinical practice, efforts over the past 5 years began to focus on finding ways to 1) facilitate CGM-guided clinical decision-making, 2) include CGM metrics as outcome measures in clinical trials and regulatory drug prescribing information, and 3) elucidate how CGM metrics can be considered diabetes quality measures. These three efforts to further enhance the clinical value of CGM for diabetes assessment and management, shown as separate steps on the roadmap, when considered together, constitute the CGM Triple Aim, as shown in Figure 2.

Figure 2.

CGM Triple Aim. ©2022 HealthPartners Institute, International Diabetes Center. All rights reserved. Used with permission.

CGM research and development did not stop with the Triple Aim, however. Innovations in diabetes technology connected CGM systems and insulin pumps with a glucose-responsive control algorithm, and thus the field of automated insulin delivery (AID) emerged. AID development successfully followed the insightful artificial pancreas roadmap championed by the JDRF’s Aaron Kowalski (8), forever changing and improving the course of type 1 diabetes management. Additionally, investigations, including clinical trials, quality improvement programs, and studies analyzing real-world data, have led to new indications for the use of CGM in people with type 2 diabetes who use insulin, as well as during pregnancy in people with diabetes.

The pace of CGM uptake in clinical practice is steadily building, but there have been detours on the journey, and even some U-turns, when we needed to circle back and update CGM metrics, find ways to better integrate data into electronic health record (EHR) systems, and find more efficient ways to use the CGM data for population health management.

Every roadmap (certainly in the era of GPS) needs a destination we are trying to reach in a timely manner. The destination of the roadmap presented here is a health care system that reaches the Diabetes Quintuple Aim, which comprises equity in diabetes care, quality of diabetes care, reduced patient burden, reduced clinician burden, and reduced costs. This is a diabetes management–focused adaptation of the broader Quintuple Aim for Health Care Improvement recently proposed by Nundy et al. (9).

It is worth briefly reviewing the progress we have made on the first nine steps along the roadmap and why steps 10 and 11 (the future steps) are still important to pursue. Exploring the CGM roadmap reveals that people with diabetes, clinicians, and even entire health care systems can benefit from learning how to use CGM effectively. We need to ensure that every person with diabetes who can benefit from CGM is given the opportunity to use this transformative technology. When traversing any roadmap, it is also important to keep an eye on the rearview mirror to appreciate the ground we have already covered and also to realize that we will likely be revisiting many of these steps again for additional refinement toward further innovation, investigation, and implementation.

CGM Metrics and Visualization: A Report to Help Standardize, Organize, and Analyze Data

In 2012, just over a decade after CGM was approved and early studies showed its promise for improving glycemic management (10,11), the first expert consensus development meeting (facilitated by the International Diabetes Center [IDC] and supported by the Leona M. and Harry B. Helmsley Charitable Trust) was held to begin the process of standardizing CGM metrics and CGM data reporting (12). This meeting resulted in the development of the one-page, three-panel ambulatory glucose profile (AGP) report. This effort built on the original AGP concept, which was developed by Mazze et al. (13) in 1987 using BGM data.

The CGM metrics and AGP report were refined in a series of additional consensus development meetings. Through these meetings, participating experts agreed on 10 core clinical CGM metrics in 2017 (14–16), and, in 2019, CGM targets were added for the five key “times in ranges” metrics (TIR [70–180 mg/dL], time above range [>180 and >250 mg/dL], and TBR [<70 and <54 mg/dL]) (17). The core metrics, targets, and AGP data visualization report (Figure 3) were then incorporated into the American Diabetes Association’s (ADA’s) Standards of Medical Diabetes Care—2020 and have been updated in subsequent years (18).

Figure 3.

AGP report three-step action plan. ©2022 HealthPartners Institute, International Diabetes Center. All rights reserved. Used with permission.

As consensus was building for the CGM metrics and data visualization report, efforts were also underway to develop a method for systematically analyzing CGM data, starting with the IDC’s nine-step guide to CGM interpretation (19–21), followed by the five-step DATAA Model (22). In response to busy clinicians who wanted to use CGM but were hoping for an even simpler or faster approach to CGM data analysis, a three-step AGP interpretation guide was presented (23), published (24,25), and revised (26) by the IDC between 2020 and 2022. Over the next year, other simplified CGM interpretation guides followed (27).

The simplified three-step approach to AGP interpretation from 2021 asked three questions of clinicians, progressing down the three panels of the AGP report: 1) Is there a problem with glucose control? 2) Where is the problem with glucose control? and 3) What adjustments in medications or lifestyle are needed to reach optimal glucose control?

Because the term “glucose control” did not represent ideal patient-centered language (28) and the key message of immediate action was not prominent enough, the IDC has since reframed the three-step CGM analysis to be much more action-oriented, with the goal of overcoming therapeutic inertia (Figure 3). The current three steps that are suggested for rapid interpretation of an AGP report and intended to lead to action can be summarized by the directive phrase “Determine Where to Act,” as follows:

1. Determine whether action is needed.

   • Review the times in ranges with a focus on TIR and TBR.
   • Action is needed if both TIR and TBR are not both at target (>70% and <4%, respectively). Goal: more green, less red (MGLR).
2. Where is action needed?

   • Review the AGP curve and daily views graphs.
   • Action is needed at the time of day when the glucose profile is furthest from being flat, narrow, and in range (FNIR).
   • Always start by addressing any hypoglycemia first (shown in red in the current version of the AGP report included in the ADA’s 2023 Standards of Care [29]).
3. Act on the data.

   • Adjust lifestyle and medications and address diabetes distress and SDOH.
   • Adjustments should be made right away, with timely follow-up adjustments scheduled (in clinic or virtually).
   • Follow up: adjust, adjust, adjust, until MGLR targets are reached and the AGP curve is approaching FNIR.

Acting on CGM Data: Thinking Fast and Slow

People with diabetes need to learn how to effectively act on the CGM data they see in real time, and clinicians need to determine where to act on the CGM data they review retrospectively, either in person or remotely. I refer to these actions as either “acting fast” or “acting slow,” as described by the Nobel laureate Daniel Kahneman in his book, Thinking Fast and Slow (30).

Fast thinking or action occurs when people with diabetes look at real-time CGM data minute by minute and day by day, responding to high and low glucose values or alerts using the CGM trend arrows displayed on their smartphone app to guide their corrective actions. Fast-ish thinking can also include making individual changes in behavior based on postprandial and post-exercise CGM data. It is remarkable how fast CGM metrics improve in clinical trials after starting people with diabetes on real-time CGM. In an analysis by Raghinaru et al. (31) of data from eight randomized trials, TIR was found to have improved within the first week of starting CGM to the level it would remain, on average, throughout the duration of the trials. TBR usually took ∼2 weeks to reach stable improved levels in these trials.

Slow thinking or action occurs with the retrospective analysis of AGP reports (via the three-step Determine Where to Act process, with the aim of achieving MGLR and FNIR and following up to adjust as needed). Such retrospective analysis is usually visualized on a computer or other larger monitor. Ideally, a shared decision-making session between the clinician and the person with diabetes would follow each AGP report analysis to reach agreement on the best therapy changes to make and a follow-up plan for further adjustments, as needed to avoid therapeutic inertia. Although some improvement in TIR and TBR occurs very quickly when starting real-time CGM (described above as fast and fast-ish thinking), to reach optimal CGM targets and A1C goals, there is added benefit to performing retrospective analysis of CGM data (i.e., slow thinking) and making adjustments as needed through shared decision-making. Additional study is needed to determine the appropriate educational strategies and management tools, including smartphone apps, and the ideal mix of face-to-face and virtual visits needed to optimize the use of real-time (fast) and retrospective (slow) CGM data.

If discrete CGM metrics and AGP reports can be automatically incorporated into the EHR with just an order placed in an individual’s electronic chart (requiring a U-turn back to step 2 of the roadmap), this ability not only helps clinic workflow and communication among team members (as in roadmap step 4), but also allows for population health and case management guided by a clinician, clinic, pharmacy, or health plan. Population-level CGM data, along with EHR demographics, medication history, and laboratory results combined with health plan claims data will allow for real-world CGM cost-effectiveness studies in the future (roadmap step 5). Although this step may be considered a future component of the CGM roadmap, some important diabetes registries are already starting to develop ways to streamline the pull of CGM data, and approaches to direct EHR integration of discrete CGM metrics have already been presented by the IDC (32) and highlighted in the consensus document on comprehensive CGM EHR integration (33).

CGM Metrics to Guide Management and Serve as Quality and Regulatory Measures

Over the past 15–20 years since the approval of CGM in the United States, as outlined in roadmap steps 1–5, clinical trials and analyses of existing diabetes datasets have shown CGM to be effective in improving glycemic management (i.e., improving A1C or GMI) and achieving more TIR and less TBR (34,35) and correlated with fewer long-term diabetes complications (3,4,36) and reduced diabetes distress (37). After showing CGM to be an effective management tool in clinical trials, the next phase of focused CGM work, happening now, is to achieve the CGM Triple Aim (roadmap steps 6–8). This effort involves developing tools to automate CGM data analysis and decision support (e.g., to suggest adjustments to diabetes medications based on CGM data), adding CGM as an end point in more clinical trials, and establishing CGM metrics as diabetes quality measures. Progress is being made on CGM-guided medication decision-support tools (for both insulin and noninsulin agents) (38), including CGM-guided nutrition selection (39). Personalizing medication and lifestyle adjustments based on CGM metrics, while also addressing diabetes distress and SDOH will start to move us toward a type of precision diabetes management (roadmap step 6). It stands to reason that, if clinicians use CGM metrics to make treatment choices, they will want CGM data to be accepted as outcome metrics in drug comparison trials and to have CGM data displayed in drug prescribing information (roadmap step 7) (40). The third part of the CGM Triple Aim (roadmap step 8) is to establish CGM metrics such as the GMI (41), TIR, TBR, and a measure of diabetes distress as digital quality metrics to be included in quality measurement sets such as HEDIS (the Healthcare Effectiveness Data and Information Set) from organizations such as the National Committee for Quality Assurance (42) and others (43). If these sections of the roadmap are completed, health care organizations and clinicians will be more likely to implement the tools needed to optimize their CGM-derived diabetes quality metrics.

While we were all learning how to organize and analyze standalone CGM data (roadmap steps 1–8), major efforts were also underway to improve CGM sensor accuracy and develop integrated CGM sensors that could be linked to insulin pumps via control algorithms in AID systems (44). Digital diabetes ecosystems and models for effective virtual diabetes care (roadmap step 9) are now established (45,46).

Expanding Indications, New Investigations, and Additional Applications

Many CGM studies are now underway or planned to expand the indications of who may benefit from the use of CGM (roadmap step 10). CGM use in pregnancy is increasing after positive early trials (47–49). Further investigations are needed to establish the value of CGM used early in pregnancy as a predictor of risk for developing gestational diabetes and the optimal use of AID systems during pregnancy. There is also great interest in CGM in the hospital setting, particularly after the difficulties encountered in monitoring glucose levels in the inpatient setting during the coronavirus disease 2019 pandemic. Improved CGM utilization in the inpatient setting also has the potential to mitigate dangers associated with inpatient hypoglycemia and hyperglycemia (50). The use of CGM in the hospital may be of particular importance when admission plasma glucose levels differ substantially from known previous glucose levels (i.e., the admission A1C) (51).

Step 10 of the CGM roadmap also highlights the need for studies to determine whether people with type 2 diabetes who are not using insulin and individuals with prediabetes can benefit from CGM. To date, there has been only one multicenter randomized controlled trial in people with non–insulin-treated type 2 diabetes (52), but almost all the data available from nonrandomized trials, registries, case management programs, and patient and clinician surveys suggest that this cohort is likely to benefit significantly from the personalized lifestyle insights and increased motivation to make healthy changes that CGM can yield (53).

We are just beginning to understand what CGM metrics and glucose profiles might look like for people with prediabetes. We need to know that these metrics and profiles may vary depending on which of the three main methods of diagnosing prediabetes (oral glucose tolerance testing, fasting glucose testing, or A1C) is used. Recently, there has been a call to update the consensus guidance on how best to diagnose prediabetes and also a suggestion to even do away with the imprecise term “prediabetes” and replace it with a calculation of individuals’ personal risk of developing diabetes calculated from glycemic, sociodemographic, and clinical data (54). Perhaps CGM metrics or some type of artificial intelligence (AI) analysis of a CGM profile will prove to be the most accurate assessment of individuals’ true glycemic status and risk for progression.

In pregnancy and in individuals with non–insulin-treated type 2 diabetes or prediabetes, we will need to adopt some of the new nomenclature for CGM metrics currently being introduced. These terms include “time in tight range” (TITR; 70–140 mg/dL) and, for prediabetes, we may even want to evaluate what I suggest we call “time in very tight range” (TIVTR; 70–120 mg/dL). It would then be appropriate to name the currently accepted CGM pregnancy target range “time in pregnancy range” (TIPR; 63–140 mg/dL) with a possible goal of (90% TIPR) and, if tighter glycemic management in pregnancy becomes desirable, we could include “time in tight pregnancy range” (TITPR; 63–120 mg/dL). These new, more specific, metrics, if accepted, would need to be incorporated into AGP reports as a menu of target range options.

The Future Is Now: Exploring New Analytes, Noninvasive Sensors, and AI Data Interpretation

Step 2 outlines an exciting set of planned super highways on the CGM roadmap that many in the field are hoping will be achieved, to some extent, in the not-too-distant future. Adding another metabolic analyte, such as the ketone body β-hydroxybutyrate, which could be continually measured in interstitial fluid along with glucose (and called “continuous ketone monitoring” [CKM]) is showing early promise, and we await pivotal trials (55–57). CKM on its own or combined with CGM in an accurate, affordable system for continuous glucose and ketone monitoring (CGKM) may facilitate the studies needed to demonstrate the safety of and obtain regulatory approval in the United States for the use of sodium–glucose cotransporter 2 inhibitors as adjunctive therapy for people with type 1 diabetes and known heart failure or chronic kidney disease (58) or for individuals with type 1 diabetes who are not reaching their glycemic goals on insulin therapy alone. CGKM may also prove helpful in the management of people with diabetes on very-low-carbohydrate diets, those with recurrent episodes of diabetic ketoacidosis, and those using an AID system who have more frequent insulin infusion site occlusions and failures than is typical.

A spot has also been saved on this CGM roadmap, assuming innovation prevails, for when we someday move from minimally invasive to noninvasive CGM systems. A recent review of the past 2 decades of work on noninvasive CGM (59) concludes that there are four methodologies of noninvasive glucose detection that may have the potential to eventually progress to an “efficient, affordable, accurate, and pain-free” way of monitoring glucose and guiding diabetes management. These four methods are optical spectroscopy, photoacoustic spectroscopy, electromagnetic sensing, and nanomaterial-based sensing. Like almost every other step on the CGM roadmap, any successful innovation would be followed by investigation and then implementation.

Finally, it seems clear that we will be exploring applications of AI to better understand CGM data and how data patterns can interact with genomic, proteomic, and metabolomic data to enhance precision diabetes diagnosis and care (60). We may need to generate additional CGM data–related nomenclature, such as a glycemic subset of metabolomics that could be called “glucomics.”

As these next-generation innovations evolve, we need to refine and better implement the previous 10 steps of the CGM roadmap to help more health care systems achieve the Diabetes Quintuple Aim, which is the ultimate destination for our roadmap to the effective use of CGM.

The Roadmap Destination: Achieving the Diabetes Quintuple Aim With a Focus on Equity

The most respected guide to improving health care systems overall was established in 2008 by Berwick et al. (61). Known as the Healthcare Triple Aim, it includes 1) improving the patient experience, 2) improving the quality of care, and 3) reducing costs. Diabetes is a significant component of every health care system, particularly when we consider that 25% of adults ≥65 years of age have diabetes, that their care costs twice as much as someone without diabetes, and that the overall costs of diabetes care accounts for ∼25% of all U.S. health care dollars spent per year.

Thus, it seems reasonable to establish a parallel Diabetes Triple Aim (Figure 4). The first component of this Diabetes Triple Aim is reducing the burden of people living with diabetes (diabetes distress). The second component is improving the quality of diabetes care, as defined by the CGM metrics of GMI, TIR, and TBR or A1C and including attention to the use of organ-protecting medications for people with diabetes who have known cardiovascular disease, heart failure, or chronic kidney disease. Today, high-quality diabetes care also includes addressing obesity, blood pressure, cholesterol, smoking cessation, and any SDOH that may be impeding optimal care. The third component of the Diabetes Triple Aim is reducing cost, which would best be achieved by overcoming therapeutic inertia and reducing acute and chronic diabetes complications. All three components of the Diabetes Triple Aim are affected by the effective use of CGM.

Figure 4.

Transforming diabetes care step by step: the Diabetes Triple Aim, Quadruple Aim, and Quintuple Aim. Adapted from refs. 9, 61, and 62.

In 2014, the Healthcare Triple Aim was expanded to the Quadruple Aim, with the fourth aim of reducing clinician burden, in recognition of the need for more efficient workflows in clinical practice (62). Efficient workflows are also an essential component of sustainable CGM implementation, as shown in the Diabetes Quadruple Aim (Figure 4). This fourth aim can be addressed in both specialty practice and primary care (63,64) by instituting measures such as a clear process for onboarding and supporting durable use of CGM, integration of CGM in the EHR, professional education on rapid AGP analysis such as the previously described Determine Where to Act method, implementation of CGM-guided decision-making tools, and a support team who can provide timely follow-up and who can use billing codes for remote patient monitoring (65).

Finally in 2022, in recognition of the large gap in equity of care for many medical conditions, including diabetes, a proposal was made for health care systems to move to a Quintuple Aim by adding a specific goal of establishing equity in care (66). Unfortunately, there may be no better example than diabetes to highlight the inequitable nature of care delivered and, of most relevance to this discussion, the vastly different levels of technology prescribing and implemented for people with diabetes based on race/ethnicity or income. Thus, equity is also addressed in the Diabetes Quintuple Aim (Figure 4). Studies are now showing that CGM can be started remotely and managed effectively by telehealth (67), which may broaden the adoption of this important tool to anyone with a smartphone, particularly now that Medicare and Medicaid have expanded CGM coverage. The treatment of other chronic diseases may benefit from laying out a similar roadmap for effective care, as well as defining overarching aims.

It may have taken 25 years, but with repeated cycles of innovation, investigation, and implementation, CGM is helping to transform diabetes management. Let’s aim for the stars as we work together to complete this CGM roadmap and not stop until we achieve all five components of the Diabetes Quintuple Aim: equity, quality, reduced burden for people with diabetes, reduced clinician burden, and reduced cost of diabetes care.