Residency · Residency · Endocrinology

Continuous Glucose Monitoring and Insulin Pump Therapy

Continuous Glucose Monitoring (CGM)

Technology Overview

Continuous glucose monitoring has fundamentally transformed diabetes management by providing near-real-time glucose data that captures the full dynamic range of glycemic excursions invisible to traditional fingerstick monitoring. CGM systems measure interstitial glucose every 1-5 minutes via a small subcutaneous sensor, typically inserted into the abdomen or upper arm. It is important to understand that interstitial glucose lags behind blood glucose by approximately 5-15 minutes, a delay that encompasses both the physiological time required for glucose equilibration between the vascular and interstitial compartments and the processing time of the sensor electrochemistry and algorithm. Modern factory-calibrated systems have largely replaced earlier generations that required multiple daily fingerstick calibrations, dramatically improving user experience and adoption. Sensor glucose accuracy is quantified by the mean absolute relative difference (MARD), which represents the average percentage deviation between sensor readings and simultaneous reference blood glucose values; lower MARD indicates better accuracy, and current state-of-the-art devices achieve MARD values of 8-10%, a level of accuracy that supports clinical decision-making including insulin dosing.

Current CGM Systems

Real-Time CGM (rtCGM)

Real-time CGM systems continuously display glucose data and provide customizable alerts for hypoglycemia and hyperglycemia, enabling proactive rather than reactive glucose management. The Dexcom G7 offers 10-day sensor wear with no calibration requirement, a MARD of 8.2%, a 30-minute warmup period, compatibility with automated insulin delivery (AID) systems, and the ability to display on an Apple Watch. The Dexcom G6, its predecessor with 10-day wear and a MARD of 9.0%, remains widely used particularly in AID systems including Control-IQ, Loop, and AndroidAPS. The Medtronic Guardian 4 features 7-day sensor wear with no calibration, a MARD of 8.7%, and is paired with the MiniMed 780G pump system using the SmartGuard algorithm. The Abbott FreeStyle Libre 3 provides 14-day sensor wear with no calibration, boasts the lowest MARD at 7.9%, offers real-time alerts, is the smallest sensor currently available, and features continuous Bluetooth streaming to the receiver or smartphone. The Senseonics Eversense E3 is uniquely an implantable sensor placed subcutaneously on the upper arm by a clinician, offering 6-month wear with a MARD of 8.5%; it requires an external transmitter adhesive and provides distinctive on-body vibration alerts, making it the only implantable CGM system available.

Intermittently Scanned CGM (isCGM / "Flash Glucose Monitoring")

The FreeStyle Libre 2 represents the intermittently scanned CGM category, offering 14-day wear with a scan-to-read interface supplemented by optional alarms and achieving a MARD of 9.2%. Its lower cost compared to rtCGM has facilitated broader adoption, but it requires deliberate scanning by the user to obtain glucose data, as no passive continuous data stream occurs without active scanning. This limitation reduces its utility for AID system integration and may render it less effective than rtCGM for hypoglycemia prevention in certain populations, particularly those with hypoglycemia unawareness.

CGM Metrics and Ambulatory Glucose Profile (AGP)

International Consensus on Time in Range (2019)

The 2019 International Consensus on Time in Range established standardized CGM-derived metrics that have become primary endpoints in diabetes management alongside HbA1c. | CGM Metric | Definition | Target | Clinical Correlation |

Time in Range (TIR), defined as the percentage of time spent between 70-180 mg/dL, has a target of greater than 70%, which corresponds approximately to an HbA1c of 7.0%. Time below range (TBR) at level 1 (below 70 mg/dL) should be less than 4%, while TBR at level 2 (below 54 mg/dL, representing clinically significant hypoglycemia) should be less than 1%. Time above range (TAR) at level 1 (above 180 mg/dL) should be less than 25%, and TAR at level 2 (above 250 mg/dL) should be less than 5%. The glucose management indicator (GMI) provides an estimated HbA1c derived from CGM mean glucose data using the formula GMI = 3.31 + 0.02392 multiplied by mean glucose in mg/dL. The coefficient of variation (CV) serves as a measure of glycemic variability, with a target of less than 36%; values exceeding 36% indicate unstable glycemia and identify patients at increased risk of hypoglycemia regardless of their mean glucose or HbA1c.

AGP Report Interpretation

The Ambulatory Glucose Profile (AGP) is a standardized 14-day report that distills CGM data into an interpretable visual format. It displays the median glucose as a central line, with the 25th-75th percentile band representing typical glycemic variation and the 5th-95th percentile band capturing outlier excursions. Narrow bands indicate stable, predictable glycemia, while wide bands reflect significant glycemic variability that warrants investigation and intervention. Pattern recognition is a key skill in AGP interpretation, enabling identification of recurrent phenomena such as the dawn phenomenon (early morning glucose rise driven by growth hormone and cortisol surges), postprandial spikes (indicating carbohydrate-insulin mismatch or timing issues), and nocturnal hypoglycemia (which may be asymptomatic and clinically dangerous). The target glucose range of 70-180 mg/dL is shaded on the report for immediate visual reference.

<image>An annotated Ambulatory Glucose Profile (AGP) report showing a 14-day CGM tracing. The x-axis shows time of day (midnight to midnight), y-axis shows glucose in mg/dL (40-350). Display the median line (dark blue), 25th-75th percentile band (medium blue shading), and 5th-95th percentile band (light blue shading). Show the target range (70-180 mg/dL) as a green shaded horizontal band. Annotate key patterns: (1) dawn phenomenon with glucose rising from 5-8 AM, (2) postprandial spike after lunch exceeding 250 mg/dL, (3) overnight hypoglycemia dipping below 54 mg/dL at 3 AM. Below the graph, show a summary statistics panel: TIR 62%, TBR <70 5%, TBR <54 2%, TAR >180 33%, TAR >250 8%, GMI 7.5%, CV 38%, mean glucose 172 mg/dL. Color code metrics that are outside target in red.</image>

Clinical Evidence for CGM

The evidence base supporting CGM use in diabetes management is robust and continues to expand. The DIAMOND trial demonstrated that real-time CGM (Dexcom G4) in type 1 diabetes patients on multiple daily injections (MDI) achieved a 0.6% HbA1c reduction compared to self-monitoring of blood glucose (SMBG) while also reducing hypoglycemia. The GOLD trial similarly showed a 0.4% HbA1c reduction with real-time CGM versus SMBG in type 1 diabetes, along with fewer hypoglycemic episodes. The IMPACT trial evaluated the FreeStyle Libre against SMBG in type 1 diabetes and found no significant HbA1c difference but a striking 38% reduction in time spent below 70 mg/dL, highlighting the particular value of CGM for hypoglycemia reduction even when average glucose remains unchanged. The FLASH-UK trial extended these findings to type 2 diabetes patients on insulin, demonstrating a 0.5% HbA1c reduction with the FreeStyle Libre. Perhaps most compelling was the MOBILE trial, which demonstrated that real-time CGM (Dexcom G6) in type 2 diabetes patients on basal insulin only achieved a remarkable 1.1% HbA1c reduction versus SMBG, with 59% of CGM users achieving TIR greater than 70% compared to only 8% with SMBG. The RELIEF trial provided important real-world evidence from France showing that FreeStyle Libre use was associated with a 50% reduction in acute diabetes events including DKA and hypoglycemia-related hospitalizations.

CGM in Special Populations

CGM technology has demonstrated particular value in several specific clinical contexts. In pregnancy, the Dexcom G7 and G6 are approved for use, and the CONCEPTT trial provided compelling evidence that CGM in type 1 diabetes pregnancy reduced large-for-gestational-age births (53% versus 69%), neonatal hypoglycemia, and neonatal intensive care unit admissions. In the hospitalized setting, evidence is growing but CGM has not yet become standard of care; the FDA approved inpatient CGM use during the COVID-19 pandemic as a strategy to reduce healthcare worker exposure, and ongoing studies continue to define its role. In pediatric patients, CGM is widely used with particular benefit for hypoglycemia detection and the ability for parental remote monitoring. Among elderly patients and those with type 2 diabetes on basal insulin, the MOBILE study supports CGM use, though the cost-effectiveness in this population continues to be debated.

Insulin Pump Therapy (CSII)

Technology Overview

Insulin pump therapy, also known as continuous subcutaneous insulin infusion (CSII), delivers rapid-acting insulin continuously through a subcutaneous catheter, replacing the need for multiple daily injections. The pump provides insulin through two mechanisms: a programmable basal rate that delivers small amounts of insulin continuously throughout the day (which can be set at different rates for each hour to account for circadian variation in insulin requirements) and patient-initiated bolus doses for meals and corrections. The bolus calculator, a sophisticated feature of modern pumps, uses programmed parameters including the insulin-to-carbohydrate ratio (ICR), insulin sensitivity factor (ISF), target glucose, and estimated insulin-on-board to calculate precise bolus recommendations. Infusion sets consist of a cannula (available in Teflon or steel, with 90-degree or angled insertion options) connected to tubing that delivers insulin from the pump reservoir; these sets require replacement every 2-3 days to prevent site-related complications. Common infusion sites include the abdomen (most frequently used due to consistent absorption), upper buttocks, upper arm, and thigh, with systematic rotation essential to prevent lipohypertrophy.

Current Pump Systems

Traditional Pumps (Tubed)

The Medtronic MiniMed 780G is a tubed pump integrated with the Guardian 4 sensor that employs the SmartGuard AID algorithm, which delivers auto-correction boluses every 5 minutes based on CGM data. Users can set adjustable glucose targets ranging from 100-120 mg/dL, and the active insulin time parameter is customizable. The Tandem t:slim X2 with Control-IQ features a touchscreen interface and integration with the Dexcom G6/G7 sensor, using the Control-IQ AID algorithm that incorporates both predictive low-glucose suspend and auto-correction boluses, with dedicated Sleep and Exercise activity modes for different physiological states.

Patch Pumps (Tubeless)

The Omnipod 5 is a tubeless patch pump integrated with the Dexcom G6/G7 sensor, utilizing the SmartAdjust AID algorithm with a customizable glucose target range of 110-150 mg/dL. The auto-mode continuously adjusts basal delivery and provides auto-correction boluses without the encumbrance of tubing. The Omnipod DASH is the non-AID version of the Omnipod platform, controlled via Bluetooth from a phone or personal diabetes manager (PDM).

Emerging Systems

Several next-generation pump systems are reshaping the technological landscape. The Beta Bionics iLet, also known as the Bionic Pancreas, represents a paradigm shift in AID design by requiring only body weight for initialization, completely eliminating the need for ICR and ISF programming. Its adaptive algorithm learns the user's insulin needs over time. The insulin-only version received FDA clearance in 2023, while a dual-hormone version incorporating both insulin and glucagon is in development. The Tandem Mobi is the smallest tubed pump available, featuring AID capability with the Control-IQ algorithm and smartphone control. The Sequel twiist is a next-generation patch pump currently in development.

Pump Advantages vs MDI

Insulin pump therapy offers several advantages over multiple daily injections. Flexible basal rate profiles allow precise management of phenomena such as the dawn effect (by programming higher early-morning basal rates) and exercise-related hypoglycemia (by programming temporary basal reductions). Modern pumps offer dosing increments as small as 0.025-0.05 units, enabling precise titration particularly important in insulin-sensitive children and lean adults. The integrated bolus calculator with insulin-on-board tracking reduces the risk of insulin stacking and simplifies dosing decisions. Pumps generally achieve reduced glycemic variability compared to MDI. The need for only an infusion set change every 2-3 days, compared to 4-6 daily injections, reduces injection burden. Extended bolus and dual-wave bolus features allow insulin delivery to be stretched over time, matching the delayed absorption of high-fat and high-protein meals.

Pump Disadvantages and Risks

Pump therapy carries distinct disadvantages and risks that must be weighed against its benefits. The financial burden is considerable, with pump devices costing $5,000-$8,000 and ongoing supply costs of approximately $1,500-$3,000 annually. Infusion set failures due to occlusion, dislodgement, or cannula kinking can cause rapid hyperglycemia because there is no long-acting insulin depot as a safety net, as exists with MDI using basal insulin. This same absence of a subcutaneous long-acting reservoir means that DKA can develop faster with pump interruption than with MDI. Skin-related issues including adhesive contact dermatitis, lipohypertrophy from repeated site use, and infusion site infections are common complaints. Body image concerns, device burden, and alarm fatigue represent psychosocial challenges that can reduce quality of life for some users. A significant learning curve and the need for ongoing diabetes self-management education are necessary for safe and effective pump use.

Automated Insulin Delivery (AID) / Hybrid Closed-Loop Systems

How AID Works

Automated insulin delivery systems integrate CGM data with algorithmic control of insulin pump delivery, creating a feedback loop that automates basal insulin adjustment. CGM glucose data streams to the algorithm (housed either in the pump processor or a smartphone), which calculates the optimal insulin delivery based on current glucose, glucose trend, predicted trajectory, and estimated insulin-on-board. The algorithm adjusts the basal rate every 5 minutes and may deliver auto-correction boluses when glucose exceeds a threshold. These systems are designated "hybrid" closed-loop because manual meal boluses are still required; a fully closed-loop system that autonomously manages both basal and meal-related insulin delivery is not yet standard. A critical safety feature is predictive low-glucose suspend, which halts insulin delivery when the algorithm predicts impending hypoglycemia and automatically resumes delivery once glucose begins to rise.

Landmark AID Trials

The clinical evidence supporting AID systems is compelling. The landmark 2019 New England Journal of Medicine trial of the Tandem Control-IQ system demonstrated that randomization to the AID system increased TIR from 61% to 71% compared to sensor-augmented pump therapy, with reduced time below range and particularly impressive overnight TIR of 76%, with improvements observed across all age groups. The Omnipod 5 pivotal trials showed TIR improvement from 64.7% to 73.9% in adults, with similar improvements in pediatric populations. The MiniMed 780G studies demonstrated TIR improvement to approximately 75%, enabled by the auto-correction bolus feature that activates every 5 minutes. The iLet (Bionic Pancreas) trial, published in the New England Journal of Medicine in 2022, demonstrated TIR improvement from 65% to 73% and HbA1c reduction from 7.9% to 7.3% using a weight-only initialization approach that eliminates the need for ICR and ISF programming, making AID technology accessible to patients and providers who find traditional pump setup daunting.

AID Optimization Tips

Maximizing the benefit of AID systems requires attention to several practical considerations. Accurate meal bolusing remains critical, as AID algorithms cannot fully compensate for missed or significantly inaccurate boluses. Pre-bolusing 15-20 minutes before meals, when feasible, substantially reduces postprandial glucose spikes. Exercise or Activity mode should be activated proactively, ideally 30-60 minutes before planned physical activity, to allow the algorithm time to reduce insulin delivery before exercise begins. Users should avoid stacking manual correction boluses on top of the algorithm's auto-corrections, which can precipitate hypoglycemia. Accurate carbohydrate counting, potentially aided by carbohydrate counting applications, is essential for optimal bolus accuracy. Infusion site issues must be addressed promptly, as DKA develops more rapidly in pump users without a long-acting insulin depot. The glucose target should be set to the most aggressive setting tolerated, typically 100-110 mg/dL for most adults, to maximize time in range.

<image>A system diagram showing how an automated insulin delivery (AID) system works. Show three connected components: (1) CGM sensor on the body with wireless signal showing glucose data transmission, (2) Algorithm/controller (shown as either a pump processor or smartphone) receiving CGM data and making calculations with labels showing: current glucose, glucose trend, insulin on board, target glucose, and predicted glucose trajectory, (3) Insulin pump delivering adjusted basal rate and auto-correction boluses through infusion set/patch. Show the feedback loop with arrows: CGM → Algorithm → Pump → Patient → CGM. Include a separate arrow showing the patient manually entering meal boluses. Add callout boxes showing specific adjustments: predictive low suspend at 70 mg/dL predicted, increased basal when glucose rising above target, auto-correction bolus when glucose >threshold. Use clean technology illustration style.</image>

Do-It-Yourself (DIY) AID Systems

Overview

The do-it-yourself AID movement emerged from the patient community under the banner of #WeAreNotWaiting, reflecting frustration with the pace of regulatory approval for closed-loop systems. These open-source AID algorithms, developed and maintained by people with diabetes and their families, represent a remarkable intersection of patient advocacy and technological innovation. OpenAPS, the original DIY system, runs on a Raspberry Pi microcomputer and executes its algorithm locally. Loop is an iOS-based system that runs on an iPhone and is compatible with select Medtronic pumps or the Omnipod platform, making it the most widely used DIY system. AndroidAPS provides a comparable solution for Android smartphone users and is compatible with multiple pump platforms. Tidepool Loop is pursuing formal FDA clearance for the Loop algorithm running on the Omnipod, which would bridge the gap between DIY innovation and regulatory approval. None of these systems are currently FDA-approved, and patients who use them assume personal responsibility for their operation. Endocrinologists increasingly encounter patients using these systems and should be prepared to provide guidance. The evidence base, while primarily observational, consistently shows TIR exceeding 70% with reduced hypoglycemia. The CREATE trial, a randomized controlled trial evaluating Loop versus standard care, demonstrated significant TIR improvement with no safety signal.

CGM-Guided Therapy Without Pump

CGM on MDI

The benefits of CGM are by no means limited to pump users. Substantial evidence from trials including DIAMOND and MOBILE demonstrates meaningful glycemic improvement with CGM in patients managing their diabetes with multiple daily injections. The advent of "smart pens" has further enhanced MDI therapy by integrating dose logging, last-dose memory, dose calculator applications, and CGM integration for insulin dosing suggestions. Devices such as the NovoPen 6/Echo Plus and the InPen/Companion Medical pen represent this category, while connected insulin pen caps offer clip-on devices that add dose tracking capabilities to standard insulin pens. The combination of smart insulin pens with CGM is increasingly approaching pump-level glycemic outcomes in motivated patients.

Dosing Decision Support

CGM-integrated applications such as the InPen app with Guardian integration and Dexcom Clarity provide bolus recommendations based on current CGM glucose, trend, and programmed parameters. Pattern analysis tools within these platforms identify recurrent hypoglycemia and hyperglycemia patterns, guiding clinicians in adjusting ICR and ISF values. Insulin titration algorithms embedded in some applications suggest basal insulin dose changes based on CGM data trends, making dose optimization more accessible and systematic.

Practical Considerations

Insurance and Access

CGM coverage is generally available for type 1 diabetes and is increasingly extended to type 2 diabetes patients on insulin therapy. Medicare Part B covers Dexcom and FreeStyle Libre devices for insulin-using patients. Pump coverage is usually available for type 1 diabetes but varies considerably by insurance carrier for type 2 diabetes, often requiring prior authorization. Cost barriers remain significant for uninsured patients, who may face monthly expenses of $300-$600 for CGM supplies and $400-$800 for pump supplies.

Patient Selection for Pump/AID

Appropriate candidates for pump and AID therapy include all patients with type 1 diabetes (who are potential candidates by virtue of their insulin dependence), type 2 diabetes patients on intensive insulin therapy, patients with frequent hypoglycemia or hypoglycemia unawareness, those with a significant dawn phenomenon or glycemic variability, and women planning pregnancy. Relative contraindications include poor engagement with diabetes self-management, inability to perform basic device troubleshooting, severe psychiatric illness that limits safe device use, and unwillingness to wear devices continuously. All pump candidates require comprehensive diabetes self-management education covering carbohydrate counting, alarm troubleshooting, and DKA prevention protocols.

Skin and Adhesive Issues

Contact dermatitis and adhesive reactions represent an increasingly recognized barrier to device adherence. Patch testing may identify specific allergens, with isobornyl acrylate (IBOA) identified as a common allergen in Dexcom sensor adhesives. Mitigation strategies include the application of skin barrier wipes (such as Cavilon or SkinTac), use of undertapes (Tegaderm, Hypafix), application of overlay patches for improved adhesion, and thin application of topical corticosteroids under the sensor. Lipohypertrophy from repeated injection or infusion site use impairs insulin absorption and introduces unpredictable glucose variability; systematic site rotation is essential to prevent this complication.

Key Clinical Pearls

• Time in Range (TIR >70%, 70-180 mg/dL) is now a primary glycemic outcome alongside HbA1c; a 10% increase in TIR corresponds to approximately 0.5-0.8% reduction in HbA1c
• AID systems require accurate meal boluses to perform optimally; the "closed loop" is only closed for basal adjustment, not meal dosing; patient education on carb counting and pre-bolusing remains critical
• CGM benefits are not limited to pump users; the MOBILE trial demonstrated striking HbA1c improvement (1.1%) with CGM in T2D patients on basal insulin only
• The iLet (Bionic Pancreas) eliminates the need for ICR/ISF programming (weight-only initialization), making AID accessible to patients and providers who find traditional pump setup daunting
• Infusion set failure is the Achilles heel of pump therapy; any unexplained hyperglycemia >250 mg/dL should prompt immediate infusion set change and correction by injection (pen or syringe) while troubleshooting
• Coefficient of variation (CV) <36% on CGM is an underappreciated metric; CV >36% indicates high glycemic variability and predicts hypoglycemia risk regardless of mean glucose or HbA1c

References

1. Battelino T, et al. "Clinical Targets for Continuous Glucose Monitoring Data Interpretation: Recommendations from the International Consensus on Time in Range." Diabetes Care. 2019;42(8):1593-1603.
2. Brown SA, et al. "Six-Month Randomized, Multicenter Trial of Closed-Loop Control in Type 1 Diabetes (Control-IQ)." N Engl J Med. 2019;381(18):1707-1717.
3. Martens T, et al. "Effect of Continuous Glucose Monitoring on Glycemic Control in Patients with Type 2 Diabetes Treated with Basal Insulin (MOBILE)." JAMA. 2021;325(22):2262-2272.
4. Russell SJ, et al. "Multicenter, Randomized Trial of a Bionic Pancreas in Type 1 Diabetes." N Engl J Med. 2022;387(13):1161-1172.
5. Holt RIG, et al. "The Management of Type 1 Diabetes in Adults: A Consensus Report by ADA and EASD." Diabetes Care. 2021;44(11):2589-2625.

Read this lecture as Markdown