How Well Do Continuous Glucose Monitoring Systems Perform During Exercise?

Abstract

Continuous glucose monitoring (CGM) has transformed the management of type 1 diabetes (T1D). Devices are now available that monitor glucose without the need for finger sticks for calibration. Use of CGM has been shown to improve time in target range while reducing hemoglobin A1c levels and risk of hypoglycemia.^1
–3 CGM has enabled predictive low glucose suspend systems and automated insulin delivery (AID).^4
–6 These systems are designed to reduce the burden of diabetes management and improve glycemic outcomes. Even with AID, optimal glucose control during exercise has proven to be a challenge. Automated insulin delivery systems are continuing to evolve to better address exercise-related dysglycemia^7,8 and have been tested under very challenging circumstances including adolescents skiing and snowboarding.⁹

Recent data from the T1D Exchange registry demonstrated that the use of CGM has increased substantially, with 38% of individuals in the registry now using CGM.¹⁰ With the expanding use of CGM, it is important to understand the performance of these devices under challenging circumstances, such as exercise. This issue of Diabetes Technology & Therapeutics includes reports of four studies on the use of CGM in people with T1D during exercise.^11

–14 Forlenza et al.¹⁴ present data from a hybrid AID study, which included two aerobic exercise sessions. Zaharieva et al.¹² and Larose et al.¹³ present data on CGM accuracy during aerobic exercise, and Li et al.¹¹ studied the effects of high-intensity interval training (HIIT). A summary of the findings from the latter three studies is presented in Table 1, which highlights that exercise significantly increased the time delay between the reference glucose and the CGM readings.

Table 1.

Summary of Continuous Glucose Monitoring Time Lag and Accuracy Reported in Three Studies^11
–13

Study	Type of exercise	Exercise duration	Lag time	Error grid analysis and ARD	Bias
Li, et al.¹¹ CGM: Dexcom G4 insulin delivery: MDI	High-intensity interval training	25 min	35 min to reach half-maximal glucose	Surveillance error grid: 84.4% to 90.4% in no risk zone before and after exercise; 65.5% during exercise and early recovery (table 2¹¹) Parkes error grid: 85.9% to 91.7% in zone A before and after exercise; 54.6% during exercise (table 2¹¹) Mean ARD: 10.4% before exercise; 17.8% during exercise and early recovery (table 2¹¹)	“Significant bias” of plasma glucose estimation with mean bias of 35.3 mg/dL during exercise and early recovery (table 2¹¹)
Zaharieva et al.¹² CGM: Dexcom G5 or G4 (505 algorithm) insulin delivery: insulin pump	Aerobic on treadmill @ 45%–55% VO_2max; comparison of effects of exercise or meal	60 min × 3 sessions with different basal rates	11.5 min (table 2¹²)	Mean ARD: 7.6% during mealtime; 12.9% during exercise (table 3¹²)	Mean bias—6.6 mg/dL (table 2¹²) Positive bias starting 15 min into exercise until end of exercise, and negative bias during recovery
Larose et al.¹³ CGM: Dexcom G4 (505 algorithm) insulin delivery: insulin pump	Aerobic on ergocycle @ 60% VO_2max	45 min × 3 sessions with different timing of basal rate reductions before exercise start		Median ARD: 8.4% during rest; 16.8% during exercise	Mean bias 14.2 mg/dL during exercise (table 1¹³)Sensor bias persisted for full 45 min of exercise session

ARD, absolute relative difference; CGM, continuous glucose monitoring; MDI, multiple daily injections.

Forlenza et al.¹⁴ present results from a supervised 54-h hybrid AID study in 12 adults with T1D. The AID system used in the study consisted of a modified personal diabetes manager (PDM), Omnipod, Windows 10 tablet running a personalized model predictive control algorithm, and Dexcom G4 CGM system (software 505). To challenge the system, participants performed moderate aerobic exercise on each of two afternoons for ∼40 min. The study tested two strategies to reduce the risk of exercise-related hypoglycemia. Each strategy was applied 90 min before exercise. On the first afternoon, the glucose set point was raised from 130 to 150 mg/dL. On the second afternoon, the basal rate was reduced by 50%. As noted by the authors, the need to make an adjustment 90 min before exercise is not ideal and requires patient engagement. Adjustments were made well in advance of exercise because of the prolonged action of subcutaneous insulin.¹⁵

Three of 12 participants developed hypoglycemia after exercise with the increased glucose set point and 1 of 12 participants with the reduced basal rate. These numbers were likely impacted by providing supplemental carbohydrate before the start of exercise to achieve a glucose level at the time of initiation of exercise >120 mg/dL for five participants on day 1 and for five participants on day 2. There was no nocturnal hypoglycemia on the nights after either intervention. This finding is of importance as glucose levels tend to drop overnight after aerobic exercise, likely due to repletion of muscle glycogen.¹⁶

This study by Forlenza et al.¹⁴ was not powered to assess superiority of one strategy of reducing exercise-related hypoglycemia over another. The mean glucose levels were similar during the ∼40 min of exercise, but dropped to a mean of 100 mg/dL during the 60-min period after exercise when an increased glucose set point was used as compared with a higher mean glucose of 122 mg/dL when a reduction of basal rate was used. There was significant variability across individuals, which may have been related, in part, to the varying types of exercise performed. Overall, these results are suggestive that a reduction in basal rate may be more effective than an increased glucose set point. Both strategies were deemed safe for managing moderate aerobic exercise. A combination of the two strategies might be even more effective for the prevention of hypoglycemia. One would need to optimize the parameters (lead time for initiation before exercise, change in glucose target level (set point), percentage reduction of basal rate) for both types of interventions or for their simultaneous use. Optimal parameters might depend on the subject and the type, duration, and intensity of exercise. A combined strategy was not tested in this study. This pilot study also did not assess other types of exercise.

The glycemic response to exercise varies depending on multiple factors, including type of exercise, intensity, and duration.¹⁷ Therefore, a single type of intervention is not likely to be appropriate for all types of exercise. Resistance exercise and high-intensity intervals often result in a rise in glucose. It might be relatively easy to control exercise-related hyperglycemia using AID, by increasing rates of insulin delivery in response to rising glucose levels. Additional studies are needed to better address this situation. In all, this feasibility study shows very promising results with a mean time in range (70–180 mg/dL) of 85.1% and mean time in level 1 hypoglycemia (<70 mg/dL) of 1.8%. This study adds to the growing experience with and refinement of AID systems being developed for commercialization.

Zaharieva et al.¹² assessed the accuracy of the Dexcom G4 (505 algorithm) (n = 4) and the Dexcom G5 (n = 13) in adults with T1D using insulin pump therapy during and after 60 min of moderate-intensity aerobic exercise on three occasions. Accuracy of the CGM was assessed through capillary blood glucose (CBG) using the built-in FreeStyle glucose meter in the Omnipod PDM. There were four principal findings from their analysis: (1) CGM lagged significantly behind CBG by an average of 12 min, (2) during the 12 episodes of hypoglycemia that occurred as measured by CBG, the corresponding CGM value during these episodes was higher on all occasions (mean glucose was 81 mg/dL by CGM as compared with 60 mg/dL by CBG), (3) there were greater discrepancies between CGM and CBG values at the extremes of glucose levels (<60 and >200 mg/dL), and (4) there was significant variability within estimates of lag and bias. The authors proposed that this variability could be attributed to intersubject variability, errors related to the CGM sensors, or both. The accuracy as measured by mean absolute relative difference (ARD) worsened during exercise (mean ARD of 13% during exercise as compared with 8% during recovery). This decline in accuracy can be largely explained by CGM values lagging behind the CBG values, as shown in figure 2 of their article.¹² However, there were other times where there was significant CGM bias which was not readily explainable by a time lag (figure 1B of their article).¹²

In a separate article in this issue, Larose et al.¹³ performed a secondary analysis of data from 22 people with type 1 diabetes who performed 45 min of moderate aerobic exercise on three occasions. CBG was measured using Contour Next meters. Similar to Zaharieva et al.,¹² Larose et al.¹³ found median ARD increasing from 8.4% before exercise to 16.8% during exercise. One of the key messages from this article is that sensor bias can develop very quickly during exercise. Sensor bias peaked at a mean value of 18 mg/dL within the first 15 min of exercise and persisted until the end of exercise. Unfortunately, this study did not evaluate CGM delay as CGM sensor data were not collected after the exercise period.

Also in this issue, Li et al.¹¹ investigated the accuracy of the Dexcom G4 CGM system as compared with venous plasma glucose in 17 people with T1D being managed with multiple daily injections of insulin. These individuals performed 25 min of HIIT while fasting. As expected, HIIT resulted in hyperglycemia for most patients, reaching a mean glucose of 216.0 mg/dL after exercise. During HIIT exercise, the CGM glucose value lagged behind YSI glucose value with a mean delay of 35 min, and the half-maximal glucose value showed a negative bias of 35.3 mg/dL. In the context of this delay, sensor accuracy worsened during exercise as measured by mean ARD (10.4% before exercise as compared with 17.8% during exercise).¹¹

This series of articles demonstrate that CGM readings lag significantly behind reference glucose both during aerobic exercise when glucose levels tend to fall rapidly and also during HIIT exercise when glucose levels are usually rising. Different methods to obtain reference glucose values were used in each of these studies, and this may have modestly impacted the results. A study comparing venous glucose, arterial–venous glucose, and CBG demonstrated that arterialized venous glucose was significantly higher than venous glucose, but the choice of reference did not significantly impact CGM accuracy.¹⁸

As noted by Larose et al., the influences on sensor bias are multiple and complex. CGM systems are known to be less accurate at times of rapid glucose rate-of-change. For example, Pleus et al. found the mean ARD of the Dexcom G4 system increased from 11.3% to 12.6% when glucose was stable [−1 to 0 mg/(dL·min) and 0 to 1 mg/(dL·min) respectively] to 24.9%–29.6% at times of rapid rates of change [<−3 and >+3 mg/(dL·min) respectively].¹⁹ Although time lag is a major issue impacting accuracy, interstitial fluid glucose is not just a shifted-in-time blood glucose.²⁰ Exercise is also a unique situation as it impacts the volume and fluid distribution within the interstitial compartment, the rate and direction of flow between the vasculature, interstitial fluid, and lymphatics, and affects both noninsulin-mediated and insulin-mediated glucose uptake by exercising tissues. All these effects are likely to contribute to the creation of gradients between interstitial and blood glucose levels.²¹ Exercise stimulates endogenous glucose production,²² and there is another physiologic lag due to glucose transport time between blood and interstitial compartments.^22,23

It is critical for health care providers and patients to understand the potential delay and inaccuracy of CGM during exercise. As Larose et al.¹³ recommend, patients should seriously consider raising the threshold of their hypoglycemia alarm during aerobic exercise, as was used successfully in the study of Forlenza et al.¹⁴ This will not resolve the issue of CGM inaccuracy during HIIT exercise when CGM tends to underestimate rather than overestimate blood glucose. Failure to detect the occurrence of hyperglycemia during HIIT exercise should be less of a safety concern than failure to detect hypoglycemia during aerobic exercise.

The glucose data presented in this series of four articles were predominantly obtained using Dexcom G4 (with and without 505 software) CGM systems, which are less accurate than currently available sensors.^24
–26 Only limited data were obtained using the Dexcom G5. These studies did not examine the accuracy, lag time, and other characteristics of several other currently available sensors such as Dexcom G6, Medtronic Guardian-3, Abbott FreeStyle Libre, or Senseonics Eversense. A study of the accuracy of the Dexcom G6 reported a mean ARD of 9% with only a 4-min lag time.²⁷ This study did not assess the impact of exercise on the accuracy of the Dexcom G6 system. A recent study showed reduced accuracy of the FreeStyle Libre during 55 min of aerobic exercise with a median ARD of 22%.²⁸

Additional studies are required to characterize how well current CGM systems perform during exercise. This is of particular importance as CGM systems are being used by patients to make treatment decisions, so-called nonadjunctive use, and in some cases are also being used to control AID, as in the study of Forlenza et al.¹⁴ The accuracy of CGM systems has dramatically improved over the past decade, and hopefully future advances will further improve accuracy during exercise. The studies of Li,¹¹ Zaharieva,¹² and Larose,¹³ and the results of studies in the future, may provide a basis for developing models to estimate blood glucose levels utilizing the interstitial fluid glucose levels obtained with CGM systems.

Hopefully, a progressively increasing percentage of patients with T1D will be able to use AID systems in the near future. In the meantime, there are pragmatic questions to be addressed for patients using multiple daily injections and insulin pump therapy: how often should they check capillary glucose levels before, during, and after exercise? Should one adjust the insulin basal rate in anticipation of exercise, and if so, by how much, and how long before the exercise? How long should a reduced basal rate remain in effect? In view of the time lag and/or measurement errors during exercise as reported in this series of articles, how should one set the thresholds for hypoglycemia and hyperglycemia alarms, and at what glucose levels should one consume carbohydrate, and how many grams of carbohydrate should be consumed per “unit” of exercise? What thresholds and adjustments should be used by children and adolescents who may be monitored remotely by parents and/or other caregivers? These questions remain to be answered.

In the meantime, patients and their caregivers can use a variety of strategies, including carbohydrate intake before anticipated or scheduled exercise to bring the initial glucose level to a safe level, reduction of basal insulin infusion rate, increasing the glucose set point during exercise and by time of day, and/or elevation of the threshold for hypoglycemia alerts and alarms above the usual 70 mg/dL level. If a person has a highly standardized exercise routine (e.g., running, swimming, or bicycling for a set duration of time, distance, and speed), then it may be possible to fine tune these factors to provide good to excellent glycemic control with minimal risk. When the form of exercise is not highly repeatable, then inputs from accelerometers, pedometers, and heart rate may be more important, and a hybrid closed loop system with an individualized model predictive control algorithm could become even more desirable, and someday may become generally accepted and widely used.

Footnotes

Author Disclosure Statement

J.R.C. has a financial interest in Pacific Diabetes Technologies, Inc., a company that may have a commercial interest in the results of this type of research and technology. J.R.C. has also consulted for Dexcom, Inc. D.R. has consulted for Eli Lilly and Co., Inc.

References

Beck

, Riddlesworth

, Ruedy

, et al.: Effect of continuous glucose monitoring on glycemic control in adults with type 1 diabetes using insulin injections: the DIAMOND randomized clinical trial. JAMA, 2017; 317:371–378.

Haak

, Hanaire

, Ajjan

, et al.: Flash glucose-sensing technology as a replacement for blood glucose monitoring for the management of insulin-treated type 2 diabetes: a multicenter, open-label randomized controlled trial. Diabetes Ther, 2017; 8:55–73.

Olafsdottir

, Polonsky

, Bolinder

, et al.: A randomized clinical trial of the effect of continuous glucose monitoring on nocturnal hypoglycemia, daytime hypoglycemia, glycemic variability, and hypoglycemia confidence in persons with type 1 diabetes treated with multiple daily insulin injections (GOLD-3). Diabetes Technol Ther, 2018; 20:274–284.

Bekiari

, Kitsios

, Thabit

, et al.: Artificial pancreas treatment for outpatients with type 1 diabetes: systematic review and meta-analysis. BMJ, 2018; 361:k1310.

Buckingham

, Raghinaru

, Cameron

, et al.: Predictive low-glucose insulin suspension reduces duration of nocturnal hypoglycemia in children without increasing ketosis. Diabetes Care, 2015; 38:1197–1204.

Forlenza

, Li

, Buckingham

, et al.: Predictive low-glucose suspend reduces hypoglycemia in adults, adolescents, and children with type 1 diabetes in an at-home randomized crossover study: results of the PROLOG trial. Diabetes Care, 2018; 41:2155–2161.

Castle

, El Youssef

, Wilson

, et al.: Randomized outpatient trial of single- and dual-hormone closed-loop systems that adapt to exercise using wearable sensors. Diabetes Care, 2018; 41:1471–1477.

DeBoer

, Chernavvsky

, Topchyan

, et al.: Heart rate informed artificial pancreas system enhances glycemic control during exercise in adolescents with T1D. Pediatr Diabetes, 2017; 18:540–546.

Breton

, Chernavvsky

, Forlenza

, et al.: Closed-loop control during intense prolonged outdoor exercise in adolescents with type 1 diabetes: the artificial pancreas ski study. Diabetes Care, 2017; 40:1644–1650.

10.

Foster

, Beck

, Miller

, et al.: State of type 1 diabetes management and outcomes from the T1D exchange in 2016–2018. Diabetes Technol Ther, 2019; 21:66–72.

11.

, Riddell

, Potashner

, et al.: Time lag and accuracy of continuous glucose monitoring during high intensity interval training in adults with type 1 diabetes. Diabetes Technol Ther, 2019; 21:286–294.

12.

Zaharieva

, Turksoy

, McGaugh

, et al.: Lag time remains with newer real-time continuous glucose monitoring technology during aerobic exercise in adults living with type 1 diabetes. Diabetes Technol Ther, 2019; 21:313–321.

13.

Larose

, Rabassa-Lhoret

, Roy-Fleming

, et al.: Changes in accuracy of continuous glucose monitoring using Dexcom G4 Platinum over the course of moderate intensity aerobic exercise in type 1 diabetes. Diabetes Technol Ther, 2019; 21:364–369.

14.

Forlenza

, Buckingham

, Christiansen

, et al.: Performance of Omnipod personalized model predictive control algorithm with moderate intensity exercise in adults with type 1 diabetes. Diabetes Technol Ther, 2019; 21:265–272.

15.

Home

: The pharmacokinetics and pharmacodynamics of rapid-acting insulin analogues and their clinical consequences. Diabetes Obes Metab, 2012; 14:780–788.

16.

McMahon

, Ferreira

, Ratnam

, et al.: Glucose requirements to maintain euglycemia after moderate-intensity afternoon exercise in adolescents with type 1 diabetes are increased in a biphasic manner. J Clin Endocrinol Metab, 2007; 92:963–968.

17.

Riddell

, Gallen

, Smart

, et al.: Exercise management in type 1 diabetes: a consensus statement. Lancet Diabetes Endocrinol, 2017; 5:377–390.

18.

Kropff

, van Steen

, deGraaff

, et al.: Venous, arterialized-venous, or capillary glucose reference measurements for the accuracy assessment of a continuous glucose monitoring system. Diabetes Technol Ther, 2017; 19:609–617.

19.

Pleus

, Schoemaker

, Morgenstern

, et al.: Rate-of-change dependence of the performance of two CGM systems during induced glucose swings. J Diabetes Sci Technol, 2015; 9:801–807.

20.

Cobelli

, Schiavon

, Dalla Man

, et al.: Interstitial fluid glucose is not just a shifted-in-time but a distorted mirror of blood glucose: insight from an in silico study. Diabetes Technol Ther, 2016; 18:505–511.

21.

Moser

, Yardley

, Bracken

: Interstitial glucose and physical exercise in type 1 diabetes: integrative physiology, technology, and the gap in-between. Nutrients, 2018; 10:E93.

22.

Basu

, Johnson

, Kudva

, et al.: Exercise, hypoglycemia, and type 1 diabetes. Diabetes Technol Ther, 2014; 16:331–337.

23.

Basu

, Dube

, Slama

, et al.: Time lag of glucose from intravascular to interstitial compartment in humans. Diabetes, 2013; 62:4083–4087.

24.

Nakamura

, Balo

: The accuracy and efficacy of the Dexcom G4 platinum continuous glucose monitoring system. J Diabetes Sci Technol, 2015; 9:1021–1026.

25.

Matuleviciene

, Joseph

, Andelin

, et al.: A clinical trial of the accuracy and treatment experience of the Dexcom G4 sensor (Dexcom G4 system) and Enlite sensor (guardian REAL-time system) tested simultaneously in ambulatory patients with type 1 diabetes. Diabetes Technol Ther, 2014; 16:759–767.

26.

Peyser

, Nakamura

, Price

, et al.: Hypoglycemic accuracy and improved low glucose alerts of the latest Dexcom G4 platinum continuous glucose monitoring system. Diabetes Technol Ther, 2015; 17:548–554.

27.

Shah

, Laffel

, Wadwa

, et al.: Performance of a factory-calibrated real-time continuous glucose monitoring system utilizing an automated sensor applicator. Diabetes Technol Ther, 2018; 20:428–433.

28.

Moser

, Eckstein

, Mueller

, et al.: Impact of physical exercise on sensor performance of the FreeStyle Libre intermittently viewed continuous glucose monitoring system in people with type 1 diabetes: a randomized crossover trial. Diabet Med, 2019; 36:606–611.