Background: It is generally known that, during exercise, physiological conditions may lead to significant difference of glucose concentration in the tissue when compared to blood. During several camps, the readings of real-time continuous glucose monitoring (rtCGM) measuring interstitial fluid glucose (IG) were compared to that of fingerstick SMBGs (self-monitoring glucose concentrations in the blood—BG). Deviations up to 40% were recorded at times. The choice of the sensor insertion site may have an impact on the observed deviations. Localized glucose utilization in a specific muscle group may result in marked differences between the IG and BG. A pilot study was done to assess differences in rtCGM accuracy with respect to SMBG when sensors are inserted in different sites during aerobic exercise.
Methods: We enrolled 10 individuals for a 50 mile bike ride. 7 participants had Type 1 Diabetes (T1D), and 3 were healthy volunteers. Study participants wore 3 Dexcom G6 sensors simultaneously, inserted (A) in the upper arm, (B) in the abdomen, (C) in the thigh. Patients were required to take a fingerstick at the beginning of the ride, and every 30 minutes during the ride. For improving the accuracy of SMBG, a BG was taken twice from the same sample. A total of 11 samples per participant were collected. Accuracy was assessed in terms of Mean Absolute Relative Difference (MARD). In addition, the average and standard deviation (SD) of the errors of all samples (Bias) were shown. Diabetes therapy was provided during the study via the patient’s own rtCGM sensor.
Results: A summary of the participants’ characteristics and the accuracy metrics is reported in the Table 1—MARD was different for the 3 different sites, with the arm being the most accurate. Our results indicate that wearing the sensor on the abdomen or the thigh during aerobic exercise (bike riding) may result in sensor readings that seem to have a minimal low bias, as can be seen from Table 1 where the Average % of deviation (Bias) is reported. The Bias was doubled and negative, which means that the CGM values are too low compared to the BG.
Table 1.
Participants’ characteristics and accuracy metrics of different sensor wearing sites.
| Participants | T1D | Healthy volunteers | ||||
|---|---|---|---|---|---|---|
| Characteristics | 3 Female/4 male | 2 Female/1 male | ||||
| 4 Open artificial pancreas system | ||||||
| 3 with CGM supported pump therapy | ||||||
| Age 46 ± 10 years, BMI 23 ± 3 kg/m2, HbA1c 6.3 ± 0.8% | Age 39 ± 10 years, BMI 22 ± 3 kg/m2, HbA1c 5.3 ± 0.2% | |||||
| Location | Arm | Abdomen | Thigh | Arm | Abdomen | Thigh |
| Number of readings | 77 | 33 | ||||
| MARD % | 11 | 15 | 13 | 10 | 12 | 22 |
| Bias % | 4.05 | −9.9 | −9.3 | −9.8 | −8.7 | −23.5 |
| SD | 18.1 | 20.4 | 14.9 | 9.3 | 13.2 | 16 |
Accuracy was assessed in terms of Mean Absolute Relative Difference (MARD)
Conclusion: Our pilot study indicates that during aerobic exercise involving sustained bike riding, the rtCGM worn on the arm provided the most accurate readings. This may help patients in the choice of the best site insertion, when they plan to go for exercise. Wearing the sensor on the thigh is not recommended as it provides the least accurate readings (note: the thigh is not an approved sensor wear site per the rtCGM label). No significant difference was found in the sensor accuracy between healthy participants and participants with T1D. In our study, the MARD values in the area of the upper arm and abdominal area were not so different from the values reported by the manufacturer.
Guillot et al. 1 shows very similar deviations in aerobic and anaerobic sessions over 30 minutes in his exercise study. But the sensor position is not specified. Also in the study of Li et al., 2 deviations during aerobic exercise of up to 18% during the training and 11% in the phase 45 minutes afterwards are named. In a study of pregnant women in the 1st trimester, 3 the MARD values were calculated as a function of the different sensor positions with abdomen, upper buttock and upper arm, in which the upper arm also showed up as the most accurate with 8.7%.
Footnotes
Abbreviations: BG, glucose concentrations in the blood; Bias, errors of all samples; IG, interstitial fluid glucose; MARD, mean absolute relative difference; rtCGM, real-time continuous glucose monitoring; SD, standard deviation; SMBG, self-monitoring glucose concentrations in the blood; T1D, diabetes mellitus type 1.
Declaration of Conflicting Interests: The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding: The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Dexcom
ORCID iDs: Michael Müller-Korbsch 
https://orcid.org/0000-0002-5815-8279
Gersina Rega-Kaun, MD
https://orcid.org/0000-0002-3110-481X
References
- 1. Guillot FH, Jacobs PG, Wilson LM, et al. Accuracy of the Dexcom G6 glucose sensor during aerobic, resistance, and interval exercise in adults with type 1 diabetes. Biosensors. 2020;10(10):138. doi: 10.3390/bios10100138 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Li A, Riddell MC, Potashner D, Brown RE, Aronson R. Time lag and accuracy of continuous glucose monitoring during high intensity interval training in adults with type 1 diabetes. Diabetes Technol Ther. 2019;21(5):286-294. doi: 10.1089/dia.2018.0387 [DOI] [PubMed] [Google Scholar]
- 3. Castorino K, Polsky S, O’Malley G, et al. Performance of the Dexcom G6 continuous glucose monitoring system in pregnant women with diabetes. Diabetes Technol Ther. 2020;22(12):943-947. doi: 10.1089/dia.2020.0085 [DOI] [PMC free article] [PubMed] [Google Scholar]