Glucose Differences Between Continuous Glucose Monitor Brands and Application Sites

Introduction

Continuous glucose monitoring devices measure glucose levels via a filament inserted in the interstitial fluid. The continuous glucose monitor (CGM) market has quadrupled in the past six years, likely due to improved glycemic control compared to self-monitored blood glucose [1, 2]. Of the over 15 CGM systems with available data, Dexcom G6 (Dexcom Inc., San Diego, CA) and FreeStyle Libre Pro (Abbott Laboratories, Abbott Park, IL) are currently prominent in the market [3].

Growing evidence supports that switching from intermittently scanned continuous glucose monitors (isCGMs) to real-time continuous glucose monitors (rtCGMs) results in improved glycemic control, time in hypoglycemia, and time in range profile [4–7]. The key difference driving these improvements is that isCGM sensors, like FreeStyle Libre Pro, need to be scanned to view glucose values whereas rtCGM sensors, like Dexcom G6, automatically send glucose values to the reader and have optional alerts for hypoglycemic glucose readings. However, the impact that isCGM and rtCGM measurement differences has on these improvements is uncertain. Simultaneously worn FreeStyle Libre Pro and Dexcom G6 would provide insight into CGM brand performance differences as time-matched CGM measurements would be drawing from the same blood glucose value, making the readings directly comparable. Currently, there is limited reported data on glycemic control and time in range from simultaneously worn isCGMs and rtCGMs [8, 9].

Rotating CGM application sites is a common practice to reduce skin irritation; however, its impact on glycemic control interpretation needs further exploration. Recent studies have identified a trend of interarm differences in simultaneously worn FreeStyle Libre Pro CGMs [10–12]. No data could be found on whether this interarm difference translates in the Dexcom G6 CGM. Additionally, a study comparing the Dexcom G6 arm to abdomen CGM placement found no clinically meaningful differences [13]. Brand differences and all three of these application site comparisons will be further explored in this case report. We hypothesize that glucose measurement differences between CGM brands and CGM application sites may affect glycemic control interpretation.

Hypothesis 1: Differences between CGM brands may affect interpretation of glycemic control

We conducted a case study in which two adult males wore a total of five CGM devices each: three Dexcom G6 (upper left arm, upper right arm, abdomen) and two FreeStyle Libre Pro (upper left arm, upper right arm) devices simultaneously. The devices were worn simultaneously for 10 days. Both males were healthy and were not on any medication. Data points were time-matched to a maximum time difference of 5 min. All FreeStyle Libre Pro data from the first 12 h after application was dropped as the warm-up period is known to be less accurate [14]. No readings are given during the Dexcom G6 2-hour calibration period [15]. Guidelines have established CGM clinical targets as time below range (TBR, <70 mg/dL) of <4% per day, an increase in time in range (TIR, 70–180 mg/dL) by 5%, and time above range (TAR, >180 mg/dL) of <25% per day [16].

The glucose measurements in the right arm of Dexcom G6 were higher by an average of 48 ± 13 mg/dL than the glucose measurements in the right arm of FreeStyle Libre Pro (n = 1131, p < 0.001). The glucose measurements in the left arm of Dexcom G6 were higher by an average of 43 ± 17 mg/dL than the glucose measurements in the left arm of FreeStyle Libre Pro (n = 981, p < 0.001) (Figure 1). The right-arm TBR was 0% for Dexcom G6 and 36.43% for FreeStyle Libre Pro. The right-arm TIR was 98.85% for Dexcom G6 and 63.57% for FreeStyle Libre Pro. The left-arm TBR was 0% for Dexcom G6 and 32% for FreeStyle Libre Pro. The left-arm TIR was 98.47% for Dexcom G6 and 68% for FreeStyle Libre Pro. The TAR for each arm is listed in Table 1.

Differences between time-matched, same-arm CGM device readings surpassed clinical benchmarks several times over. Switching from FreeStyle Libre Pro to Dexcom G6 in these two patients would cause hypoglycemia control assessment to be changed from poor to optimal despite no actual change in glucose. This large difference may be explained by differing directions, CGM overestimation versus underestimation, of blood glucose measurements. When compared to the standard-of-care blood glucose measurements, Dexcom G6 showed a mean absolute relative difference (MARD) of 9.3% and FreeStyle Libre Pro showed a MARD of 12.3% [14, 15]. An absolute difference of 9.3% and 12.3% for a blood glucose of 200 mg/dL would translate to 18.6 mg/dL and 24.6 mg/dL differences, respectively. If one CGM overestimated and the other CGM underestimated blood glucose, it would be possible for a 43.2 mg/dL difference between the two sensors for the same blood glucose value. This may be a conservative estimate as this hypothetical calculation used the manufacturer-provided MARD values. At home, the use of CGM devices and lower and higher ends of glucose readings are reported to have higher MARD values relative to the manufacturer MARD estimate [15–17].

Although the sample size is small, a meaningful indicator is evident, which warrants further assessment of this hypothesis. Evaluating differences in average glucose and time in range as a supplement to MARD would provide a more complete picture of CGM brand differences.

Hypothesis 2: Differences between application sites may affect interpretation of glycemic control

Notable differences between FreeStyle Libre Pro glucose readings in the right and left arms have emerged in the recent literature. A study of 72 subjects who wore a FreeStyle Libre on each arm simultaneously reported a mean precision absolute relative difference (PARD) of 8.6% among 49,806 data pairs [15]. This PARD value suggests that differences may exist between arms but does not give a complete picture of how this difference can impact glycemic control interpretation. Three unique studies evaluating simultaneously worn FreeStyle Libre Pro found significant interarm differences in average glucose readings. A study of two subjects and 1920 time-matched CGM data pairs showed that the right-arm glucose reading was higher than the left-arm glucose reading by an average of 8 mg/dL (p ≤ 0.001) [10]. A study of 10 subjects and 9830 time-matched CGM data pairs showed that the right-arm glucose reading was higher than the left-arm glucose reading by an average of 3.8 ± 7.8 mg/dL (p ≤ 0.001). The TBR was 12.5% and 18.4% for the right and left arms, respectively (p ≤ 0.001) [11]. A study of 46 subjects and 41,288 time-matched CGM data pairs showed that the right-arm glucose reading was higher than the left-arm glucose reading by an average of 3.7 ± 10.6 mg/dL (p ≤ 0.001), with a PARD of 9.8%. The TBR was 22.1% and 29.8% for the right and left arms, respectively (p ≤ 0.001) [12]. A combined analysis of the previous studies yielding 54 subjects and 51,118 data pairs reported the right-arm glucose levels to be higher than the left arm with a difference of 3.7 ± 10.2 mg/dL (p < 0.001). The TBR was 20.2% and 27.6% for the right and left arms, respectively (p < 0.001), and the TIR was 76.2% and 69.1% for the right and left arms, respectively (p < 0.001), resulting in a TBR/TIR difference of approximately 7% between the two arms [18]. There is a clear trend across these four studies that the left-arm glucose is consistently lower than the right-arm glucose.

In this case report, this trend was further evaluated and significant interarm FreeStyle Libre Pro differences were seen. Among FreeStyle Libre Pro application sites, the right-arm glucose was higher than the left-arm glucose (n = 1914) by an average of 2 ± 9 mg/dL (p < 0.001). The TBR was 29.31% for the right arm and 38.55% for the left arm. The TIR was 70.69% for the right arm and 61.44% for the left arm. Both arms had a TAR of 0% (Table 1).

This data supports the growing literature that clinically significant interarm differences can exist for FreeStyle Libre Pro, with the right-arm glucose being higher than the left-arm glucose. Across the current literature and this case report, the TBR has been higher in the left arm by a range of 5.9% to 9.24%, well above the 4% TBR clinical target. Exceeding this clinical target means that a patient who switches CGM placement from the left arm to the right arm could go from not well controlled to controlled and vice versa despite no actual change in the glucose level. Differences do not appear to be influenced by interarm differences in muscle mass, arm exercise, fat composition, or dominance [11, 12]. The explanation behind this trend needs further investigation, and there is currently an ongoing study that will further evaluate this trend [19].

This case report additionally explored whether different application sites of Dexcom G6 would affect glycemic control interpretation. Between the three Dexcom G6 application sites, a statistically significant difference was evident (p < 0.001 for ANOVA). The Dexcom G6 average glucose measurements (n = 1642) were higher in the right arm (5 ± 12 mg/dL) and the left arm (6 ± 9 mg/dL) than the abdomen (p < 0.001 for both comparisons). The average difference between right- and left-arm glucose was 0 ± 10 mg/dL (p = 0.59). The TBR was 0% for the right arm, 0% for the left arm, and 0% for the abdomen. The TIR was 99.15% for the right arm, 98.66% for the left arm, and 99.76% for the abdomen. The TAR was 0.91% for the right arm, 1.34% for the left arm, and 0.24% for the abdomen (Table 1).

A previous study using Dexcom G4 Platinum found that the glucose reading accuracy of the upper arms was comparable to the accuracy of the abdomen [20]. Although the arms had significantly higher average glucose readings than the abdomen average glucose readings, application sites of Dexcom G6 did not appear to make substantial changes to CGM clinical targets. Unlike FreeStyle Libre Pro, there were no meaningful differences between right and left arms. Both systems report similar precisions, FreeStyle Libre PARD of 8.6% and Dexcom G6 PARD of 8.9%, so system precision may not be the strongest driving factor behind this discrepancy. Additionally, mechanism of action does not seem to explain this discrepancy either as both systems measure glucose concentrations using a similar enzymatic electrochemical reaction with glucose oxidase [14, 15].

This case report adds to the growing evidence behind the hypothesis that interarm differences between FreeStyle Libre Pro may exist. Although the Dexcom G6 application site did not impact glycemic control interpretation, further exploration on Dexcom G6 arm versus abdomen differences is warranted as a significant difference between average glucose readings was observed.

Discussion and perspectives

Significant differences between CGM brands and right- versus left-arm application site of FreeStyle Libre Pro have been observed. These differences are clinically meaningful, based on the CGM clinical targets and guidelines. The data on CGM brand differences is limited by a small sample size, and data presented from this case report on CGM brand differences should be hypothesis-generating only. Furthermore, this report looked at interbrand and placement site differences, not accuracy, as a standard blood glucose control was not used. Additionally, sensor-to-sensor variability when wearing the CGM on the same arm needs further exploration. Growing data exists on right- versus left-arm glucose differences in FreeStyle Libre Pro, and further studies evaluating this hypothesis are in progress. This case report and the existing literature suggest that implications of FreeStyle Libre Pro interarm differences should be considered when assessing glycemic control in practice and in future clinical trials. Additionally, a notable discrepancy in CGM brand differences was observed, and further study of this hypothesis is required.

Acknowledgment

The authors thank Aydin Shek, PharmD student, for his support on this manuscript.

Conflict of interest

The authors declare no conflict of interest.

Disclaimer

The views expressed in this material are those of the authors and do not reflect the official policy or position of the US Government, the Department of Defense, the Department of the Air Force, or the University of the Pacific.