Data Obtained with Early Generations of CGM Sensors: Comment on Pleus et al.

Two recent publications,^1,2 including one in J Diabetes Sci Technol, ¹ from Pleus, Freckmann and colleagues report the outcomes of a single-arm study, comparing the FreeStyle® Libre system against the Dexcom G5® continuous glucose monitoring (CGM) system, in which 24 adults with type 1 diabetes (T1D) wore both sensors in parallel over 8 calendar days. The overall mean absolute relative differences (MARD) of the systems were 12.5% (FreeStyle Libre) and 13.2% (G5) but there was marked intraday variability of MARD, particularly before and after meals. ¹ Both systems were identical in measuring time in range (TIR) 70-180 mg/dL, ² but differed significantly in time below range (TBR) <70 mg/dL, <54 mg/dL and time above range (TAR) >180 mg/dL. They conclude that the analytical performance of both systems was variable as a consequence of activities of daily life, and that differences in measuring %TBR and %TAR may lead to distinct therapy recommendations, with implications for the health of users. We must challenge these conclusions.

Firstly, the FreeStyle Libre sensors being tested in the 2018 study were operating with the earliest generation algorithm for glucose sensing. ³ Pleus et al. ¹ assessed overall MARD of this first-generation algorithm as 12.5%, based on up-to 157 paired readings, compared to the published MARD of 11.4%, based on 12,172 reference pairs. ³ However, the FreeStyle Libre sensor algorithm was updated in 2020, with superior MARD of 9.2% in adults, based on 18,926 paired observations. ⁴

The updated algorithm also invalidates the conclusions reported by Freckmann et al. ² Concurrence data for the earlier generation algorithm indicated that FreeStyle Libre sensors had a low rate of readings in the euglycemic range when reference blood glucose was in the hypoglycemic range, ⁵ and that it was more common that a low-glucose sensor reading would prompt safe and timely action when blood glucose was still euglycemic. This would explain the data presented by Freckmann et al. In contrast, the latest FreeStyle Libre sensor algorithm has improved performance across all glucose ranges, ⁴ such that 98.4% of FreeStyle Libre sensor readings <70 mg/dL are within ± 20 mg/dl of paired blood-glucose reference values and 95% of readings >180 mg/dL are within ± 20% of reference values. Similarly, the lag time between blood and interstitial-fluid glucose has been reduced to 2.4 mins (±4.6). ⁴ These updates have impacted the %TIR, %TBR and %TAR glucose profiles for users such that there is no longer a labelling requirement for users to undertake a confirmatory blood-glucose test for low-glucose readings. This refutes any concerns about patient safety. Pleus’ and Freckmann’s conclusions, based on a single 2018 study,^1,2 are not relevant to comparative clinical decision-making and CGM-sensor choices in 2021. Objective review of differences between CGM devices, including the FreeStyle Libre system, are important to maintain confidence in their efficacy and widespread clinical use in diabetes care. However, as CGM technologies undergo rapid changes and improvements, it is important that the generation of CGM sensors is taken into account when discussing clinical decision making.