Regarding “The Performance and Usability of a Factory-Calibrated Flash Glucose Monitoring System,” by Bailey et al.

Dear Editor:

In The Performance and Usability of a Factory-Calibrated Flash Glucose Monitoring System, Bailey et al. ¹ claim that factory-calibrated Abbott FreeStyle Libre continuous blood glucose monitoring systems (CGMs) are sufficiently accurate that patients would not benefit significantly from additional calibrations based on discrete capillary blood glucose (BG) measurements during the course of their use. This is also the position maintained by Abbott Diabetes Care in their customer literature and in on-line communications with users of their LibreLink smartphone application software. ² But the data cited by Bailey et al., ¹ as well as those from my own measurements as a user of both the 10- and 14-day Libre sensors, are inconsistent with this important and highly promoted conclusion.

Bailey et al. ¹ address only the performance of the 14-day Libre sensor that is currently on the market. They do not provide access to the original CGM-BG comparison data upon which they base their analysis, but from the statement in the article that 85.5% of the comparisons lie within the ±20% Zone A of the Clarke Error Grid, and assuming that the data are normally distributed, one can conclude that the overall standard deviation of their CGM-BG comparisons is ∼13%. They also characterize the between-sensor variability in terms of the mean slope of the CGM versus measured BG value as being ∼8.7% (8.5%, 8.6%, and 8.9% for three sensor lots).

Because errors combine as the squares of their standard deviations, the component of the 13% overall standard deviation of an individual CGM-BG comparison that is not due to the 8.7% between-sensor variability is 9.7%. Accordingly, dividing by the square root of the number of observations to determine the standard error of the mean offset for a single sensor, two or more comparisons stand to improve upon the observed 8.7% between-sensor variability. And just one comparison each day, made during times of low interstitial fluid variability (as measured by the CGM) to account for the time lag in the CGM measurement, will reduce the between-sensor uncertainty to ∼2.6% over the 14-day lifetime of these sensors, an improvement of more than threefold.

My own comparisons of both the 10- and 14-day Libre sensors (using either the Abbott reader or the LibreLink smartphone reader) with the highly rated Abbott FreeStyle Lite BG meter ³ have produced somewhat larger but still comparable sensor offsets, sometimes exceeding 30 mg/dL. Such large differences could be the result of sensor mistreatment (e.g., high storage temperature) before purchase, or electronic component failure, but they are nonetheless real-world experiences that could significantly reduce the level of care that users of these sensors achieve.

With the discontinued 10-day Libre sensor, it was possible for users to apply offset corrections based on their own BG measurements using third-party smartphone applications that recognize this need. But the new 14-day Libre sensor uses a proprietary near-field communication protocol that prevents the use of third-party applications, and thus forces users to rely upon the Abbott LibreLink application that does not offer a calibration option.

Although factory calibration of the 14-day Libre sensor is acceptable according to the liberal constraints of the Consensus and Clarke Error Grid metrics used by Bailey et al., ¹ statistics show that these CGMs can perform significantly more accurately with even a moderate amount of additional calibration. Furthermore, the user community is adversely affected by the lack of a calibration option in the Abbott LibreLink application software, and by the use of a proprietary communications protocol in the 14-day Libre sensor that excludes the engagement of third-party application developers.