Response to Seibold: Data Obtained With Early Generations of CGM Sensors: Comment on Pleus et al.

In his comment on 2 of our works, Seibold ¹ questions the applicability of our study’s results to the current use of continuous glucose monitoring (CGM) systems in light of updated products. In his opinion, this invalidates the findings of our studies.

Generally speaking, specific product updates can affect sensor performance. Alva et al., ² whom Seibold references as claim for a 2020 update of the FreeStyle Libre system, do not make explicitly clear whether the update affected only FreeStyle Libre 2 or also the first-generation system that was used in our studies. In mid-2019, an update for the first-generation system was released, which could be related to the study results reported by Alva et al., as the clinical trials seemed to have been performed before the mid-2019 update.

In addition, information is missing regarding where sensor currents are converted to glucose concentrations. This could occur either in the sensors, which are replaced every 14 days, or in the readers, which can last for years. If conversion occurs in the readers or smartphone app, a newly introduced algorithm might only spread slowly among established users. The aforementioned mid-2019 update of the first-generation system, for example, required users to update their readers or apps to allow better system performance, implying that conversion occurs in the readers or apps. With smartphone apps, there is a risk of system requirements (version of the operating system) becoming an issue and users may have to decide between updating the app and keeping their current phone. Sensors for the first-generation system are still available at least in the German webshop. ³

More importantly, however, Seibold fails to substantiate his claim that with the new algorithm, the differences in clinically relevant metrics and the variability in performance throughout the day do no longer occur. Since the technology used in different CGM systems typically is proprietary information, the expectation that there no longer exist noticeable or clinically relevant differences in performance between CGM systems should be backed up by Seibold. Such substantiation is especially relevant in light of 2 recent publications: Nørgaard et al. ⁴ found relevant differences in time spent in different glucose ranges when pregnant women simultaneously wore the first-generation system (2019-updated version) and a real-time CGM system. Szadkowska et al. ⁵ observed significant differences in performance based on glucose rates of change even with the 2019-updated version of the first-generation system. Glucose rates of change are affected by meal intake and physical activity and thus can vary during the day, which we identified as influencing factors in our analysis.

Although we certainly can understand Seibold’s argument that “the generation of CGM sensors [has to be] taken into account when discussing clinical decision making”, we do not fully agree. With increasingly short life cycles of medical devices, this approach would render manufacturer-independent research pointless. Furthermore, as algorithms are proprietary and CGM metrics are increasingly often used for modifications of diabetes therapy, and in absence of internationally recognized standards defining minimum criteria for CGM performance, manufacturer-independent research is still essential.