Response to Dr. Seibold's Comment on Puhr et al. (DOI: 10.1089/dia.2019.0360)

Iappreciate the concerns raised by Dr. Seibold in his correspondence regarding our comparison ¹ of continuous glucose monitoring (CGM)-based glycemic outcomes with the Dexcom G5 and G6 CGM systems. I agree that the value of alerts for impending hypoglycemia is of considerable interest and am pleased to respond.

One concern shared by Dr. Seibold is that the sample size in the work in question is different from the sample sizes used in other descriptions of the urgent low soon (ULS) alert. Interested readers are invited to compare the different inclusion criteria and observation windows in the cited works. ^{1 –3} A second of his concerns is our inability to comment on the hypoglycemia (un)awareness of the population. The last sentence of our article mentions the need for further studies in this regard; however, we made no assertions about the utility of the ULS alert as a function of Gold score. A third concern is that our conclusions are “not based solely on actual CGM glucose values” when in fact the only data we analyzed were estimated glucose values provided by our CGM systems, agnostically and automatically uploaded to the Dexcom Cloud.

More interestingly, Dr. Seibold raises the question of whether the hypoglycemia mitigation seen in the G6 data could be attributable to performance differences, such as a tendency of G5 to over-report hypoglycemia or a tendency of G6 to under-report hypoglycemia, or both. Separately conducted accuracy studies in which G5 and G6 data were compared against reference YSI values ^4,5 show the contrary. For YSI values in the 61–80 mg/dL range, 27.9% of G6 values (vs. 22% of G5 values) were erroneously in the 40–60 mg/dL range. If a G5 and a G6 system were worn concurrently, it is likely that the G6 system would report more values in the hypoglycemic range than the G5. Thus, hypoglycemia reductions observed after transitioning from G5 to G6 are unlikely to be attributable to device performance differences. A separate publication in which we compared hypoglycemic exposure among G6 users with the ULS feature enabled or disabled ² also supports the conclusion that use of the ULS feature (rather than simply switching from G5 to G6) mitigates hypoglycemia.

The notion that predictive alerts can help people avoid hypoglycemia was also supported by a recent publication from Medtronic ⁶ using data from their Guardian Connect CGM system. This system's predictive alerts were found to help prevent some real-world hypoglycemic excursions. By contrast, the FreeStyle Libre glucose monitoring system (Abbott), which does not provide alerts in the absence of a user-initiated scan, did not reduce the incidence of severe hypoglycemia in adults with a recent history of clinically significant biochemical or symptomatic hypoglycemia. ⁷

Alerts for impending hypoglycemia are a simple, plausible, and efficient way for people to minimize their exposure to low glucose concentrations, and I remain confident in the data and conclusions of our cited work.