Journal of the Endocrine Society Study Examines Optimal Hypoglycemia and Hyperglycemia Alarm Thresholds in 95 Type 1s – January 21, 2020

Average Dexcom CGM user had hypo alarm threshold of 73 mg/dl, hyper threshold of 205 mg/dl; optimal hypo threshold of 75 mg/dl, hyper: 170 mg/dl

The Journal of the Endocrine Society published “Alarm Settings of Continuous Glucose Monitoring Systems and Associations to Glucose Outcomes in Type 1 Diabetes,” in late 2019, estimating that the ideal thresholds for hypoglycemia and hyperglycemia alarms were 75 mg/dl and 170 mg/dl, respectively. The observational study took place at the University of Utah and included 95 type 1 patients using Dexcom G4 (7%), G5 (78%), and G6 (15%). At the one time-point data was collected, average Time-in-Range and A1c were 62.8% and 7.4%, respectively. Times spent above 180 mg/dl and 250 mg/dl were 33.6% and 11.2%, and time spent below 70 mg/dl and 54 mg/dl were 3.9% and 1.1%. The mean hypoglycemia alarm thresholds set by users was 73 mg/dl; participants with their alarms set ≥73 mg/dl spent lower percentages of time below 70 mg/dl than those with alarms set <73 mg/dl (2.3% vs. 4.7%; p=0.005) – that is the equivalent of nearly 35 minutes per day! This is a big difference and a good “hack” for patients who aren’t troubled by alarms – Kelly looked at her alerts and found 70 mg/dL and 180 mg/dL, and because the next “level” she could select on the G6 was 75 mg/dL, she did, and she will report back. She also changed the 180 mg/dL alarm to 170 mg/dL.

Notably, participants with alarms set ≥73 mg/dl also spent less time <54 mg/dl (0.5% vs. 1.5%; p=0.016) – that’s the 14 minutes per day. Somewhat surprisingly, the participants with higher hypo alarm thresholds (≥73 mg/dl) also had higher mean glucose levels (175 vs. 153 mg/dl) and lower Time in Range (57% vs. 68%) – according to the authors, this differential ws due to the more time they spent in hyperglycemia as a “price” of hypoglycemia reduction.

On the hyperglycemia end, 83 participants had hyperglycemia alarms enabled with a mean threshold of 205 mg/dl. As we would imagine would be expected, those with alarms set below 205 mg/dl spent less time >180 mg/dl (30% vs. 37%; p=0.01) and had a lower mean glucose level (155 vs. 172 mg/dl; p=0.02). The Time-in-Range between patients with lower and higher thresholds of hyperglycemia (i.e. <205 mg/dl vs. >205 mg/dl) was statistically indistinguishable.

Over the two-week study period, CGMs generated ~1.4 alarms/day, 42% of which were hypoglycemia alarms. Using receiver operating characteristic (ROC) curves, the authors identified optimal hypo- and hyperglycemia alarm thresholds. No optimal hypo alarm threshold was found for patients to achieve the <4% time below 70 mg/dl consensus goal – the authors state that other techniques are likely needed for the population to reach this goal. However, the optimal threshold of 75 mg/dl was found for achieving more stringent goals of <2% and <1% time below 70 mg/dl. Similarly, no optimal alarm setting was found to reach the <25% time above 180 mg/dl consensus goal; however, 170 mg/dl was an optimal threshold for the more stringent <15% and <5% time above 180 mg/dl goals. The authors did warn against reducing hyperglycemia alarm thresholds too far for fear of alarm fatigue. In the study cohort, participants with hyperglycemia alarm thresholds <205 mg/dl had more than double the number of hyperglycemia alarms as those with thresholds ≥205 mg/dl (2.2 vs. 1 alarm/day). While that is understandable, we’re slightly surprised that anyone chose a hyperglycemia alarm above the target – we would assume that better alarms are being created that are easier to respond to.

Broader patient-reported outcomes were not measured in this study. As this is an observational study, the effects of the alarms are not confirmed, and a prospective clinical trial would be required to effectively evaluate the ideal alarm settings in patients.


--by Ani Gururaj, Albert Cai, and Kelly Close