The Relationship Between Percent Time <70 mg/dL and Percent Time <54 mg/dL Measured by Continuous Glucose Monitoring

Abstract

Objective:

While it is recognized that there is a strong relationship between the amount of time glucose levels are <70 mg/dL (T^<70) and the amount of time <54 mg/dL (T^<54), the association has not been well quantified.

Methods:

Datasets with Dexcom continuous glucose monitoring (CGM) data from nine type 1 diabetes randomized trials were pooled to evaluate the relationship between CGM-measured T^<70 and T^<54. Penalized B-spline regression lines were fitted to assess the relationship between T^<70 and T^<54 for blinded CGM use, unblinded CGM use without an automated insulin delivery (AID) system, and unblinded CGM use with an AID system.

Results:

For blinded data, the T^<54 : T^<70 ratio varied from 19% when the amount of T^<70 was <1% to 44% when the amount of T^<70 was ≥7% whereas for unblinded data the ratio varied from 15% to 42%, respectively. When T^<70 was 4%, the predicted T^<54 was 1.18%, 0.94%, and 0.91% for the blinded, unblinded, and AID data, respectively (P<0.001 comparing blinded versus unblinded and AID).

Conclusions:

The T^<54 : T^<70 ratio increases with greater T^<70, and the ratio generally is higher with blinded than unblinded CGM data, with the latter appearing to be similar to AID system data. The finding of greater T^<54 for a given T^<70 with blinded CGM data is presumed to be due to an action being taken by the unblinded CGM user and/or by the AID system to minimize hypoglycemia which will have the effect of reducing the amount of T^<54.

Introduction

Continuous glucose monitoring (CGM) has provided the ability to assess the frequency of hypoglycemic glucose levels. CGM-measured time <70 mg/dL (T^<70) and time <54 mg/dL (T^<54) have become the standard hypoglycemia metrics for patient care and for clinical trial outcomes. Glucose levels in the range of 54–69 mg/dL are referred to as level 1 hypoglycemia signifying a hypoglycemia alert, whereas glucose levels <54 mg/dL are referred to as level 2 signifying clinically meaningful hypoglycemia due to an association with impaired glucose counter-regulation and reduced hypoglycemia awareness, cognitive function impairment, an increase in cardiac arrhythmias and death, and other effects.¹ The frequency of T^<54 has been shown to be associated with an increased risk of a clinical severe hypoglycemia event.²

Although it is recognized that there is a strong relationship between the amount of time glucose levels are <70 mg/dL and the amount of time <54 mg/dL, the association has not been well quantified. When only T^<70 is known for a patient, it would be useful to be able to estimate the T^<54 as clinical guidelines are being developed based on minimizing T^<54.

We combined data sets from multiple studies to evaluate the relationship between CGM-measured T^<70 and T^<54. We hypothesized that the relationship might differ for blinded versus unblinded CGM and with use of an automated insulin delivery (AID) system.

Methods

We evaluated CGM data from nine studies with 886 type 1 diabetes (T1D) participants and two studies with 173 type 2 diabetes (T2D) participants, using datasets from trials coordinated by the Jaeb Center that have been deidentified and are publicly available at https://public.jaeb.org/datasets/diabetes or by request. The amount of hypoglycemia present in the two T2D data sets (DIAMOND and MOBILE) was too little for a meaningful analysis as the vast majority had <1% T^<70. Therefore, the analyses included only T1D data sets (Supplementary Table S1).

We limited the data sets to those with CGM data from a Dexcom sensor for consistency. Blinded CGM data were obtained from 599 participants using a G4 sensor with 505 software or G5 sensor. Unblinded CGM data without an AID system were obtained from 604 participants using a G4 with 505 software or G5 sensor and 78 participants using a G6 sensor. AID system data from the t:slim X2 insulin pump with Control-IQ Technology (Tandem Diabetes Care, San Diego, CA) and using a G6 sensor were obtained from 190 participants.

The T^<70 and T^<54 was calculated for blinded periods with at least 240 h of CGM data and for unblinded and AID periods with at least 336 h of CGM data. Penalized B-spline regression lines were fitted to assess the relationship between T^<70 and T^<54 for the different types of device use. Bootstrapping was performed to test whether the estimated T^<54 was different for blinded versus unblinded periods and blinded versus AID periods when T^<70 was 4%. Analyses were replicated in subgroups based on age (<18 years and ≥18 years).

Results

The blinded CGM data set included data from 599 participants (92 pediatrics [<18 years] and 507 adults [≥18 years]), the unblinded CGM data set included data from 682 participants (148 pediatrics and 534 adults), and the AID data set included data from 190 participants (109 pediatrics and 81 adults).

Plots of T^<54 versus T^<70 for the blinded, unblinded (without an AID system), and AID data are shown in Figure 1 and the regression lines for all three are shown in Figure 1D. Table 1 summarizes the data to show the ratio of T^<54 to T^<70 in categories based on the amount of T^<70, which is illustrated in Figure 2. It is shown that the T^<54: T^<70 ratio increases with greater T^<70, and the ratio generally is higher with blinded than unblinded CGM data, with the latter appearing to be similar to the AID data.

FIG. 1.

Plots of T^<54 versus T^<70 with smoothing regression lines—for the blinded (A), unblinded (B), AID (C), and all three data (D). Note the axes are changed for (D). When T^<70 is 4%, the predicted values for T^<54 for blinded, unblinded, and AID are 1.18%, 0.94%, and 0.91%, respectively. AID, automated insulin delivery.

FIG. 2.

Plots of means values for T^<54 versus T^<70 for the blinded, unblinded, and AID data. Smoothing regression lines of the ratio of means T^<54/T^<70 according to T^<70 shown.

Table 1.

Ratio of T^<5 ⁴ to T^<7 ⁰ Means in Categories Based on the Amount of T^< ⁷ ⁰

T^<70	Blinded				Unblinded				AID
T^<70	No. of participants	Mean T^<70	Mean T^<54	Ratio of means ^a T^<54/T^<70	No. of participants	Mean T^<70	Mean T^<54	Ratio of means ^a T^<54/T^<70	No. of participants	Mean T^<70	Mean T^<54	Ratio of means ^a T^<54/T^<70
<1%	64	0.5%	0.1%	18.6%	122	0.5%	0.1%	14.6%	74	0.6%	0.1%	13.2%
1%–<2%	71	1.5%	0.4%	27.4%	134	1.5%	0.3%	17.3%	46	1.5%	0.2%	16.0%
2%–<3%	73	2.5%	0.7%	25.9%	121	2.5%	0.5%	19.2%	43	2.4%	0.4%	18.3%
3%–<4%	67	3.5%	1.0%	28.8%	93	3.4%	0.8%	23.1%	14	3.3%	0.7%	21.1%
4%–<5%	72	4.5%	1.3%	29.9%	77	4.5%	1.1%	23.6%	7	4.5%	1.1%	23.4%
5%–<6%	40	5.5%	1.7%	31.5%	37	5.5%	1.6%	29.3%	4	5.5%	1.3%	24.2%
6%–<7%	52	6.6%	2.5%	38.1%	34	6.5%	1.9%	29.4%	2	6.6%	1.7%	25.1%
≥7%	160	11.9%	5.3%	44.1%	64	11.0%	4.6%	41.7%	0

Ratios calculated using exact values and may differ from the ratio of the rounded values.

AID, automated insulin delivery.

For blinded data, the ratio varied from 19% when the amount of T^<70 was <1% to 44% when the amount of T^<70 was ≥7%, whereas for unblinded data the ratio varied from 15% to 42%, respectively. When T^<70 is 4%, the predicted T^<54 is 1.18%, 0.94%, and 0.91% for the blinded, unblinded, and AID groups, respectively (P < 0.001 for both blinded vs. unblinded and blinded vs. AID comparisons). Results for adult and pediatric participants appear in Supplementary Table S2 and results for G4/G5 and G6 users appear in Supplementary Table S3.

Owing to clinical interest, plots of T^<70 versus T^<54 for the blinded, unblinded (without an AID system), and AID data and the regression lines are shown in Supplementary Figure S1. For the blinded group, the T^<70: T^<54 ratio decreased from 5.4 when T^<70 was <1% to 2.6 when T^<70 was 6 to <7%; for unblinded group, ratio decreased from 6.8 to 3.4; and for AID group, ratio decreased from 7.6 to 4.0 (Supplementary Table S4 and Supplementary Fig. S2). When T^<54 is 1%, the predicted T^<70 is 3.6%, 3.9%, and 4.1% for the blinded, unblinded, and AID groups, respectively (P < 0.001 for blinded vs. unblinded and P = 0.01 for blinded vs. AID).

Discussion

T^<54 has become the primary CGM-measured metric for evaluating clinically meaningful hypoglycemia, both for clinical care and as an outcome measure in clinical trials. T^<70 has importance as an indicator of the risk for more significant hypoglycemia. Our analysis quantifies the relationship of T^<70 and T^<54.

We found that the ratio of T^<54 to T^<70 increases as the amount of T^<70 increases for blinded CGM data, unblinded data (without an AID system), and AID system data. We also found that the ratio is higher for blinded CGM data than unblinded or AID system data. This means that for a given amount of T^<70, the amount of T^<54 is significantly higher with blinded CGM than with unblinded CGM with or without an AID system. This is consistent with the preanalysis hypothesis that when the CGM glucose level drops to <70 mg/dL or is predicted to drop to <70 mg/dL, an action is taken by the CGM user and/or by the AID system to minimize hypoglycemia, which will have the effect of reducing the amount of T^<54. This also implies greater T^<70 for a given T^<54 for unblinded CGM or AID system data compared with blinded data. Results were reasonably similar in adults and youth and when restricting to the same type of Dexcom sensor.

The strong relationship between the hypoglycemia metrics has been reported previously. Rodbard found a strong correlation between CGM-measured time <50 mg/dL (T^<50) and time <80 mg/dL (T^<80).³ In addition, greater time spent in hypoglycemia has been associated with greater glucose coefficient of variation.^4,5 Our results suggest that these relationships may slightly differ for blinded CGM, unblinded CGM (without an AID system), and AID system data.

Interpretation of the results should consider that the analysis cohort included only individuals with T1D as there was too little hypoglycemia in the available T2D data sets for a meaningful analysis. In addition, all the CGM data were from Dexcom sensors. The results were similar when repeating analyses for the same version of the Dexcom sensor, but results might not be the same with other sensors.

Clinicians may find these data particularly useful when T^<70 is known but T^<54 is not. In such cases, an estimate of the amount of T^<54 that is likely to be present, or an estimate of the amount of T^<70 before T^<54 will likely exceed the goal of <1%, can be determined based on the current amount of T^<70 and on whether the data were collected with a blinded or unblinded sensor.

Footnotes

Authors' Contributions

R.W.B., P.C., and R.M.B. wrote/edited the article. D.R. performed statistical analysis and wrote/edited the article.

Author Disclosure Statement

R.W.B. reports no personal financial disclosures but reports that his institution has received funding on his behalf as follows: grant funding and study supplies from Tandem Diabetes Care, Beta Bionics, and Dexcom; study supplies from Medtronic, Ascencia, and Roche; consulting fees and study supplies from Eli Lilly and Novo Nordisk; and consulting fees from Insulet, Bigfoot Biomedical, vTv Therapeutics, and Diasome. D.R. and P.C. have no financial disclosures. R.M.B. has received research support, has acted as a consultant, or has been on the scientific advisory board for Abbott Diabetes Care, Ascensia, Bigfoot Biomedical, Inc., CeQur, DexCom, Eli Lilly, Hygieia, Insulet, Medtronic, Novo Nordisk, Onduo, Roche Diabetes Care, Sanofi, United Healthcare, Vertex Pharmaceuticals, and Zealand Pharma. Dr. Bergenstal's employer, nonprofit HealthPartners Institute, contracts for his services for which he receives no personal income.

Funding Information

Supported through the Jaeb Center for Health Research Foundation.

Supplementary Material

Supplementary Table S1

Supplementary Table S2

Supplementary Table S3

Supplementary Table S4

Supplementary Figure S1

Supplementary Figure S2

References

Glucose concentrations of less than 3.0 mmol/l (54 mg/dl) should be reported in clinical trials: A joint position statement of the American diabetes association and the European association for the study of diabetes. Diabetes Care, 2017; 40(1):155–157.

Fiallo-Scharer

, Cheng

, Beck

, et al. Factors predictive of severe hypoglycemia in type 1 diabetes: Analysis from the juvenile diabetes research foundation continuous glucose monitoring randomized control trial dataset. Diabetes Care, 2011; 34(3):586–590.

Rodbard

. Hypo- and hyperglycemia in relation to the mean, standard deviation, coefficient of variation, and nature of the glucose distribution. Diabetes Technol Ther, 2012; 14(10):868–876.

Rodbard

. Glucose time in range, time above range, and time below range depend on mean or median glucose or hba1c, glucose coefficient of variation, and shape of the glucose distribution. Diabetes Technol Ther, 2020; 22(7):492–500.

Toschi

, Slyne

, Sifre

, et al. The relationship between cgm-derived metrics, a1c, and risk of hypoglycemia in older adults with type 1 diabetes. Diabetes Care, 2020; 43(10):2349–2354.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.02 MB

0.01 MB

0.19 MB

0.09 MB