The Development of New Composite Metrics for the Comprehensive Analytic and Visual Assessment of Hypoglycemia Using the Hypo-Triad

Abstract

Background:

Quantifying hypoglycemia has traditionally been limited to using the frequency of hypoglycemic events during a given time interval using data from blood glucose (BG) testing. However, continuous glucose monitoring (CGM) captures three parameters—a Hypo-Triad—unavailable with BG monitoring that can be used to better characterize hypoglycemia: area under the curve (AUC), time (duration of hypoglycemia), and frequency of daily episodes below a specified threshold.

Methods:

We developed two new analytic metrics to enhance the traditional Hypo-Triad of CGM-derived data to more effectively capture the intensity of hypoglycemia (IntHypo) and overall hypoglycemic environment called the “hypoglycemia risk volume” (HypoRV). We reanalyzed the CGM data from the ASPIRE In-Home study, a randomized, controlled trial of a sensor-integrated pump system with a low glucose threshold suspend feature (SIP+TS), using these new metrics and compared them to standard metrics of hypoglycemia.

Results:

IntHypo and HypoRV provide additional insights into the benefit of a SIP+TS system on glycemic exposure when compared to the standard reporting methods. In addition, the visual display of these parameters provides a unique and intuitive way to understand the impact of a diabetes intervention on a cohort of subjects as well as on individual patients.

Conclusion:

The IntHypo and HypoRV are new and enhanced ways of analyzing CGM-derived data in diabetes intervention studies which could lead to new insights in diabetes management. They require validation using existing, ongoing, or planned studies to determine whether they are superior to existing metrics.

Keywords

composite metric continuous glucose monitoring hypoglycemia diabetes

Self-monitoring of blood glucose (SMBG) is the standard method by which the vast majority of patients assess their diabetes control. By virtue of the episodic nature, the limited number of times per day that it is actually performed, and the infrequent testing at night, SMBG can provide only a partial view of the frequency and severity of hypoglycemia. While multiple studies have used the nadir of the glucose level to differentiate between mild and severe hypoglycemia,^1-3 it is not possible to infer the intensity of hypoglycemia from SMBG because the duration of hypoglycemia is not known. This limits our ability to interpret the effect of various interventions for improving glycemic control as well as our understanding of the short- and long-term risks associated with hypoglycemia. On the other hand, continuous glucose monitoring (CGM) collects data on the frequency, duration and severity of hypoglycemia whether or not it is symptomatic.^4-6 The hypoglycemia triad (Hypo-Triad) consists of the three metrics that are usually reported in trials using CGM—area under the curve (AUC), time in hypoglycemia, and frequency of hypoglycemic excursions per day. However, it is unclear which individual metric or combination of metrics of the Hypo-Triad best characterizes the clinical and pathophysiologic impact of hypoglycemia. Figure 1 shows an example of a CGM tracing with two hypoglycemic episodes with different characteristics. Therefore, we developed two new CGM-derived metrics from this Hypo-Triad—intensity of hypoglycemia (IntHypo) and hypoglycemia risk volume (HypoRV). This approach provides both a numeric value and a clear visual representation of hypoglycemia which is missing when using standard metrics. Using the data from the ASPIRE In-Home trial we compared these new metrics to the standard reported metrics of hypoglycemia.

Figure 1.

Example of a CGM tracing showing two episodes of hypoglycemia with different clinical manifestations. The symptomatic hypoglycemia at 14:30 is significantly longer and shows a significantly higher AUC than the mild hypoglycemia around 18:30. C_gluk;min = nadir of sensor glucose; T = time <70 mg/dL; AUC = area under curve <70 mg/dL.

Methods

We used the existing CGM data from the ASPIRE In-Home trial to assess the new metrics and then compared them to the standard hypoglycemic metrics as reported in that study.⁷ The design and results of this trial have been previously reported. In summary, the ASPIRE In-Home trial was a multicenter, randomized, controlled, 3-month study of the effectiveness of a sensor-integrated pump system which had a low glucose threshold suspend feature (SIP+TS) compared a sensor-augmented pump system which lacked the threshold suspend feature (SAP-TS) in reducing nocturnal hypoglycemia in 247 patients with type 1 diabetes. Subjects were eligible for randomization if they had ≥2 episodes of nocturnal hypoglycemia at baseline determined by masked CGM. The use of SIP+TS resulted in a decrease in AUC <70 mg/dL of 38% at night and 31% all day compared to the SAP-TS group (Table 1). There was a 32% reduction in the frequency of hypoglycemia <70 mg/dL at night (10 pm to 8 am) and a 30% reduction all day. There was no significant change in hemoglobin A1C (A1C) in either group.

Table 1.

Comparison of the New Versus Traditional Metrics of Hypoglycemia in Analyzing the Results of the ASPIRE In-Home Trial.

	IntHypo <70 mg/dL (%)	HypoRV <70 mg/dL (%)	AUC <70 mg/dL (%)	Frequency of episodes per day <70 mg/dL (%)	Time <70 mg/dL (%)
SIP+TS	–44.8	–64.3	–30.7	–34	–50.7
SAP-TS	–5.2	–3.5	+6.3	–7.8	–7.4
Absolute difference	39.7	60.8	37.0	26.6	43.3

SIP+TS = sensor integrated pump system with low glucose threshold suspend feature; SAP-TS = sensor-augmented pump system without low glucose threshold suspend feature; AUC = area under the curve; IntHypo = intensity of hypoglycemia; HypoRV = hypoglycemia risk volume.

The Hypo-Triad-Based Model

The three elements of the Hypo-Triad used in this model are the mean AUC in the glucose range <70 mg/dL, the mean time <70 mg/dL, and the mean rate of hypoglycemia <70 mg/day. We defined the intensity of hypoglycemia (IntHypo) as a combination of AUC and the time per day of glucose values <70mg/dL. This creates a vector spanning the surface of both parameters in a Cartesian coordinate system (red arrow in Figure 2A). The magnitude of this vector is calculated as:

Figure 2.

The intensity of hypoglycemia (IntHypo) is the vector between the AUC and time in the glucose range <70 mg /dL (A). The hypoglycemia risk volume (HypoRV) is expressed as the product of IntHypo vector and the rate of hypoglycemia per day (B).

IntHypo = | AUCxTime | = . \sqrt{{AUC}^{2} {+ Time}^{2}}

The third element of the Hypo-Triad is the frequency of hypoglycemic episodes per day. Incorporating this into the Cartesian coordinate system results in a hypoglycemia risk volume (“HypoRV”) (Figure 2B green circle). We presume that the IntHypo represents the effect of hypoglycemia on glucose counterregulation including stimulation of glucagon and epinephrine, the depletion of glycogen stores, and the mitigation of glycemic variability. We postulate that the HypoRV represents the overall experience of hypoglycemia or the hypoglycemic environment. This is demonstrated in Figure 3 with hypoglycemia with high intensity but low frequency. We postulate that high-intensity/low-frequency hypoglycemia is potentially more deleterious than hypoglycemia with low intensity but high frequency. Hypoglycemia of high intensity and low frequency represents clinically “severe” hypoglycemia which is associated with an increased the risk of fatal events.⁸ We also investigated which of the hypoglycemic metrics correlates best with the coefficient of variation (CV), a marker of glycemic variability, where CV = 100 × (mean value / standard deviation).⁹

Figure 3.

Graphic representation of the low-intensity/high-frequency (A) and high-intensity/low-frequency (B) hypoglycemia. The putative effect of low intensity-high frequency hypoglycemia on glucose counter-regulation, restoration of glycogen stores, and glycemic variability is smaller than that of high-intensity/low-frequency hypoglycemia.

Results: Evaluation of Data From the ASPIRE In-Home Study With the Hypo-Triad Model

The SIP+TS system significantly reduced all-day IntHypo by 44.8% from 1620 ± 1397 mg/dL × min² to 895 ± 676 mg/dL × min² (P < .001) and the all-day HypoRV by 64.3% from 1935 ± 2989 mg/dL × min² to 690 ± 975 mg/dL × min² (P < .001) in the ASPIRE In-Home study. In the SAP-TS cohort, both parameters decreased only slightly by 5.2% (IntHypo) and 3.5% (HypoRV) (Figures 4 and 5). The frequency of excursions events decreased by 37.6% (5.0 to 3.3) and 20.2% (5.1 to 4.7) in the SIP+TS and SAP-TS, respectively. We compared the results using these new metrics to those presented in the original report (Table 1). The IntHypo and HypoRV provide supplementary and, arguably, more clinically important information about the results of this trial. In particular, the HypoRV demonstrates that the SIP+TS had an even more profound effect on the hypoglycemic environment that the subjects experienced over the 3 months of the trial than originally reported. The correlation with CV was the highest with IntHypo <70 mg /dl (r = .78), intermediate with AUC <70 mg /dl (r = .74), and smallest with time <70 mg /dl (r = .66) and the HypoRV (r = .64).

Figure 4.

Comparison of IntHypo (left) and HypoRV (right) in patients from the ASPIRE In-Home study: SAP-LGS = sensor-augmented pump without low glucose suspend also referred to as threshold suspend (SAP-TS) in the text; SAP+LGS is sensor-augmented pump with low glucose suspend also referred to as sensor-integrated pump with threshold suspend (SIP+TS) in the text. The data were compared at the beginning (baseline) and at the end of the investigation (month 3).

Figure 5.

Group data of all-day IntHypo and HypoRV at beginning and end of study in patients from the ASPIRE In-Home study. Top: Control group (SAP-TS). Below: Intervention group (SIP+TS). The data of the entire groups are presented at the beginning and at the end of the study.

The IntHypo and HypoRV of three patients from the SIP+TS cohort are shown in Figure 6 to illustrate the diversity of responses that were seen in the ASPIRE In-Home study and how individual responses can be visually captured. In the subject in Figure 6 Top, the use of the low glucose threshold suspend system resulted in large and highly significant decrease of both parameters. The subject in Figure 6 Middle has a modest reduction of the time per day and frequency of excursions <70 mg/dL with a slight increase in the AUC <70 mg/dL. This results in a small increase in the IntHypo and a small reduction of HypoRV. Finally, the subject in Figure 6 Bottom illustrates a case where the IntHypo decreases but the HypoRV increases the latter driven by an increase in the frequency of excursions per day <70 mg/dL from 0.28 to 0.36.

Figure 6.

Examples of patients from the intervention group (SIP+TS) of the ASPIRE In-Home trials. Top: Patient has a marked decrease of IntHypo and HypoRV. Middle: HypoRV improves but IntHypo is slightly increased. Bottom: IntHypo is decreased but HypoRV is slightly increased. Note that each image has a different scale.

Discussion

We developed a new approach that integrates the three standard CGM-derived metrics of AUC, duration of hypoglycemia and frequency of excursions per day (the Hypo-Triad) into two composite metrics. These new metrics are the intensity of hypoglycemia (IntHypo) and risk volume of hypoglycemia (HypoRV). We have taken this approach because these new metrics can clarify our understanding of hypoglycemia (and potentially hyperglycemia) that are not possible using more standard metrics such as AUC and percentage hypoglycemia. The AUC represents the integral over glucose values over the time. If the glucose level was constant and the AUC is known, the amount of time can be derived. However, glucose levels constantly vary. Under those circumstances, there are situations where equivalent AUCs have very different times.

We initially thought that it would be best to combine all three metrics to define a space vector leading from the zero point to the top corner of HypoRV. However, we felt that the most important consideration was the impact of the metric on the glucose regulation particularly glucose variability (GV) The frequency of hypoglycemia has a small effect on GV especially if the hypoglycemia has a low intensity. If the frequency of hypoglycemia is high, the risk of these events is high (they occur with high probability). But that doesn’t mean that the impact on GV is high. In addition, we selected AUC and time as the components of IntHypo rather than other combinations, for example, AUC and frequency or time and frequency, because physiologically high intensity hypoglycemia is mitigated by hepatic and renal endogenous glucose production assuming adequate glycogen stores. This occurs during high intensity hypoglycemia especially if these episodes have a long duration. Thus, the combination of AUC and time determines the intensity of hypoglycemia. This combination has also the highest correlation to the %CV (r = .75-.80). However, a vector of frequency of hypoglycemia × time correspond to the total time in the hypoglycemic range. Its correlation to %CV is lower (r = .5-.6) as is the metric of frequency of hypoglycemia × AUC (r = .5-.6). Therefore, we believe that the IntHypo regarding the impact on the stability of glucose and HypoRV as a risk for hypoglycemia is the most appropriate combination of metrics to describe intensity.

Consequently, these CGM-derived metrics provide a more in-depth understanding of the significance of hypoglycemia than the traditional method of displaying the data. Traditional metrics such as AUC and frequency of hypoglycemic events can be misleading. For example, mild hypoglycemia over 2 hours may have the same AUC as severe hypoglycemia over 30 minutes. Yet, the effect on a patient’s clinical symptoms and counter-regulatory responses is likely to differ.

It appears that the best metric for describing the immediate clinical impact of hypoglycemia is the intensity of hypoglycemia (IntHypo), while the HypoRV, which includes the frequency of hypoglycemia events, provides additional context about the overall hypoglycemic experience. Thus, we believe that these new metrics are complementary to the traditional metrics and advance our understanding of the effect of various interventions on the hypoglycemia environment. For example, by analyzing the data from the ASPIRE In-Home study with this approach, we found that each of these new metrics demonstrated a greater impact of a low glucose threshold suspend system when compared to traditional metrics. The apparent enhanced sensitivity of these metrics, particularly the HypoRV, in reflecting the hypoglycemia experience may be due to the use of multiple (composite) metrics drawn from the traditional metrics of hypoglycemia. At the present time, we do not know the clinical implications, if any, of displaying the data in this new way. For example, these new metrics may be more predictive of the development of the severe short-term complications of hypoglycemia like cardiac arrhythmias. Studies combining continuous glucose and cardiac rhythm monitoring have shown that 62% of nighttime hypoglycemic episodes were associated with cardiac arrhythmias and prolongation of the QT interval.^10,11 These cardiac arrhythmias may lead to the so-called “dead-in-bed” syndrome.^12-14 Validation of these new metrics requires their application to existing, ongoing, or future studies of the clinical effects of hypoglycemia to determine if they are superior to traditional metrics.

The relationship of glycemic variability to long-term complications is controversial.^15-17 Some believe the oxidative stress engendered by glycemic variability increases the risk cardiovascular events.^18,19 These new metrics may prove to be more relevant markers for analyzing the effects of hypoglycemia on long-term diabetes complications. The new analytic approach presented herein needs to be validated by comparing it to the traditional metrics using data from other completed studies. In addition, it can be further validated by applying it to future studies where it can be used to evaluate if IntHypo or HypoRV correlates better than traditional metrics with quality of life, costs, short-term and long-term complications. Finally, a similar approach can be used to define the metrics of hyperglycemia to determine if these new analytic metrics correlate any better with long-term complications of diabetes than current metrics. These novel metrics can then be incorporated into other approaches to understand overall glycemic control such as the glucose pentagon.²⁰

We did not compare the IntHypo and HypoRV to the low blood glucose index (LBGI) as described by Kovatchev et al in this analysis.²¹ The LBGI is calculated from the logarithmic transformation of blood sugar values and has been adapted to CGM-derived data by Fabris and colleagues.²² However, to calculate the LBGI from our data set we would have to be selective about what data should be extracted from the CGM profiles. It is not clear what values are the best, for example, the minimum of each glucose fluctuation in CGM profile, all severely hypoglycemic values, and so on. From our point of view the selection criteria are arbitrary and the resultant LBGI would depend on the choice of those criteria. On the other hand, there are clearly defined CGM parameters by CGM, for example, AUC, the amount of time in the hypoglycemic range and the frequency of hypoglycemia, which are included in our model.

Footnotes

Abbreviations

A1C, hemoglobin A1C; ASPIRE, Automation to Simulate Pancreatic Insulin Response; AUC, area under the curve; BG, blood glucose; CGM, continuous glucose monitoring; CV, coefficient of variation; GV, glucose variability; HypoRV, hypoglycemia risk volume; IntHypo, intensity of hypoglycemia; LBGI, low blood glucose index; LGS, low glucose suspend; SIP, sensor-integrated pump; SIP-LGS, sensor-integrated pump without low glucose suspend; SIP+LGS, sensor-integrated pump with low glucose suspend; SMBG, self-monitoring of blood glucose.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Diabetes Control and Complications Trial Research Group. Hypoglycemia in the Diabetes Control and Complications Trial. Diabetes. 1997;47:271-286.

Kristensen

Beck-Nielsen

Nørgaard

Tarnow

Insulin analogues and severe hypoglycemia in type 1 diabetes. Diabetes Res Clin Pract. 2012;96:17-23.

Pedersen-Bjergaard

Pramming

Heller

et al . Severe hypoglycemia in 1076 adult patients with type 1 diabetes: influence of risk markers and selection. Diabetes Metab Res Rev. 2004;20:479-486.

Battelino

Phillip

Bratina

Nimri

Oskarsson

Bolinder

Effect of continuous glucose monitoring on hypoglycemia in type 1 diabetes. Diabetes Care. 2011;34:795-800.

van Beers

CAJ

DeVries

Kleijer

et al . Continuous glucose monitoring for patients with type 1 diabetes and impaired awareness of hypoglycemia (IN CONTROL): a randomised, open-label, crossover trial. Lancet Diabetes Endocrinol. 2016;4:893-902.

Beck

; for JDRF CGM Study Group. The effect of continuous glucose monitoring in well-controlled type 1 diabetes. Diabetes Care. 2009;32:1378-1383.

Bergenstal

Klonoff

Garg

et al . Threshold-based insulin-pump interruption for reduction of hypoglycemia. N Engl J Med. 2013;369:224-232.

McCoy

Van Houten

Ziegenfuss

Shah

Wermers

Smith

SA.

Increased mortality of patients with diabetes reporting severe hypoglycemia. Diabetes Care. 2012;35:1897-1901.

Kovatchev

Cox

Farhy

Straume

Gonder-Frederick

Clarke

WL.

Episodes of severe hypoglycemia in type 1 diabetes are preceded and followed within 48 hours by measurable disturbances in blood glucose. J Clin Endocrinol Metab. 2000;85:4287-4292.

10.

Veglio

Sivieri

Chinaglia

Scaglione

Cavallo-Perin

QT interval prolongation and mortality in type 1 diabetic patients. A 5-year cohort prospective study. Diabetes Care. 2000;23:1381-1383.

11.

Novodvorsky

Bernjak

Chow

et al . Diurnal differences in risk of cardiac arrhythmias during spontaneous hypoglycemia in young people with type 1 diabetes. Diabetes Care. 2017;40:655-662.

12.

Tattersall

Gill

GV.

Unexplained deaths of type 1, diabetic patients. Diabet Med. 1991;8:49-58.

13.

Thordarson

Sovik

Dead in bed syndrome in young diabetic patients in Norway. Diabet Med. 1995;12:782-787.

14.

Sartor

Dahlquist

Short-term mortality in childhood onset insulin-dependent diabetes mellitus: a high frequency of unexpected deaths in bed. Diabet Med. 1995;12:607-611.

15.

Lachin

Bebu

Bergenstal

et al . Association of glycemic variability in type 1 diabetes with progression of microvascular outcomes in the Diabetes Control and Complications Trial. Diabetes Care. 2017;40:777-783.

16.

Hirsch

Glycemic variability and diabetes complications: Does it matter? Of course it does!

Diabetes Care. 2015;38:1610-1614.

17.

Bergenstal

RM.

Glycemic variability and diabetes complications: Does it Matter? Simply put, there are better glycemic markers!

Diabetes Care. 2015;38:1615-1621.

18.

Monnier

Colette

Wojtusciszyn

et al . Toward defining the threshold between low and high glucose variability in diabetes [published online ahead of print December 30, 2016]. Diabetes Care.

19.

Shu-Hua

Tao

et al . Impact of admission glycemic variability, glucose and glycosylated hemoglobin on major adverse cardiac events after acute myocardial infarction. Diabetes Care. 2013;36:1026-1032.

20.

Thomas

Schonauer

Achermann

et al . The “Glucose Pentagon”: assessing glycemic control of patients with diabetes mellitus by a model integrating different parameters from glucose profiles. Diab Tech Ther. 2009;11:399-409.

21.

Kovatchev

Cox

Gonder-Frederick

Young-Hyman

Schlundt

Clarke

Assessment of risk for severe hypoglycemia among adults with IDDM. Validation of the low blood glucose index. Diabetes Care. 1998;21:1870-1875.

22.

Fabris

Patek

Breton

MD.

Are risk indices derived from CGM interchangeable with SMBG-based indices?

J Diab Sci Tech. 2016;10:50-59.