The Comprehensive Glucose Pentagon: A Glucose-Centric Composite Metric for Assessing Glycemic Control in Persons With Diabetes

Abstract

Background:

Composite metrics have the potential to provide more complete and clinically useful information about glycemic control than traditional individual metrics such as hemoglobin A1C, %/time/area under curve of hypoglycemia and hyperglycemia.

Methods:

Using five key metrics that are derived from continuous glucose monitoring, we developed a new, multicomponent composite metric, the Comprehensive Glucose Pentagon (CGP) that demonstrates glycemic control both numerically and visually. Two of its axes are composite metrics—the intensity of hypoglycemia and intensity of hyperglycemia. This approach eliminates the use of the surrogate marker, hemoglobin A1C (A1C), and replaces it with glucose-centric metrics.

Results:

We reanalyzed the data from two randomized control trials, the STAR 3 and ASPIRE In-Home studies using the CGP. It provided new insights into the effect of sensor-augmented pumping (SAP) in the STAR 3 trial and sensor-integrated pumping with low-glucose threshold suspend (SIP+TS) in the ASPIRE In-Home trial.

Conclusions:

The CGP has the potential to enable health care providers, investigators and patients to better understand the components of glycemic control and the effect of various interventions on the individual elements of that control. This can be done on a daily, weekly, or monthly basis. It also allows direct comparison of the effects on different interventions among clinical trials which is not possible using A1C alone. This new composite metric approach requires validation to determine if it provides a better predictor of long-term outcomes than A1C and/or better predictor of severe hypoglycemia than the low blood glucose index (LBGI).

Keywords

composite metrics glycemic control hypoglycemia continuous glucose monitoring hemoglobin A1C

Composite outcomes have been used to provide an integrated and more complete understanding in a variety of circumstances. In sports (National Football League quarterback rating) and culinary circles (wine points), composite metrics are frequently used comparators. Among the most commonly used medical composite metrics are the APGAR score (a 5-component compilation of complexion, pulse, reflex activity and respiratory effort in newborns) and the Glasgow Coma Scale (a 3-component metric of eye, verbal and motor responses). Yet, there has been no composite metric that has been widely adopted for the description of glycemic control. Since the publication of the results of Diabetes Control and Complications Trial,¹ the diabetes community has used hemoglobin A1C (A1C) as the principal metric to establish therapy goals, evaluate the benefits of pharmacologic, technologic and psycho-social interventions, and attain regulatory approval for new pharmaceuticals and devices. The advantage of using this surrogate metric which reflects mean plasma glucose levels is that it reflects the risk of micro- and macrovascular outcomes.^2,3 Nevertheless, A1C is an imperfect metric and its exigencies have been well-described.^4-6 In addition, A1C reflects only mean glucose levels and does not provide information about the frequency, duration and severity of hypoglycemia and hyperglycemia, the time in range, and glucose variability. These metrics are critical to the understanding of short-term glycemic control and may have implications for long-term complications.

The ability of continuous glucose monitoring (CGM) to collect key metrics of glycemic control (including mean glucose levels) allows us to go beyond A1C in understanding most, if not all, the important variables comprising glycemic control. By utilizing CGM-derived data, we can develop glucose-centric, composite metrics that eliminate the chore of having to individually compare individual aspects of glycemic control. Indeed, there has been a recent interest in providing a more comprehensive description of glycemic control that includes hypoglycemia and time in range as a supplement to A1C as a performance metric and the basis for regulatory approval.^7-9 The European Medicines Agency encourages the use of CGM in the determination of overnight profiles, postprandial hyperglycemia and hypoglycemia rates in clinical trials,¹⁰ although it is not clear to what degree these data contribute to regulatory approval and labeling.

There have been several attempts to develop and display CGM-based composite metrics. Such metrics can be grouped into those that are purely visual,^11-13 purely numeric,^14,15 and a combination of visual and numeric.¹⁶ We have developed a new composite metric, the CGM-derived Glucose Pentagon (Comprehensive Glucose Pentagon or CGP), that includes all the key metrics of glycemic control: mean sensor glucose, glucose variability, severity of hypo- and hyperglycemia, and time out of range (the inverse of time in range), while eliminating A1C as an integral component. We have reanalyzed data from the Sensor-Augmented Pump Therapy for A1C Reduction (STAR 3) trial¹⁷ and the Automation to Simulate Pancreatic Insulin Response (ASPIRE) In-Home trial¹⁸ to validate this approach and determine if additional insights beyond those initially reported in those studies can be obtained using the CGP. Based on these analyses, we believe that the CGP can also be used to better evaluate glycemic control with CGM-derived data in individual subjects. In clinical practice, the CGP should be able to compare an individual patient’s overall glycemic control in short time periods, for example, daily, weekly, and monthly, which is not possible with A1C.

Methods

The five axes of CGP are shown in Figure 1. Compared to the original Glucose Pentagon model,¹⁶ the A1C axis has been eliminated and the metric of glycemic variability is now % coefficient of variation (%CV) instead of standard deviation (SD). There are three new axes: (1) time out of range (ToR)—the reciprocal of time in range (so expressed to permit increasing/decreasing pentagon area as TIR worsens/improves) which is defined as: 1440_min – time in range _min (TIR_min); (2) the intensity of hyperglycemia or IntHyper (in place of AUC >160 mg/dl and time >160 mg/dl) which is expressed as $\sqrt{A U C^{2} + T i m e^{2}}$ and, (3) the intensity of hypoglycemia or IntHypo expressed as $\sqrt{A U C^{2} + T i m e^{2}}$ , since there was no previous axis reflecting hypoglycemia . Only the mean glucose level is retained from the original Glucose Pentagon.

Figure 1.

The Comprehensive Glucose Pentagon (CGP). Compared to the original Glucose Pentagon model, the new axes include the intensity of hyperglycemia (in place of AUC >160 mg/dl and time >160 mg/dl), intensity of hypoglycemia (no previous hypoglycemia axis), time out of range (1440 min – time in range) (in place of A1C because of the known discrepancies between A1C and mean glucose.⁶ Mean glucose remains as one of the axes. The green pentagon is the area of an average person without diabetes and has an area = 466 mm².

The approach to scaling of the mean glucose and glycemic variability axes have been previously described in detail.¹⁶ Briefly, the mean glucose scale—transformed from the relationship of A1C to mean glucose by the formula Average Glucose_(mg/dl) = 28.7 × A1c − 46.7⁶—is weighted to reflect the risk of developing microvascular complications as originally reported in the DCCT study. Since there were different curvilinear relationships between the risk of complications and A1C, we have used the average risk of all the complications as they relate to A1C. No similar relationship of glycemic variability (either SD or %CV) to long-term complications exists so we have used a linear scale based on the published relationship of oxidative stress and glycemic variability.^19,20 For the hypoglycemia and hyperglycemia axes, the CGP displays the intensity of hypoglycemia (IntHypo) and intensity of hyperglycemia (IntHyper) which are the lengths of the vector of two standard metrics of hypoglycemia and hyperglycemia—area under the curve (AUC) ≤70 mg/dL and time/day ≤70 mg/dL and AUC ≥180 mg/dL and time/day ≥180 mg/dL, respectively. The scaling of IntHypo and IntHyper are adapted from Kovatchev.²¹ He previously described risks for severe hypoglycemia and hyperglycemia in the low blood glucose index (LBGI) and high blood glucose index (HBGI), respectively. The scaling of the ToR axes is also adapted from Kovatchev²¹ but because we do not qualitatively differentiate the low and high values of ToR, we calculate both with the same exponential function.

The scalings of the five axes of the CGP are described by the following formulae:

Mean axes: × [mm] = [(Mean glucose − 90) × 0.0217]^2.63 + 14

CV axes: × [mm] = (CV − 17) × 0.92 + 14

IntHyper: × [mm] = (IntHyper_mg/dL × 0.000115)^1.51 + 14

IntHypo: × [mm] = e^{(IntHypo × 0.00057)} + 13

ToR axes: × [mm] = (ToR × 0.00614)^1.581 + 14

The area of the CGP is calculated by summing the areas of five 72° triangles according to the following:

A_CGP = A_ToR-CV + A_CV-IntHYPO + A_{IntHYPO-IntHYPER} + A_{IntHYPER-Mean glucose} + A_{Mean glucose-ToR,} where:

A_x-y = ½ a × b × sin γ (γ = angle between the axes, each 72°;

a, b = length of each axis; A = area of triangle; and,

x-y = length of axis for each area).

The length of all axes is 76 mm. The green pentagon has an area of 466 mm² and represents the glycemic status of someone without diabetes. Based on the limited number of studies that have done CGM in nondiabetic subjects, we chose the following values for the normal glucose pentagon: Mean sensor glucose = 90 mg/dl, SD = 15 mg/dl resulting in a CV 16.7%, IntHypo = 0, IntHyper = 0, ToR = 0.^22,23 By setting that area to equal 1, the area of any CGM-derived set of data permits a direct comparison of the glycemic control of an individual person or group of persons with diabetes to those without diabetes and is postulated to represent a relative risk of a combination of hypoglycemia and long-term complications on a scale of 1 to 5, called prognostic glycemic risk (PGR) parameter. We have arbitrarily assigned descriptions of risk to the PGR as follows: If the PGR ≤ 2.0, the patient has a very low risk for diabetes short- and/or long-term complications; for 2.0 < PGR ≤ 3.0 the risk is low; for 3.0 < PGR ≤ 4.0 the risk is moderate; 4.0 < PGR ≤ 4.5 high and > 4.5 extremely high.

We used data from two studies using CGM, the STAR 3 and ASPIRE In-Home trials, to determine how to scale the axes. The STAR 3 and ASPIRE In-Home trials have been previously described in detail.^17,18 Briefly, in the STAR 3 trial, 495 adults and children were randomized to therapy with multiple daily injections (MDI) and glucose monitoring by fingerstick-obtained blood samples or continuous subcutaneous insulin infusion with glucose monitoring by CGM or sensor-augmented pumping (SAP) for 1 year. The A1C (the primary outcome metric) improved by 0.8% in the SAP group and by 0.2% in the MDI group (P < .0001). The rate of severe hypoglycemia (needing assistance and a documented BG value of <50 mg/dL) was not significantly different between the two groups. CGM was performed on all subjects at baseline, 6 and 12 months. The AUC <50 mg/dL and <70 mg/dL were not different between the groups, but the AUC >180 mg/dL and >250 mg/dL were significantly improved in the SAP group compared to MDI (P < .0001). The ASPIRE In-Home trial randomized 247 adolescent and adult subjects to sensor-augmented pump without a threshold suspend feature (SAP-TS) or sensor-integrated pump with a threshold suspend feature (SIP+TS) for 3 months. Those with the SIP+TS had a reduction in AUC of nocturnal hypoglycemia (the primary outcome metric) by 37.5% compared to SAP alone (P < .001) and a reduction in the number of sensor glucose values <50 mg/dL, <60 mg/dL, and <70mg/dL at night and also day plus night (P < .001). There was no significant change in the A1C in either group. Thus, the STAR 3 and ASPIRE In-Home trials provide an opportunity to compare composite outcomes in studies that differ both in types of interventions (SAP vs MDI and SAP-TS vs SIP+TS), as well as in the types of outcomes (improvement in A1C without a reduction in hypoglycemia in STAR 3 and a reduction in hypoglycemia without a change in A1C in ASPIRE In-Home).

Given the normal distribution of patient values, we used paired t-tests to compare the changes between baseline and end of study for both STAR 3 and ASPIRE In-Home studies.

Results

The CGM-derived data from the STAR 3 and ASPIRE In-Home trials are shown in Tables 1 and 2. The standard glucose metrics of the SAP and MDI cohorts in the STAR 3 trial demonstrate that those subjects had much less well-controlled diabetes at baseline than those in the ASPIRE In-Home trial. For example, mean glucose was 184 mg/dL and 185 mg/dL versus 147 mg/dL and 147 mg/dL, respectively and A1C was 8.3% and 8.3% versus 7.3% and 7.2%, respectively at baseline. The glucose pentagon area and most of its components—time out of range, intensity of hyperglycemia, intensity of hypoglycemia, and CV—were also worse in STAR 3 subjects compared with those subjects in the ASPIRE In-Home trial. Conversely, the intensity of hypoglycemia, at baseline, was worse in the both groups in the ASPIRE In-Home trial compared with both STAR 3 groups. The PGR at baseline in each trial is: is 2.41 (SAP group) and 2.72 (MDI group) in the STAR 3 trial and 2.27 (SIP+TS) and 2.42 (SAP-TS) in the ASPIRE In-Home trial. Table 3 shows the percentage differences in the glucose pentagon area between the baseline and end of study for both the STAR 3 and ASPIRE In-Home studies.

Table 1.

STAR 3 Study Results.

STAR 3	SAP		P	MDI		P
STAR 3	Baseline	12 months	P	Baseline	12 months	P
Mean glucose (mg/dl)	184.1 ± 30.1	164.7 ± 28.0	<.01	185.0 ± 30.4	182,2 ± 35.4	.151
CV (%)	33.2 ± 7.1	33.0 ± 6.7	.44	28.2 ± 11.6	37.2 ± 7.5	<.01*
Time out of range (1440 -TIR) (min)	708.3 ± 220.1	532.8 ± 236.8	<.01	708.9 ± 215.0	676.1 ± 263.4	.071
Intensity HYPER (mg/dl × min²)	46408 ± 28425	29269 ± 22650	<.01	47537 ± 27917	46163 ± 33756	.40
Intensity HYPO (mg/dl × min²)	517 ± 794	495 ± 695	.67	546 ± 744	517 ± 801	.57
HypoRiskVolume (mg/dl × min²)	315 ± 728	377 ± 859	.58*	416 ± 912	421 ± 729	.47
Pentagon area (CI**)	1123 ± 457 (1048, 1198)	1002 ± 294 (955, 1049)	<.01	1268 ± 503 (1165, 1371)	1221 ± 452 (1128, 1314)	.068
PGR (CI**)	2.41 ± 0.98 (2.25, 2.57)	2.15 ± 0.63 (2.05, 2.25)	<.01	2.72 ± 1.08 (2.50, 2.94)	2.62 ± 0.97 (2.42, 2.82)	.07
SD (mg/dl)	60.5 ± 11.6	62.1 ± 11.7	<.01	42.9 ± 20.8	67.2 ± 15.8	<.01
HbA1c (%)	8.27 ± 0.53	7.50 ± 0.80	<.01	8.30 ± 0.52	8.07 ± 0.93	<.01

ASPIRE, Automation to Simulate Pancreatic Insulin Response; CV, coefficient of variation; HYPER, hyperglycemia; HYPO, hypoglycemia; MDI, multiple dose insulin; PGR, prognostic glycemic risk; SAP, sensor-augmented pump; STAR, sensor-augmented pump therapy for A1C reduction.

Value is higher after 12 month in comparison to baseline.

CI, confidence interval (95%).

Table 2.

ASPIRE In-Home Study Results.

ASPIRE	SIP+TS		P	SAP-TS		P
ASPIRE	Baseline	12 weeks	P	Baseline	12 weeks	P
Mean glucose (mg/dl)	146.6 ± 20.2	158.1 ± 20.5	<.01*	147.2 ± 21.2	155.6 ± 27.0	<.01*
CV (%)	41.4 ± 6.0	40.2 ± 6.3	.09	41.8 ± 5.9	41.8 ± 6.1	.49
Time out of range (1440-TIR) (min)	507.5 ± 191.8	551.0 ± 176.3	<.01*	503.2 ± 155.2	561.8 ± 203.1	<.01*
Intensity HYPER (mg/dl × min²)	19670 ± 14193	24891 ± 14027	<.01*	20632 ± 14014	24908 ± 16332	<.02*
Intensity HYPO (mg/dl × min²)	1620 ± 1397	895 ± 676	<.01	1642 ± 1392	1557 ± 1163	.30
HypoRiskVolume (mg/dl × min²)	1935 ± 2989	690 ± 975	<.01	1601 ± 2030	1545 ± 4127	.43
Pentagon area (CI**)	1058 ± 280 (1008, 1108)	1095 ± 256 (1048, 1142)	.14*	1128 ± 433 (1052, 1204)	1170 ± 363 (1106, 1234)	.21*
PGR (CI**)	2.27 ± 0.60 (2.16, 2.38)	2.35 ± 0.55 (2.25, 2.45)	.14*	2.42 ± 0.93 (2.26, 2.58)	2.51 ± 0.78 (2.37, 2.65)	.21*
SD (mg/dl)	60.8 ± 12.3	63.7 ± 13.5	<.01*	61.5 ± 12.2	64.9 ± 14.1	<.01*
HbA1c (%)	7.26 ± 0.70	7.25 ± 0.66	.04	7.21 ± 0.74	7.17 ± 0.82	<.03

ASPIRE, Automation to Simulate Pancreatic Insulin Response; CV, coefficient of variation; HYPER, hyperglycemia; HYPO, hypoglycemia; PGR, prognostic glycemic risk; SAP-TS, control group: sensor-augmented pump without threshold suspend feature (interruption of insulin delivery at a fixed threshold sensor glucose value); SD, standard deviation; SIP+TS, intervention group: sensor-integrated pump with threshold suspend feature (interruption of insulin delivery at a fixed threshold sensor glucose value); TIR, time out of range.

Value is higher after 12 weeks in comparison to baseline.

CI, confidence interval (95%).

Table 3.

Comparison of Glucose Pentagon Areas for the STAR 3 and ASPIRE In-Home Studies.

	Baseline	End of study
	Pentagon area (CI*)	Pentagon area (CI*)	% change	P
STAR 3
SAP	1123 (1048, 1198)	1002 (955, 1049)	−10.8	<.01
MDI	1268 (1165, 1371)	1221 (1128, 1314)	−3.7	.07
ASPIRE In-Home
SAP-TS	1128 (1052, 1204)	1170 (1106, 1234)	1.0	.21
SAP+TS	1058 (1008, 1108)	1095 (1048, 1142)	1.0	.14

ASPIRE, Automation to Simulate Pancreatic Insulin Response; MDI, multiple dose insulin; SAP, sensor-augmented pump; SAP-TS, control group: sensor-augmented pump without threshold suspend feature (interruption of insulin delivery at a fixed threshold sensor glucose value); SIP+TS, intervention group: sensor-integrated pump with threshold suspend feature (interruption of insulin delivery at a fixed threshold sensor glucose value); STAR, sensor-augmented pump therapy for A1C reduction.

In the STAR 3 trial, the interventions demonstrated improved overall glycemic control as exemplified by the reduction in the CGP areas. In the ASPIRE In-Home trial, the CGP area and therefore the PGR did not significantly change. However, the reason(s) for the resultant CGP areas in the intervention versus control groups were different in each study. In the STAR 3 trial, the reduction of the CGP area in the SAP cohort resulted from a significantly greater improvement in three metrics: mean glucose, ToR and IntHypo, compared to the MDI group. In the ASPIRE In-Home trial, the SIP+TS cohort had a significant decrease in the IntHypo, but this was offset by an increase in the IntHyper. The PGR was already low and not significantly changed in either group (Figures 2 and 3). Examples of an individual from each study’s intervention arm is shown in Figure 4 (STAR 3) and Figure 5 (ASPIRE In-Home). The examples from individual patients illustrate the capability of the CGP to clearly illustrate what elements of glycemic control were affected by each intervention and demonstrate the trade-offs that sometimes exist between short-term and long-term risk reduction.

Figure 2.

Results of the STAR 3 trial. Upper panel (left): baseline data from the MDI cohort. Upper panel (right): end-of-study (12 month) data from the MDI cohort. Lower panel (left): baseline data in the SAP cohort. Lower panel (right): end-of-study (12 month) data in the SAP cohort. The green pentagon represents an average person without diabetes (area = 1).

Figure 3.

Results of the ASPIRE In-Home trial. Upper panel (left): baseline data from the SAP-TS cohort. Upper panel (right): end-of-study (12 weeks) data from the SAP-TS cohort. Lower panel (left): baseline data in the SIP+TS cohort. Lower panel (right): end-of-study (12 weeks) data in the SIP+TS cohort. The green pentagon represents an average person without diabetes (area = 1).

Figure 4.

The green pentagon represents a person without diabetes (area = 1). The baseline pentagon is the pink area and represents the area relative to normal = 2.86; the 1 year pentagon is the dark red area relative to normal = 2.48 in a patient randomized to receive the sensor-augmented pump (SAP) in the STAR 3 trial. After 1 year mean value of glucose concentration, ToR (1440 – time in range) and intensity of hypoglycemia decreased significantly, but the glycemic variability (%CV) was increased. The intensity of hyperglycemia was nearly the same (39579 mg/dl × min² in baseline vs 38926 mg/dl × min², after 12 month). The A1C deteriorated from 8.0 to 8.4%, however, the prognostic risk parameter improved from 2.86 to 2.48. This demonstrates improved short-term glycemic control despite worse A1C. The green pentagon is the area of an average person without diabetes.

Figure 5.

The green pentagon represents a person without diabetes (area = 1). The baseline pentagon (orange) has a high GPR (4.35). The 12-week pentagon (blue) has a low GPR (2.31) in a patient randomized to receive the sensor-augmented pump with the low glucose threshold suspend (SAP + TS) feature, in the ASPIRE In-Home trial. There was a rise in A1C from 5.9% to 6.5% in this patient yet the patient’s overall glucose control improved. The green pentagon is the area of an average person without diabetes.

Because the data analyzed in this study represent that of a large patient population, it is not surprising that there is a good correlation between the mean glucose and the intensity of hyperglycemia and the ToR in these studies (r = .75-.85). However, for an individual patient, the correlation may be significantly lower especially in a patient who has a high IntHypo and/or high IntHyper.

Discussion

The CGP extends the novel approach of the original Glucose Pentagon in several ways. First, it establishes a composite metric that is entirely glucose-centric by eliminating the need for a laboratory derived metric, that is, A1C. This accomplishes several important goals. It circumvents the issue that an A1C may represent widely different mean glucose levels between individuals.^6,24 For example, an A1C of 8% can reflect a mean glucose between 120 mg/dL and 200 mg/dL. These discrepancies are thought to be due to genetic and/or clinical differences in glycation rates, red cell survival, and ethnicity among others. Second, since the CGP can be calculated at frequent intervals it permits a short-term analysis of glycemic control, which A1C cannot provide because it is slow to change. By incorporating a program that calculates the CGP into CGM software, it could be accessed by the patient and/or health care provider on a daily, weekly or monthly basis. Third, the CGP can both numerically and pictorially identify the component(s) of glycemic control that are most in need of improvement by either a review of the analytical data that comprises the CGP or by visual review. Fourth, it incorporates two key metrics, time in range (expressed in the CGP as its reciprocal, ToR) and intensity of hypoglycemia, which are important correlates of quality of life. At the August 2016 Food and Drug Administration Outcomes Measures Beyond Hemoglobin A1C workshop, data from a 3,455-patient online survey of people with both type 1 and type 2 diabetes conducted by the diaTribe Foundation and dQ&A showed that time in range ranked first in what was most impactful in the daily lives of those individuals regardless of therapy.²⁵ In addition, the DAWN2 found that 56% of persons living with diabetes felt very worried about the risk of hypoglycemia and 39% reported that taking diabetes medications interfered with their quality of life.²⁶ The use of IntHypo and IntHyper goes beyond the standard metrics of AUC, %Hypo and %Hyper. The AUC represents the integral over glucose values over the time. If the glucose level was constant and the AUC is known, the duration of hypo- or hyperglycemia can be derived. However, glucose levels constantly vary. Under those circumstances, there are situations where equivalent AUC’s have very different durations (Figure 6). Thus, the intensity of hyper- or hypoglycemia are calculated as the length of the vector between AUC and time using the formula $\sqrt{A U C^{2} + T i m e^{2}}$ .

Figure 6.

The upper figure represents a condition where glucose levels are stable while the lower figures demonstrate how equivalent AUC’s can have very different durations depending on the variability of the glucose levels.

Finally, the CGP incorporates several parameters that have not been previously considered in composite metrics. The first is the intensity of hypoglycemia, which is not the product of duration and AUC, but rather the vector of these two metrics as described in the companion paper.²⁷ The inclusion of duration in both of the intensity terms provides additional weighting of this important metric, which is generally expressed as simple individual terms of frequency, duration, and/or AUC. There have been numerous ways to define hypoglycemia both clinically and numerically and but no consensus has been reached until recently. In the most recent American Diabetes Association’s Standards of Medical Care in Diabetes 2017²⁸ and in an accompanying consensus document by the International Hypoglycemia Study Group,²⁹ there are revised definitions of clinical and biochemical hypoglycemia However, the biochemical definitions are unidimensional when obtained by blood glucose testing and thus provide no information about the depth or duration of hypoglycemia over time. Even when using CGM, the American Diabetes Association and the International Hypoglycemia Study Group propose that a static measure of CGM-derived sensor glucose level of 54 mg/dL for 20 minutes defines biochemical hypoglycemia.^28,29

The intensity of hyperglycemia is the second new parameter. An approach like that used to develop the intensity of hypoglycemia was used to calculate it. Finally, we use ToR (the inverse of time in range) for one of the axes. Expressing it as ToR permits an improvement in glycemic control to be displayed as a reduction in the length of the axis and thus a reduction of the CGP area.

The CGP is still slightly weighted toward hyperglycemia because of the nonnormal distribution of glucose. However, we believe that log transformation of the data would make it difficult for the typical reader to understand such metrics. Investigators may choose to explore that option in the future in determining whether it correlates better with outcomes than the present composition of the CGP.

In developing the original Glucose Pentagon model, Thomas and colleagues included primarily metrics that reflected hyperglycemia, as these could be related to the microvascular outcomes results of the Diabetes Control and Complications trial. This permitted scaling of the axes in proportion to the exponential curves relating A1C and mean glucose to the risk of complications. They were cautious in their conclusions by suggesting that the clinical relevance of their model would have to be independently validated against future studies with the effects of an intervention on hard clinical outcomes. We also feel that the CGP approach needs validation in subsequent clinical trials to determine if the current approach to scaling is correct and to assess the utility of this approach in predicting short- and long-term clinical outcomes. Thomas and colleagues included the metric of glycemic variability (SD) assuming that it was linearly correlated to the risk of complication based on in vitro and limited in vivo studies on the relationship of glycemic variability and markers of oxidative stress.¹⁶ In the CGP, we have changed from SD to %CV to accommodate the nonnormal distribution of glucose. In addition, CV, unlike SD, is independent of mean glucose. Including a glycemic variability axis is intended to reflect the importance of this metric in short-term control of diabetes. However, we recognize that there continues to be controversy over the relationship of glycemic variability to long-term microvascular complications.^30-32 Preclinical experiments have demonstrated that variable glucose levels significantly increase marker of oxidative stress in human umbilical endothelial cells as well as increase the production of VEGF by retinal epithelium.^33-35 Thus, ToR is less important than %CV in this model. While it will be logistically and financially unlikely that DCCT and UKPDS-like trials will be repeated, we would encourage the pharmaceutical industry to incorporate CGM into the cardiovascular outcomes trials that are now mandated by the Food and Drug Administration for all new diabetes drugs. These trials often include microvascular complications as well. This would permit a better understanding of how CGM-derived, glucose-centric composite metrics affect clinical outcomes .

Based on the analyses presented herein, we believe that the CGP provides additional insights about the STAR 3 and ASPIRE In-Home trial that were either missing from and/or difficult to appreciate in the reports of those trials. In addition, this approach permits a straight forward comparison of cohorts in different trials as well as changes within the trials. In addition to its use in assessing the results of clinical trials, the CGP can provide a short-term clinical metric for use by both patients and their health care providers that supplements the A1C. Lipska and Krumholtz have recently questioned whether or not A1C or, indeed, any metric of glycemic control are the appropriate metrics for determining benefit in clinical trials.⁵ Whether or not an approach which incorporates multiple metrics of glycemic control like the CGP may better predict outcomes must be determined by applying it to long-term clinical trials and to large databases which include data from insurance claims, electronic record information along with CGM data. This will allow refinement of the CGP that may include alternate metrics and/or different scaling based on such data.

Footnotes

Abbreviations

A1C, hemoglobin A1C; APGAR, appearance, pulse, grimace, activity, respiration; ASPIRE, Automation to Simulate Pancreatic Insulin Response; AUC, area under the curve; BG, blood glucose; CGM, continuous glucose monitoring; CGP, Comprehensive Glucose Pentagon; CV, coefficient of variation; DAWN2, desires, attitudes, wishes and needs 2; HBGI, high blood glucose index; HypoRV, hypoglycemia risk volume; IntHypo, intensity of hypoglycemia; LGS, low glucose suspend; LBGI, low blood glucose index; MDI, multiple dose insulin; PGR, prognostic glycemic risk; SAP, sensor-augmented pump; SD, standard deviation; SIP, sensor-integrated pump; SIP+LGS, sensor-integrated pump with low glucose suspend; SIP-LGS, sensor-integrated pump without low glucose suspend; STAR, sensor-augmented pump therapy for A1C reduction; TiR, time in range; ToR, time out of range.

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. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N Eng J Med. 1993;329:977-986.

Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications (DCCT/EDIC) Study Research Group. Update on cardiovascular outcomes at 30 years of the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications Study. Diabetes Care. 2014;37:39-43.

UK Prospective Diabetes Study (UKPDS) Group. Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). Lancet. 1998;352:837-853.

Wilson

Xing

Beck

et al . Hemoglobin A1c and mean glucose in patients with type 1 diabetes. Diabetes Care. 2011;34:540-544.

Lipska

Krumholtz

Is Hemoglobin A1c the right outcome for studies of diabetes?

J Amer Med Assoc. 2017;317(10):1017-1018.

Nathan

Kuenen

Borg

et al . Translating the A1C assay into estimated average glucose values. Diabetes Care. 2008;31:1473-1478.

Rodriguez-Gutierrez

Lipska

McCoy

Ospina

Ting

Montori

VM.

Hypoglycemia as a quality measure in Diabetes Study Group. Brit Med J. 2016;352:i1084.

Rodriguez-Gutierrez

Ospina

Shah

et al . Inclusion of hypoglycemia in clinical practice guidelines and performance measure in the care of patients with diabetes. JAMA Int Med. 2016;176(11):1714-1716.

Food and Drug Administration. Diabetes outcome measures beyond hemoglobin A1c workshop, 2016. Available at: https://www.fda.gov/drugs/newsevents/ucm499281.htm. Accessed April 4, 2017.

10.

European Medicines Agency Guideline on clinical investigation of medicinal products in the treatment or prevention of diabetes mellitus CPMP/EWP/1080/00 Rev. 1 Committee for Medicinal Products for Human Use (CHMP), 2012. Available at: http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2012/06/WC500129256.pdf. Accessed March 31, 2017.

11.

Rodbard

Evaluating quality of glycemic control: graphical displays of hypo- and hyperglycemia, time in target range, and mean glucose. J Diabetes Sci Technol. 2015;9:56-62.

12.

Liebl

Henrichs

Heinemann

et al . Continuous glucose monitoring: evidence and consensus statement for clinical use. J Diabetes Sci Technol. 2013;7:500-519.

13.

Russell

El-Khatib

Sinha

et al . Outpatient glycemic control with a bionic pancreas in type 1 diabetes. N Eng J Med. 2014;371(4):313-325.

14.

Augstein

Heinke

Vogt

et al . Q-Score: development of a new metric for continuous glucose monitoring that enables stratification of antihyperglycaemic therapies. BMC Endocr Disord. 2015;15:22.

15.

Vigersky

Escaping the A1c-centric world in evaluating diabetes interventions. J Diabetes Sci Technol. 2015;9:1148-1151.

16.

Thomas

Schönauer

Achermann

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

17.

Bergenstal

Tamborlane

Ahmann

et al . Effectiveness of sensor-augmented insulin pump therapy in type 1 diabetes. N Engl J Med. 2010;363(4):311-320.

18.

Bergenstal

Klonoff

Garg

et al . Threshold-based insulin-pump interruption for reduction of hypoglycemia. N Engl J Med. 2013;369(3):224-232.

19.

Monnier

Emilie Mas

Ginet

et al . Activation of oxidative stress by acute glucose fluctuations compared with sustained chronic hyperglycemia in patients with type 2 diabetes. JAMA 2006;295:1681-1687.

20.

Ceriello

Kumar

Piconi

Esposito

Giugliano

Simultaneous control of hyperglycemia and oxidative stress normalizes endothelial function in type 1 diabetes. Diabetes Care. 2007;30:649-654.

21.

Kovatchev

. Metrics for glycaemic control—from HbA_1C to continuous glucose monitoring [published online ahead of print March 17, 2017]. Nat Rev Endocrinol. doi:10.1038/nrendo.2017.3.

22.

Freckmann

Hagenlocher

Baumstark

et al . Continuous glucose profiles in healthy subjects under everyday life conditions and after different meals. J Diabetes Sci Technol. 2007;1:695-703.

23.

Salkind

Huizenga

Fonda

Walker

Vigersky

RA.

Glycemic variability in nondiabetic morbidly obese persons: results of an observational study and review of the literature. J Diabetes Sci Technol. 2014;8(5):1042-1047.

24.

Gubitosi-Klug

Beck

Gal

, for T1D Exchange. Symposium presented at: 76th Scientific Session of the American Diabetes Association; June 10-14, 2016; New Orleans, LA.

25.

. Closer Look. Available at: https://www.closeconcerns.com/knowledgebase/r/01354d4c Accessed August 29, 2016.

26.

Dawn Study. Available at: http://www.dawnstudy.com/dawn2/dawn2_-study-results.html. Accessed January 23, 2017.

27.

Thomas

Shin

Jiang

McMahon

Kolassa

Vigersky

. The Development of New Composite Metrics for the Comprehensive Analytic and Visual Assessment of Hypoglycemia Using the Hypo-Triad. J Diabetes Sci Technol. 2017.

28.

American Diabetes Association’s standards of medical care in diabetes—2017. Diabetes Care. 2017;40(suppl 1):S1-S135.

29.

International Hypoglycemia Study Group. 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;40:155-157.

30.

Hirsch

IB.

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

Diabetes Care. 2015;38:1610-1614.

31.

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(6):777-783.

32.

Ceriello

Kilpatrick

ES.

Glycemic variability: both sides of the story. Diabetes Care. 2013;36(suppl 2):S272-S275.

33.

Piconi

Quagliaro

Da Ros

et al . Intermittent high glucose enhances ICAM-1, VCAM-1, E-selectin and interleukin-6 expression in human umbilical endothelial cells in culture: the role of poly(ADP-ribose) polymerase. J Thromb Haemost. 2004;2(8):1453-1459.

34.

Risso

Mercuri

Quagliaro

Damante

Ceriello

Intermittent high glucose enhances apoptosis in human umbilical vein endothelial cells in culture. Am J Physiol Endocrinol Metab. 2001;281:E924-E930.

35.

Sun

Wang

Intermittent high glucose enhances cell proliferation and VEGF expression in retinal endothelial cells: the role of mitochondrial reactive oxygen species. Mol Cell Biochem. 2010;343(1-2):27-35.