It seems obvious that there should be some level of glycemic control as a clinical requirement for CL control systems. Deciding what measure of glycemic control to select may be problematic. Glycosylated hemoglobin is the obvious choice because most reports of glycemic control include HbA1c as an outcome variable. Recent reports have established the relationship between current laboratory and point-of-care assays and older DCCT standards and have demonstrated a decrease in the calibration variability among various laboratories.30
In addition, a large clinical study has established the relationship between HbA1c and average BG.31
The average HbA1c in the DCCT conventional treatment group was 9.4% (adolescents, 9.8%), while that in the intensive treatment group was 7.5% (adolescents, 8.1%).10,32
As presented in
, average HbA1c in several large databases is approximately 8.2%. Thus, it seems appropriate that an HbA1c value no higher than 8.2% should be a noninferiority clinical criterion for CL control systems.
Superiority outcome criteria for HbA1c could be any HbA1c less than the noninferiority value of 8.2%. However, it is also reasonable to suggest that the HbA1c superiority criterion be the achievement of the American Diabetes Association’s HbA1c goals of 7.0% for adults, 7.5–8.5% in toddlers and preschoolers (ages 0–6 years), <8.0% in school-aged children (6–12 years), and <7.5% in adolescents (13–19 years).33
Such goals should be achievable in well-motivated compliant individuals given the results of the STAR 3 trial, in which the average HbA1c achieved was 7.3% in those over 19 years old and 7.9% in those 7–19 years old.27
demonstrates the clinical problem associated with arbitrarily selecting an HbA1c level as a criterion for CL control systems. This figure represents the CGM tracings of two individuals with identical HbA1c values. It is apparent to even the casual observer that the glycemic control of these individuals differs significantly. While controversy still exists over the clinical significance of glycemic variability to short- and long-term health among T1DM individuals, it is clear that CGM has made it possible to observe this phenomenon to an extent unavailable previously.34–37
We will never be able to know the influence of variability on the results of the DCCT based on the data collected during that study.10
Given that glycemic variability is greater among persons with T1DM than among those with no diabetes, it seems reasonable that a goal of clinical care would be to reduce variability as much as would be feasible.36,37
Deciding how to analyze variability is not an easy task.38
It is important to note that standard deviation (SD) is not a recommended statistic to describe BG variability because the BG scale is asymmetric; BG values are not normally distributed. Therefore, SD is influenced more by hyperglycemia than by hypoglycemia. Statistics such as MAGE (mean amplitude of glycemic excursions), which are based on SD, are similarly insensitive to hypoglycemia. However, SD is an appropriate statistic for use when describing the rate of change of BG scale and the stability of CL control overtime.
Continuous glucose monitor profiles of two patients with T1DM and identical HbA1c values.
Perhaps the simplest way to characterize variability is by recording time spent within a target range, for instance 70–180 mg/dl, as well as time spent below and above that range. Such computations may be tedious but sophisticated CGM data recorders can provide this information easily. The Diabetes Research in Children Network study group has demonstrated an increase in the percentage (52–60%) of BG values with the range of 70–180 mg/dl among children using a CGM system over 12 weeks.39
Others have shown a 21% reduction in time spent with BG <55 mg/dl, a 23% reduction in time spent with BG >240 mg/dl, and a 26% increase in time spent within a range of 81–140 mg/dl among adults using a CGM system for as short a time period as 3 days.24
In the JDRF CGM study, patients >25 years old reduced their mean min/day with BG >180 mg/dl while increasing their mean min/day with BG between 71 and 180 mg/dl.26
In the STAR 3 trial, adults and children reduced the area under the glucose curve measured when BG >180 mg/dl.27
More sophisticated statistics can be determined from CGM data that can provide important information regarding variability and risk assessment for both high and low BG.38
The blood glucose risk index (BGRI), the sum of the low blood glucose index (LBGI) plus the high blood glucose index (HBGI), provides a measure of the extent and frequency of BG fluctuations. Studies have demonstrated that the LBGI can predict 40–50% of the variance in the prediction of future low BG (BG < 70 mg/dl), while the HBGI correlates with postprandial BG levels and HbA1c. Visual interpretations of CGM data include histograms of rate of BG change and Poincaré plots of BG (t(i-1)
) vs BG (ti
), which demonstrate the stability of the BG system. Control variability grid analysis permits graphical representation of minimum vs maximum BG levels over a designated period of time.40
It remains to be determined which of these representations of variability will emerge as a standard.