Diabetes Technol Ther. 2013 Jun;15(6):448-54.
Evaluation of the mean absolute glucose change as a measure of glycemic variability using continuous glucose monitoring data.
Kohnert KD, Heinke P, Fritzsche G, Vogt L, Augstein P, Salzsieder E.
Gerhardt Katsch Institute of Diabetes, Karlsburg, 17495, Germany. email@example.com
Klaus-Dieter Kohnert, M.D., Ph.D.
Institute of Diabetes “Gerhardt Katsch” Karlsburg, Germany
Dept of Clinical Diabetes Research
Greifswalder Str. 11e
17495 Karlsburg, Germany
The mean absolute glucose (MAG) change, originally developed to assess associations between glycemic variability (GV) and intensive care unit mortality, has not yet been validated. We used continuous glucose monitoring (CGM) datasets from patients with diabetes to assess the validity of MAG and to quantify associations with established measures of GV.
SUBJECTS AND METHODS:
Validation was based on retrospective analysis of 72-h CGM data collected during clinical studies involving 815 outpatients (48 with type 1 diabetes and 767 with type 2 diabetes). Measures of GV included SD around the sensor glucose, interquartile range, mean amplitude of glycemic excursions, and the continuous overlapping net glycemic action indices at 1, 3, and 6 h. MAG was calculated using 5-min, 60-min, and seven-point glucose profile sampling intervals; correlations among the variability measures and effects of sampling frequency were assessed.
Strong linear correlations between MAG change and classical markers of GV were documented (r=0.587-0.809, P<0.001 for all), whereas correlations with both glycosylated hemoglobin and mean sensor glucose were found to be weak (r=0.246 and r=0.378, respectively). The magnitude of MAG change decreased in a nonlinear fashion (P<0.001), as intervals between glucose measurements increased. MAG change, as calculated from 5-min sensor glucose readings, did reflect relatively small differences in glucose fluctuations associated with glycemic treatment modality.
MAG change represents a valid GV index if closely spaced sensor glucose measurements are used, but does not provide any advantage over variability indices already used for assessing diabetes control.
Glycemic variability is an important measure to characterize the dynamics of glucose profiles. It has been shown to predict hypoglycemia in type 1 as well as in type 2 diabetes mellitus and was found to be associated with mortality in intensive care unit patients. Glycemic variability may add or modify the risk of diabetes complications, but conclusive evidence for its role was not yet provided. When considering a role for glycemic variability in the development of diabetes complications, one should acknowledge that variation in glucose homeostasis is a physiological process. In the normoglycemic state; however, variability is small in amplitude and relatively uncorrelated, whereas in diabetes, amplitudes are greater with retention of interdependence of adjacent glucose profile values. It is thus mandatory to choose appropriate measures of glycemic variability restricted to glucose fluctuation that exceed those observed in normoglycemic subjects.
With the aim of accurately assessing glycemic variability, several clinically useful indices have been proposed and thoroughly characterized by Rodbard (1). The fact that the number of glycemic variability indices is still growing indicates that no consensus could be achieved regarding which ones are the most useful for analysis of glucose profiles. Novel measures, taking a time component into account, have been proposed to assess glycemic variability. One of these includes the mean absolute glucose (MAG) change, which was used to assess glycemic variability from infrequently sampled glucose values mainly in intensive care (2). However, as this measure integrates minor as well as major glucose swings, it is questionable whether it is an appropriate index for assessment of glycemic variability. Since the applicability of MAG for continuous glucose monitoring has not previously been established and a thorough validation was lacking, we first scrutinized whether MAG results were influenced by the number of glucose measurements. Indeed, we found that MAG results were highly dependent on the number of glucose measurements, such that MAG values dropped from 288 to 24, to seven measurements per 24 h by roughly 25%. We further revealed interesting correlations between MAG and classical indices of glycemic variability, such as standard deviation (SD), the mean amplitude of glycemic excursions (MAGE) and interquartile range (IQR). For example, when comparing the variability indices derived from CGM measurements (288 measurements per 24 h), SD has a high correlation with MAGE and IQR (r = 0.931 and 0.860, respectively), but a markedly lower correlation with MAG (r = 0.587).
Since appropriate therapies have been shown to be capable of reducing glycemic variability in diabetes patients, we also performed a detailed analysis of various antihyperglycemic agents on changes that occurred in the magnitude of MAG. Although we observed a progressive increase between different treatment groups, as with classical indices, when moving from diet alone, to oral drug monotherapy, to combination oral therapy, to oral drugs combined with insulin, to insulin monotherapy, the mean values for MAG shown in Figure 1 did not greatly differ across the treatment groups. Use of the classical variability indices, however; provided better assessment of therapeutic effects. We further demonstrated that the sensitivity to detect pharmacological effects decreased for the indices of glycemic variability on an ordinal scale: IQR>SD>MAGE>CONGA1>MAG (Figure 2). As can be seen, the mean percentage difference for IQR, SD, and MAGE were of similar magnitude (27, 25, and 22%, respectively) but clearly smaller for MAG (14%). In view of requirements for optimized diabetes control, it appears important to choose the most sensitive indices for evaluation of glycemic variability.
The importance of this large cross-sectional study is threefold: First, it shows that MAG, if computed from continuous glucose monitoring profiles with 5-min sampling intervals, is closely related with widely used indices of glycemic variability but less with conventional measures of overall glycemic control (mean glucose and HbA1c). However, MAG results are highly dependent on the frequency of glucose sampling―lower numbers of glucose measurements used in conventional blood glucose profiles result in approximately four-fold lower MAG values, as compared to those derived from continuous glucose monitoring.
Second, irrespective of the kind of index, there is a progressive increase in glycemic variability in patients with type 2 diabetes when moving from diet to insulin therapy and finally to patients with type 1 diabetes. This is largely attributable to a progressive decline of postprandial ß-cell function in type 2 diabetes (3) and the almost complete loss of ß-cell reserve in type 1 diabetes. Among the glycemic variability indices used in our study, the MAG index is the least sensitive one to differentiate between therapies on lowering glycemic excursions. With regard to the limitations of MAG, it does not qualify for measurements of clinically relevant glucose excursions.
Third, our study has clearly shown that determination of glycemic variability requires continuous glucose measurement. Infrequently sampled glucose measurements at different time points provide poor assessment of glycemic variability and make comparison of therapeutic interventions unreliable.
- Rodbard D 2011 Glycemic variability: measurement and utility in clinical medicine and research―one viewpoint. Diabetes Technol Ther13:1-4
- Hermanides J, Vriesendorp TM, Bosman RJ, Zandstra DF, Hoekstra JB, DeVries JH 2010 Glucose variability is associated with intensive care mortality. Crit Care med 38: 838-842
- Kohnert KD, Freyse EJ, Salzsieder E 2012 Glycemic variability and ß-cell dysfunction. Current Diab Res 8: 345-354
Figure 1. Increase of glycemic variability, using magnitude of mean absolute glucose (MAG) change as index, in patients with patients with type 2 (colored columns) when moving from diet, to oral antidiabetic drugs, to oral therapy combined with insulin, to insulin treatment alone, and to type 1 diabetes (T1DM, white column). Note that MAG decreases upon widening of glucose sampling intervals. Oral antidiabetic drugs (OAD), OAD + insulin (OAD + ins), and insulin (ins).
Figure 2. Mean percentage difference in glycemic variability between antidiabetes treatment categories for interquartile range (IQR), standard deviation around mean sensor glucose (SD), mean amplitude of glucose excursions (MAGE), continuous overall net glycemic action (CONGA1), and mean absolute glucose (MAG) change. Among these indices, MAG shows the lowest sensitivity to detect pharmacological effects on glycemic variability.