Differences in the Mean Amplitude of Glycemic Excursions Calculated by Two Automated Methods

Dear Editor:

Glycemic variability has been implicated in the development of diabetes-related late complications. ^1,2 Service et al. ³ first described the mean amplitude of glycemic excursions (MAGE) as the arithmetic mean of the blood glucose level increases or decreases, from blood glucose nadirs to peaks, or vice versa, when both ascending and descending segments exceeded the value of 1 SD of the blood glucose for the same 24-h period. MAGE has been widely used in the literature and has been considered the “gold standard” for measuring glycemic variability by some authors, as a relationship has been found between the MAGE value and the urinary excretion of 8-isoprostaglandin-F2α, a marker of oxidative stress. ⁴ Continuous glucose monitoring (CGM) systems can assess glucose variability more accurately than self-monitoring of blood glucose measurements. The graphical approach to calculate MAGE from CGM data is operator-dependent, so several computer programs have been developed in order to standardize MAGE calculation. ^{5 –8}

The aim of the study was to calculate the MAGE value in CGM data from type 1 diabetes patients by two different computerized methods and to compare these MAGE results in order to verify if both methods generate similar values when applied to the same datasets.

We performed CGM with the IPro™2 (Medtronic MiniMed, Northridge, CA) for 4.2±1.5 (range, 1–6) days in 57 patients with type 1 diabetes, 28 males and 29 females, 26±15 (range, 8–69) years of age, with a diabetes duration of 11.3±7.5 (range, 2–41) years. The study followed the ethical standards of Declaration of Helsinki. Only complete 24-h CGM profiles, including 288 readings, were considered. Each CGM trace was calibrated by at least four finger-prick blood glucose measurements. Only those recordings with a mean absolute difference between CGM and capillary blood glucose of <28% were included. In total, 248 24-h monitorings were analyzed. Each 24-h CGM profile was retrospectively analyzed, and MAGE was calculated by GlyCulator software ⁵ (MAGE1) and by the EasyGV^© program ⁶ (MAGE2), both of which are available online (available at, respectively, www.pediatria.umed.pl/team/glyculator and www.easygv.co.uk).

Data were analyzed using SPSS software version 15.0 (SPSS, Inc., Chicago, IL). Correlations between continuous variables were analyzed by Pearson's test. A value of P<0.05 was considered statistically significant.

Mean glucose and SD levels were identical when both calculation programs were used, which confirmed that the analyzed readings were the same. MAGE1 and MAGE2 correlated positively (r=0.663; P<0.0005). Mean MAGE1 was 167.19±51.89 mg/dL, and mean MAGE2 was 126.44±52.59 mg/dL (P<0.0005). The MAGE1/SD and MAGE2/SD ratios were 2.92±0.36 and 2.21±0.61, respectively. The MAGE1/MAGE2 ratio was 1.45±0.54. Of the differences between MAGE1 and MAGE2, 214 (86.3%) were positive, whereas 34 (13.7%) were negative, all but one between −1 and −48 mg/dL. The frequency and percentage of results in each group of absolute differences are shown in Table 1. In 9.3% (n=23) of the calculations, the absolute difference between MAGE1 and MAGE2 was >100 mg/dL. These datasets showed higher SD and higher variation coefficient than the datasets in which the differences between MAGE1 and MAGE2 were lower than 100 mg/dL (78.74±15.82 mg/dL vs. 56.01±18.34 mg/dL [P<0.0005] and 43.70±12.98% vs. 37.13±12.34% [P=0.016], respectively).

OPEN IN VIEWER

Table 1.

Absolute Difference in Mean Amplitude of Glycemic Excursions Between Two Methods of Calculation

Absolute difference (mg/dL)	n	%
0–50	163	65.7
51–100	62	25
101–150	19	7.7
>150	4	1.6

We conclude that MAGE results from both automated methods correlate positively, although MAGE2 underestimates the results in relation to MAGE1. Most of the differences between MAGE1 and MAGE2 are slight, but in some cases the differences can be significant. The identification of blood glucose peaks and nadirs in CGM traces can be subjective as the MAGE calculation can be troublesome. It is not possible to determine whether the two programs are considering the same peaks and nadirs for any particular dataset. Higher glycemic variability, with more peaks and nadirs in the CGM profiles, could be responsible for the two methods giving the higher differences in the results. The method used to calculate MAGE from CGM data should be routinely specified, and methods for MAGE calculations should be validated in order to allow the comparison of glycemic variability indexes in different studies.