Can Continuous Glucose Monitoring Systems Predict Glycemia?

Background

G lucose control improves in adults with type 1 diabetes (T1D) using a continuous glucose monitoring system (CGMS), ^1,2 but its value for the prevention of hypoglycemia is a matter of debate. ³ The accuracy is lower than with the capillary measurements by self-monitoring of blood glucose (SMBG), and there is a lag time in the interstitial glucose measurement, so the capillary glucose value is still recommended for acute treatment decisions. ⁴ An interesting feature of a CGMS is that it provides a direction of the interstitial glucose change, indicated by an arrow on the screen of the CGMS devices. It may help the patient to perform proactive adaptations, especially to prevent hypoglycemia, as proposed in some educational algorithms. ⁵ However, we do not know whether the CGMS arrows really predict the subsequent capillary glucose level. To answer this question, as a part of their CGMS educational program, our patients with T1D measured capillary glucose levels at time 0 and 15 min after noting their interstitial glucose value and its direction.

Subjects and Methods

The 23 subjects were 12 women and 11 men, 40±13 years old. They had had T1D for 19±12 years, and their hemoglobin A1c level was 8.5±1.5%. All were on an intensive insulin regimen: 13 on continuous subcutaneous insulin infusion and 10 on four insulin injections per day at least. After their admission to the Diabetology Unit, they were provided with glucose sensors, by switching to a Veo™ pump (Medtronic Diabetes, Minneapolis, MN) for the subjects treated by continuous subcutaneous insulin infusion or a Freestyle^® Navigator (Abbott Diabetes Care, Alameda, CA) for the others. The patients used the glucose meters of the Freestyle Navigator to calibrate it, and their own glucose meters to calibrate the Medtronic CGMS. All the glucose meters were checked for accuracy with control solutions according to the recommendation of the manufacturer prior to entry into the study. On the morning of the day after admission, the capillary glucose measured with the glucose meters was 178±108 mg/dL, compared with blood glucose levels measured by the glucose oxidase method of 184±118 mg/dL; they did not differ significantly, and they were well correlated (r=0.99).

Our structured educative program began with learning how to insert and calibrate the sensor according to the manufacturer's instructions. Then the patient tested the accuracy of the sensor compared with simultaneous capillary glucose measurements, before learning how to set alarms and to perform reactive and proactive adaptations. At this accuracy-testing step, the patient noted the CGMS arrow and was requested to measure interstitial and capillary glucose levels 15 min later to check whether the arrow had predicted the glucose level course. No changes in treatment were instigated between the two glucose measurements, except in the event of hypoglycemia with clinical symptoms. We asked the patients to perform as many observations (one observation=one arrow and two interstitial plus two capillary glucose levels) as necessary to obtain their opinion on the reliability of the CGMS arrows.

The results are expressed as mean±SD values. The interstitial and capillary glucose levels were compared by paired t tests. The glucose level courses (Time+15 min – Time 0) were compared between arrows by analysis of variance with Bonferroni's correction. As the arrow distribution was one-fifth descending, three-fifths stable, one-fifth ascending, the capillary glucose courses were categorized as quintiles before performing a crosstab analysis by χ² test. The analysis was performed using SPSS software (version 10.0; SPSS, Inc., Chicago, IL).

Results

The 23 subjects performed 242 observations (11±6 observations per subject). Three additional observations were excluded from the analysis because the subjects had intervened owing to a manifestation of hypoglycemia. The interstitial glucose levels (139±57 mg/dL) were lower (P<0.0001) but well correlated (n=484; r=0.78; P<0.0001) with the capillary levels (155±59 mg/dL).

The first arrow was descending in 55 observations, stable in 147, and ascending in 40. The outcome of the glucose levels in the three situations is summarized in Table 1. The 55 descending arrows had subsequently lower capillary glucose levels, significantly lower than the stable (P<0.001) and ascending (P<0.01) arrows, which did not differ significantly from one another. Similar glucose changes were observed after double arrows. Two-thirds of the stable arrows corresponded correctly to the three median quintiles of subsequent capillary glucose levels. Gross errors included six descending arrows in the most increased capillary glucose and five ascending arrows in the most decreased capillary glucose, representing 4.5% of the observations, but there were large proportions of stable capillary glucose levels after ascending (60%) and descending (45%) arrows.

These analysis were also separately performed for the Medtronic (n=140 observations) and Abbott (n=102 observations) sensors, leading to the same significantly more decreasing capillary glucose after descending arrows (−14±18 mg/dL for Medtronic and −17±40 mg/dL for Abbott), significant crosstab analysis (by χ² test: P=0.004 for Medtronic and P=0.04 for Abbott), and similar numbers of gross errors (seven of 140 for Medtronic and four of 102 for Abbott).

These differences were also significant when the analyses were repeated on the 63 observations with an initial interstitial glucose below 100 mg/dL. The 15 of 63 descending arrows had lower subsequent capillary glucose levels: −10±16 mg/dL versus +5±16 mg/dL (n=41) (P<0.01) for stable arrows and +11±27 mg/dL (n=7) (P<0.05) for ascending arrows.

Discussion

The prediction of glucose levels has long been a concern for researchers. ⁶ With appropriate models ⁷ and algorithms, ⁸ CGMS can be used for predictive monitoring. ^9,10 For the patients with T1D, the arrows on the CGMS screen provide additional information. They depict a trend from past and present interstitial measurements, but they also suggest what the next glucose will be, which may allow anticipative adaptations. Traditional SMBG cannot provide this information because of random errors when capillary glucose levels are measured repeatedly at brief intervals. Our use of SMBG as a reference is a limitation because SMBG data points can carry errors, which are amplified when their difference is taken to assess the trends over 15 min: these errors may account for some of the 4.5% gross errors of arrows that we found. SMBG is, however, what the patients use in real life for practical decisions, and our finding that a descending arrow is more often than not followed by a lower capillary glucose level than other arrows is of practical importance for the patients who want to prevent hypoglycemia.

The value of glucose trends from CGMS has been examined by other authors: a rate accuracy, “R,” was defined, ¹¹ criticized, ¹² and compared, with no major difference between the presently affordable devices. ¹³ As documented for the point glucose value, ¹⁴ the accuracy of the “R” determination is far from perfect: the agreement between simultaneously worn CGMS devices is low in conditions of rapid glucose change. ¹⁵ The low point-accuracy probably explains why the benefit of CGMS for the detection of hypoglycemia is not a universal finding. ^16,17 Good algorithms are probably critical to provide pertinent arrows from glucose trends: Dassau et al., ¹⁸ who included glucose trends in their hypoglycemia prediction algorithm, could, however, predict 91% of experimentally produced hypoglycemia episodes in 22 subjects with T1D. Our demonstration that a descending arrow was associated with a further decreasing capillary glucose when interstitial glucose was below 100 mg/dL also argues for the interest of CGMS in the prevention of hypoglycemia in real-life situations. The discrimination of later increase or decrease of capillary glucose by the arrows of the CGMS is, however, far from perfect, with 4.5% gross errors: a decrease after an ascending arrow, or the opposite. Patients who use a CGMS must be aware of this uncertainty, for their safety and to prevent loss of motivation, and we feel that it is beneficial for them to gain this experience as part of their CGMS education.