Continuous Subcutaneous Glucose Monitoring Improved Metabolic Control in Pediatric Patients With Type 1 Diabetes: A Controlled Crossover Study

METHODS

The CGMS sensor monitors interstitial glucose levels (2.2–22 mmol/L) in subcutaneous tissue every 10 seconds and records an average value every 5 minutes.¹⁴ The subcutaneous glucose oxidase electrode is viable for up to 72 hours. Data are downloaded to a computer providing a continuous tracing of BG values in the form of daily BG profiles and a summary table of average glucose levels, glucose ranges, and standard deviations. The lag time between CGMS and BG after a meal has been found to be 4 minutes in rise and 9 minutes in fall.¹⁵ Local anesthetic EMLA Cream (Astra, Södertälje, Sweden) was used before all insertions. Most used a special instrument (Senserter) to minimize insertion pain.

RESULTS

We registered 298 sensor profiles (both open and blind arms included) with an average sensor life of 2.1 ± 1.0 days. This corresponded to 642 days of glucose measurements (partial and complete days added to 24-hour periods). In both the open and blinded study arms there was a decrease in HbA₁C during the first 3-month period. After crossover, the now open arm had a further decrease in HbA₁C (caused mostly by a decrease in nocturnal high subcutaneous glucose readings), whereas the now blind arm increased again (Fig 1). Altogether there was a significant decrease in HbA₁C during the open arm when there was access to the CGMS data from 7.70% to 7.31% (P = .013; Fig 2), whereas the decrease during the “blind” arm was from 7.75% to 7.65% (P = not significant). There were on average 1.5 episodes/day of daytime high subcutaneous glucose (>15 mmol/L, duration 126 ± 33 minutes, 19.4% of total time) and .6 episodes/night (duration 177 ± 83 minutes, 25.5% of total time). Low glucose values were measured quite frequently, as exemplified in Fig 3. Twenty-six of 27 patients experienced daytime low subcutaneous glucose (<3.0 mmol/L, .8 episodes/day, duration 58 ± 29 minutes, 5.5% of total time) and all patients had at least 1 nighttime episode of low subcutaneous glucose (Fig 3; .4 episodes/night, duration 132 ± 81 minutes, 10.1% of total time) during the study. We found no difference in low subcutaneous glucose frequency between the 2 treatment arms, nor between MDI and continuous subcutaneous insulin infusion users.

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

HbA₁C (Swedish calibration, ∼1.2% below DCCT standard) values during the study. After 12 weeks, the treatment arms crossed over. Comparisons between the start and end of the 12-week periods are done with the 2-sided t test. The P values above the graph refer to the blinded sensors, below the graph to the open sensors.

• Download figure
• Open in new tab
• Download powerpoint

Fig 2.

HbA₁C values during the open and blind arm for all 27 patients. The difference between baseline and 12 weeks was significant for the open arm (7.70% vs 7.31%; P = .013; 2-sided t test) but not for the blind. The P values in the figure refer to the difference between the arms, which was significant at 12 weeks (7.65% vs 7.31%; P = .011; 2-sided t test).

• Download figure
• Open in new tab
• Download powerpoint

Fig 3.

All patients showed low (<3 mmol/L) nocturnal subcutaneous glucose on 1 or more occasions, and all but 1 had at least 1 episode of low daytime subcutaneous glucose. This chart is from an 11-year-old girl with HbA₁C of 5.7%.

Fig 4 illustrates how individual patients registered abnormal but quite systematic glucose profiles measured by the sensor, which led to recommendations of treatment changes and improved glucose profiles as a consequence. Examples of changes in treatment are both increased and decreased bedtime insulin dose/nighttime basal rate, change in premeal bolus dose and timing, changed type of basal insulin in MDI, and change in the amount of extra insulin used to correct high BG levels. Dawn phenomena, defined as a nighttime subcutaneous glucose level of >3.5 mmol/L (absence of hypoglycemia), followed by a spontaneous rise in subcutaneous glucose of 7 mmol/L in the early morning hours (Fig 5), were found in 5.3% of the total number of recorded days (in 10 patients). Somogyi phenomena, defined as a similar spontaneous rise of subcutaneous glucose during night but preceded by a low subcutaneous glucose (<3.5 mmol/L), occurred in 13.3% of the days (in 17 patients). Somogyi phenomena were less common in the pump patients (5/13 vs 12/14 patients, P = .018, χ²). Daytime rebound phenomena (Fig 6) with a rapid increase (within 3 hours) of subcutaenous glucose from <3.5 mmol/L with >10 mmol/L were found in 25.7% of the days (in 24 patients). Two patients experienced severe hypoglycemia during the study: 1 during the blind arm and 1 during the open.

• Download figure
• Open in new tab
• Download powerpoint

Fig 4.

Modal day view from a 13-year-old boy with HbA₁C of 7.3%. The variations over the day and night are sometimes very similar for the 3 testing days. Glucose concentration in millimoles per liter.

• Download figure
• Open in new tab
• Download powerpoint

Fig 5.

Example of “Dawn phenomenon” in a 16-year-old girl with HbA₁C of 7.4%.

• Download figure
• Open in new tab
• Download powerpoint

Fig 6.

A 9-year-old girl with HbA₁C 7.0% showing 2 rebound phenomena (arrows). It is unlikely that the very rapid increase in BG is caused only by eating as a girl of this weight (36.6 kg) would need ∼50 g of quick-acting sugars to accomplish a rise of close to 20 mmol/l in such a short time.

The glucose sensors gave important information, but there were several practical problems. A total of 104 of the 298 (34.9%) sensors diplayed a 3-day curve, whereas 154 of the 298 (51.7%) curves gave information during 1 to 2 days. In 25 of the 298 (8.4%) cases, no sensor curves were generated. In some cases, the cable had to be exchanged. A 7-year-old girl left the study after having tried the sensor during the first 3-day period, as she disliked wearing it.

DISCUSSION

Our results indicate that changes in intensive insulin therapy based on CGMS profiles improved metabolic control in pediatric patients with type 1 diabetes. As the device was completely new for us, one could expect less effect on HbA₁C during the first period, when many initial practical problems (such as forgetting to calibrate the monitor or enter self-monitoring of BG readings in the appropriate time frame) were encountered. We were surprised and cautioned by the many low subcutaneous glucose periods, not least during nighttime, which led to decreased insulin doses in the open arm during the first 3-month period. Our lack of experience of CGMS before the study, leading to insufficient/ineffective information to the patients/parents, may explain some of the practical problems. This led to some difficulties with the interpretation of the glucose sensor profiles, especially in the beginning of the study. Despite such initial problems, we were able to decrease HbA₁C significantly during the open arm when we received information about the glucose profiles, without increasing the frequency of hypoglycemia. Improvements in HbA₁C, even in patients initially randomized to the blinded study arm, may be attributed to a study effect related to more frequent visits to the diabetes team and enthusiasm to see CGMS profiles in the future open study arm. For most patients, the number of 7-point profiles taken during the study months did not differ from their regular routines. It is remarkable that HbA₁C decreased as the large number of periods with very low subcutaneous glucose led to decreased insulin doses in several patients, especially during the first 3-month period. The return of HbA₁C to baseline levels in the group that ended with the blind arm indicates a need for repeated use of the sensor, especially in this group of patients including many teenagers with unstable diabetes control.

The number and duration of low subcutaneous glucose readings is surprising and perhaps even scary. Other authors have also found a surprisingly high frequency of nighttime low subcutaneous glucose after only 2 to 3 days use of CGMS: 67.8% of patients aged 2 to 18 years¹¹ and 83% of patients aged 3 to 29 years.¹³ There are several possible explanations. Glucose in subcutaneous tissue fluid has been said to be in rapid steady state with BG, but we cannot be sure that this is the case during very low glucose concentrations when glucose might be distributed to the organs with greatest need, eg, the brain.¹⁸ If subcutaneous glucose does not reflect BG in the very low range this would be a great pity, as a subcutaneous sensor then cannot be used as an alarm preventing hypoglycemia. Does sleeping on your belly affect subcutaneous blood flow, thus lowering the registered values? A third, interesting, possibility would be that our present paradigm regarding BG concentrations is wrong. Thus, perhaps BG concentrations might go down to 2 to 3 mmol/L without causing neither counter regulation nor harm? This would actually fit with our own observations (J. Ludvigsson, unpublished observations) that many healthy school children may have BG values down to 2.7 mmol/L, without showing any symptoms of hypoglycemia. Interestingly, when we tried the glucose sensor on the staff before the study, several nondiabetic adults noticed readings below 2.2 mmol/L during the night. More studies are needed, but so far we have to act aiming at prevention of hypoglycemia.^6,19

Our study showed wide and rapid glucose fluctuations in all patients. Whether spontaneous increase of BG in the early morning should be defined as Dawn phenomena, Somogyi phenomena, or are just consequences of fluctuations of insulin concentrations cannot be answered by our study. Elevations of glucose levels in the early morning were found in 22% of pediatric patients after using CGMS at 1 occasion.¹³ We can only notice that such rapid fluctuations exist in children and adolescents. Daytime so-called rebound phenomena are even more common, sometimes seemingly caused by hormonal counterregulation, but at other times probably caused by eating too much extra food to cure hypoglycemia. This knowledge can lead to an insight which improves both treatment routines and metabolic control.

Some patients did not want to try the device at all as it meant new pricks, a subcutaneous catheter and an instrument connected to the body, which may be disturbing during sports, baths, and perhaps during sleep. However, most patients accepted the device, and in most cases we got very informative glucose profiles during 48 to 72 hours. Too many interruptions of the glucose curves, unexpected alarms, and nonfunctioning wires certainly diminished the enthusiasm, but these practical problems have decreased with more practical experience of the method, and could perhaps be overcome with improvements of the device.

CONCLUSIONS

Our experience of CGMS in diabetic children and teenagers is mainly good and we expect this device to be an important complement of our treatment choices and a useful educational tool. The sensor gave valuable information that we have never had before. We are now turning this device into clinical practice. We recommend the use of CGMS in patients with elevated HbA₁C, a worrying increase of HbA₁C, in patients with known or suspected hypoglycemia during the night or tendency to severe hypoglycemia, patients who lack motivation for self-monitoring of BG or who do not understand how to interpret BG profiles and how to adjust insulin dose, and finally in patients who want to learn more about the effect of meals or physical exercise on their BG.