A Pilot Study of the Continuous Glucose Monitoring System

RESEARCH DESIGN AND METHODS

Patient selection

Fifty pediatric patients with type 1 diabetes were recruited for this study during routine clinical visits to the Comprehensive Diabetes Center at Childrens Hospital Los Angeles during a 6-month period at the end of 1999 and the beginning of 2000. Patients were eligible to enter if they had an HbA_1c value >8.0% (n = 35) associated with one of the following clinical problems: elevated fasting blood glucose levels or suspected dawn phenomenon; glycemic excursion with exercise; insulin dosage >1–1.5 units · kg^–1 · day^–1 for prepubertal and pubertal children, respectively; widely fluctuating blood glucose levels; or erratic response to blood glucose correction. Subjects with recurrent hypoglycemia, hypoglycemia unawareness, or suspected nocturnal hypoglycemia (n = 12) combined with an HbA_1c level ≤8.0% were also eligible. After being shown the sensor and receiving an explanation as to what a 3-day sensor-wear entailed, all subjects agreed to wear the sensor for a 3-day period. Pump patients were given the option to insert a second sensor. A total of 47 subjects/parents agreed to participate and signed informed consent. The mean age of the subjects was 11.8 ± 4.6 years (range 2.7–29.1). There were 18 male and 29 female subjects. The patients had a mean duration of diabetes of 5.5 ± 3.5 years (range 0.8–20) and a mean baseline HbA_1c of 8.6 ± 1.6% (6.3–12.9). A total of 24 subjects took three to four insulin injections per day, and the remaining 23 subjects used CSII.

Study protocol

Patients and parents were brought to the outpatient clinic at Childrens Hospital Los Angeles for a 2-h training and evaluation session. Sensor placement was done by one of two Certified Diabetes Educators after pretreatment of the abdominal skin where the sensor was placed with EMLA Cream (Astra Pharmaceuticals, Wayne, PA). Calibration of the sensor was accomplished by following the protocol established and outlined in the MiniMed CGMS manual. Four monitors were used during the study.

During continuous glucose monitoring, patients/parents were instructed to measure a minimum of four finger-stick blood glucose levels per day and to record glucose values, meals, insulin doses, exercise periods, and symptomatic hypoglycemia in a logbook. They were also instructed on how to code these events into the monitor.

At the completion of the CGMS period, the system was returned and the data downloaded via the com-station to determine glucose patterns. Logbooks were also analyzed for glucose patterns. A high glucose pattern was defined if at least 2 of 3 days or >50% of glucose levels were >150 mg/dl, and a low-glucose pattern was defined if at least 2 of 3 days or >50% of values were <70 mg/dl during one of the eight time periods. The eight time periods were: breakfast, postbreakfast, lunch, postlunch and afternoon snack, dinner, bedtime, midnight, and 3:00 am. The number of high and low patterns determined by logbooks were compared with the number of high and low patterns seen with the CGMS.

The CGMS patterns were reviewed with the patient/parent, and recommended changes in the diabetes regimen to improve glycemia were discussed. The following nine types of recommendations were made from CGMS analyses: 1) change in one or more bolus or rapid-acting insulin dosage; 2) change in one or more basal or intermediate/long-acting insulin dosage; 3) change in the correction algorithm (3); 4) alteration in the approach to exercise to include change in carbohydrate management (11); 5) reinforcing treatment algorithm for hypoglycemia to include rechecking, adding extra carbohydrate, and protein (11); 6) change in early morning basal rate (pump patients) or bedtime insulin dosage and/or timing (injection patients) due to the dawn phenomenon; 7) increase in meal bolus and timing of injection after the meal (injection patients) or adding square-wave bolus (pump patients) for high-fat meals; 8) increase in meal insulin dosage for high glycemic foods; and 9) referral for counseling to improve adherence with diabetes regimen. Data were analyzed to determine the percentage of patients advised to perform each of these nine recommendations. HbA_1c levels were obtained with the DCA 2000 (Bayer, Tarrytown, NY) (normal range 3–6%) 3 months before CGMS, at the time of sensor placement, and at 3 and 6 months post sensor-wear.

Descriptive statistical analysis was performed on the sample population. This was done to determine the frequency of sensor findings, differences between information obtained from the CGMS and logbooks, and treatment recommendations. Repeated measures of analysis of variance (ANOVA) were used to evaluate change in HbA_1c in subjects. The paired Student’s t test was used to determine differences in HbA_1c levels before and after sensor placement.

RESULTS

All subjects tolerated the CGMS for at least 3 days. The mean duration of sensor-wear was 69.5 ± 28 h (range 29–111). Patients coded a mean of 49 ± 24 events (meals, exercise, and symptomatic hypoglycemia) into the monitor during CGMS use. There was a mean of 18.5 ± 8.0 finger-stick glucose levels recorded in logbooks, resulting in a mean of 4.3 ± 0.7 measurements/day. There was a mean of 1.1 ± 0.9 interruptions in the sensor tracings per subject that were <3 h in duration and did not interfere with data analysis.

Table 1 shows the number of high and low patterns that could be determined using the CGMS as compared with that from the logbook for all subjects and for those on CSII and injection therapy. As shown in Table 1, high and low pattern identification was greater with the CGMS (191 patterns) than with logbook records (42 patterns) for all patients and for those using injection (120 patterns with CGMS and 30 with logbook) and insulin pump therapy (71 patterns with CGMS and 12 with logbook). Overall, there were 1.8 times more abnormal patterns detected for subjects on injection therapy compared with subjects using insulin pump therapy (Table 1).

With the CGMS, a mean of 1.17 ± 1.3 asymptomatic nocturnal hypoglycemic events during the bedtime, night, and early morning time periods were noted per patient, with 83% of subjects having at least one episode. Ten (22%) subjects were found to have an elevation in glucose level during the early morning period that was consistent with the dawn phenomenon. During sensor-wear, 15 (32%) patients were found to have a high breakfast pattern, 15 (32%) had a high lunch pattern, and 12 (25%) had a high dinner pattern, whereas 2 (4%) had a low breakfast pattern, 3 (6%) had a low lunch pattern, and 2 (4%) had a low dinner pattern. In contrast, there were no low postmeal patterns. Eight (17%) patients had a high postbreakfast pattern, 11 (23%) had a high postlunch/afternoon snack pattern, and 9 (19%) had a high bedtime pattern.

Figure 1 illustrates the percentage of patients with whom a recommendation to alter their diabetes regimen was discussed after using the CGMS. As shown, of the nine types of recommendations of how to alter the diabetes regimen, the most common was to make a change in one or more of the subject’s basal insulin dosages for those on CSII and in one or more of the long- or intermediate-acting insulin doses for those taking injections. This was followed by recommendations to change one or more bolus or rapid-acting insulin dose, to change hypoglycemia treatment, and to change the correction algorithm. There was a mean of 3.3 ± 0.8 recommendations per patient.

Table 2 shows the HbA_1c results 3 months before, at baseline, and 3 and 6 months post-CGMS for all patients and those using CSII and injection therapy. Repeated measures of ANOVA showed significance in HbA_1c at these intervals (P = 0.04). Post hoc analysis showed a significant change between time of sensor placement and 3 months post-CGMS as in table 2 (P = 0.03, paired Student’s t test). Comparing HbA_1c levels at the time of sensor placement and at 3 months, 27 patients (25 with HbA_1c >8.0% and 2 with HbA_1c ≤8.0% at baseline) had a reduction in HbA_1c by a mean of 0.66%. These subjects had a mean baseline HbA_1c of 9.08 ± 1.1%. Sixteen subjects (seven with HbA_1c >8.0% and nine with HbA_1c ≤8.0% at baseline) had an increase in HbA_1c by a mean of 0.43%. Their mean baseline HbA_1c was 7.88 ± 0.9%. In four children (three with HbA_1c >8.0% and one with ≤8.0% at baseline), there was no change in HbA_1c between 3 months before and 3 months after continuous glucose monitoring.

CONCLUSIONS

This study showed that the CGMS was well tolerated and beneficial in pediatric type 1 diabetic patients. Patients, some of whom were very young, were able to wear the sensor without problems, and events such as meals, exercise, and symptoms of hypoglycemia were coded in the sensor during daily activities as advised. The glucose tracings derived from the CGMS enabled diabetes care providers as well as patients and their families to discern distinct patterns of abnormal glycemia regardless of whether they were using CSII or insulin injection therapy. The information obtained led to recommendations for adjustment of the diabetes regimen. Follow-up studies determined that these alterations led to improved glycemic control as assessed by mean HbA_1c levels. Although HbA_1c levels were statistically unchanged when retrospectively compared 3 months before and at the time of CGMS placement, these values were improved at 3 months post-CGMS and sustained at 6 months. Because HbA_1c levels had been previously correlated with the number of finger-stick glucose levels obtained in a French study that evaluated 2,579 children with type 1 diabetes (12), it was anticipated that the larger number of glucose measurements done with the CGMS would improve the glycemic outcome of pediatric subjects.

The CGMS revealed that one or more basal, bolus, and correction insulin dosage was not optimal in a large number of patients. At least one-quarter to one-third of subjects were found to have a high premeal pattern (at one or more premeal time periods), and 17–23% were found to have a high postmeal pattern (at one or more postmeal time periods). These findings necessitated increasing one or more basal and bolus insulin dosage in 87 and 70% of subjects, respectively, as well as increasing the dose of insulin used to correct a high glucose level outside of the target range in 32% of patients. It is likely that these recommendations on how to decrease the large number of high pre- and postmeal patterns positively impacted on the follow-up HbA_1c. At the present time, it is controversial whether fasting or postprandial glycemia has more impact on diabetes control (13). With follow-up sensor-wear, it might be possible to determine the relative importance of pre- versus postmeal glucose levels on diabetes outcome. Therefore, the CGMS might prove to have a role in clinical diabetes research that is aimed at addressing areas of controversy in diabetes management.

As recently described by Bode et al. (14), during sensor-wear, nocturnal hypoglycemia and the dawn phenomenon are frequently found but are not apparent with finger-stick monitoring alone because finger-sticks are rarely done at those times. In our patients, the CGMS confirmed what has been described in previous research studies, which showed a high occurrence of asymptomatic nocturnal hypoglycemia in pediatric subjects who measured nighttime blood glucose levels as part of a research protocol (15,16). In our subjects, the finding of nighttime hypoglycemia lead to the recommendation to decrease basal insulin rates during the night for pump patients and evening intermediate-acting insulin for patients on injection therapy. Reducing insulin dosages might have helped decrease further episodes of hypoglycemia by ameliorating hypoglycemia unawareness, which results in part from recurrent low blood glucose (17).

We were surprised by the number of recommendations that were made to alter the diabetes regimen in addition to changing insulin dosages. Over one-half of our subjects did not appropriately manage hypoglycemia. This could be discerned when hypoglycemia was detected by a finger-stick glucose measurement and then subsequent glucose levels were evaluated by the CGMS. Approximately 30% of subjects did not recheck to insure that hypoglycemia had been resolved. In addition to this pattern of persistent hypoglycemia, at least one-quarter of the time, a recurrent episode of hypoglycemia occurred after treatment.

The CGMS data also revealed the frequent occurrence of hypoglycemia associated with exercise in many of our subjects. Some subjects experienced hyperglycemia after the completion of exercise as a result of excess carbohydrate ingestion at the time of hypoglycemia or as a result of disconnecting the insulin pump during exercise. The CGMS data indicated that if glycemic control is to be optimized in the future, patients/families must consider whether the food that is being ingested is high in fat or carbohydrates in order to appropriately alter insulin doses and insulin delivery to account for meal composition. Finally, it should be noted that fewer abnormal patterns were detected in subjects using insulin pumps compared with those taking insulin injections. This suggests that better glycemic control might be achieved with CSII compared with multiple-injection therapy.

Although there was a significant change in HbA_1c over the time intervals, the mean decrease for our study population was modest (0.2 and 0.3% at 3 and 6 months, respectively). These findings were similar to those of Chase et al. (10). This can be explained by the fact that one-quarter of the subjects entered the study with HbA_1c values ≤8.0%. Many subjects in this group had either a small rise (n = 9) or no change in HbA_1c (n = 1) after 3 months, presumably due to a decrease in hypoglycemia. On the other hand, 25 patients with elevated baseline HbA_1c levels >8.0% (mean 9.08%) had a decrease in HbA_1c after the diabetes regimen was altered (mean decrease 0.66%). It is difficult to be sure that the overall improvement in HbA_1c and diabetes management in our subjects that was apparent at 3 months post-CGMS and still evident at 6 months post-CGMS was due to instituting the recommendations that came from sensor-wear itself. Our subjects showed no change in HbA_1c over the 3-month time period before sensor use despite undergoing a multidisciplinary clinic visit that was focused on altering the diabetes regimen to improve glycemic control. However, they did experience improvement in HbA_1c after a self-management counseling session in which the CGMS was used. This suggests that using the sensor and altering each subject’s diabetes regimen as a result of the information obtained from the sensor tracing was an effective method for decreasing HbA_1c.

The data provided by the CGMS for 3 days should be used to help optimize glycemia in pediatric patients with type 1 diabetes. The wealth of information obtained showed pre- and postprandial glucose levels, incidence and timing of hypo- and hyperglycemia, and the effect of exercise and food composition. These data were used to make specific recommendations that improved HbA_1c values in subjects for a 3- to 6-month time period. Whether a more prolonged benefit will be appreciated or repeat continuous glucose monitoring will be required needs to be investigated.