The use of CGM to identify hypoglycemia and glycemic patterns in congenital hyperinsulinism

Introduction

Patients with congenital hyperinsulinism (HI) can experience significant morbidity, mainly secondary to hypoglycemia. In patients with HI, inappropriately elevated insulin concentration leads to hypoglycemia. Insulin suppresses ketogenesis, hence depriving the body and brain of other sources of energy. This can have profound consequences for the neonate and infant’s brain development, whether the disease is transient or permanent. Studies have shown neurological impairment with associated MRI findings as high as 50 % in children with HI [1, 2].

Prompt diagnosis of hypoglycemia should be the primary focus in management of infants with HI, and treatment should be tailored to each patient’s glycemic patterns. Typically, hypoglycemia in children and adolescents with HI is treated with carbohydrate supplementation and/or medications that inhibit insulin secretion such as diazoxide, which works to keep the K+ ATP channels open, or octreotide, a somatostain analogue. Point-of-care testing (POCT) using capillary glucose meters is the standard of care for blood glucose monitoring in patients with HI. The frequency varies from one to more than eight measurements per day according to the perceived severity of the disease, number of symptomatic episodes, and the age of the patient. At our tertiary academic care center, most patients with HI are seen every 3–6 months.

Intermittent POCT only provides a snapshot information of a patient’s glycemic trends. Even if measurements are performed multiple times per day, they do not accurately reflect glycemic patterns and may fail to detect hypoglycemia, particularly when they occur overnight or in patients that are asymptomatic. In addition, POCT can be inconvenient and add significant distress and burden for patients and their caregivers. In newborns and infants, heel lance, the routine method of drawing blood, is recognized as a painful procedure, and is usually done without analgesia or sedation [3].

Continuous glucose monitoring (CGM) may provide a solution to some of these challenges. These sensors provide continuous interstitial glucose readings. For Dexcom G6^®, glucose readings are reported every 5 min, for a total of 288 values/day. Most studies assessing the accuracy of these devices have been conducted in people living with diabetes. Whereas older device generations such as Dexcom G5^® had a high false positive rate of hypoglycemia [4] and needed calibration with POCT by the user, Dexcom G6^® has shown more precise and accurate results during hypoglycemia, specifically with a mean absolute difference (MAD) of 0.43 mmol/L when blood glucose is between 3.0 and 3.9 mmol/L, and a MAD of 0.6 mmol/L, when blood glucose is less than 3.0 mmol/L [5]. In patients with diabetes, CGM has been shown to improve glycemic control and reduce the frequency of hypoglycemia, and is now considered standard of care [6].

Recent studies have evaluated the use of CGM in pediatric patients with HI. Braune et al. [7] used Dexcom G6^® in a newborn with transient hyperinsulinism to decrease the frequency of POCT. A randomized, controlled, multicenter study in preterm neonates showed that CGM could help decrease exposure to prolonged or severe hyperglycemia and hypoglycemia in this at-risk group [8]. Worth et al. [9] evaluated a large dataset of pediatric patients with HI using either Dexcom G4^® or G6^® and found that patients experienced a mean of 1.3 hypoglycemic events per day (and 2.6 hypoglycemic events when not on therapy). Despite its use in this patient population, CGM has not been approved since there remain challenges with the use of CGM, particularly related to accuracy [10], [11], [12].

To better understand the frequency of hypoglycemia and glycemic patterns in patients with HI on stable medical therapy, we performed a pilot study and hypothesized that children and adolescents with HI on stable medical therapy can still experience asymptomatic hypoglycemia not recognized by routine POCT, which can be detected using CGM.

Materials and methods

Study design and participants

This is a prospective, pilot observational study in children 0–18 years old with HI followed by the endocrinology division at British Columbia Children’s Hospital (BCCH), a tertiary pediatric academic center located in Vancouver, Canada.

All children under the age of 18 years with a genetic or clinical diagnosis of HI under medical treatment with diazoxide or octreotide were eligible to be enrolled in the study, and were identified from the endocrinology database at BCCH. Exclusion criteria included the use of CGM 60 days prior to the start of the study, and changes in medical therapy or a hospital admission 15 days prior to the start of the study.

After identification, each chart was manually reviewed to determine eligibility. A letter of study introduction was sent to all potential participants and/or their caregivers. They were then contacted by the research coordinator who presented the research project. Participants were enrolled after informed consent and assent (as applicable) were obtained.

The duration of the study was 20 days. The study timeline is shown in Figure 1. At their first visit to the study site, participants had a blinded Dexcom G6^® sensor inserted. They were also given a new glucometer, Contour Next^®, and asked to monitor their blood glucose, as advised by their primary endocrinologist, and to continue the same medical therapy. The blinded use of Dexcom G6^® was mandatory for the first 10 days of the study. At the second visit (after 10 days), data was uploaded to the Dexcom software, Clarity, and reviewed by the investigators. If prolonged or severe hypoglycemia episodes were detected by CGM (BG <2.8 mmol/L for 60 min or BG <2.0 mmol/L for 30 min), families were offered to use an unblinded CGM for the remaining 10 days. If there was no prolonged or severe hypoglycemia detected, participants continued using the blinded CGM. At the third visit (end of the 20-day study period), CGM data was uploaded and reviewed again on Clarity. Contour Next^® POCT data was also obtained from all participants. The participants and their primary endocrinologists were provided with a copy of the CGM data to guide dose adjustments if necessary.

Figure 1:

Study timeline.

Results

Forty patients with HI were identified. A total of 10 participants ages 3 months to 17 years were enrolled, and nine completed the 20-day study. The study flow chart is illustrated in Figure 2. One withdrew from the study after four days, as the caregiver believed the child’s increased fussiness was secondary to the sensor. The demographic data is reported in Table 1. The median age was 5.6 years (Q1: 2.8, Q3: 14.1), and half of the participants were male (n=5). The individual daily average glycemic profile of each participant is available in the Supplemental Material, Appendix 1. Analysis of these profiles showed that overall, participants remained in range more than 99 % of the time, aside from two participants who were in range 90 % and 95 % of the time.

Figure 2:

Flow chart diagram.

On average, POCT was performed 1.2 (SD 0.80) times a day. No participant had their CGM unblinded after 10 days, since there was no prolonged or severe hypoglycemia detected. During the study period, a total of 26 hypoglycemia episodes were recorded by CGM in six participants, for a median of two episodes per participant (Q1: 0, Q3: 3). Four of these participants (66.7 %) were under 6 years old. One participant had nine episodes (34.5 %, AE), and another had seven episodes (26.9 %, AG). The average duration of total hypoglycemia episodes was 26.5 min, with a median of 15 min (Q1: 10.0, Q3: 28.8.). The distribution of hypoglycemia is reported in Table 2. The average blood glucose value of all hypoglycemia events was 2.9 mmol/L (Q1: 2.8, Q2: 2.9), equal to the median. The lowest blood glucose measured via CGM was 2.3 mmol/L. Only one episode of hypoglycemia was detected by POCT (2.7 mmol/L). 50 paired values between CGM and POCT BG data were collected. The MARD was 11.63 % (SD: 7.9, 95 % CI: [6.2, 17.1]) and the MAD was 0.56 mmol/L (SD: 0.43, 95 % CI: [0.26, 0.86]).

CGM results demonstrated a common glycemic pattern of a mild decrease in BGs after midnight followed by an increase in the morning around 0800, with 45 % of the hypoglycemia episodes occurring in the early morning hours between 0600 and 0800. Some specific individual analysis of glycemic patterns in our participants are described below the Supplemental Material, Appendix 1.

1. A 3-year-old female (AE) had three hypoglycemic events all happening between 0600 and 0900 over the duration of the study, but was otherwise close to the lower threshold of 3.3 mmol/L overnight. After reviewing the CGM data and discussing with the family, the primary endocrinologist increased their diazoxide dose.

2. A 1-year-old male (AA) showed a stable glycemic pattern overnight with consistently higher BGs ranging between 9.0 and 11.0 mmol/L throughout the day. No blood glucose measurements were below 5.0 mmol/L. This participant withdrew from the study after four days and was lost to follow-up.

3. A 17-year-old male (AJ) showed a consistent pattern of one or two hyperglycemic events with a sharp increase above 10 mmol/L daily. At visit 2, it was noted that the morning (close to 9 am) peak in BG corresponded to the participant taking their medication along with a morning snack. The late-night peak (close to midnight) in BG corresponded to their late-night snacking habits. This participant was the only one on octreotide (0.94 mcg/kg/day), which was taken in the morning, after a discussion with their primary endocrinologist, the octreotide dose was reduced.

Discussion

The results of this pilot study revealed a total of 26 hypoglycemia episodes (less than 3.3 mmol/L for a minimum of 10 min) detected by CGM in a total of six participants (60 %), ranging from two to nine episodes per participant during the 20-day study. Four of these participants (66.7 %) were under 6 years old. Forty-five percent of the hypoglycemia episodes detected by CGM occurred in the early morning hours between 0600 and 0800. This finding agrees with Worth et al. [14], who studied the glycemic patterns in children with HI (n=23) and idiopathic ketotic hypoglycemia (n=24) using both Dexcom G4^® and G6^®. They identified the early morning period (0300–0700) as being the most conducive to hypoglycemic events, when the risk is usually 2–3 times higher than other times of the day [14].

In our study, younger children with HI seem to be more prone to hypoglycemia episodes. This is likely because their medications require frequent dose adjustments as they grow. POCT BG was not systematically measured at the time of the undetected hypoglycemia episodes noted on CGM; hence, it is difficult to conclude how many of these episodes were true hypoglycemic events or compression lows that can occur when a child is laying on their sensor site when asleep. However, the glycemic patterns detected by CGM did lead to medication dose adjustments in our study: one patient had their octreotide dose reduced by 40 %, one was weaned off diazoxide completely, and another had their diazoxide dose increased by 25 %.

Worth et al. [14] showed that in children with HI, hypoglycemia episodes detected by Dexcom G6^® were sustained for an average of 35 min. This agrees with our results, which showed a positively skewed distribution of time spent in hypoglycemia with an average of 26.5 min (SD 40.9) and a median of 15 min (Q1: 10.0, Q3: 28.8). This is because 33 % (2/6) of our participants had majority of the hypoglycemia episodes (16/26, 62 %).

CGM accuracy in HI is limited. The use of Dexcom G4^® and G5^® in patients with HI has shown MARD values ranging between 11.0 and 17.5 % [4, 10, 15, 16]. However, these analyses have also included normoglycemic values, overestimating the accuracy of CGM specifically in hypoglycemia. Worth et al. [11] showed a MARD of 19.3 % using Dexcom G6^® in patients with HI; however, when the true negative values defined as sensor BG and POCT BG >4.0 mmol/L were removed, the MARD increased to 23.2 %, demonstrating a further reduction in CGM accuracy. In our study, the MARD was 11.63 % and the MAD was 0.56 mmol/L. Though the calculated MARD is low compared to what has been reported in the literature, this was because the MARD was calculated almost exclusively with euglycemic values, as only one of all POCT BG measurements was below 3.3 mmol/L. We did not have enough POCT BG measurements during hypoglycemia detected via CGM, which could be attributed to the use of a blinded CGM.

With the use of Dexcom G6^®, there can still be false positives, often associated with compression lows overnight, which can result in increased parental anxiety. Further, Worth et al. [11] showed a high false negative rate of 7.2 % in patients with HI using Dexcom G6^®, though this number could be higher since the incentive to check a POCT BG during normoglycemia on CGM is low. Our study highlights that the goal of using CGM is not to replace POCT in patients with HI, but rather to have a clinical adjunct method to detect asymptomatic hypoglycemia, as shown in other studies [11, 15, 17], and to identify blood glucose trends to guide medication dose adjustments. This would be helpful, especially in younger children with transient hyperinsulinism or those who need frequent medication dose adjustments as they grow, as also concluded by Ranannavar et al. [4].

CGM is not a licensed device and has not been validated for patients with hyperinsulinism. Risks regarding the accuracy of CGM needs to be considered such as the suboptimal sensitivity to detect hypoglycemia, which provides a false reassurance to patients with HI. There is also an increased incidence of false positives with compression lows. However, this should not deter clinicians from using CGM to identify blood glucose trends. Despite the limited size of our study, the use of CGM helped guide dose adjustments in 30 % of our population. It is worth considering CGM trials periodically to detect glycemic patterns and frequency of hypoglycemia episodes in patients with HI to better guide medication dose adjustments.

Conclusions

The results of our pilot study identified undetected hypoglycemia episodes in patients with HI with the use of Dexcom G6^®, of which most of these episodes occurred in the early morning hours. It also helped us identify blood glucose trends that led to medication dose adjustments in our study. POCT BG monitoring along with the periodic use of an unblinded CGM could better guide the frequency of hypoglycemia episodes and aid in identifying blood glucose trends for medication dose adjustments, especially in the pediatric population with hyperinsulinism.