Continuous Glucose Monitoring in Patients With Postbariatric Hypoglycemia: Effect on Hypoglycemia and Quality of Life

Abstract

Context

Postbariatric hypoglycemia (PBH), complicating up to one-third of bariatric surgeries, is characterized by repeated episodes of severe hypoglycemia and hypoglycemia unawareness that threaten patient safety and impair quality of life.

Objective

We tested the hypothesis that use of a continuous glucose monitor (CGM) would reduce hypoglycemia and improve quality of life in patients with PBH.

Design

In a crossover design, 14 patients with diagnosed PBH were assigned in random order to sequential treatment with unblinded CGM or blinded CGM/no alarms for 10 days each. Glucose and quality of life measures were compared between the 2 periods.

Setting

Outpatient.

Outcomes

Hypoglycemia measured by fingerstick blood glucose in response to symptoms or CGM alarm and CGM glucose values; quality of life measures included dietary liberalization and hypoglycemia-related worries/behaviors captured by the Hypoglycemia Fear Survey-II.

Results

Baseline frequency of hypoglycemic events, disability, and hypoglycemia-related worries were high. Symptom-triggered hypoglycemic events confirmed by fingerstick glucose were reduced 6-fold (P = .008) and the glucose nadir measured by CGM was >8 mg/dL higher (P = .005) during unblinded use of CGM compared to the blinded comparison period. Hypoglycemia Fear Survey-II scores improved significantly in response to unblinded CGM use compared to the blinded control period (P = .026). The intake of carbohydrate-containing meals increased without increasing rate of postprandial hyper- or hypoglycemia.

Conclusion

Use of unblinded CGM in patients with PBH reduces frequency and severity of hypoglycemia and improves quality of life by decreasing hypoglycemia-related worries and enabling a less restrictive diet. CGM should be considered a first-line treatment for patients with PBH.

Postbariatric hypoglycemia (PBH) is a rare disease entity that continues to grow in prevalence due to the increasing use of bariatric surgery for durable resolution of obesity and diabetes. Occurring in a subset of patients months to years after surgery, PBH is characterized by episodes of postprandial hyperinsulinemic hypoglycemia occurring 1 to 3 hours after meals. In patients with gastric bypass or vertical sleeve gastrectomy, rapid transit of oral nutrients, and carbohydrates in particular, stimulate high levels of GLP-1 from enteroendocrine cells in the distal small and large intestine, which in turn leads to overstimulation of pancreatic insulin secretion, followed by hypoglycemia [1]. Severe hypoglycemia occurs several times per week to multiple times per day, causing seizures, loss of consciousness, falls, motor vehicle accidents, coma, and potentially death [2]. In addition, more than 60% of patients are unable to work, drive, or care for themselves and their children [2]. The majority avoid leaving their homes due to considerable anxiety and fear of hypoglycemia, leading to major disruptions in quality of life for them and for their support networks [2]. Many rely on another family member or hired caregiver to be present at all times to assist with hypoglycemia episodes. There are no approved treatments. Because the condition does not remit over time, the toll on quality of life can be substantial.

Continuous glucose monitoring (CGM) has been shown to significantly decrease the frequency of hypoglycemia in both type 1 and type 2 diabetes [3, 4]. PBH patients have particularly severe hypoglycemia that, compared to patients with insulin-treated diabetes, is more frequent, comes on more rapidly, and is characterized by a higher prevalence of hypoglycemia unawareness [2, 5]. CGM is currently only approved for diabetes management: its use is not approved for hypoglycemia alone. Given the strong data that use of CGM can prevent hypoglycemic events in patients with diabetes [6], and the frequency and severity of hypoglycemia in patients with PBH [2], it is critically important to determine whether the use of CGM can reduce hypoglycemia and related disturbances in quality of life in patients with PBH.

Only 1 prior study has evaluated the potential of CGM to prevent hypoglycemia in patients with PBH [7]. Results showed reduction in hypoglycemia, hyperglycemia, and glycemic variability, but 2 different CGMs were used, wear time differed, and all patients underwent blinded prior to unblinded treatment. Furthermore, patients were on varied medications used off label and they were not required to adopt dietary therapy prior to participation. Last, this study did not routinely capture lifestyle or psychological measures such as hypoglycemia-related worries/behavior changes. We concurrently conducted a rigorous study that entailed the use of the same CGM device in all patients for both arms, random order of assignment to blinded vs unblinded treatment arms, prohibition of medications that might affect glucose levels, and adoption of medical nutrition therapy throughout. We systematically assessed quality of life measures including dietary liberalization using quantitative data from CGM and hypoglycemia-related fears using a standardized questionnaire after each treatment arm. Endpoints included both hypoglycemia and quality of life measures.

Methods

Study Design

The study was approved by the Stanford Internal Review Board and all participants provided written informed consent. Patients with confirmed PBH were screened for eligibility and if enrolled, were assigned to 2 treatment periods given sequentially but in random order to avoid confounding by treatment order: (1) a 10-day period with use of unblinded Dexcom G6 (the intervention of interest) and (2) a 10-day control period without use of unblinded Dexcom G6 (Fig. 1). To ensure consistency of data collection in both treatment phases, glucose metrics for analysis were obtained from a blinded Dexcom G6 Pro that was used for both treatment periods. Thus, during the unblinded period, participants wore both the unblinded Dexcom G6, which provided real-time glucose trends and alarmed for glucose <65 mg/dL, and the blinded Dexcom G6 Pro for data collection. During both periods, the participants were instructed to check fingerstick blood glucose (FSBG) if they experienced symptoms of hypoglycemia and to record the value in an event log at the time of symptoms. In addition, during the unblinded period, participants were instructed to also check FSBG and log an event if the CGM read <65 mg/dL or the alarm sounded (there was no alarm during the blinded period). Hypoglycemia fear (Hypoglycemia Fear Survey-II [HFS-II]) [8] survey were administered after each of the 2 periods. All patients were on medical nutrition therapy (low carbohydrate diet with recommended carbohydrate grams/meal not to exceed 30 and strict avoidance of sugar and processed foods) and were advised not to change diet throughout the entire study. Because use of unblinded CGM may have led to subconscious dietary liberalization, number of glucose excursions >40 mg/dL above baseline on CGM tracings was quantified as an estimate of carbohydrate-containing meals/snacks (dietary liberalization), and statistical adjustment applied for residual differences between periods. Hypoglycemic events were quantified and categorized as <70, <54, and <40 mg/dL or need for third-party assistance, in line with current definitions of levels 1, 2, and 3 events [9, 10]. These events included those triggered by symptoms or CGM alarm and captured by FSBG, as well as those identified by CGM alone. Blinded CGM (Dexcom Pro) data were also compared during the 2 periods with respect to time spent in hypoglycemia <70, <54, and <40 mg/dL.

Figure 1.

Study design overview. Participants were first screened, randomized, and trained in CGM use. Subjects were then assigned to begin with either a 10-day unblinded CGM (Dexcom G6) or 10-day blinded CGM (Dexcom G6 Pro) period. Each monitoring phase was followed by completion of the Hypoglycemia Fear Survey-II (HFS-II), an event diary, and training (if applicable) with a new CGM device. All participants crossed over to the alternate CGM modality for another 10-day period. The study concluded with repeated HFS-II assessments, event diary entries, and CGM data download. Blinded CGM data (Dexcom G6 Pro) was used for all outcome analyses to ensure consistency.

Participants

Participants were recruited from referring physicians in the San Francisco Bay area and from our research registry of more than 80 patients with PBH who have consented to be contacted for research studies. Patients were screened for eligibility by video and informed written consent was obtained by screen sharing the consent form on RedCap, which the participant signed digitally. This method of screening and consenting was approved by our institutional review board to maximize patient safety during COVID, when this study took place. Eligible participants were required to be between 18 and 70 years of age and in stable health, to have had Roux-en-Y bariatric surgery at least 12 months prior to enrollment, stable body weight (<5-lb weight change over the prior 2 months), and confirmed PBH defined as: history of plasma glucose <54 mg/dL, presence of Whipple triad, and exclusion of other causes by treating physician prior to enrollment. Patients with a history of fasting hypoglycemia <54 mg/dL were included only if they had undergone evaluation for insulinoma with a 72-hour fast. Those with and without fasting hypoglycemia were required to have had evaluation for other causes, including morning cortisol, liver and kidney function tests, sulfonylurea screen, anti-insulin antibody, and IGF2 if indicated based on medical history. All were screened for use of alcohol and medications associated with hypoglycemia. Participants were not eligible if they were on medications known to alter glucose metabolism (eg, acarbose, metformin, diazoxide, octreotide, calcium-channel blockers, GLP1 agonists, DPP4 inhibitors, SGLT2 inhibitors, sulfonylureas, and insulin), but were allowed to participate if they underwent a washout period of at least 5 half-lives of the medication. Patients on glucocorticoids or stimulants used for weight loss or attention deficit disorder and those with illicit substance abuse or any alcohol (>1 drinks per day) were excluded. Stable dose of antidepressant or thyroid medication was allowed. All had received counseling by a registered dietitian on proper medical nutrition therapy. Other exclusions included acute COVID-19 infection within the prior 3 months or current symptoms of long-COVID, acute illness of any kind, major organ disease, pregnancy/lactation, lack of comfort in using the CGM or prior skin reaction to the Dexcom CGM, inability to participate in training, and unstable psychiatric condition.

Training on CGM Use and Symptom-triggered Event Logging

This study was conducted largely remotely during the COVID-19 epidemic of 2021 through 2022. Enrolled participants were scheduled for a live video training meeting with the study coordinator after being sent the CGM devices and glucometer (Bayer Contour) to their residence. A blinded Dexcom G6 Pro was worn by all participants for data collection, and during the unblinded period, a Dexcom G6 (identical to the commercially available model) was also worn so that the participant could respond to real-time glucose values and alarm. During the first training session, the coordinator explained the study design, goal, and surveys, and reviewed symptoms and signs of hypoglycemia and how to treat it. Although no specific dietary intervention was included as part of the study, all patients were counseled on medical nutrition therapy for PBH, which included limiting carbohydrate intake to a maximum of 30 g per meal and 20 g per snack, avoiding simple sugar and processed foods, and excluding alcohol. Participants put on their CGM device(s) under the coordinator's supervision. Due to loose skin (typical of postbypass surgery) in the abdomen, all CGMs were placed on the back of the upper arm between the triceps and deltoid muscle in the fatty part of the skin. Participants assigned to the unblinded period were instructed on how to view their Dexcom G6 glucose values in real time using either the Dexcom app or a Dexcom receiver. The unblinded CGMs were set to alarm the participant of a low glucose reading at <65 mg/dL. This would alert the participant to check their FSBG with a glucometer and fill out an event log in RedCap to enter the CGM reading, glucometer reading, time, date, prior food intake, exercise, and if any assistance was necessary. Participants starting both treatment periods were instructed to check FSBG and fill out the event log when experiencing hypoglycemic symptoms. Participants in both periods were allowed to self-rescue for hypoglycemic symptoms or if the CGM or FSBG indicated glucose <54 mg/dL or a rapid drop that indicated they would drop below this threshold without treatment (unblinded group only). After completion of the first 10-day period, participants had another video meeting with the coordinator to remove the CGM transmitter(s), which was mailed back in a prepaid USPS envelope or dropped off in person to the study team. The participants then put on a new CGM device, under video supervision, to start the next period of the study, without washout in between. Symptoms of hypoglycemia were again reviewed, and participants reminded not to alter their diet. For those starting the unblinded period, instructions on how to view glucose and event logging instructions were provided as previously. After the second period, participants had a final video visit to ensure their CGM device was removed correctly and mailed back to the study team.

Symptom Questionnaires for Eligibility and Cohort Description

At screening, participants were asked to complete multiple surveys through RedCap to collect clinical and demographic information and to characterize severity of disease for purposes of eligibility screening and cohort description. These included a survey for: (1) demographics and concomitant medications and (2) characterization and quantification of hypoglycemia including presence of Whipple triad, frequency of hypoglycemic events, categorized as <70, <54, and <40 mg/dL or need for third-party assistance, in line with current definitions of level 1, 2, and 3 events [9], type of hypoglycemia symptoms—autonomic (sweating, shaking, palpitations, hunger), neuroglycopenic (blurred vision, confusion, drowsiness, odd behavior, speech difficulty, incoordination, dizziness, inability to concentrate or complete mental tasks), or malaise (nausea, headache).

To assess the impact of unblinded CGM use on quality of life, the hypoglycemia fear survey (HSF-II) [7, 8] was administered (online) at the end of each study period to determine the impact of hypoglycemia on quality of life, including frequency of rescue behaviors (eating snacks, limit driving, measure blood glucose more than 6 times per day, etc.) and worries about their hypoglycemic events. Answers to all were recorded on a scale from 0 (never) to 4 (very often) for individual questions and summed as a composite score for purposes of quantification.

CGM Data Upload and Cleaning

Once the CGM transmitters were received by the study team, the data were uploaded to Dexcom Clarity through a Dexcom receiver. The CGM Dexcom Pro transmitters were then returned to the study team and the raw data were exported to Excel for analysis. We performed a thorough data-cleaning process to ensure accuracy and consistency, which included checking for missing or erroneous values, validating timestamps, and confirming that all data points were within the expected range for glucose readings. For data analysis, we used all the data captured by the CGM without excluding any data, as this decision was made to maximize the data available and avoid potential bias. To account for variations in the duration of data capture, we adjusted the analysis based on the total minutes of data captured, ensuring standardization and comparability across different participants and recording periods. Hypoglycemic events included those captured by FSBG (see the following section) and by CGM. CGM hypoglycemic events were defined as periods in which CGM glucose levels dropped below a specified threshold (70, 54, or 40 mg/dL) for at least 15 consecutive minutes. We manually identified and calculated these events by reviewing filtered CGM data, excluding any periods that lasted fewer than 15 minutes. The glucose nadir (lowest) CGM glucose level during each event was recorded. CGM hypoglycemic events did not require the presence of symptoms or FSBG verification, since the blinded group did not have alarms to trigger FSBG and it is known that hypoglycemia unawareness is common. The extracted events were cross-referenced with the patient-reported event diary, using date and time, and classified based on whether participants experienced hypoglycemic symptoms, received an alarm notification from the CGM, or had neither. Separately from event identification, we used CGM data to calculate the total time spent in hypoglycemia below three glucose thresholds (70, 54, and 40 mg/dL). To do this, we exported raw CGM data from Dexcom Clarity into an Excel workbook, where we applied Excel's filtering functions to identify readings that fell below or exceeded each threshold. Similarly, we evaluated hyperglycemia using 140 and 180 mg/dL. We also analyzed diurnal and nocturnal hypoglycemic events. Diurnal hypoglycemia was defined as the period starting with the first morning glucose spike and ending 3 hours after the last evening glucose spike. Nocturnal hypoglycemia was defined as the period starting 3 hours after the beginning of the last glucose spike (from dinner or a bedtime snack) and ending with the first morning glucose spike. This approach enabled us to accurately and systematically quantify both hypoglycemic and hyperglycemic events, as well as the total duration of time spent within specific glucose thresholds, ensuring a robust, unbiased, and systematic analysis of the CGM data.

Quantification of Dietary Liberalization

Because patients wearing the unblinded CGM may have subconsciously liberalized their diet due to feeling safer, we estimated dietary liberalization by quantifying, for each period, the number of glucose spikes, which reflect increased oral nutrient (containing carbohydrates) consumption. A glucose spike was defined as an increase in blood glucose CGM of ≥40 mg/dL above baseline within the course of an hour. A glucose spike was resolved once the glucose returned to baseline and remained there for a minimum of 30 minutes. A second elevation after 30 minutes was recorded as a separate event and an elevation within 30 minutes was considered part of the initial event. Glucose spikes following a qualifying CGM- or FSBG-confirmed hypoglycemic events (recorded in patient event log) were not counted.

Hypoglycemic Events (FSBG and CGM Events)

Participants were asked to report symptom-triggered events defined as symptoms of hypoglycemia or an unblinded CGM low value or alarm (set to blood glucose <65 mg/dL) and confirm hypoglycemia with FSBG and log the symptoms and FSBG value in RedCap. For each event logged, participants were asked: (1) Did you experience any symptoms/signs of hypoglycemia (low blood sugar)? Yes/No; (2) Did your alarm alert you of a low blood sugar or risk of a low blood sugar (eg, urgent low soon)? Yes/No. Events were classified as follows:

• Symptom-only events were those in which participants reported feeling symptoms but did not hear an alarm (answered yes to question 1 but no to question 2). Participants were asked to log their glucometer reading (FSBG), describe their symptoms, and log their CGM reading if in the unblinded phase.

• Alarm events were defined as instances in which the Dexcom G6 CGM triggered an alarm due to a glucose reading below 65 mg/dL, regardless of whether the participant experienced any symptoms. Specifically, these events occurred when participants answered “yes” to the alarm sounding in the RedCap survey. During these events, participants were instructed to log their glucometer reading (FSBG), the CGM reading, and confirm whether or not they experienced any symptoms associated with the alarm event.

• “CGM-only” hypoglycemic events are those in which participants did not report any symptoms and did not report the alarm alerting them (answered no to questions 1 and 2 on the RedCap Event Diary) but the Dexcom G6 Pro collected glucoses below the threshold for at least 15 minutes. Nothing was required to be logged in RedCap and FSBG was not required.

• Total hypoglycemic events are cumulative (without duplicating) CGM only events + symptom only events + alarm events. This measure includes all events that the Dexcom G6 Pro collected regardless of symptoms or an alarm.

Time Spent in Hypoglycemia by CGM Capture of Minutes Spent Below Glycemic Thresholds

As described previously, CGM data captured the minutes during Dexcom wear that were below specified thresholds of <70, 54, and 40 mg/dL, adjusted for minutes of data capture during the 10-day wear period to adjust for any lapses in due to Bluetooth connection or other interruption in data capture. Time below each glucose threshold is presented as absolute time (adjusted for minutes of data capture) as well as time adjusted for dietary liberalization defined by number of glucose spikes >40 mg above baseline glucose as described previously and presented as time below threshold per glucose spike.

Data Analysis

All data were normally distributed. Student paired t-tests were used to compare glycemic and survey endpoints between the 2 arms of treatment. P < .05 was considered statistically significant. Adjustments for minutes of CGM capture and if applicable, dietary liberalization, were applied to the data values prior to statistical analysis.

Results

Fifteen patients enrolled. One CGM was lost in the mail and the patient was excluded from analyses. The 14 patients analyzed included 13 patients who had undergone Roux-en-Y gastric bypass and 1 patient with Nissen fundoplication. The majority were female. As shown in Table 1, 43% had been hospitalized for hypoglycemia and 57% were disabled. Among the 14 participants, the majority (42.9%) experienced low blood glucose (<54 mg/dL) on a daily basis and 14.3% experienced it more than once daily. Weekly occurrences were reported by 35.7% of participants, and monthly occurrences were reported by 7.1%. Common autonomic symptoms reported as occurring at least once per week included sweating (71.4%), palpitations (64.3%), and shaking and hunger (both 57.1%). Neuroglycopenic symptoms were also reported, with confusion, inability to concentrate/haziness, and behavioral changes as the highest, with each reported weekly in 35.7% of participants. Drowsiness and motor incoordination each affected 28.5% and speech difficulty 21.4%. Importantly, more than 71% reported that due to neuroglycopenic symptoms, they would have benefitted from third-party assistance at least once per week.

Table 1.

Summary of participant demographics and clinical characteristics (n = 14)

Variable	Sample (mean ± SD)
Height (cm)	166.3 ± 5.36
Weight (kg)	76.1 ± 13.5
BMI (kg/m2)	27.8 ± 3.97
Age (y)	47.5 ± 13.1
Sex (female/male)	14/0
Years after bariatric surgery	8 ± 5.5
Race (Caucasian/Hispanic/Asian)	11/2/1
Surgery type (RYBG/Nissen)	13/1
Previous hospitalization due to hypoglycemia	6/14 (42.8%)
Unable to work due to hypoglycemia	8/14 (57.1%)
Frequency of glucose <54 mg/dLa
>1× daily	2/14 (14.3%)
Daily	6/14 (42.9%)
Weekly	5/14 (35.7%)
Monthly	1/14 (7.1%)
Frequency of autonomic symptomsa reported weekly
Sweating	10/14 (71.4%)
Shaking	8/14 (57.1%)
Palpitations	9/14 (64.3%)
Hunger	8/14 (57.1%)
Frequency of neuroglycopenic symptomsa reported weekly
Confusion	5/14 (35.7%)
Drowsiness	4/14 (28.5%)
Speech difficulty	3/14 (21.4%)
Inability to concentrate/haziness	5/14 (35.7%)
Behavioral changes	5/14 (35.7%)
Incoordination	4/14 (28.5%)
Would have benefitted from third-party assistance	10/14 (71.4%)

Abbreviations: BMI, body mass index; RYBG, Roux-en-Y gastric bypass.

^aEpisodes per week by reported prior to study entry.

Comparison of hypoglycemic events in the 2 treatment periods (Table 2) revealed that the number of symptom-only (no alarm) events confirmed by FSBG <70 mg/dL was 6-fold lower during the unblinded vs the blinded period (P < .008). Symptom-triggered events with glucose <54 and <40 mg/dL events were infrequent and not significantly different between treatment periods. Alarmed events were not compared as the blinded group did not have alarms. CGM-defined hypoglycemic events <70, 54, or 40 mg/dL did not differ significantly between treatment periods, but the glucose nadir was significantly higher in the unblinded phase compared to the blinded phase, reaching a difference of >8 mg/dL (P < .005). Time in minutes spent below each hypoglycemic threshold (<70, 54, and 40 mg/dL) as calculated from CGM tracings did not differ significantly between the blinded and unblinded periods (Table 3), but when separated into diurnal and nocturnal time frames, a consistent trend was seen in the diurnal (but not the nocturnal) time frame toward less time spent in hypoglycemia in the unblinded vs the blinded treatment period, although the difference did not reach statistical significance (Table 4).

Table 2.

Hypoglycemic events and nadir (CGM and FSBG) during an event (n = 14)

Category of eventa	Glucose thresholdb	Blinded (mean ± SD)	Unblinded (mean ± SD)	P value
Symptom only (no alarm triggered)	<70 mg/dLc	2.5 ± 2.87	0.43 ± 0.76	.008
	<54 mg/dL	0.14 ± 0.36	0.14 ± 0.36	1.000
	<40 mg/dL	0 ± 0	0 ± 0	N/A
	CGM nadir (mg/dL)	54.28 ± 9.4	55.8 ± 10.2	.946
	FSBG (mg/dL)	60 ± 2.92	61.2 ± 6.47	.837
Alarm (with or without symptoms)	<70 mg/dL	No alarms in the blinded phase	1.71 ± 1.14	N/A
	<54 mg/dL		0.29 ± 0.61	N/A
	<40 mg/dL		0.21 ± 0.43	N/A
	CGM nadir (mg/dL)		57.4 ± 6.0	N/A
	FSBG (mg/dL)		57.69 ± 6.9	N/A
CGM only (no alarm, no symptoms)	<70 mg/dL	6.07 ± 7.3	6.93 ± 9.86	.470
	<54 mg/dL	1.43 ± 2.28	0.86 ± 1.17	.336
	<40 mg/dL	0.36 ± 0.93	0.07 ± 0.27	.165
	CGM nadir (mg/dL)	52.6 ± 5.6	60.9 ± 5.6	.005
	FSBG (mg/dL)	N/A	N/A	N/A
Total (all events)	<70 mg/dLd	8.57 ± 6.6	8.57 ± 9.9	1.000
	<54 mg/dL	1.57 ± 2.3	1.29 ± 1.3	.513
	<40 mg/dL	0.36 ± 0.93	0.29 ± 0.47	.775
	CGM nadir (mg/dL)	53.5 ± 5.4	57.84 ± 4.5	.037
	FSBG (mg/dL)	60 ± 2.92	57.9 ± 7.1	.915

Abbreviations: CGM, continuous glucose monitoring; FSBG, fingerstick blood glocose; N/A, not available.

^aA single event is defined by at least 15 consecutive minutes spent below corresponding threshold with recovery time above 70 mg/dL for at least 15 minutes signifying the end of the event.

^b<70 = lowest point in the event is above 54 mg/dL but under 70 mg/dL; <54 = lowest point in the event is below 54 mg/dL but above 40 mg/dL; <40 = lowest point in the event is below 40 mg/dL.

^c<70 threshold includes events <54 and events <40 (all inclusive).

^dThis includes symptom triggered events, alarm triggered events, and CGM only events without duplicates.

Table 3.

Time spent in hypoglycemia as captured by continuous glucose monitoring with adjustment for dietary liberalization (mean ± SD)

Glucose threshold	Proportion of time spent below glucose threshold (minutes/minutes monitored)			Adjusted (for dietary liberalization) proportion of time spent below glucose threshold (minutes/minutes monitored/spikea)
	Blinded	Unblinded	P value	Blinded	Unblinded	P value
<70 mg/dL	0.022 ± 0.026	0.029 ± 0.043	.19	0.0005 ± 0.0006	0.0005 ± 0.0007	.27
<54 mg/dL	0.006 ± 0.012	0.009 ± 0.015	.11	0.0002 ± 0.0003	0.0002 ± 0.0003	.39
<40 mg/dL	0.003 ± 0.006	0.002 ± 0.004	.81	0.00007 ± 0.0002	0.00007 ± 0.0002	.47

^aA spike is defined by an increase in blood glucose of 40 mg/dL or more over the course of an hour which gives evidence to dietary liberalization.

Table 4.

Time spent in hypoglycemia as captured by continuous glucose monitoring with adjustment for dietary liberalization separated by diurnal and nocturnal (mean ± SD)

	Proportion of time spent below glucose threshold (minutes/minutes monitored)			Adjusted (for dietary liberalization) proportion of time spent below glucose threshold (minutes/minutes monitored/spikec)
Diurnal onlya	Blinded	Unblinded	P value	Blinded	Unblinded	P value
<70 mg/dL	0.025 ± 0.031	0.018 ± 0.015	.44	0.0005 ± 0.0007	0.0003 ± 0.0003	.27
<54 mg/dL	0.010 ± 0.017	0.006 ± 0.011	.53	0.0002 ± 0.0003	0.0001 ± 0.0002	.39
<40 mg/dL	0.002 ± 0.009	0.001 ± 0.004	.52	0.00007 ± 0.0002	0.00002 ± 0.00007	.47

Nocturnal onlyb	Blinded	Unblinded	P value	Blinded	Unblinded	P value
<70 mg/dL	0.016 ± 0.025	0.034 ± 0.090	.36	0.0004 ± 0.0005	0.0006 ± 0.0017	.41
<54 mg/dL	0.001 ± 0.004	0.011 ± 0.034	.32	0.00002 ± 0.00008	0.0002 ± 0.0006	.33
<40 mg/dL	0.001 ± 0.004	0.003 ± 0.007	.47	0.00002 ± 0.00008	0.00005 ± 0.0001	.54

^aDiurnal was defined as the beginning of the first morning glucose spike and ending 3 hours after the last evening glucose spike.

^bNocturnal was defined as glucose capture starting 3 hours after the beginning of the last glucose spike (dinner or bedtime snack) and ending with the beginning of the first morning glucose spike.

^cA glucose spike is defined by elevation of at least 40 mg/dL above baseline over the course of an hour.

In addition to less frequent hypoglycemia, time spent in hyperglycemia >180 mg/dL was significantly decreased in the unblinded vs the blinded period, with a trend toward a decrease in time spent >140 mg/dL as well (Table 5). Further comparison of CGM patterns during the 2 treatment periods revealed a significant increase in dietary liberalization, defined as oral intake that led to >40 mg /dL glucose rise above baseline, during the unblinded vs the blinded period, as shown in Fig. 2. Notably, 11 of 14 participants exhibited a higher number of glucose spikes in the unblinded period, providing further evidence of dietary liberalization when participants had real-time access to their glucose data.

Table 5.

Time spent in hyperglycemia as captured by continuous glucose monitoring with adjustment for dietary liberalization (mean ± SD)

	Proportion of time spent above glucose threshold (minutes/minutes monitored)			Adjusted (for dietary liberalization) proportion of time spent above glucose threshold (minutes/minutes monitored/spikea)
	Blinded	Unblinded	P value	Blinded	Unblinded	P value
>180 mg/dL	0.044 ± 0.035	0.034 ± 0.028	.13	0.001 ± 0.0008	0.0007 ± 0.0006	.02
>140 mg/dL	0.158 ± 0.110	0.116 ± 0.048	.09	0.004 ± 0.003	0.002 ± 0.001	.08

^aAdjusted for number of gluose spikes indicating carbohydrate-containing meals consumed: a glucose spike is defined as glucose elevation ≥40 mg/dL above baseline within the course of 1 hour.

Figure 2.

Dietary liberalization defined by frequency of glucose spikes (n = 14). The total number of glucose spikes was significantly higher during the unblinded continuous glucose monitoring (CGM) period compared to the blinded period (P = .005), suggesting potential dietary liberalization when patients were aware of their glucose readings in real time. Each dot represents an individual participant. A glucose spike was defined as an increase in CGM glucose of ≥40 mg/dL above baseline within 1 hour, which resolved when glucose returned to baseline and remained there for at least 30 minutes. Bars represent mean ± SD.

It is important to note that CGM alone captured 3- to 4-fold more events (defined as below 70 or 54 mg/dL for more than 15 minutes) than did the alarm and symptom-triggered FSBG combined. And the alarm detected 2- to 4-fold more events than did symptoms alone. In this study, the unblinded group had 0.43, 1.71, and 6.73 events of glucose <70 mg/dL per 10-day period that were detected, respectively, by symptoms only, alarm ± symptoms, and CGM alone. For glucose events <54 mg/dL, the frequency of detection by each method was 0.14, 0.29, and 0.86 events/10-day period, respectively. The blinded group showed a similar pattern, with 3-fold more detection of glucose events <70, 10-fold more detection of glucose events <54, and 30-fold more detection of glucose events <40 mg/dL on CGM as compared with symptoms alone.

Psychological distress related to hypoglycemia was captured via the HFS-II survey, which was completed by 100% of participants. Scores on the HFS-II survey were high during the blinded period with a mean composite Worry score of 52.3 and a mean Behavior score of 37.3. The composite score and individual question scores of the HFS-II survey indicated significantly lower levels of worry during the unblinded period (Table 6). Overall, the total Worry score was significantly lower in the unblinded group (mean score 43.8 vs 52.3, P = .026), indicating that participants in the unblinded group experienced less frequent worry/fears across various hypoglycemia-related scenarios compared to the blinded group. Improvement was present in 15/18 measures, reaching statistical significance in 6 and near-significance in 2 additional measures. Worries spanned multiple key areas, including being unaware of low blood glucose (P = .029), passing out in public (P = .026), having a hypoglycemic episode when alone (P = .044), getting a bad evaluation or being criticized (P = .044), difficulty thinking clearly when responsible for others (P = .016), and low blood glucose interfering with important tasks (P = .027). In contrast, worries about not having food available, embarrassing oneself in social situations, and appearing drunk or stupid did not significantly differ between the 2 groups. The scores for individual measures on the behavior portion of HFS-II did not differ significantly between the blinded vs unblinded periods (Table 7).

Table 6.

Hypoglycemia fear survey, worries (Hypoglycemia Fear Survey-II) (mean ± SD)

How often do you worry about:	Blinded	Unblinded	P value
Being unaware of low blood glucose	3.6 ± 0.737	3.1 ± 0.834	.029
Not having food or snacks available	2.6 ± 1.175	2.3 ± 1.234	.416
Passing out in public	2.5 ± 1.642	1.5 ± 1.187	.026
Embarrassing yourself in a social situation	2.7 ± 1.291	2.3 ± 0.884	.304
Having a hypoglycemic episode when you're alone	2.9 ± 1.302	2.3 ± 0.883	.044
Appearing drunk or stupid	2.6 ± 1.454	2.1 ± 1.187	.15
Losing control	2.3 ± 1.234	2 ± 0.925	.237
Not having anyone to help you during a hypoglycemic episode	2.6 ± 1.447	2.6 ± 1.183	.751
Having a hypoglycemic episode while driving	2.1 ± 1.407	2.1 ± 1.373	.719
Making mistakes or having accidents	2.6 ± 1.175	2.1 ± 0.961	.107
Getting a bad evaluation or being criticized	2.7 ± 1.163	2.1 ± 1.187	.044
Difficulty thinking clearly when responsible for others	2.6 ± 1.175	2 ± 1.099	.016
Feeling lightheaded or dizzy	3.1 ± 0.961	2.5 ± 1.014	.334
Accidentally injuring yourself or others	2.5 ± 1.187	1.9 ± 0.925	.149
Permanent injury or damage to your health or body	2.9 ± 1.486	2.1 ± 1.552	.061
Low blood glucose interfering with important tasks	3.4 ± 0.828	3.1 ± 0.990	.027
Becoming hypoglycemic while sleeping	3.1 ± 0.990	3.3 ± 0.774	.67
Getting emotionally upset and difficult to deal with	3 ± 1.069	2.4 ± 1.060	.168
Total	52.3 ± 16.3	43.8 ± 12.4	.026

Answers are on a scale of 0-4; 0 = never, 1 = rarely, 2 = sometimes, 3 = often, 4 = very often.

Table 7.

Hypoglycemia fear survey, behaviors (Hypoglycemia Fear Survey-II) (mean ± SD)

What you do during your daily routine to avoid low blood sugar	Blinded	Unblinded	P value
Eat snacks	3.1 ± 0.915	3.1 ± 0.884	.581
Try to keep my blood glucose above 70 mg/dL	3 ± 1.309	3.3 ± 0.816	.313
Measure my blood glucose 6 or more times per day	2.6 ± 1.549	2.4 ± 1.454	.334
Take someone with me when I go out	2.4 ± 1.502	2.6 ± 1.352	.118
Limit my out-of-town travel	2.3 ± 1.579	2.3 ± 1.533	1
Limit driving	2.3 ± 1.534	2.3 ± 1.543	.334
Avoid visiting friends	2.1 ± 1.598	1.8 ± 1.373	.207
Stay home more than I would like	2.3 ± 1.718	2.1 ± 1.457	.384
Limit my exercise/physical activity	2.7 ± 1.223	2.9 ± 1.125	.581
Make sure others are around	2.3 ± 1.335	2.5 ± 1.246	.334
Avoid sex	1.9 ± 1.438	2.4 ± 1.502	.289
Keep my blood glucose higher than usual in social situations	2.5 ± 1.407	2.8 ± 1.320	.216
Keep my blood glucose higher than usual when doing important tasks	2.6 ± 1.404	2.9 ± 1.187	.164
Have other people check on me several times during the day or night	1.5 ± 1.356	2.1 ± 1.486	.104
Carry glucose tabs or other high-sugar snacks	3.6 ± 0.737	3.5 ± 1.125	.774
Total	37.3 ± 13.6	38.3 ± 13.9	.216

Answers are on a scale of 0-4; 0 = never, 1 = rarely, 2 = sometimes, 3 = often, 4 = very often.

Discussion

The results of this study, in which (unblinded) use of CGM with real time glucose monitoring and alarms set to <65 mg/dL was compared to a control condition in which blinded CGM was used only for data collection, demonstrated that in patients with PBH, use of an unblinded CGM significantly reduced the frequency of symptomatic hypoglycemia and improved quality of life. The 4-fold increase in alarm-detected events relative to symptom-detected events during the unblinded period suggests that events were detected earlier, before they got to the point of symptoms. This is also supported by the significantly higher glucose nadir during unblinded use vs blinded use of the CGM in which the alarm was disabled. A 16% increase in glucose nadir (8.3 mg/dL) is clinically important, particularly as the mean glucose nadir during the blinded period was 52.6 mg/dL, which is in the range known to cause cognitive dysfunction. Although not statistically significant, total minutes spent low on CGM during daytime hours, when patients are eating and therefore experience the highest frequency of postprandial hypoglycemic episodes, showed a trend toward less time spent below all 3 glycemic thresholds (<70, <54, and <40 mg/dL) during the unblinded vs blinded CGM use. Level 3 events (need for third-party assistance) are also important to quantify: in the current study, we did not evaluate this endpoint, but CGM-detected events <40 mg/dL, which would normally be associated with a level 3 presentation, were 5-fold reduced and reduction in time (minutes on CGM) spent <40 mg/dL showed a nonsignificant trend during the diurnal monitoring. Features specific to CGM technology that likely contributed to prevention of hypoglycemia include real-time monitoring in which a glucose spike and crash may be visualized as they are occurring, allowing the patient to eat preventively or “self-rescue” as the glucose is falling, as well as the alarm feature, neither of which require symptoms to be present in order to identify a current or impending hypoglycemic event. Furthermore, a significant reduction in time spent high, defined as >180 mg/dL, during unblinded vs blinded CGM use indicates that use of CGM may have informed dietary choices for less glucose-elevating foods, which in turn would have the effect of reducing hypoglycemia since higher glucose peaks and higher dietary sugar/processed carbohydrates provoke hyperinsulinemia and consequent postprandial hypoglycemia [1].

The results of this study also demonstrate that while the alarm was better than symptoms at detecting hypoglycemic events, CGM values alone were 17-fold more sensitive than symptom-triggered FSBG and 4-fold more sensitive than alarm-triggered FSBG in detecting hypoglycemic events with glucose <70 and <54 mg/dL, regardless of whether the patient was blinded or unblinded during CGM use. This indicates that even use of alarms may fail to detect the glucose nadir, likely due to the real-time (every 5 minutes) sampling that can be done with CGM compared to less frequent snapshots with FSBG, which even when done in response to an alarm may miss clinically important (defined as glucose <54 mg/dL) hypoglycemic events. This is important for research studies/clinical trials that are attempting to capture clinically important events to quantify the prevalence of hypoglycemia and/or response to investigational therapies. Of note, while sensitivity is greater, the accuracy of CGMS is lower than FSBG, particularly in the hypoglycemic range, as measured by the mean absolute relative difference. For Dexcom G6 the mean absolute relative difference is 18% in the hypoglycemic range vs 9% in the normoglycemic range [11]. Thus, in the clinical setting, if a patient is asymptomatic it may be prudent to confirm a low CGM reading with a FSBG, although as stated previously, the FSBG may actually miss the glucose nadir. Potential considerations in clinical trial design include use of alarm- or symptom-triggered FSBG rather than relying on symptoms alone to detect events, confirming CGM low values with FSBG, repeating FSBG to try to capture the nadir, and including CGM alone as a primary or secondary endpoint.

A second important finding of this study is improvement in quality-of-life measures in association with unblinded CGM use. Specifically, there was a significant reduction in multiple measures of hypoglycemia-related worries as captured by the HFS-II survey. This is extremely important given the substantial physical and psychological consequences of hypoglycemia faced by patients with PBH, and the toll they take on quality of life. In patients with diabetes, recurrent severe episodes of hypoglycemia can lead to behavioral changes including food avoidance, social isolation, cognitive impairment, unawareness of hypoglycemia, and decreased quality of life [6, 12]. Because of these negative consequences, patients may develop psychological fear of hypoglycemia, leading to phobia, anxiety, depression, and disordered eating [13]. Surveys of patients with PBH, including this study, demonstrate severe psychological consequences of hypoglycemia. In the current cohort, 57% of patients were unable to work due to hypoglycemia, 35% reported behavioral changes, and 71% reported that they would have benefitted from third-party assistance. In a prior study of 32 patients with PBH [2], 100% reported that repeated hypoglycemia negatively affected their quality of life, 93% considered themselves to be disabled, 42% were unable to work, and 10% reported requiring assistance to perform activities of living. Additionally, 65% reported depression, 56% reported memory problems, and 36% reported anxiety as a consequence of their hypoglycemic disorder [2]. Not surprisingly, the scores on a standardized questionnaire that evaluates hypoglycemia-related fears and behavior changes: HFS-II scores were extremely high in our study (52 for worries and 37 for behaviors), which compares to 18 and 11 in insulin-treated patients with diabetes [14, 15]. The Cummings study also reported a high HFS-II score in their patients, with mean scores for worries and behavior of 46 and 24, respectively [7]. Our results are the first to demonstrate significant improvement in multiple HFS-II worry measures, as well as composite worry score as a result of CGM use. Scores decreased in 15/18 measures, with 6 reaching statistical significance and 2 reaching borderline significance, and composite score decreasing from 52.3 to 43.8 (P = .026). Behavior scores did not exhibit statistically significant changes (Table 7). Our results differ from the Cummings paper, which showed a trend toward worse scores on HFS-II for both worry and behavior sections, possibly indicating that CGM use led to higher awareness of hypoglycemia and more conservative behaviors, but in that study only a small subset of the participants received follow-up surveys after the CGM intervention resulting from staff oversight. In patients with type 1 diabetes, use of CGM has also been shown to increase hypoglycemic confidence and reduce feelings of disease-related and interpersonal distress [15].

In addition to decreased hypoglycemia-related worries, we found evidence of dietary liberalization without worsening of hypoglycemia. The amount of dietary carbohydrates that are tolerated by patients with PBH is quite low and difficult to maintain. In patients with PBH, it has been shown that consuming carbohydrates aggravates hyperinsulinemic hypoglycemia [16] and dietary recommendations suggest 30 g per meal and 20 g per snack [17], which meets the minimum daily carbohydrate intake of 130 g recommended by the American Nutrition Society [18]. Some patients must restrict carbohydrates even more severely, however, to avoid hypoglycemic episodes; thus, many patients with PBH are on far fewer daily carbohydrates than the minimum recommended intake, which is required for brain function and prevention of muscle breakdown and ketosis [18]. The severe dietary restrictions on diet may be why in the unblinded group, the use of CGM was associated with dietary liberalization despite instruction to keep diet constant throughout the study. This is viewed as a positive, as most patients struggle with maintaining extremely low carbohydrate intake, leading to inability to exercise, food-related anxiety, and difficulty maintaining social relationships with family and friends that entail eating together. Thus, increasing carbohydrate tolerability in the setting of extreme restriction, without aggravating hyper and hypoglycemia, is congruent with nutritional recommendations for a healthy diet.

The results of this study extend those of the only other study [7] that has evaluated use of CGM in patients with PBH. This study, also a sequential design in which patients with PBH wore CGM during “masked” and “unmasked” periods of study, also showed a significant reduction in hypoglycemia, with a reduction in hypoglycemic events and percent of time spent low (<54 and <40 mg/dL). The Cummings study differed from ours in that patients were all assigned to the blinded period before the unblinded, which might have introduced systematic bias related to treatment order. Furthermore, the patients were allowed to continue medications to treat hypoglycemia and used a combination of Dexcom G4 and G6, which the study team was able to mask and unmask, whereas our patients were all drug-free, assigned to blinded vs unblinded arms in random order, and used the Dexcom G6 Pro for all data collection including the unblinded period in which patients also wore the unblinded G6, which constituted the study intervention of interest. Both our study and the previously published study also showed less time spent in hyperglycemia >180 mg/dL during the unblinded period, which indicates that patients used the CGM to identify and avoid foods that lead to extreme hyperglycemia, which may have also contributed to reduced hypoglycemia since it is known that higher blood glucose spikes predict more severe crashes [1]. The data from the 2 studies combined provide strong support for use of CGM to prevent hypoglycemia in patients with PBH.

Strengths of the current study include a rigorous study design that minimizes bias. To ensure that differences in the CGM model did not introduce bias, we used the blinded Dexcom G6 Pro for all CGM data used in analyses, even though G6 was used during the unblinded period to inform patients about their glucose in real time. Our patients were assigned to treatment period in random order, thus removing the possibility of bias due to treatment order. We also removed potential confounding by medication use, as all our patients were required to be off glucose-modifying therapies during the study. Last, all our patients had already received counseling on following a low carbohydrate diet that was relatively devoid of simple sugars and processed foods, and patients were instructed not to change their diet throughout the 2 treatment periods. A second strength of this study is the capture of quality of life changes using a validated questionnaire on hypoglycemia-related fears. Limitations include the small sample size, but given this is a rare disease, all published studies are generally small and female-dominant. The small size of the cohort likely limited our ability to detect statistical significance in several metrics in which a clear difference was seen. Our cohort included only 1 male, mirroring the higher proportion of women who undergo bariatric procedures. Finally, we cannot rule out that changes in dietary habits contributed to the differences seen during the 2 treatment periods, even though patients were instructed not to change their diet. Indeed, if changes in diet did contribute to improved hypoglycemia in the unblinded period, it is likely that the real-time monitoring and detection of glucose spikes served as a learning tool to help patients avoid foods that “spiked” their glucose and thus averted some of the postprandial crashes that typically follow spikes.

In summary, results of this study provide important data that support extending the use of CGMs to patients with PBH. Currently, these devices are only approved by insurance for use in patients with a diagnosis of insulin-treated diabetes, in large part to prevent hypoglycemia. The first demonstrated clinical benefit of CGMs was reduction of hypoglycemia in patients with insulin-treated diabetes, which has now been shown in multiple studies [3, 4, 19]. This is due to not only the alarm function, but also real-time viewing of glucose patterns and the ability to detect a steep decline before hypoglycemia even occurs. Patients with PBH have extremely rapid glucose drops and have more frequent episodes of clinically important hypoglycemia than do patients with diabetes. For example, among insulin-treated patients with type 1 and type 2 diabetes, the incidence of symptomatic hypoglycemia is 1 to 2 times per week on average, with incidence of severe hypoglycemia (requiring assistance) 1 to 1.7 episodes per year, which affects approximately 30% of patients [13, 20, 21], whereas 100% of patients with PBH experience glucose <54 mg/dL (level 2 hypoglycemia) at least monthly, and in the current cohort, 42% experienced level 2 hypoglycemia on a daily basis, and 14% experienced it multiple times per day. Furthermore, 72% reported level 3 events weekly, which is similar to the 82% incidence of level 3 events reported by Cummings et al [7]. Furthermore, the prevalence of hypoglycemia unawareness is higher in PBH (37-69%) [2, 7] than in type 1 diabetes (29%) [22], highlighting the increased need for real-time monitoring with an alarmed system. Thus, the data presented here provide strong evidence that use of CGM, through real-time glucose monitoring and alarms that detect lows or impending lows, as well as detection of postprandial highs that enable patients to avoid glucose-spiking foods that precipitate lows, can prevent both frequency and severity of hypoglycemic events. Additionally, use of CGM reduced hypoglycemia-related fear and anxiety, thus improving quality of life. Last, patients on extremely carbohydrate-restricted diets may be confident enough while wearing CGM to liberalize their diet without worsening hypoglycemia, likely due to learning which carbohydrates are tolerable and when to self rescue for impending lows. In summary, use of CGM reduces hypoglycemia and improves quality of life in patients with PBH.

Abbreviations

CGM

continuous glucose monitoring

FSBG

fingerstick blood glucose

GLP-1

glucagon-like peptide-1

HFS-II

Hypoglycemia Fear Survey-II

PBH

postbariatric hypoglycemia

Funding

American Diabetes Association (1-19-ICTS-073 to T.M.) and Dexcom (provided continuous glucose monitors to T.M.).

Disclosures

N.T.: None. S.R.: None. T.S.: None. C.M.C.: Scientific Advisory Board; consultant; inventor on patents: Amylyx Pharmaceuticals. T.M.: Research support from Vogenyx, Recordati; steering Committee for Recordati; Advisory Board and inventor on patent for Amylyx.

Data Availability

Some or all datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.

Clinical Trial Information

ClinicalTrials.gov ID: NCT04615546.