Insulin Response to Oral Stimuli and Glucose Effectiveness Increased in Neuroglycopenia Following Roux-en-Y Gastric Bypass

Abstract

Objective

Hyperinsulinemic hypoglycemia with neuroglycopenia is a rare complication following Roux-en-Y gastric bypass (RYGB) surgery for weight management. This study evaluates insulin secretion and action in response to oral and intravenous stimuli in persons with and without neuroglycopenia following RYGB.

Design and Methods

Cross-sectional cohort studies were performed at a single academic institution to assess insulin secretion and action during oral mixed meal tolerance test (MMTT) and intravenous glucose tolerance test (IVGTT).

Results

Insulin secretion was increased more following oral mixed meal than intravenous glucose in individuals with neuroglycopenia compared to the asymptomatic group. Reduced insulin clearance did not contribute to higher insulinemia. Glucose effectiveness at zero insulin, estimated during the intravenous glucose tolerance test, was also higher in those with neuroglycopenia. Insulin sensitivity did not differ between groups.

Conclusions

Increased beta cell response to oral stimuli and insulin-independent glucose disposal may both contribute to severe hypoglycemia after Roux-en-Y gastric bypass.

Introduction

Hyperinsulinemic hypoglycemia with neuroglycopenia is a rare but increasingly recognized complication of Roux-en-Y gastric bypass (RYGB) surgery (1, 2). Prevalence rates are estimated between 0.1-0.36% of post-RYGB patients (3, 4, 5), although prevalence may be higher (6). Symptoms typically become manifest 1-3 years postoperatively, after weight loss has stabilized, but may occur as long as 20-years later (7). Therapeutic approaches to reduce frequency and severity of hypoglycemia include medical nutrition therapy with controlled portions of low glycemic index carbohydrates, inhibition of intestinal α-glucosidase to slow carbohydrate absorption and thus reduce postprandial insulin secretion, and somatostatin analogues, diazoxide and/or calcium channel blockers to reduce insulin secretion (8, 9, 10). Continuous glucose sensors with alarms for hypoglycemia may have clinical value. Partial pancreatectomy, performed in some patients in attempt to reduce insulin secretion, is not always successful (1, 11). Some patients have responded to enteral nutrition delivered via the bypassed stomach (12) or to gastric restriction procedures (13).

Given the difficulty of achieving glycemic control in patients with neuroglycopenia following RYGB, it is important to better understand mechanisms responsible for severe hypoglycemia. Studies have highlighted contributions of increased insulin secretion in response to meals, potentially mediated in part by increased secretion of incretin hormones glucagon-like peptide-1 (GLP-1) and gastric inhibitory polypeptide (GIP) (14, 15) and potentiated by reduced insulin clearance (15). The importance of gut-derived signals extrinsic to β-cells is also supported by the normal insulin response seen with nutrient delivery to the remnant stomach (12). By contrast, other data support a role for increased β-cell mass and/or intrinsic function.. Pancreatic pathology from patients who underwent partial pancreatectomy demonstrates increased β-cell mass and/or increased nuclear diameter (1, 2, 16, 17). To better understand physiologic differences in persons with post-RYGB hypoglycemia associated with neuroglycopenia, we examined glucose and insulin responses to oral and intravenous stimuli in cohorts with and without neuroglycopenia following RYGB.

Methods

Subjects

Twenty-six persons with prior RYGB for morbid obesity were recruited from clinical practice. Seventeen manifest clinically significant hypoglycemia (neuroglycopenia) defined by documented hypoglycemia associated with altered mental status requiring assistance, with or without seizure. Nine subjects who had undergone uncomplicated RYGB who were weight stable for 6 months or more, and had no history of diabetes or glucose intolerance, and denied hypoglycemic symptoms (asymptomatic, controls) were also studied. Consecutive patients were enrolled and classified based solely on medical history. Exclusion criteria included heart failure, liver or kidney disease, malignancy, acute infection or injury, pregnancy and use of medications known to affect insulin secretion or action. Medications including alpha-glucosidase inhibitors, somatostatin analogues, diazoxide, calcium channel inhibitors were held for two days prior to evaluations. The Joslin Diabetes Center Institutional Review Board approved the study. Written informed consent was obtained from participants.

Mixed Meal Tolerance Test (MMTT)

Subjects were instructed to consume at least 200-g carbohydrate for 3-days before visits. Height and weight were measured using a wall-mounted stadiometer (Holtain Ltd., Crymych, UK) and electronic scale (model 0501; Acme Scale Co., San Leandro, CA), and sitting blood pressure was measured. Fasting blood samples were obtained before a liquid mixed meal (Ensure®, 9 g protein, 40 g carbohydrate, 6 g fat, 240 ml; Abbott Laboratories, Abbott Park, IL). Additional blood samples were collected at 10, 20, 30, 60, and 120 minutes for glucose, insulin, and C-peptide. Corrected insulin response (CIR= l₃₀/(G₃₀ × (G₃₀ - 70))) (18) and composite insulin sensitivity index $\left(\text{CISI}=\frac{10000}{\sqrt{(\text{fasting glucose}\times \text{fasting insulin})\times (\text{mean OGTT glucose})\times (\text{mean OGTT insulin})}}\right)$ (19) were calculated. Insulin secretion rates (ISR) were calculated from plasma C-peptide using I(nsulin-)SEC(retion) (ISEC, Version 3.4a, Hovorka, 1994) and population estimates of C-peptide kinetics (20). Insulin clearance was calculated by dividing area under the curve (AUC) of the insulin secretion rate from zero to 120 minutes, by AUC for insulin over the same time interval (15). Early (0-30 minutes) and late (60-120 min) insulin clearance rates were also calculated. Dumping score was calculated during MMTT using a formula reflecting change in pulse and hematocrit (Hct) as indicator of plasma volume: score =685 × (1-[Hct_basal (100- Hct_20min)/ Hct_20min × (100- Hct_20min)] + 100 (Pulse_20min -Pulse_basal)/ Pulse_basal (21).

Insulin-modified frequently sampled intravenous glucose tolerance test (IVGTT)

Participants returned on a separate day, after overnight fast. Intravenous catheters were placed in the antecubital veins for blood sampling and glucose and insulin injections. Dextrose (0.5 g/kg, 30% solution) was administered over 3-minutes and insulin (0.03 U/kg body weight) was injected intravenously at time 19-minutes. Blood was sampled twice before glucose bolus and at 2, 3, 4, 5, 6, 8, 10, 12, 14, 19, 22, 23, 24, 25, 27, 30, 40, 50, 60, 70, 80, 90, 100, 120, 140, 160 and 180 minutes following dextrose injection (22, 23). MINMOD MILLENNIUM (version 6.02; kindly provided by R.N. Bergman, Cedars-Sinai, Los Angeles, CA) provided estimates of acute insulin secretory response to glucose (AIRg), insulin sensitivity (S_I), glucose effectiveness (S_G) and glucose effectiveness at zero insulin (GEZI). The disposition index (DI), a measure of insulin secretion for prevailing insulin resistance, was calculated as the product of AIRg and S_I. The apparent distribution space of glucose (Vg) was calculated as Vg= 300 × body weight (kg)/G₀, where G₀ is the initial glucose concentration during IVGTT (23). Homeostasis model assessment: insulin resistance (HOMA-IR) and beta-cell function (HOMA-β) were calculated (24). The metabolic clearance rate of insulin (MCRI) during IVGTT was calculated as the ratio of insulin dose over incremental insulin AUC from 20 to 180 minutes (25, 26).

Assays

Glucose was measured by glucose oxidation, fasting cholesterol and high-density lipoprotein (HDL) by cholesterol esterase, triglycerides via hydrolysis to glycerol and free fatty acids (Beckman Synchron CX3delta and CX9, Beckman Coulter, Brea, CA), and hemoglobin A_1c by high-performance liquid chromatography (HPLC) (Tosho 2.2, Tosoh Bioscience, San Francisco, CA) in Joslin Diabetes Center's clinical laboratory, and hematocrit by ZIPocrit (LW Scientific, Lawrenceville GA). Immunoassays were performed in duplicate, including radioimmunoassay (RIA) for insulin and C-peptide (Diagnostic Systems Laboratories, Webster, TX).

Statistical Analysis

Results are presented as mean ± standard error. Primary comparisons were performed between groups with and without neuroglycopenia using Student's t-test, or linear mixed models repeated measures (MMRM) for variables measured multiple times after mixed meal or intravenous glucose. Nonparametric variations (Mann-Whitney U-test) were used when data departed from normal distribution. Exploratory analysis evaluated subsets of the neuroglycopenia group. Analysis was performed using SPSS (SPSS Inc., Version 17.0. Chicago, IL). Results were considered significant for two-tailed P-values less than 0.05.

Results

Study cohorts include 17 persons with neuroglycopenia post-RYGB, five with a previous history of type 2 diabetes (T2D), and 9 asymptomatic participants of similar age, gender, time post-surgery, and reduction in body mass index with no prior diabetes (Table 1). Neuroglycopenic symptoms typically occurred within 3-hours of meal ingestion and resolved with carbohydrate intake (Table 2). There were no differences in glucose or insulin, estimates of insulin secretion or action (HOMA-β or HOMA-IR), or insulin secretory rates (ISR_B) in the fasting state (Table 1).

Table 1.

Clinical and metabolic characteristics of study subjects.

	Neuroglycopenia A (n=17)	Neuroglycopenia –No prior diabetes A (n=12)	Neuroglycopenia –MMTT Hypoglycemia A (n=7)	Asymptomatic A (n=9)	P-valueB	P-valueC	P-valueD
Demographics
Age (years)	46.9 ± 9.8	46.3 ± 10.9	45.3 ± 10.8	48.8 ± 10.5	0.662	0.612	0.525
Gender (Male/Female)	1/16	0/12	0/7	3/6	0.065E	0.063E	0.213E
Pre-op diagnosis of type 2 diabetes	5/17	NA	1/7	0/9	0.129E	NA	0.438E
Years post RYGB	4.93 ± 2.12	5.12 ± 1.68	4.37 ± 1.60	3.44 ± 2.23	0.099	0.057	0.346
Current BMI (kg/m2)	32.6 ± 6.0	32.5 ± 6.1	32.5 ± 6.2	28.4 ± 5.3	0.091	0.124	0.173
Reduced BMI (kg/m2)	18.7 ± 7.9	15.2 ± 4.0	15.3 ± 2.15	16.4 ± 3.8	0.407	0.502	0.500
Waist/Hip ratio	0.89 ± 0.83	0.87 ± 0.08	0.87 ± 0.09	0.90 ± 0.10	0.766	0.455	0.489
Systolic Blood Pressure (mmHg)	125.0 ± 16.9	123.6 ± 17.2	123.0 ± 14.4	120.3 ± 14.6	0.490	0.653	0.720
Diastolic Blood Pressure (mmHg)	75.5 ± 11.2	75.3 ± 10.4	74.4 ± 12.6	74.0 ± 9.1	0.727	0.776	0.938
Fasting Assessments
HbA1c (%)	5.40 ± 0.30	5.43 ± 0.29	5.46 ± 0.26	5.59 ± 0.37	0.168	0.268	0.729
Cholesterol (mg/dl)	163.6 ± 21.5	166.4 ± 23.0	154.3 ± 18.2	162.0 ± 29.3	0.876	0.703	0.553
Triglycerides (mg/dl)	90.6 ± 34.3	90.7 ± 40.0	86.4 ± 32.8	76.7 ± 19.8	0.310	0.348	0.471
HDL-C (mg/dl)	58.2 ± 17.2	62.1 ± 19.0	58.6 ± 11.8	66.0 ± 23.8	0.347	0.679	0.463
Fasting glucose (mg/dl)G	74.0 ± 5.9	75.7 ± 5.7	73.0 ± 3.7	72.5 ± 4.5	0.521	0.179	0.933
Fasting insulin (μU/ml)G	4.53 (3.83, 8.13)	5.37 (3.83, 10.71)	3.98 (2.90, 6.45)	3.58 (2.62, 10.32)	0.628F	0.270F	0.672F
HOMA-β (% mU/ml)G	184.8 (145.5, 220.5)	168.0 (108.2,355.7)	148.3 (96.0,387.0)	209.3 (82.4, 359.0)	0.235F	0.943F	0.958F
HOMA-IR (mg×μU×ml-2×103)G	0.93 (0.66, 1.54)	1.02 (0.64, 2.13)	0.72 (0.53, 1.10)	0.64 (0.47, 1.93)	0.808F	0.255F	0.672F
Fasting ISR (pmol×(kg×min)-1)H	2.60 ± 1.35	2.17 ± 0.91	2.23 ± 0.51	2.12 ± 0.73	0.336	0.898	0.983
Dumping Score (%)H	80.5 ± 43.8	83.7 ± 43.4	92.8 ± 42.8	68.5 ± 48.8	0.527	0.459	0.314
Vg (kg×106)G	111.0 ± 25.6	102.5 ± 18.5	106.3 ± 21.3	124.4 ± 27.3	0.228	0.041	0.172

^AMean ± SD or Median (IQR);

^B-DP-value by independent sample t-tests unless specified;

^BP-value between Neuroglycopenia group and Asymptomatic group;

^CP-value between Neuroglycopenia with no prior history of diabetes subgroup and Asymptomatic group;

^DP-value between Neuroglycopenia with hypoglycemia during mixed meal tolerance test (MMTT) subgroup and Asymptomatic group;

^EChi-Square;

^FMann-Whitney U-test;

^GCalculated on Day of Intravenous Glucose Tolerance Test;

^HCalcuated on Day of Mixed Meal Glucose Tolerance Test; to convert HbA_1c to International Federation of Clinical Chemistry (IFCC) standard millimole per mole, use (HbA_1cminus 2.15) multiplied by 10.919; Scientific International conversion factors: to convert cholesterol and HDL-C to millimoles per liter, multiply by 0.0259;triglycerides to millimoles per liter, 0.0113; glucose to millimoles per liter, 0.0555; insulin to picomole per liter, 6.945. RYGB, Roux-en-Y gastric bypass; BMI, body mass index; HbA_1c, hemoglobin A_1c; HDL-C, high-density lipoprotein cholesterol; HOMA-β, homeostasis model assessment of beta-cell function; HOMA-IR, homeostasis model assessment of insulin resistance; ISR, insulin secretion rate; Vg, the apparent distribution space of glucose.

Table 2.

Clinical presentation of study participants with neuroglycopenia.

Subject	Time from surgery to first neuroglycopenic event (yr)	Clinical symptoms	Symptom frequency	Postprandial delay (h)	Lowest glucose level at neuroglycopenic event (mg/dl)	Medical treatment
1	5.2	Confusion	daily	1	30	Medical nutrition therapy only
2	3.0	Confusion	1-2 × weekly	1	23	Acarbose
3	7.0	Syncope, loss of consciousness	weekly	2	33	Miglitol
4	1	Weak, sweaty, headache	4 × weekly to 3 × daily	1-2	40	Pramlintide, acarbose, octreotide
5	2.0	Seizure, loss of consciousness	3-4 × weekly	2	44	Medical nutrition therapy only
6	4.0	Presyncope	2 × weekly to 6 × daily	0.3-1	28	Acarbose, octreotide
7	0.8	Seizures, loss of consciousness	weekly	2	26	Acarbose, diazepam,octreotide, clonazepam
8	1.0	Confusion	3 × weekly	1	25	Diazoxide, acarbose, octreotide
9	2.5	Lightheaded, fatigue	1-2 × weekly	2-3	29	Acarbose, nifedipine,diazepam
10	1.9	Loss of consciousness	2-3 × daily	2-3	50	Acarbose
11	3.3	Visual blurred, sweaty	3-4 × weekly	1-2	30-40	Medical nutrition therapy only
12	1.0	Confusion	1-2 × weekly	1	41	Acarbose
13	2.9	Visual blurred, sweaty, shaky	2-3 × daily	2-3	39	Acarbose, octreotide,diazoxide
14	6.6	Seizures	3 × weekly	2-3	26	Acarbose
15	6.0	Dizziness, assistance from others	3 × daily to 1 × weekly	1	68	Medical nutrition therapy only
16	2.0	Loss of consciousness	weekly	0.5-1	52	Acarbose, octreotide
17	2.0	Seizures, loss of consciousness	2-3 × daily	2	39	Acarbose, octreotide

Participants were not symptomatic for hypoglycemia during mixed meal tolerance testing. In response to oral mixed meal, there were no differences in mean glucose concentrations, either overall for the 120 minutes of evaluation (P_(MMRM) =0.916), or at any individual time-point (Figure 1A). However, 7/17 (41%) of the neuroglycopenic group had a glucose level below 60 mg/dl [3.33 mmol/L] at 60 and/or 120 minutes, while none of the nine asymptomatic participants had a glucose level below this threshold (χ²P=0.058). One of these 7 individuals with lower glucose levels achieved had diabetes prior to RYGB. C-peptide levels were higher (P_(MMRM) =0.028) and insulin tended to be higher (P_(MMRM) =0.077) in neuroglycopenic compared to asymptomatic groups (Figure 1B-C). Difference in C-peptide was greatest at 60-120 minutes after meal ingestion, approximately 47% higher in the neuroglycopenic group (AUC C-peptide _{(60-120 min)} 7.5 ± 2.7 versus 5.1 ± 2.7 min×10^-4), neuroglycopenic vs asymptomatic, P=0.043)(Figure 1B, Table 3). Corrected insulin response (CIR) (18) was 75% greater (P=0.027), and both C-peptide-to-glucose (P=0.023) and insulin-to-glucose ratios (P=0.047) were higher in neuroglycopenic compared to asymptomatic groups (Figure 1D-F). In contrast, estimated insulin secretion rates were similar between groups (Figure 1G). Insulin clearance did not differ between groups over the entire mixed meal study, the early (0-20 minute), or late (60-120 minute) assessment intervals (Table 3). The composite insulin sensitivity index (CISI) (19) did not differ between groups (Figure 1H). Dumping score also did not differ between groups (Table 1).

Figure 1.

Mixed meal tolerance test (MMTT)-derived data and calculated values for (A) glucose (mg×dl^-1), (B) C-peptide (ng×ml^-1), (C) insulin (μU×ml^-1), (D) the corrected insulin response to glucose (CIR₃₀, 10×mU×ml×mg^-2),(E) C-peptide/glucose ratio (×10^-4), (F) insulin/glucose ratio (μU×10²×mg^-1), (G) insulin secretion rate (ISR, pmol×(kg×min)^-1), and (H) the composite insulin sensitivity index (CISI, dl×ml×(μU×mg)^-1) are provided. P-values were calculated by linear mixed models repeated measures analysis [A-C and E-G]. Boxplots with median (line), mean (x), interquartile range (box) and minimum and maximum (whiskers) are shown [D and H]. P-values were calculated by independent sample t-tests [D] and nonparametric variations (Mann-Whitney U-test) [H].

Table 3.

Baseline and metabolic response to mixed meal and intravenous glucose tolerance tests.

	Neuroglycopenia A (n=17)	Neuroglycopenia –No prior diabetes A (n=12)	Neuroglycopenia –MMTT hypoglycemiaA (n=7)	Asymptomatic A (n=9)	P-valueB	P-valueC	P-valueD
Mixed Meal Tolerance Test
CIR30 (10×mU×ml×mg-2)	0.022 ± 0.018	0.025 ± 0.018	0.036 ± 0.018	0.012 ± 0.006	0.027	0.046	0.013
CISI [dl×ml×(μU×mg)-1]E	6.02 (4.18,7.51)	6.46 (5.17,7.60)	6.07 (2.59,7.50)	6.73 (4.03,17.24)	0.186	0.256	0.239
Peak postprandial glucose (mg×dl-1)	150.1 ± 27.6	153.7 ± 29.4	140.6 ± 20.2	147.7 ± 19.1	0.815	0.601	0.484
Peak postprandial insulin (μU×ml-1) E	162.9 (130.3, 282.0)	175.7(121.8,284.5)	163.9(103.5,249.2)	125.3 (76.4, 190.0)	0.225	0.136	0.050
AUC C-peptide/glucose (min×10-4)	15.0 ± 4.82	15.5 ± 5.40	17.9 ± 4.31	11.2 ± 4.80	0.085	0.073	0.023
AUC C-peptide/glucose (60-120 min) (min×10-4)	7.52 ± 2.72	7.63 ± 3.14	8.91 ± 2.55	5.10 ± 2.72	0.043	0.075	0.034
Fasting ISR [pmol×(kg×min)-1]	2.60 ± 1.35	2.17 ± 0.91	2.23 ± 0.51	2.12 ± 0.73	0.336	0.898	0.739
Peak ISR [pmol×(kg×min)-1]	32.0 ± 9.27	33.1 ± 8.61	35.8 ± 6.69	28.5 ± 9.60	0.379	0.320	0.110
Insulin clearance [pmol×(kg×min)-1]	0.228 ± 0.087	0.221 ± 0.082	0.202 ± 0.079	0.251 ± 0.110	0.567	0.435	0.335
Insulin clearance (0-30)min [pmol×(kg×min)-1]	0.222 ± 0.084	0.213 ± 0.081	0.190 ± 0.087	0.253 ± 0.108	0.439	0.351	0.233
Insulin clearance (60-120)min [pmol×(kg×min)-1]	0.339 ± 0.284	0.260 ± 0.171	0.436 ± 0.372	0.356 ± 0.284	0.886	0.344	0.631
Intravenous Glucose Tolerance Test
Acute Insulin Response (AIRg) (μU×min×ml-1)E	668.3 (251.2, 777.8)	568.2 (268.1, 836.7)	737.9 (234.6,1353.9)	222.7 (165.6, 394.7)	0.024	0.025	0.091
Insulin Sensitivity Index (SI) [103× (mU×min)-1]E	2.61 (1.57, 3.92)	2.99 (2.22, 4.16)	2.30 (1.93,3.63)	3.92 (3.18, 5.03)	0.056	0.155	0.050
Glucose estimate at time zero (G0) (mg×dl-1)E	231.6 (214.0, 265.8)	249.6 (231.5, 272.7)	246.5 (224.9,265.3)	192.8 (185.0, 210.4)	0.001	0.001	0.003
Glucose effectiveness at zero insulin (GEZI) (min-1)E	0.02 (0.02, 0.03)	0.025 (0.02, 0.03)	0.03 (0.02,0.03)	0.01 (0.01, 0.02)	0.018	0.008	0.008
Disposition Index (DI) E	989.5 (599.5, 1413.3)	1168.2 (950.9, 2831.2)	1618.8 (340.0,3111.2)	844.5 (748.6, 109.9)	0.374	0.102	0.266
Glucose Effectiveness (SG) [min-1]E	0.02 (0.02, 0.03)	0.03 (0.02, 0.03)	0.03 (0.02,0.03)	0.02 (0.01, 0.02)	0.091	0.025	0.011
AUC Insulin/Glucose (μU×min×mg-1×102)E	47.4 (35.0, 67.8)	54.5 (33.2, 76.5)	74.1 (64.4, 122.2)	32.3 (26.2, 45.1)	0.049	0.076	0.039
Δ AUC Insulin/Glucose (μU×min×mg-1×102)E	31.0 (24.6, 45.2)	37.5 (21.1, 62.7)	44.2 (27.9, 46.1)	21.0 (15.0, 27.7)	0.048	0.105	0.040
AUC Insulin/Glucose (1-10 min) (μU×min×mg-1×102)E	2.81 (1.44, 4.12)	2.78 (1.56, 4.13)	3.56 (1.42, 5.46)	1.51 (1.18, 2.44)	0.095	0.082	0.204
ΔAUC Insulin/Glucose (1-10 min) (μU×min×mg-1×102)E	1.87 (0.77, 3.54)	1.68 (0.82, 3.79)	2.99 (0.78, 3.94)	1.03 (0.79, 1.86)	0.244	0.305	0.165
AUC Insulin/Glucose (120-180 min) (μU×min×mg-1×102)E	7.12 (5.20, 10.45)	6.74 (3.91, 11.23)	6.00 (3.10, 12.40)	3.51 (2.35, 7.45)	0.046	0.102	0.244
ΔAUC Insulin/Glucose (120-180min) (μU×min×mg-1×102)E	2.45 (1.79, 3.74)	2.39 (0.70, 2.74)	2.40 (0.27, 9.52)	0.85 (0.25, 2.41)	0.062	0.217	0.247
Metabolic Clearance Rate of Insulin (MCRI) (l×min-1)E	0.72 (0.52, 1.07)	0.66 (0.38, 1.08)	0.83 (0.54, 1.03)	0.96 (0.52, 1.09)	0.609	0.414	0.427

^AMean ± SD or Median (IQR);

^B-DP-value by independent sample t-tests unless specified;

^BP-value between Neuroglycopenia group and Asymptomatic group;

^CP-value between Neuroglycopenia with no prior history of diabetes subgroup and Asymptomatic group;

^DP-value between Neuroglycopenia with hypoglycemia during mixed meal tolerance test (MMTT) subgroup and Asymptomatic group;

^EMann-Whitney U-test; Scientific International unit conversion factors: to convert glucose to millimoles per liter, multiply by 0.0555; insulin to picomole per liter, 6.945. CIR₃₀, corrected insulin response at 30 min; CISI, composite insulin sensitivity index; AUC, area under the curve; ISR, insulin secretion rate.

The IVGTT was successfully completed in all individuals; no study was terminated prematurely for hypoglycemia after administration of exogenous insulin. Although the dose of dextrose administered did not differ (0.5 g/kg), serum glucose concentrations achieved were higher in neuroglycopenic compared to asymptomatic groups (P_(MMRM) =0.007), particularly in the early time period (Figure 2A). G₀, the estimated initial glucose concentration, was also higher in the neuroglycopenia group (median 231.6 (IQR 214.0 to 265.8) versus 192.8 (IQR 185.0 to 210.4) mg/dl, neuroglycopenic versus asymptomatic respectively, P=0.001). There was no difference in estimated volume of distribution (Vg) between groups. Insulin concentrations achieved acutely (before exogenous insulin administration) and the acute insulin response to glucose (AIRg) (P=0.024) during IVGTT were also higher in the neuroglycopenia group (Figure 2B, 2D). As insulin sensitivity index (S_I) tended to be lower in neuroglycopenic patients, the disposition index did not differ between groups (Figure 2E-G). However, those with neuroglycopenia tended to distribute along the DI curve with greater insulin secretion and resistance compared to lower insulin secretion and resistance in the asymptomatic group (Figure 2G).

Figure 2.

Intravenous glucose tolerance test (IVGTT)-derived values for (A) glucose (mg×dl^-1), (B) insulin (μU×ml^-1), (C) insulin/glucose ratio (μU×10²×mg^-1), (D) acute insulin response to glucose (AIRg, μU×min×ml^-1), (E) insulin sensitivity index [S_I, l0³× (mU×min)^-1], (F) disposition index (DI), (G) the relationship between AIRg and S_I, in individuals with neuroglycopenia (●) or in asymptomatic (○) persons, (H) glucose effectiveness at zero insulin (GEZI, min^-1) and (I) glucose effectiveness (S_g, min^-1),are provided. P-values were calculated by linear mixed models repeated measures analysis [A-C]. Boxplots with median (line), mean (x), interquartile range (box) and minimum and maximum (whiskers) are shown with P-values calculated by nonparametric variations (Mann-Whitney U-test) [D-F and H-I]. *P<0.05 versus asymptomatic group.

Analysis of additional insulin secretion markers during IVGTT demonstrated the ratio of change in insulin to change in glucose did not differ between groups in the first 10 minutes following intravenous dextrose. However, the area under the curve (AUC) insulin to glucose was increased overall (P=0.049) and approximately double in the last hour of observation (120-180 minutes) (AUC Insulin/Glucose _{(120-180 min)} median 7.1 (IQR 5.2 to 10.5) versus 3.5 (IQR 2.4 to 7.5) μU×min×mg^-1×10², P=0.046) in the neuroglycopenic compared with asymptomatic group (Table 3). Higher insulin could not be explained by reduced insulin clearance, as metabolic clearance rate of insulin during IVGTT did not differ between groups (Table 3). Additionally, glucose effectiveness at zero insulin (GEZI) was higher in neuroglycopenic compared with asymptomatic groups (median 0.02 (IQR 0.02 to 0.03) versus 0.01 (IQR 0.01 to 0.02) min^-1, neuroglycopenic versus asymptomatic, Mann-Whitney P=0.018), with similar trend for glucose effectiveness (S_G) (Figure 2G-H). In consideration of the potential role of glucose effectiveness on insulin secretion, we performed an exploratory analysis and found S_G inversely correlated with HOMA-β (r=-0.541, P=0.025) but not AIRg (r=0.146, P=0.576) in the neuroglycopenic group. The opposite was true in the asymptomatic group where S_G did not correlate with HOMA-β (r=0.354, P=0.350) but did correlate with AIRg (r=0.764, P=0.017).

Neuroglycopenia Subgroup Analysis

Secondary analysis was performed excluding those with prior diabetes in the neuroglycopenia group (Table 1). Again, glucose concentrations during MMTT were similar between groups (Table 3, Figure 3). C-peptide, insulin, ratios of both C-peptide and insulin to glucose, and corrected insulin response (CIR) were higher in those with neuroglycopenia compared to those without. During IVGTT, the acute insulin response to glucose (AIRg) was higher in neuroglycopenic individuals, but disposition index (DI) did not differ. Glucose effectiveness at zero insulin (GEZI) was higher, and insulin independent glucose disposal (glucose effectiveness, S_G) emerged as significantly higher in those with compared to those without neuroglycopenia. Estimates of insulin clearance did not differ between groups.

Figure 3.

Comparison of neuroglycopenic subjects without diabetes prior to RYGB (NG-No-DM) (●) and asymptomatic (ASx) subjects (○). Mixed meal tolerance test (MMTT) [A-G] and Intravenous Glucose Tolerance Test (IVGTT)-derived [H-L] values for (A) glucose (mg×dl^-1), (B) C-peptide (ng×ml^-1), (C) insulin (μU×ml^-1), (D) C-peptide/glucose ratio (×10^-4), (E) insulin/glucose ratio (μU×10²×mg^-1), (F) insulin secretion rate (ISR, pmol×(kg×min)^-1), (G) the corrected insulin response to glucose (CIR₃₀, 10×mU×ml×mg^-2), (H) acute insulin response to glucose (AIRg, μU×min×ml^-1), (I) insulin sensitivity index [S_I, 10³× (mU×min)^-1], (J) disposition index (DI), (K) glucose effectiveness at zero insulin (GEZI, min^-1), and (L) glucose effectiveness (S_G, min^-1) are provided. P-values were calculated by linear mixed models repeated measures analysis [A-F]. Boxplots with median (line), mean (x), interquartile range (box) and minimum and maximum (whiskers) are shown. P-values were calculated by independent sample t-tests [G] and nonparametric variations (Mann-Whitney U-test) [H-L].

A second subgroup analysis was performed considering those that manifest glucose ≤60 mg/dl during MMTT (Table 1). In this subset defined by hypoglycemia during MMTT, glucose concentrations were lower compared to the asymptomatic group (P=0.021)(Table 3, Figure 4). C-peptide, insulin, ratios of both C-peptide and insulin to glucose, and the beta cell estimate CIR were likewise higher those with neuroglycopenia compared to those without. During IVGTT, the acute insulin response to glucose (AIRg) tended higher, but the group was more insulin resistant by the index S_I (P=0.050), again with no difference in the disposition index (DI). Estimates of insulin clearance did not differ. However, glucose effectiveness at zero insulin (GEZI) was higher, and glucose effectiveness (S_G) was also significantly greater in those with neuroglycopenia.

Figure 4.

Comparison of participants with neuroglycopenia who manifest glucose concentratons less than or equal to 60 mg/dl during mixed meal tolerance test (NG-hypo-MMTT) (●) and asymptomatic (ASx) subjects (○). Mixed meal tolerance test (MMTT) [A-G] and Intravenous Glucose Tolerance Test (IVGTT)-derived [H-L] values for (A) glucose (mg×dl^-1), (B) C-peptide (ng×ml^-1), (C) insulin (μU×ml^-1), (D) C-peptide/glucose ratio (×10^-4), (E) insulin/glucose ratio (μU×10²×mg^-1), (F) insulin secretion rate (ISR, pmol×(kg×min)^-1), (G) the corrected insulin response to glucose (CIR₃₀, 10×mU×ml×mg^-2), (H) acute insulin response to glucose (AIRg, μU×min×ml^-1), (I) insulin sensitivity index [S_I, 10³× (mU×min)^-1], (J) disposition index (DI), (K) glucose effectiveness at zero insulin (GEZI, min^-1), and (L) glucose effectiveness (S_G, min^-1) are provided. P-values were calculated by linear mixed models repeated measures analysis [A-F]. Boxplots with median (line), mean (x), interquartile range (box) and minimum and maximum (whiskers) are shown. P-values were calculated by independent sample t-tests [G] and nonparametric variations (Mann-Whitney U-test) [H-L].

Discussion

This study was performed to evaluate physiologic differences in response to oral and intravenous stimuli in post-RYGB patients with and without severe hypoglycemia with neuroglycopenia. We did not find differences in basal estimates of insulin secretion or action, consistent with clinical observations that hypoglycemia occurs postprandially. However, in response to mixed meal, C-peptide-to-glucose ratio, insulin-to-glucose ratio, and corrected insulin response (CIR) were all higher in patients with neuroglycopenia as compared to those without neuroglycopenia. We find no difference in insulin clearance to account for higher insulin concentrations following mixed meal. With intravenous glucose stimulus, the acute insulin response (AIRg) was also higher in those with neuroglycopenia. However, this higher insulin response was proportionate to the higher glucose concentration achieved at early timepoints during the test, and/or degree of insulin resistance, indicating that intrinsic β-cell secretion is appropriate. Together, these data suggest the pancreatic β-cell response is exaggerated following oral but not with intravenous stimulation.

Our findings are consistent with other reports of exaggerated β-cell response to oral stimuli, and with those that find exaggerated incretin secretion or response in individuals with post-RYGB hypoglycemia (8, 14, 27, 28). Like others (15) we do not find differences in insulin secretion rates following oral stimulation in those with and without neuroglycopenia post-RYGB. Our findings that the β-cell response is not disproportionally exaggerated in response to bolus intravenous glucose, when considered in response to glucose level achieved and magnitude of insulin resistance, are similar to those seen with graded intravenous glucose infusion (29); this prior investigation did not compare response to oral and intravenous stimuli. In contrast to others (15), we do not find differences in insulin clearance rates. Both investigations share similar limitations on estimates of insulin clearance, which may have bias given non-steady state rates of secretion and clearance, and methods which use peripheral (post-hepatic) insulin concentrations. Absence of an exaggerated β-cell response following intravenous glucose is also consistent with the relative normalization of glucose and insulin profiles in response to mixed meal administration via gastrostomy tube into the remnant stomach, as compared to oral administration (12). Previous reports of increased nuclear diameter in β-cells from affected individuals raised the possibility of intrinsic β-cell hyperfunction (16), but our findings of normal response to intravenous glucose stimuli do not support marked cell-autonomous perturbations of β-cell responses to glucose. Together, our findings suggest that functional, reversible abnormalities in β-cell response to orally-administered nutrients contribute substantially to hypoglycemia, but these defects are unlikely to be fully β-cell-autonomous.

Our studies are the first to demonstrate glucose effectiveness at zero insulin to be higher in those with neuroglycopenia post-RYGB, indicating that additional mechanisms of insulin-independent glucose disappearance from the circulation could contribute to more profound hypoglycemia. These include accelerated glucose uptake into multiple tissues, potentially via increased glucose transport or phosphorylation, glycogen synthesis, glucose oxidation, or activity of the pentose phosphate shunt or hexosamine biosynthesis pathways (reviewed in (30)). Interestingly, hyperinsulinemic hypoglycemia due to insulinoma has not been associated with differences in glucose effectiveness (31). These observations are of particular interest given recent demonstration of increased intestinal glucose uptake in rodents post-RYGB (32), suggesting that dysregulation of insulin-independent glucose uptake could be a consistent feature of post-RYGB hypoglycemia. Alternatively, increased bile acids occurring after gastric bypass (33) might induce increased energy expenditure (34). Furthermore, relationships between glucose effectiveness and fasting (HOMA-β) and dynamic (AIRg) estimates of β-cell function differed between neuroglycopenic and asymptomatic groups. Increased insulin sensitivity does not appear to contribute to accelerated glucose disappearance, as the neuroglycopenic group was not more insulin sensitive and in fact tended to be more insulin resistant.

Differences in β-cell functional responses to oral compared with intravenous glucose tolerance test could be due to differences in nutrient composition, with mixed nutrient composition of the MMTT compared with only glucose administered during IVGTT. Stronger potentiation of β-cell function has been demonstrated in healthy persons following mixed meal compared with oral glucose alone (35), potentially due to the added effect of amino acids and fatty acids. However, administration of simple glucose challenge is difficult in post-RYGB patients given the profound dumping and vasomotor response that could also confound potential findings if oral glucose alone were used.

Differences between dynamic insulin response to mixed meal compared to intravenous glucose could also result from route of administration, as incretin action would not contribute to enhanced insulin secretion following intravenous glucose. Incretins were not measured in this study, but other studies have suggested GLP-1 contributes to improved β-cell function post-RYGB (27, 28). Several studies show GLP-1 levels are higher in patients with neuroglycopenia following RYGB (14, 15) although this is not consistently observed (28). Other gut incretins or metabolites could also contribute, but to date have not been found to be higher in patients with post-RYGB hypoglycemia.

Multiple estimates of insulin sensitivity, assessed by fasting (HOMA-IR) or dynamic response to mixed meal (CISI) or IVGTT (S_I), demonstrate no difference in insulin action between neuroglycopenic and asymptomatic groups. However, the two dynamic measures CISI and S_I both trend toward greater insulin resistance, not sensitivity, in the neuroglycopenic group. We acknowledge that euglycemic-hyperinsulinemic clamps, considered gold-standard for whole-body glucose disposal measures, might detect differences with sufficient sample sizes. However, it is possible that chronic hyperinsulinemia in post-RYGB hypoglycemia could itself contribute to insulin resistance, via downregulation of insulin receptor expression or signaling (36), as seen in transgenic mice overexpressing the insulin gene (yielding two to four-fold elevations in plasma insulin) which develop insulin resistance. Similarly, patients with sustained elevations in plasma insulin due to insulinoma have resistance to administered insulin (31, 37, 38). Alternatively, it is possible that patients developing post-bypass hypoglycemia have higher insulin resistance due to their tendency to higher body weight or potentially due to underlying genetic background. Notably, these patterns are distinct from those in persons with spontaneous postprandial reactive hypoglycemia, who have higher insulin sensitivity (S_I) compared to healthy controls (39).

Participants were classified based solely on clinical history. Metabolic responses of glucose, insulin secretion and action, and glucose effectiveness all fell along a continuum and were not distinctly different based on clinical symptomology, suggesting that additional factors not measured including alternate fuel availability, counter-regulation, or neuronal adaptation may impact awareness and cognitive response to hypoglycemia, or that multiple subtle metabolic differences along a normal spectrum of response when present concurrently contribute to the severity of clinical symptoms. One recent study suggests patients with history of symptoms suggestive of hypoglycemia perceive symptoms but do not have physiologic differences (40). However, none of these individuals manifest neuroglycopenia. None of our neuroglycopenic participants were symptomatic during the MMTT, as previously reported (14). There are many potential explanations, but all are speculative. Potential factors contributing to greater severity of hypoglycemia in the ambulatory as compared with the research setting include: 1) meal size and composition, 2) increased gastric distention with solid foods, and 3) increased physical activity. Repeated asymptomatic bouts of hypoglycemia, with secondary loss of adrenergic symptoms, may contribute to the frequency and severity of more profound hypoglycemia. Furthermore, altered consciousness in the ambulatory setting may also reflect volume shifts and relative hypotension in the postprandial state.

Interestingly the glucose level achieved following intravenous glucose bolus administration was higher in patients with neuroglycopenia. This suggests either differences in volume of distribution or diminished immediate glucose clearance. The estimated volume of distribution was not different using established formulas, although these may still be imprecise. Glucose effectiveness was higher while insulin sensitivity tended to be lower, potentially yielding opposing effects on glycemia. Thus the mechanism(s) for higher glucose achieved following intravenous injection cannot be explained at this time. Evaluation of differences in hepatic glucose uptake or insulin suppression of hepatic glucose output might be warranted, although others have not demonstrated differences in endogenous glucose production, either in the basal state or after oral stimulus, in patients with and without hypoglycemia following RYGB (27).

Finally two subgroup analyses, one excluding those with diabetes prior to RYGB and one including only those with hypoglycemia manifest during MMTT, showed highly consistent findings of both increased insulin response to oral compared to intravenous stimuli and enhanced insulin-independent glucose disposal.

In summary, we find higher index of β-cell function in response to oral mixed meal in patients with neuroglycopenia compared to those without neuroglycopenia post-RYGB. In contrast, the insulin secretory response to intravenous glucose is proportionate to the glucose stimulus and to the level of insulin resistance. These findings support a role for gut-derived peptides or metabolites to contribute to increased insulin secretion in the postprandial state, and indicate hypoglycemia is not solely due to intrinsic excessive β-cell function, at least in the majority of individuals. Whether some patients could have additional β-cell defects, either genetic or acquired, which contribute to insulin secretory dysfunction, remains an unanswered question. Finally, increased glucose effectiveness may promote glucose uptake and/or disposal and further contribute to hypoglycemia, providing additional potential novel therapeutic targets.

Bullets.

1. What is known?

• Hyperinsulinemic hypoglycemia with neuroglycopenia is a rare complication following Roux-en-Y gastric bypass (RYGB) surgery for weight management.

• Increased insulin secretion in response to meals, mediated in part by increased secretion of the incretin hormones glucagon-like peptide-1 (GLP-1), contributes to postprandial hyperinsulinemic hypoglycemia.

• The contribution of intrinsic beta cell function and insulin-dependent or -independent glucose disposal to hyperinsulinemic hypoglycemia remains incompletely understood.

2. What does this study add?

• Persons with hyperinsulinemic hypoglycemia with neuroglycopenia demonstrate increased beta-cell response to oral stimuli compared to asymptomatic persons who have also previously had RYGB.

• The acute insulin secretory response to intravenous glucose is higher in the neuroglycopenic compared with the asymptomatic group, but with the trend for lower insulin sensitivity index there is no difference between groups in the disposition index.

• Insulin-independent glucose disposal may also contribute to severe hypoglycemia after Roux-en-Y gastric bypass.

Abbreviations

AIRg

Acute insulin secretory response to glucose

ASx

Asymptomatic

AUC

Area under the curve

BMI

Body mass index

CIR

Corrected insulin response

CISI

Composite insulin sensitivity index

DI

Disposition index

G₀

Glucose concentration at time zero

GEZI

Glucose effectiveness at zero insulin

GIP

Gastric inhibitory polypeptide

GLP-1

Glucagon-like peptide-1

HbA1c

Hemoglobin A1c

Hct

Hematocrit

HDL

High-density lipoprotein

HOMA-β

Homeostasis model assessment of beta-cell function

HOMA-IR

Homeostasis model assessment of insulin resistance

HPLC

High-performance liquid chromatography

ISEC

Insulin secretion

ISR

Insulin secretion rate

IVGTT

Intravenous glucose tolerance test

MCR_I

Metabolic clearance rate of insulin

MMRM

Mixed model repeated measures

MMTT

Mixed meal tolerance test

NG

Neuroglycopenia

RYGB

Roux-en-Y gastric bypass

S_I

Insulin sensitivity

S_G

Glucose effectiveness

Vg

The apparent distribution space of glucose