Unraveling, contributing factors to the severity of postprandial hypoglycemia after gastric bypass surgery

M.D

Postprandial hypoglycemia is an increasingly recognized late metabolic complication of bariatric surgery also known as post-bariatric hypoglycemia (PBH) [1,2]. The condition affects approximately 30% of patients undergoing Rouxen-Y gastric bypass (RYGB) surgery [3] and is characterized by hypoglycemic episodes occurring w90-120 minutes after meals. PBH is the result of a dysregulated glucose-insulin homeostasis. Previous work suggested a higher insulin exposure following meal intake with the involvement of several factors such as increased insulin secretion, driven by high incretin levels (another hallmark of PBH) and/or intrinsic b-cell alterations as well as diminished hepatic insulin extraction [4,5]. Differential contribution of these factors may explain why PBH manifests with varying degrees of biochemical severity, ranging from mild transient events to more pronounced episodes requiring self or third party treatment. A better understanding of this clinical heterogeneity has the potential to develop more targeted clinical management strategies and improve safety of affected patients.
The objective of this work is to contrast parameters of glucose-insulin homeostasis in 2 distinct phenotypes of PBH (mild versus moderate hypoglycemia) based on nadir plasma glucose.

Study design and population
The retrospective analysis included data from studies conducted at the University Hospital Bern (Switzerland). This study involved adults (age 18 years) who underwent RYGB .12 months ago. Patients with documented evidence of the Whipple's triad (interstitial or plasma glucose ,54mg/dL at time of hypoglycemia symptoms, relieved by correction) were recruited. The respective medical diagnosis was performed on the grounds of objective assessments (supervised diagnostic test or blinded continuous glucose monitoring with symptom tracking during routine care) by a physician outside the team of investigators. Key exclusion criteria were a history of diabetes (glycated hemoglobin 6.5% [48mmol/mol]) and medication interfering with glucose-insulin homeostasis (e.g. acarbose). Protocols were approved by the local Ethics Committee and registered on clinicaltrials.gov (NCT03609632 and NCT04330196). All participants provided written informed consent.

Provocative test, blood sampling and laboratory methods
After an overnight fast and 24 hours avoidance of physical activity, alcohol, and caffeine intake, participants underwent an oral glucose tolerance test (75g of glucose in 200ml water ingested within 5 min in an upright position). During the 48 hours before the OGTT (oral glucose tolerance test), participants were instructed to adhere to a weight-maintaining diet (calculated using their estimated REE [resting energy expenditure] multiplied by a physical activity level of 1.3) and refrained from alcohol, caffeine, and physical activity. During this period, participants were fitted with a continuous glucose monitor and were instructed to correct sensor glucose values of less than 54 mg/dL. A peripheral intravenous cannula was inserted for regular blood sampling. Plasma glucose was determined immediately after sampling using the Biosen C-Line glucose analyzer (EKF-diagnostic, Barleben, Germany). Blood samples were kept on ice until centrifugation and plasma was stored at -80 C until analysis. Commercial immunoassays were used to quantify insulin (Elecsys Insulin, Cobas, Roche Diagnostics, Mannheim, Germany), C-peptide

Stratification based on postprandial nadir plasma glucose
Based on plasma glucose nadir during the OGTT, participants were divided into the following 2 groups: level 1 (L1: ,70mg/dL and 50mg/dL) and level 2 (L2: ,50mg/dL) PBH, in line with recently published international consensus guidelines [6].

Calculation of indices of glucose-insulin homeostasis
Insulin exposure was calculated as the area over the baseline insulin concentration curve (iAUC insulin). Indices of glucose-insulin homeostasis were calculated using a modified version of the Oral Minimal Model method [7].
In particular, the oral model of GLP-1 action [8] coupled with the Oral C-peptide Minimal model [9] were used to assess the overall effect of glucose and GLP-1 on b-cell responsivity (F), as well as the b-cell responsivity to glucose alone (F Glu ) and the b-cell sensitivity to GLP-1 (p). GLP-1 exposure (L), i.e. the L-cell responsivity to glucose in the gut, was estimated by the area under the above basal GLP-1 concentration. Finally, the overall effect of GLP-1 on insulin secretion was estimated by the product of b-cell sensitivity to GLP-1 and GLP-1 exposure (PI 5 p *L).
In addition, combining the Oral C-peptide Minimal model [9] with a model of insulin kinetics [7], first-pass hepatic insulin extraction (HE) was estimated, while assuming post-hepatic insulin clearance determined from anthropometric characteristics [10]. C-peptide kinetics (required for the Oral C-peptide Minimal model identification) were estimated by exploiting a recently proposed methodology in this population [9].
To complete the picture of glucose-insulin control, the Oral Glucose-Minimal Model [7] was used to estimate insulin sensitivity (SI) and the rate of gastro-intestinal glucose absorption (Ra) from postprandial glucose and insulin data. Of note, data was right censored following hypoglycemia (i.e. data following hypoglycaemia were not considered to fit the model). An overview of the estimated indices, together with their physiological meaning is provided in Table 1.

Statistical analysis
Indices were compared between the groups using the Wilcoxon rank-sum test. Results are reported as median (interquartile range) unless otherwise specified. P values ,.05 were considered as statistically significant. Statistical analyses were performed with R (version 4.0.2).

Population
The L2 and L1-PBH groups comprised 11 and 14 participants, respectively. Participants (19 females and 6 males) were aged 44.0 years (32.2; 47.7), had a current body mass index (BMI) of 27.7kg/m 2 (23.4; 32.3), and a presurgery BMI of 41.1 kg/m 2 (39.6; 43.8). Median duration since surgery was 6 years (5; 7). No significant differences in participant characteristics were observed between the 2 groups. Further, sociodemographic and metabolic characteristics are reported in the Supplemental Material. One patient was treated with acarbose at the time of study inclusion (the drug was stopped 5 days before the experimental visit as defined in the study protocol).

Glucose, insulin and C-peptide profiles
Nadir glucose was 42.7mg/dL (36.6; 47.5) and 57.4mg/ dL (52.8; 63.4) in the L2-and L1-PBH group, respectively. Peak glucose levels and glucose exposure (iAUC Glucose ) did P values were computed using the Wilcoxon rank sum test. Differences represents the Hodges-Lehman estimator. * n 5 10, One subject was excluded due to poor model fit. y n 5 13, One subject was excluded due to absence of GLP-1 data.
not significantly differ between the 2 groups, nor did timeto-peak and time-to-nadir glucose. Median insulin exposure (iAUC Insulin ) was 138% higher in the L2-PBH group versus L1-PBH group (P , .001). Similarly, the L2-PBH group showed higher peak insulin levels than the L1-PBH group (P 5 .003) while time-to-peak insulin concentration was not significantly different between the groups (Table 2). Mean time course of plasma glucose, insulin and C-peptide concentrations for the 2 groups are shown in Figure 1.

Indices of glucose-insulin homeostasis
Results of indices of glucose-insulin homeostasis for both groups and between-group differences are provided below. The exact numbers are reported in Table 1.

b-cell function
Both indices of b-cell function (b-cell responsivity to glucose alone [F Glu ] and b-cell responsivity to glucose potentiated by GLP-1 [F]) were higher in the L2-PBH group versus L1-PBH group (P 5 .008 and P 5 .003, respectively).

Effect of GLP-1
GLP-1 exposure [L] was higher in L2-PBH compared to L1-PBH (P 5 .021). Differences in b-cell sensitivity to GLP-1 (p) and in the effect of GLP-1 on insulin secretion (PI) were not significantly different between the 2 groups (P 5 .111 and P 5 .204, respectively).

Discussion
In the present work, we compared indices of glucoseinsulin homeostasis between RYGB individuals with mild versus moderate PBH according to nadir plasma glucose following an oral glucose load. Based on the models applied in the present study, the higher insulin exposure in patients with lower glucose nadir appears to be driven by an increased b-cell function [F] and is possibly further compounded by a reduced hepatic insulin extraction (although the difference was not statistically significant in the present study). Patients with lower nadir plasma glucose also demonstrated higher GLP-1 exposure [L], an observation that is in line with several previous studies [11,12]. Greater GLP-1 exposure is most likely caused by an accelerated nutrient flux and digestion [13,14] and/or alterations in enteroendocrine cells and gut microbiome [15] in more severely affected PBH patients. Despite the higher GLP-1 exposure, we found no evidence for an increased incretin effect ([PI] or the effect of GLP-1 on insulin secretion) using our modeling approach. Lack of the stimulatory effect of GLP-1 must be seen in the context of possible reduced b-cell sensitivity to GLP-1 [p] (albeit lower, no statistical significance was found in the present work). Although the explanation for these findings remains speculative, GLP-1 receptor desensitization in an attempt to prevent hypoglycemia appears to be a possible mechanism and has been previously reported, although in slightly different settings [16]. Of note, 2 previous studies explored the effect of GLP-1 on postprandial insulin secretion using GLP-1 receptor blockade with exendin-9-39 (Ex-9) and found conflicting results [5,17]. Following the ingestion of a mixed-meal with clamped hyperglycemic plasma glucose levels, reduction in insulin secretion with Ex-9 versus saline infusion was comparable between PBH and asymptomatic patients following RYGB. Conversely, in a second study which involved Ex-9 in a mixed-meal test setting without a concomitant hyperglycemic clamp, the reduction in postprandial insulin secretion was significantly higher in PBH compared to asymptomatic patients, indicating a role for a heightened GLP-1 effect in the pathogenesis of PBH [5]. Taken together, the heterogeneity in study findings indicates that the role of GLP-1 in the pathogenesis of PBH remains a matter of debate.
In contrast to the responsiveness to GLP-1, responsivity to glucose alone (F Glu ) was higher in patients with more pronounced PBH. Higher beta-cell responsivity to glucose may reconcile with previously reported postpancreatectomy findings showing diffuse islet hyperplasia and expansion of beta cell mass in patients with severe PBH [18]. However, subsequent histological examinations were unable to confirm increases in beta-cell mass or formation in pancreatic specimens from PBH [19]. Thus, increased beta-cell responsivity may be rather explained by alterations of beta-cell function (e.g. insulin secretory rate per cell) rather than total cell mass, but more research is needed to make reliable statements regarding intrinsic regulator of beta-cells in PBH.
Finally, our observations do not support the notion that lower nadir glucose plasma is linked with higher insulin sensitivity [SI].
The strengths of the present work lie in the standardized setting on well-characterized individuals (e.g. stratification of participants based on objective criteria). We acknowledge the following limitations: our exploration of underlying mechanisms focused on specific aspects of glucose-insulin homeostasis and other contributors to insulin release (e.g. other peptide hormones or bile acid) were not addressed. We want to emphasize that the selected features are unlikely to cover the entire spectrum of underlying pathophysiological mechanismsor, nor was the study designed to uncover all of these. The applied mathematical representation of the glucose-homeostasis has well-recognized drawbacks such as fixing post-hepatic insulin clearance to population value and utilization of a method to estimate C-peptide kinetics that was only validated in silico [9]. Furthermore, due to the small sample size and the explorative nature of the study, the findings should not be overstated.

Conclusions
In conclusion, our results suggest that PBH patients with lower postprandial nadir glucose values relate to higher insulin exposure mainly caused by a glucose-, rather than GLP-1 stimulated heightened insulin secretion. These findings further deepen our understanding of the mechanisms behind the heterogeneity of PBH and may inspire future targeted therapeutic approaches.

Data Availability
All datasets analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.