Study participants

A total of 24 non-Hispanic White participants with a history of type 2 diabetes who underwent RYGB 1–4 years before enrolment completed the assessments. During the OGTT, 12 (50%) participants had PPHG between 150 and 255 min (PPHG group), while 12 (50%) participants did not experience PPHG (control group). In the PPHG group, 11 participants had a plasma glucose nadir of <3.0 mmol/l with no or mild hypoglycaemia symptoms (EHSS ≤20 points, ≤8 point increase from baseline), while one had a glucose nadir of 3.2 mmol/l with more severe hypoglycaemia symptoms (EHSS 40 points, 29 point increase from baseline). PPHG participants were more frequently women and had better glucose tolerance than control participants, while the two groups were similar in terms of age, BMI, fat distribution, post-surgery weight loss, lifestyle habits, diabetes duration, BP, lipid profile, kidney function and liver enzyme levels (Table 1).

Table 1 Clinical and metabolic characteristics of study participants Glucose profile, glucose metabolic fluxes and hormone profiles

Fasting and post-OGTT plasma glucose concentrations were lower in the PPHG group than the control group (Fig. 2a, b). Whole-body GCl was similar between groups at fasting, but consistently higher (mean difference ± SE in AUC0–180 min +261±72 ml min−1 kg−1 × min) throughout the OGTT in the PPHG group (Fig. 2c, d). EGP was slightly less suppressed during the first hour of the OGTT in the PPHG group, but it did not increase during hypoglycaemia compared with the control group (Fig. 2e, f). The RaO was similar between groups in both the early and late phases of the OGTT (Fig. 2g, h).

Fig. 2

Plasma glucose, glucose metabolic fluxes and hormone profiles. Profiles and post-OGTT AUCs of plasma glucose (a, b); tracer-derived GCl (c, d), EGP (e, f) and RaO (g, h); and plasma insulin (i, j), C-peptide (k, l), GLP-1 (m, n) and GIP (o, p) in individuals with a history of type 2 diabetes with PPHG or without (control) after RYGB during a 75 g OGTT. Glucose tracer data are available for 23 of 24 participants. Data are mean ± SEM indicated by error bars or shaded areas. In (a), (c), (e), (g), (i), (k), (m) and (o), repeated measures were tested by two-way ANOVA; the effects of group (G) and group by time interaction (G×T) and statistically significant p values in pairwise comparisons (*p<0.05) are reported. In the other panels, group differences were tested using the Mann–Whitney test

Fasting plasma insulin levels were similar between groups; however, the glucose-stimulated insulin response was significantly higher in the PPHG group during the first 2 h of the OGTT (Fig. 2i, j). Fasting and post-load plasma C-peptide, GLP-1 and GIP concentrations were similar between groups (Fig. 2k–p).

Beta cell secretory function

Despite the different glucose levels achieved between the PPHG and control groups, the glucose-stimulated ISR was virtually identical between the two groups over the first 2 h of the OGTT, whereas it returned to fasting values appropriately faster in the PPHG group in the second half of the OGTT (Fig. 3a–c). Consistently, the dose–response curve between the observed plasma glucose level and resulting ISR demonstrated enhanced beta cell function in the PPHG group (interaction effect: p<0.0001) (Fig. 3d), which was not fully captured by any single model-derived parameter of beta cell function (Fig. 3e–h).

Fig. 3

Beta cell function, insulin clearance and insulin sensitivity. Profile of the ISR estimated from C-peptide deconvolution (a); fasting and post-glucose AUCs of the ISR (b, c); relationship between the ISR and plasma glucose concentration (d); model-derived variables of beta cell function, including β-GS (e), ISR at 5 mmol/l glucose (ISR@5) (f), the potentiation factor ratio (g) and β-RS (h); fasting and total endogenous insulin clearance (i, j); and tracer-derived indexes of whole-body insulin sensitivity (k, l) and hepatic insulin resistance (m) in individuals with a history of type 2 diabetes with PPHG or without (control) after RYGB during a 75 g OGTT. Data are mean ± SEM indicated by shaded areas or error bars. In (a), (c) and (k), repeated measures were tested by two-way ANOVA; the effects of group (G) and group by time (G×T) or group by glucose (G×Gluc) interactions are reported. In the other panels, group differences were tested using the Mann–Whitney test

Insulin clearance

Insulin clearance was similar between groups during fasting (Fig. 3i) but markedly lower in the PPHG group over the OGTT (Fig. 3j), contributing to the relative glucose-stimulated hyperinsulinaemia in PPHG participants.

Insulin sensitivity

Tracer-derived whole-body insulin sensitivity (GCl/Ins) was similar between groups at fasting and immediately after glucose ingestion (Fig. 3k); however, it was significantly higher in the PPHG group in the later hours of the OGTT (Fig. 3l). The OGIS index calculated over the first 120 min of the OGTT was similar between groups (mean ± SEM: PPHG 417±22, control 401±56; p=0.713), as was hepatic insulin resistance (EGP×ISR) (Fig. 3m).

Counterregulatory hormones and lipolysis

Plasma glucagon levels were consistently lower in the PPHG group than in the control group during the OGTT (group effect: p=0.0006) and were unaffected by acute hypoglycaemia (Fig. 4a, b). There were no group differences in adrenaline, noradrenaline and cortisol levels at fasting or during the OGTT (Fig. 4c–e). NEFA levels were similar between groups at fasting and were equally suppressed in both groups after the oral glucose load (Fig. 4f, g).

Fig. 4

Counterregulatory hormone responses and lipolytic effects. Plasma profiles and post-OGTT AUCs of glucagon (a, b); plasma concentrations at fasting, glucose peak and glucose nadir of adrenaline (c), noradrenaline (d) and cortisol (e); and plasma profiles and AUC of NEFAs (f, g) in individuals with a history of type 2 diabetes with PPHG or without (control) after RYGB during a 75 g OGTT. Data are mean ± SEM indicated by error bars. In panels (a), (c), (d), (e), and (f), repeated measures were tested by two-way ANOVA; the effects of group (G) and group by time interaction (G×T) are reported. In (b) and (g), group differences were tested using the Mann–Whitney test

Pathogenetic mechanisms of PPHG

In bivariate regression analysis, the post-load glucose nadir was strongly associated with GCl and EGP (Fig. 5a, b), but not with RaO (r=−0.227, p=0.297). Logistic multivariable regression analysis adjusted for fasting plasma glucose and sex, which differed between groups, confirmed an independent effect of GCl on the glucose nadir, but not of EGP (Fig. 5c). In turn, GCl was modulated by both plasma insulin levels (insulin AUC0–120 min) and whole-body insulin sensitivity (OGIS index) to a similar extent (Fig. 5d). We observed no significant relationships between EGP and counterregulatory glucagon responses, insulin sensitivity (OGIS index) or AUCs of insulin, ISR, GLP-1, GIP or NEFAs (p>0.200 in all analyses). Moreover, the glucagon AUC was not related to the glucose nadir (r=0.036, p=0.867).

Fig. 5

Pathophysiological determinants of glucose nadir and glucose clearance. Linear correlations of post-load glucose nadir with the AUCs of tracer-derived GCl (a) and EGP (b), and multivariable analyses showing the independent effects of distinct glucose metabolic fluxes on glucose nadir (c) and of plasma insulin levels (insulin AUC0–120 min) and whole-body insulin sensitivity (OGIS index) on GCl AUC0–180 min (d) in individuals with a history of type 2 diabetes with PPHG or without (control) after RYGB during a 75 g OGTT. Best-fit lines (continuous), 95% CIs (dotted), Spearman correlation coefficients (r) and p values are shown in (a) and (b). Shaded areas in (a) and (b) indicate glucose nadir values below 3.0 mmol/l

Group differences were attenuated after excluding the five participants in the control group with persistence or relapse of type 2 diabetes (ESM Fig. 2). However, PPHG participants still had a significantly higher GCl AUC0–180 min than control participants (p=0.036), which was confirmed as the main determinant of glucose nadir among glucose metabolic fluxes (ESM Fig. 2b–d).

Hypoglycaemia awareness

A documented history of hypoglycaemia after RYGB was reported by eight (33.3%) participants, including six (50.0%) in the PPHG group and two (16.7%, p=0.193) in the control group. Among them, impaired hypoglycaemia awareness was identified in three (50.0%) or four (66.7%) of the participants in the PPHG group using the Gold or Clarke hypoglycaemia questionnaires, respectively, while all control participants had preserved hypoglycaemia awareness. Symptoms of PPHG were similar between groups at fasting and increased equally during the OGTT (Table 2). At fasting, VAS ratings for hunger, desire to eat, fullness and prospective food consumption were similar between groups (Table 2). At glucose nadir, hunger and desire to eat increased to the same extent in both groups, while fullness decreased numerically (p=0.103) and prospective food consumption increased numerically (p=0.080) to a similar extent between groups.

Table 2 Symptoms and signs of PPHG from fasting to glucose nadir ECG and haemodynamic responses

At fasting, the morphology and duration of ECG waves, intervals and rhythm were within the normal range and similar between groups (Table 2). At glucose nadir, there were no significant alterations in static or dynamic ECG parameters suggestive of counterregulatory sympathetic activation. The two groups experienced a similar reduction in systolic BP at glucose nadir, whereas diastolic BP did not change (Table 2).

Psychometric tests

At fasting, the two groups showed similar accuracy and reaction times on the Sternberg memory tasks and the four-choice reaction time test (Table 2). At glucose nadir, there was a similar accuracy reduction in the Sternberg memory test between groups, with no significant changes in reaction time, while results in the four-choice test did not change.

Free-living correlates of PPHG

On 14 day CGM under free-living conditions, the PPHG group spent on average 9% more time below the range of glucose normality (3.9–10.0 mmol/l), experienced more hypoglycaemic events (<3.9 mmol/l) and had lower average glucose values than the control group (Table 3). After exclusion of participants with current type 2 diabetes, the time below range of the control group increased slightly (6% [9%]) while remaining numerically lower than in the PPHG group (p=0.057). Analyses of food diaries showed similar energy and macronutrient intake between groups (Table 3). However, the habitual number of meals was significantly higher in PPHG participants, who also reported a numerically longer daily eating window (p=0.057).

Table 3 Free-living glycaemic management and dietary habits