Introduction

Recent years have seen a marked increase in the number of laparoscopic sleeve gastrectomy (LSG) operations performed for obesity and obesity-associated type 2 diabetes (T2DM) [1, 2]. Previous studies have shown that LSG produces weight loss and an improvement in metabolic outcomes comparable to biliopancreatic diversion (BPD) and Roux-en-Y gastric bypass (RYGB). A meta-analysis of 11 randomised clinical trials (RCTs), comprising of 765 participants, comparing at least 2 of 3 bariatric surgical procedures (LSG, RYGB and LAGB) concluded that RYGB and LSG results in similar weight loss, both of which were superior to LAGB [3]. Tsoli et al. demonstrated that LSG (n = 12) was comparable to BPD (n = 12) in % excess weight loss (%EWL), diabetes remission and enhanced postprandial glucagon-like peptide-1 (GLP-1) response at 1 year in patients with obesity-associated T2DM [4]. Similarly, Nosso et al. reported that LSG (n = 19) was comparable to RYGB (n = 14) in weight loss and diabetes remission rate, but the improvement in the postoperative postprandial GLP-1 response was more pronounced in the RYGB group at 1 year [5]. While the observations on the effect of LSG on GLP-1 response were consistent with previous studies, the data on GIP response remains inconclusive. Some studies have described no changes in GIP response [6] while others have reported a decrease [5]. It is important to note that in all the published studies relating to LSG, the follow-up period was up to 1 year. There is limited literature on the long-term effect of LSG on glucose homeostasis, GLP-1 and GIP response in individuals with T2DM or impaired glucose tolerance.

Previously, we have demonstrated that LSG was associated with an improvement in glucose homeostasis and an increased GLP-1 response at 1 and 6 months, and these results were comparable to the BPD group [6]. We hypothesise that these improvements would be preserved in the longer term (approximately 4 years). The primary aim of this current study was to examine static and dynamic changes in glucose, insulin, C-peptide, GLP-1 and GIP in participants with T2DM and impaired glucose regulation at a mean follow-up period of 4 years after LSG. The secondary aim was to examine long-term incretin hormone changes in participants who underwent BPD.

Methods

Study Participants

Approval for the study was obtained from the Local Research Ethics Committee (LREC reference 06/WMW02/7). Participants (n = 40) who took part in a previously published initial study [6] (pre-operative, 1-month and 6-month post-operative) were invited (by post, email or telephone call) to return at a mean follow-up of 4 years (range 2–7 years). There were 26 participants in the LSG group and 14 in the BPD group. Ten participants in the LSG group and 6 in the BPD group agreed to participate. Sixteen participants in LSG group and 2 in the BPD group were lost to follow-up or declined to participate in the follow-up study. Six participants were deceased in the BPD group. The study sample has been previously described [6, 7]. In brief, the inclusion criteria at the outset included both genders, age between 20 and 60 years and BMI ≥ 40 kg/m2. All participants had previously diagnosed T2DM or diagnosed during an oral glucose tolerance test (OGTT) at the start of the study or impaired glucose regulation according to the American Diabetes Association (ADA) criteria [8]. Buse’s consensus (2009) criteria was used to define diabetes remission [9]. Complete remission is defined as HbA1c < 6% and fasting plasma glucose (FPG) < 5.6 mmol/L, and off medication and partial remission is defined as HbA1c 6.0–6.4% and FPG 5.6–6.9 mmol/L and off medication for 1 year.

Study Design

The study was a non-randomised prospective study. All the participants underwent a standardised 75 g OGTT (122 mls of Polycal 61.9 g/100 ml of glucose, Nutricia Clinical Care, Trowbridge, UK) pre-operatively and post-operatively at 1 month, 6 months and 4 years. Pre-operative tests were performed 1 month prior to the surgery. All participants were asked to fast from the midnight before the test, and all diabetes-related medications were omitted for 24 h before OGTT. During the OGTT, blood samples were collected for measurements of glucose, insulin, C-peptide, GLP-1 and GIP at time 0, 15, 30, 45, 60 and 120 min. All samples were collected on ice, centrifuged within 1 h of collection and stored at − 80 °C until analysis. At the time of the pre-operative and post-operative OGTT, clinical and biochemical information were obtained. Clinical measurements included weight, height, body mass index (BMI), waist circumference and blood pressure. Biochemical data such as HbA1c and lipid profile were analysed within the local hospital accredited laboratory.

Measurement of Insulin, C-Peptide, Insulin Sensitivity and Beta Cell Function

Insulin was measured using an Invitron Insulin ELISA kit. The analytical sensitivity of the Invitron insulin assay was 0.02 mU/L and the dynamic range of this assay is 0.02–250 mU/L. The inter-assay coefficient of variation was ≤ 7.1. Cross reactivities (CR) of related proteins were as follows: CR 1.2% with intact proinsulin; and 0% with C-peptide. C-peptide was measured with an Invitron C-peptide kit. The analytical sensitivity of the Invitron C-peptide assay was 5.0 pmol/L. The dynamic range of this assay is 5.0–5000 pmol/L. There was 2% cross reactivity with intact proinsulin but no cross reactivity with insulin.

Insulin sensitivity and beta cell function were measured with homeostasis model assessment (HOMA) by using measurements of fasting glucose and C-peptide concentrations. These were calculated by using the Oxford University online calculator (https://www.dtu.ox.ac.uk/homacalculator/ accessed 01 June 2015). HOMA was first developed by Matthews and colleagues in 1985 [10] and updated to a HOMA2 computer model in 1996 [11]. HOMA provides three measures: HOMA-%B (estimated steady state beta cell function), HOMA-%S (insulin sensitivity) and HOMA-IR (insulin resistance). These measures have been validated and shown to correlate with clamp-derived studies [11].

Measurement of Total GLP-1 and Total GIP

Total GLP-1 was quantitatively measured using the EMD Millipore Total GLP-1 ELISA Kit. The antibody pair used in this assay measures GLP-1 (7-36) and (9-36) and has no significant cross-reactivity with GLP-2, GIP, glucagon and oxyntomodulin. The sensitivity of this assay was 1.5 pM and the approximate range of this assay was 4.1 to 1000 pM. The intra- and inter-assay coefficients of variation were ≤ 2% and ≤ 12% respectively. Total GIP was measured using the EMD Millipore Human GIP (total) ELISA Kit, which reacts fully with intact GIP (1-42) and the NH2-terminally truncated metabolite GIP (3-42). The assay does not significantly cross-react with glucagon, oxyntomodulin, GLP-1 and GLP-2. The sensitivity of this assay was 4.2 pg/mL (20 μL sample size). The appropriate range of this assay is 4.2 to 2000 pg/mL. The intra- and inter-assay coefficients of variation were ≤ 8.8% and ≤ 6.1% respectively.

Statistical Analysis

Statistical analysis was performed using SPSS (version 22, SPSS Inc., Chicago). The normality of data was assessed by the Shapiro-Wilk test. Continuous data with a normal distribution are presented as mean and standard deviation, and data that did not have a normal distribution are described as median and interquartile range. We used ANOVA with post-hoc analyses for data with a normal distribution and Friedman test for data without a normal distribution. In all cases, p < 0.05 is considered to be significant.

Results

Participants’ Characteristics

There were 10 participants (6 females, a mean age of 52 ± 6 years) in the LSG group and 6 (5 females, a mean age of 48 ± 7 years) in the BPD group with follow-up data available for analysis. The mean duration of follow-up for the LSG group was 4.0 ± 1.5 years and for the BPD group 7.2 ± 0.6 years. The baseline and follow-up characteristics of both groups are shown in Table 1. Eight participants had T2DM in the LSG group and all 6 participants in the BPD group. The median duration of T2DM was 19 months in the LSG group and 41 months in the BPD group.

Table 1 Weight, blood pressure and lipid profile pre-operatively and 1, 6 months and 4–7 years after LSG and BPD
Full size table

Changes in Weight, Blood Pressure and Lipid Profile

In both the LSG and BPD groups, an initial improvement in clinical measures of obesity at 1 and 6 months were maintained at long-term follow-up (Table 1). A reduction in both systolic and diastolic blood pressure was noted at 4 years in the LSG group. There was a significant increase in total cholesterol and LDL-cholesterol along with HDL-cholesterol following LSG whereas a reduction in total cholesterol, LDL-cholesterol and HDL-cholesterol was observed following BPD. However, triglyceride level in both groups did not show any significant changes.

Changes in Glucose-Insulin Homeostasis

LSG Group

FPG level was lower at 4 years compared to baseline but the result did not reach statistical significance. Two-hour plasma glucose (2 h PG) and HbA1c were significantly lowered at 4 years compared to baseline. Four participants (50%) (1 complete +3 partial) achieved remission of T2DM at 4 years. Preoperatively, two participants were treated with insulin and three treated with oral diabetes agents/ GLP-1 agonist. Postoperatively, one received insulin and two received oral diabetes agents. None of these participants were on long-acting oral diabetes agents. Fasting insulin, 2-h insulin, fasting C-peptide and 2-h C-peptide levels were lower at 4 years but did not reach statistical significance. A significant reduction in HOMA-IR and HOMA-%B and a significant increase in HOMA-%S were observed (Table 2).

Table 2 Static measurements of glucose-insulin homeostasis and incretin hormones in the LSG and BPD groups
Full size table

BPD Group

There was no change in FPG at 7 years. However, a reduction in 2-h PG and HbA1c were observed. Two participants (33%) achieved complete diabetes remission. Two out of the 6 participants were on oral agents postoperatively, whereas all received oral diabetes agents preoperatively. A numerical but non-significant reduction in fasting insulin, 2-h insulin and HOMA-%IR and a non-significant increase in HOMA-%B and HOMA-%S were observed at 7 years. There was no change in fasting C-peptide and 2-h C-peptide at 7 years (Table 2).

Dynamic Measurements of Glucose, Insulin and C-Peptide

LSG Group

At 4 years, there was a significant reduction in the mean plasma glucose at 120 min but not at 0, 15, 30, 45 and 60 min. The peak glucose response was shifted to the left (peak at 45 min at 4 years and at 60 min at baseline) (Fig. 1). The median AUC measurements for glucose did not show any significant changes compared to baseline (Table 3). With respect to dynamic insulin changes, there were no significant changes in the mean insulin at all 6 time points as well as the AUC measurements (AUC 0-60, AUC 0-120). The median AUC measurements for C-peptide also did not show any changes at 4 years from the baseline.

Fig. 1
figure 1

Changes in glucose, insulin, C-peptide, GLP-1 and GIP during the OGTT before and 1 month, 6 months and 4–7 years after LSG and BPD. Value represents mean ± standard error. GLP-1, glucagon-like peptide-1; GIP, glucose-dependent insulinotropic polypeptide; LSG, laparoscopic sleeve gastrectomy; BPD, biliopancreatic diversion

Full size image
Table 3 Dynamic measurements of glucose, insulin, C-peptide, GLP-1 and GIP within LSG and BPD groups
Full size table

BPD Group

At 7 years, there was a significant reduction in the mean plasma glucose at 120 min but not at the remaining time points. The peak glucose response was shifted to the left (peak at 30 min at 7 years and at 60 min at baseline) (Fig. 1). The median AUC measurements of glucose did not show any significant changes compared to baseline. There were also no changes in AUC measurements of insulin and C-peptide at 7 years. The BPD group had significant lower C-peptide AUC measurements compared to the LSG group pre-operatively, 1 and 6 months and 7 years post-operatively (Table 3).

Changes in Incretin Hormones

LSG Group

Fasting GLP-1 concentrations were not different at 4 years compared to baseline (Table 2). A significant increase in postprandial GLP-1 responses achieved at 1 and 6 months was not observed at 4 years. There were no changes in the GLP-1 AUC 0-60 and AUC 0-120 at 4 years post-operatively (Table 3). On the other hand, a non-significant increase in fasting GIP and postprandial GIP responses were observed at 4 years. There was a significant increase in the mean GIP concentrations at all postprandial sampling time points at 4 years, in contrast to 1 and 6 months (Fig. 1). In line with this, there were significant increase in the GIP AUC 0-60 and non-significant increase in the GIP AUC 0-120 at 4 years post-operatively.

BPD Group

There was a significant increase in fasting GLP-1 at 7 years but there was no change in fasting GIP (Table 2). A significant increase in postprandial GLP-1 response was observed at 7 years compared to baseline. However, there were no significant changes in GIP response (Table 3; Fig. 1).

Discussion

LSG has become a popular bariatric surgical procedure due to its comparable effectiveness and safety profile. In line with previous studies [3, 12,13,14,15], we observed significant improvements in markers of glycaemia as well as a significant weight loss at 4 years. There was a significant reduction in HbA1c and 2-h PG levels. One of the eight participants achieved complete diabetes remission. The complete remission rate in our study is comparable to other published studies using ADA criteria. Mingrone et al. reported 0% complete remission rate 5 years after RYGB (n = 19) and BPD (n = 19) [16]. Previous studies have shown that LSG has no significant impact on lowering total cholesterol and LDL-C but is associated with an increase in HDL-C [17,18,19]. Van Osdal et al. showed that the proportion of participants with a total cholesterol concentration [≥ 200 mg/dL (5.2 mmol/L)] and LDL-C concentration values [≥ 130 mg/dL (3.4 mmol/L)] was greater in the LSG group than the RYGB group (30% vs 10%, p < 0.001 for total cholesterol; 24% vs 4%, p,0.001 for LDL-C). They also observed that the LSG group had higher proportion of patients with the recommended HDL-C values than the RYGB group (57% vs 52%, p = 0.68) [19]. In our study, there was a significant increase in LDL-C and HDL-C at 4 years. Regarding triglyceride level, which is related to the degree of weight loss, we did not find any significant changes despite a significant weight loss at 4 years. Along with a significant weight loss, we observed a non- significant reduction in fasting C-peptide and fasting insulin. Of note, elevated levels of fasting insulin and C-peptide are a key feature of individuals with insulin resistance [11]. Previous studies have demonstrated that LSG is associated with a reduction in insulin resistance [20]. In line with this, we observed a significant improvement in insulin sensitivity (reduction in insulin resistance) at 1 and 6 months. At 4 years, insulin sensitivity (HOMA%-S) decreased compared to the 1- and 6-month’s values but was still greater than the baseline level. Similarly, insulin resistance as measured by HOMA-IR increased by 4 years but remained lower than the baseline value. These observations might be explained by weight regained at 4 years (mean BMI was 36.9 kg/m2 at 6 months and 39.1 kg/m2 at 4 years). With respect to the effect of bariatric surgery on beta cell function, previous studies have demonstrated an improvement in beta cell function in the early post-operative period [21] but limited improvement in the long-term [22]. In line with this, we observed a significant increase in HOMA-%B at 1 and 6 months, but a significant decrease at 4 years and a downward trend of HOMA-%B from 1-month post-operatively to 4-year post-operatively.

With respect to BPD, our observations of improvement in measures of obesity, glycaemic control, fasting GLP-1 and postprandial GLP-1 responses were in line with previous published literature. Since BPD is a malabsorptive procedure, it is associated with a significant improvement in lipid profile. In line with this, we observed a significant decrease in total cholesterol at 1 month, 6 months, 7 years and LDL-cholesterol at 1 month and 6 months.

While there is abundant evidence on the metabolic outcomes of LSG, there is limited literature available on the detailed examination of long-term effect of LSG on glucose metabolism and incretin hormone response. Most clinical studies have follow-up periods between 6 weeks and 12 months [4, 13, 23]. Our previous study described a significant improvement in static and dynamic measurements of glucose at 1 and 6 months in 22 participants with impaired glucose regulation [6]. In the current study, we observed a reduction in 2-h PG and HbA1c but no significant changes in glucose AUC measurements in both groups. The mean plasma glucose concentrations at all postprandial sampling time points except 120 min were not different at 4 years compared to baseline. The AUC measurements for glucose (0-60min) (0-120min) did not show any significant changes, suggesting that the glucose intolerance state of the participants at 4 years was similar to baseline. Time-related relapse of T2DM following bariatric surgery is inevitable [24]. This study’s observations raised the questions of whether the glucose AUC measurements are better than static glucose measurements to detect impaired glucose regulation? Of interest, published studies have suggested that the 2-h PG alone cannot predict the full picture of glucose intolerance and the glucose AUC measures are better than 2-h PG [25].

Both LSG and BPD are associated with an accelerated gastric emptying [26]. Studies have suggested that gastric emptying time has an impact on glucose homeostasis and is related to the peak circulation glucose contraction after an oral glucose load. The higher initial gastric emptying rate is positively correlated with the peak postprandial glucose concentration [27]. In line with this, we observed that the peak glucose response was shifted to the left in both LSG and BPD group. The time to reach peak glucose concentration after an oral glucose load was 60 min in both groups at baseline and 45 min in the LSG group and 30 min in the BPD group at long-term follow-up visit.

An enhanced GLP-1 response is thought to be one of the mechanisms responsible for the resolution of T2DM following bariatric surgery [28, 29]. Previous studies demonstrated that LSG results in an increased GLP-1 response and a T2DM complete remission rate comparable to BPD [4] and RYGB [12]. Tsoli et al. reported that both BPD and LSG are associated with an increase GLP-1 response at 1 month, 3 months and 12 months in 24 participants with T2DM [4]. Papamargaritis et al. reported that the postprandial GLP-1 responses were significantly increased at 6 weeks following LSG and remained at this level at 1 year [30]. In contrast to previous observations of studies with short-term follow-up, there were no significant changes in fasting and postprandial GLP-1 responses in the LSG group at 4 years. One of the possible explanations for our findings is a type 2 error. The other possible explanation is the use of an OGTT instead of a mixed meal tolerance test. The majority of previous published studies have measured a meal-stimulated GLP-1 [5, 12, 13], while others have examined glucose-stimulated GLP-1 [4, 6] Of interest, Nannipieri and colleagues reported an increased GLP-1 response at 15 days but an attenuated GLP-1 response at 1 year following RYGB (n = 23) and LSG (n = 12) in patients with T2DM [13]. This study used a mixed meal test. Our findings question the role of GLP-1 in the mechanism of long-term diabetes remission. Of note, Steven and colleagues reported that calorie restriction but not GLP-1 explains the acute improvement in glycaemic control following RYGB in subjects with T2DM [31].

With regard to GIP, an increase in fasting GIP as well as postprandial GIP response was observed in the LSG group at 4 years. However, there was no significant changes in GIP response at 1 month and 6 months. Previous studies have described conflicting observations on the GIP response following LSG. Some studies have described no changes in GIP response [6] while others have reported a decrease [5]. On note, these studies examining GIP response have follow-up period of up to 12 months only. The findings of our current study suggest that GIP might play a role in the long-term diabetes outcome.

Although GLP-1 and GIP are derived from the intestine, they have different sites of secretion within the gut: GLP-1 is mainly secreted from ileal L-cells, while GIP is secreted from the K-cells of the duodenum and jejunum. There might be differences in GLP-1 and GIP responses relating to changes in gut anatomy following different bariatric procedures [29]. Our study showed that the BPD group maintained the improvement in GLP-1 response at 7 years, while the LSG group showed no change at 4 years from baseline. Of importance, BPD involves considerable alteration in the anatomy of the gut, which may result in direct exposure of nutrients to the ileal L-cells and would therefore produce a greater postprandial GLP-1 response compared to LSG. On the other hand, LSG would produce a greater postprandial GIP response because of the expedited delivery of nutrient to the duodenum and the jejunum where K-cells are populated. In line with this, a greater GIP response was observed in the LSG group at 4 years.

To our knowledge, our study is the first to examine the glucose, incretin hormone response 4 years after LSG in participants with impaired glucose regulation. We describe the novel findings of no change in GLP-1 response but an exaggerated GIP response along with an improvement in glycaemic control at 4 years following LSG. These observations require replication in other bariatric cohorts with longer follow-up.

There are limitations to the current study. Firstly, our study had a small sample size, particularly for the BPD group. The second limitation was that there was a considerable drop-out rate at long-term follow-up with 22.5% (18/40) of participants declining to take part or being lost to follow-up, and 15% (6/40) were deceased. We have examined whether the pre-operative clinical and biochemical characteristics of the participants were related to their long-term follow-up attendance and survival status. There were no differences in baseline weight, blood pressure, HbA1c and lipid profile (except triglyceride level) amongst those who participated in the study, compared to those who were lost-to-follow-up and deceased. All 6 deceased participants underwent BPD, had a significant higher level of triglyceride and lower fasting C-peptide. The causes of death were infection (n = 3), suicide (n = 1) and unknown (n = 2). The third limitation was a variation in follow-up duration in two groups. The BPD group had longer follow-up than the LSG group. Despite this, we observed that the BPD group had a significant increase in GLP-1 level compared to baseline, whereas the LSG group had no change in GLP-1 level. The fourth limitation was the study design, which was a nonrandomised prospective study. At the time of initial study, participants who had BMI > 50 kg/m2 were usually offered BPD as per local protocol. Within the current study, we did not plan at the outset to measure glucagon and other gut hormones such as ghrelin, neuropeptide Y.

Conclusion

The current study supports the role of LSG in the management of obesity-associated T2DM in the long-term. Despite a non-significant weight regain, increase in HbA1c and a decline in HOMA measures compared to the 6-month values, the study observed that LSG was associated with a significant reduction in weight and glycaemic measures compared to baseline values. This work also provides new insight relating to the long-term effect of LSG on incretin hormone responses, which requires further evaluation.