Introduction
Obesity is a major risk factor for development of type 2 diabetes mellitus (T2DM)—and a key target in its treatment. Bariatric surgery, and in particular Roux-en-Y gastric bypass (LRYGB), induces marked and well-characterized effects on glycemic control that are superior to best medical treatment [
1]. Bariatric surgery has therefore recently been recommended as part of the diabetes treatment, extending to patients with T2DM with body mass index (BMI) below 35 kg/m
2 [
2,
3]. The observed effects on glucose control are rapid and obviously in part non-weight dependent [
4].
A non-weight-dependent mechanism after surgery involving bypass of the small intestine is a rapid and pronounced postprandial increase of incretin hormones, e.g., glucagon-like peptide (GLP)-1 [
5,
6]. Despite its “simplistic” nature without intestinal bypass and despite incomplete elucidation of the underlying mechanisms, laparoscopic sleeve gastrectomy (LSG) seems to have good effects on glucose control as well, at least in the middle–long term. Some data indicate that gastric emptying is not delayed after LSG, suggesting that the operation does not exert its effects solely through “restriction.” [
7‐
9] Supporting this idea is the observation in pregnant rat dams showing that they can increase food intake substantially even after LSG surgery [
10]. Despite the lack of intestinal bypass, it seems that release of several incretins, such as GLP-1 and peptide YY (PYY), is increased in a similar way after LSG as after LRYGB, at least in non-diabetic patients [
11]. A potential mechanism could be, in line with the situation after LRYGB, that, due to the reduced reservoir capacity of the stomach, undigested food enters the small intestine more rapidly, which in turn increases stimulation of the hormone-producing entero-endocrine cells of the intestinal mucosa.
The aim of this study was to compare and characterize, in detail, the changes in glycemic control in patients with T2DM and obesity who undergo LRYGB or LSG. In order to differentiate between weight- and non-weight-dependent effects, measurements were performed at various times after surgery.
Methods
Patients
Thirty-four subsequent patients with T2DM who fulfilled the inclusion criteria were asked to participate and were included from the waiting list for bariatric surgery at two sites in Sweden, the Sahlgrenska University Hospital, Gothenburg, and Ersta Hospital, Stockholm. The inclusion criteria were BMI 35–50 kg/m
2, age 18–60 years, and T2DM requiring any available diabetes medications, but not only dietary regimen. Exclusion criteria were inability to understand the Swedish language or to adhere to study instructions. At Sahlgrenska Hospital, all patients were randomized to either method (seven LRYGB, five LSG). At Ersta Hospital (12 LRYGB, 10 LSG), the patients were designated for either LRYGB or LSG depending on the patients’ preference and the surgeons’ judgment, e.g., severe reflux disease was considered a contraindication for LSG. There was no systematic bias for the choice of one or the other method. Patients were asked to daily record their fasting (FBG) and 90 min postprandial glucose levels (PPBG; after individual breakfast), as well as all use of diabetes medications during 2 weeks before and 2 weeks after surgery. Postoperative adjustments of diabetes medications were handled by the physician usually in charge of the patient’s diabetes treatment. Weight, waist circumference, and height were recorded with patients wearing only underwear after an overnight fast. The usual 2-week preoperative low-energy diet (LED) was omitted in order not to influence preoperative or early postoperative glycemic control [
12]. Eighteen patients underwent LRYGB and 15 underwent LSG. One patient who was not operated on by the intended technique at Sahlgrenska was excluded. One patient was lost to follow-up after the 3-week follow-up. The total follow-up rate was 94% (see Supplementary Fig.
1 for flow chart of inclusion). Data for GLP-1 and GIP were lost for four patients in the LSG group at 12 months; therefore, data are shown for 11 LSG patients in Figs.
6d and
7d. All patients were of Caucasian origin. The study was registered at
ClinicalTrials.gov identifier NCT01984762.
Study Ethics
This study was approved by the Regional Ethical Review Board of Gothenburg (study reference number 016-12), and the study was conducted according to the principles of the Helsinki declaration. All patients gave written informed consent.
Modified Oral Glucose Tolerance Test (MOGTT)
MOGTT with 30 g of glucose (compared to standard OGTT 75 g) dissolved in 150 ml water was performed at four occasions: 3 weeks preoperatively and 2 days, 3 weeks, and 12 months postoperatively. The glucose dose was reduced in order to, as far as possible, avoid side effects such as vomiting and dumping in the postoperative phase. MOGTT was performed after a 12-h fast. Patients did not take any antidiabetic medications, and no long-, intermediate-, or fast-acting insulin was administered on the morning of the test. Glucose, insulin, glucagon-like peptide (GLP)-1, and gastric inhibitory peptide (GIP) levels were measured at 0, 15, 30, 60, 90, 120, 150, and 180 min after intake of the oral glucose dose. In case the glucose level declined to basal level at any time before 180 min, the blood sampling was terminated at that time.
Surgery
All patients underwent a laparoscopic LRYGB or LSG by experienced bariatric surgeons. LRYGB was performed by a five-port technique with the use of a linear stapler for both the gastrojejunostomy and the jejuno-jejunostomy. A running suture was used for closure of the remaining defects as previously described [
13]. The length of the Roux limb (alimentary limb) was typically 120 cm and the biliopancreatic limb 50 cm. For LSG, a linear stapler was used for resection of the stomach over a 35–36 F bougie from 3 to 5 cm proximal to the pylorus to 1–2 cm distal to the angle of His. Postoperative dietary recommendations included intake of fluids and semi-solid food for 2 weeks, and solid food thereafter. Patients were instructed to adhere to a diet rich in protein and with a calorie content of approximately 800–1000 kcal for the first six postoperative weeks.
Glucose, Insulin, GLP-1, and GIP Measurements
At the Sahlgrenska University Hospital, blood glucose was measured using StatStrip according to the manufacturer’s instructions (Nova Biomedical, Waltham, MA, USA). At Ersta Hospital, blood glucose was measured using the YSI Model 2300 Stat Plus glucose analyzer according to the manufacturer’s instructions (Yellow Springs Instruments, OH, USA). Insulin was measured by an electrochemiluminescence immunoassay “ECLIA” using a Cobas e immunoassay analyzer according to the manufacturer’s instructions (Roche Diagnostics, Dublin, Ireland). The intra-assay variation of the insulin measurement was 1.1–1.4% (CV) and the inter-assay variation 3.5–3.7% (CV). The cross-reactivity with IGF-1 was 0.04%. GLP-1 and GIP were measured using ELISAs according to the manufacturer’s instructions (Merck Millipore, Solna, Sweden; Human total GIP ELISA, product number EZHGIP-54K, and Multi Species GLP-1 Total ELISA, product number EZGLP1T-36K). The intra-assay variation for GLP-1 was 1–2% (CV) and the inter-assay variation < 12% (CV). For GIP, the intra-assay variation was 3–8.8% (CV) and the inter-assay variation 1.8–6.1% (CV).
Calculation of HOMA-IR and HOMA-B (%)
HOMA-IR, an index of insulin resistance, was calculated based on the morning fasting plasma insulin and blood glucose levels using the formula: HOMA-IR = (insulin (μIU/ml) × FBG (mmol/l))/ 22.5. HOMA-B (%) is a measure of pancreatic β-cell activity and was calculated using the formula: ((20 × insulin (μIU/ml))/(FBG (mmol/l)) – 3.5)% [
14]. It should be noted that the assessment of the dynamics of beta cell functioning in the stability of a fasting state and using a single mathematical model such as HOMA-B is less reliable than the assessment of a relatively stable factor such as insulin resistance [
15].
Statistics
All statistics were performed using Prism 5 and 7 (version 5.0a and 7.0a). Logarithmic transformation was performed where indicated in order to obtain equal variances. Area under the curve for glucose, insulin, GLP-1, and GIP during the oral glucose tolerance tests was calculated using the trapezoidal rule. One-way ANOVA with Dunnett’s post hoc test was used for analyzing changes from baseline to the different postoperative times within groups. Paired and unpaired Student’s t test was used for single comparisons between baseline and postoperative values and between the groups, respectively. χ
2 test was used to compare numbers of patients on/off diabetes medications 12 months after surgery. A P value of < 0.05 was considered statistically significant.
Discussion
Sleeve gastrectomy (LSG) is currently one of the most common bariatric operations worldwide, although the long-term effects on body weight and diabetes have not yet been characterized in detail. A number of studies focusing on long-term effects are ongoing [
16‐
19], but very few data on the very early weight-independent effects of LSG have been presented in patients with obesity and T2DM. Therefore, we performed a comparative study between LSG and LRYGB, to examine the early postoperative effects on glycemic control, supplemented by the effects after 1 year.
The most important finding of the present study in patients with T2DM was that the early effect of SG on glucose control and insulin secretion was equal compared to that of RYGB. Although based on a relatively limited number of patients, we found that RYGB elicited a statistically significant increase in the GLP-1 response at 3 weeks. However, this did not translate into a greater improvement of glucose metabolism variables in the RYGB group at that moment compared to the LSG group. This is well in line with experimental data from animal studies, showing that even complete absence of GLP-1 does not change the effect of LSG on weight decrease and glucose metabolism [
20]. In relation to LRYGB, it has been speculated that decreased release of a hitherto uncharacterized anti-incretin or “decretin” factor from the bypassed foregut, which would suppress insulin release, could be of importance [
21]. If this factor is of importance after LSG as well, it would most probably be released from the area along the major curvature of the stomach, which is removed, rather than from the duodenum since this is not bypassed in LSG. Moreover, this would also fit with the rapid recurrence of diabetes in RYGB patients or diabetic rats where food is reintroduced to the remnant stomach via a gastrostomy cannula [
22,
23]. On the other hand, many other factors have been suggested to influence the metabolic response after bariatric surgery, e.g., bile acids and changes of the microbiota [
24]. Our study did not focus on the latter aspects, and as such, a contribution of these parameters cannot be excluded nor proven. Moreover, calorie restriction is known to be a potent insulin sensitizer [
25]. The improvement in glucose metabolism, observed in both groups, could be influenced by the preoperative fasting and postoperatively decreased caloric intake. The authors assume caloric intake was similarly decreased within the first days and weeks after SG and RYGB despite the fact that the two procedures are anatomically completely distinct.
In corroboration with previous studies [
17,
18,
26,
27], the weight loss was statistically significantly lower after LSG compared to LRYGB 1 year after surgery. Despite this and despite the subtle differences in the insulin and GLP-1 curves, mentioned above, these differences did not translate into differences in clinically relevant diabetes measures or diabetes medication use between the groups at any time during the first postoperative year. The fact that both groups lost more than 20% of their initial body weight after 1 year corresponds to the observed improvement in glucose metabolism. These improvements can be equally important when the weight loss has been achieved by non-dietary measures, as shown in the Look AHEAD study [
28].
Postoperative vomiting and nausea (PONV) is more commonly reported after LSG than after LRYGB. Accordingly, three of the LSG patients were not able to ingest the entire 30 g glucose dose at the first 2-day MOGTT. This did, however, not influence the data presented in any significant way since the results were the same irrespective if these patients were included or not. On post-op day 2, HOMA-IR was significantly reduced in both groups, whereas fP-insulin and HOMA-B (%) were significantly reduced only in the LRYGB patients. On the other hand, the peak AUC 0–60 min for insulin was increased only in the LSG patients at this time. Although we cannot exclude that the lack of differences between groups was due to a type II error, these findings might suggest that the anti-diabetes effects of these two operations are distinct, with a somewhat more pronounced effect by LRYGB on insulin sensitivity, and by LSG on β-cell activity.
The measurements at 1 year postoperatively were performed in order to evaluate the combined effect of weight-dependent and non-weight-dependent mechanisms. There were no significant differences in glycemic control between LSG and LRYGB at 1 year, as assessed by HbA1c levels, MOGTT measurements, the patients self-reported FBG and PPBG, or the numbers of patients off anti-diabetes medications. The exception being HOMA-B (%), that still was significantly decreased after LRYGB but not LSG. The effect of LSG, with removal of a major part of the stomach, could be related to the decreased reservoir capacity and the transfer of undigested food to the small intestine where the release of hindgut factors, e.g., GLP-1 and GIP, is stimulated and could exert beneficial effects on insulin and glucagon release. The levels of incretin hormones, in response to a somewhat higher oral glucose load, were previously shown to be relatively similar in
non-diabetic RYGB and SG patients [
11]. In the present study, however, the GLP-1 stimulatory effect of LSG was transient and started to subside already 3 weeks after surgery. Whether this reflects a true difference in GLP-1 response after LSG compared to LRYGB in patients with or without diabetes is not entirely clear. An alternative explanation is that the differences in GLP-1 response are associated with the use of a lower energy load in our study (30 g glucose vs. a liquid test meal containing 15 g carbohydrates, 25 g protein, and 28 g fat).
Limitations of this study include the relatively small numbers of patients, which obviously is associated with a risk that we were unable to statistically demonstrate some additional true differences between the groups. Moreover, all patients were not randomized. Anthropometric data were not different in any aspect between the groups at baseline. To assess glucose metabolism, ingestion of a low-dose glucose drink was used. This choice was made to avoid vomiting and dumping, despite the fact that this low-dose drink has not been extensively validated. A similar load of glucose has, however, previously been shown to induce a satisfactory response in intestinal hormones after bariatric surgery [
29]. Moreover, using a glucose drink as opposed to a mixed meal test limits the interpretability as human data indicate proteins are also able to elicit a dose-dependent incretin effect [
30]. Detailed food diaries, in particular during the first 3 weeks after surgery, were not collected. Therefore, the authors cannot be certain that patients consumed the recommended post-op diet. Recent studies suggest that differences in glycemic control between the groups may appear with longer follow-up time [
19].
Strengths of this study were the careful monitoring of glucose metabolism at several times both very early and up to a year after surgery by MOGTT, and the “real-life” self-monitoring data from patients during 2 weeks before and 2 weeks after surgery on glucose levels at fasting and postprandially after breakfast.
In conclusion, in this study, LSG exerted early beneficial effects on glycemia that were nearly indistinguishable from those after LRYGB. GLP-1 secretion in response to an oral glucose load was only transiently increased after LSG, in contrast to LRYGB where GLP-1 peak levels were still increased at 1 year postoperatively. The BMI loss after 1 year was lower after LSG than LRYGB. Elucidation of other mechanisms that contribute to the positive glycemic effects of LSG may hold the key to understanding how patients can be helped to maintain long-term glycemic control after bariatric surgery.