Pancreatic Effects of Liraglutide or Sitagliptin in Overweight Patients With Type 2 Diabetes: A 12-Week Randomized, Placebo-Controlled Trial

Glucagon-like peptide 1 (GLP-1)–based drugs—that is, GLP-1 receptor agonists and dipeptidyl peptidase 4 (DPP-4) inhibitors—are increasingly prescribed for the management of type 2 diabetes (1). A decade after drug approval, however, the long-term pancreatic safety of these antihyperglycemic drugs is still not fully established (2).

Shortly after their introduction, sporadic cases of acute pancreatitis were described for GLP-1 receptor agonists (3), and a study using an adverse event reporting system found strong associations with acute pancreatitis and pancreatic cancer for both drug classes (4). These warnings led to numerous studies evaluating potential risks. Several administrative database studies demonstrated an association between GLP-1–based drugs and acute pancreatitis (2,5). Also, in experimental animal models, GLP-1–based therapies induced pancreatitis, ductal cell proliferation, and premalignant lesions (6). A recent meta-analysis of three large-scale, randomized, placebo-controlled trials with DPP-4 inhibitors demonstrated a modest, yet significant, increase in acute pancreatitis risk (risk ratio 1.8; 95% CI 1.1–2.8) (7). However, many other studies were neutral, including database studies, clinical trials, and animal studies (2,8). For example, the largest observational database study, including 12,868 patients, did not find an increased risk of pancreatitis (9). Prospective studies of pancreatic cancer are lacking, given the low prevalence and long exposure required to develop this condition. However, a recent case-control study did not find an increased pancreatic cancer risk compared with sulfonylurea derivatives (10).

Human mechanistic studies could support the debate on pancreatic safety, but, to date, the pancreatic effects of these drugs have been selectively investigated, with a focus on endocrine physiology. To determine pancreatic safety requires a broader view, one encompassing three complementary aspects: 1) plasma enzyme concentrations, as a marker of inflammation or leakage; 2) parameters representing pancreatic exocrine function, such as intraduodenal lipase activity, pancreatic fluid secretion, and fecal pancreatic enzyme concentrations; and 3) morphological features, including pancreatic volume, degree of steatosis, and main pancreatic duct (MPD) diameter (11). So far, only effects on plasma lipase and amylase concentrations have been studied, providing consistent evidence of a drug-induced increase (12–14). The current study was designed to perform an integrated assessment of the effects of the GLP-1 receptor agonist liraglutide and the DPP-4 inhibitor sitagliptin on pancreatic physiology and morphology in type 2 diabetes.

This was a 12-week, randomized, placebo-controlled, double-blind, parallel-group intervention trial in which subjects received the GLP-1 receptor agonist liraglutide (Novo Nordisk A/S, Bagsvaerd, Denmark), the DPP-4 inhibitor sitagliptin (Merck & Co, Kenilworth, NJ), or matching placebos. The study was carried out at the Diabetes Center of the VU University Medical Center. The protocol has previously been described in detail (15). The study was approved by the ethics review board of the VU University Medical Center (approval number 2012/391), registered at clinicaltrials.gov (NCT01744236), and conducted in accordance with the Declaration of Helsinki and the International Conference on Harmonization of Good Clinical Practice. All patients provided written informed consent before participation.

Fifty-six Caucasian patients with type 2 diabetes were recruited by newspaper advertisements. Inclusion criteria were age 35–75 years (women had to be postmenopausal, defined as no menses for >1 year); HbA_1c 6.5–9.0% (48–75 mmol/mol); using a stable dose of metformin and/or a sulfonylurea derivative for at least 3 months; and a BMI of 25–40 kg/m². Exclusion criteria were treatment with insulin or GLP-1–based therapy; history of pancreatic, hepatic, renal (estimated glomerular filtration rate <60 mL/min/1.73 m²), or cardiovascular disease; allergy to any of the test substances; and inability to undergo MRI.

The trial pharmacist randomized patients using computer-generated numbers. Patients were assigned in a 1:1:1 ratio, with a block size of six, to receive liraglutide 1.8 mg, sitagliptin 100 mg, or placebo. Both patients and trial physicians were blinded to study group assignments. Novo Nordisk provided pens prefilled with liraglutide or placebo, and ACE Pharmaceuticals B.V. (Zeewolde, the Netherlands) encapsulated sitagliptin or placebo tablets. Study drugs were taken once daily, in the evening. Liraglutide (and placebo) injections were started at a dose of 0.6 mg for the first week, 1.2 mg for the second week, and 1.8 mg for the remaining 10 weeks. In the case of drug intolerance, time between doses could be extended and/or the dose could be reduced. After completion of the study, compliance was established by counting the remaining study drugs (16).

The primary end point was first defined as changes in fecal elastase-1. Later, it was expanded to include other exocrine pancreatic function tests: intraduodenal lipase activity (¹³C-mixed triglyceride [MTG] breath test), pancreatic fluid secretion (secretin-enhanced magnetic resonance cholangiopancreatography), and fecal chymotrypsin concentrations. This adaptation was made because the fecal elastase-1 test is not considered to be sensitive enough to detect minor changes. The protocol was adjusted well before the statistical analyses were conducted (see the Supplementary Data). Secondary end points included plasma lipase, amylase, and trypsinogen concentrations; urine trypsinogen concentrations; and morphological features (pancreatic volume, steatosis, MPD diameter), as evaluated by MRI. In addition, gastric emptying (acetaminophen absorption test) was measured.

End points were measured after a 4-week run-in period. Blood was drawn at baseline and after 2, 6, and 12 weeks of treatment. Stool and urine were collected at baseline and after 12 weeks. Also, at both time points, two study visits were planned for the ¹³C-MTG breath test and acetaminophen absorption test on one day and MRI on another (in random order). For each visit, patients were instructed to arrive in a fasted state and withhold all medication, except metformin. Study medication was taken the evening before the study visit. (The assessment techniques are described in detail in the Supplementary Data').

Patients received a kit containing sample tubes and were carefully instructed in fecal collection and storage. Within 2 days, samples were frozen at −20°C until determination of elastase-1 (ELISA techniques; normal value >200 μg/g), whereas chymotrypsin concentrations were measured within 7 days (colorimetric method; normal value >3 U/g).

Plasma lipase (pancreas-specific; normal value <70 U/L) and amylase (α-amylase, measuring pancreatic and salivary amylase; normal value <100 U/L) were measured using standardized enzymatic techniques, according to the International Federation of Clinical Chemistry. Plasma and urine trypsinogen were determined by sandwich ELISA (normal values unknown for plasma; urine, <50 ng/mL).

Before this visit, subjects were instructed not to take any products that are naturally enriched with ¹³C for 2 days. Upon arrival, blood (acetaminophen concentrations) and breath (¹³C) samples were collected. Subsequently, a mixed meal (420 kCal, 22.4 g fat, 38.6 g carbohydrates, and 14.6 g protein) containing 250 mg ¹³C-MTG (Euriso-Top, Saint-Aubin Cedex, France) and 1.5 g liquid acetaminophen (62.5 mL Daro Paracetamol; Remark Groep, Rogat, the Netherlands) was consumed within 15 min. Breath samples were collected every 30 min for 6 h, and blood samples were taken every 30 min for 3 h. Biochemical analysis techniques are described in the Supplementary Data.

The MRI protocol has been described previously (17). In short, after administering 300 mL of oral ferumoxsil suspension (Lumirem; Guerbet, Gorinchem, the Netherlands), a contrast agent that enhances pancreatic duct visibility, imaging was performed using a 1.5-T MRI system (Magnetom Avanto; Siemens Healthcare, Erlangen, Germany). First, scans were acquired in the unstimulated state. Subsequently, 1 CU/kg of secretin (Secrelux; Sanochemia Diagnostics, Vienna, Austria) was administered intravenously over 60 s, after which imaging was repeated to assess secretin-induced bicarbonate and fluid production.

Images were analyzed using dedicated software, Onis version 2.4 (DigitalCore Co., Tokyo, Japan) and Sante DICOM Editor version 3.1 (Santesoft LTD, Athens, Greece). Signs of pancreatitis were evaluated (18). As detailed in the Supplementary Data, pancreatic volume, fat content, total pancreas-secreted volume, secretion speed, time to reach Matos-3 (filling beyond the inferior duodenal genu), and MPD diameter were measured.

Glucose and HbA_1c were measured at baseline and after 12 weeks of treatment using the Gluco Quant-hexokinase method on a Modular P analyzer (Roche Diagnostics, Basel, Switzerland) and high-performance liquid chromatography, respectively.

The sample size was calculated as the primary outcome measure, based on an acute intervention study (19) in which GLP-1 peptide infusion reduced pancreatic exocrine function by 40% (SD 35%). With a parallel-group design, an α of 0.05, and power (1 − β) of 80%, we needed 13 participants per group, comparing liraglutide or sitagliptin with placebo. To account for dropouts, and to increase the power to find smaller changes, we included 20 participants per group.

All data were double-entered into an electronic data management system (OpenClinica version 3.3; OpenClinica LLC, Waltham, MA) and exported to the study database. Baseline characteristics are presented as means ± SDs. End point data are presented as placebo-corrected means with 95% CIs or, in the case of a non-Gaussian distribution, as medians (interquartile ranges).

To test treatment effects versus placebo, multivariable regression models (for singly measured continuous data) and linear mixed models (for repeatedly measured continuous data) were used in the per-protocol population. Treatment with liraglutide or sitagliptin was added to the model as the dummy variable, thereby integrally correcting for multiple use of the placebo group. In addition, corresponding baseline variables were added as covariates to correct for baseline differences. When variables demonstrated a non-Gaussian distribution, log-transformation was applied. In linear mixed-model analyses, time was added as a fixed factor. The outcome of interest was the intervention-by-time interaction. A Mann-Whitney U test was performed for end points for which the time to reach an event was measured on an ordinal scale (for example, time to reach Matos-3). Finally, correlation analyses were performed between treatment-induced changes in enzyme levels, exocrine function, and MRI parameters using the Spearman correlation technique. All analyses were performed using SPSS 22.0 (IBM/SPSS Inc., Chicago, IL), and a two-sided P value ≤0.05 was considered to be statistically significant.

Between July 2013 and August 2015, 56 participants were randomized: 19 to liraglutide, 20 to sitagliptin, and 17 to placebo (Fig. 1). In the sitagliptin group, one patient withdrew from the trial because of adverse events (dizziness and pollakiuria). In the liraglutide group, the dose was reduced to 0.6 mg for one patient. Drug compliance was similar for patients treated with liraglutide (mean proportion of taken pens 101.6% and pills 98.9%), sitagliptin (pens 100.1%, pills 98.6%), and placebo (pens 100.9%, pills 98.7%); none of the subjects had compliance <90%. Baseline characteristics were similar in the three groups (Fig. 1). HbA_1c improved with both liraglutide (placebo-corrected mean difference −1.3% [95% CI −0.9 to −1.7]; P < 0.001) and sitagliptin (−0.8% [−0.4 to −1.2]; P = 0.001). Moreover, liraglutide and sitagliptin decreased fasting glucose by −1.7 mmol/L (−0.8 to −2.6 mmol/L; P = 0.001) and 1.8 mmol/L (−0.8 to −2.7 mmol/L; P < 0.001), respectively. Liraglutide tended to reduce body weight (−1.7 kg [−3.6 to 0.3]; P = 0.09), whereas sitagliptin had a neutral effect on weight (−0.9 kg [−2.7 to 1.0]; P = 0.37). Gastrointestinal adverse effects (nausea, diarrhea) were reported by 12 patients taking liraglutide, 2 patients taking sitagliptin, and no patients receiving placebo treatment. Other (minor) adverse effects were experienced equally among the treatment groups. No cases of pancreatitis were observed.

After 12 weeks of treatment, lipase activity, measured as cumulative ¹³C recovery, was similar to that of placebo (liraglutide −0.4% [95% CI −4.9 to 4.3], P = 0.88; sitagliptin −1.5% [−6.1 to 3.0], P = 0.50) (Fig. 2A). Maximal ¹³C recovery speed and time were also unaffected. Sitagliptin increased the total secreted pancreatic fluid volume (16.3 mL [−0.3 to 32.9]; P = 0.05), partly driven by a decrease with placebo treatment, whereas liraglutide had no effect (−0.2 mL [−16.8 to 16.5]; P = 0.99) (Fig. 2B). Neither drug changed the maximal secretion speed (Fig. 2B) or time to reach maximal secretion speed or Matos-3 (data not shown). Fecal elastase-1 concentrations were not affected by either liraglutide (+152.1 µg/g [−383.8 to 688.1]; P = 0.58) or sitagliptin (+196.8 µg/g [−324.6 to 718.2]; P = 0.46) (Fig. 2C). Also, fecal chymotrypsin concentrations in both treatment groups were similar to those in the placebo group after 12 weeks, although after 2 weeks, an increase was observed with sitagliptin (11.9 U/g [−0.1 to 24.0]; P = 0.05).

After 12 weeks, no differences in plasma lipase concentrations were observed between the GLP-1–based treatment groups and placebo, although a trend toward higher concentrations in the sitagliptin group was found (placebo-corrected increase of 17.7 U/L [95% CI −3.1 to 38.4]; P = 0.10) (Fig. 3). In addition, at 6 weeks, lipase concentrations were significantly higher with liraglutide than placebo (23.5 U/L [2.1–44.8]; P = 0.03). Plasma amylase concentrations were no different from those in the placebo group after 12 weeks of treatment. However, liraglutide tended to increase amylase after 6 weeks (10.3 U/L [−0.2 to 20.9]; P = 0.06), whereas in the sitagliptin group, amylase concentrations were higher than those in the placebo group at weeks 2 and 6 (13.7 U/L [3.4–23.9], P = 0.03; and 11.8 U/L [1.5–22.2], P = 0.01, respectively). Plasma trypsinogen concentrations were higher with both drugs after 12 weeks of treatment compared with placebo (liraglutide 34.6 µg/mL [15.1–54.2], P = 0.001; and sitagliptin 23.9 µg/mL [4.9–42.9], P = 0.01). These increases reached significance after 2 weeks of treatment. Neither drug increased urinary trypsinogen >50 ng/mL.

No signs of pancreatitis (edema, infiltration) were present on MRI after 12 weeks. Pancreatic steatosis was not significantly altered by liraglutide (−2.4% [95% CI −6.4 to 1.6]; P = 0.24), but was reduced by sitagliptin (−4.2% [−8.1 to −0.3]; P = 0.04). A modest trend toward increased pancreatic volume was seen with both liraglutide (7.7 cm³ [−1.2 to 16.6]; P = 0.09) and sitagliptin (6.8 cm³ [−1.9 to 15.6]; P = 0.12) (Fig. 4) compared with placebo. The MPD diameter (before and after secretin stimulation) did not differ between treatment groups and the placebo group (Fig. 4).

In an analysis of all treatment groups combined, a change in pancreatic volume was associated with an increase in amylase levels (R = 0.353; P = 0.01), but not lipase levels (R = 0.111; P = 0.45) (Supplementary Fig. 1). Finally, liraglutide or sitagliptin did not change gastric emptying, as measured by the maximal acetaminophen concentration (C_max), the time to reach the maximal concentration (T_max), and area under the curve, compared with placebo (Supplementary Fig. 2).

In patients with type 2 diabetes, 12 weeks of treatment with the GLP-1 receptor agonist liraglutide or the DPP-4 inhibitor sitagliptin induced a modest increase in trypsinogen levels. No significant increase in amylase or lipase levels was found after 12 weeks, although liraglutide transiently increased lipase and sitagliptin transiently increased amylase. Sitagliptin decreased steatosis and marginally increased pancreatic exocrine fluid secretion. No other morphological or functional changes were observed.

The transient increase in plasma enzyme concentrations in the earlier weeks of the intervention confirms the results of many observational studies and randomized clinical trials (12–14). We show that this increase occurs after 2 weeks and also includes trypsinogen. In contrast to a recent trial, in which the same dose of liraglutide elevated lipase and amylase levels for up to 56 weeks (13), our enzyme increase was not sustained. This may be explained by fluctuations in lipase and amylase levels during GLP-1–based treatment, as was shown in several prolonged intervention studies (12–14). Alternatively, effects on these enzyme levels may wane over time. We cannot explain why the increase in trypsinogen levels was sustained while the increases in lipase and amylase levels were transient. One reason could be the ability of acinar cells to differentially secrete these enzymes, as illustrated by the influence of nutrients on enzyme composition. While dietary lipids predominately stimulate lipase, carbohydrates induce more amylase secretion (21). GLP-1 potentially stimulates trypsinogen secretion more than it does the other enzymes.

While liraglutide had no effect on pancreatic exocrine function, sitagliptin modestly increased secretin-stimulated pancreatic fluid secretion after 12 weeks, without having any effect on intraduodenal lipase activity or fecal enzyme concentrations. Secretin-enhanced magnetic resonance cholangiopancreatography assesses ductal bicarbonate and fluid secretion, whereas the other two parameters measure acinar enzyme secretion, which could explain the observed difference. Alternatively, the effect of sitagliptin may only be present after maximal pancreatic stimulation, as is induced by secretin infusion. Also, since pancreatic duct cells express DPP-4, local inhibition by sitagliptin may alter intracellular processes, thereby increasing bicarbonate secretion (22). The meaning of the transient increase in fecal chymotrypsin concentrations after 2 weeks of sitagliptin is unclear; this effect was not paralleled by the more sensitive fecal elastase-1 test. This temporarily higher level may have been caused by the large day-to-day variation (∼30%) of chymotrypsin (23). Also, DPP-4 is known to degrade chymotrypsin, and thus reduced intestinal degradation as a result of DPP-4 inhibition may have added to the tendency of sitagliptin to increase fecal chymotrypsin concentrations (24). Notably, the effects of sitagliptin on bicarbonate secretion and chymotrypsin were small and the 95% CI wide.

The lack of an effect of liraglutide treatment on exocrine secretion is in contrast with several previous studies in which infusion of GLP-1 instantly decreased meal-stimulated exocrine secretion up to 40% (19,25). In those studies, however, the effect of GLP-1 was likely caused by a reduction in gastric emptying, which reduces duodenal acidity and nutrient loading, both of which are important triggers of pancreatic secretion. Long-acting liraglutide has no effect on gastric emptying after prolonged intervention (2), as also demonstrated in the current study, which explains why we observed no effect on the meal-stimulated ¹³C-MTG breath test parameters. Short-acting GLP-1 receptor agonists retain their gastric inhibitory effect (2); therefore, the results of this test may have been different if a short-acting agent had been used. Importantly, however, the current data confirm our previous finding that the GLP-1 receptor agonist exenatide does not affect secretin-stimulated intraduodenal pancreatic fluid secretion (17). Secretin infusion bypasses the effect of GLP-1 on gastric emptying, and the direct effect of GLP-1 on pancreatic exocrine secretion can be measured. Since neither acute exenatide (17) nor 12-week liraglutide affected secretin-stimulated pancreatic secretion, it is unlikely that GLP-1 (receptor agonists) has direct effects on pancreatic secretion.

Using MRI techniques, we found no significant changes in pancreatic morphology after either treatment. However, pancreatic volume tended to increase with liraglutide and, to a lesser extent, with sitagliptin. Pancreatic volume may be expanded by cellular hyperplasia/hypertrophy, increased steatosis, or edema. We did not observe the latter two conditions, although discrete changes may have been missed on MRI. In animal studies, an increase in pancreatic weight was found after treatment with a GLP-1 receptor agonist (8,26). A study of mice demonstrated that 14 days of treatment with the GLP-1 receptor agonist exendin-4 or liraglutide increased pancreatic weight by stimulating acinar protein production without inducing changes in ductal or islet cells (26). Increased acinar productivity can induce cellular growth and thus increase both plasma enzyme concentrations and pancreatic volume, which may explain our observed correlation between changes in amylase levels and pancreatic volume, and the tendency of liraglutide to increase pancreatic volume.

The reduction of pancreatic steatosis with sitagliptin is of interest. Pancreatic steatosis is a relatively unexplored phenomenon with an unknown pathophysiology. However, because it has been associated with β-cell dysfunction, exocrine insufficiency, acute pancreatitis severity, and pancreatic fistula after pancreatic surgery (27), further study should assess whether this modest decrease in pancreatic fat content has clinical relevance.

In this study, increases in plasma enzyme concentrations and exocrine fluid secretion were not accompanied by clinical signs or symptoms of pancreatitis. The implications of these modest changes remain speculative. Elevations in enzyme levels caused by GLP-1–based drugs do not predict pancreatitis events, and enzyme levels rapidly normalize after treatment cessation (28). On the other hand, an analogy may exist with cerulein-induced pancreatitis, whereby high-dose cerulein induces acute pancreatitis and a lower dose causes cellular injury and, eventually, chronic pancreatitis (29). In turn, chronic inflammation may lead to pancreatic cancer (29). In this context, the modest increase in exocrine secretion with sitagliptin may support the findings of a recent meta-analysis demonstrating an increased risk of acute pancreatitis with DPP-4 inhibitors (7). Large-scale prospective data on GLP-1 receptor agonists do not suggest pancreatic adverse effects (30–32). Evidence of pancreatic cancer risk is hampered by a lack of long-term studies.

This study has several limitations. First, the study has a relatively small sample size. Importantly, this mechanistic trial was not designed to assess pancreatic adverse events, which would require a much larger study population and longer follow-up. The power calculation was based on a study demonstrating an acute and profound inhibition of exocrine secretion (19), and was supported by studies demonstrating extensive inhibitory effects of GLP-1 on other proximal gastrointestinal organs. However, the effects in the current study were not as strong as expected, creating risk of type 1 and 2 errors. Second, while the study population is representative of patients with diabetes who receive GLP-1–based therapies in a clinical setting, we did not assess patients at high risk of developing pancreatitis. It has been suggested that pancreatic adverse events occur more readily in patients with previous pancreatic damage or with risk factors such as alcoholism or a family history of pancreatic disease (33). Third, in this study, most subjects used metformin, which could have attenuated effects; an animal study showed that pancreatic adverse effects of GLP-1–based therapies were ameliorated by concomitant use of metformin (34). Fourth, the observed lowered glucose may have affected potential effects of GLP-1–based therapies on the exocrine pancreas (35). However, while interesting from a mechanistic point of view, the net effect is more important clinically. Finally, we did not measure exocrine function by duodenal aspiration, which is considered the gold standard test. Because of the large reserve capacity of the exocrine pancreas, only profound changes can be measured. Thus, more subtle fluctuations may not have been captured.

In conclusion, in this comprehensive study of patients with type 2 diabetes, treatment with a GLP-1 receptor agonist or a DPP-4 inhibitor increased serum trypsinogen concentrations after 12 weeks, but not lipase or amylase concentrations. Furthermore, our study suggests an association between treatment-induced changes in plasma enzyme concentrations and pancreatic volume expansion. Pancreatic exocrine function was unaffected, with the exception of exocrine fluid secretion, which was stimulated by sitagliptin. It seems unlikely that these subtle changes will induce pancreatitis after long-term treatment, but this requires further study.

Acknowledgments. This article is published in memory of Prof. Michaela Diamant, whose experience and expertise were crucial for the design of this study.

Funding. The research leading to these results was funded by the European Community’s Seventh Framework Programme (FP7/2007-2013), under grant agreement no. 282521 (the SAFEGUARD project). Sanochemia Diagnostics Deutschland GmbH provided the Secrelux vials. Novo Nordisk provided prefilled liraglutide and liraglutide-placebo pens.

Duality of Interest. The VU University Medical Center received research grants from AstraZeneca, Boehringer Ingelheim, Novo Nordisk, and Sanofi (to M.H.H.K. and M.D.). M.J.B. received speaker and consultancy fees from Boston Scientific, Cook Medical, and SOCAR. M.J.B. also received research grants from Boston Scientific, Cook Medical, and Pentax. No other conflicts of interest relevant to this article were reported.

The funders had no role in designing the study; in collecting, analyzing, and interpreting data; in writing the article; or in deciding to submit the article for publication.

Author Contributions. M.M.S. and L.T. developed the study protocol, performed the measurements and analyses, and wrote the manuscript. M.H.A.M., I.C.P.v.d.B., and K.E.W.V. performed measurements, contributed to the discussion, and wrote the manuscript. M.H.H.K., T.H., M.J.B., D.H.v.R., and D.L.C. contributed to the discussion and edited the manuscript. M.D. developed the study protocol. M.M.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.