The Human GLP-1 Analogs Liraglutide and Semaglutide: Absence of Histopathological Effects on the Pancreas in Nonhuman Primates

Pancreas safety has become a subject of much debate concerning dipeptidylpeptidase-4 (DPP-4) inhibitors (DPP-4is) and glucagon-like peptide-1 (GLP-1) receptor agonists (GLP-1RAs). This concern is partly based on the well-described effect of GLP-1 to induce growth of pancreatic β-cells (1). Both drug classes increase effective GLP-1 levels, but to a different degree, and have different modes of action; thus, it is important to differentiate between them, particularly when considering mechanistic hypotheses for potential safety concerns or signals. GLP-1RAs mediate their effects directly through the GLP-1R (2,3). Although increased levels of GLP-1 and glucose-dependent insulinotropic peptide (GIP) are considered important parts of the mechanism of action of DPP-4is (4), DPP-4 is known to degrade many other hormones (5). DPP-4is have been shown to increase GLP-1, GIP, and peptide YY (6,7). This complicates the understanding of both desired and potentially undesired effects of this class of compounds. Within GLP-1RAs, differences exist in safety-related parameters. One subgroup is the exendin-4–based drugs with exenatide and lixisenatide, which are structurally distinct from human GLP-1. Owing to the low amino acid homology to native human GLP-1, these medications are associated with an increased number of immune reactions that are, however, all of a relatively mild form, that is, mostly antibody development, injection site nodules, and loss of efficacy (8,9). The other subgroup is based on human GLP-1 and contains liraglutide, taspoglutide, and larger covalently conjugated molecules such as albiglutide and dulaglutide. Clinical development of taspoglutide was stopped due to severe immune-related side effects, with cases of anaphylactic shock, possibly caused by the formulation (8,10). No such findings have been reported with other GLP-1–based analogs.

Semaglutide is a once-weekly GLP-1 analog that is in phase 3 clinical development (11). Where liraglutide is acylated with a palmitic acid and has an extra amino acid as a spacer between the palmitic acid and the Lys26, where the fatty acid is attached, semaglutide is acylated with a stearic diacid at Lys26 but has a much larger synthetic spacer and is, furthermore, modified for DPP-4 stability in position 8, where the amino acid α-aminobutyric acid has been introduced.

In support of drug development and regulatory approval for treatment of chronic diseases, repeat dose toxicity and carcinogenicity studies are conducted. These studies are performed at different dosing durations to support the different phases of clinical development and with doses aiming to obtain exposure several multiples higher than the clinically relevant exposure with the aim to identify potential drug-related organ toxicity and carcinogenicity. Repeat dose toxicity studies are typically conducted in a rodent and a nonrodent species. For liraglutide and semaglutide, Cynomolgus monkeys were chosen as the nonrodent species. Repeat dose toxicity studies in nonhuman primates are designed to screen for potential hazards and are not designed or statistically powered for identification of differences in the incidence or severity of individual organ changes. The number of nonhuman primates per group is limited to three to five for ethical reasons (12). Because of the statistical limitations, standardization of the examinations is critical: If pathological findings are identified at a frequency or severity exceeding those in the in-study control group, they are often compared with historical control data to assist interpreting the significance of the finding. This principle also applies to organ weights. Histopathological findings in the pancreas from the repeat-dose studies in rodents and nonhuman primates and from carcinogenicity studies in rodents with liraglutide have been published previously (13). Liraglutide was not found to have a causal relationship to any histopathological findings.

Some studies in rats and mice have shown an increased pancreas weight induced by DPP-4is or GLP-1RAs (14,15), and a recent ex vivo study with human pancreata suggested an increase in glucagonomas as well as increased pancreas weight (16). Here, pancreas weight in Cynomolgus monkeys is reported for liraglutide in toxicology studies with 4, 13, 52, and 87 weeks’ dosing, and for semaglutide in toxicology studies with 13 and 52 weeks’ dosing, as well as a full histopathological evaluation of these same studies, except the 87 weeks’ study, which has been reported previously (13). For liraglutide, a full quantitative histological assessment of the endocrine as well as the exocrine pancreas was also performed in the 52 weeks’ study.

The research design and methods for liraglutide studies in Cynomolgus monkeys have been described previously (13). All animals were examined daily for clinical signs in the in-life phase. Studies with semaglutide were generally performed similarly, and both followed international guidelines provided by International Conference of Harmonization. Dose levels, duration, and group sizes for all studies are described in Fig. 1. Compounds were administered as subcutaneous injections.

Terminal body weight was obtained in sedated animals immediately before they were killed. The entire pancreas of each animal was examined macroscopically for any abnormalities, excised, cleaned of fat and connective tissue, and weighed.

A transverse section from the midpart of the pancreas from all animals was sampled, and this section and the rest of the pancreata were fixed in 10% neutral buffered formalin for at least 48 h. The sections were dehydrated and paraffin-embedded according to standard histopathological procedures. According to international standard practices, one section per animal was cut at a nominal thickness of 4–5 µm and stained with hematoxylin and eosin (17). The slides were read unblinded because this is recommended by toxicopathology experts as a way to increase the chance of separating subtle changes from normal background changes (18). Further details of the methodology have been previously described (13). The pancreas specimens from animals killed at termination of dosing in the 52-week liraglutide study were evaluated by quantitative histology (all groups for α-cells and Ki-67, and only vehicle and high-dose groups for β-, δ-, pancreatic polypeptide [PP], ductal, and acinar cells). The cranial and caudal remnants were sectioned longitudinally, cut into ∼40 pieces, and distributed to four capsules, with one-fourth in each, according to the smooth fractionator principle (19,20). Processing and staining for detection of β-, α-, δ-, and PP cells was done as previously described (21). The reactivity of the primary antibodies to insulin, glucagon, somatostatin, and PP in monkey pancreas had been tested individually to validate the method. Costaining for duct plus acinar cells using mouse anti–CK-7 (Dako, Glostrup, Denmark) and rabbit anti–a-amylase (Calbiochem, Darmstadt, Germany) plus Ki-67 using polyclonal rabbit anti–Ki-67 (Novus Europe, Cambridge, U.K.) followed the same principles as above. Stained slides were scanned in a Hamamatsu NanoZoomer 2.0-HT (Hamamatsu, Hamamatsu City, Japan). Images were subsequently analyzed automatically in the Visiopharm Integrator System (Visiopharm A/S, Hørsholm, Denmark) image analysis program.

Data on pancreas weights were collected from the contract laboratories where the study of interest was conducted to ensure matching for origin, age range, and environmental conditions to the highest possible extent. Historical control data on pancreas weights from animals used as vehicle controls in other studies of 4, 13, and 39–52 weeks’ duration was used for comparison with liraglutide: 4 week (n = 29 males and 28 females as historical controls), 13 week (n = 19 males and 19 females as historic controls), and 52 week (n = 22 males and 17 females as historical controls). No historical control data were available for comparison with the 87-week study because this is not a standard-length toxicology study.

Measurements were performed on samples obtained by a systematic uniform random sampling technique. The mean value for each animal was calculated relative to the total tissue area counted for each estimate. Volume fractions were measured in percentage of the total pancreas volume. Total cell mass in milligrams was calculated by multiplying the volume fraction with the total pancreas weight.

The liraglutide data on pancreas weight were, per study protocol, analyzed by ANCOVA on body weights and pancreas weights combining both sexes and using day 0 body weight and terminal body weight, respectively, as covariates. These prespecified models were slightly different between studies because they were data-driven. Only the 52-week study reported an increase in pancreas weight when this analysis was used (22). To obtain consistency across studies and thus enhance comparisons, post hoc modeling was performed using the same models in all studies. In the prespecified analyses, terminal body weight was taken into consideration in the statistical models. However, that is problematic because liraglutide lowers body weight. The post hoc analysis is thus a one-way ANOVA of the pancreas weight for each study and sex separately. In case of variance inhomogeneity, measured by means of the Bartlett and Brown-Forsythe tests, a Kruskal-Wallis test was performed to evaluate overall group effects. Because statistically significant effects of liraglutide were seen in some studies for some of the doses, a one-way ANOVA was also performed for the relative pancreas weight to further examine a possible effect. The Dunnett multiple comparison test was used as posttest after ANOVA in cases where the overall group effect was significant. Parameters from quantitative histology were analyzed by the Student t test for each sex, except for α-cell mass, where one-way ANOVA was used. P < 0.05 was considered statistically significant. Data are presented as mean ± SEM unless otherwise stated.

Pancreatic weight data from liraglutide studies are shown in Fig. 1. A significant increase that was apparently not dose-related was found in the 4-week study in male animals (P = 0.036 by ANOVA), whereas no increase was seen in females in the same study. No statistically significant differences were found for any dose level in male or female animals after 13 or 87 weeks of dosing. Similarly, in the 52-week study, no statistically significant differences were found for the male animals but a significantly higher pancreas weight, apparently dose-related, was found in the female high-dose group compared with the in-study control group (P = 0.007 by ANOVA).

This comparison did not show any statistically significant differences in pancreas weight for any dose level in males or females from the 4- and 13-week studies or males in the 52-week study. When compared against historical controls in females from the 52-week study, a statistically significant difference across groups was found (P = 0.005 by ANOVA), but posttest showed that only the in-study controls exhibited a statistically significant different (lower) pancreas weight than the historical controls (n = 4 and 17, respectively; P = 0.04).

An ANOVA analysis of the relative pancreas weight showed a significant increase compared with in-study controls in the male mid- and high-dose groups and in the female high-dose group in the 52-week liraglutide study (data not shown). There were no statistically significant differences in the 4-, 13-, or 87-week liraglutide studies.

Additional animals were included to explore reversibility of potential findings (n = two males and two females, in the control and high-dose groups in the 13- and 52-week studies for liraglutide and semaglutide). These additional animals were dosed for the same duration as all other animals in the study but were kept for additional 4 weeks after end of treatment before being killed. In the 52 weeks’ study, there was no longer any difference in pancreas weight versus controls after the 4-week treatment-free period (data not shown).

Histological examination of the pancreata from the 4-, 13-, and 52-week studies is shown in Table 2. Representative histological sections of the different compartments of pancreas after 52 or 87 weeks of dosing are illustrated in Figs. 2 and 3. The histological examination did not reveal treatment-related differences between dosed and control animals at any time point. The endocrine pancreas revealed well-demarcated islets with normal pale islet cells after 52 and 87 weeks of dosing (Fig. 2). In the exocrine pancreas, all ducts appeared normal, both the large main duct with the high columnar epithelium and abundant surrounding connective tissue (shown in Fig. 3, left side panels, and in higher magnification in Supplementary Fig. 1) and the medium-sized interlobular ducts with the lower cuboidal epithelium and less surrounding connective tissue (Fig. 3, right side panels). The small intercalated ducts with flat epithelium and no or sparse surrounding connective tissue also appeared normal (not shown in figures). The acinar cell parenchyma consisted of normal pyramid-shaped cells where the apical part was filled with eosinophilic zymogen granules and the basophilic basal part contained the nucleus, as shown in Fig. 3 and in Supplementary Fig. 2.

A quantitative histological assessment of the pancreas in the 52-week liraglutide study was conducted to evaluate whether there were changes in the mass of pancreatic cell types that could not be identified by the qualitative histological analysis. Table 1 shows absolute mass of β-, α-, δ-, PP, duct, and acinar cells. Liraglutide-dosed monkeys showed no significant differences in any of these measures, except for absolute duct cell and acinar cell mass, which was significantly increased in the female high-dose group. When the proportion of cells per volume of pancreas was evaluated, no changes were found for ductal cell volume in males (liraglutide 6.14 ± 0.59 vs. control 6.71 ± 0.62%, P = 0.53) or females (7.40 ± 0.63 vs. 6.55 ± 1.29%, P = 0.57) or for acinar cell volume in males (91.4 ± 0.9 vs. 88.7 ± 0.8%, P = 0.06) or females (90.3 ± 0.4 vs. 90.0 ± 0.7%, P = 0.71) high-dose compared with vehicle. Thus, the increased pancreas weight was a balanced increase of the exocrine pancreas, with no apparent change in the ratio of ductal to acinar tissue. To further evaluate if α-cell mass specifically was changed by liraglutide, the low- and medium-dose liraglutide groups were also evaluated quantitatively. Liraglutide did not change α-cell mass in males in the low- or middle-dose groups (30.0 ± 8.4 and 24.2 ± 3.3, respectively, vs. 29.9 ± 3.6 mg for controls, from Table 1) or females (16.0 ± 1.6 and 20.4 ± 2.0, respectively, vs. 22.6 ± 3.3 mg for controls, from Table 1), with P = 0.74 and P = 0.22 for males and females across groups by ANOVA.

Examples of β- and non–β-cell staining (Fig. 4) and α-cell and proliferation (Ki-67) (Fig. 5, and higher magnification in Supplementary Fig. 3) in representative sections of pancreata from males and females in the vehicle and the liraglutide high-dose group of the 52-week dosing study are shown. In control and high-dose animals, glucagon staining showed a high and variable number of α-cells in islets (typically ∼50% of non–β-cells were α-cells). Small numbers of single cells or small clusters of glucagon-positive cells associated with ducts were seen in the exocrine pancreas (Fig. 5). Very few cells in the endocrine and exocrine pancreas were positive for Ki-67, and there was no apparent difference between the liraglutide and control groups (data not shown). As a positive control for proliferation, lymph nodes present in 15 of the total of 32 pancreata showed strong labeling of many cells in germinal centers and also some single cells (shown as inserts in Fig. 5).

Pancreas weights from semaglutide studies are shown in Fig. 1. In the 13-week study, there was no statistically significant treatment-related effect on pancreas weight in males, whereas a significant difference was seen across groups in the females (P = 0.02 by ANOVA). However, by posttest, no significant difference was found between treated groups and the control group, and no apparent dose-related effects were seen. In the 52-week study, no statistically significant differences across study groups were observed for pancreas weight; the highest pancreas weights were found in the control groups.

Histological examination of the pancreas from the 13- and 52-week studies revealed common background findings of minimal to mild severity and with a focal distribution. There were no signs of treatment-related effects. The data are reported in Table 2.

Reported here are further data from nonhuman primate studies conducted with liraglutide as a supplement to previously published data on pancreas histology in mice, rats, and nonhuman primates (13). An apparent dose-related increase in absolute pancreas weight was found in females in one of four monkey studies with liraglutide and in none of two monkey studies with semaglutide, and an increase that did not appear to be dose-related was found in a 4-week liraglutide study, in males only (Table 3). A comparison with historical control data was made. This is a common way of setting toxicological data into perspective and may be especially useful in nonhuman primate studies where the number of animals is low due to ethical reasons. The histological analysis of the pancreas from the liraglutide and semaglutide studies did not reveal any potentially adverse findings that could be related to treatment (e.g., pancreatitis, inflammatory cell infiltrations, or hyperplasia). Overall, this led to the conclusion that an increase in pancreas weight cannot be ruled out, but no consistent dose-related increases in pancreas weight were seen across the liraglutide and semaglutide studies in monkeys. In combination with the lack of treatment-related histopathological changes, the data showed no adverse effects on the pancreas by liraglutide dosing of up to 60-times the clinically relevant exposure for up to 87 weeks or semaglutide dosing for up to 52 weeks in monkeys.

The studies reported here have a high relevance for humans because the pancreata of nonhuman primates are closely related to humans anatomically and physiologically. Additionally, the morphology of nonhuman primate islets is like that seen in humans and different from rodents (23). Quantitative histology was used to assess changes in the mass of different pancreatic tissue components, taking the three-dimensional structure of the organ into consideration (24). An inherent weakness of these studies is the relatively limited number of nonhuman primates per group because number of animals is constrained for ethical reasons. However, this report is based on data from 90 and 48 animals dosed with liraglutide and semaglutide, respectively. Another limitation is that only one transverse section was examined per animal. As a consequence thereof, the statistical power is low. Three of the four studies that have described adverse effects of DPP-4is or GLP-1RAs on the pancreas were performed in rodents (15,25,26). Thus, at the current point in time with the very few studies available in nonhuman primates, the cumulative data in these studies have a strong relevance for the assessment of adverse pancreas effects in humans, despite the relatively limited number of animals per group.

A recent study with human pancreata from patients previously treated with sitagliptin or exenatide (seven sitagliptin, one exenatide) reported an increased pancreas weight (16). A number of important considerations with the study design may have affected the results: the groups were unbalanced, with an 18-year age difference between the groups, and there was no attempt to control for type of diabetes, weight, age, or sex. Pancreas weight depends on body weight, stage of diabetes, age, and sex (27,28). These methodological problems are clearly elucidated in two related commentaries/reviews (29,30).

The data presented here do not show an increase in cell replication or number of α-cells caused by liraglutide treatment. Quantitative histology of the pancreas from the 52-week study with liraglutide demonstrated an increase of exocrine pancreas tissue with an apparently unchanged ratio between acinar and ductal cells and with normal tissue architecture. Cell proliferation in the pancreas measured by Ki-67 appeared unchanged by liraglutide; very low proliferation rates were found in all animals. There appeared to be no change of β-, δ-, or PP cell mass by 52 weeks of liraglutide treatment in the high-dose group and there was apparently no change in α-cell mass or indication of proliferation of α-cells in any of the three liraglutide-dosed groups in the 52-week study. The islets from the control and the liraglutide groups in the 52-week study had ∼50% non–β-endocrine cells, with a substantial portion of those being α-cells (25–30% of endocrine cell mass), less δ-cells (∼15%), and only a few PP cells (4–7%). The fraction of the four endocrine cell types in our study is in agreement with data from human pancreata and a descriptive study in Cynomolgus monkeys (31–34). The intraislet organization of the β- and non–β-cells was random, with no clear rodent-like mantle and core, but with a more complex subunit structure of mantles and cores, as characteristic for nonhuman primates and humans (23,35,36). In all groups, including the controls, the distribution of α-cells was identical in islets and islets-like structures of variable size. The control and liraglutide-dosed groups both showed a number of single glucagon-positive cells in the exocrine area, single cells and small clusters of glucagon-positive cells associated to the epithelial lining of both main and smaller ducts. This finding of small clusters and single endocrine cells is normal in Cynomolgus monkeys (34). A similar pattern with single cells in the exocrine areas and associated to ductal structures was observed for β-cells and less frequently with δ-cells and PP cells, with no differences between groups. An apparently dose-related increase in absolute pancreas weight was found in one of four studies, and only in one sex, in nonhuman primate studies with liraglutide. Despite this apparently dose-related increase in pancreas weight, no histopathology was associated thereto, and there were no pancreatic intraepithelial neoplasia lesions (13).

The four studies that have suggested adverse effect of DPP-4is or GLP-1RAs on the pancreas have described or discussed increased risk for pancreatitis, metaplasia, or inflammation and pancreatic adenocarcinomas and glucagonomas (15,16,25,26). In contrast, hundreds of other studies have investigated effects of these drugs on the pancreas but have not reported adverse effects; a few are referenced here for liraglutide (37–39). A study, performed after others had reported potential adverse findings, confirmed the absence of pathology in diabetic rats and also did not show any regional differences in the pancreas induced by liraglutide when the pancreas was divided into four regions and examined by stereology (40). A recent publication used a human islet amyloid polypeptide transgenic model similar to one of the earlier studies, just in mice instead of in rats, treats the animals for 1 year instead of 12 weeks, used 20–25 animals in each group instead of 8, and found no pancreas pathology associated with sitagliptin (25,41). The U.S. Food and Drug Administration recently published that it reassessed data from 50 GLP-1 based therapeutics and found no changes indicating pancreatic injury (42). To understand whether there are any potential adverse effects of GLP-1RAs on the human pancreas, the expression pattern of GLP-1R may be important. It has recently been recognized that most studies measuring the GLP-1R may be invalid because the antibodies used are not specific for GLP-1R (30,43,44). G-protein–coupled receptors are notoriously known for this problem (45–47). Some scientific journals have provided new guidance for validation experiments that must be available for reliable documentation of expression of a G-protein–coupled receptor (48). However, valid studies are available documenting GLP-1R expression (49). These studies measure receptor expression by ligand binding and show that pancreatic adenocarcinomas do not express GLP-1R (49,50). Thus, from a molecular target point of view, it seems unlikely that GLP-1R agonism should directly worsen or induce pancreatic adenocarcinomas when such tumors do not express GLP-1R.

On the basis of the totality of information available to Novo Nordisk, the available information is insufficient to confirm or exclude an association between liraglutide and pancreatitis, and there is no evidence that it increases the risk of pancreatic cancer in patients with type 2 diabetes. The Liraglutide Effect and Action in Diabetes: Evaluation of Cardiovascular Outcome Results—A Long Term Evaluation (LEADER; NCT01179048) will prospectively evaluate the overall safety of liraglutide. The trial has enrolled 9,340 patients with type 2 diabetes and a high cardiovascular risk profile. Patients are randomized 1:1 in a double-blind study design to liraglutide or placebo and will be followed up for a minimum of 42 months for the primary end point of adjudicated macrovascular events, including nonfatal myocardial infarction, stroke, or cardiovascular death. Adjudication of all adverse reactions related to pancreatitis and any neoplasm is an integral part of the protocol throughout the duration of the LEADER study. Randomized, controlled, long-duration trials with independent adjudication are the only way to evaluate rare side effects, as also recently mentioned by Kahn (29) and Drucker (30). The LEADER study will report in 2016.

Acknowledgments. The authors thank Charles Pyke, Novo Nordisk, for careful reading of the manuscript and interpretations of the methodology used.

Duality of Interest. Novo Nordisk markets liraglutide for the treatment of diabetes and has semaglutide in phase 3 clinical development. All authors are full-time employees of Novo Nordisk and hold minor share portions as part of their employment. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. C.F.G. performed the quantitative histology. A.-M.M. collected and reviewed parts of the data and took part in drawing conclusions from the liraglutide studies. I.T. reviewed the histopathological data and selected the photographs. N.C.B.N. took part in drawing conclusions from the liraglutide studies. Z.S. designed the semaglutide studies and concluded the results for those. L.B.K. took part in drawing conclusions from the results of the studies and wrote the major part of Abstract, Introduction, and Discussion sections. M.O.L. wrote the major part of the Research Design and Methods and Results sections and some parts of the Abstract and Discussion and performed the statistical analysis. All authors reviewed the manuscript. L.B.K. 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.