Introduction
Patients with schizophrenia have a reduced life expectancy of approximately 20 years compared to the general population.Reference Laursen, Nordentoft and Mortensen 1 The increased mortality is primarily caused by cardiovascular disease,Reference Correll, Carvalho and Fornaro 2 likely due to a combination of genetic, lifestyle, and treatment factors.Reference Laursen, Nordentoft and Mortensen 1 , 3-5 Metabolic disturbances, obesity, and type 2 diabetes mellitus (T2D) are all known risk factors for cardiovascular disease but also established adverse effects from the antipsychotic medications used to treat most patients with schizophrenia spectrum disorders. 6-8 Most recently, a Danish population-based cohort study demonstrated that a high endogenous risk for T2D in patients with schizophrenia was further increased after antipsychotic treatment,Reference Rajkumar, Psych and Horsdal 9 and it is well established that the antipsychotic medications, olanzapine and clozapine give rise to the greatest weight gain and metabolic disturbances.Reference Leucht, Cipriani and Spineli 10
Glucagon-like peptide-1 (GLP-1) is an incretin hormone secreted from the L-cells of the small intestine in response to food intake. It has a variety of physiological targets, including decreasing appetite, inhibiting gastric emptying, and lowering blood glucose concentrations by increasing insulin secretion from β-cells and decreasing glucagon secretion from α-cells of the pancreas in a glucose-dependent manner. 11-13 GLP-1 receptor agonists (GLP-1RA) mimic the effects of naturally occurring GLP-1 and are approved for the treatment of diabetes and obesity.Reference Marso, Daniels and Brown-Frandsen 14 , Reference Marso, Bain and Consoli 15 In a recent randomized double-blinded placebo-controlled trial, we demonstrated that the GLP-1RA liraglutide vs placebo significantly reduced glucometabolic disturbances and bodyweight in prediabetic, overweight/obese schizophrenia-spectrum disorder patients treated with the antipsychotic compounds, clozapine or olanzapine.Reference JR, Vedtofte and MSL 16 After 16 weeks of treatment, glucose tolerance normalized in 63.8% of the liraglutide group compared to only 16% in the placebo group, and a greater body weight reduction of 5.3 kg with liraglutide compared to placebo was found. A recent patient-level meta-analysis confirmed the efficacy of GLP-1RA for antipsychotic-related weight gainReference Siskind, Hahn and Correll 17 ; however, the exact mechanism for these favorable effects is yet to be elucidated.
The role of inflammation in T2D is well-established, 18-21 and elevated levels of proinflammatory cytokines TNF-α, IL-1, and IL-6Reference Eguchi and Nagai 20 , 22-25 are found in diabetic and obese individuals. Furthermore, interference with cytokine signaling pathways, for example, the introduction of an IL-1 receptor antagonist, has shown promising effects, which emphasize inflammation as a potential therapeutic target in T2D.Reference Larsen, Faulenbach and Vaag 26 Evidence suggests that biomarkers of inflammation are elevated already in the prediabetic state 27-29 and can be used to evaluate the progression of the disease.Reference Grossmann, Schmitt and Zeller 30 Due to the pleiotropic effects of GLP-1, it has been suggested that GLP-1RA may affect inflammatory pathways, 31-33 but the evidence is divergent. While a large clinical study did not find any impact of GLP-1RA on inflammation,Reference Courrèges, Vilsbøll and Zdravkovic 34 other studies found decreased levels of several proinflammatory biomarkers after GLP-1RA treatment, including TNF-α, IL-1, and IL-6, 35-39 and a preclinical study found higher concentrations of IL-1 and IL-6.Reference Shirazi, Palsdottir and Collander 40
In this study, we aim to further investigate the underlying mechanism of the beneficial effects of the GLP-1RA liraglutide in prediabetic, antipsychotic-treated patients with schizophrenia by measuring the serum concentrations of proinflammatory cytokines before and after 16 weeks of treatment. Additionally, we will compare proinflammatory cytokine levels from the prediabetic group with a control group comprised of normal glucose-tolerant (NGT) patients with schizophrenia. We hypothesize that proinflammatory biomarker concentrations are higher in prediabetes compared to NGT patients, and that treatment with liraglutide alleviates the inflammatory process.
Methods
Study design and population
The design and population of the original trial have been described previously.Reference JR, Vedtofte and MSL 16 , Reference Larsen, Vedtofte and Holst 41 In short, the study was a randomized double-blind, placebo-controlled trial (RCT) conducted from May 1, 2013, through February 25, 2016. At baseline, patients were randomized 1:1 to 16 weeks of treatment with subcutaneous liraglutide or placebo (saline injection) provided in prefilled pen injectors. Patients were diagnosed with a schizophrenia-spectrum disorder (schizoaffective disorder excluded) according to ICD-10 (International Statistical Classification of Diseases and Related Health Problems, 10th revision) or DSM-IV (The Diagnostic and Statistical Manual of Mental Disorders, 4th edition) and received stable treatment with the antipsychotic compounds clozapine or olanzapine for at least 6 months (without dose changes for at least 30 days). Eligible patients were 18 to 65 years old, fulfilled criteria of prediabetes, and had a body mass index (BMI) ≥27 kg/m2. Prediabetes was defined as either: (1) fasting plasma glucose (FPG) 6.1 to 6.9 mM or (2) glycated hemoglobin A1c (HbA1C) between 6.1% and 6.4% (43.2-46.4 mmol/mol) or (3) impaired glucose tolerance with a 2-hour plasma glucose (PG) above 7.8 mM during a 75-g oral glucose tolerance test (OGTT). Key exclusion criteria were T2D (HbA1C ≥ 6.5% or 47.5 mmol/mol), treatment with antidiabetic medications, substance abuse, or serious somatic illnesses (eg, pancreatitis or cardiovascular disease).
Screen failures for the RCT (patients without prediabetes) were included in the present study as a control group for baseline comparisons of prediabetic vs NGT patients.
Assessments
A detailed description of the assessments included in the RCT has been published previously.Reference JR, Vedtofte and MSL 16 At baseline following ≥10 hours of fasting, potential study participants underwent blood sampling, including HbA1C and a 75-g OGTT, as well as measurements of blood pressure, height, body weight, and waist circumference. The 4-hour OGTT value from the original investigation was not available for the NGT group, as the OGTT was stopped after 2 hours, if prediabetes was not confirmed at that time. Additionally, quality of life, disease severity, daily function, and alcohol use were assessed using different rating scales, that is, the Schizophrenia Quality of Life Scale,Reference Wilkinson, Hesdon and Wild 42 Clinical Global Impressions Severity Scale,Reference Busner and Targum 43 Global Assessment of Functioning,Reference Aas 44 and Alcohol Use Disorders Identification Test.Reference Saunders, Aasland, Babor, De La Fuente and Grant 45
Blood sampling
All study participants had to be fasting at the time of blood samples, which were taken from 9 to 10 AM. In the prediabetic group, fasting blood samples were taken at baseline and repeated after 16 weeks of treatment. For the NGT group, blood samples were only taken at baseline, since they were not eligible for enrollment in the trial. Blood serum was obtained after letting blood coagulate for a minimum of 1 hour, followed by centrifugation to separate serum from cellular components. Serum samples were immediately stored at −80°C during the baseline and end of treatment visit (week 16).
Analysis of samples
Serum samples were analyzed at the Institute of Immunology and Microbiology, University of Copenhagen, by use of multiplexed immunoassays (V-Plex Proinflammatory Panel 1 (human), Mesoscale Discovery© [MSD]). The assay was analyzed using the principle of electrochemiluminescence. The analytes in the panel included 10 cytokines: interferon-γ (IFN-γ), tumor necrosis factor-α (TNF-α), interleukin 1β (IL-1β), IL-2, IL-4, IL-6, IL-8, IL-10, IL-12p70, and IL-13. The assay is validated according to fit-for-purpose principlesReference Lee, Devanarayan and Barrett 46 and design control procedures of the company, that is, MSD. 47 For further details, including assay sensitivity, precision, and stability, please see the assay protocol. 47
Initially, an assay was tested for sensitivity. After this, samples were tested in separate 96-well plates in singlets. The first 8 wells (A1-A8) contained seven standard solutions plus a zero standard that generated a unique 8-point standard curve for each plate. To minimize the effects of variation, each plate contained samples from NGT patients as well as prediabetic patients treated with either liraglutide or placebo. The experiment included the following steps: adding samples, adding detection antibodies, adding reading buffer, and finally analyzing the plate with an MSD instrument that applies voltage to the plate. This process generates an emission of light from each well whose intensity provides a quantitative measure for each analyte. All experimental procedures were performed following the protocol provided by the manufacturer. 47
Data processing
Raw data from the assay were processed using the MSD DISCOVERY WORKBENCH® analysis software, version 4.0. The software calculates the concentrations of cytokines from the 8-point standard linear regression made for each plate and thereafter subtracts the background value. Since samples were tested in singlets, the coefficient of variation was not calculated.
Statistical methods
For statistical comparison, the analyses were divided into two subanalyses. One analysis investigated prediabetic vs NGT patients at baseline, and one analysis investigated the treatment effect of 16 weeks’ treatment with liraglutide vs placebo in prediabetic patients.
Baseline characteristics in the two groups in each subanalysis were compared using Student’s t-tests for continuous variables and χ 2-tests for categorical variables. The cytokine concentrations in NGT and prediabetic patients were compared using Student’s t-tests. The cytokine concentrations in the placebo and liraglutide group were compared individually using a mixed model analysis of covariance (ANCOVA). The model included the relevant biomarker as the dependent variable and treatment group as the independent variable. Covariates that differed significantly between the groups and the baseline value of the relevant biomarker were adjusted for in the ANCOVA.
All statistical tests were conducted with IBM SPSS Statistics 25. For all statistical tests, a P-value below .05 was considered significant. The level of significance was adjusted for multiple comparisons by the Bonferroni correction.
In cases of concentrations below the detection limit of the assay, the lowest measured concentration of the relevant biomarker in the same experimental group was imputed instead, since these reflect physiological concentrations more accurately than no measurements.Reference Kwak and Kim 48
Outliers were identified as values higher or lower than three times the interquartile range (IQR). These values were substituted with the highest or lowest value measured within 3 × IQR, according to the principle of winsorization.Reference Kwak and Kim 48 As with concentrations below the detection limit of the assay, this was done for each biomarker separately in each of the groups.
Results
Study population
Of the 97 patients that completed the original trial,Reference JR, Vedtofte and MSL 16 77 patients had their blood samples from the baseline and week 16 analyzed; 37 patients received liraglutide and 40 patients received placebo, see Figure 1. In the prediabetic group, samples from 20 patients (10 in each group) were not analyzed because of missing informed consent for further research use or due to insufficient sample material. In the NGT group, 13 patients were excluded for the same reasons. One patient was excluded after measurements because of antipsychotic treatment with neither clozapine nor olanzapine. Altogether 31 patients were included as NGT.
Figure 1. Flowchart of study population.
Measurements of IL-2 were excluded from the analysis because 33.5% of the measurements were below the assay detection limit. For all the cytokines, including IL-2, a total of 6.9% were below the detection limit. See Supplementary Tables for further details.
Baseline demographics
Prediabetes status
The patients in the NGT group (n = 31) were compared with all the prediabetic patients (n = 77) (Table 1). The prediabetes criteria, that is, impaired glucose tolerance during the 75 g OGTT, HbA1C, and FPG, were all significantly different for NGT vs prediabetic patients (2-h PG during the 75 g OGTT: 6.5 vs 10.0 mmol, P < .001; HbA1C: 5.4% vs 5.6% [35.2 vs 37.3 mmol/mol], P = .005; FPG: 5.2 vs 5.8 mmol, P < .001). Moreover, there was a significantly greater waist circumference in the prediabetic patient group (116.1 vs 108.9 cm, P = .009). Otherwise, the groups were comparable.
Table 1. Demographics and Clinical Characteristics at Baseline for Patients in the Diabetes Status Analysis and Patients in the Treatment Group Analysis.
Abbreviations: AUDIT: Alcohol Use Disorders Identification Test; BMI: body mass index; CGI-S: Clinical Global Impressions Scale-Severity; GAF: Global Assessment of Functioning; HbA1C: glycated hemoglobin type A1c; mM: mmol/l; OGTT: oral glucose tolerance test; SD: standard deviation; SQLS: Schizophrenia Quality of Life Scale.
a For statistical comparison, student’s t-tests were used for continuous variables and χ 2-test for categorical variables (sex, antipsychotic treatment).
b To convert HbA1C measurements to mmol/mol, please subtract 2.15 and multiply by 10.929.
c Scores range from 0 to 100, with higher scores indicating poorer quality of life.
d Scores range from 0 to 100, with higher scores indicating a higher function of daily living.
e Scores range from scores range from 0 to 7, with higher scores indicating illness severity.
f P < .05.
Treatment group
The patients in the placebo group (n = 40) were compared with the patients in the liraglutide group (n = 37) (Table 1). In the original study by Larsen et al., there were no statistically significant differences between the two treatment groups at baseline, except for antipsychotic treatment. A higher proportion of patients were treated with clozapine in the placebo group, and a higher proportion of patients were treated with olanzapine in the liraglutide group. This was still the case in the present analysis (P = .025).
Normal glucose tolerant group vs prediabetic group
The cytokines IFN-γ, IL-4, and IL-6 were significantly higher in the prediabetic group compared to the NGT group, and this remained significant after Bonferroni corrections (Table 2). IFN-γ showed a mean difference of 0.81 pg/ml (95% confidence interval [CI]: 0.23; 1.39; P = .001), IL-4 showed a mean difference of 0.01 pg/ml (95% CI: 0.01; 0.02; P < .001), and IL-6 showed a mean difference of 0.28 pg/ml (95% CI: 0.14; 0.41; P < .001). The cytokines TNF-α, IL-1β, IL-8, IL-10, IL-12p70, and IL-13 showed no differences between the groups (P = .102, P = .806, P = .381, P = .853, P = .411, and P = .269, respectively).
Table 2. Baseline Biomarker Concentrations in Normal Glucose Tolerance Group vs Prediabetic Group
Abbreviations: IFN, interferon; IL, interleukin; NGT, normal glucose tolerance; PD, prediabetes; SEM, standard error of the mean; TNF, tumor necrosis factor.
a Differences between normal glucose tolerance and prediabetes were statistically compared using an independent t-test. To adjust for multiple comparisons, Bonferroni correction was applied.
b The P-value remained significant after Bonferroni corrections with P < .006 (0.05/number of tests n = 9).
Placebo group vs liraglutide group
When comparing the estimated treatment differences for cytokine concentrations after 16 weeks of treatment between placebo and liraglutide groups, there were no significant differences for any of the nine cytokines (Table 3).
Table 3. Cytokine Concentrations in Placebo Group vs Liraglutide Group at Baseline and After 16 Weeks of Treatment
Abbreviations: IFN, interferon; IL, interleukin; SEM, standard error of the mean; TNF, tumor necrosis factor.
a Changes in concentrations from baseline to week 16 between liraglutide and placebo groups were statistically compared using a mixed model analysis of covariance. The model includes the baseline value of the relevant biomarker alongside antipsychotic treatment as a covariate.
Discussion
Serum concentrations of the proinflammatory cytokines: IFN-γ, TNF-α, IL-1β, IL-4, IL-6, IL-8, IL-10, IL-12p70, and IL-13 were measured in prediabetic patients with schizophrenia treated with clozapine or olanzapine before and after treatment with the GLP-1RA liraglutide.
Prediabetic patients had approximately twofold higher concentrations of cytokines IFN-γ, IL-4, and IL-6 compared to NGT patients, indicating increased inflammation already in the prediabetic state. The remaining six biomarkers did not differ significantly between the prediabetic and NGT group, though earlier studies show elevated levels of TNF-α and IL-1 associated with diabetes.Reference Akash, Rehman and Liaqat 23 , Reference Herder, Dalmas, Böni-Schnetzler and Donath 24 However, our findings are consistent with recent studies showing elevated levels of various proinflammatory cytokines, including IL-1, IL-4, IL-6, IL-10, and IL-13 27-29 , Reference Ligthart, Ghanbari and Brahimaj 49 in prediabetic patients in accordance with our present data in patients with schizophrenia-spectrum disorder. Interestingly, a large German cohort study with 15 000 individuals demonstrated gradually elevated biomarker profiles, including CRP, IL-1RA, and IL-18, across patients progressing from normoglycemia to prediabetes and diabetes.Reference Grossmann, Schmitt and Zeller 30 Our current results support the hypothesis that inflammation is involved in the development of T2D.
Although more prediabetic patients in the original sample became normal glucose tolerant (63.8% vs 16%), the intervention did not affect the serum levels of the cytokines significantly. These results do not support the hypothesis, that the beneficial effects of liraglutide demonstrated in the original trial can be partially explained by an antiinflammatory mechanism. However, earlier studies reported that GLP-1 receptor stimulation decreased plasma concentrations of IL-6 and TNF-α indicative of an anti-inflammatory effect.Reference Chaudhuri, Ghanim and Vora 37 , Reference von Scholten, Persson and Rosenlund 39 Additionally, Chaudhuri et al. found decreased mRNA concentrations of IL-1β and TNF-α in isolated mononuclear cells after treatment with GLP-1RA exenatide. Other studies with similar findings were not comparable since only acute effects of GLP-1RAs were investigated or because of more in-vitro approaches, for example, one study showed decreased concentrations of IL-6 after a 4-hour infusion of synthetic human GLP-1 (7-36 amide) in obese patients with T2D vs healthy controls.Reference Daousi, Pinkney, Cleator, Wilding and Ranganath 35 Another study found reduced levels of IL-1β and TNF-α and increased IL-10 after 4-hour exposure to exenatide; however, the measurements were performed in cultured macrophages from healthy nondiabetic individuals.Reference Bułdak, Machnik and Jakub Bułdak 36 An 8-week trial found that liraglutide reduced concentrations of IL-1β, IL-6, and TNF-α, measured by enzyme linked immunosorbent assay (ELISA) in blood mononuclear cells isolated from obese patients with T2D.Reference Hogan, Gaoatswe and Lynch 38 The sample size in our study may have been too small to detect differences, and the analyte was serum, which could explain the diverging results. However, our results are in alignment with findings from a large RCT, which showed no effect on IL-6 or TNF-α concentrations after 14 weeks of liraglutide treatment measured with ELISA in serum from 165 patients with T2D and albuminuria.Reference Courrèges, Vilsbøll and Zdravkovic 34
Before discarding the hypothesis of an antiinflammatory effect of GLP-1RAs, other explanations should be considered first. IL-6 is known to be a potent proinflammatory cytokine associated with T2D but has also been shown to induce GLP-1 production in pancreatic α-cells, possibly to compensate for the impaired insulin secretion in β-cells.Reference Akbari and Hassan-Zadeh 22 The specific tissue as a determining factor has also been demonstrated in a preclinical study by Shirazi et al., where central stimulation with GLP-1 agonist, exendin-4, resulted in increased expression of both IL-1β and IL-6 in the hypothalamus and hindbrain of rats, indicating that GLP-1 and these cytokines may be positively associated.Reference Shirazi, Palsdottir and Collander 40 Another possible explanation could be specific signaling mechanisms of the cytokines, for example, IL-6 can have proinflammatory effects if it acts through trans signaling but, on the other hand, have antiinflammatory effects if activated via classic signaling.Reference Akbari and Hassan-Zadeh 22 IL-4 is also known to have both proinflammatory effects in asthmaReference Steinke 50 and antiinflammatory properties concerning the pathogenesis of T2D.Reference Pirola and Ferraz 21 These examples illustrate the complexity in the workings of cytokines, which may explain why we did not find any differences in their peripheral levels, and thus further research in their tissue-specific and signaling-dependent effects is warranted.
The strengths of this study include the comparability of the groups, and although the sample size of prediabetic patients was relatively small, the equally small sample of NGT patients ensured that differences obtained were not a result of an overpowered control population. Furthermore, the multiplex assay is a well-tested and effective method for investigating many cytokines simultaneously. In addition, human serum was used and long-term changes over 16 weeks of intervention were examined, representing the most accurate in-vivo perspective.
The limitations of the study are as follows: First, the original RCT had a different primary aim than this present study, that is, testing the effects of liraglutide vs placebo in prediabetic patients with schizophrenia-spectrum disorders on stable treatment with clozapine or olanzapine. Second, the reduced sample size may reduce the power of the analyses. Thirdly, the selected group of patients limits the study’s generalizability to patients with schizophrenia and prediabetes. Another risk is that cytokine levels may have been confounded by the schizophrenia disorder itself and the antipsychotic medication. A growing effort is being dedicated in finding biomarkers for schizophrenia 51-54 with evidence indicating that there may be an association between inflammation and schizophrenia, but the relationship remains unclear.Reference Steiner, Bernstein and Schiltz 55 Similarly, antipsychotics, including clozapine and olanzapine, have been shown to both increase and decrease inflammatory biomarker levels.Reference Li, Peng and Li 56 , Reference Romeo, Brunet-Lecomte, Martelli and Benyamina 57 In our study, there is a risk that a preexisting proinflammatory state in patients with schizophrenia additionally reinforced by antipsychotic medication may have hindered the GLP-1 treatment to exert an adequate antiinflammatory response. Ideally, this uncertainty might have been ameliorated by cytokine measurements in larger samples of healthy controls and patients with prediabetes only, whereas antipsychotics’ impact on proinflammatory cytokine levels would require patients with schizophrenia naïve to antipsychotic treatment and measurements following treatment initiation.
Conclusion
Prediabetic vs NGT patients with schizophrenia treated with clozapine or olanzapine had increased serum concentrations of proinflammatory cytokines, IFN-γ, IL-4, and IL-6. These findings further substantiate the link between inflammation and T2D and are in concordance with the presented hypothesis. Treatment with GLP-1RA liraglutide did not affect serum levels of the investigated cytokines, which is in discordance with the last part of our hypothesis, namely that GLP-1RAs ameliorate prediabetic states by way of their potential antiinflammatory effects. Further testing of this hypothesis in larger numbers of individuals and with more control groups is needed.
Disclosures
Dr. Marco Møller reports receiving a 1-year pregraduate research grant from Lundbeckfonden during conducts of the study to cover salary expenses.
Dr. Vilsbøll reports receiving lecture fees from Amgen, AstraZeneca, Boehringer Ingelheim Pharmaceuticals, Bristol-Myers Squibb, Eli Lilly and Company, Merck Sharp & Dohme, Novo Nordisk A/C, Sanofi, and Zealand Pharma and serving on the advisory boards of AstraZeneca, Boehringer Ingelheim Pharmaceuticals, Bristol-Myers Squibb, Eli Lilly and Company, Merck Sharp & Dohme, Novo Nordisk A/C, and Sanofi.
Dr. Fink-Jensen reports sponsoring the study and receiving an unrestricted research grant from Novo Nordisk A/S.
Simon Fredholm, Mathias Ebbesen Jensen, Gitta Wörtwein, Julie Rask Larsen, and Niels Ødum have declared that there are no conflicts of interest in relation to the subject of this study.
Supplementary Materials
To view supplementary material for this article, please visit http://doi.org/10.1017/S1092852920002217.
