Article Text

Ceramides and phospholipids are downregulated with liraglutide treatment: results from the LiraFlame randomized controlled trial
  1. Emilie H Zobel1,
  2. Asger Wretlind1,
  3. Rasmus S Ripa2,
  4. Viktor Rotbain Curovic1,
  5. Bernt J von Scholten1,3,
  6. Tommi Suvitaival1,
  7. Tine W Hansen1,
  8. Andreas Kjær2,
  9. Cristina Legido-Quigley1,4,
  10. Peter Rossing1
  1. 1Steno Diabetes Center Copenhagen, Gentofte, Denmark
  2. 2Department of Clinical Physiology, Nuclear Medicine & PET and Cluster for Molecular Imaging, Rigshospitalet, Copenhagen, Denmark
  3. 3Novo Nordisk AS, Bagsvaerd, Denmark
  4. 4King’s College, London, UK
  1. Correspondence to Dr Emilie H Zobel; emilie.hein.zobel{at}regionh.dk

Abstract

Introduction Treatment with glucagon-like peptide-1 receptor agonists (GLP-1 RAs) can reduce risk of cardiovascular disease (CVD) in persons living with type 2 diabetes, however the mechanisms explaining this cardiovascular benefit are still debated. We investigated changes in the plasma lipidome following treatment with the GLP-1 RA liraglutide.

Research design and methods In a double-blind placebo-controlled trial, we randomized 102 persons with type 2 diabetes to liraglutide or placebo for 26 weeks. Fasting blood plasma was collected at baseline and at end-of-treatment. The lipidome was measured using liquid-chromatography-coupled mass-spectrometry as a secondary end point in the study. Treatment response of each lipid was tested with lipid-specific linear mixed-effect models comparing liraglutide with placebo. Bonferroni p<7.1e-03 was employed. The independence of the findings from clinical covariates was evaluated with adjustment for body mass index, HbA1c, fasting status, lipid-lowering treatment and change in lipid-lowering treatment during the trial.

Results In total, 260 lipids were identified covering 11 lipid families. We observed significant decreases following liraglutide treatment compared with placebo in 21 lipids (p<7.1e-03) from the following lipid families: ceramides, hexocyl-ceramides, phosphatidylcholines, phosphatidylethanolamines and triglycerides. We confirmed these findings in adjusted models (p≤0.01). In the liraglutide-treated group, the individual lipids were reduced in the range of 14%–61% from baseline level, compared with 19% decrease to 27% increase from baseline level in the placebo group.

Conclusions Compared with placebo, liraglutide treatment led to a significant downregulation in ceramides, phospholipids and triglycerides, which all are linked to higher risk of CVD. These findings were independent of relevant clinical covariates. Our findings are hypothesis generating and shed light on the biological mechanisms underlying the cardiovascular benefits observed with GLP-1 RAs in outcome studies, and further strengthen the evidence base for recommending GLP-1 RAs to prevent CVD in type 2 diabetes.

Trial registration number NCT03449654.

  • glucagon-like peptide 1
  • lipids
  • diabetes mellitus
  • type 2

Data availability statement

Data are available upon reasonable request. The datasets analyzed during the current study are not publicly available due to the risk of patient re-identification. De-identified participant data or anonymised clinical study reports can be obtained from the first author upon reasonable request. Necessary data protection agency and ethical committee approvals must be provided in compliance with relevant legislation.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Significance of this study

What is already known about this subject?

  • Glucagon-like peptide-1 receptor agonists (GLP-1 RAs) can reduce the high risk of cardiovascular disease in persons living with type 2 diabetes.

  • The mechanism behind this effect is largely unknown.

What are the new findings?

  • This is the first randomized placebo-controlled clinical trial to assess GLP-1 RAs’ effect on the plasma lipidome using untargeted liquid-chromatography-coupled mass-spectrometry.

  • The present results suggested that treatment with the GLP-1 RA liraglutide led to a downregulation in ceramides, phospholipids and triglycerides.

  • Ceramides, phospholipids and triglycerides are lipids linked to risk of cardiovascular disease.

How might these results change the focus of research or clinical practice?

  • These results are hypothesis generating and may inspire research that aims to understand the cardiovascular mode-of-action of GLP-1 RAs to further investigate GLP-1 RAs’ lipid-regulating effect.

Introduction

Recent outcome studies have demonstrated that treatment with glucagon-like peptide-1 (GLP-1) receptor agonists can reduce the high risk of cardiovascular disease in persons living with type 2 diabetes.1–4 To date, there is no convincing evidence that points to a single mechanism as dominant for this risk reduction.5

Favorable within-trial changes have been demonstrated in traditional cardiovascular risk factors including glycemia, body weight, systolic blood pressure, urinary albumin-to-creatinine ratio and low-density lipoprotein (LDL)-cholesterol. A recent analysis of The Liraglutide Effect and Action in Diabetes: Evaluation of Cardiovascular Outcome Results trial evaluated these cardiovascular risk factors as potential mediators of the effect of the GLP-1 receptor agonist liraglutide on major cardiovascular events.6 The analysis indicated potential mediation by HbA1c and to a lesser extent by urinary albumin-to-creatinine ratio. However, the authors concluded that HbA1c was unlikely to be a true mediator based on the totality of evidence including other cardiovascular outcome trials, where the choice of therapies used to reduce HbA1c, and not just HbA1c reduction by itself, have important impact on cardiovascular outcomes.6

A mediation analysis can only evaluate candidate mediators that were measured during the trial. LDL-cholesterol was not found to be a mediator, but the lipid metabolism associated with type 2 diabetes and cardiovascular disease is complex, and other lipids play direct roles in development of atherosclerosis. With new lipidomic technologies using mass spectrometry, it is now achievable to study the whole range of lipids (lipidome) in a biological system.7 Lipidomics has been applied to discover new disease biomarkers and to unravel mechanisms and pathways underlying various diseases and therapies.

In a randomized, double-blind, placebo-controlled trial, we analyzed the lipidome before and after treatment with the human GLP-1 receptor agonists liraglutide as a secondary end point. Our aim was to explore downstream effects of liraglutide on lipid metabolism that could link to the cardiovascular benefit.

Research design and methods

Study design and participants

The trial was conducted between October 26, 2017 and August 16, 2019 at the Steno Diabetes Center Copenhagen. We included men and women with type 2 diabetes in a randomized, double-blind, placebo-controlled, parallel-group trial. Participants were recruited from the outpatient clinic at the Steno Diabetes Center Copenhagen and through newspaper advertisements.

Inclusion criteria were as follows: age >50 years; HbA1c ≥ 48 mmol/mol (6.5%); estimated glomerular filtration rate ≥30 mL/min/1.73 m2 (estimated by Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) formula); stable glucose-lowering and cholesterol-lowering treatment (minimum 4 weeks). Key exclusion criteria were: type 1 diabetes; chronic or previous acute pancreatitis; other treatment (90 days prior to screening) which in the investigator’s opinion could interfere with the effect of liraglutide (including oral glucocorticoids, calcineurin inhibitors, dipeptidyl peptidase 4 inhibitors, GLP-1 receptor agonists); cancer or disorders (except for conditions associated with type 2 diabetes history), which could interfere with the results of the trial; clinical signs of diabetic gastroparesis; previous bowel resection; impaired liver function; inflammatory bowel disease; weight >150 kg (full list see online supplemental materials).

Supplemental material

The primary end point of the trial was change in vascular inflammation assessed using ¹⁸F-fluorodeoxyglucose PET/CT,8 and here we report results from the secondary end point of change from baseline to week 26 in the lipidome for the liraglutide-treated group compared with placebo. The protocol for this analysis was developed before the last participant was enrolled in the study.

Randomization and masking

All participants were randomized in a 1:1 ratio to daily subcutaneous injections of liraglutide or matching placebo for 26 weeks. All investigators and participants were blinded to treatment allocation. Treatment pens and computer-generated allocation sequence were provided by Novo Nordisk (Bagsvaerd, Denmark) and two clinicians not otherwise involved in the trial handled the treatment allocation. EHZ enrolled participants and assigned participants to interventions.

Procedures

Starting dose was 0.6 mg/day and was escalated to 1.2 mg/day after 1 week and to 1.8 mg/day after 2 weeks. Maintenance-dose was 1.8 mg/day. We allowed the dose escalation to be flexible in order to reach the maximum tolerated dose for each participant. Participants were instructed to be fasting for 4 hours prior to the study visits. Blood samples were collected at randomization and after 26 weeks treatment and were stored at −80°C until analysis, which was 4 months after last patient last visit.

Lipid quantification

Plasma samples were prepared and analyzed at the Steno Diabetes Center Copenhagen. Lipids were extracted using a modified Folch extraction method9 and were analyzed by ultra-high-performance liquid chromatography coupled with quadrupole time-of-flight mass spectrometry. Samples were analyzed in a randomized order, with QC pooled plasma samples. Details can be found in the online supplemental materials.

The acquired data were preprocessed using MZmine 2 (V.2.28)10 and postprocessed in R V.3.5.1 (R Core Team, Vienna, Austria) by normalization to internal standards, batch correction, truncation of outliers, imputation of missing values and log10 transformation.11

Statistical analysis

The lipidome analysis was prespecified and hypothesis generating. The sample size was a consequence of the sample size of the main study and no specific sample size estimation was done for this substudy.

Clinical data were assessed with SAS software (V.9.4; SAS Institute, Cary, North Carolina, USA). All data analysis and visualization related to the lipidome was done with R (V.3.5.1; R Core Team).

For the clinical data, continuous variables are given as means with SD, non-normally distributed variables are summarized as medians with IQR and categorical variables are reported as percentages. Unpaired t-test and χ2 test was used to test differences in clinical characteristics between the two groups at baseline, and unpaired t-test was also used to compare change from baseline to end-of-treatment in HbA1c, weight and clinical lipid measurements between the two groups.

To identify which lipid levels changed in response to treatment, linear mixed-effect models were built for each individual lipid species to explain observed standardized levels with time, treatment and their interaction as fixed effects and participant ID as random effect. A full list of lipids, estimated coefficients and p values was reported. Lipid class-level changes were assessed with the same approach of modeling on the average level in each lipid class. The independence of the findings from clinical covariates was evaluated with adjustment for body mass index (BMI), HbA1c, fasting status, lipid-lowering treatment and change in lipid-lowering treatment during the trial by adding each of these covariates as additional independent variables to the linear mixed-effect model. Using similar approach, we evaluated the independence of the findings from the baseline measurement of total triglycerides.

Models were fitted using the limma12 and lme4 packages.13 Nominal associations at p<0.05 were reported for exploratory associations at the lipid class level also reported as family level. Bonferroni p<7.1e-03 was employed for the single lipid linear mixed-effect models. Results were visualized as a lipidome-wide heatmap using the LipidomeR package.14

Selected lipid species (n=21) with a response to the treatment in the linear mixed-effect models were investigated further: for each lipid species, relative change from baseline to end-of-treatment in each of the two arms was calculated. The result was visualized as a diverging bar graph with the ggplot2 package.15 Baseline lipid levels were compared between participants with and without a history of cardiovascular disease. Standardized mean difference of the selected lipids was calculated using the ‘effsize’ package.16 Linear regression was used to investigate associations between the selected individual lipids and sex, age, BMI, HbA1c and lipid-lowering treatment at baseline. Mediation analysis was performed with linear regression models, the significance and the tested mediation effect was calculated with the ‘mediation’ package in R,17 using bootstrapping to simulate 500 samples.

Data availability statement

The datasets analyzed during the current study are not publicly available due to the risk of patient re-identification. De-identified participant data or anonymized clinical study reports can be obtained from the first author on reasonable request. Necessary data protection agency and ethical committee approvals must be provided in compliance with relevant legislation.

Results

Participants

The study included 102 participants, and all were eligible for lipidomic analysis. Five participants did not have blood samples available for lipidomic analysis at end-of-treatment: one participant randomized to liraglutide dropped out of the study after experiencing gastrointestinal side effects, and one participant randomized to liraglutide discontinued the study treatment due to gastrointestinal side effects, remained in the study, but did not wish to have blood samples taken at the end-of-treatment visit. Three participants randomized to placebo did not have blood samples taken at end-of-treatment: one dropped out of the study due to concurrent cancer disease, one discontinued study treatment after experiencing shoulder pain in relation to the first study visit, remained in the study, but did not wish to have blood samples taken at the end-of-treatment visit, and one did not show up for the visit where the end-of-treatment blood samples were taken due to work-related stress. These five participants were only included in the baseline analysis.

The population included 16 (15.7%) women and the mean (SD) age was 66 (8.2) years, median (IQR) known diabetes duration was 10.9 (5.7; 18.2) years, HbA1c 58.4 (10.1) mmol/mol and BMI 29.9 (4.6) kg/m2. The baseline clinical characteristics of the 51 participants receiving liraglutide and the 51 receiving placebo were balanced (table 1), except for triglycerides (mean (SD): liraglutide 2.1 (1.2) vs placebo 1.6 (0.78) mmol/L, p=0.01). A history of cardiovascular disease (defined as history of acute myocardial infarction, percutaneous coronary intervention, coronary artery bypass graft, stroke, peripheral arterial thrombosis, claudication and/or nitroglycerin requiring angina pectoris) was reported in 23 (22.6%) participants at baseline (balanced between groups, p=0.24).

Table 1

Characteristics of the participants at baseline

Mean changes (95% CI) over the 26 weeks for the liraglutide-treated group compared with placebo in body weight were −3.7 (−4.8 to −2.6) vs −0.18 (−0.76 to 0.40) kg (p<0.0001); and HbA1c −5.1 (−8.0 to −2.0) vs −0.08 (−1.9 to 1.7) mmol/mol (p=0.006).8 Mean changes (95% CI) for the liraglutide-treated group compared with placebo for clinical lipid measurements were as follows: LDL-cholesterol 0.06 (−0.17 to 0.30) vs −0.08 (−0.29 to 0.13) mmol/L (p=0.37);8 HDL-cholesterol 0.06 (0.004 to 0.12) vs 0.05 (−0.02 to 0.11) mmol/L (p=0.72); total cholesterol 0.08 (−0.18 to 0.35) vs −0.03 (−0.26 to 0.21) mmol/L (p=0.54); triglycerides −0.11 (−0.44 to 0.21) vs 0.01 (−0.19 to 0.21) mmol/L (p=0.52).

Lipids reduced by liraglutide treatment compared with placebo

In total, 260 lipids were identified (figure 1, online supplemental table S1) covering 11 lipid families including ceramides, diglycerides, hexocyl-ceramides, lactosyl ceramides, lysophosphatidylcholines, lysophosphatidylethanolamines, phosphatidylcholines, phosphatidylethanolamines, phosphatidylinositoles, sphingomyelins and triglycerides. Of the 260 lipids, 21 were significantly decreased after liraglutide treatment compared with placebo (figure 1, online supplemental table S1). These 21 lipids were from the following lipid families: ceramides, hexocyl-ceramides, phosphatidylcholines, phosphatidylethanolamines and triglycerides. We observed a pattern with noticeably reduction in short ceramides and large, poly-unsaturated triglycerides (figure 1). The size of the reductions was ranging from 14% to 61% from baseline level, in the liraglutide-treated group, compared with 19% decrease to 27% increase from baseline level in the placebo group, as visualised in figure 2. We confirmed the reduction in these 21 individual lipids in models adjusted for BMI, HbA1c, fasting state (4 were not fasting for baseline measurements, and 3 were not fasting for end-of-treatment measurements), lipid-lowering treatment at baseline (yes or no) and change in lipid-lowering treatment during the study period (n=9) (p≤0.01).

Figure 1

Lipidome-wide responses to liraglutide treatment compared with placebo. Lipid species grouped in panels according to lipid families (title of each panel). X-axis is the number of fatty-acid double bonds for the specific species (level of unsaturation) and Y-axis is the number of carbon atoms (total fatty acid chain length), and therefore each cell in the heat maps represents one lipid species (one individual lipid). Color indicates the coefficients from lipid-specific linear mixed-effect models. Blue colors represent lipids that are reduced in level by liraglutide treatment compared with placebo. Red colors represent lipids that are increased in level by liraglutide treatment compared with placebo. *p<0.05; **p<0.01. Cer, ceramides; DG, diglycerides; HexCer, hexocyl-ceramides; LacCer, lactosyl-ceramides; LPC, lysophosphatidylcholines; LPE, lysophosphatidylethanolamines; PC, phosphatidylcholines; PE, phosphatidylethanolamines; PI, phosphatidylinositoles; SM, sphingomyelins; TG, triglycerides.

Figure 2

Percentage change after 26 weeks in average amount of individual lipids and in the traditional clinical lipid measurements following liraglutide and placebo treatment. Included are the 21 individual lipids significantly reduced by liraglutide compared with placebo and the traditional clinical lipid measurements, including HDL-cholesterol, LDL-cholesterol, triglycerides and total cholesterol. The traditional clinical lipid measurements were not significantly reduced by liraglutide compared with placebo treatment. Cer, ceramides; HDL, high-density lipoprotein; HexCer, hexocyl-ceramides; LDL, low-density lipoprotein; PC, phosphatidylcholines; PE, phosphatidylethanolamines; TG, triglycerides.

To provide a context for the 21 lipids that changed significantly in the liraglutide-treated group, we examined how they were related to selected baseline characteristics. At baseline, there were few correlations between these 21 individual lipids and sex, age, BMI, HbA1c and lipid-lowering treatment (online supplemental table S2). The baseline levels of these 21 individual lipids were generally higher in participants with versus without a history of cardiovascular disease (figure 3).

Figure 3

Baseline level of individual lipids for participants with versus without a history of cardiovascular disease. Shown are selected lipids that were significantly reduced by liraglutide compared with placebo treatment for participants with (n=23) vs without a history of cardiovascular disease (CVD). Levels are log10 transformed. Levels were generally higher for participants with a history of cardiovascular disease, although none of the differences were significant. A history of cardiovascular disease was defined as a history of acute myocardial infarction, percutaneous coronary intervention, coronary artery bypass graft, stroke, peripheral arterial thrombosis, claudication and/or nitroglycerin requiring angina pectoris. Cer, ceramides; PC, phosphatidylcholines; PE, phosphatidylethanolamines; TG, triglycerides.

We also analysed changes on a lipid family level for the liraglutide-treated group compared with placebo, and observed significant decreases in 5 of the 11 lipid families namely ceramides (26 weeks average reduction liraglutide vs placebo: −27% vs 8%, p=0.01), hexocyl-ceramides (−28% vs −10%, p=0.02), phosphatidylcholines (−5% vs 2%, p=0.04), phosphatidylethanolamines (−14% vs 7%, p=0.004) and triglycerides (−27% vs 14%, p=0.03).

The primary end point of the LiraFlame trial, change in vascular inflammation assessed using ¹⁸F-fluorodeoxyglucose PET/CT, was unchanged.8

Sensitivity analysis

In a sensitivity analysis, we adjusted for the clinical measurement of total triglycerides at baseline, since this level was not balanced between the two groups, despite randomization. The change in the 21 lipid levels in the liraglutide-treated group remained significant compared with placebo after this adjustment, the highest adjusted p value was 0.009 (PE(O-34:2)/PE(P-34:1)).

Moreover, we performed a sensitivity analysis, where zero change from baseline to end-of-treatment was assumed for the five participants with lipidomics data missing at end-of-treatment. Overall results were similar, but six single lipids (TG(58:10); TG(60:11); Cer(d38:1); TG(54:6); TG(55:4) and PE(O-34:2)/PE(P-34:1)) were no longer significantly reduced by liraglutide treatment compared with placebo after multiple adjustment (online supplemental table S3).

Mediation analysis

In a mediation analysis including the top 10 most associated lipids, we tested if the observed effect of liraglutide on these lipids were mediated through change in BMI or HbA1c. None of the p values for mediation effect were significant (p≥0.06; online supplemental table S4).

Discussion

In this randomized clinical trial, we demonstrate that individual ceramide and phospholipid lipid species were reduced in the liraglutide-treated group compared with placebo. In addition, we confirmed a reduction in individual triglyceride lipid species in the liraglutide-treated group and we were able to demonstrate that this was primarily a downregulation of large, poly-unsaturated triglycerides. In support of this, when analyzing the lipid species as lipid families, we observed significant reductions in the ceramide, hexocyl-ceramide, phosphatidylcholine, phosphatidylethanolamine and triglyceride lipid family in the liraglutide-treated group compared with placebo. Moreover, mediation analyses indicated that the lipid-regulating effect of liraglutide was likely not mediated by change in BMI or HbA1c.

Ceramides, phospholipids and triglycerides are lipids identified as biomarkers—and potential causes—of cardiovascular disease. Given the robustness and magnitude of the reduction in these pro-atherogenic biomarkers observed, we consider that our findings are interesting, and that liraglutides effect on lipids should be explored further and could improve our understanding of the mechanisms explaining the cardiovascular protection observed with human GLP-1 receptor agonists.

We investigated 260 lipids of which 21 was significantly reduced in the group treated with liraglutide compared with placebo. In the discussion, we focus only on the lipids that changed.

Liraglutide, ceramides and atherosclerosis: a pathway to reduced risk of cardiovascular disease?

As a novel finding, we demonstrated that ceramide levels were reduced in the group treated with liraglutide. Ceramides are a class of bioactive lipids implicated in cardiovascular disease and with interesting similarities to LDL-cholesterol. Ceramides are present in the serum in much lower concentrations than cholesterol (approximately 1/1000) and can be measured with sensitive techniques such as mass spectrometry. Like cholesterol, ceramides (the backbone of all sphingolipids) accumulate in atherosclerotic lesions.18 Experimental data have revealed how ceramides are formed at the surface of atherogenic lipoproteins via sphingomyelinase activity.19 20 An increased ceramide content promotes lipoprotein aggregation, which in turn promotes the subendothelial retention or trapping of lipoproteins within the vessel wall, a key event in early atherogenesis.19 20 Ceramides are involved in the transcytosis of lipoproteins across the endothelium21 and are further perceived as an important second messenger in several aspects of the inflammatory process.18 Several studies have demonstrated that higher circulating levels of ceramides are risk factors for future cardiovascular events in apparently healthy individuals, in individuals with known coronary artery disease as well as in persons with diabetes.22–28 Ceramides were recently reported as part of the Cardiovascular Event Risk Test (CERT) and the derivative CERT2 risk-score developed and validated in large cohorts by Hilvo et al as a simple cardiovascular risk estimation score based on ceramides and phospholipids.29 30

In ApoE−/− mice and rabbits fed a high fat and cholesterol diet, inhibiting the biosynthesis of ceramides ameliorates atherosclerosis and arterial stiffness.31 32 Evidence that therapeutic interventions can reduce ceramides in humans is sparse. The PREDIMED trial recently revealed a strong association between ceramides, and risk of cardiovascular disease that was mitigated by a Mediterranean dietary intervention.33

Liraglutide and phospholipids

We report significant reductions for the liraglutide-treated group compared with placebo in single-lipid phospholipids from the two lipid families phosphatidylethanolamines and phosphatidylcholines. In comparisons with ceramides, the body of literature that links phospholipids to cardiovascular disease is limited, but emerging. A study in a case-cohort (n=3779) subset from the Action in Diabetes and Vascular Disease: Preterax and Diamicron-MR trial demonstrated that higher levels of phospholipids among other lipids were associated with risk of cardiovascular morbidity and mortality28 and as already mentioned, phospholipids were part of the recently published CERT2 cardiovascular risk estimation score.29

Liraglutide and large triglycerides

Others have demonstrated, before us, that liraglutide lowers the triglyceride level measured as a traditional clinical lipid.5 34 35 We identified 88 distinct triglyceride lipid species, which enabled us to add new information that especially large, poly-unsaturated triglycerides were reduced in the liraglutide-treated group. The precise mechanism behind liraglutide’s effect on triglycerides has yet to be determined. One hypothesis that has been put forward is that GLP-1 receptor signaling mediates a decreased secretion of apoB48-containing chylomicron particles in the intestinal mucosa, which subsequently reduces the intestinal absorption of triglycerides.34 36

The overall size of triglyceride reduction with liraglutide has been reported as modest when triglycerides were measured as a traditional clinical lipid.5 35 37 Our data indicate that this modest reduction could cover more pronounced reductions in large, poly-unsaturated triglycerides together with less pronounced effects on other triglyceride lipid species. Our findings were independent of changes in BMI and thus weight changes cannot explain this finding.

It remains controversial that triglyceride reduction could translate into a substantial reduced cardiovascular risk. A long-standing association exists between triglycerides and cardiovascular disease in observational studies, but, whether the triglyceride level is a risk marker for development of cardiovascular disease or more directly promote cardiovascular disease has been discussed for decades.38 Mendelian randomization studies support a casual role of triglycerides for risk of cardiovascular disease.39 The many randomized clinical trials using medications that lower triglycerides, such as fibrates, niacin and omega-3 fatty acids, have failed to show conclusive evidence of further cardiovascular risk reduction after LDL-cholesterol levels were ‘optimally controlled’.40 However, a positive outcome has recently been reported in the Reduction of Cardiovascular Events with Icosapent Ethyl–Intervention Trial (REDUCE-IT).41 In REDUCE-IT, the omega-3 fatty acid product icosapent ethyl reduced the cardiovascular risk regardless of the triglyceride level attained, contributing to the ongoing debate whether triglyceride lowering per se reduces the cardiovascular risk and pointing out to the importance of determining single lipid biochemistry.

Strengths and limitations

Our study is hypothesis generating. Only similar analysis of a cardiovascular outcome trial could provide evidence that liraglutide prevents cardiovascular events by decreasing ceramides and phospholipids. By using lipidomics we were able to investigate liraglutide’s possible effect on a broad panel of lipid families and individual lipids. The strengths of our study include the double-blinded, randomized, placebo-controlled design and the robust technology applied to measure a comprehensive panel of lipids. Participants were instructed to fast for 4 hours. We can speculate that 12–14 hours overnight complete dietary restriction would give lower overall levels of the lipids measured, including triglycerides, but as we are measuring changes from baseline within individuals, we do not think this impacted our findings. Despite the randomized design, we observed a higher baseline triglyceride level in the liraglutide-treated group compared with placebo. As we are measuring changes from baseline within individuals, we do not think this impacts our findings for the single-lipid triglycerides, but we cannot rule out that it is easier to clear more triglycerides with higher levels. We confirmed our findings in a sensitivity analysis, adjusted for the clinical measurement of total triglycerides at baseline. Lipidomics is a tool to study fingerprints with hundreds of lipids but currently it is platform-dependent and does not detect all the lipids in the lipidome. Other relevant variables (ie, glycemic control, body weight) were improved in the liraglutide-treated group, and therefore it is not possible to assume a direct relationship between the observed changes in the single lipids and liraglutide therapy. In order to evaluate if the effect on the lipidome is a direct effect of liraglutide, or secondary to weight loss or improved glycemic control following liraglutide treatment, an active comparator should have been included (ie, insulin as a comparator for liraglutides effect on glycemic control or diet-induced weight loss as a comparator for liraglutides effect on body weight).

Conclusion

Ceramides, phospholipids and triglycerides were downregulated in the group treated with liraglutide compared with placebo. Lipids which all are linked to risk of cardiovascular disease. This lipid-regulating effect of liraglutide should be examined further and may contribute to the cardiovascular benefits observed in outcome studies.

Data availability statement

Data are available upon reasonable request. The datasets analyzed during the current study are not publicly available due to the risk of patient re-identification. De-identified participant data or anonymised clinical study reports can be obtained from the first author upon reasonable request. Necessary data protection agency and ethical committee approvals must be provided in compliance with relevant legislation.

Ethics statements

Patient consent for publication

Ethics approval

The study was approved by the local ethics committee (H-16044546) and the Danish Medicines Agency (2016110109) and was performed in compliance with the principles of the Declaration of Helsinki.

Acknowledgments

The authors are grateful to all the participants and acknowledge the work of study nurse L. Jelstrup and laboratory technicians D. Riis, T. R. Juhl and J. A. Hermann (Steno Diabetes Center Copenhagen, Gentofte, Denmark).

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Presented at Presented at EASD 2020 and EDNSG 2021.

  • Contributors EHZ, AW, RSR, BJvS, TS, TWH, AK, CL-Q and PR contributed to study design and data interpretation. EHZ recruited participants to the study. EHZ, VRC and TWH contributed to running of the study and data collection. AW performed the lipidomics sample analysis. AW and TS processed and analyzed the lipidomics data. AW, TS and CL-Q interpreted the lipidomics data. EHZ performed statistical analysis of the clinical data. EHZ drafted the manuscript and the final version was critically reviewed and approved by all authors. Guarantor statement: EHZ is the guarantor of the 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.

  • Funding The study was funded by Novo Nordisk and Skibsreder Per Henriksen, R. og hustrus fund. Steno Diabetes Center Copenhagen and Department of Clinical Physiology, Nuclear Medicine & PET, Rigshospitalet have provided internal funding.

  • Competing interests AW, VRC, TS and CL-Q declares no competing interests. AK has received consultancy fees from Novo Nordisk. RSR, BJvS, TWH and PR have shares in Novo Nordisk and BJvS and EHZ is now an employee of Novo Nordisk, but work related to this article was conducted while EHZ was employed by Steno Diabetes Center Copenhagen. PR has received the following: consultancy and/or speaking fees (to Steno Diabetes Center Copenhagen) from AbbVie, Astellas, AstraZeneca, Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, Eli Lilly, MSD, Novo Nordisk and Sanofi Aventis; research grants to institution from AbbVie, AstraZeneca and Novo Nordisk.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.