Greater glucagon-like peptide-1 responses to oral glucose are associated with lower central and peripheral blood pressures

Brachial blood pressure was measured three times with the participant in a sitting position after 10 min of rest (Omron M6 comfort, Omron Healthcare, Milton Keynes, UK) and reported as an average. Aortic pulse wave velocity (PWV) and central blood pressure were assessed with the participant in a supine position after 10 min of rest using a SphygmoCor apparatus (version 8, Atcor Medical, West Ryde, NSW, Australia) and a tonometer, which captured wave forms at the carotid, femoral and radial artery. Using simultaneous ECG monitoring, the time from the R wave to the arrival of the pulse wave at the carotid and femoral artery, respectively, was captured. An anthropometer (Seca, Medical Scales and Measuring Systems, Hamburg, Germany) was used to measure the distance from the suprasternal point to the carotid artery recording site and from the suprasternal point to the femoral artery recording site to avoid overestimation of the distance. The ratio between the distance and the time defined the PWV. The PWV was measured twice between the right carotid and right femoral artery and in case of a difference of more than 0.5 m/s, a third measure was taken. It was reported as an average of the two closest measures. Peripheral pressure waveforms assessed at the radial artery were used to estimate central blood pressure [17]. The average of two measures were used.

A 3-point standard 75 g oral glucose tolerance test (OGTT) was performed, with blood sampling before (0 min) and 30 and 120 min after glucose intake. Plasma glucose was measured using one of two different methods; Hitachi 912 system (Roche Diagnostics, Mannheim, Germany) or Vitros 5600 system (Ortho Clinical Diagnostics, Illkirch Cedex, France). Subsequently, Vitros values were converted to Hitachi values [18]. HbA1c was measured by high-performance liquid chromatography (TOSOH G7, Tokyo, Japan) [18]. Plasma levels of total GLP-1 (intact and metabolites) were measured in EDTA plasma stored at − 80 °C using RIA with analytical detection limit around 1 pmol/l and with intra- and interassay coefficients of variation of 6.0% and 15%. To diminish methodological variances, all samples were analysed after study completion during 2 consecutive months. Furthermore, quality controls and batches for all reagents were identical for all sets [18].

GLP-1 response was defined as the incremental area under the curve (iAUC), from start of the OGTT to 120 min after glucose intake, using the GLP-1 measurements before glucose intake and after 30 and 120 min. Using the trapezoid rule, iAUC was calculated as total area under the curve subtracted the fasting GLP-1 value × 120 min (participants with iAUC ≤ 0 are excluded from analyses). Associations between GLP-1 release and central haemodynamics were analysed in separate models using linear regression analysis. Prior to analysis, iAUC for GLP-1(iAUC_GLP-1) was log₂-transformed to fulfil the requirement of a normal distribution of the model residuals. All analyses were adjusted for age and sex. Analyses of PWV were additionally adjusted for mean blood pressure and heart rate at time of PWV measurement. Additional analyses adjusting for age, sex and either iAUC insulin, fasting plasma glucose, BMI or HOMA-IR were performed for association between iAUC GLP-1 and central and brachial blood pressures. The study population was categorized into quartiles according to the iAUC GLP-1 response and the lowest and highest quartiles were compared with regards to blood pressure measures (adjusting for age and sex). Sub-analysis of the study population was made dividing the population into a normo glucose (fasting glucose < 5.6 mmol/l) group and a pre-diabetes risk (fasting glucose ≥ 5.6 mmol/l) group and association to blood pressures were investigated (adjusting for age and sex). Analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA) and figures were created in R version 3.2.2 (The R Foundation for Statistical Computing). A P-value of < 0.05 was considered significant.

The participants were divided into quartiles according to their iAUC GLP-1 response. The group with the highest quartile of GLP-1 response was associated with significant lower central systolic (− 4.94 (95% CI − 8.56 to − 1.31) mmHg) and diastolic (− 3.05 (95% CI − 5.29 to − 0.80) mmHg) blood pressure as well as significant lower brachial systolic (− 5.18 (95% CI − 8.94 to − 1.42) mmHg) and diastolic (− 2.96 (95% CI − 5.26 to − 0.67) mmHg) blood pressure, compared to the group with the lowest quartile of GLP-1 response.

Several parameters have been proposed to affect the secretion and concentration of GLP-1 including glucose [19], fat [19], bile acids [20], inflammatory markers [21, 22], 3-deoxyglucosone [23] (generated from carbohydrates) and the presence of obesity [24] and type 2 diabetes [18].

We find that there is an association between a greater GLP-1 response during an OGTT and lower central and peripheral blood pressure. Our results remain significant even after adjusting for iAUC insulin, fasting glucose, and for most analyses also for BMI and HOMA-IR, which are all factors known to have an impact on GLP-1 secretion. Interestingly, when dividing participants into quartiles according to their GLP-1 response we found that the group with the highest GLP-1 response was associated with approximately 5 mmHg lower blood pressures compared to the group with the lowest GLP-1 response. Many individuals with obesity and/or type 2 diabetes have decreased GLP-1 secretion [18, 24, 25] and these individuals also often have increased blood pressure [26, 27]. Furthermore, weight loss increases GLP-1 response and decreases blood pressure [25, 28]. Thus, according to our results, it could be speculated whether the decreased GLP-1 response could contribute to the increased blood pressure in this population. Together with our analyses adjusted for several parameters, this implies that endogenous GLP-1 may have an effect on blood pressure in a dose–response dependent manner with the greater the GLP-1 response the lower the blood pressure as also observed for the dose-response dependently lowering of blood glucose [29, 30].

We observe that the quartile of individuals with the highest GLP-1 response have a ~ 5 mmHg lower blood pressure compared to the quartile of individuals with the lowest GLP-1 response. A large meta-analysis found that the degree of blood pressure reduction was proportional to the risk of major cardiovascular disease events, stroke, heart failure, and all-cause mortality [31]. Interestingly, a reduction of 4.5 mmHg in systolic blood pressure has been associated with lower risk of cardiovascular event, all-cause mortality and cardiovascular mortality [32]. Thus our finding of a 5 mmHg difference in blood pressure between GLP-1 response groups is of clinical relevance with regards to mortality. In comparison, DPP-4 inhibitors reduce systolic blood pressure by 2.5–2.8 mmHg [33‐35] and GLP-1 RAs reduce systolic blood pressure by 1.2–4 mmHg [4, 5, 36]. GLP-1 RAs [4, 5, 37] have, furthermore been found to reduce risk of major adverse cardiac events which has not been observed for DPP-4 inhibitors [38‐40].

The GLP-1 receptor has been located on endothelial cells in the vasculature [41] and a direct effect of GLP-1 on vascular smooth muscle has also been proposed [42]. This may be one of the mechanisms by which GLP-1 RA treatment decreases blood pressure. Endothelial dysfunction, characterized by reduced nitric oxide (NO) production, is a precursor of atherosclerosis and a risk factor for CVD [43]. GLP-1 treatment promotes vasodilation by production of NO [44]. Insulin has also been shown to induce NO production [45], however, the association in our study remained significant when adjusting the analyses for insulin (see Fig. 2), indicating that GLP-1 works independently of insulin on the blood pressure-lowering mechanism. Infusion of GLP-1 in physiological doses exerted vasodilatory actions, decreased diastolic blood pressure and increased heart rate whereas systolic blood pressure was unchanged in a study of 26 healthy individuals [15]. Furthermore, treatment with GLP-1 RAs have shown to improve arterial stiffness in type 2 diabetes [46], which is augmented by hypertension [47]. Similarly, treatment with DPP-4 inhibitors have shown to improve intima-media thickness [48], arterial stiffness [49] and diastolic dysfunction [50, 51], which are all factors associated with blood pressure [52, 53]. In our study of individuals without diabetes and hypertension we did not find an association between GLP-1 response and PWV (Table 2), a measurement of arterial stiffness, or central pulse pressure (Fig. 2), which is determined by arterial stiffness [47]. To further support the effect of GLP-1 RAs on endothelial cells, liraglutide has also been found to have anti-restenotic effects mediated by endothelial NO [54]. Endothelial cells do not only play a role in NO production, they also influence on the angiogenesis process [55] and liraglutide has been found to promote angiogenesis after stroke [56] opening up for the possibility that GLP-1-based treatment potentially could be applied as a neuroprotective drug as well as a cardioprotective, anti-hyperglycemic and anti-obesity drug.

Only few studies have previously analysed the association between endogenous GLP-1 levels and cardiovascular disease. In agreement with our findings, an inverse correlation (r = − 0.26, P = 0.0085) between GLP-1 secretion and systolic blood pressure was found by Yoshihara et al. [57]. However, active GLP-1 (the levels of which are most often below detection levels of the assays) rather than total GLP-1 was measured in this study of 103 individuals with an ELISA kit. Another study reported a positive association between levels of GLP-1 and coronary artery disease [58]. However blood samples were gathered randomly, i.e. not necessarily in a fasting state, from 303 persons and GLP-1 levels (active GLP-1 and metabolites) were measured by ELISA in serum. Since random blood sampling is a major limitation to the study above comparability to other studies is likely to be difficult. Furthermore, it is worth noting that the specificity, sensitivity, and validity can vary considerably between the ELISA kits available for GLP-1 analyses [59].

Our physiological study includes more than 800 participants making its size a major strength, outnumbering previous studies [57, 58]. Total GLP-1 (not active/intact GLP-1) was measured with the use of a well-documented RIA with high specificity and sensitivity [60]. Measuring GLP-1 in this manner provides more information than solely measuring intact GLP-1 as GLP-1 metabolites also are included in the analyses [18]. Because of the rapid metabolism of the hormone, estimation of its secretion must be made with an assay for total GLP-1. A limitation of this study, however, is the cross-sectional design, allowing us to find associations between GLP-1 and central and peripheral haemodynamic measures whereas we cannot conclude on causality. However, our findings are in agreement with observations from GLP-1 treatment studies on lowering hypertension and cardiovascular mortality [4, 5]. Haemodynamic measurements were collected only in a fasting state. Therefore, it was not possible to investigate the acute effects of GLP-1 secretion on blood pressure and PWV. Other studies have found no change [61] or increase [62] in blood pressure after acute infusion of GLP-1. In general, there does not seem to be consensus about whether blood pressure changes during an OGTT. One study found an elevation in systolic blood pressure of more than 4 mmHg after glucose load [63] while other studies find either small increases [64] or no differences [65]. Lastly, our sub-analysis dividing participants into a normal fasting glucose group and a pre-diabetes group did not reveal any significant differences between the groups, which may be due to the lower number of participants in each group.