The impact of prediabetes and diabetes on endothelial function in a large population-based cohort

Abstract

Background

Diabetes and prediabetes are well-recognized risk factors for cardiovascular disease (CVD) and are marked by vascular endothelial dysfunction (ED). However, there is a scarcity of thorough population-based studies examining ED in individuals with diabetes/prediabetes free from manifest CVD. Here, we examined the association between ED assessed by reactive hyperaemia index (RHI) in the finger and diabetes/prediabetes in a large middle-aged population cohort.

Methods

Within the Malmö Offspring Study, following the exclusion of participants <30 years and participants with prevalent CVD, 1384 participants had complete data on all covariates. The RHI was calculated using pulse amplitude tonometry. ED was defined as RHI < 1.67. Multivariable logistic and linear regression models were conducted to investigate associations between ED and RHI with diabetes and prediabetes.

Results

The study population had a mean age of 53.6 ± 7.6 years (53% women). In study participants with manifest diabetes (n = 121) and prediabetes (n = 514), ED was present in 42% and 25% respectively, compared to 23% in those with normal glucometabolic status. In multivariable logistic regression analyses, prevalent diabetes was significantly associated with ED (OR 1.95; 95%CI 1.57-3.39; p = 0.002), as well as with lower RHI (β-coeff. -0.087; p = 0.002). However, prediabetes showed no association with neither ED nor RHI.

Conclusion

In a population free from CVD, vascular endothelial dysfunction was primarily associated with manifest diabetes, but not with prediabetes, implying that finger ED may develop when diabetes is established, rather than being an early sign of glucose intolerance. Further research is needed to explore whether addressing glucose intolerance could potentially delay or prevent vascular ED onset.

Plain language summary

What is the context?

Diabetes and prediabetes are known to increase the risk of cardiovascular disease (CVD) through a condition called vascular endothelial dysfunction (ED). However, there is a lack of comprehensive studies on ED in individuals with diabetes/prediabetes who do not already have CVD. In this study, we investigated the association between ED, assessed using the reactive hyperaemia index (RHI) in a finger, and diabetes/prediabetes in a large group of middle-aged individuals.

What is new?

We conducted this study within the Malmö Offspring Study, involving 1384 participants who were over 30 years old and did not have pre-existing CVD. The average age of the participants was 53 years, with 53% being women. Among those with diagnosed diabetes (121 individuals) and prediabetes (5141 individuals), 42% and 25% respectively showed signs of ED, compared to 23% in those with normal glucose metabolism. In our analyses, we found that established diabetes was significantly associated with ED, as well as with lower finger RHI values. However, prediabetes did not show any significant association with either ED or RHI.

What is the impact? In a healthy population without pre-existing CVD, vascular endothelial dysfunction was predominantly linked to diagnosed diabetes, rather than prediabetes. This suggests that ED may develop once diabetes is established, rather than being an early indicator of glucose intolerance. Further research is necessary to investigate whether addressing glucose intolerance could potentially delay or prevent the onset of vascular ED.

Keywords:

• Cardiovascular disease
• Diabetes
• Endothelial dysfunction
• Prediabetes
• Reactive hyperaemia index

Introduction

Cardiovascular disease (CVD) represents the primary cause of both mortality and morbidity in individuals with type 2 diabetes mellitus (T2DM), accounting for over half of all fatalities.¹ The impact of CVD in T2DM is substantial, encompassing both macrovascular and microvascular complications such as myocardial infarction, hypertension, peripheral vascular disease, retinopathy, kidney failure, and neuropathy.² The endothelium assumes a central role in the pathophysiology of CVD in T2DM by regulating key processes including inflammation, coagulation, vasodilation, and oxidative stress.^3–6 The occurrence of endothelial dysfunction (ED) in T2DM⁷ is well-documented, manifesting when there is a reduction in nitric oxide production or availability, or an imbalance in endothelial factors regulating relaxation and vasoconstriction. This impairs the endothelium’s ability to respond to shear stress brought about by alterations in blood flow. ED has also been observed in prediabetes^8–11, but existing studies are primarily based on relatively small sample sizes and in individuals with concurrent CVD. Further, the evaluation of ED in T2DM has been conducted using a diverse range of methods.

By the use of peripheral arterial tonometry (PAT), the endothelial function can be assessed by measuring finger arterial dilation in response to post-occlusion ischemia (reactive hyperaemia index (RHI)).¹² A proposed RHI cut-off value of <1.67 indicates endothelial dysfunction^13–15 and has been found to independently predict future cardiovascular events in T2DM with albuminuria and chronic kidney disease.¹⁶ In addition, findings suggest that decreased RHI is a powerful predictor for the development of T2DM and prediabetes, regardless of other established cardiovascular risk factors for diabetes.¹⁷ This implies that vascular ED may be detected prior to the onset of diabetes, and may therefore precede, rather than being solely a consequence of T2DM and prediabetes.

Timely recognition of individuals exhibiting ED, including those in the prediabetes stage, holds promise for identifying individuals susceptible to CVD development. Therefore, within the scope of this study, we aim to investigate the association between T2DM and prediabetes with ED in a sizable cohort of individuals free from CVD, encompassing varying ages and risk CVD profiles.

Methods

Study population

The Malmö Offspring study (MOS) is a population-based study that includes adult participants, including children (generation 2) and grandchildren (generation 3) of individuals previously examined within the framework of the Malmö Diet and Cancer–Cardiovascular Cohort (MDC-CC; generation 1; n= 6103).^18,19 Data collection in the MOS study was carried out between years 2013 to 2021, rendering a study sample of 5269 individuals (2884 children (generation 2); 2392 grandchildren (generation 3), and 3 individuals that were both children/grandchildren to any of participants in MDC-CC). Endothelial function was assessed in generation 2 only. Glycated haemoglobin (HbA1C) was added to the test protocol in 2015. After the exclusion of participants below the age of 30 years based on our prior study demonstrating that low RHI in young, healthy individuals may not consistently signify genuine ED but rather an artifact,²⁰ participants with prevalent CVD and missing values on relevant co-variates, 1384 eligible participants remained. (Figure 1).

Figure 1. Flowchart of the study population.

The Regional Ethical Review Board in Lund has provided ethical approval for this study (DNR. 2012/594). All participants provided informed consent, and the study adheres to the principles outlined in the Helsinki Declaration for ethical research involving human subjects.

Assessment of endothelial function

Endothelial function was assessed using the EndoPat® device (Itamar, Israel). The procedure involved placing a cuff on the non-dominant upper arm and placing the index or middle fingers in pneumo-electric tubes (probes) to record continuous arterial pulsatile volume changes. After a 5-minute period of rest, the cuff was inflated to 60 mmHg above baseline systolic blood pressure, but not less than 200 mmHg. If necessary, the pressure was increased to a maximum of 300 mmHg. After occlusion of 5 minutes, the pressure of the cuff was released, and the subsequent arterial dilation, assessed as an increase in the signal amplitude, was recorded for another 5 minutes. The RHI was calculated as the ratio of post-occlusion to pre-occlusion signal amplitudes.

Assessment of arterial stiffness

Arterial stiffness was assessed utilizing the Sphygmocor device (AtCor, Australia). This assessment involves measuring carotid-femoral pulse wave velocity (c-f PWV) along the abdominal aorta. Pulse curves are obtained from both the carotid and femoral arteries. The device then calculates the time elapsed from the R-wave, recorded by an ECG, to the onset of the pulse wave in both the carotid and femoral arteries. The distance, determined as a direct line from the carotid to the femoral artery and then multiplied by 0.8, is manually inputted into the program in accordance with established guidelines. This process is performed twice, and if the discrepancy between the two measurements exceeds 0.5 m/s, a third measurement is conducted. The final c-f PWV value used in analysis represents the average of these measurements.¹⁸

Clinical assessment

The clinical assessment included data on medical history and lifestyle which was obtained through a self-administered questionnaire. Information regarding prevalent cases of myocardial infarction, heart failure, or heart valve surgery, as well as previous instances of atrial fibrillation and stroke, was obtained from regional and national registers utilizing the relevant ICD-10 codes. Weight (kg) and height (m) were measured in light indoor clothing to calculate BMI (kg/m²). Resting blood pressure (mmHg) was measured as a mean of two readings in the supine position after 5 min rest by an automatic device (Omron^®). Prevalent diabetes was defined as either self-reported diagnosis of type 2 diabetes, use of antidiabetic medication recorded in Swedish Prescribed Drug Register, or fasting plasma glucose ≥7 mmol/L on two separate occasions, or HbA_1c ≥48 mmol/mol. Prevalent prediabetes was defined as a fasting plasma glucose between 5.6 and 6.9 mmol/L, or HbA_1c between 39 and 47 mmol/mol.²¹ Prevalent CVD was defined as having either prior myocardial infarction, prior heart failure or heart valve surgery, or prior stroke. Blood samples were collected after an overnight fast and stored at -80 °C. Smoking was defined as self-reported current smoking. Hypertension was defined as any of systolic blood pressure >140 mmHg, diastolic blood pressure >90mmHg, or antihypertensive treatment. Data on antihypertensive treatment (AHT) were obtained through the Swedish Prescribed Drug Register and were defined as use of any antihypertensive agent.

Laboratory

Upon study visit, fasting blood samples were collected. Participants were advised to fast from 10 p.m. the night before. Subsequently, the fasting blood samples are analysed for lipids, glucose, HbA_1c, creatinine and cystatin-C at the Department of Clinical Chemistry, Malmö. Estimated glomerular filtration rate mean (eGFR_mean) was based on cystatin-C and creatinine, where relative GFR estimates based on creatinine and cystatin-C were calculated. The average of these two estimates is reported as eGFR_mean.²²

Statistics

The study participants were divided into two groups based on the prevalence of ED (RHI <1.67), or not, and the differences in continuous variables between the groups were compared using Students t-test for normally distributed variables or Mann Whitney U-test for skewed variables. Pearson’s Chi-square test was used to compare categorical variables. RHI in participants that are normoglycemic, participants with prediabetes and participants with diabetes was compared using Tukey’s HSD post hoc test. The results are presented as means (standard deviation; SD) and medians (25-75 interquartile range). Univariate linear regression analyses were deployed to explore associations between age, SBP, BMI, PWV, fasting plasma glucose, and RHI (dependent variable). Univariate logistic regressions were deployed to explore associations between fasting plasma glucose, eGFR, age, smoking status, and ED. Variables with a skewed distribution (fasting plasma glucose, RHI and eGFR) were ln-transformed prior to analyses. Prior to analyses of associations between prediabetes and ED, subjects with diabetes were excluded. Logistic regressions were utilized to investigate the associations between prediabetes and diabetes with ED prevalence. Adjustment for age and sex was performed in Model 1. Further adjustment for smoking, BMI, systolic blood pressure (SBP), c-f PWV, low-density lipoprotein cholesterol (LDL-c), eGFR and AHT was carried out in Model 2. Linear regressions were carried out to analyse the associations between prevalent DM and prediabetes with RHI, adjusted according to the models described above. The statistical analysis was conducted using SPSS version 28.0, and a p-value of <0.05 was considered statistically significant.

Results

Patient characteristics

The study population had a mean age of 53.6 (±7.6)years and approximately half (53%) were women. ED was present in 354 subjects (25.6%). Subjects with ED had higher BMI, but lower systolic and diastolic blood pressure, higher prevalence of diabetes. Further, subjects with ED had higher eGFR, higher fasting plasma glucose and HbA_1c, but lower HDL-C, and were treated with antihypertensive agents to a greater extent (Table 1). Figure 2 illustrates the relationship between RHI and HbA_1c in participants with and without diabetes. Population characteristics stratified on glycaemic status (normoglycemic/prediabetes/diabetes) are presented in Supplementary Table S1.

Figure 2. Scatterplot illustrating the relationship between RHI and HbA1c in participants with and without diabetes.

Table 1. Characteristics of study participants.

	All participants	Participants without ED	Participants with ED
	n = 1384	n = 1030	n = 354	p
CLINICAL
Age (years)	53.6 (±7.6)	53.4 (±7.6)	54.28 (±7.8)	0.112
Sex (women)	735 (53.1)	555 (53.9)	180 (50.8)	0.323
BMI (kg/m^2)	27.0 (±4.7)	26.7 (±4.7)	27.8 (±4.7)	<0.001
SBP (mmHg)	121 (±16)	122 (±17)	118 (±13)	0.001
DBP (mmHg)	76 (±10)	77 (±10)	75 (±8)	<0.001
Hypertension (n(%))	354 (25.6)	267 (25.9)	87 (24.6)	0.617
PWV (m/s)	8.2 (±1.5)	8.3 (±1.6)	8.1 (±1.5)	0.176
Prediabetes (n(%))	514 (37.1)	386 (37.5)	128 (36.2)	0.529
Diabetes (n(%))	121 (8.7)	70 (6.8)	51 (14.4)	<0.001
Diabetes type (n(%))				0.905
Type 2	74 (61.1)	43 (61.4)	31 (60.2)
Type 1	2 (0.02)	1 (0.01)	1 (0.01)	–
LADA	2 (0.02)	1 (0.01)	1 (0.01)	–
Secondary	1 (0.01)	0 (0.01)	1 (0.02)	–
Unknown	42 (34.7)	24 (34.2)	18 (35.3)	–
Smoking (n(%))	172 (12.4)	123 (11.9)	49 (13.8)	0.350
AHT (n(%))	268 (14.4)	175 (13.0)	93 (17.8)	<0.001
Beta receptor blocking agents (n(%))	165 (11.9)	112 (10.9)	53 (15)	0.040
Calcium channel blockers (n(%))	128 (9.2)	88 (8.5)	40 (11.3)	0.123
RAS inhibitors (n(%))	209 (15.1)	138 (13.4)	71 (20.1)	0.003
RHI	2.00 (1.65-2.48)	2.24 (1.97-2.65)	1.46 (1.37-1.58)	<0.001
LABORATORY
LDL-c (mmol/L)	3.5 (±0.9)	3.5 (±0.9)	3.4 (±1.0)	0.215
HDL-c (mmol/L)	1.7 (±0.5)	1.7 (±0.5)	1.6 (±0.5)	<0.001
eGFR (L/min/1.73m^2)	80 (73-87)	80 (73-87)	84 (75-91)	0.436
FPG (mmol/L)	5.3 (5.0-5.8)	5.3 (4.9-5.8)	5.4 (5.0-5.9)	0.011
HbA1c (mmol/mol)	35 (33-38)	35 (33-38)	36 (33-39)	0.005

Values are means (±standard deviations), medians (25-75 interquartile range) or numbers (%). BMI – body mass index; SBP – systolic blood pressure; DBP – diastolic blood pressure; LDL-c – low density lipoprotein cholesterol; LADA – latent autoimmune diabetes of adults; HDL-c – high density lipoprotein cholesterol; PWV – pulse wave velocity; eGFR –estimated glomerular filtration rate, Lund-Malmö revised equation²²; AHT – antihypertensive treatment; RHI – reactive hyperaemia index; FPG – fasting plasma glucose; HbA_1c – glycated haemoglobin.

Association between diabetes and prediabetes, and endothelial dysfunction

In univariate linear regressions, RHI was positively associated with PWV (0.010, p < 0.041), but negatively with BMI (-0.011, p < 0.001), and fasting plasma glucose (-0.185; p < 0.001). In univariate logistic regression analyses, fasting plasma was associated with ED (OR 3.20; 95%CI 1.73-5.94; p < 0.001,), but no association with ED was observed, neither for age (OR 1.01; 95%CI 0.99-1.03; p = 0.112), nor smoking (OR 1.19; 95%CI 0.83-1.69; p = 0.350). In the multivariable logistic regression analyses, prevalent DM exhibited associations with ED (OR 1.95; 95%CI 1.26-2.99, p = 0.002; Table 2). However, no significant associations were found between prediabetes and ED (Table 2). Similarly, in the multivariable linear regression models, prevalent diabetes demonstrated an inverse association with RHI (β-coeff. -0.087; p = 0.002; Table 3) whereas prediabetes exhibited no significant associations with RHI, when RHI was treated as a continuous variable.

Table 2. Associations between prediabetes and diabetes and endothelial dysfunction (finger RHI <1.67).

	Diabetes		Pre-diabetes
	OR (95%CI)	p	OR (95%CI)	p
Unadjusted	2.31 (1.57-3.39)	<0.001	1.09 (0.84-1.41)	0.529
Model 1	2.21 (1.50-3.27)	<0.001	1.07 (0.82-1.39)	0.629
Model 2	1.95 (1.26-2.99)	0.002	1.07 (0.81-1.41)	0.624

Values are odds ratios (OR) and 95% confidence intervals. Model 1 is adjusted for age and sex. Model 2 is adjusted for age, sex, BMI, smoking status, systolic blood pressure, carotid-femoral pulse-wave velocity, low-density lipoprotein cholesterol, estimated glomerular filtration rate and antihypertensive treatment.

Table 3. Associations between prediabetes and diabetes with reactive hyperaemia index as a continuous variable.

	Diabetes		Pre-diabetes
	B	p	B	p
Unadjusted	−0.125	<0.001	−0.015	0.377
Model 1	−0.119	<0.001	−0.016	0.344
Model 2	−0.087	0.002	−0.015	0.379

Values are unstandardized beta coefficients (β). Model 1 is adjusted for age and sex. Model 2 is adjusted for age, sex, BMI, smoking status, systolic blood pressure, carotid-femoral pulse-wave velocity, low-density lipoprotein cholesterol, estimated glomerular filtration rate and antihypertensive treatment.

Univariate associations between specific antihypertensive agents and ED are presented in Supplementary Table S2, where significant positive association were observed between beta-receptor blocking agents and agents acting on the renin-angiotensin system and ED, but no association was observed for calcium-channel blockers and ED.

Discussion

In this large cross-sectional, population-based, observational study consisting of participants free from CVD, we established that ED, defined as finger RHI < 1.67, is predominantly associated with prevalent type 2 diabetes. Conversely, individuals with impaired glucose tolerance (commonly referred to as prediabetes) have not yet exhibited notable signs of ED.

To maintain vascular homeostasis, endothelial cells release several molecules with the ability to alter the vascular tone. The presence of vascular risk factors such as hypertension, diabetes mellitus and smoking damages the endothelial cells and thereby also lowering NO synthesis which has previously been reported to decrease during ageing.³⁴ Previous findings showing that infusion of NO synthase inhibitor dampened the PAT hyperaemic response suggest that the increased pulse amplitude following hyperaemic flow has to be partly dependent on NO bioavailability.³⁵ The unexpected observation of higher systolic blood pressure in individuals without ED as compared to individuals with ED is most likely attributed to a higher use of antihypertensive agents in individuals with hypertension (13% vs. 18%), a treatment effect that reduces the PAT response.

In accordance with previous studies, we were able to demonstrate that metabolic risk factors for CVD such as higher BMI and diabetes were associated with lower RHI.³³ Oxidative stress and lower endothelium-derived nitric oxide in obesity and diabetes have previously been associated with lower vessel distensibility and structural remodelling.^36–38 Additionally, diabetes creates an environment of heightened oxidative stress and inflammation, generating harmful molecules that further impair endothelial function.^39,40 This dysfunction contributes to a prothrombotic state, increasing the risk of blood clots. Furthermore, the endothelium’s barrier function is compromised, potentially leading to vessel leakage and tissue damage. Lastly, diabetes interferes with the formation of new blood vessels, hindering tissue repair.⁴¹ These mechanisms collectively underpin the development of cardiovascular complications and microvascular issues seen in diabetes.

In this study, the presence of ED was associated with lower PWV which appears contradictory. One potential explanation for this unexpected outcome could be that individuals with ED might have been prescribed medications that influence arterial stiffness, potentially lowering PWV without corresponding impact on the endothelial function assessed through RHI. Moreover, PWV measurement reflects the overall stiffness of larger arteries, indicating systemic arterial stiffness. Conversely, RHI evaluates localized endothelial function, in this case the fingertip’s microvasculature. This localized assessment might pinpoint specific areas or smaller vessels where ED persists despite the observed improvement in systemic arterial stiffness indicated by lower PWV. This paradoxical scenario arises where improved systemic arterial stiffness (reflected in lower PWV) might not correspond to a similar enhancement in localized endothelial function assessed by RHI. Another possible explanation can possibly be found in the shortcomings of the PAT-methodology, itself. The use of PAT has been widely studied and is considered a useful method to assess ED, at least in middle-aged or elderly subjects, or in patients with established disease. However, this cohort consisted of younger and middle-aged individuals (30-50 years) free from manifest CVD.

The inverse associations observed in the current study between PAT response and some cardiovascular risk factors suggest that its usefulness for risk stratification in healthy, middle-aged, or younger patients may still be limited. Further studies are warranted to fully comprehend whether CV risk factors impair endothelial nitric oxide bioavailability and thereby PAT response.

We believe that the observed associations between beta-receptor blocking agent and agents acting on the renin-angiotensin system usage and ED may stem from the frequent coexistence of these two conditions.⁴²

Strengths and limitations

A notable strength of our study lies in the large population-based cohort consisting of middle-aged participants free from manifest CVD, not often screened with PAT in previous studies that mostly recruited patients. Further, we could account for a comprehensive set of cardiovascular risk factors, including the assessment of arterial stiffness in our analyses. Assessment of PAT response in the current study is limited to small arteries in the fingertip. However, arterial physiology and mechanisms of vessel distensibility may vary considerably between different vascular beds which affects the interpretation of the clinical utility. One limitation to consider is that in a middle-aged population free from CVD, the technical device and its findings may not necessarily indicate genuine ED. Instead, they could be influenced by other age-related factors, such as small artery diameter, finger volume, heightened sensitivity to stress during examination, and occlusion-related discomfort, potentially resulting in a condition referred to as "pseudo-ED”, found in younger subjects, which we reported on in a previous study.²⁰ To mitigate this potential confounding factor, individuals under the age of 30 were excluded from the study, as they have been shown to exhibit manifestations consistent with "pseudo-ED”.²⁰ Further, the precision of prediabetes definition might be an limitation, since it is based on a single glucose measurement, potentially contributing to a non-significant association between RHI and prediabetes. This study is cross-sectional by nature, limiting the capacity to make causal inferences. Finally, the study is confined to a single centre, thereby restricting the generalizability of findings to diverse populations and ancestral groups.

Conclusion

In this population-based study, endothelial dysfunction was mainly associated with established diabetes, but not prediabetes, in middle-aged subjects free from cardiovascular disease. This suggests that finger ED may develop as diabetes progresses, rather than being an early sign of glucose tolerance issues. Further research is needed to confirm causality and explore whether addressing prediabetes could potentially delay or prevent endothelial dysfunction onset.

Declaration of Interests Statement

The authors have no conflicting interests to report.

Data availability statement

The dataset for this manuscript cannot be made publicly available because of ethical and legal restrictions. Requests to access the dataset should be directed to the Chair of the Steering Committee for the Malmö cohorts, see instructions at https://www.malmo-kohorter.lu.se/english.

Additional information

Funding

Hjärt-Lungfonden; Hulda och E Conrad Mossfelts Stiftelse för Vetenskaplig Forskning Inom Hjärt- och Kärlsjukdomarnas Område; Region Skåne; Vetenskapsrådet; The Malmö Offspring Study (MOS) has been funded by the Research Council of Sweden (Grant 521-2013-2756), the Heart and Lung Foundation (Grant 20150427), and by funds obtained from the local Region Skane County Council (ALF) to PM Nilsson. In addition, funding has been obtained from Ernhold Lundströms Stiftelse, donations to the Skane University Hospital, a research grant from Oral Health Related Research by Region Skane to Malmo University, Faculty of Odontology, Periodontal Department, and the European Research Council. Amra Jujic received funding from Hulda and E Conrad Mossfelt foundation.