Type 2 diabetes and HbA_1c are independently associated with wider retinal arterioles: the Maastricht study

We used data from The Maastricht Study, an observational, prospective population-based cohort study. The rationale and methodology have been described previously [10]. In brief, the study focuses on the aetiology, pathophysiology, complications and comorbidities of type 2 diabetes, and is characterised by an extensive phenotyping approach. All individuals aged between 40 and 75 years and living in the southern part of the Netherlands were eligible to participate. Participants were recruited through mass media campaigns and from the municipal registries and the regional Diabetes Patient Registry by postal mailing. Recruitment was stratified according to known type 2 diabetes status, with an oversampling of individuals with type 2 diabetes for reasons of efficiency. The present report includes cross-sectional data from the first 3451 participants, who completed the baseline survey between November 2010 and September 2013. The baseline examinations of each participant in the study were performed within a time window of 3 months (except for some participants in whom fundus photography was initially unavailable or in whom photos were of low quality; in these participants, fundus photography was obtained later; see below). The study was approved by the medical ethical committee of the Maastricht University Medical Center (NL31329.068.10) and the Minister of Health, Welfare and Sports of the Netherlands (permit 131088-105234-PG). All participants gave written informed consent. From the initial 3451 participants included, those with types of diabetes other than type 2 diabetes were excluded (n = 41). Of the remaining 3410 participants, retinal microvascular diameter data were available for 2924 participants, 48 of whom had data missing for one or more covariates. The main reasons for missing data were logistic (no equipment, no trained researcher available or technical failure), contraindications for the eye drops or fundus photographs of insufficient quality. The retinal microvascular diameter study population thus consisted of 2876 participants (ESM Fig. 1); fundus photography was obtained within the time window of 3 months in 2700 participants and after a mean of 47 months (range 34–57) after the date on which the retinal measurement was planned in 176 participants.

To assess glucose metabolism status, all participants (except those who used insulin) underwent a standardised 2 h 75 g OGTT after an overnight fast. For safety reasons, participants with a fasting glucose level above 11.0 mmol/l, as determined by a finger prick test, did not undergo the OGTT. For these individuals, fasting glucose level and information about diabetes medication use were used to assess glucose metabolism status. Glucose metabolism status was defined according to the WHO 2006 criteria as normal glucose metabolism (NGM, fasting glucose <6.1 mmol/l; 2 h postload glucose <7.8 mmol/l), impaired fasting glucose and/or impaired glucose tolerance (combined as prediabetes, fasting glucose 6.1–7.0 mmol/l or 2 h postload glucose 7.8–11.1 mmol/l) and type 2 diabetes (fasting glucose ≥7.0 mmol/l or 2 h postload glucose ≥11.1 mmol/l) [11].

All participants were asked to refrain from smoking and drinking caffeine-containing beverages for 3 h before the measurement. Participants were allowed to consume a light meal (breakfast or lunch) low in fat content at least 90 min before the start of the measurement [12]. For retinal measurements, fundus photography of both eyes was performed 15 min after the pupils had been dilated with tropicamide 0.5% and phenylephrine 2.5% (wt/vol.).

All fundus photographs were taken with an auto-focus, auto-shot and auto-tracking fundus camera (Model AFC-230; Nidek, Gamagori, Japan) in an optic disc-centred field of view of 45° in a darkened room. Static retinal vessel analysis (one image of the left or right eye was randomly chosen per participant) was performed using the retinal health information and notification system (RHINO) software developed by the RetinaCheck group of the Technical University of Eindhoven (Eindhoven, the Netherlands) [13, 14]. Optic disc detection and arteriole/venule classification were corrected manually. Retinal vessel diameters were measured at 0.5–1.0 disc diameter away from the optic disc margin and were presented as central retinal arteriolar equivalent and central retinal venular equivalent (CRAE and CRVE, respectively) in measurement units (MU). The scale factor is based on the optic disc diameter, which is assumed to be 1800 μm [15], i.e. 1 MU = 1 pixel size× 1800 μm/pixel size of optic disc diameter. CRAE and CRVE represent the equivalent single-vessel parent diameter for the six largest arterioles and largest venules in the region of interest, respectively. The calculations were based on the improved Knudtson–Parr–Hubbard formula [16].

Fundus photographs of insufficient quality, e.g. obstructed by lashes or defocused, were evaluated and discussed with a second observer and excluded on mutual agreement. We calculated the intraclass correlation coefficients for CRAE and CRVE to assess the agreement between analyses of the RHINO software with vs without manual identification of arterioles and venules using 2556 images. The intraclass correlation coefficient of CRAE was 0.910 and that of CRVE was 0.897.

History of cardiovascular disease, duration of diabetes, physical activity (h/week), smoking status (never, former, current) and alcohol intake (none/low/high) were assessed by questionnaire [10]. Use of lipid-modifying, antihypertensive and glucose-lowering medication was assessed during a medication interview in which the generic name, dose and frequency were recorded [10]. We measured weight, height, BMI, waist circumference, office and ambulatory 24 h blood pressure, plasma glucose levels, serum creatinine, 24 h urinary albumin excretion (twice), peripheral vibration perception threshold, HbA_1c and plasma lipid profile, as described elsewhere [10]. eGFR (in ml min⁻¹ 1.73 m⁻²) was calculated with the Chronic Kidney Disease Epidemiology Collaboration equation based on both serum creatinine and serum cystatin C [17]. The presence of retinopathy was assessed in both eyes by use of fundus photographs taken with the same fundus camera (Model AFC-230; Nidek, Gamagori, Japan) as used for measurement of retinal microvascular diameters [10]. Plasma biomarkers of inflammation included high-sensitivity C-reactive protein, serum amyloid A (SAA), IL-6, IL-8 and TNF-α and were measured in EDTA plasma samples with commercially available 4-plex sandwich immunoassay kits (Meso Scale Discovery, Rockville, MD, USA).

Multiple linear regression analysis was used to determine the association of glucose metabolism status (NGM, prediabetes and type 2 diabetes) and measures of blood glucose (HbA_1c, fasting glucose, 2 h post-load glucose levels) with retinal vessel diameters. For linear trend analyses, the categorical variable glucose metabolism status (NGM = 0, prediabetes = 1, and type 2 diabetes = 2) was used in the regression models. To estimate the difference in retinal microvascular diameters between individuals with prediabetes and type 2 diabetes compared with NGM, we performed analyses with dummy variables for prediabetes and type 2 diabetes. We used the likelihood ratio test to compare models in which glucose metabolism status was treated as a categorical or continuous variable [18]. Model 1 was adjusted for age and sex; Model 2 was additionally adjusted for cardiovascular risk factors that have previously been associated with retinal microvascular diameters (waist circumference, smoking status, office systolic blood pressure, use of antihypertensive and/or lipid-modifying drugs, fasting triacylglycerols and total- to HDL-cholesterol ratio). We also performed a range of additional analyses (see Results for details). A standardised sum score was calculated for plasma markers of inflammation as follows: for each individual biomarker, a z score was calculated according to the formula (individual value – population mean)/population standard deviation and the resulting individual biomarker z scores were then averaged. A p value of <0.05 was considered statistically significant. Interactions of glucose metabolism status and measures of blood glucose with sex and left vs right eye (with regard to the associations between glucose metabolism status with retinal diameters) were tested by incorporating interaction terms (e.g. prediabetes × sex) in the regression models. A p for interaction of <0.10 was considered statistically significant. Statistical analyses were performed by use of the Statistical Package for Social Sciences (Version 25.0; IBM, Chicago, IL, USA), except for the likelihood ratio test, which was performed using Stata (Version 14.1; StataCorp, College Station, TX, USA).

Table 1 shows the general characteristics of the study population stratified by glucose metabolism status. The study population consisted of 2876 individuals (98.6% white) with a mean age of 59.8 ± 8.2 years; 51.2% were men, and 28.3% had type 2 diabetes (by design), including both previously diagnosed type 2 diabetes (24.5%) and newly diagnosed type 2 diabetes (3.7%). Individuals with type 2 diabetes and prediabetes, compared with those with NGM, were older (p<0.001, ANOVA test), more often male (p<0.001, χ² test) and a current smoker (p<0.001), and had a higher BMI (p<0.001), waist circumference (p<0.001), systolic and diastolic blood pressure (p<0.001 for both), fasting plasma glucose (p<0.001), 2 h post-load glucose (p<0.001), HbA_1c (p<0.001) and triacylglycerol levels (p<0.001), lower level of physical activity (p<0.001) and lower eGFR (p<0.001). The group of individuals with missing data on retinal microvascular measurements or covariates were generally quite similar to those included, but had a higher total- to HDL-cholesterol ratio, more current smokers and insulin use, and smaller CRAE (ESM Table 1).

Retinal arterioles were wider (CRAE measured in MU) in individuals with type 2 diabetes compared with those with NGM in the age- and sex-adjusted model (Model 1: β = 2.29 [0.52, 4.06]; Fig. 1a). The association of prediabetes with CRAE was non-significant (Model 1: β = 0.42, [95% CI −1.73, 2.57]; Fig. 1a), whereas CRAE showed a linear trend across glucose metabolism status (Model 1: p for trend = 0.013; p for likelihood ratio test = 0.49). After further adjustment for cardiovascular risk factors (Model 2), the difference in CRAE between type 2 diabetes and NGM became somewhat larger (prediabetes β = 0.62 [−1.58, 2.83]; type 2 diabetes β = 2.89 [0.69, 5.08]; Fig. 1a). The linear trend for CRAE across glucose metabolism status remained (Model 2: p for trend = 0.013; p for likelihood ratio test = 0.43).

Retinal venules were wider (CRVE measured in MU) in individuals with prediabetes and type 2 diabetes, compared with those with NGM, in the age- and sex-adjusted model (Model 1: prediabetes β = 3.84 [0.50, 7.18]; type 2 diabetes β = 4.68 [1.93, 7.43]; Fig. 1b), and the CRVE showed a linear trend across glucose metabolism status (Model 1: p for trend = 0.001; p for likelihood ratio test = 0.36). The difference in CRVE was attenuated and non-significant after adjustment for cardiovascular risk factors (Model 2: prediabetes β = 2.40 [−1.03, 5.84]; type 2 diabetes β = 2.87 [−0.55, 6.29]; Fig. 1b). CRVE no longer showed a linear trend across glucose metabolism status (Model 2: p for trend = 0.083; p for likelihood ratio test = 0.55).

Higher levels of HbA_1c were associated with greater CRAE after adjustment for age and sex (Model 1; Figs. 2a and 3a), and also after further adjustment for cardiovascular risk factors (Model 2; Fig. 2a). Higher levels of fasting glucose and 2 h post-load glucose were not statistically significantly associated with greater CRAE (Fig. 2a).

Higher levels of HbA_1c were associated with greater CRVE after adjustment for age and sex (Model 1; Figs. 2b and 3b). The association was attenuated and non-significant after further adjustment for cardiovascular risk factors (Model 2; Fig. 2b). Higher fasting glucose and 2 h post-load glucose were not statistically significantly associated with greater CRVE (Fig. 2b).

Further analyses to assess the robustness of our observations are described in the ESM results (ESM Tables 2–7); in general, these analyses confirmed the observations reported above. To explore whether retinal diameters are intrinsically linked, we analysed venular diameters as a function of arteriolar diameters. We found that retinal arteriolar diameters were positively associated with retinal venular diameters after adjustment for age and sex (β = 1.09 [1.05, 1.13], p < 0.001; ESM Fig. 2). The association remained similar after further adjustment for height, body surface area, systolic blood pressure and HbA_1c level (β = 1.09 [1.05, 1.13], p < 0.001). In addition, to explore whether the associations of glucose metabolism status and measures of blood glucose with CRAE and CRVE are linked, we additionally adjusted for CRVE in models of CRAE and for CRAE in models of CRVE. We found that further adjustment for CRVE attenuated the difference in CRAE in prediabetes and type 2 diabetes vs NGM (prediabetes β = −0.46 [−2.03, 1.11]; type 2 diabetes β = 1.60 [0.03, 3.16]; p for trend = 0.073), while further adjustment for CRAE completely eliminated the difference in CRVE (prediabetes β = 1.72 [−0.73 to 4.18]; type 2 diabetes β = −0.28 [−2.73, 2.16]; p for trend = 0.984). Similarly, the association of HbA_1c with CRAE was attenuated after further adjustment for CRVE, while the association with CRVE was eliminated after further adjustment for CRAE (Model 3; Fig. 2).

We did not find any significant associations between duration of type 2 diabetes and retinal microvascular diameters (available for n = 673 individuals; ESM Table 8).