Introduction
The worldwide epidemic of diabetes and its complications necessitates identification of early pathophysiological changes in the development of complications, as an essential requirement for risk assessment as well as for the design of interventions.
In recent studies, retinal microvascular diameters have been shown to be closely related to incidence of complications of diabetes, including retinopathy, nephropathy and stroke [
1,
2], suggesting a role for early retinal changes in assessment of risk of these complications. However, the association of (pre)diabetes with retinal microvascular diameters remains unclear. Although type 2 diabetes has been associated with wider retinal arterioles in the majority of studies, an association of type 2 diabetes with wider retinal venules has only been found in non-white individuals [
3‐
9]. Moreover, few studies have investigated the association of prediabetes with retinal microvascular diameters; those that have, also found that prediabetes was associated with wider retinal venules in non-white individuals only [
3‐
6]. However, these studies [
3‐
6] had significant limitations. For example, only one study used the gold standard of OGTT, rather than fasting glucose level or random glucose level, to define prediabetes and diabetes in a multi-ethnic population [
5]. In addition, none adjusted for use of medication that is associated with glucose metabolism and microvascular function, such as antihypertensive drugs.
As it has been suggested that (pre)diabetes may be associated with both wider retinal arterioles and venules, the question arises whether, and, if so, how, these changes are related. Theoretically, wider arterioles could lead to wider venules via transmittance of greater microvascular pressure. Alternatively, or additionally, wider venules (if they are proven to be a marker for arteriole–venule shunting) could lead to wider arterioles through local tissue hypoxia. Recent studies on a multi-ethnic Asian population found that the association of (pre)diabetes with wider retinal arterioles was independent of retinal venular diameters, while the association with wider retinal venules was not independent of retinal arteriolar diameters [
4,
7,
8], which supports the hypothesis that (pre)diabetes-associated retinal venular dilation is linked to retinal arteriolar dilation. However, these associations have not been studied in a white population.
Therefore, in this population-based cohort study, we investigated the associations of OGTT-based glucose metabolism status (normal glucose status, prediabetes, type 2 diabetes) and measures of blood glucose with retinal microvascular diameters in a predominantly white population, taking into account a broad array of potential confounders. In addition, we explored whether retinal arteriolar and venular diameters were mutually related.
Discussion
This study shows that type 2 diabetes, higher levels of HbA1c and, possibly, prediabetes are associated with wider retinal arteriolar diameters in a predominantly white population. Notably, the associations with retinal arteriolar diameters are independent of a broad array of potential confounders. These findings indicate that retinal microvascular changes already occur prior to the diagnosis of type 2 diabetes. In addition, retinal arteriolar diameters are associated with retinal venular diameters, independently of age, sex, height, body surface area, blood pressure and blood glucose, which suggests a close link between arteriolar and venular dilation in general and, thus, also in (pre)diabetes.
Our results indicate that type 2 diabetes and, possibly, prediabetes are independently associated with wider retinal arteriolar diameters, which is consistent with previous cross-sectional studies [
3‐
9]. Compared with these studies, we used OGTT and HbA
1c, which are more accurate measurements for classifing glucose metabolism status than measuring fasting glucose, random glucose, and HbA
1c levels only [
19]. In addition, we showed that the associations were independent of a broad array of cardiovascular risk factors. Notably, we found that age, sex and systolic blood pressure had strong confounding effects. For example, older age, male sex and higher blood pressure were associated with both narrower arterioles and (pre)diabetes. Unsurprisingly, eliminating these confounders through statistical adjustment reversed the direction of association between (pre)diabetes and retinal arteriolar diameters (Table
1). Note also that we included ambulatory 24 h blood pressure as a confounder in our additional analyses, as it is more accurate than office blood pressure [
20] and has not been used in previous studies [
3‐
9], making residual confounding by inaccurately measured blood pressure much less likely in our study. With regard to the outcomes, we measured diameters with semi-automated software (RHINO), which was validated manually and had a relative error that was comparable to that of Interactive Vessel Analysis (IVAN) software [
21]. Finally, we used linear trend analyses, as we hypothesised that the difference in retinal microvascular diameters from NGM to prediabetes to type 2 diabetes is of a continuous nature. The results of these analyses favour the interpretation that arteriolar widening occurs in both type 2 diabetes and prediabetes. In support of this, HbA
1c, a continuous measure of blood glucose, was significantly associated with retinal arteriolar diameters. Although we cannot exclude the possibility that there is no true association between prediabetes and greater arteriolar diameter, we attribute the lack of statistical significance of the difference between prediabetes and NGM with regard to retinal arteriolar diameters to a type 2 statistical error, because the power of between-group comparisons was reduced compared with the power of trend analyses.
Retinal arteriolar dilation in (pre)diabetes is thought to be a result of impaired arteriolar autoregulation [
22]. Lacking neuronal innervation, retinal arterioles are affected mainly by local autoregulation through the release of vasoactive substances by microvascular endothelium and the myogenic response of smooth muscle cells [
23]. For example, in retinal arteriolar smooth muscle cells, hyperglycaemia and hypoxia can cause endothelin-1 resistance and inhibit Ca
2+ influx channels [
24,
25]. In addition, death and insufficient renewal of endothelial cells, smooth muscle cells and pericytes can further weaken arteriolar wall and boost dilation.
The associations of prediabetes and type 2 diabetes with retinal venular diameters were directionally similar to those for arterioles even though they were not statistically significant after adjustment for cardiovascular risk factors. The non-significance of the associations may be explained by four factors. First, the relatively larger measurement error of venular vs arteriolar diameters [
21] decreases the precision of the association with blood glucose and thus increases the confidence interval [
26]. Second, our additional analyses (Fig.
2 and ESM Fig.
2) are consistent with the concept that (glucose-related) arteriolar widening drives venular widening to an important extent, possibly by greater transmission of blood pressure. Such mediation will tend to bias the association between blood glucose and venular diameters towards the null [
27]. Third, longitudinal studies [
28‐
30] have suggested that widening of retinal venules may also occur before prediabetes, which may reduce the difference in venular diameter between (pre)diabetes and NGM. Fourth, our fully adjusted model may have been over-adjusted as a result of the inclusion of waist circumference, since obesity may be on the causal pathway between (pre)diabetes and retinal venular dilation [
31].
In general, studies on the associations between diabetes or blood glucose and retinal venular diameters have not shown consistent results [
3‐
9], although venular widening has been much more consistently observed among Asian populations [
3,
4,
6‐
8] than among white populations [
3,
5,
9]. These inconsistent results may be attributed to ethnicity but also to differences in classification of glucose metabolism status, insufficient adjustment for confounding and different types of software used.
The pathophysiological mechanisms that explain retinal venular dilation in (pre)diabetes remain unclear. Retinal venules have been proposed to dilate in response to diabetes-associated inflammation [
3,
32], but the association was unchanged after adjustment for inflammation in our study and a previous study [
3]. As alluded to above (Fig.
2 and ESM Fig.
2), our results are consistent with the hypothesis that retinal venular dilation is, at least in part, a direct consequence of arteriolar dysfunction [
24]. However, we cannot exclude the possibility that, alternatively or additionally, wider venules could lead to wider arterioles, for example, through arteriole to venule shunting and local tissue hypoxia.
Retinal arteriolar dilation is associated with progression of retinopathy [
33] and presence of neuropathy [
34]. Retinal venular dilation is similarly associated with incidence and progression of retinopathy [
35], incidence of nephropathy [
36], prevalence and incidence of stroke [
37,
38] and progression of cerebral small vessel disease [
39]. Taken together with our findings, these results may explain why such complications are commonly present at diagnosis of type 2 diabetes or sometimes before. Retinal microvascular dilation seems to be reversible [
40‐
42]; however, whether this improvement in retinal microvascular dilation will translate into an improved prognosis with respect to complications of diabetes needs further investigation.
Strengths of our study include the population-based design with oversampling of individuals with type 2 diabetes; the use of OGTT to characterise glucose metabolism status; the extensive phenotyping, which enables detection of independent associations after extensive adjustments for potential confounders; and the broad array of additional analyses, which gave deeper insight into the associations. Our study also has limitations. First, the cross-sectional data cannot definitively establish a causal link between (pre)diabetes and retinal microvascular features. Nevertheless, there is extensive evidence that hyperglycaemia causes microvascular dysfunction, and that the association may in fact be bidirectional [
43,
44]. Second, our study population was 40–75 years of age, predominantly white, with relatively well-controlled blood glucose and cardiovascular risk factors, which should be taken into consideration when the findings are extrapolated to other populations. Third, although treating glucose metabolism status as a continuous variable increased statistical power to detect the associations of (pre)diabetes with retinal microvascular diameters and the results of likelihood ratio test confirmed the feasibility of this approach, it may also introduce bias into the estimates of associations, which are largely influenced by the difference in retinal microvascular diameters between the two extreme groups, i.e. the NGM and type 2 diabetes groups. Fourth, our fully adjusted model may have been over-adjusted, as a result of the inclusion of waist circumference, since for retinal venular diameters, and therefore the association of (pre)diabetes with retinal microvascular diameters may have been underestimated. Fifth, although investigational procedures were standardised, participants were allowed a light meal, which will increase variation in retinal microvacular diameters and thus bias associations towards the null. Sixth, although we adjusted for major potential confounders, there is still a possibility of residual confounding by variables that were not included in the analyses.
In summary, this study has demonstrated that type 2 diabetes, higher levels of HbA
1c, and, possibly, prediabetes are associated with wider retinal arterioles, independent of major cardiovascular risk factors, in a predominantly white population. These results support the ‘ticking clock’ hypothesis, which postulates that microvascular dysfunction precedes the clinical diagnosis of type 2 diabetes [
43,
44], and may partly explain the occurrence of complications related to microvascular dysfunction in prediabetes and in early type 2 diabetes. Thus, microvascular dysfunction can be considered an early marker of (pre)diabetes and a potential target for intervention.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.