Retinal Vessel Diameter as a Clinical Predictor of Diabetic Retinopathy Progression: Title and subTitle BreakTime to Take Out the Measuring TapeRetinal Vessel Diameter

Irregular dilation of the retinal venules (venous beading) has long been known to be an important clinical sign of diabetic retinopathy severity and a signal of impending progression to vision-threatening stages of proliferative disease.¹ Traditionally, ophthalmologists have been able to identify prominent venous beading clinically, but they may not be able to distinguish milder and more generalized dilatation of the venules with precision.

Improvements in digital retinal photography and imaging technology now allow objective documentation of various retinal vascular characteristics. One of the key developments in the past 2 decades is the measurement of retinal vessel diameter from fundus photographs, first reported in the Atherosclerosis Risk in Communities Study in the 1990s.² This technique has now been used in many epidemiological and clinical studies around the world to examine the relationship between the quantitatively measured retinal vessel diameter and the risk of diabetes mellitus,^{3 - 5} diabetic retinopathy,^{6 - 8} and various systemic vascular diseases.^{9 - 11}

What have these investigations found? First, the measurement of retinal vessel diameter from fundus photographs is precise, with good intersession and intrasession reliability.² Second, the relationships of retinal arteriolar and venular diameter to various systemic risk factors differ remarkably. Although retinal arteriolar diameter is strongly related to blood pressure (with a narrower arteriolar diameter seen in patients with hypertension), venular diameter is associated with metabolic risk factors, venous dilatation linked with hyperglycemia, higher body mass index, and systemic markers of inflammation (eg, C-reactive protein).^{12 - 13} Thus, studying retinal arteriolar and venular diameters may provide a unique in vivo opportunity to differentiate specific microvascular processes occurring in the eye.¹⁴ Finally, somewhat unexpectedly, studies suggest that retinal venular diameter is a stronger predictor of major cardiovascular disease events (including stroke and coronary heart disease) than arteriolar diameter.^{11 ,15}

There is now also evidence that the measurement of retinal vessel diameter may provide additional clinical information for the prediction of diabetic retinopathy. A larger retinal venular diameter or a generalized venular dilatation has been shown to predict the progression of diabetic retinopathy, independent of other known risk factors such as duration of disease and glycemic levels.^{6 ,8} In this issue of the Archives, Roy and colleagues¹⁶ extend this area of research. They examined the relationship between retinal arteriolar diameter and venular diameter, measured using the Atherosclerosis Risk in Communities protocol at baseline, and the subsequent progression of diabetic retinopathy in a cohort of 468 African American individuals with type 1 diabetes. The study found that a larger retinal venular diameter was significantly associated with 6-year progression to either proliferative diabetic retinopathy or proliferative retinopathy with high-risk characteristics, while controlling for baseline risk factors. These results provide additional data to a previous report of white persons with type 1 diabetes in the Wisconsin Epidemiologic Study of Diabetic Retinopathy⁶ in which a larger retinal venular diameter, measured using a similar computer-assisted technique as in the study by Roy et al,¹⁶ was also significantly associated with 4-year progression of diabetic retinopathy and incidence of proliferative retinopathy.

The underlying pathophysiological mechanisms explaining these observations are still unclear, and 3 interrelated hypotheses are discussed by Roy and colleagues.¹⁶ Retinal venous dilatation may indicate ocular ischemia,¹⁷ systemic inflammation,^{12 - 13 ,18} (an emerging pathogenic factor involved in retinopathy), and/or endothelial dysfunction.¹⁴ The most convincing evidence regarding endothelial dysfunction are new studies showing that the vasodilatory response of the retinal vessels to flicker-light stimulation, a process mediated by nitric oxide and correlated with brachial artery flow-mediated vasodilation,¹⁹ is blunted in persons with diabetes compared with nondiabetic controls and is further impaired with increasing diabetic retinopathy severity.²⁰ It is unclear why retinal arteriolar diameter, which appears to predict the development of diabetes itself as well as the onset of mild retinopathy, is not associated with progression to more severe stages of retinopathy.^{1 ,3 - 4 ,14} Clearly, further mechanistic studies are needed to investigate the link between retinal vessel diameter and the natural history of diabetes and its microvascular complications.

Regardless of underlying mechanisms, how are these findings useful for clinicians? Ophthalmologists will welcome more precise methods to determine which diabetic patient is more likely to progress to high-risk retinopathy. Traditional systemic risk factors (ie, duration of diabetes and glycemic and blood pressure control) may provide some information on a person's odds of progression.¹ Most ophthalmologists also schedule follow-up appointments for diabetic patients based on the severity of retinopathy at the time of examination (eg, 6-12 months for mild retinopathy and 2-4 months for severe nonproliferative retinopathy). Will it be possible for ophthalmologists to now use a desktop software, perhaps linked to a fundus camera, to measure retinal vessel diameter and then use the additional information to further refine the probability of diabetic retinopathy progression?

Although such a concept is appealing, translating the current research findings to clinical practice is likely to take longer. First, the current range of software programs for measuring retinal vessel diameter, such as the one used in the study by Roy et al,¹⁶ are not fully automated or easy to use without standardized protocols, training, and additional input by technicians. New software programs incorporating automated optic disc detection, vessel identification, and vessel differentiation (arteriole vessel vs venule vessel) are being developed. Such automated software programs will likely improve precision and reliability but are not yet widely available.

Second, the performance of retinal vessel imaging in clinical practice and for risk stratification requires a series of data analysis steps. Although the study by Roy et al¹⁶ has shown that there is a prospective association between retinal vessel diameter and diabetic retinopathy progression, what is further needed is to clarify that adding such a novel test (ie, measuring retinal vessel diameter) incrementally improves prediction over traditional methods (eg, based on severity of retinopathy at presentation), reclassifies a patient's risk, changes treatment options, improves outcomes, and is cost-effective.²¹ Such analyses may include in prediction models the calculation of absolute risks (eg, 5-year risk of proliferative retinopathy in persons with a certain range of retinal venular diameter), positive and negative predictive values, and change in the area under the receiver operating characteristic curve for retinal vessel diameter. For the prediction of stroke, a recent study showed that the inclusion of retinal venular diameter containing traditional stroke risk factors resulted in the reassignment of 10% of people at intermediate risk into different, mostly lower, risk categories.²² Thus, knowing the retinal vessel diameter allowed reassignment of stroke risk in 1 of 10 patients, beyond the information from traditional risk factors such as blood pressure and diabetes status. These types of analyses have not been conducted for diabetic retinopathy prediction.

The retinal blood vessels are accessible to direct noninvasive visualization, providing a unique opportunity to study subtle, early microvascular changes prior to clinically significant end points such as diabetic retinopathy. The recognition that patients with different retinal vessel sizes have different rates of progression to proliferative diabetic retinopathy is important and substantiates clinical observations. Further development of retinal vascular imaging technologies may allow improved prediction of retinopathy progression in patients with diabetes, and ultimately the risk.

Correspondence: Dr Wong, Singapore Eye Research Institute, National University of Singapore, 11 Third Hospital Ave, Singapore 168751, Singapore (ophwty@nus.edu.sg).

Financial Disclosure: None reported.

Funding/Support: This work was funded by grants from the National Medical Research Council (NMRC/STaR/0003/2008), Singapore, and the National Health and Medical Research Council (52993 and 475617), Australia.