patients. Whilst there are methods to adjust for difference in measurement using ocular biometric data (e.g. refractive power and axial length) [55], most were designed for telecentric cameras. For non-telecentric cameras, Rudnicka and colleagues [56] have described a method to adjust for magnification using film-based photographs, but it is not well investigated for digital fundus images. Further studies defining normal range are crucial for a clinical risk prediction model.

To account for magnification effects and allow for comparison of measurements of retinal vessel calibre changes between individuals and between studies, AVR has been mainly used in earlier studies [50, 53]. However, AVR has a significant limitation that it cannot distinguish whether it represents ‘arteriolar narrowing’ or ‘venular dilation’. There have been accumulated observations that different systemic diseases or conditions appear to be associated with specific calibre changes in arterioles and venules respectively. Whilst smaller retinal arteriolar calibre is associated with hypertension and may even precede clinical hypertension development, larger retinal venular calibre has been associated with inflammation, smoking, hyperglycaemia, obesity and dyslipidemia. These observations suggest that changes in retinal arteriolar and venular calibre may reflect different aspect of pathophysiological processes underlying the associated systemic diseases. Therefore, combining these two components into one estimate of the AVR without consideration of separate arteriolar or venular calibre measurements would therefore lose this important information.

8.2.3Newer Retinal Imaging for Morphologic Features of Retinal Vasculature

There are newer imaging for morphological or geometrical retinal vascular features [55], including parameters to quantify retinal vascular fractal dimension (Fig. 8.2) [57, 58], retinal vascular tortuosity and branching angles and patterns [59], promising to predict CVD risk. These parameters have a potential to represent abnormal haemodynamic state of microvasculature because our vascular system is under constant remodelling

adopting for changing environment (e.g. change in blood pressure, velocity, viscosity). This adaptation process might result in change in vessel morphology such as calibre, tortuosity and branching patterns, which are thought to be secondary to provide information on subtle but global changes in vascular physiology or vascular health in whole body [60]. These vascular parameters may be a sensitive marker for the early prediction of CVD [61].

8.3Associations of Retinal Imaging and CVD Risk

8.3.1Retinal Microvascular Signs

(Table 8.1)

8.3.1.1 Risk of Pre-clinical CVD

Studies suggest that retinal vascular signs are associated with presence of subclinical or pre-clinical CVD. For example, in the ARIC study, increased carotid artery intima-media wall thickness was positively associated with severity of retinopathy (10% more likely to have more severe retinopathy per 0.1-mm thickness) independent of diabetes duration, glucose and blood pressure [62].

8.3.1.2 Risk of Stroke

There is substantial evidenced that retinal microvascular signs are associated with presence and risk of clinical stroke. In the ARIC study [63], persons with retinopathy signs at baseline were two to three times more likely to have incident stroke over 3.5 years. Presence of arterio-venous nicking was also associated with 60% higher risk of stroke independent of conventional stroke risk factors. In the same study, persons with both retinopathy and white matter lesions (WML) in the brain (an ischaemic complications of cerebral microvascular disease) had 18 times higher risk of incident stroke than those without either retinopathy or WMLs [36]. In persons with diabetes, diabetic retinopathy was associated with twofold increased risk of ischaemic stroke over 8 years independent of conventional stroke risk factors [16]. In the Blue Mountains Eye Study (BMES) [64], stroke events were more frequently found in persons with retinopathy (5.7%), with

Fig. 8.2 Fractal dimensions of the retinal vascular patterns in the area surrounding the optic disc. (a) Original retinal image of image B. (b) Vessel identified from image

A (fractal dimensions = 1.50). (c) Original retinal image of image D. (d) Vessel identified from image C (fractal dimensions = 1.46)

moderate/severe arterio-venous nicking (4.2%), or with focal arteriolar narrowing (7.2%), compared with those without retinopathy (1.9%) over 7 years follow-up. After adjusting for stroke risk factors, presence of retinopathy was significantly associated with 1.7-fold increased risk of stroke in non-diabetic population. Interestingly, this association was stronger in those without severe hypertension (relative risk of 2.7) or in persons with multiple retinal microvascular signs (relative risk of 2.7). Focal arteriolar narrowing or

arterio-venous nicking was not independently associated with combined stroke events after adjustment for stroke risks.

8.3.1.3 Risk of Coronary Heart Disease

There is less consistent data regarding a strong relationship between retinal microvascular signs and risk of coronary heart disease (CHD). In the LipidResearchClinic’sCoronaryPrimaryPrevention Trial [65], a cohort study of middle-aged men

with hypertension and hyperlipidemia, the presence of retinopathy was associated with twofold risk of CHD, and presence of either generalized or focal arteriolar narrowing was associated with threefold increased risk of CHD. In the Beaver Dam Eye Study (BDES) [66], non-diabetic adults with retinal haemorrhages had 2.4 times higher risk of ischaemic heart disease mortality in 5 years; presence of moderate retinopathy was associated with threefold higher risk of mortality from ischaemic heart disease in 14 years of fol- low-up. In the ARIC study [67], persons with retinopathy had a twofold higher risk of incidence of chronic heart failure compared with those without retinopathy, independent of conventional cardiovascular risk factors. In persons with type 2 diabetes in the ARIC study [68], presence of diabetic retinopathy was associated with twofold increased risk of incident CHD over 8 years and threefold increased risk of

mortality from CHD independent of other cardiovascular risk factors.

8.3.2Retinal Vessel Biometry

(Table 8.2)

8.3.2.1 Risk of Hypertension

The BDES has reported that persons with hypertension have more retinal arteriolar abnormalities [69] and more likely to develop over 5 years [70] compared to those without hypertension. Retinal arteriolar narrowing has long been regarded as an early sign of hypertension secondary to chronic exposure to increased blood pressure [71–73] and is inversely related to higher blood pressure levels [74]. This finding subsequently confirmed in multiple population-based studies [75–78]. In addition, recent epidemiologic studies have demonstrated that retinal artery narrowing is not

only reflecting a chronic exposure to hypertension as a result, but is antecedent of developing hypertension. Hypothesis underneath is that primary feature of essential hypertension is increased peripheral vascular resistance in small vessels elsewhere in the body, and this might be observed in retinal microvasculature as well. In longitudinal epidemiologic studies, it has been shown that change in retinal vessel calibre provided the first prospective clinical evidence showing that narrower arteriolar calibre preceded the development of clinical hypertension and was not purely a secondary response to established hypertension [79]. Subsequent studies such as the BMES have since demonstrated that retinal arteriolar narrowing precedes by years the development of hypertension in initially persons with normal blood pressure level [80, 81].

8.3.2.2 Risk of Stroke

There is now good evidence that narrower retinal arteriolar calibre as well as wider retinal venular calibre is associated with stroke and other cerebrovascular diseases. In the ARIC study, smaller retinal AVR was reported to be associated with increased risk of stroke [63]. This association was confirmed in the CHS [82], but persons with larger retinal venular calibre were also more likely to have a higher incidence of stroke compared to persons with smaller retinal venular calibre. The association of stroke with wider retinal venular calibre further confirmed in the Rotterdam study [83] where larger venular diameters were also associated with 12% higher risk of stroke and 15% higher risk of cerebral infarction, whilst retinal arteriolar narrowing was neither related to the risk of stroke nor to the risk of cerebral