Incidental observations of remarkable degrees of straightness or tortuosity of retinal blood vessels may be evaluated as potential signs of morbidity, but there is little evidence to support that any diagnostic value can be inferred from non-prospective funduscopic observation of single individuals. Considerable variation in retinal vessel tortuosity can be seen in young healthy subjects. Increased retinal arterial tortuosity has been suggested to be associated with hypertension.1 2 A recent study demonstrated that less tortuous retinal arteries were associated with increased risk of death from ischaemic heart disease, independent of systolic blood pressure.3 No twin study has previously been performed on the normal variation in retinal vessel tortuosity and its association with genetic and systemic factors before. To evaluate assessment of retinal vessel tortuosity as a means of monitoring systemic health, it is essential to know what determines the layout of retinal vessels in healthy young subjects with no significant systemic disease. The aim of the present study was to assess the relative contribution of genetic and environmental factors on retinal vessel tortuosity, and the association between tortuosity and various health indices in healthy young to middle-aged adults.

MATERIALS AND METHODS

Participants

The study was a cross-sectional study of 57 monozygotic (MZ) and 52 dizygotic (DZ) same-sex twin pairs, aged 20 to 46 years, all Caucasian and living in Denmark. The subjects were recruited from the population-based Danish Twin Registry.4 The twins participated in a larger study of the metabolic syndrome (The Geminakar Study).5 Only subjects in self-assessed good health were invited to participate, and exclusion criteria were pregnancy, breastfeeding, known diabetes or cardiovascular disease, and conditions preventing them from completing a bicycle test. Sampling from this group was stratified according to age and gender, to get an equal distribution of twin pairs along the age span in the two zygosity groups. DNA-based microsatellite markers (AmpFISTR Profiler Plus Kit; PE Applied Biosystems, Perkin Elmer, Foster City, CA) were used to determine zygosity of the twins. Twin pairs where both twins lived on the island of Sjaelland were invited to participate in a separate ophthalmic examination, for which 114 pairs of twins volunteered. Ophthalmic exclusion criteria included cataract and lens opacities near the optical axis of the eye, other manifest eye disease, and unclear fundus photographs, such findings leading to the exclusion of both twins in a pair. The study was approved by the Medical Ethics Committee of Copenhagen County and followed the tenets of the Declaration of Helsinki, including informed consent.

Clinical examination

Study examinations that have previously been described in detail elsewhere included oral glucose tolerance testing, blood glucose and insulin levels, assessment of smoking habits, systolic and diastolic blood pressure measurement, blood sampling, measurement of waist circumference, height and weight.5 6 Body mass index (BMI) was calculated as weight (kg) divided by square of height (m). Systolic and diastolic blood pressures were measured after 30 minutes’ rest, in a sitting position, using a conventional mercury sphygmomanometer and hands-free stethoscope by trained nurses. Measurements were taken three times, and the mean of the three measurements was used for analysis. The mean arterial blood pressure (MAP) was calculated as diastolic blood pressure plus 33% of the difference between systolic and diastolic blood pressure. Hypertension was defined as systolic blood pressure ⩾140 mm Hg or diastolic blood pressure ⩾90 mm Hg and/or current use of antihypertensive medicine. None of the participants had arterial hypertension. Five participants in the eye study met one criterion for diabetes mellitus according to current WHO7 criteria: four had fasting glucose ⩾6.1 mmol/l (whole capillary blood), and one had 2 h glucose ⩾11.1 mmol/l. None of these subjects had any symptoms of diabetes prior to or during the study, and none had diabetic retinopathy. All five subjects were included in the study. The eye examination included refraction, determination of visual acuity, pupil dilation, slit-lamp biomicroscopy and fundus photography. Best corrected visual acuity was determined using a Snellen decimal projection chart, and all eyes included had a visual acuity of at least 0.9.

Fundus photography

Digital grey-scale fundus photographs (20° and 50°, TIFF 1024×1024 pixels), centred on the macula and optic disc, after pupil dilation (with phenylephrine hydrochloride 10% and tropicamide 1%), were obtained using a fundus camera (model TRC-50X; Topcon, Tokyo) equipped with a digital back piece (MEGAPLUS model 1.4; Eastman Kodak, San Diego, CA) and a PC-based image-management system (Ophthalmic Imaging Systems, Sacramento, CA). A green filter (filter: Wratten 54; Eastman Kodak, Rochester, NY) for “red-free” photographs was used to enhance the sharpness and contrast of the blood vessels.

Fundus grading

Tortuosity was assessed by visual grading of one fovea-centred and one disc-centred fundus image from each eye displayed on a computer display and viewed independently by two graders. The graders were medical doctors (first and second authors of this paper) with 3 years of experience in ophthalmology research, especially trained in evaluating retinal blood vessels in fundus photographs. Retinal vessel tortuosity was evaluated simultaneously for both of a subject’s eyes, and reported as the average of the right and left eyes, because the correspondence between right and left eye was very high in all subjects (weighted kappa = 0.74). A grading scale was developed for the study based upon a preliminary inspection of the study photographs, which indicated that an acceptable reproducibility could be achieved using a three-level scale for arterial tortuosity, whereas the use of four or more levels led to inconsistent grading of a large proportion of cases. The variation in venous tortuosity was found to be much smaller than that of the arteries and too small to warrant an attempt at grading the entire study data set. The four fundus quadrants were evaluated separately, and the number of inflections of the upper and lower nasal arteries and upper and lower temporal arteries and their first-order branches was counted. The grading levels for retinal arterial tortuosity were: Category 1 (predominantly straight arteries in all four quadrants); Category 2 (wavy, mild to moderate tortuosity with one or two inflections of at least one major artery in one to three quadrants); Category 3 (prominent tortuosity, with two or more inflections of at least one artery in all four quadrants; fig 1). The first author examined all images twice at an interval of 2 months in random order, masked to zygosity, relatedness and values of all covariates. The second author examined all images once, masked to zygosity, relatedness and values of covariates. The three grading sessions were compared and were found to disagree for 71 of 218 subjects. The fundus photographs from these 71 subjects were then consensus-graded by the two graders in a joint session. The intra- and intergrader reproducibility was 0.79 (CI₉₅ 0.73, 0.86, weighted kappa, n = 218) and 0.68 (CI₉₅ 0.60, 0.77, weighted kappa, n = 218), respectively, demonstrating good agreement. The equation modelling yielded heritability estimates of 82–87% for arterial tortuosity, for the three different grading sessions, indicating that the intra- and intergrader variation did not affect the heritability results significantly. We reported only results from the final consensus grading.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Sample photographs from the right and left eyes of three subjects each representing a typical example of retinal arterial tortuosity of category 1 (straight), 2 (wavy) and 3 (tortuous).

Retinal vessel measurement

Retinal vessel calibres were assessed using a custom-developed semiautomated computer algorithm, described in detail elsewhere.8 The central retinal artery equivalent (CRAE) and the central retinal vein equivalent (CRVE) were calculated using the formulae described by Knudtson et al.9

Statistical analysis

The classical twin model is based on the assumption that MZ twins are approximately 100% genetically identical, for which reason all observed differences between two twins in a pair are attributable to environmental factors. DZ twins share on average 50% of their genes. The extent to which MZ twins are more alike than DZ twins is therefore assumed to reflect a genetic influence on the phenotype in question. Polychoric correlations, conditional probabilities, heritability estimates and regression coefficients for various covariates were estimated by use of structural equation modelling using the MX computer software.10 The polychoric correlations were estimated because the dependent variable, tortuosity, was categorical, and so we built our estimates on the assumption that there was an underlying normally distributed continuous variable, which we could not observe directly. The polychoric correlations correspond to the intrapair correlations (Pearson correlation between two twins in a pair) for continuous traits. We assume that the categorical variable (tortuosity) changes value when the underlying variable goes beyond the threshold values (there are two such thresholds, because there are three tortuosity categories; straight, wavy and tortuous). The model estimates the polychoric correlations for MZ and DZ twins respectively. If the polychoric correlations are significantly higher for MZ twins than for DZ twins, this indicates that genetics has an influence on the phenotype in question. Conditional probabilities estimate the probability that a twin has for example straight arteries if the co-twin has straight arteries. If there is a genetic effect on the phenotype in question, the MZ concordant pairs have higher estimates of conditional probability than DZ concordant pairs, because this indicates that MZ twin pairs are more alike than DZ twin pairs, for the given phenotype. Concordant means that both twins in a pair have the same phenotype.

Heritability is the proportion of phenotypic variance that can be accounted for by genetic differences among individuals for a particular trait in a particular population at a particular time. The genetic contribution can be divided into an additive genetic variance component A, representing the influence of alleles at multiple loci acting in an additive manner, and a non-additive genetic variance component D, representing intragene interactions (dominance) and inter-gene interactions (epistasis). The intrapair variance due to the influence of the environment can be subdivided into a common environmental variance component C, representing environmental factors shared by twins during intrauterine life and early childhood (a source of similarity), and an unshared environmental variance component E, representing environmental factors within families that are specific for each twin (hence, a source of dissimilarity which also includes random factors and measurement error).11 Structural equation modelling was used to fit ACE and ADE models to the observed data. The model with the lowest Akaike’s information criterion (AIC) was used to determine the best-fitting model. AIC = −2×log likelihood+2×(number of free parameters in the model). To adjust for the effects of age, gender and other covariates, allowance was made for a linear effect on the thresholds in the model, resulting in additional regression coefficients to be estimated. Due to a lack of statistical power and for technical reasons, it was decided to adjust for a maximum of three covariates per analysis. Covariate selection was prioritised by first adjusting for covariates with an established relation to vascular morphology, such as age, gender and blood pressure. BMI was adjusted for because it was found to be significantly associated with tortuosity by use of multiple regression analysis (SAS software, version 9.1). Since MAP and BMI were correlated, we estimated the heritability, adjusting for MAP and BMI separately. Reported p values for the comparison of MZ and DZ twins were found to be robust with respect to deviations of data distribution from normality and the presence of intrapair correlation (STATA software, version 9). Adjustment for refraction was performed by excluding all twin pairs (six pairs) with a difference of more than 4 dioptres between twin 1 and 2. The level of statistical significance was set at 0.05.

RESULTS

The study population of 57 MZ and 52 DZ same-sex twin pairs of both sexes aged 20–46 years demonstrated comparable distributions of age, gender, blood pressure, smoking habits and other study variables between the two subpopulations (table 1). Retinal arterial tortuosity analysis demonstrated that 79 subjects (36.2%) had predominantly straight arteries (category 1), 110 (50.5%) had wavy arteries (category 2), and 29 (13.3%) had tortuous arteries (category 3). For MZ twins, 40 pairs had the same retinal artery tortuosity category, and 17 pairs had different tortuosity categories. For DZ twins, 26 pairs had the same retinal artery tortuosity category, and 26 pairs had different tortuosity categories. Thus, more MZ pairs than DZ pairs shared the same tortuosity category, and the agreement was higher for straight and wavy arteries than for tortuous arteries; this is statistically more correctly illustrated by use of conditional probabilities (see tables 2 and 3). The conditional probabilities were higher for concordant MZ twin pairs than for concordant DZ twin pairs, indicating a genetic effect on retinal arterial tortuosity. For example, if a MZ twin had tortuosity category 1, the probability that the co-twin also had category 1 was 0.72 (tables 2 and 3).

The diagonal cells in tables 2 and 3 represent the concordant twin pairs. For example, for MZ twins, the probability that a twin has category 3 is 0.63 if the co-twin has category 3, and for DZ twins, the probability that a twin has category 3 is only 0.20 when the co-twin has category 3.

Estimates for polychoric correlations (intrapair correlations) were significantly higher for MZ twins, 0.81 (CI₉₅ 0.62, 0.92) compared with DZ twins, 0.40 (CI₉₅ 0.05, 0.67, p = 0.303, probability of MZ and DZ twins being identical). This means that MZ pairs were more alike than DZ pairs, again demonstrating a genetic influence.

Structural equation modelling estimated the heritability of retinal arterial tortuosity to be 82% (CI₉₅ 64, 92%), unshared environmental factors accounting for the remaining 18% (CI₉₅ 8, 36%) of the variation in tortuosity (table 4). When including MAP as a covariate, the estimated regression coefficient was −0.02 (CI₉₅ −0.03, −0.001). When including BMI, the estimated regression coefficient was −0.06 (CI₉₅ −0.11, −0.02). Other indices of cardiovascular health: age, gender, systolic blood pressure, fasting blood glucose, total cholesterol, low-density lipoprotein, high-density lipoprotein, triglycerides, smoking (pack years), retinal artery diameter and retinal vein diameter were not significantly associated with retinal arterial tortuosity. Adjustment for refractive errors did not attenuate the results, and so these results are not included in this communication.

DISCUSSION

The results of the present study demonstrated that wide variations in retinal arterial tortuosity may mostly be explained by the effect of genetic factors, which had a prominent impact in a study population that was free from arterial hypertension. Retinal arterial tortuosity demonstrated a heritability of 82%. Analysis of a wide range of systemic factors revealed significant statistical effects of MAP and BMI. The regression coefficients for these correspond to a decrease in MAP of 0.02 mm Hg for a one-step increase in retinal arterial tortuosity and a decrease in BMI of 0.06 kg/m2 for a one-step increase in tortuosity, both effects being of null clinical significance in this normotensive and healthy study population. Age, gender, cardiovascular risk factors and environmental factors such as smoking and diet (represented by blood lipids) were not associated with retinal arterial tortuosity in the present study. Our findings are in agreement with an earlier study that found no effect of age on retinal arterial tortuosity in subjects without hypertension.12

The exact mechanisms that underlie the development and the later modulation of the vascular network are largely unknown. In response to increased transmural pressure, the canine carotid artery has been shown to dilate and elongate.13 Qualitatively, this is what one would expect of a passive elastic tube.14 Observation of such a passive behaviour may indicate that the autoregulatory capacity has been exhausted. The same behaviour can be observed in the retinal arteries of newborns with acute perinatal distress, systemic hypoxia and lactic acidosis.15 Preterm birth is also associated with permanently increased retinal arterial and venous tortuosity, and reduced numbers of vascular branching points, independently of retinopathy of prematurity.16 Relief of acute fetal distress is followed by decreasing retinal arterial tortuosity.17 Metabolic alterations and consequent disease can also increase retinal vascular tortuosity, as seen in diabetic retinopathy.18

The role of arterial hypertension in modulating retinal arterial dilation and tortuosity is of particular interest. In monkeys and rats, high blood pressure has been shown to be associated with increasing dilation and tortuosity of retinal arteries.19 20 This finding has been replicated in humans in two cross-sectional studies, one of which also found an effect of increased pulse pressure.1 2 In the present cross-sectional study, which did not include subjects with arterial hypertension, we found the opposite relation, namely decreasing retinal arterial tortuosity with increasing blood pressure, although the change in blood pressure was so close to zero that it had null clinical significance in this normotensive population. A recent study found that ischaemic heart disease (IHD) was associated with reduced retinal arterial tortuosity, independent of arterial blood pressure.3 It is an interesting parallel and supports the finding that in the present study, increasing BMI, a risk factor for IHD, was also associated with decreasing retinal arterial tortuosity. Although the association with BMI was so weak that it had null clinical significance in this healthy population, it cannot be excluded that there may be a more clinically significant association in larger populations with higher BMI.

It is important to emphasise that the subjects were young and healthy, because when analysing genetic and environmental factors, ageing and disease processes should not cause any bias to the results. It is hypothesised that most phenotypes vary in expression with age, and a trait may be influenced by different genes at different ages; therefore using MZ and DZ twins instead of first- and second-degree relatives has the advantage of removing the age confounder from consideration.

It may have been a limitation that we applied a categorical visual grading procedure to quantify variations in retinal arterial tortuosity. Since retinal vascular tortuosity is a continuum, semiautomatic digital image analysis may have the potential to provide better measures of tortuosity and make it easier to compare results with other studies. In the literature, however, there is no standard method for measuring vessel tortuosity, but Hart et al21 have described an automatic method for vessel tortuosity which has been used by Witt et al.3 We did not have access to this type of method. Even though a study uses a quantitative method, there may still be a high intergrader variability, because the results depend on where in the fundus the grader chooses to measure and where to cut up the vessels in smaller segments. In this study, we believe that the human eye/brain is just as good or maybe even better at classifying tortuosity from fundus photographs than a semiautomatic computer program, especially because the results were meant for estimation of heritability, where it is not necessary to compare the results with other population studies. Evaluating each quadrant of the eye and counting the number of inflections of first- and second-order arteries made it very easy to assess the three different categories. Clinical assessment of retinal arterial tortuosity based on a single examination appears to have little diagnostic value, however, because we do not know if the grade of tortuosity is a constantly changing parameter. Longitudinal studies based on repeated fundus photographs over many months and years are likely to be more sensitive than cross-sectional population studies in showing effects of current ocular and systemic health parameters on vascular morphology.

In summary, the present study contributed further evidence of a pronounced genetic effect on ocular morphology, as has been shown for retinal vessel diameters,8 cilioretinal arteries,22 small hard drusen23 and retinal nerve-fibre layer thickness in this same twin population.24 We also found that there was a large variation in tortuosity of retinal arteries in normotensive, young, healthy subjects, stressing the importance of inter-individual comparisons in other populations, too. The extensive inter-pair variation in vessel tortuosity supports the hypothesis that comparison against a baseline fundus photograph is a prerequisite for using evaluation of retinal arterial tortuosity to assess the impact of arterial blood pressure and other health indices. A longitudinal follow-up study of this twin population will allow us to determine the influence of genetic factors and blood pressure on the age-related changes in retinal vascular morphology.