Retinal Arteriolar Narrowing and Risk of Diabetes Mellitus in Middle-aged Persons FREE

Type 2 diabetes mellitus affects up to 15 million persons in the United States and is a leading cause of morbidity and mortality in middle-aged persons.¹ Although the main physiological abnormalities are insulin resistance and hyperglycemia, the specific underlying mechanisms determining these changes remain uncertain.² Microvascular disease has been hypothesized to contribute to the development of diabetes.^{2 - 3} This is based on studies that demonstrate microvascular abnormalities (eg, impaired microvascular reactivity and flow in the skin and skeletal muscles) in persons with type 2 diabetes^{4 - 5} and in persons at high risk of developing diabetes, such as those with impaired glucose tolerance^{6 - 8} and first-degree relatives of persons with diabetes.^{8 - 9}

However, these studies are cross-sectional investigations among highly selected patient groups, often atypical of the general population.^{4 - 9} Prospective or population-based data are unavailable, largely because changes in the microcirculation are difficult to evaluate outside of experimental settings.¹⁰ The retinal arterioles offer an excellent opportunity to explore, noninvasively, the prognostic importance of microvascular disease.¹¹ In the Atherosclerosis Risk in Communities (ARIC) study, we developed a technique to quantify narrowing of the retinal arterioles, based on measuring their diameters on digitized photographs.¹² We have found retinal arteriolar narrowing to be strongly related to concurrent blood pressure and, independently, to past blood pressure,¹³ markers of inflammation,¹⁴ and risk of stroke.¹⁵

In this study, we examined the relation of retinal arteriolar narrowing to incident diabetes in middle-aged persons free of this condition in the ARIC study.

The ARIC study¹⁶ included 15 792 women and men 45 to 64 years of age at recruitment in 1987 through 1989. Population samples were selected from 4 US communities: Forsyth County, North Carolina; Jackson, Miss (black participants only); suburbs of Minneapolis, Minn; and Washington County, Maryland. Participants underwent 3 yearly follow-up examinations, in 1990 through 1992 (93% return rate), 1993 through 1995 (86% return rate), and 1996 through 1998 (81% return rate).

Retinal photographs were taken at the third examination (1993-1995).¹² Of the 12 887 participants who returned for this examination, we excluded 38 whose race was neither black nor white, 42 nonwhite residents in Minneapolis and Maryland (to permit stratification by race and field center), 2399 with prevalent diabetes, and 19 with retinal vascular occlusions, leaving 10 389 eligible for this study. Of these, 1372 had no retinal photographs or ungradable photographs and 1024 did not return for the fourth examination (1996-1998) or had incomplete data to confirm a new diagnosis of diabetes, leaving 7993 who provided data for this study. Comparisons of characteristics between participants included (n = 7993) and excluded (n = 2396) indicated that those included were younger and more likely to be white, had lower systolic and diastolic blood pressures, and lower fasting glucose levels, body mass indexes (BMIs), and waist-hip ratios (data not shown).

The retinal photography procedure has been described previously.¹² Briefly, photographs of the retina were taken of 1 randomly selected eye after 5 minutes of dark adaptation. These photographs were digitized by a high-resolution scanner and the diameters of individual arterioles and venules coursing through a specified zone surrounding the optic disc were measured on the computer monitor by trained graders masked to subject identity.¹² The individual arteriolar and venular diameters were combined into summary measures and expressed as an arteriole-to-venule ratio (AVR).¹² The AVR accounts for magnification differences between photographs and is distributed normally in the general population.¹² An AVR of 1.0 indicates that, on average, retinal arteriolar diameters are the same as venular diameters, while a smaller AVR represents narrower arterioles, since venular diameters vary little.^{12 - 13} The intragrader and intergrader reliability coefficients for AVR were 0.84 and 0.79, respectively.¹²

Trained graders also evaluated retinal photographic slides for focal lesions, including signs typical of diabetic retinopathy (eg, microaneurysms, retinal hemorrhages), arteriovenous nicking, and focal arteriolar narrowing, according to a standardized protocol.¹² These lesions were defined as present if graded as either definite or probable. Intragrader and intergrader κ statistics ranged from 0.61 to 1.00.¹²

Methods of ascertainment and diagnosis of diabetes in the ARIC study have been previously published.¹⁷ Participants were asked to fast for at least 12 hours before morning blood collection. Glucose was processed via a modified hexokinase/glucose-6-phosphate dehydrogenase procedure.

Diabetes mellitus was defined if any of the following criteria, adapted from the 1997 American Diabetes Association guidelines,¹⁸ were met: fasting (≥8 hours) serum glucose levels of at least 126 mg/dL (7.0 mmol/L), casual (fasting <8 hours) glucose levels of at least 200 mg/dL (11.1 mmol/L), use of diabetic medications, or physician-diagnosed diabetes. Individuals were defined as having incident diabetes if they did not develop diabetes through the third examination but met any of these criteria at the fourth examination.

Participants underwent an extensive interview, physical examination, and laboratory investigations. Race was self-reported. A positive family history of diabetes was defined by participant report of diabetes in either biological parent. Physical activity was characterized by sports and leisure activity indexes (range, 0-5) described elsewhere.¹⁹ Blood pressure was taken with a random-zero sphygmomanometer and the mean of the last 2 measurements was used. Mean arterial blood pressure was computed as two thirds of the diastolic value plus one third of the systolic value, and the average of this over the first 3 examinations (ie, 6-year mean arterial blood pressure) was included as a covariate in the assessment of the independence of retinal arteriolar narrowing with incident diabetes.^{14 - 15} Height and weight were taken with participants in scrub suits and BMI was calculated. Waist-hip ratio was computed as the circumference of the waist (umbilical level) divided by the hips (maximum buttocks). Blood collection and processing for fasting insulin, total cholesterol, high-density lipoprotein cholesterol, triglycerides, white blood cell count (WBC), plasma fibrinogen, factor VIII, and von Willebrand factor (vWF) are described elsewhere.²⁰ Average internal carotid intima-media wall thickness (IMT) was obtained from standardized B-mode ultrasonograms.²¹ All variables were measured at the third examination, except for insulin, WBC, fibrinogen, factor VIII, vWF, and carotid IMT, which were measured at the first examination, and mean arterial blood pressure, which was averaged over the first 3 examinations.

The AVR was categorized into quartiles (with the first quartile indicating the most severe arteriolar narrowing and the fourth the least) and also analyzed as a continuous variable (per 1-SD difference in AVR). We used analysis of covariance to compare the AVR and its components (summary measures of retinal arteriolar and venular diameters) between persons who did and did not develop diabetes. We used logistic regression models to estimate the odds ratios (ORs) and 95% confidence intervals (CIs) of incident diabetes, comparing a given quartile of AVR vs the fourth quartile or a 1-SD difference in AVR. We initially adjusted for age (years), sex, race, and field center. In multivariable models, we additionally adjusted for 6-year mean arterial blood pressure, fasting glucose levels, fasting insulin levels, family history of diabetes, BMI, waist-hip ratio, sports and leisure activity indexes, high school education, pack-years of cigarette smoking, alcohol consumption status (ever, never), total and high-density lipoprotein cholesterol levels, and triglyceride levels. In a separate model, we also adjusted for WBC (1000 cells/mm³), plasma fibrinogen, factor VIII, vWF, and carotid IMT, since these are possible confounders.¹⁴

We performed the following supplementary analyses. First, we used a higher cutoff (≥141 mg/dL [7.8 mmol/L]) for the fasting glucose value required for the diagnosis of diabetes (alternate diabetes definition). Second, we repeated analyses excluding persons with signs typical of diabetic retinopathy (eg, microaneurysms), since these persons may have preexisting diabetes. Third, we evaluated associations separately in persons with and without impaired fasting glucose levels (110-124 mg/dL [6.1-6.9 mmol/L]) at the third examination and tested interactions with other diabetes risk factors by stratification and by inclusion of cross-product terms in the logistic regression.

The baseline mean AVR in the population was 0.843 (SD, 0.08; median, 0.844; range, 0.57-1.22). Table 1 compares baseline characteristics of persons with retinal AVR falling in the first quartile (0.57-0.79) with those in the fourth (0.91-1.22). In general, persons in the first compared with the fourth AVR quartile were older, more likely to be men and to be black, and, after adjusting for age, sex, race, and field center, more likely to have a poorer diabetes risk profile.

Over a median follow-up of 3.5 years (range, 0.7-5.5 years), 291 (3.6%) persons developed diabetes. The incidence of diabetes increased from 2.4% to 5.2% with decreasing quartiles of AVR (Table 2). After adjustment for age, sex, race, and field center, persons in the lowest compared with the highest AVR quartile were twice as likely to develop diabetes (OR, 2.09; 95% CI, 1.47-2.98). This association was attenuated but still significant after additional adjustment for fasting glucose and insulin levels, family history of diabetes, and other risk factors (OR, 1.71; 95% CI, 1.13-2.57). These associations were somewhat stronger when we used an alternate definition of incident diabetes (fasting glucose ≥141 mg/dL [7.8 mmol/L] used as the cutoff) (Table 2).

When persons with signs of diabetic retinopathy (n = 382) were excluded, the association was similar (OR, 1.75; 95% CI, 1.15-2.66, comparing the first with the fourth AVR quartile). Further adjustment for WBC, plasma fibrinogen, factor VIII, vWF, and carotid IMT in the multivariable model resulted in a slightly stronger association (OR, 1.98; 95% CI, 1.15-3.41, comparing the first with the fourth AVR quartile).

The results of AVR analyzed as a continuous variable are presented in Table 3. Each 1-SD decrease in the AVR (a decrease of 0.08) in the total sample was independently associated with a 26% increase in risk of diabetes (OR, 1.26; 95% CI, 1.09-1.46). This association was generally similar in people stratified into "high-risk" and "low-risk" baseline characteristics, being possibly stronger in black than in white participants and in persons with lower BMIs and waist-hip ratios. Analysis repeated using the logarithmic transformation of AVR did not improve the fit of the models presented (data not shown).

Focal signs of hypertensive retinopathy, such as arteriovenous nicking and focal arteriolar narrowing, were not related to incident diabetes (adjusted OR, 0.94; 95% CI, 0.65-1.35, and adjusted OR, 1.12; 95% CI, 0.78-1.59, respectively).

In this prospective cohort study of middle-aged persons without diabetes, retinal arterioles were significantly narrower in persons who subsequently developed diabetes during the ensuing 3.5 years compared with those who did not. After controlling for fasting glucose level and insulin level 6 years prior, family history of diabetes, blood pressure, adiposity, physical activity, and other known risk factors, retinal arteriolar narrowing (as reflected by a lower AVR) was independently associated with increased risk of diabetes. This association persisted with different diabetes definitions, and was seen even among people at lower risk of developing this condition, including those without a family history of diabetes, those without impaired fasting glucose, those physically more active, and those with lower measures of adiposity.

Retinal arteriolar narrowing is a marker of microvascular damage from aging, hypertension, inflammation, and other processes.¹¹ It reflects intimal thickening and medial hyperplasia, hyalinization, and sclerosis seen histopathologically.¹¹ Because similar arteriolar changes associated with hypertension are well documented elsewhere in the body,^{22 - 23} the retinal arterioles appear to offer insights into the state of the systemic arterioles in health and disease.

Our finding provides prospective clinical evidence to support a key hypothesis in the pathogenesis of diabetes. The microcirculation, estimated using measures of microvascular reactivity and blood flow in the skin and skeletal muscles, is known to be insulin sensitive (ie, insulin stimulates microvascular dilation and flow).²⁴ Several cross-sectional studies have demonstrated microvascular alterations in persons with type 2 diabetes^{4 - 5} and in those at high risk of developing diabetes, including persons with impaired glucose tolerance,^{6 - 8} first-degree relatives of persons with diabetes,^{8 - 9} persons who are obese,²⁵ and persons with hypertension.^{26 - 27} Because these microvascular alterations may result in a reduced ability of insulin to mediate glucose uptake in skeletal muscles, microvascular disease has been suggested to play a causal role in the development of diabetes.^{2 - 3} Our study now suggests that arteriolar narrowing precedes the onset of diabetes, and may even play a role in its initial development. This relationship was independent of impaired fasting glucose, a family history of diabetes, and obesity and hypertension.

Our data also offer insights for a number of diverse observations regarding the epidemiology of diabetes. Although hypertension^{28 - 29} and cigarette smoking^{30 - 31} are related to the risk of diabetes, the underlying mechanisms are unclear. Microvascular alterations (eg, increased arteriolar resistance and reduced microvascular flow) have also been shown to precede hypertension^{26 - 27} and have therefore been hypothesized to explain the excess risk of diabetes in persons with hypertension.³² The risk of diabetes associated with cigarette smoking may be related to inflammation and consequent microvascular injury, since inflammation itself appears to play an important role in the development of diabetes.^{33 - 34} We have previously demonstrated that retinal arteriolar narrowing is related to long-term average blood pressure levels¹³ and independently related to cigarette smoking and markers of inflammation.¹⁴ Thus, it is possible that arteriolar narrowing, resulting from hypertension, cigarette smoking, inflammation, and other unmeasured processes, may be a common pathophysiological link to diabetogenesis.

The strengths of the current study include its population-based design, the quantitative and masked evaluation of retinal arteriolar diameters, standardized identification of incident diabetes cases, and detailed information on risk factors. However, some limitations should be highlighted. First, given the imprecision of a single glucose determination and the relatively short follow-up, misclassification of diabetes may occur. Thus, arteriolar narrowing may be a marker of underlying diabetes in persons not yet meeting the diagnostic criteria. However, this association was independent of fasting glucose, insulin measured 6 years earlier, and other risk factors; persisted with a different diabetes definition; and was observed in persons without signs of diabetic retinopathy (a marker of underlying disease). Moreover, the fact that the associations were largely similar in those at higher and lower risk of diabetes probably minimizes the likelihood of misclassification biases. Second, selection bias may have masked some associations or accentuated others. Retinal photographs were taken 6 years into the ARIC study, and if persons with arteriolar narrowing at risk of developing diabetes were more likely to die prior to photography, these associations could be falsely attenuated. Third, we have only shown a short-term association between AVR and incident diabetes; further study is required to determine whether longer-term associations exist. Finally, no data on the use of different vasodilator medications were available.

In conclusion, this population-based study documents a prospective association of retinal arteriolar narrowing to incident diabetes mellitus in middle-aged persons, independent of known risk factors. This finding suggests that microvascular processes may play a role in the development of diabetes.