The Relation of Retinal Vessel Caliber to the Incidence and Progressionof Diabetic Retinopathy: Title and subTitle BreakXIX: The Wisconsin Epidemiologic Study of Diabetic Retinopathy FREE

Despite the efficacy of glycemic and blood pressure control and photocoagulationtreatment in reducing visual loss in persons with diabetes, retinopathy remainsan important cause of visual loss.¹ Early detectionby means of dilated eye examinations and close follow-up of those with advancedretinopathy has been advocated to minimize visual loss in persons with diabetes.² Hyperglycemia, hypertension, microalbuminuria status,duration of diabetes, and pregnancy have been used to define increased riskfor progression of retinopathy and to determine the frequency of such examinations.³ Other than retinopathy severity, no ocular factorshave been consistently associated with the risk for progression to proliferativeretinopathy and incidence of macular edema. Dilated retinal venules have beenconsidered to be common in persons with diabetes, but were difficult to assessconsistently and were not included in schemes classifying the severity ofdiabetic retinopathy.⁴ Nevertheless, dilatedretinal venules have been associated inconsistently with an increased riskfor progression to proliferative retinopathy in persons with diabetes.^{5 - 12} However,most of these observations have been made on selected small samples usingophthalmoscopy, and few have come from population-based studies.

In the Atherosclerosis Risk in Communities Study, we developed a techniqueto quantify vessel caliber in the retinal arterioles and venules based onmeasuring their diameters on digitized fundus photographs.¹³ Usingthis technique, our group recently reported that narrowed retinal arterioleswere associated with 3-year incidence of diabetes in the Atherosclerosis Riskin Communities Study, suggesting that microvascular caliber changes occurin the preclinical stages of diabetes.¹⁴ Thepurpose of this report is to describe the relation of retinal arteriolar andvenular caliber and arteriole-venule ratio (AVR) with the incidence and progressionof diabetic retinopathy and the incidence of macular edema in a large population-basedstudy of persons with younger-onset (type 1) diabetes.

The population, which has been described in previous reports, consistedof a probability sample selected from 10 135 diabetic patients who receivedprimary care in an 11-county area in southern Wisconsin from 1979 to 1980.^{15 - 20} Thissample was composed of a younger-onset group (all patients diagnosed as havingdiabetes before 30 years of age who took insulin [n = 1210]) and an older-onsetgroup.

Data from participants in the younger-onset group who underwent at least1 follow-up examination form the basis of this report. Of this group, 996participated in the baseline examination (1980-1982),¹⁶ 891in the 4-year follow-up,¹⁷ 765 in the 10-yearfollow-up,¹⁹ and 634 in the 14-year follow-up.²⁰ Reasons for nonparticipation and comparisons betweenparticipants and nonparticipants at baseline and all the follow-up examinationshave been presented elsewhere.^{15 - 20} Theprincipal reason for nonparticipation was death, accounting for 64 nonparticipantsat the 4-year follow-up, an additional 86 at the 10-year follow-up, and anadditional 64 at the 14-year follow-up.

The baseline and follow-up examinations were performed in a mobile examinationvan in or near the cities where the participants resided. All examinationsfollowed a similar protocol that was approved by the institutional human subjectscommittee of the University of Wisconsin–Madison. The pertinent partsof the ocular and physical examinations included measurement of blood pressure,²¹ dilation of the pupil, 30° stereoscopic colorfundus photography of 7 standard fields,²² asemiquantitative determination of protein levels in the urine using a urinechemistry test strip (Labstix; Ames, Elkhart, Ind), and determination of glycosylatedhemoglobin A₁ levels from a capillary blood sample.^{23 - 24}

A structured interview was conducted by the examiners and included questionsabout specific medications for control of hyperglycemia and blood pressure.Any question about medication use was verified by a physician's report.

Grading protocols have been described in detail elsewhere^{17 ,25} andare modifications of the Early Treatment Diabetic Retinopathy Study (ETDRS)adaptation of the modified Airlie House classification of diabetic retinopathy.^{26 - 27} Interobserver and intraobserver variationsand the validity of the systems have been evaluated, and the results havebeen presented elsewhere.^{17 ,25 ,27 - 28}

For each eye, the maximum grade in any of the 7 standard photographicfields was determined for each of the lesions and used in defining the retinopathylevels varying from level 10 (no retinopathy) to level 60 or greater (proliferativeretinopathy); definitions have appeared elsewhere.^{19 ,27} Theretinopathy level for a participant was derived by concatenating the levelsfor the 2 eyes, giving the eye with the higher level greater weight. Thisscheme provided a 15-step scale. For purposes of classification, if the retinopathyseverity could not be graded in an eye, it was considered to have a scoreequivalent to that of the other eye.

The incidence of any retinopathy was estimated from all persons whohad no retinopathy at the baseline examination (severity level, 10/10) andwho participated in the follow-up examination(s). Progression to proliferativeretinopathy was estimated from all persons who were free of this complicationat the baseline examination. For persons with nonproliferative or no retinopathy,progression was defined as the first instance of an increase in the severityof retinopathy by 2 steps or more from the baseline level at any of the follow-upexaminations.

Macular edema was defined as thickening of the retina with or withoutpartial loss of transparency within 1 disc diameter (DD) from the center ofthe macula²⁹ or the presence of focal photocoagulationscars in the macular area associated with a history of development of macularedema as documented by stereoscopic fundus photographs. Clinically significantmacular edema was based on the detailed gradings and was defined as the presenceof any one of the following: retinal thickening at or within 500 µmof the center of the macula; hard exudates at or within 500 µm of thecenter of the macula if associated with thickening of the adjacent retina;and/or a zone or zones of retinal thickening 1 disc area in size, at leastpart of which was within 1 DD of the center.²⁹ Wheneverwe found new signs of photocoagulation scars in the macular area in the absenceof macular edema and we had not previously documented macular edema by gradingfundus photographs taken at an earlier examination, we obtained fundus photographsfrom the participant's ophthalmologist. In the absence of fundus photographs,we obtained medical records documenting that macular edema due to diabeteshad been present before the focal (or grid) photocoagulation. In situationswhere participants gave a history of laser photocoagulation but no signs oftreatment burns were seen, we requested information from the treating ophthalmologistto verify that such treatment had been performed and to ascertain whethermacular edema had been present before focal laser treatment. If macular edemacould not be graded in an eye, the individual was assigned the score of theother eye. The incidence of macular edema was estimated from data for allpersons who had no macular edema and had not been treated previously withphotocoagulation at the baseline examination, and who participated in at least1 follow-up examination.

Diameters of retinal vessels were measured after converting the photographsof field 1 to digital images. All arterioles and venules were measured inthe area between 0.5 and 1 DD from the optic disc margin using a computer-assistedprogram. Computer-assisted measurements of individual arterioles and venuleswere each combined according to formulas developed by Parr and Spears^{30 - 31} and Hubbard et al¹³ toprovide the average diameters of retinal arterioles (central retinal arteriolarequivalents [CRAE]) and venules (central retinal venular equivalents [CRVE])in that eye. These were then expressed as an AVR. An AVR of 1.0 indicatesthat, on average, retinal arteriolar diameters are the same as venular diameters,whereas a smaller AVR represents narrower arterioles or larger venules. Forthe primary analyses, data from both eyes were combined by computing the meanif both eyes were gradable (n = 754) or by using the value of the single gradableeye (n = 117). The range of the CRAE was 117.1 to 306.6 µm; of the CRVE,167.6 to 325.2 µm; and of the AVR, 0.587 to 1.119.

Age was defined as the age at the time of the baseline examination in1980 to 1982. Age at diagnosis of diabetes mellitus was defined as the ageat the time the diagnosis was first recorded by a physician on the patient'smedical chart or in a hospital record. The duration of diabetes was the timefrom the age at diagnosis to the age at the baseline examination.

The mean systolic and diastolic blood pressures at baseline were theaverages of the last 2 of 3 measurements obtained according to the protocolof the Hypertension Detection and Follow-up Program.²¹ Inindividuals 25 years or older, hypertension at baseline was defined as a meansystolic blood pressure of at least 160 mm Hg, a mean diastolic blood pressureof at least 95 mm Hg, and/or a history of antihypertensive medication at thetime of examination; in younger persons, it was defined as a mean systolicblood pressure of at least 140 mm Hg, a mean diastolic blood pressure of atleast 90 mm Hg, and/or a history of antihypertensive medication at the timeof examination. Pulse pressure at baseline was defined as the systolic minusthe diastolic blood pressure. Mean arterial blood pressure at baseline wasdefined as diastolic blood pressure +(systolic blood pressure − diastolicblood pressure)/3. Body mass index at baseline was defined as weight in kilogramsdivided by the square of height in meters. Proteinuria was defined as a urineprotein concentration of 0.03 g/dL or greater.

We computed the 4-, 10- and 14-year cumulative incidences. Some participantswho were observed at the 4- or 10-year examinations and were still at riskfor development of an end point did not participate in the later examinations.Thus, these are censored observations. To compute cumulative 14-year rateswhile still using the information contained in these censored observations,the product-limit method was used.³² To testfor trends in the rates of incidence or progression and to compute relativerisks, we used the Mantel-Haenszel method³³ stratifiedon the 3 follow-up periods. Multivariable analyses were based on logisticregression for 4-year end points and the discrete linear logistic model for10- and 14-year end points.^{34 - 35} Continuousvariables were used as such. For the retinal vessel variables, the means ofthe eyes were used as the per person measurements. The variables used in allthe models for the whole cohort, except for macular edema as an end point,were duration of diabetes, sex, glycosylated hemoglobin level, mean arterialblood pressure, antihypertensive medication use, and severity of retinopathyat baseline. As the risk factors of interest (ie, the retinal vessel measurements)and the end points (ie, incidence and progression of retinopathy) are eyespecific, we also performed analyses using the eye-specific data. We usedgeneralized estimating equation models to evaluate these relationships.³⁶

Characteristics of the cohort at baseline are presented in Table 1. Persons were excluded if theydid not participate in the 1984-1986 follow-up (n = 105) or did not have gradableretinal vessels in at least 1 eye (n = 20). Persons included in the studywere younger, had shorter duration of diabetes and lower systolic and diastolicblood pressure, and were less likely to have hypertension, gross proteinuria,and proliferative retinopathy.

The 4-, 10-, and 14-year cumulative incidence and progression of retinopathyand incidence of macular edema appear in Table 2. The cumulative incidence and rates of progression of retinopathywere high and increased with longer follow-up. By the 14th year of follow-up,85.7% of the cohort had had progression of their retinopathy, proliferativeretinopathy had developed in 37.4%, and clinically significant macular edemahad developed in 17.2%.

The crude associations of retinal vascular measures at baseline withincidence and progression of retinopathy at 4, 10, and 14 years are presentedin Figure 1, Figure 2, Figure 3, and Figure 4. Multivariable models controllingfor factors previously shown to be related to the end points of interest arepresented for the 4-year data in Table 3. The CRAE, CRVE, and AVR at baseline were not associated with incidenceof retinopathy (Figure 1 and Table 3). There were few differences whenthe multivariable model was rerun excluding mean arterial blood pressure orantihypertensive medication use (data not shown).

Larger CRAE and CRVE were associated with greater 4-year progressionof retinopathy (Figure 2). The associationsremained while controlling for sex, duration of diabetes, glycosylated hemoglobinlevel, mean arterial blood pressure, use of antihypertensive medications,and baseline retinopathy severity level (Table 3). Interactions of CRAE, CRVE, and AVR with baseline retinopathylevel were examined. There was evidence of interactions between baseline retinopathyand CRVE (P = .06) and AVR (P =.04). The direction of this interaction between retinopathy and CRVE is suchthat at higher levels of retinopathy, the direct relationship between CRVEand progression is greater. For AVR, at higher levels of retinopathy, theinverse relationship between AVR and progression is greater. The associationswere no longer statistically significant after 10 years of follow-up.

Larger CRVE and smaller AVR were strongly associated with greater 4-,10-, and 14-year incidence of proliferative retinopathy (Figure 3). The CRAE was not associated. While controlling for sex,duration of diabetes, glycosylated hemoglobin level, mean arterial blood pressure,use of antihypertensive medications, and retinopathy severity level, proliferativeretinopathy was 4 times as likely to develop at 4 years in participants inwhom the CRVE was in the fourth quartile range at baseline (95% confidenceinterval [CI], 1.50-12.19) compared with participants in the first quartilerange (Table 3).

Larger CRVE and smaller CRAE at baseline were associated with greater4-year and 14-year incidence of macular edema, respectively (Figure 4). Smaller AVR was associated with greater incidence ofmacular edema at 4 and 10 years but not at 14 years of follow-up. Multivariableanalyses, controlling for glycosylated hemoglobin, gross proteinuria, andretinopathy severity level at baseline, showed that these relationships withincident macular edema were no longer statistically significant for CRAE,CRVE, and AVR (data not shown).

Findings were similar when we used generalized estimating equations(data not shown). There were no significant interactions with hypertensivestatus at baseline (data not shown).

Glycosylated hemoglobin and the retinal vessel variables are importantvariables with missing values (4.6% and 12.6%, respectively). To evaluatethe effect this missing information has on the results, the multivariablemodels were rerun with these cases included with indicator variables to denotethe state of the relevant variable being missing. This procedure resultedin no substantive changes in the associations between the retinal vessel variablesand any of the end points (data not shown).

The data reported herein provide unique long-term population-based informationregarding the relation of retinal vessel caliber to the incidence and progressionof diabetic retinopathy and macular edema during a 14-year follow-up. Usinga new computer-assisted technique to quantify measurement of the retinal arteriolesand venules previously used in the Atherosclerosis Risk in Communities Study,¹³ we report that, in persons with younger-onset diabetes,greater vascular retinal caliber is not associated with incidence of retinopathy,larger arteriolar and venular diameters are associated with progression ofretinopathy, and larger venular diameter at baseline is associated with theincidence of proliferative diabetic retinopathy.

The association in the Wisconsin Epidemiologic Study of Diabetic Retinopathy(WESDR) of larger venular diameter with progression of retinopathy and incidenceof proliferative retinopathy, independent of retinopathy severity, is consistentwith data from earlier studies.^{4 - 12 ,37} Retinalvenous caliber abnormalities were important predictors of visual loss dueto progression of retinopathy in the Diabetic Retinopathy Study (DRS).³⁸ Retinal venous beading, an irregular venular dilation,was significantly associated with progression of proliferative disease inthe DRS and ETDRS.^{26 - 27} Basedon these observations, the presence of venous beading was used to define severenonproliferative retinopathy severity levels 47 and 53 in the ETDRS severityscale.²⁷ In the present study in the WESDR,the association of our measurements of venous caliber with progression ofretinopathy remained independent of baseline retinopathy, hypertension status,and glycemic control, suggesting that larger venous caliber provides informationindependent of the current ETDRS system used to classify retinopathy severitylevel. Increase in venular diameter in eyes with retinopathy is thought toresult from retinal hypoxia³⁹ and lactate accumulationresulting from hyperglycemia.⁴⁰ Successfultreatment of proliferative diabetic retinopathy with panretinal photocoagulationhas been shown to be accompanied by a reduction of venular caliber, thoughtto result from reduction of retinal hypoxia.^{10 ,41 - 43}

We hypothesized that before the onset of retinopathy, eyes in diabeticsubjects with narrower retinal arterioles, dilated venules, and smaller AVRwould be associated with an increased risk for incident retinopathy. Thiswas based on a number of observations. First, retinal arteriolar narrowingis associated with hypertensive and arteriosclerotic microvascular changes,endothelial dysfunction, and inflammatory changes, and these factors are thoughtto be involved in the pathogenesis of diabetic retinopathy.^{44 - 47} Second,microvascular compliance is lower in diabetic subjects without retinopathycompared with nondiabetic subjects.⁴⁸ Third,retinal venular dilation had been observed in some studies in diabetic personswithout retinopathy.^{49 - 52} Analternative hypothesis was that before the onset of retinopathy, eyes in diabeticsubjects with dilated arterioles would be at increased risk for incident retinopathydue to a breakdown in autoregulation.⁵³ However,we found no association of arteriolar or venular caliber with the incidenceof retinopathy. Our findings regarding a lack of an association of incidentretinopathy with venular caliber are consistent with data from earlier cross-sectionalstudies in which venular diameter was similar in the eyes of nondiabetic personsand diabetic persons without retinopathy.^{8 ,10 ,54 - 55} In45 children with type 1 diabetes mellitus, Falck and Laatikainen¹⁰ reportedhigher but statistically nonsignificant venular diameter at baseline in eyeswith incident retinopathy compared with eyes without incident retinopathy.They found that other signs of retinopathy developed in those eyes with greaterthan 10 µm of venous dilation during the follow-up period more oftenthan in patients with less or no change in the venous caliber. We have notmeasured change in venous caliber over time in our study. Our findings suggestthat retinal vascular caliber measurements in those without retinopathy havelittle prognostic value before the development of retinopathy in persons withtype 1 diabetes.

The strength of our study includes its large size, the broad distributionof severity of retinopathy at baseline, the defined and uniform follow-upintervals, and the low refusal rate. In addition, standardized protocols ofmeasurement, including computer-assisted measurement of retinal vessel caliberand objective recording of diabetic retinopathy and macular edema using stereoscopicfundus photographs of 7 standard fields, were consistent over time. Gradingof fundus photographs was performed in masked fashion using a standard classificationsystem. However, caution must be observed in interpreting the findings inthe present study. First, relationships may have been attenuated by selectivesurvival. Retinopathy severity and retinal arteriolar narrowing have beenshown to be related to mortality in persons in the diabetic population.^{56 - 57} This would reduce the associationsbetween arteriolar narrowing and progression of disease and incident proliferativeretinopathy, especially in those followed up for a long time. Second, we didnot control for variations in pulsatility associated with the cardiac cycle.This has been estimated to cause up to a 6% variation for arteriolar and venulardiameters.^{58 - 59} If we assume thatthe photographs were taken at random during the cardiac cycle, the resultingincrease in variability of the gradings might result in an attenuation ofour findings. Third, inability to find some associations of retinal vascularcaliber width with incident or progressed retinopathy may be due, in part,to the limited area near the disc in which these blood vessels are measured.

Measurement of venous caliber may provide additional information regardingprogression of and risk for development of proliferative retinopathy thanretinopathy severity itself in persons with type 1 diabetes mellitus. However,vessel caliber is not predictive of incident retinopathy. Further study isneeded to examine whether retinal vessel caliber provides more informationabout systemic microvascular and macrovascular disease than severity of retinopathyin persons with type 1 diabetes.

Corresponding author: Ronald Klein, MD, MPH, Department of Ophthalmologyand Visual Sciences, University of Wisconsin–Madison, 610 N Walnut St,450 WARF, Madison, WI 53726-2336 (e-mail: Kleinr@epi.ophth.wisc.edu).

Reprints not available from the authors.

Submitted for publication February 3, 2003; final revision receivedJune 9, 2003; accepted September 10, 2003.

This study was supported by grants EY03083 and HL59259 from the NationalInstitutes of Health, Bethesda, Md (Drs R. Klein and B. E. K. Klein), andin part by a Senior Scientific Investigator Award from Research to PreventBlindness, New York, NY (Dr R. Klein).

We are grateful to the 452 Wisconsin physicians and their staffs whoparticipated in and supported this study.