Retinal Vessel Calibers Predict Long-term Microvascular Complications in Type 1 Diabetes: The Danish Cohort of Pediatric Diabetes 1987 (DCPD1987) Free

Diabetic microvascular complications, namely diabetic peripheral neuropathy (DPN), diabetic nephropathy (DN), and diabetic retinopathy (DR), are common in type 1 diabetes (1) despite advances in metabolic care. Identifying new predictors of microvascular disease could be helpful for early individual risk stratification, which may provide better opportunity for timely implementation of effective interventions. For this purpose, the retinal vasculature provides a unique opportunity to assess vascular health directly and noninvasively in vivo.

Studies have investigated how retinal vessel calibers are associated with microvascular complications in both type 1 and type 2 diabetes (2–14). The evidence to date suggests that wider retinal venular diameters are associated with presence of both DN and severe levels of DR in several cross-sectional studies (3,8–10) and with incident DN (4,5) and progression to severe DR (13,14) in prospective studies. No relations between retinal venular diameters and DPN have been shown thus far (2,4,6,7).

Concurrently, narrower arteriolar diameters have been linked to severe DR (3,8) and DN (2,3) in cross-sectional reports, but the results for DPN have been inconclusive (2,6,7). Two prospective studies proposed that in type 1 diabetes, wider arteriolar diameters are associated with an increased risk of incident DR (11,12), whereas other groups showed no longitudinal correlation between arteriolar diameters and diabetic microvascular complications (4,5,13).

Because most previous studies on retinal vascular calibers and diabetes complications are cross-sectional or short-term follow-up studies, there is very limited information on the predictive value of these measurements. Prospective studies with long follow-up periods are needed to define the link between retinal vascular calibers and diabetic microvascular complications.

The aim of this study was to investigate the predictive value of retinal vascular calibers on microvascular complications in a 16-year prospective study of a young population-based Danish cohort with type 1 diabetes.

The participants in this study were identified from a nationwide population-based pediatric cohort of Danish children with type 1 diabetes who were initially studied in 1987–1989 (n = 720), the Danish Cohort of Pediatric Diabetes 1987 (DCPD1987). Detailed characteristics of this cohort have been reported elsewhere (15–22). Of this cohort, 339 were included in a baseline examination in 1995 and thus eligible for 16-year follow-up in 2011. Of these, 15 (4.4%) were excluded for follow-up due to missing baseline retinal photographs, 13 (3.8%) had died, and 63 (18.6%) were unreachable because they had emigrated or could not be contacted due to research protection. Of the remaining 248 patients, 185 (74.6%) participated at the follow-up in 2011 and 63 (25.4%) declined to participate. The study was performed in accordance with the criteria of the Helsinki II Declaration and was approved by the local scientific ethics committee. All patients gave a written informed consent at baseline and at follow-up examinations.

In 1995, information on sex, age, diabetes duration, HbA_1c, blood pressure, BMI, vibration perception threshold (VPT), determination of mean albumin excretion rate (AER), and presence of DR was collected. The methods used have been described elsewhere (16); however; baseline micro- and macroalbuminuria were defined as a mean AER of 20–200 µg/min and >200 μg/min, respectively, in at least two timed overnight urine collections.

Color retinal photographs were taken after pupillary dilation by certified operators using 40–60° retinal cameras in accordance with the European Diabetes Study Group (EURODIAB) recommendations (23). The film slides were later digitalized with the DigitDia 5000 (Reflecta, Rottenburg, Germany) film scanner. Two photographs, a macular-temporal field and a disc-nasal field, were taken of each eye. The latter was used for analyses of vessel diameters.

A high intereye correlation has been shown previously, and consequently, we only used the right eye for assessment of vessel calibers (24). If the photograph of the right eye was not of sufficient quality for grading, the left eye was used. Retinal vascular calibers were assessed by the same certified and validated grader with a computer-based program following a previously validated protocol (24,25). Diameters of all vessels coursing completely through a zone 0.5–1 disc diameter from the disc margin were measured with IVAN image analysis software (Department of Ophthalmology Visual Science, University of Wisconsin, Madison, WI).

The calibers of the central retinal artery and vein were estimated using the “big-6 formula,” a method that combines diameter measurements of the six largest arterioles and the six largest venules into the central artery and vein equivalents (CRAE and CRVE, respectively) (26). In other words, the diameter of the six largest arterioles in the specified zone is used to calculate the diameter of the central retinal artery as it enters from the optic disc. CRAE is therefore an estimated central retinal artery diameter based on measurements of its branches. The method for calculating CRVE is similar.

For assessment of baseline DR, all photos were graded by the same certified grader by a modified Early Treatment of Diabetic Retinopathy Study (ETDRS) protocol, allowing for nonstandard photography (15,27).

The examinations took place between 1 January and 1 November 2011. The participants underwent a clinical examination where VPT was measured in triplicate on the apex of the right first toe with a hand-held biothesiometer (Bio-Medical Instrument Company, Newbury, OH). The mean of the last two measurements was calculated, and a mean VPT of >25 volts was considered abnormal. Participants with an abnormal mean VPT were classified as having DPN.

Spot urine samples were collected and analyzed centrally to determine the urine albumin-to-creatinine ratio (ACR). Participants with ongoing infections or who were menstruating were asked to hand in urine samples at a later time. Laboratory measurements of albumin and creatinine were performed on an Abbott Architect analyzer by immunoturbidimetric and enzymatic assays, respectively. A spot urine ACR was then calculated and microalbuminuria was defined as ACR = 30–299 mg/g and macroalbuminuria as ACR ≥300 mg/g. Participants with macroalbuminuria at follow-up, or who had had a renal transplant or had received dialysis, were classified as having DN. We used the Danish Patient Registry to identify those who had dialysis/renal transplant among the patients who declined to participate.

Participants underwent an ophthalmological examination, including retinal photography where seven 45° fields were taken of each eye in accordance with the ETDRS standards (28). Nonstereoscopic, digital color retinal photographs were performed in mydriasis (Tropicamide 1% and epinephrine 10%) by the same trained and certified operator using 3D OCT-2000 Spectral Domain OCT (Topcon, Tokyo, Japan). Images were captured at a resolution of 4.288 × 2.848 pixels. The retinal photographs were graded by the same certified grader by a modified ETDRS scale (15,27). Participants were classified as having proliferative diabetic retinopathy (PDR) if they had ETDRS level 61 or higher on the eye included in the vessel analysis. We used the Danish Patient Registry to identify those who had had panretinal photocoagulation among the patients who declined to participate to get an estimate of the incidence rate of PDR in this group.

Categorical data are presented as percentages, whereas continuous data are presented as means ± SD. We used the Mann-Whitney U test for differences between two groups, the Cuzick test for trend for several groups, and the Spearman rank correlation to test for associations between two continuous variables.

Multiple logistic regression analyses with backward selection were performed to estimate odds ratios (ORs) for incident DPN, DN, and PDR. All baseline variables (sex, age, diabetes duration, HbA_1c, blood pressure, BMI, VPT, mean AER, and level of retinopathy) were included in each of these models.

All participants with VPT <25 volts at baseline were considered at risk for incident DPN, and all participants without PDR at baseline were at risk for incident PDR. Patients with baseline mean AER >200 µg/min (overt DN) were excluded from analyses on renal function at follow-up.

The 95% CI are given for estimates of ORs and were considered statistically significant when they did not cross 1.0. Findings with a P value <0.05 were considered statistically significant.

For analyses on cutoff points for CRAE and CRVE, we generated a series of dichotomous variables for each 10-μm increase in diameter (i.e., over/under CRAE = 140 µm, over/under CRAE = 150 μm, etc.). We ran multiple logistic regressions for every new CRAE and CRVE variable, for each end point, and plotted receiver operating characteristics curves for each regression model to determine the area under the curve (AUC). We furthermore tested the equality of the highest versus the lowest AUC for each outcome with respect to both CRAE and CRVE cutoffs to determine if there was a significant difference in the predictive value of the models. If a significant difference was found between the lowest versus the highest AUC, we tested the second lowest AUC against the highest; if this difference was significant, we proceeded on to the third lowest versus highest, etc., to establish the span of the investigated threshold.

For the cutoffs producing the highest AUCs, the ORs were used to calculate an estimate of the risk increase (as a percentage) of a given outcome for participants below vs. above this threshold.

All statistical calculations were performed with STATA 11.1 software (StataCorp LP, College Station, TX).

All 185 participants (49.7% males) had at least one gradable baseline retinal photo for vessel analyses, and thus, none were excluded for this reason. The mean age and diabetes duration at baseline of the participants were 21.0 and 13.5 years, respectively. A comparison was done between the participants and the patients who were available for follow-up but declined to participate (Table 1). The patients who declined had significantly higher HbA_1c values at baseline than the participants, but the groups did not differ in any other respect.

The range of CRAE and CRVE were 123.8–197.4 µm and 175.2–296.2 µm, respectively, for the 185 participants and were 133.6–190.5 µm and 200.8–307.6 µm, respectively, for the 63 who declined. The distribution of CRAE and CRVE in these two groups is shown in Fig. 1.

Table 2 demonstrates how various baseline characteristics of the participants were associated with CRAE and CRVE when looked at separately. CRVE was significantly related to the level of retinopathy, diabetes duration, VPT, and systolic blood pressure, when analyzed one by one. Furthermore, there was a borderline association with diastolic blood pressure. CRAE was significantly related to systolic and diastolic blood pressure.

DPN was evaluated in 162 patients at follow-up who had had VPT <25 volts at baseline in 1995. Of these, 17 (10.5%) had developed DPN during the 16 years.

As reported in Table 3, multiple logistic regressions found CRAE (OR 2.96 per 10-μm decrease; 95% CI 1.36–6.45) and CRVE (OR 1.52 per 10-μm increase; 95% CI 1.01–2.30) were both significantly associated with the 16-year incidence of DPN. Other baseline risk factors associated with DPN were increasing age (OR 1.62 per 1-year increase; 95% CI 1.12–2.33), male sex (OR 3.62 vs. female sex; 95% CI 1.03–12.68), increasing HbA_1c (OR 3.53 per 1% increase; 95% CI 1.67–7.47), increasing diastolic blood pressure (OR 2.77 per 10-mmHg increase; 95% CI 1.16–6.61), and increasing VPT (OR 1.62 per 1-volt increase; 95% CI 1.14–2.30).

A total of 166 participants without macroalbuminuria at baseline in 1995 handed in urine samples at follow-up and were included in analyses of vessel diameters and incident DN. Of these, 13 patients (7.8%) had developed macroalbuminuria, had had a kidney transplant (n = 2), and/or had received dialysis (n = 4) at follow-up. Among the 63 patients who declined to participate at follow-up, 1 had had a kidney transplant, and none had received dialysis.

CRAE (OR 2.63 per 10-μm decrease; 95% CI 1.09–6.36) and CRVE (OR 1.76 per 10-μm increase; 95% CI 1.05–2.94) were both significantly associated with the 16-year incidence of DN in the multiple logistic regression model (Table 3). Of the other risk factors we adjusted for in the model, only increasing diastolic blood pressure (OR 4.06 per 10-mmHg increase; 95% CI 1.21–13.61) and increasing levels of HbA_1c (OR 2.18 per 1% increase; 95% CI 1.04–4.58) were associated with DN.

One of the 185 participants had PDR at baseline in 1995 and was excluded from this analysis. Of the 184 patients at risk, 50 (27.2%) progressed to PDR during the 16-year period on the eye chosen for vessel analyses. Among the 63 patients who declined participation in 2011, 26 (41.3%) had had panretinal photocoagulation performed in at least one eye.

As reported in Table 3, multiple logistic regressions found CRAE (OR 1.56 per 10-μm decrease; 95% CI 1.07–2.27) and CRVE (OR 1.36 per 10-μm increase; 95% CI 1.05–1.76) were both significantly associated with the 16-year incidence of PDR. Other baseline risk factors associated with PDR were HbA_1c (OR 2.44 per 1% increase; 95% CI 1.61–3.69), diastolic blood pressure (OR 1.95 per 10-mmHg increase; 95% CI 1.05–3.62), and baseline retinopathy at ETDRS level 35 (OR 6.15 vs. level 10; 95% CI 1.84–20.60). Baseline retinopathy at levels 20 and 43 versus level 10 was not significantly associated with incident PDR.

We used the following dichotomous variables for CRAE: over/under 140, 150, 160, 170, and 180 µm, and for CRVE: over/under 210, 220, 230, 240, 250, 260, 270, and 280 μm. The calculated AUCs for each multiple regression are reported in Table 4.

For CRAE, the highest AUC was produced with the cutoff 170 μm with respect to incident DPN and DN and with 180 μm for incident PDR. For each outcome, no significant difference was found when comparing the highest AUC with the lowest (P = 0.10, 0.10, and 0.18, respectively). Participants with CRAE of less than 170 μm had an increased risk of incident DPN and DN of 803% and 1,090%, respectively, compared with participants above this threshold (DPN: OR 9.03 [95% CI 1.25–65.3]; DN: OR 11.90 [95% CI 1.03–141.2]). Participants below CRAE 180 μm had a 436% increased risk of incident PDR compared with participants above (OR 5.36; 95% CI 1.15–24.9).

For CRVE, the highest AUC was seen with the cutoff 260 μm with respect to incident DPN and with 250 µm for DN and PDR. No significant difference was found when the highest AUC was compared with the lowest for all three outcomes (P = 0.25, 0.25, and 0.13, respectively).

Participants with CRVE above 260 μm had an increased risk of incident DPN of 592% compared with participants below this threshold (OR 6.92; 95% CI 1.32–36.4). Participants above CRVE 250 μm had a 341% and 224% increased risk of incident DN and PDR, respectively, compared with participants below (DN: OR 4.41 [95% CI 1.21–16.1]; PDR: OR 3.24 [95% CI 1.15–9.11]).

In this prospective study of young patients with type 1 diabetes, changes in retinal vascular caliber (wider venular caliber and narrow arteriolar caliber) were consistently associated with the 16-year incidence of the major diabetic microvascular complications, namely DPN, DN, and PDR.

Most previous studies have examined these complications separately, and no previous studies have investigated the relation between retinal vessel calibers and DPN prospectively; with cross-sectional reports inconclusive (2,4,6,7). Our results show a strong association between DPN and both narrow arteriolar and wide venular diameters, despite the relatively small number of cases.

A possible link between retinal vessel diameters and renal disease was illustrated earlier in the Wisconsin Epidemiologic Study of Diabetic Retinopathy, where larger retinal venular caliber was independently associated with the 10-year incidence of DN in type 2 diabetes (4) and the 16-year incidence in type 1 diabetes (5). Nevertheless, they found no associations with arteriolar caliber. Our data support the concept that retinal venular dilation is related to the development of DN in people with type 1 diabetes. Moreover, we observed a relation between narrower arteriolar calibers and the 16-year incidence of nephropathy, which has not been previously demonstrated longitudinally, but a similar association was presented in cross-sectional reports (2,3).

Hypertension is known to accelerate the progression of renal disease in patients with diabetes (29). In addition to the well-known association between narrow arteriolar calibers and coexisting hypertension (30–32), a meta-analysis and data pooling study showed that narrow retinal arteriolar calibers appear to precede clinical hypertension (33). Therefore, a plausible explanation for the link between narrower arteriolar caliber and incident DN seen in this population, compared with previously studied cohorts, is a higher prevalence of hypertension after the baseline examination.

Large venular calibers (3,8–10) and narrow arteriolar calibers (3,8) were formerly associated with the presence of severe DR in cross-sectional reports. We showed the same relation for the 16-year progression to PDR. Large venular calibers have previously been related to progression to severe DR in some studies with shorter follow-up periods (13,14). In contrast to our data, two prospective studies found larger arteriolar caliber was associated with an increased risk of incident DR (11,12), but another group could not show this correlation (13). However, these studies focused on the short-term incidence of DR and general progression of the disease and not incident PDR. Relations between hypertension and DR have been demonstrated earlier (34). As with our findings on retinal arteriolar narrowing and incident DN, we speculate that a high prevalence of hypertension after baseline examinations partly explains the association between narrower arteriolar caliber and PDR seen in our cohort.

The underlying pathophysiological mechanisms of the venular dilation observed in this study are not fully known. One hypothesis is retinal hypoxia (35,36). Another possible mechanism is inflammation, because larger retinal venules were linked to higher levels of inflammatory biomarkers (37,38). In addition, studies on vasodilatory response to flicker-light stimulation, a process reflecting the endothelial function of retinal vessels, propose that larger retinal venular and arteriolar calibers indicate endothelial dysfunction (39,40). Our findings on arteriolar calibers do not confirm this, though. It is possible that mechanisms stronger than endothelial dysfunction, such as hypertensive mechanisms, have a superior effect on the arterioles in our cohort.

In the pursuit for cutoff values, participants with CRAE values below 170–180 µm seemed to be at higher risk of microvasculopathy than those with higher values. However, this was only an observed trend, because we could not show a significant difference between the highest AUCs and even the lowest ones. Thus, it cannot be concluded that these thresholds are the best in predicting microvascular disease.

We similarly observed the highest AUCs for CRVE cutoffs in an area of 250–260 µm but found no significant difference compared with the lowest AUCs. We speculate that the relatively low number of participants in the study might be the reason these observed trends were not significant. Correspondingly, the number of participants is a plausible explanation for the relatively high-risk increase calculated for each outcome for patients below/above a given threshold compared with the opposite group. The logistic regression with CRAE and CRVE categorized by a chosen threshold produced rather large CIs, and these numbers should therefore be interpreted with caution. Future studies are clearly needed to establish the cutoffs and the risk estimations.

The strengths of this study include the population-based cohort, the long follow-up time, and use of quantitative methods for assessment of vascular calibers by standardized protocols. None of the participants had ungradable retinal photographs, and all were graded by the same, certified grader.

Most participants had little or no vascular disease at baseline, which provides a better estimate of the predictive value of retinal vessel caliber because progression to the end-stage of any complication is a major increase.

A limitation to this study is the relatively small number of patients; however; our results were statistically significant irrespective of this.

Intermediate measurements on the cohort would provide valuable information on the timeline in the development of complications, but such data are not available, unfortunately.

Nephropathy at baseline and follow-up was assessed with different methods: mean AER was used at baseline, and ACR was used at follow-up. Classification issues have been shown when these two methods are compared (41); however, only smaller differences were observed for the classification of macroalbuminuria that we used for defining DN in this study, and we therefore speculate that this is a minor issue.

Assessing DR from two fundus photos at baseline compared with seven at follow-up make underestimation of DR at baseline a possibility. Nevertheless, it has previously been shown that the two EURODIAB fields are acceptable for grading DR compared with the seven ETDRS fields (23).

In conclusion, our study supports that changes in retinal vascular calibers are preclinical biomarkers of early diabetic neuropathy, nephropathy, and retinopathy in individuals with type 1 diabetes. This could possibly be an indication of a common pathogenic pathway for diabetic microvascular complications, and thus, retinal imaging could provide a noninvasive instrument to identify patients at risk for microvasculopathy (42).

Funding. This work was supported by Fight for Sight, Denmark; Synoptik Foundation; Medivit Aps; Gangsted Foundation; Foundation of Karen Svankjaer Yde; Lykfeldts Grant; The A.P. Moeller Foundation for the Advancement of Medical Science; The Region of Southern Denmark; and The University of Southern Denmark.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. R.B. contributed to the study concept and design, contributed to the acquisition of data, performed data analyses, and wrote the initial draft of the paper. M.L.R. and L.H. contributed to the acquisition and interpretation of data and revised the paper critically for intellectual content. U.F.-O. contributed to the acquisition of data. B.S.O. and H.B.M. contributed to the study concept and design, contributed to the acquisition and interpretation of data, and revised the paper critically for intellectual content. T.Y.W. and T.P. contributed to the interpretation of data and revised the paper critically for intellectual content. J.G. contributed to the study concept and design, contributed to the interpretation of data, and revised the paper critically for intellectual content. All authors approved the final version of the paper. R.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes full responsibility for the integrity of the data, and the accuracy of the data analysis.

Prior Presentation. An abstract of this study was presented as a Poster Presentation at the Association for Research and Vision in Ophthalmology 2014 Annual Meeting, Orlando, FL, 4–8 May 2014.