Proliferative Diabetic Retinopathy in Type 2 Diabetes Is Related to Coronary Artery Calcium in the Veterans Affairs Diabetes Trial (VADT)

Data for this study derive from baseline examinations of participants in the Risk Factors, Atherosclerosis and Clinical Events in Diabetes (RACED) study (12), a seven-site substudy of the VADT. A detailed description of the VADT with exclusion and inclusion criteria has been previously described (13). Approximately 95% of all subjects who were recruited into the VADT study, at sites participating in the RACED study, also agreed to receive coronary calcium scans.

The VADT baseline examination included a medical history, physical examination, and collection of blood for measurement of traditional cardiovascular risk factors. Height and weight were measured to the nearest 0.1 cm and 0.5 kg, respectively, and BMI (weight in kilograms divided by the square of height in meters) was calculated. Information regarding current medical health status including history of diabetes, hypertension, prior CVD, and medication use was collected by a questionnaire administered by research staff as previously described (13). A non-Hispanic white variable was generated because this race-ethnicity grouping effectively identifies those with greater CAC in this and other cohorts (12,14,15). All laboratory assays, including plasma total cholesterol, triglycerides, HDL cholesterol concentrations, and A1C, were measured in the central laboratory at Tufts University. Lipid values were assayed using standard enzymatic methods, and LDL cholesterol was calculated using the Friedewald equation. A1C was measured using an immunoaffinity method that was referenced against the national standard methodology as derived from the Diabetes Control and Complications Trial (13). Urinary protein and creatinine were measured on random morning urine samples, and an albumin-to-creatinine ratio was calculated.

Patients consented to a set of seven-field standard 30° stereoscopic color photographs of both eyes at the baseline exam according to the Diabetes Retinopathy Study protocol (13). The Fundus Photograph Reading Center of the University of Wisconsin assessed the photographs using the Early Treatment Diabetic Retinopathy Study modification of the Airlie House classification and severity scale; the score from the eye with the most severe retinopathy was used (16).

CAC was determined by electron-beam computed tomography cardiac scanning using Imatron C150XL scanners (GE Imatron, South San Francisco, CA) as previously described (12,17). Readers at the centralized reading center, which were blinded to the demographic and clinical information, performed calcium scoring. A threshold of 4 pixels and 130 Hounsfield units was used for identification of calcified lesions. Each focus exceeding the minimum volume criteria was scored using the algorithm developed by Agatston et al. (18). Total coronary calcium scores were determined by summing individual lesion scores from each of four anatomic sites (left main, left anterior descending, circumflex, and right coronary arteries). A calibration phantom was scanned under the chests of each participant at each scanning center to allow calibration of the images to identical standards, as previously described (12,17).

Statistical analyses were performed with the SAS statistical package (release 8.2; SAS Institute, Cary, NC). For the description of the data parameters with a normal distribution, means ± SD are reported. Parameters with a skewed distribution are reported as medians (interquartile range), and proportions are given for categorical variables. Significant differences between the levels of retinopathy were assessed using one-way ANOVA or the Kruskal-Wallis where appropriate. Linear and logistic regression models were used to investigate the association of CAC with risk factors, including retinopathy. The log of CAC + 1 was used to include patients with a score of zero in the linear regression models. Univariate linear regression analysis was applied to examine the association of CAC with each covariate. Stepwise regression models were created to find appropriate and parsimonious subsets of traditional risk factors to include in the initial multivariate linear (and logistic) regression analyses in order to investigate the independent association of retinopathy with CAC. Subsequent regression analyses contained additional risk factors to provide more complete models for presentation.

Our study population included 204 subjects with type 2 diabetes with a mean diabetes duration of 12.3 ± 8.3 years and had characteristics very similar to those of the overall VADT cohort (38). The majority of subjects were male (95%) and non-Hispanic white (70%), with a mean age of 62.0 ± 9.2 years. Prevalence of hypertension (80%) and a prior history of smoking (70%) was high in this population, and 39% had some prior history of CVD. Shown in Table 1 are characteristics of the population across categories of retinopathy. Age, BMI, history of smoking, hypertension, and CVD history, as well as values of A1C and LDL and HDL cholesterol, were not significantly different between groups. While total cholesterol and triglyceride levels did vary between groups, only duration of diabetes and insulin use showed a clear increasing trend across categories (from no retinopathy to PDR).

To investigate the potential relationship between retinopathy and CAC, we first determined that retinopathy (Early Treatment of Diabetic Retinopathy Study scores) was significantly correlated with CAC (r = 0.19, P = 0.006). We then examined the median CAC score across retinopathy categories (Fig. 1). For those with no retinopathy, the median CAC score was 197; for microaneurysm only, 229; for mild NPDR, 364; and for moderate to severe NPDR, 300. Most striking was the significantly higher CAC values in individuals with PDR: 981 (P < 0.01).

To determine whether the association between PDR and CAC was independent of other standard risk factors for these conditions, we first performed multivariable linear regression analyses. To avoid over-fitting of the models, we chose to initially limit the number of variables included to those with evidence of relevant model effects. Therefore, stepwise variable selection was initially performed to find a parsimonious subset of risk factors to include along with PDR in predicting CAC. After adjustment for these factors (age, non-Hispanic white status, HDL cholesterol, insulin use, and prior CVD events), PDR was significantly (P = 0.047) associated with CAC (Table 2). This translated into an ∼3.2-fold increase in individuals’ CAC scores if PDR was present. Interestingly, the β coefficients and P values for the other variables changed very little (data not shown) when PDR was included in the model, suggesting that the association between PDR and CAC does not appear to be mediated through these other risk factors. The addition of variables such as BMI, smoking, hypertension history, A1C, diabetes duration, lipid levels, and albumin-to-creatinine ratio in the models was found to not contribute to the extent of CAC and increased the strength of the PDR-CAC relationship (Table 2).

We also explored the relationship of PDR with common clinical categories of CAC (0–10, 11–100, 101–400, and >400), recognizing that prior studies have demonstrated that individuals with CAC >400 are at very high risk for prevalent and incident CVD. Impressively, in 80% of the subjects with PDR, CAC was >400 (P = 0.001), and 100% of the subjects with PDR had CAC scores >250.

To determine whether this strong association between PDR and the highest risk category of CAC scores was explained by factors other than retinopathy, we performed multivariable logistic regression analyses. Of the many standard risk factors assessed in these models, age, non-Hispanic white status, prior CVD events, and PDR were all significantly associated with CAC >400 (Table 3). In fact, individuals with PDR were over sixfold more likely to have CAC >400 than those with no PDR, even after adjustment for other CVD risk factors. Similar or higher odds ratios for PDR were also present in models that excluded women or all subjects with known CVD disease or when a less selective model was used that included many additional risk factors (Table 3).

Our data show a rather striking relationship between diabetic retinopathy, especially PDR, and coronary calcium deposition, a reliable estimate of coronary atherosclerosis. This relationship was present whether CAC was assessed as a continuous outcome or in clinical categories. In fact, in this cohort of 204 patients, 80% of the individuals who had PDR had CAC scores >400. Thus, the presence of PDR was associated with a greater than sixfold increased risk of having CAC >400, a value that is increasingly recognized as placing individuals at a particularly high risk for future clinical cardiovascular events (19,20).

Because of the moderate sample size of this substudy, variable selection procedures were initially used in order to discard less statistically important factors and thus identify relevant factors that could be included, along with PDR, in models to predict CAC. In these analyses and in subsequent more complete models, it did not appear that BMI, hypertension, lipid levels (other than HDL cholesterol), A1C, smoking history, or even the more novel risk factors plasminogen activator inhibitor-1 or fibrinogen (data not shown) were useful in predicting CAC. Although plasma cholesterol levels and blood pressure abnormalities have been associated with CAC in some but not all studies (21–27), these associations may be more difficult to identify in an older cohort of subjects with diabetes that receives numerous medications for these conditions. Consistent with other reports (12,17,27–29), age, HDL cholesterol, race-ethnicity status, and prior CVD were each significantly related to CAC. Interestingly, baseline use of insulin, possibly because it indicates long-standing and complicated diabetes, was also a significant predictor of coronary atherosclerosis burden. However, PDR remained associated with CAC, even after adjusting for these factors. Thus, the relationship between PDR and CAC score could not fully be accounted for by traditional historical or laboratory-measured cardiovascular risk factors. These data, therefore, provide additional support for the hypothesis that PDR and CAC share a common underlying pathophysiology that may be mediated, at least in part, by nontraditional risk factors.

One potential implication of a strong and independent relationship between retinopathy, especially PDR, and severe coronary atherosclerosis is that the noninvasive measurement of PDR, in combination with standard risk factors, may provide a relatively useful assessment of clinically relevant atherosclerosis. This possibility deserves evaluation in additional studies.

These results are consistent with several prior studies demonstrating that microvascular disease is related to both macrovascular disease events and overall mortality (2,3). In fact, substantial evidence now exists that both microalbuminuria and retinopathy are independent risk factors for CVD events and mortality (2–5,7–11,30). Although the studies in type 1 diabetes are limited in number (9,31), this relationship between retinopathy and CVD is present in both type 1 and type 2 diabetic populations and appears strongest for proliferative retinopathy (4,5,7–11).

Our study has several unique advantages compared with previous studies. Whereas many of the prior studies only used opthalmoscopic exams or one- or two-field retinal photographs of one or occasionally both eyes, the current study used the gold standard seven-field stereoscopic color photographs of both eyes to provide a more sensitive and comprehensive assessment of retinopathy. In addition, the fundus photographs were centrally analyzed at the Retinopathy Reading Center at the University of Wisconsin, a facility with extensive experience in retinopathy assessment for many landmark clinical studies of retinopathy. Moreover, this study clearly demonstrates that there is a direct relationship of retinopathy, and PDR in particular, not just with CVD events as previously shown, but also with the burden of coronary atherosclerosis. Of note, this association remained robust and significant using several different approaches to analyze CAC. Only one other study, comprised of a small group of patients preselected for suspected CVD, has explored the relationship between retinopathy and coronary atherosclerosis and demonstrated a relationship between retinopathy, assessed during opthalmoscopic fundus exams, and coronary angiogram severity (32).

Several biochemical pathways have been proposed to account for diabetic retinopathy, including polyol accumulation, oxidative damage, formation of advanced glycation end products, stimulation of activation pathways such as PKC, and increased levels of various growth factors (33–35). Each of these pathways has also been implicated in the initiation and/or progression of atherosclerosis. In addition, ischemia of small retinal vessels is believed to be particularly relevant for the development of new vessel proliferation in the retina (36). Similarly, recent studies have also implicated the contribution of plaque ischemia and small vessel angiogenesis in progression of atherogenesis (37). Thus, there are clearly several novel pathways that may contribute to both retinopathy and atherosclerosis and may help explain our demonstration of a close relationship between PDR and atherosclerosis. Measurement of novel risk factors that will reflect activity of these pathways and processes are needed in future studies to provide further insight into the specific mechanisms that underlay the association of retinopathy with atherosclerosis.

Several limitations of this study deserve mention. Because this is a cross-sectional study, and the subcohort of the VADT has a relatively modest number of subjects with PDR that are predominantly male, we recognize that these novel findings need confirmation in larger cohorts with greater female representation and in prospective studies. Importantly, this subset of subjects with coronary calcium scans appears to have demographic and risk factor characteristics similar to those of the entire diabetic cohort of the VADT (38). Similarly, the percentages of individuals in this cohort with mild or moderate/severe NPDR and PDR are essentially the same as those recently reported in the overall VADT population (39). For these reasons, and because the substudy is being conducted at seven sites around the country, we believe that the results from these individuals are representative of the larger VA population.

In summary, these data indicate a surprising relationship between retinopathy and the extent of CAC and, if confirmed in other studies, suggest that identifying type 2 diabetic patients with PDR may help ascertain who is at uniquely high risk for clinical CVD, indicating that it may be possible to identify and treat shared risk factors for these common micro- and macrovascular complications.

Christian Meyer, MD, John Matchette, PA, Dawn Schwenke, PhD, Phoenix VAMC; Jayendrah H. Shah, MD, Southern Arizona VA Health Care System; Sundar Muduliar, MD, Robert Henry, MD, VA San Diego Healthcare System; Moti Kayshap, MD, Long Beach VAMC; Jennifer B. Marks, MD, Hermes Florez, Miami VAMC; R. Harsha Rao, MD, VA Pittsburgh Healthcare System; Nasrin Azad, MD, Lily Agrawal, MD, Hines VA Hospital; Steven Goldman, MD, Southern Arizona VA Health Care System; Michael Criqui, MPH, MD, University of California, San Diego; Robert Detrano, MD, University of California, Irvine; Mathew Budoff, MD, Harbor UCLA Medical Center; Michael Wright, MD, University of California, San Diego; George Kondos, MD, University of Illinois, Chicago Medical Center; Steven Reis, MD, University of Pittsburgh Medical Center; and Joseph Horgan, MD, Health Test Scan Center.

Financial support was provided by the Department of Veterans Affairs Cooperative Studies Program of the VA Office of Research and Development, National Institutes of Health Grant RO1067690, the Kronos Research Institute, and a clinical research award from the American Diabetes Association.

We acknowledge the contributions of the Hines VA Cooperative Studies Program Coordinating Center, the Tufts Lipid Metabolism Laboratory, and the Harbor UCLA EBCT Reading Center.

Results from this study were presented in abstract form at the 67th annual meeting of the American Diabetes Association, Chicago, Illinois, 22–26 June 2007.